Deutsche Geophysikalische Gesellschaft e.V. ... - Bibliothek - GFZ
Deutsche Geophysikalische Gesellschaft e.V. ... - Bibliothek - GFZ
Deutsche Geophysikalische Gesellschaft e.V. ... - Bibliothek - GFZ
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
<strong>Deutsche</strong><br />
<strong>Geophysikalische</strong><br />
<strong>Gesellschaft</strong> e.V.<br />
Protokoll über das<br />
23. Schmucker-Weidelt-Kolloquium für<br />
Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See,<br />
Oliver Ritter<br />
GeoForschungsZentrum<br />
Telegrafenberg<br />
14473 Potsdam<br />
28. September - 2. Oktober 2009<br />
ISSN 0946-7467<br />
herausgegeben von<br />
Ute Weckmann<br />
GeoForschungsZentrum<br />
Telegrafenberg<br />
14473 Potsdam
Widmung<br />
Wir wollen unser Kolloquium Ulrich Schmucker (1930 -2008) und Peter<br />
Weidelt (1938 – 2009) widmen und ihm den Namen „Schmucker-Weidelt-<br />
Kolloquium für Elektromagnetische Tiefenforschung“ geben.<br />
Unser Kolloquium soll Gespräche zwischen allen fördern, die mit<br />
elektromagnetischen Methoden das Innere der Erde erforschen wollen.<br />
Ulrich Schmucker und Peter Weidelt haben in diesem Kolloquium seit<br />
seiner Gründung neue und anspruchsvolle Methoden entwickelt, dort<br />
vorgestellt, oft sogar nur in diesem Kolloquium. Sie haben damit den<br />
internationalen Ruf dieses Kolloquiums begründet und das hohe<br />
intellektuelle Niveau dieser Forschungsrichtung weltweit bestimmt.<br />
Wenn wir dieses Kolloquium, das eigentlich auch ungenannt schon immer<br />
das Schmucker-Weidelt-Kolloquium war, nun auch so nennen wollen, dann<br />
hat das drei tiefe Gründe:<br />
Es sollen diese beiden Wurzeln unseres Kolloquiums und unserer<br />
Forschungsrichtung im Gedächtnis bleiben.<br />
Das Kolloquium soll uns Teilnehmer erinnern, auch in Zukunft die<br />
methodenorientierte Grundlagenforschung und ihre praktische Erprobung<br />
auf hohem Niveau als Ideal und Ziel dieses Kolloquiums anzustreben.<br />
Eine besondere Botschaft von Ulrich Schmucker und Peter Weidelt war es,<br />
gerade junge Studierende der elektromagnetischen Tiefenforschung von<br />
Anfang an in einen ernsthaften und lebhaften Kontakt mit allen<br />
Teilnehmern einzubinden, ihnen zuzuhören und zu helfen, sie zu kritisieren<br />
und Neues von ihnen zu lernen. Wenn wir diese Botschaft auch in Zukunft<br />
zum Arbeitsstil unseres Kolloquiums machen werden, wird die Entwicklung<br />
der elektromagnetischen Tiefenforschung so voller Dynamik bleiben, wie<br />
sie es bisher war.
I<br />
Vorwort<br />
Mit Erscheinen dieses Bandes und im Einvernehmen mit den Hinterbliebenen wollen wir von nun an<br />
das Kolloquium Prof. Ulrich Schmucker (1930 -2008) und Prof. Peter Weidelt (1938 – 2009) widmen.<br />
Den Wortlaut der Widmung haben wir an den Beginn des Bandes gestellt. Beide waren für die<br />
Entwicklung der elektromagnetischen Tiefenforschung weltweit von herausragender Bedeutung. Ihre<br />
Persönlichkeit und Schaffenskraft hatten aber vor allem in Deutschland einen maßgeblichen Anteil an<br />
der überaus positiven Entwicklung, die die Elektromagnetische Tiefenforschung genommen hat.<br />
Veranstalter des 23. Kolloquiums war das <strong>Deutsche</strong> GeoForschungsZentrum - <strong>GFZ</strong>. Der Tagungsort,<br />
die Heimvolkshochschule am Seddiner See, befand sich in unmittelbarer Nähe zur<br />
brandenburgischen Landeshauptstadt Potsdam, direkt am Großen Seddiner See, in einer für<br />
Brandenburg typischen Umgebung mit vielen Seen, Laub- und Kiefernwäldern. Am Kolloquium hatten<br />
sich 88 Teilnehmer angemeldet, größtenteils aus dem deutschsprachigen Raum, aber wie auch schon<br />
in den Jahren zuvor gab es auch diesmal eine ganze Reihe von Anmeldungen aus dem europäischen<br />
Ausland. Insgesamt waren 15 deutsche und 8 europäische Institutionen vertreten.<br />
Im Gedenken an Ulrich Schmucker hatten seine ehemaligen Schüler bereits ein Kolloquium vom 26 -<br />
28 Juni 2009 im Herz-Jesu Kloster in Neustadt ausgerichtet. Peter Weidelt wurde in einer Reihe von<br />
Beiträgen im Rahmen des 23. Kolloquiums gedacht und besonders gefreut hat uns, dass auch Heidi,<br />
Anne und Jan Weidelt anwesend waren. Im vorliegenden Tagungsband sind die meisten dieser<br />
Beiträge protokolliert. Allerdings haben wir uns entschlossen, sie nicht besonders hervorzuheben,<br />
sondern sie wie immer thematisch einzuordnen.<br />
Ulrich Schmucker und Peter Weidelt hatten von Beginn an in diesem Kolloquiumsband publiziert, oft<br />
sogar nur in diesem. Der diesjährige Band enthält 32 Beiträge. Es gab die übliche große Bandbreite an
Beiträgen die theoretisch- methodische Aspekte und experimentelle Untersuchungen im Hinblick auf<br />
geodynamische und angewandte Fragestellungen abdeckten. Generell war vielleicht eine Zunahme<br />
der mehr angewandten Arbeiten festzustellen, was sicherlich auch darauf zurückzuführen ist, dass<br />
die elektromagnetischen Verfahren mittlerweile fester Bestandteil der Exploration im off-shore aber<br />
auch on-shore Bereich geworden sind.<br />
Quicklebendige Debatten gab es auch im Rahmen unserer neuen Rubrik „Was Sie schon immer über<br />
die Elektromagnetik wissen wollten, sich bisher aber nicht zu fragen trauten.“, die Ute Weckmann<br />
angeregt hatte. Insbesondere sollten damit die jüngeren Teilnehmer angesprochen werden, jedoch<br />
hatten offensichtlich besonders viele der Ü40 Teilnehmer noch jahrelang unbeantwortet gebliebene<br />
Fragen. Manchmal war für die Beantwortung der Fragen auch etwas mehr Zeit zum Nachdenken<br />
nötig, weshalb wir eine nach dem Kolloquium entstandene Antwort in diesen Band ebenfalls<br />
aufgenommen haben.<br />
Für die finanzielle Unterstützung danken wir auch unseren Sponsoren: den Firmen KMS und<br />
Metronix. Herzlich bedanken möchten wir uns bei Frau Hannelore Gendt (<strong>GFZ</strong>) für die Hilfe bei<br />
administrativen und finanziellen Aufgaben. Herrn Palm und dem Daten- und Rechenzentrum sei für<br />
die Einrichtung des Internetportals gedankt. Roxana Barth und Gregor Willkommen haben für den<br />
reibungslosen Ablauf während des Kolloquiums gesorgt und auch tatkräftig bei der<br />
Zusammenstellung dieses Bandes geholfen. Natürlich ließ sich ein solches Kolloquium nur mit der<br />
Unterstützung der gesamten MT Arbeitsgruppe des <strong>GFZ</strong> durchführen. Ein herzliches Dankeschön<br />
dafür; ebenso an Herrn Bertelmann von der <strong>Bibliothek</strong> für das Hosting der Protokollbände.<br />
Oliver Ritter und Ute Weckmann<br />
Potsdam, 6. Juni 2010
II<br />
Inhaltsverzeichnis / table of contents<br />
Rita Streich & Michael Becken, EM fields generated by finite-length wire sources in 1D media:<br />
comparison with point dipole solutions ............................................................................................1<br />
M. Afanasjew, R.-U. Börner, M. Eiermann, O. G. Ernst, S. Güttel & K. Spitzer, 2D Time Domain TEM<br />
Simulation Using Finite Elements, an Exact Boundary Condition, and Krylov Subspace Methods 16<br />
Tilman Hanstein, TEM with anomalous diffusion in fractal conductive media. .................................... 25<br />
K. M. Bhatt, A. Hördt & T. Hanstein, Analysis of seafloor marine EM data with respect to motioninduced<br />
noise ................................................................................................................................ 33<br />
K. M. Bhatt, A. Hördt, P.Weidelt & T. Hanstein, Motionally Induced Electromagnetic Field within the<br />
Ocean................... .......................................................................................................................... 46<br />
S. Kütter, A. Franke-Börner, R.-U. Börner & K. Spitzer, Three-dimensional finite element simulation of<br />
magnetotelluric fields incorporating digital elevation models ...................................................... 60<br />
M. Becken, R. Streich & O. Ritter, Establishing Controlled Source MT at <strong>GFZ</strong> ...................................... 71<br />
Johannes Kenkel, Andreas Hördt & Andreas Kemna, 2D-SIP-Modellierung mit anisotropen<br />
Widerständen. ................................................................................................................................ 77<br />
Xiaoming Chen, Ute Weckmann, Kristina Tietze, Towards a 2D anisotropic inversion. ....................... 88<br />
Kristina Tietze, Oliver Ritter & Ute Weckmann, Substitute models for static shift in 2D. ..................... 97<br />
Andreas Hördt, Peter Weidelt & Anita Przyklenk, Die Übergangsimpedanz einer kapazitiv<br />
angekoppelten Elektrode ............................................................................................................. 102<br />
Häuserer & Junge, Eine Methode gegen auslaufende Ag/AgCl//KCl(H2O) Elektroden ...................... 115<br />
Johannes B. Stoll, Celle, Christopher Virgil, Attitude Algorithm Utilised in Mobile Geophysical<br />
Measuring Systems ...................................................................................................................... 120<br />
G. Muñoz, K. Bauer, I. Moeck, O. Ritter, Joint Interpretation of Magnetotelluric and Seismic Models<br />
for Exploration of the Gross Schoenebeck Geothermal Site ......................................................... 126<br />
Ulrich Kalberkamp, Magnetotelluric measurements to explore for deeper structures of the Tendaho<br />
geothermal field, Afar, NE Ethiopia.. ............................................................................................ 131<br />
J. H. Börner, V. Herdegen, R.-U. Börner, K. Spitzer, Electromagnetic Monitoring of CO2 Storage in Deep<br />
Saline Aquifers - Numerical Simulations and Laboratory Experiments. ....................................... 137<br />
K. Lippert, B. Tezkan, R. Bergers, M. Gurk, M. v. Papen, P. Yogeshwar, Erkundung eines Aquifers unter<br />
dem Mittelmeer vor der israelischen Küste mit LOTEM. .............................................................. 143<br />
M. von Papen, B. Tezkan, On the analysis of LOTEM time series from Israel and the preliminary 1D<br />
inversion of data ........................................................................................................................... 149<br />
P. Yogeshwar, B. Tezkan, M. Israil, Grundwasserkontamination bei Roorkee/Indien: 2D Joint Inversion<br />
von Radiomagnetotellurik und Gleichstromgeoelektrik Daten .................................................... 153<br />
Widodo, Marcus Gurk, Bülent Tezkan, Site Effect Assessment in the Mygdonian Basin (EUROSEISTEST<br />
area, Northern Greece) using RMT and TEM Soundings .............................................................. 164<br />
Gerlinde Schaumann, Annika Steuer, Bernhard Siemon, Helga Wiederhold & Franz Binot, Die<br />
deutsche Nordseeküste im Fokus von aeroelektromagnetischen Untersuchungen Teilgebiete<br />
Langeoog mit Wattenmeer und Elbemündung ............................................................................ 177
Annika Steuer, Bernhard Siemon and Michael Grinat, The German North Sea Coast in Focus of<br />
Airborne Electromagnetic Investigations: The Freshwater Lenses of Borkum ............................. 188<br />
Gerhard Kapinos & Heinrich Brasse, Some notes on bathymetric effects in marine magnetotellurics,<br />
motivated by an amphibious experiment at the South Chilean margin ...................................... 198<br />
Dirk Brändlein, Oliver Ritter, Ute Weckmann, A permanent array of magnetotelluric stations located<br />
at the South American subduction zone in Northern Chile. ......................................................... 210<br />
D. Eydam & H. Brasse, Discussion on backarc mantle melting in the central Andean subduction zone,<br />
based on results of magnetotelluric studies ................................................................................. 216<br />
Václav Červ, Světlana Kováčiková, Michel Menvielle and Josef Pek, Thin Sheet Conductance Models<br />
from Geomagnetic Induction Data: Application to Induction Anomalies at the Transition from the<br />
Bohemian Massif to the West Carpathians .................................................................................. 232<br />
Anne Neska, Subsurface Conductivity Obtained from DC Railway Signal Propagation with a Dipole<br />
Model ........................................................................................................................................... 244<br />
Anja Schäfer, Heinrich Brasse, Norbert Hoffmann and EMTESZ working group, Magnetotelluric<br />
investigation of the Sorgenfrei-Tornquist Zone and the NE German Basin ................................. 252<br />
P. Sass, O. Ritter, A. Rybin, G. Muñoz, V. Batalev and M. Gil, Magnetotelluric data from the Tien Shan<br />
and Pamir continental collision zones, Central Asia ..................................................................... 263<br />
Alexander Löwer, Audiomagnetotellurik im Hohen Vogelsberg ......................................................... 269<br />
Ute Weckmann, Carsten Scholl, Dumme Fragen ................................................................................ 277
III<br />
Teilnehmerverzeichnis<br />
YassineAbdelfettah LaboratoiredeGéothermie,Neuchâtel,Suisse<br />
FilipeAdão CentrodeGeofísicadaUniversidadedeLisboa,Lisboa,Portugal<br />
JulianeAdrian UniversitätzuKöln,InstitutfürGeophysikundMeteorologie<br />
MartinAfanasjew TUBergakademieFreiberg<br />
KatharinaBairlein TUBraunschweig<br />
RoxanaBarth UniversitätPotsdam<br />
MichaelBecken <strong>Deutsche</strong>sGeoForschungsZentrumPotsdam<strong>GFZ</strong><br />
K.MangalBhatt TUBraunschweig<br />
DirkBrändlein <strong>Deutsche</strong>sGeoForschungsZentrumPotsdam<strong>GFZ</strong><br />
AnneBublitz J.W.GoetheUniversität,FrankfurtamMain<br />
MatthiasBücker TUBraunschweig<br />
XiaomingChen <strong>Deutsche</strong>sGeoForschungsZentrumPotsdam<strong>GFZ</strong><br />
JinChen IFMGEOMAR,Kiel<br />
DanielDiaz FUBerlin<br />
SebastianEhmann TUBraunschweig<br />
DianeEydam <strong>Deutsche</strong>sGeoForschungsZentrumPotsdam<strong>GFZ</strong><br />
JohannesGeiermann igem,Bingen<br />
MichaelGrinat LeibnizInstitutfürAngewandteGeophysik,Hannover(LIAG)<br />
HannahGroßbach UniversitätzuKöln,InstitutfürGeophysikundMeteorologie<br />
MarkusGurk UniversitätzuKöln,InstitutfürGeophysikundMeteorologie<br />
VolkerHaak Blankenfelde<br />
TilmanHanstein KMSTechnologiesGmbH,Köln<br />
MichaelHäuserer J.W.GoetheUniversität,FrankfurtamMain<br />
BjörnHenningHeincke IFMGEOMAR,Kiel<br />
WiebkeHeise CentrodeGeofísicadaUniversidadedeLisboa,Lisboa,Portugal<br />
StefanHendricks AlfredWegenerInstitut,Bremerhaven<br />
PaulHofmeister TUBraunschweig<br />
SebastianHölz IFMGEOMAR,Kiel
AndreasHördt TUBraunschweig<br />
AndreasJunge J.W.GoetheUniversität,FrankfurtamMain<br />
UlrichKalberkamp FederalInstituteforGeosciencesandNaturalResources,Hannover(BGR)<br />
ThomasKalscheuer ETHZurich,Zurich,Switzerland<br />
JochenKamm UppsalaUniversity,DepartmentofEarthSciences,Uppsala,Sweden<br />
GerhardKapinos FUBerlin<br />
JohannesKenkel TUBraunschweig<br />
JudithKirchner TUBergakademieFreiberg<br />
FlorianLePape DIAS,Dublin,Ireland<br />
JochenLehmannHorn ETHZurich,Zurich,Switzerland<br />
MartinLeven InstitutfürGeophysik,UniversitätGöttingen<br />
ShengjunLiang TUBergakademieFreiberg<br />
MaikLinke TUBergakademieFreiberg<br />
KlausLippert UniversitätzuKöln,InstitutfürGeophysikundMeteorologie<br />
AlexanderLöwer J.W.GoetheUniversität,FrankfurtamMain<br />
StephanMalecki TUBergakademieFreiberg<br />
EricMandolesi DIAS,Dublin,Ireland<br />
NaserMeqbel <strong>Deutsche</strong>sGeoForschungsZentrumPotsdam<strong>GFZ</strong><br />
MarionMiensopust DIAS,Dublin,Ireland<br />
MaxMoorkamp IFMGEOMAR,Kiel<br />
DanaMoritz TUBergakademieFreiberg<br />
GerardMuñoz <strong>Deutsche</strong>sGeoForschungsZentrumPotsdam<strong>GFZ</strong><br />
LutzMütschard FUBerlin<br />
AnneNeska InstituteofGeophysics,PolishAcad.Sci.,CentralneObserwatorium<br />
Geofizyczne,BelskDuzy,Poland<br />
LaustB.Pedersen UppsalaUniversity,DepartmentofEarthSciences,Uppsala,Sweden<br />
JosefPek InstituteofGeophysics,Acad.Sci.CzechRep.,Prague,CzechRepublic<br />
AnitaPrzyklenk TUBraunschweig<br />
LasseRabenstein AWIBremerhaven<br />
TinoRadic RadicResearch,Berlin<br />
ZhengyongRen ETHZurich,Zurich,Switzerland<br />
OliverRitter <strong>Deutsche</strong>sGeoForschungsZentrumPotsdam<strong>GFZ</strong>
AnnikaRödder UniversitätzuKöln,InstitutfürGeophysikundMeteorologie<br />
EstelleRoux DIAS,Dublin,Ireland<br />
AnjaSchäfer FUBerlin<br />
GerlindeSchaumann LeibnizInstitutfürAngewandteGeophysik,Hannover(LIAG)<br />
CarstenScholl FugroElectroMagnetic,Berlin<br />
KatrinSchwalenberg FederalInstituteforGeosciencesandNaturalResources,Hannover(BGR)<br />
KlausSpitzer TUFreibergTUBergakademieFreiberg<br />
AnnikaSteuer LeibnizInstitutfürAngewandteGeophysik,Hannover(LIAG)<br />
JohannesStoll Celle<br />
KurtStrack KMSTechnologies,HoustonTX,USA<br />
RitaStreich <strong>Deutsche</strong>sGeoForschungsZentrumPotsdam<strong>GFZ</strong><br />
BülentTezkan UniversitätzuKöln,InstitutfürGeophysikundMeteorologie<br />
KristinaTietze <strong>Deutsche</strong>sGeoForschungsZentrumPotsdam<strong>GFZ</strong><br />
AndreaTreichel WestfälischeWilhelmsUniversität,Münster<br />
ChristopherVirgil TUBraunschweig<br />
MichaelvonPapen UniversitätzuKöln,InstitutfürGeophysikundMeteorologie<br />
UteWeckmann <strong>Deutsche</strong>sGeoForschungsZentrumPotsdam<strong>GFZ</strong><br />
JuliaWeißflog TUBergakademieFreiberg<br />
Widodo UniversitätzuKöln,InstitutfürGeophysikundMeteorologie<br />
GregorWillkommen UniversitätPotsdam<br />
HelmuthWinter J.W.GoetheUniversität,FrankfurtamMain<br />
TamaraWorzewski IFMGEOMAR,Kiel<br />
PritamYogeshwar UniversitätzuKöln,InstitutfürGeophysikundMeteorologie<br />
FanZhang TUBergakademieFreiberg
EM fields generated by finite-length wire sources in 1D<br />
media: comparison with point dipole solutions<br />
Rita Streich ∗† and Michael Becken ∗† ,<br />
∗ Potsdam University, Institute of Geosciences,<br />
Karl-Liebknecht-Str. 24, 14476 Potsdam-Golm, Germany<br />
† <strong>GFZ</strong> German Research Centre for Geosciences,<br />
Telegrafenberg, 14473 Potsdam, Germany<br />
ABSTRACT<br />
In present-day land and marine controlled-source electromagnetic (CSEM) surveys, EM<br />
fields are commonly generated using wires that are hundreds of meters long. Nevertheless,<br />
simulations of CSEM data often approximate these sources as point dipoles.<br />
Although this is justified for sufficiently large source-receiver distances, many real surveys<br />
include frequencies and distances at which the dipole approximation is inaccurate.<br />
For 1D layered media, EM fields for point dipole sources can be computed using<br />
well-known quasi-analytical solutions, and fields for sources of finite lengths can be<br />
synthesized by superposing point dipole fields. However, the calculation of numerous<br />
point dipole fields is computationally expensive, requiring a large number of numerical<br />
integral evaluations. We combine a more efficient representation of finite-length sources<br />
in terms of components related to the wire and its end points with very general expressions<br />
for EM fields in 1D layered media. We thus obtain a formulation that requires<br />
fewer numerical integrations than the superposition of dipole fields and permits source<br />
and receiver placement at any depth within the layer stack. Complex source geometries,<br />
such as wires bent due to surface obstructions, can be simulated by segmenting<br />
the wire and computing the responses for each segment separately. We first describe<br />
our finite-length wire expressions, and then present examples of EM fields due to finitelength<br />
wires for typical land and marine survey geometries and discuss differences to<br />
point dipole fields.<br />
INTRODUCTION<br />
Controlled-source electromagnetic (CSEM) surveys are a useful exploration tool, applicable,<br />
e.g., for exploring hydrocarbon reservoirs, geothermal reservoirs, or for characterizing<br />
and potentially monitoring sites considered for carbon sequestration. A multitude of active<br />
electromagnetic sources are available, including magnetic loops (Frischknecht et al., 1991;<br />
Spies and Frischknecht, 1991) and long wires, grounded in land-based surveys (Strack, 1992;<br />
Wright et al., 2002; Ziolkowski et al., 2007) and deployed at the seafloor (Edwards, 2005) or<br />
towed through the water (e.g., Constable and Srnka, 2007) in marine surveys. Wire sources<br />
with lengths of several 100 m are most commonly used in commercial hydrocarbon exploration<br />
because of their capability to generate three-dimensional electrical current systems<br />
sensitive to both resistive and conductive, relatively deep targets (Spies and Frischknecht,<br />
1991; Constable and Srnka, 2007).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
1
In processing, inversion and interpretation of CSEM data, the sources are commonly<br />
approximated as point dipoles (e.g., Edwards, 1997; Johansen et al., 2005; Ziolkowski et al.,<br />
2007). This is adequate for sufficiently large source-receiver distances. However, real surveys<br />
targeting increasingly deep and small structures may include distances and frequencies at<br />
which the inaccuracy of the dipole approximation is on the order of (or larger than) the<br />
target-related anomalies. In such cases, it is vital to consider the actual source geometry.<br />
For horizontally layered media, EM fields can be computed using well-known quasianalytical<br />
solutions that involve numerical evaluations of Bessel function integrals (e.g.,<br />
Weidelt, 2007; Løseth and Ursin, 2007). Using such point dipole solutions, EM fields due<br />
to finite-length wire sources can be synthesized by representing the wire as a line of point<br />
dipoles and summing the dipole fields. However, this procedure is computationally expensive,<br />
requiring many numerical integrations to calculate all of the point dipole contributions.<br />
To improve the efficiency of long-wire source simulations, Soerensen and Christensen<br />
(1994) derived integrated Hankel integral expressions for finite-length wires that barely<br />
require more integral evaluations than the computation of point dipole fields. However, their<br />
approach includes elaborate computations of filter coefficients for every source-receiver pair.<br />
Another approach that reduces the number of integrations from that required for summing<br />
dipole fields is the separation of EM fields due to long wires into contributions from the wire<br />
and its end points (Ward and Hohmann, 1987). This technique works with standard Hankel<br />
filters. Whereas previous application of this approach considered subsurface receivers and<br />
sources located at the air-ground interface (Ward and Hohmann, 1987), we have applied<br />
it to a more general 1D field formulation that permits source and receiver positions at<br />
arbitrary depths within the layer stack (Løseth and Ursin, 2007). This allows simulations<br />
of finite marine sources in addition to land sources.<br />
In this contribution, we first describe our finite-length wire representation. We then show<br />
examples of EM fields due to finite-length wires in representative land and marine settings<br />
over the frequency and distance ranges of typical CSEM surveys, and discuss differences to<br />
the fields due to infinitesimal dipoles.<br />
EM FIELD EXPRESSIONS FOR FINITE-LENGTH WIRES<br />
Expressions for the electromagnetic field due to a finite-length wire can be obtained by<br />
representing the wire as a line of infinitesimal dipoles and integrating the dipole expressions<br />
along the length of the wire. Using the nomenclature of Løseth and Ursin (2007), the EM<br />
fields for an x-directed point dipole embedded in a 1D layered medium can be expressed in<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
2
a form convenient for later integration as (see Appendix A)<br />
Ex = Idx<br />
⎛<br />
∞<br />
−μ0 ⎝ √<br />
4π pr zp 0<br />
s R<br />
z<br />
A ⎧<br />
⎨<br />
x<br />
11κJ0(κr)dκ+∂x<br />
⎩<br />
r−xs∞ <br />
μ0<br />
√<br />
r pr zp 0<br />
s R<br />
z<br />
A <br />
pr zp 11−<br />
s z<br />
˜ε r ˜ε s RA ⎫⎞<br />
⎬<br />
22 J1(κr)dκ ⎠, (1a)<br />
⎭<br />
Ey = Idx<br />
4π ∂x<br />
⎧<br />
⎨<br />
y<br />
⎩<br />
r − ys∞ <br />
μ0<br />
√<br />
r pr zp<br />
0<br />
s R<br />
z<br />
A <br />
pr zp 11 −<br />
s z<br />
˜ε r ˜ε s RA ⎫<br />
⎬<br />
22 J1(κr)dκ , (1b)<br />
⎭<br />
Ez = jIdx<br />
⎧<br />
⎨∞<br />
˜ε<br />
∂x<br />
4πω˜ε r ⎩<br />
0<br />
r ˜ε s<br />
pr zps R<br />
z<br />
D 22κJ0(κr)dκ ⎫<br />
⎬<br />
, (1c)<br />
⎭<br />
Hx = Idx<br />
4π ∂x<br />
⎧<br />
⎨<br />
y<br />
⎩<br />
r − ys ∞<br />
<br />
p<br />
r<br />
0<br />
r z<br />
ps R<br />
z<br />
D 11 −<br />
<br />
˜ε rps z<br />
˜ε spr R<br />
z<br />
D ⎫<br />
⎬<br />
22 J1(κr)dκ , (1d)<br />
⎭<br />
Hy = Idx<br />
⎛<br />
∞<br />
⎝<br />
p<br />
4π<br />
r z<br />
ps R<br />
z<br />
D 11κJ0(κr)dκ ⎧<br />
⎨<br />
x<br />
− ∂x<br />
⎩<br />
r−x s ∞<br />
p<br />
r<br />
r z<br />
ps R<br />
z<br />
D 11− <br />
˜ε rps z<br />
˜ε spr R<br />
z<br />
D ⎫⎞<br />
⎬<br />
22 J1(κr)dκ ⎠, (1e)<br />
⎭<br />
Hz = jIdx<br />
4πωμ0<br />
0<br />
yr − ys <br />
r<br />
0<br />
∞<br />
0<br />
μ0<br />
√<br />
pr zps R<br />
z<br />
A 11κ2J1(κr)dκ, (1f)<br />
where I is the source current, dx is the length of the source dipole, κ is the horizontal<br />
wavenumber, ω is the angular frequency, μ0 is the vacuum magnetic permeability,<br />
˜ε {s,r} = ε {s,r} + jσ {s,r} /ω are the dielectric permittivity and electric conductivity of the<br />
source and receiver layer, respectively, p {s,r}<br />
z<br />
= μ0˜ε {s,r} − κ 2 /ω 2 are frequency-normalized<br />
vertical wavenumbers, r = (x r − x s ) 2 +(y r − y s ) 2 is the horizontal source-receiver distance,<br />
and J0 and J1 are the zero- and first-order Bessel functions. R A 11 , RA 22 , RD 11 and RD 22<br />
are the reflection responses of the layered medium as given by Løseth and Ursin (2007, their<br />
Equation 134). These quantities are computed recursively using the properties and thicknesses<br />
of all layers and the source and receiver depths. Subscripts 11 denote the TE-mode<br />
and subscripts 22 the TM-mode responses. Slightly different expressions apply for receivers<br />
below (Equations 134a and b) and above the source (Equations 134c and d of Løseth and<br />
Ursin, 2007). In Equations (1), a spatial derivative ∂x has been retained wherever possible.<br />
To obtain electromagnetic fields for a finite-length wire, we integrate Equations (1) over<br />
thewirelength. Assumingthatthewireisparalleltothex-axis, we express the integrals as<br />
discrete sums over N wire elements of length Δx, located at (xn,y s ,z s ). The derivatives ∂x<br />
are replaced by −∂/∂Δx (Ward and Hohmann, 1987). Then the fields at receiver location<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
3
(x r ,y r ,z r )are<br />
Ex = I<br />
4π<br />
− I<br />
4π<br />
Ey = − I<br />
4π<br />
N<br />
<br />
Δx<br />
n=1<br />
2<br />
m=1<br />
0<br />
∞<br />
−κμ0<br />
√<br />
pr zps R<br />
z<br />
A 11J0(κrn)dκ<br />
(−1) m x r − xm<br />
rm<br />
∞<br />
0<br />
∞<br />
2<br />
(−1) m yr − ys <br />
m=1<br />
Ez = − jI<br />
4πω˜ε r<br />
Hx = − I<br />
4π<br />
Hy = I<br />
4π<br />
+ I<br />
4π<br />
rm<br />
2<br />
(−1) m<br />
<br />
m=1<br />
0<br />
∞<br />
<br />
2<br />
(−1) m yr − ys m=1<br />
N<br />
<br />
Δx<br />
n=1<br />
2<br />
m=1<br />
Hz = jI(yr − y s )<br />
4πμ0ω<br />
0<br />
∞<br />
<br />
pr z<br />
ps z<br />
rm<br />
(−1) m x r − xm<br />
rm<br />
N Δx<br />
n=1<br />
rn<br />
0<br />
<br />
<br />
μ0<br />
√<br />
pr zps R<br />
z<br />
A 11 −<br />
μ0<br />
√<br />
pr zps R<br />
z<br />
A 11 −<br />
p r zp s z<br />
˜ε r ˜ε r RA 22<br />
p r zp s z<br />
˜ε r ˜ε r RA 22<br />
<br />
J1(κrm)dκ, (2a)<br />
<br />
J1(κrm)dκ, (2b)<br />
˜ε rps z<br />
˜ε spr R<br />
z<br />
D 22κJ0(κrm)dκ, (2c)<br />
∞<br />
0<br />
<br />
pr z<br />
ps z<br />
R D 11κJ0(κrn)dκ<br />
∞<br />
0<br />
∞<br />
0<br />
<br />
pr z<br />
ps z<br />
R D 11 −<br />
<br />
˜ε rps z<br />
˜ε spr R<br />
z<br />
D <br />
22 J1(κrm)dκ, (2d)<br />
R D <br />
11 −<br />
˜ε rps z<br />
˜ε spr R<br />
z<br />
D <br />
22 J1(κrm)dκ, (2e)<br />
μ0<br />
√<br />
pr zps R<br />
z<br />
A 11κ 2 J1(κrn)dκ. (2f)<br />
For Ex, Hy and Hz, we obtain a contribution from each wire element. This contribution<br />
is described by the first sum in Equations 2a, 2e and 2f, with the distance between the<br />
receiver and the n th wire element given by rn = (x r − xn) 2 +(y r − y s ) 2 .Uponintegrating<br />
those terms of Equations (1) that contain derivatives ∂x, the derivatives disappear, and we<br />
obtain explicit contributions from the integration limits, i.e., the end points of the wire. In<br />
Equations (2), the end point contributions are given by the summation over m (m ∈{1, 2}),<br />
with the distances between the receiver and the wire ends denoted rm.<br />
To compute all EM field components for a finite-length wire using Equation (2), we<br />
have to evaluate three different integrals over the entire wire length. Accordingly, for a<br />
wire discretized into N elements, the number of numerical integral evaluations is approximately<br />
3N. In contrast, Equation (A-7) shows that two different integrals are required<br />
for computing only Ex for an infinitesimal dipole. The computation of all electromagnetic<br />
field components for a horizontal electric point dipole source requires numerical evaluations<br />
of eight different integrals (Løseth and Ursin, 2007). This would result in 8N numerical<br />
integrations when calculating finite-length wire fields from the contributions of N dipole<br />
elements. Compared to the simple summation of dipole fields, the separate computation of<br />
wire and end point contributions thus reduces the computational effort by more than 60%.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
4
EXAMPLES<br />
We use the expressions of EM fields due to finite-length wires given in Equation (2) to assess<br />
differences between the fields of finite sources and point dipoles for different representative<br />
experimental settings of land and marine surveys.<br />
Land survey: straight grounded wire<br />
We have simulated a land survey using the setup depicted in Figure 1. The electric conductivity<br />
model is based on the situation at the CO2 sequestration pilot site in Ketzin,<br />
Germany, using measured conductivity values and the actual depth and thickness of the<br />
layer into which CO2 is being injected (Giese et al., 2009). To simulate a realistic survey,<br />
in which sensors would typically be buried just below the surface, we placed electric and<br />
magnetic field receivers at a depth of 0.15 m. EM fields were computed for an infinitesimal<br />
dipole source at (x, y) =(0, 0) and a 1-km long finite-length wire centered at (x, y) =(0, 0).<br />
Both sources were located at a depth of 0.1 m.<br />
surface<br />
source,<br />
centered<br />
at (0,0)<br />
air<br />
1/3<br />
0.1<br />
1 S/m<br />
10 km<br />
surface<br />
receivers<br />
15 km<br />
0<br />
635<br />
650<br />
z (m)<br />
Figure 1: The 1D conductivity model and survey geometry used for simulating a land CSEM<br />
survey. The red arrow indicates the point dipole or 1000-m long wire source, centered at<br />
(x, y) =(0, 0) and located 0.1 m below the surface. Receivers are located 0.15 m below the<br />
surface.<br />
For convenience in the numerical simulations, we place the entire wire at a constant<br />
depth, either above or below the air-ground interface, although in real field surveys, grounded<br />
wires would typically be used, with the wire laid out on the surface and the end points coupled<br />
into the ground, e.g., via metal electrodes. The contribution from the wire body is<br />
typically smaller than that from the grounding points, and varies slowly as the wire depth<br />
crosses interfaces. Therefore, this simplification does not cause noticeable errors.<br />
In Figure 2, we display the electric field component Ex at a frequency of 0.1 Hz for<br />
the model depicted in Figure 1. The finite-wire and point dipole fields differ significantly<br />
withinaradiusof∼ 4 km from the source. Large relative differences also occur in the<br />
lowest-amplitude regions at oblique angles to the source; however, these are insignificant,<br />
because in these regions, Ex would not be measurable.<br />
For comparison, we show in Figure 3 the electric field for the reservoir model of Fig-<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
5
y (km)<br />
−6<br />
−3<br />
0<br />
3<br />
6<br />
(a)<br />
0 3 6 9<br />
x (km)<br />
y (km)<br />
−6<br />
−3<br />
−14 −12 −10 −8<br />
log10( | Ex [V/m] |)<br />
0<br />
3<br />
6<br />
(b)<br />
0 3 6 9<br />
x (km)<br />
y (km)<br />
−6<br />
−3<br />
0<br />
3<br />
6<br />
(c)<br />
0 3 6 9<br />
x (km)<br />
| Ex wire / Ex dipole 0.8 1 1.2<br />
|<br />
Figure 2: Electric field Ex for the configuration shown in Figure 1, a frequency of 0.1 Hz<br />
and (a) a point dipole source at (0, 0, 0.1) m and (b) a 1000-m long grounded wire extending<br />
from (−500, 0, 0.1) to (500, 0, 0.1) m. (c) shows the ratio between the finite-length wire and<br />
point dipole fields.<br />
ure 1 relative to the electric field for a model that does not contain a resistive layer. This<br />
demonstrates the size of the anomaly we would attempt to detect in the CSEM survey. The<br />
reservoir-related anomaly is smaller than the relative differences between finite-length wire<br />
and dipole fields, and overlaps spatially with the region in which the response is significantly<br />
influenced by the source geometry. This clearly indicates the importance of considering the<br />
true source geometry.<br />
Similar observations can be made over a wide frequency range. In Figure 4, we compare<br />
the electric field Ex for finite-length and point dipole sources, and for the reservoir and<br />
background models, at frequencies of 0.001 Hz and 1 Hz. Significant differences between<br />
finite-length and point dipole sources occur in a similar region as for f =0.1 Hz(compare<br />
Figures 4a and b to Figure 2c). The reservoir-related anomalies are somewhat smaller than<br />
for f =0.1 Hz, underlining again that it is vital to consider the actual source length when<br />
searching for such relatively small anomalies.<br />
In Figure 5, we display the finite-length wire fields for the Ey and Ez components,<br />
and their ratios to the respective point dipole fields. Here, the relative differences between<br />
the finite-length wire and point dipole fields are of similar size as for Ex. The regions in<br />
which the responses of finite-length wires and point dipoles differ significantly are similar,<br />
or slightly larger, in extent than for Ex. However, the amplitudes of Ez are considerably<br />
smaller than those of Ex and Ey, such that nearly the entire region in which Ez would<br />
be measurable (assuming instrument detection thresholds of ∼ 10 −14 − 10 −15 , and possible<br />
measurement of Ez in shallow boreholes) is strongly influenced by the source geometry.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
6
y (km)<br />
−6<br />
−3<br />
0<br />
3<br />
6<br />
0 3 6 9<br />
x (km)<br />
Figure 3: Electric field Ex for the configuration shown in Figure 1, a 1000-m long wire source<br />
and f =0.1 Hz, normalized by the electric field for a background model not containing a<br />
resistive layer.<br />
Required wire discretization<br />
To minimize the computational effort, we have empirically investigated the effects of wire<br />
discretization on the resulting electromagnetic field values. In Figure 6, we display differences<br />
between a reference field Ex computed for a 1000-m long wire discretized using an<br />
element length of 0.1 m, and more coarsely discretized wires with element lengths varying<br />
between 1 m and 50 m. As expected, differences between the reference field and fields computed<br />
using coarser discretizations increase as the element length increases. Nevertheless,<br />
at the lower frequency of 0.1 Hz (Figure 6a), differences are small for all tested element<br />
lengths. As expected, differences are larger at f = 100 Hz, with maximum differences<br />
occurring in the vicinity of the end point of the wire at x = 500 m.<br />
From this test, we conclude that for the lower frequency, a coarse discretization of ∼<br />
20−50 m would be sufficient, whereas for the higher frequency, the wire should be discretized<br />
using elements no longer than ∼ 2 − 5 m. These results may serve as rough guidelines for<br />
further simulation studies, but actual required discretizations are likely to depend on the<br />
total wire length and the resistivity model. To exclude any discretization-related error,<br />
all results presented here were computed using somewhat too careful discretizations with<br />
element lengths of 1 m.<br />
Non-straight grounded wire<br />
In real field surveys, surface obstacles may preclude laying out the source wire in a straight<br />
line. We have therefore studied differences between the EM fields for straight and nonstraight<br />
wire sources. EM fields for non-straight wires are calculated by segmenting the<br />
wire into straight sections, and computing the responses for each segment separately using<br />
1.2<br />
1.1<br />
1<br />
0.9<br />
0.8<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
7<br />
|Ex reservoir / Ex background |
y (km)<br />
y (km)<br />
−6<br />
−3<br />
0<br />
3<br />
6<br />
−6<br />
−3<br />
(a)<br />
0 3 6 9<br />
x (km)<br />
0<br />
3<br />
6<br />
(c)<br />
0 3 6 9<br />
x (km)<br />
y (km)<br />
y (km)<br />
−6<br />
−3<br />
0<br />
3<br />
6<br />
−6<br />
−3<br />
(b)<br />
0 3 6 9<br />
x (km)<br />
0<br />
3<br />
6<br />
0 3 6 9<br />
x (km)<br />
Figure 4: Ratio of Ex for a 1000-m long wire source to Ex for a point dipole for frequencies<br />
of (a) 0.001 Hz and (b) 1 Hz, and ratio of Ex for a 1000-m long wire source and the reservoir<br />
model shown in Figure 1 relative to Ex for a background model not containing the resistive<br />
layer for frequencies of (c) 0.001 Hz and (d) 1 Hz.<br />
Equations (2) with appropriate rotations. As an example, we have computed the responses<br />
for a wire that consists of two segments arranged at a 120 ◦ angle. This wire has the same<br />
grounding points as the 1000-m long straight wire used previously. Figure 7 shows the wire<br />
geometry and the electric field Ex for the bent wire relative to Ex for a straight wire at<br />
frequencies of 0.001 Hz, 0.1 Hz and 1 Hz.<br />
As expected, the Ex amplitudes for the bent wire are increased relative to the straight<br />
wire field at the side to which the wire is deviated (y >0), and decreased at y
y (km)<br />
y (km)<br />
−6<br />
−3<br />
0<br />
3<br />
6<br />
−6<br />
−3<br />
(a)<br />
0 3 6 9<br />
x (km)<br />
0<br />
3<br />
6<br />
(c)<br />
0 3 6 9<br />
x (km)<br />
−6<br />
−8<br />
−10<br />
−12<br />
−14<br />
−10<br />
−12<br />
−14<br />
−16<br />
−18<br />
log10(Ey [V/m])<br />
log10(Ez [V/m])<br />
y (km)<br />
−6<br />
−3<br />
0<br />
3<br />
6<br />
(b)<br />
0 3 6 9<br />
x (km)<br />
0 3 6 9<br />
x (km)<br />
Figure 5: Electric field components (a) Ey and (c) Ez for the model shown in Figure 1 and<br />
a 1000-m long wire source at f =0.1 Hz, and (b, d) the ratios of the finite-source Ey and<br />
Ez to the respective point dipole fields.<br />
pare Figure 7b to 3), and larger than the reservoir-related anomaly for f =1Hz(compare<br />
Figure 7c to 4d). These results demonstrate that at these frequencies, the EM fields are<br />
significantly influenced not only by the grounding point locations, but also by the entire<br />
wire layout. In contrast, at f =0.001 Hz, differences between bent and straight wire fields<br />
are barely visible (Figure 7a). At this low frequency, the electric field is similar to the potential<br />
field occurring in the DC limit (i.e., for f = 0). Here, the electric field is practically<br />
determined by the grounding point locations alone.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
9<br />
y (km)<br />
−6<br />
−3<br />
0<br />
3<br />
6<br />
(d)<br />
1.2<br />
1.1<br />
1<br />
0.9<br />
0.8<br />
1.2<br />
1.1<br />
1<br />
0.9<br />
0.8<br />
| Ey wire / Ey dipole |<br />
| Ez wire / Ez dipole |
log 10(| (E x coarse - Ex fine ) / Ex fine |)<br />
−2<br />
−4<br />
−6<br />
−8<br />
(a)<br />
Element length (m)<br />
1<br />
2<br />
5<br />
10<br />
20<br />
50<br />
0 5 10<br />
x (km)<br />
Figure 6: Relative differences between E coarse<br />
x<br />
−2<br />
−4<br />
−6<br />
−8<br />
(b)<br />
0 5 10<br />
x (km)<br />
computed for a 1000-m long wire using wire<br />
element lengths between 1 m and 50 m, and a reference field E fine<br />
x for a wire discretized<br />
using 0.1-m long elements, for frequencies of (a) 0.1 Hz and (b) 100 Hz. The resistivity<br />
model shown in Figure 1 was used, and field values were extracted along a line parallel to<br />
the source wire at a lateral offset of 50 m.<br />
Marine survey: floating wire<br />
Finite electric dipole sources used in marine CSEM surveys for commercial hydrocarbon<br />
exploration are typically ∼ 100 − 300 m long (Constable and Srnka, 2007). Accordingly,<br />
differences between the EM fields due to such finite-length sources and point dipole fields<br />
are expected to be somewhat smaller than those observed for the 1-km long wire considered<br />
in the land survey examples.<br />
We have computed marine EM responses for the model shown in Figure 8, which contains<br />
a 200-m water column and a stack of sedimentary layers, into which a 100-m thick resistive<br />
layer, representing a hydrocarbon reservoir, is embedded at a depth of 800–900 m below<br />
the seafloor. We simulated a point dipole source located at (x, y) =(0, 0) and a 300-m long<br />
wire source, also centered at (0, 0). Both sources were located 50 m above the seafloor, and<br />
receivers were placed 0.01 m above the seafloor.<br />
Electric field Ex data for this configuration due to the finite-length wire and point dipole<br />
sources are shown in Figures 9a and b, and finite-source Ex data for the reservoir model<br />
relative to Ex for a model not containing the resistive layer are displayed in Figure 9c. The<br />
finite-length wire and point dipole fields are nearly identical at radii larger than ∼ 1km<br />
from the source center. The anomaly due to the resistive layer is several times larger<br />
than the relative differences between the finite-length wire and point dipole fields, and is<br />
largest at distances well beyond the region significantly influenced by the source geometry.<br />
Therefore, the point dipole approximation may be adequate in this case. However, taking<br />
into account the exact source geometry may again become important when searching for<br />
smaller anomalies caused, e.g., by thinner reservoirs or relatively small 3D structures.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
10
y (km)<br />
−6<br />
−3<br />
3<br />
6<br />
(a)<br />
-500 0 500<br />
120° x (m)<br />
287<br />
y (m)<br />
0 3 6 9<br />
x (km)<br />
y (km)<br />
−6<br />
−3<br />
0<br />
3<br />
6<br />
(b)<br />
0 3 6 9<br />
x (km)<br />
y (km)<br />
| Ex bent / Ex straight 0.8 0.9 1 1.1 1.2<br />
|<br />
−6<br />
−3<br />
0<br />
3<br />
6<br />
(c)<br />
0 3 6 9<br />
x (km)<br />
Figure 7: Electric field Ex for a bent wire. The inset in (a) shows the geometry of a wire<br />
that consists of two segments and has the same grounding points as the straight wire used<br />
previously. Shown is the ratio of Ebent x for the bent wire to E straight<br />
x for the straight wire at<br />
depth z =0.15 m and frequencies of (a) 0.001 Hz, (b) 0.1 Hz and (c) 1 Hz. Gray shades in<br />
(b) and (c) roughly mark very low-amplitude regions in which Ex would not be measurable.<br />
σ (S/m)<br />
0<br />
3.6<br />
0.5<br />
0.01<br />
1<br />
10 km<br />
seafloor<br />
receivers<br />
15 km<br />
0<br />
200<br />
1000<br />
1100<br />
z (m)<br />
Figure 8: The 1D conductivity model and survey geometry used for simulating a marine<br />
CSEM survey. The red arrow indicates the point dipole or 300-m long wire source, centered<br />
at (x, y) =(0, 0) and located 50 m above the seafloor. Receivers are located 0.01 m above<br />
the seafloor.<br />
CONCLUSIONS<br />
We have presented a formulation for computing electromagnetic fields due to electric dipole<br />
sources of finite extent by splitting the responses into contributions from the wire body<br />
and its end points. Being derived from a quite general representation of point dipole fields,<br />
our finite-length wire formulation allows us to compute EM fields in 1D layered media for<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
11
y (km)<br />
−4<br />
−2<br />
0<br />
2<br />
(a)<br />
4<br />
0 2<br />
x (km)<br />
4<br />
log10(Ex wire −12 −11 −10 −9 −8<br />
[V/m])<br />
y (km)<br />
−4<br />
−2<br />
0<br />
2<br />
(b)<br />
4<br />
0 2<br />
x (km)<br />
4<br />
y (km)<br />
| Ex wire / Ex dipole 0.9 0.95 1 1.05 1.1<br />
|<br />
−4<br />
−2<br />
0<br />
2<br />
(c)<br />
4<br />
0 2<br />
x (km)<br />
4<br />
| Ex reservoir / Ex background 0.5 0.75 1 1.25 1.5<br />
|<br />
Figure 9: (a) Electric field Ex for a 300-m long wire source and the marine survey configurationshowninFigure8atf<br />
=0.1 Hz. (b) The ratio between the finite-length wire field<br />
shown in (a) and the field due to a point dipole source. (c) The finite-length wire field for<br />
the reservoir model of Figure 8 relative to the background field for a model not containing<br />
the 100-m thick resistive layer. Note the different color scales in (b) and (c).<br />
sources and receivers located at any depth. Compared to direct summation of dipole fields,<br />
we gain ∼ 60% efficiency. Implementation is straightforward, as we only require integral<br />
evaluations via standard fast Hankel transforms, for which precomputed sets of coefficients<br />
are available.<br />
The utility of the finite-length wire formulation has been demonstrated by presenting<br />
EM fields for several representative CSEM survey configurations. Our tests confirm that<br />
finite-length wire and point dipole fields can differ significantly over distance ranges reaching<br />
several times the wire length. For a simulated land CSEM survey targeting a thin resistive<br />
anomaly, comparison of the responses for a 1000-m long grounded wire source to point dipole<br />
responses shows that in this case, it is crucial to consider the actual source geometry. The<br />
anomalous response of the target structure is smaller than the relative differences between<br />
finite-length wire and point dipole fields, and overlaps spatially with the region in which<br />
the response is strongly influenced by the source geometry. Qualitatively similar deviations<br />
between finite-length wire and point dipole fields are observed over relatively wide frequency<br />
ranges and for different EM field components. At frequencies above the ‘effective’ DC limit,<br />
it is also important to consider the actual wire layout; knowledge of the grounding point<br />
positions is not sufficient for computing accurate responses.<br />
In contrast, for the marine survey simulated, using a shorter wire and thicker target layer,<br />
the wire geometry only had a relatively small impact on the responses, and anomalies due<br />
to the target layer were spatially well separated from the region significantly influenced by<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
12
the source geometry. This indicates that the actual source geometry is of minor importance<br />
for the survey considered. Nevertheless, this cannot be generalized to other surveys aiming<br />
at detecting smaller target structures that generate weaker EM field anomalies.<br />
Our 1D solution for finite-length wire sources can easily be incorporated into higherdimensional<br />
CSEM modeling and inversion algorithms. Most 2D and 3D modeling schemes<br />
use a secondary field approach, in which primary fields for a simple (e.g., homogeneous or<br />
1D) model are computed analytically, and secondary fields arising from deviations of the<br />
2D or 3D resistivity model from the background model are computed numerically. Here,<br />
finite-length sources can be included into the background field computations.<br />
ACKNOWLEDGMENTS<br />
This work was funded by the German Federal Ministry of Education and Research within<br />
the framework of the GeoEn project.<br />
APPENDIX A<br />
EM FIELD EXPRESSIONS FOR POINT DIPOLES<br />
We show the derivation of expressions for dipole EM fields in a form convenient for integration<br />
over the length of a source wire using the example of the Ex component. Analogous<br />
considerations apply for the other EM field components.<br />
Using the nomenclature of Løseth and Ursin (2007), the space-frequency domain electric<br />
field Ex for an isotropic layered medium, with source and receiver embedded at arbitrary<br />
depth, is given by the double Fourier integral<br />
Ex = − Idx<br />
8π2 ∞<br />
∞<br />
1− k2 x<br />
κ2 <br />
μ0<br />
√<br />
pr zps R<br />
z<br />
A 11 +<br />
−∞ −∞<br />
<br />
pr zps z<br />
˜ε r ˜ε s<br />
k2 x<br />
κ2 RA22 <br />
exp {j (kxx+kyy)}dkxdky, (A-1)<br />
where kx and ky are horizontal wavenumbers with κ2 = k2 x + k2 y , and the other symbols are<br />
the same as those explained for Equation (1).<br />
Transformation to cylindrical coordinates using kx = κ cos α, ky = κ sin α, x = r cos β,<br />
y = r sin β, ξ = α − β + π/2, k2 x →−∂2 x , and substituting the zero-order Bessel function,<br />
results in<br />
Ex = Idx<br />
4π<br />
⎛<br />
<br />
⎝<br />
0<br />
∞<br />
−μ0<br />
√ p r z p s z<br />
J0 (κr) = 1<br />
2π<br />
2π<br />
R A 11κJ0(κr)dκ−∂ 2 ⎧<br />
⎨∞<br />
<br />
μ0<br />
x √<br />
⎩ pr zps R<br />
z<br />
A 11−<br />
0<br />
0<br />
e jκr sin ξ dξ, (A-2)<br />
p r zp s z<br />
˜ε r ˜ε s RA 22<br />
κ J0(κr)dκ<br />
⎫⎞<br />
⎬<br />
⎠.<br />
⎭<br />
(A-3)<br />
After evaluating one of the spatial derivatives using (Ward and Hohmann, 1987)<br />
∂xJ0 (κr) =− κx<br />
r J1 (κr) , (A-4)<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
13<br />
<br />
1
we obtain<br />
Ex = Idx<br />
4π<br />
⎛<br />
<br />
⎝<br />
0<br />
∞<br />
−μ0<br />
√<br />
pr zps R<br />
z<br />
A 11κJ0 (κr)dκ + ∂x<br />
⎧<br />
⎨ <br />
x<br />
⎩ r<br />
0<br />
∞<br />
<br />
μ0<br />
√<br />
pr zps R<br />
z<br />
A 11 −<br />
p r zp s z<br />
˜ε r ˜ε s RA 22<br />
⎫⎞<br />
⎬<br />
dκ ⎠ . (A-5)<br />
⎭<br />
Equation (A-5) is used as the basis for deriving the expressions for finite-length wire fields.<br />
Further evaluation of the second spatial derivative using<br />
∂xJ1 (κr) =− x<br />
r2 J1 (κr)+ κx<br />
r J0 (κr) (A-6)<br />
results in an explicit expression for the point dipole field in terms of two different Hankel<br />
integrals:<br />
Ex = Idx<br />
4π<br />
+<br />
⎛<br />
<br />
⎝<br />
0<br />
∞<br />
x 2<br />
<br />
1 2x2<br />
−<br />
r r3 ∞<br />
− 1<br />
r2 0<br />
<br />
μ0<br />
√<br />
pr zps R<br />
z<br />
A x2<br />
11 −<br />
r2 μ0<br />
√<br />
pr zps R<br />
z<br />
A 11 −<br />
p r z p s z<br />
˜ε r ˜ε r RA 22<br />
REFERENCES<br />
p r zp s z<br />
˜ε r ˜ε r RA 22<br />
<br />
κJ0 (κr)dκ<br />
⎞<br />
J1 (κr)dκ⎠<br />
. (A-7)<br />
Constable, S., and L. J. Srnka, 2007, An introduction to marine controlled-source electromagnetic<br />
methods for hydrocarbon exploration: Geophysics, 72, WA3–WA12.<br />
Edwards, N., 2005, Marine controlled source electromagnetics: Principles, methodologies,<br />
future commercial applications: Surveys in Geophysics, 26, 675–700.<br />
Edwards, R. N., 1997, On the resource evaluation of marine gas hydrate deposits using<br />
sea-floor transient electric dipole-dipole methods: Geophysics, 62, 63–74.<br />
Frischknecht, F. C., V. F. Labson, B. R. Spies, and W. L. Anderson, 1991, Profiling methods<br />
using small sources, in Electromagnetic Methods in Applied Geophysics: Society of<br />
Exploration Geophysicists, 2, 105–270.<br />
Giese, R., J. Henninges, S. Lüth, D. Morozova, C. Schmidt-Hattenberger, H. Würdemann,<br />
M. Zimmer, C. Cosma, C. Juhlin, and CO2SINK Group, 2009, Monitoring at the<br />
CO2SINK site: A concept integrating geophysics, geochemistry and microbiology: Energy<br />
Procedia, 1, 2251–2259.<br />
Johansen, S. E., H. E. F. Amundsen, T. Røsten, S. Ellingsrud, T. Eidesmo, and A. H.<br />
Bhuyian, 2005, Subsurface hydrocarbons detected by electromagnetic sounding: First<br />
Break, 23, 3136.<br />
Løseth, L. O., and B. Ursin, 2007, Electromagnetic fields in planarly layered anisotropic<br />
media: Geophysical Journal International, 170, 44–80.<br />
Soerensen, K. I., and N. B. Christensen, 1994, The fields from a finite electrical dipole - A<br />
new computational approach: Geophysics, 59, 864–880.<br />
Spies, B. R., and F. C. Frischknecht, 1991, Electromagnetic sounding, in Electromagnetic<br />
Methods in Applied Geophysics: Society of Exploration Geophysicists, 2, 285–425.<br />
Strack, K. M., 1992, Exploration with deep transient electromagnetics: Elsevier.<br />
Ward, S. H., and G. W. Hohmann, 1987, Electromagnetic theory for geophysical applications,<br />
in Electromagnetic Methods in Applied Geophysics: Society of Exploration Geophysicists,<br />
131–311.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
14
Weidelt, P., 2007, Guided waves in marine CSEM: Geophysical Journal International, 171,<br />
153–176.<br />
Wright, D., A. Ziolkowski, and B. Hobbs, 2002, Hydrocarbon detection and monitoring<br />
with a multicomponent transient electromagnetic (MTEM) survey: The Leading Edge,<br />
21, 852–864.<br />
Ziolkowski, A., B. A. Hobbs, and D. Wright, 2007, Multitransient electromagnetic demonstration<br />
survey in france: Geophysics, 72, F197–F209.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
15
2D Time Domain TEM Simulation Using Finite Elements, an<br />
Exact Boundary Condition, and Krylov Subspace Methods<br />
M. Afanasjew 1,2 ,R.-U.Börner 1 ,M.Eiermann 2 ,O.G.Ernst 2 ,S.Güttel 2,3 , and K. Spitzer 1<br />
1 Institute of Geophysics, TU Bergakademie Freiberg, Germany, 2 Institute of Numerical<br />
Analysis and Optimization, TU Bergakademie Freiberg, Germany, 3 now at: Section of<br />
Mathematics, Université deGenève, Switzerland<br />
1 Summary<br />
In this report we present a numerical method for solving Maxwell’s equations in the time domain<br />
assuming an arbitrary two-dimensional conductivity distribution including an isolating air half-space.<br />
The method allows to carry out the computations for the subsurface only, because the air-earth interface<br />
is handled by an exact boundary condition. The spatial discretization is done with the finite element<br />
method, leading to a linear system of ordinary differential equations (ODE). We use state-of-the-art<br />
Krylov subspace methods for this ODE to advance a given initial electric field to selected times of<br />
interest. The presented theory is tested with some standard models and compared to a traditional<br />
finite difference time stepping implementation with respect to accuracy and efficiency. The results<br />
clearly demonstrate the superiority of the presented method in terms of run time given a comparable<br />
accuracy.<br />
Keywords: Analytic boundary condition, finite element method, Krylov subspace, time domain, transient<br />
EM<br />
2 Introduction<br />
The transient electromagnetic (TEM) method has become a standard technique in geophysical prospecting<br />
during the past years. It is already in wide use, e. g., for the exploration of important resources like<br />
hydrocarbons, groundwater and minerals. One important aspect here is a reliable and computationally<br />
efficient simulation of the decaying electromagnetic field, which can be leveraged to get a better understanding<br />
of field behavior in complicated real-world settings as well as a building block in inversion<br />
schemes, that ultimately aim at resolving arbitrary conductivity structures from only a few well-placed<br />
measurements.<br />
The predominant forward modeling technique in the literature is the finite difference time domain<br />
(FDTD) method, that was already introduced by Yee (1966). An explicit time-stepping technique, that<br />
already dealt with an isolating air half-space, was developed by Oristaglio and Hohmann (1984) for the<br />
two-dimensional case and later refined by Wang and Hohmann (1993) for three dimensions. Like for<br />
other explicit time-stepping methods, the size of the time steps, that the described Du Fort-Frankel<br />
scheme can stably perform, depends on the grid spacing and the lowest conductivity. Although the<br />
resistive air is already eliminated, thousands of time steps have to be performed although only a few<br />
dozen solutions are necessary to describe the decaying field.<br />
The approach taken here is based on a finite element discretization, which allows for greater flexibility<br />
when modeling complicated conductivity structures. High accuracy is obtained with less effort compared<br />
to graded tensor product grids used with finite differences. It also helps in the construction of an<br />
analytic boundary condition, avoiding a few drawbacks the implementation by Oristaglio and Hohmann<br />
(1984) has. Contrary to the finite difference approach, the matrices resulting from the discretization are<br />
symmetric and, thus, allow for a wide range of efficient and state-of-the-art time integration techniques,<br />
which can exploit this property. One such family of time integrators is based on building a Krylov<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
16
subspace and extracting approximations to the matrix exponential function from said space, thus<br />
evaluating the sought electric field directly at a given time.<br />
3 Theory<br />
3.1 Governing Equations<br />
Our governing equation derives from Maxwell’s equations in the diffusive limit, with the constitutive<br />
relations used and the magnetic field already eliminated. Thus, we write<br />
with<br />
∇× μ −1 ∇× e + ∂tσe = −∂tj e<br />
e = e(x,t) the electric field,<br />
μ = μ(x) the magnetic permeability,<br />
σ = σ(x) the electric conductivity, and<br />
j e = j e (x,t) the impressed source current density.<br />
We now restrict ourselves to the two-dimensional case (xz-plane) with infinitely long line sources<br />
perpendicular to this plane. Given these assumptions, we can express the electric field as<br />
e(x, y, z, t) =e(x, z, t) y (2)<br />
where y is the unit vector along the y-axis and e a scalar function. Equation (1) then reduces to<br />
−∇ 2 e + σμ∂te = −μ∂tj e . (3)<br />
Γ0<br />
Ω<br />
z =0<br />
Figure 1: Computational domain with an arbitrary conductivity structure. On top is the air-earth interface<br />
(bold), the remaining boundaries are subsurface boundaries.<br />
Our computational domain Ω (cf. Figure 1) is a rectangle and its top edge is aligned with the airearth<br />
interface Γ0 = {(x, z) :z =0} on which we impose an explicit boundary condition and perfect<br />
conductor boundary conditions on all other domain boundaries. It is important for these boundaries to<br />
be sufficiently far away from the sources, so that they don’t distort the propagating field.<br />
Our objective will be to compute the configuration of the electric field ei at times ti for i ∈{1, 2,...,n}<br />
givenaninitialfielde0 at t0. We use sources that are switched off at t = 0 and therefore the right hand<br />
side of (3) vanishes for t>0.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
17<br />
(1)
3.2 Air-Earth Interface<br />
We assume that e satisfies Laplace’s equation in the isolating air half-space (z 0} denotes the lower half-space.<br />
By using the exact boundary condition (6) we can write<br />
∂ta(e, ψ) σ + c(e, ψ)+<br />
+∞<br />
−∞<br />
Te(x, t) ψ(x, 0) dx = −μ<br />
+∞<br />
−∞<br />
∂tj e (x, t) ψ(x, 0) dx. (10)<br />
To be able to discretize this equation we need a computable expression for the following bilinear form<br />
b(φ, ψ) =<br />
+∞<br />
−∞<br />
Tφ(x) ψ(x) dx (11)<br />
and indeed, after some rearrangement in the Fourier domain, one obtains (Goldman et al. 1986)<br />
+∞ +∞<br />
b(φ, ψ) =− 1<br />
log(|x − y|) φ<br />
π −∞ −∞<br />
′ (x) ψ ′ (y) dxdy. (12)<br />
The discretization of this integral on the boundary of a triangulation can now be easily computed.<br />
The continuity of e allows us to use standard linear Lagrange elements on a triangulation of the<br />
computational domain. After discretizing the variational formulation we arrive at<br />
M∂te = Ke. (13)<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
18
Finite Differences For comparison, we also perform a finite difference discretization and get the ODE<br />
∂te = Ae where A is a large sparse matrix representing the discrete curl-curl operator.<br />
The following procedure, that has to be repeated in every time step, outlines the implementation of an<br />
exact boundary condition in this setting: Take the field data at the air-earth interface, interpolate it to<br />
an equidistant grid, perform an FFT to transform the data into the wave number domain, multiply<br />
every wave number with a constant (this would be a convolution in the space domain), revert the<br />
FFT and finally interpolate back to the graded grid to get the field values at a negative z, thus, inside<br />
the air half-space. Use the data resulting from this classical upward continuation in the usual finite<br />
difference stencil to compute the top layer in the next step.<br />
A variation of this approach was also implemented for the numerical examples. It combines all of the<br />
above steps into a single (dense) matrix that has the same effect but can be evaluated more efficiently.<br />
This is called matrix upward continuation later on.<br />
Details about the finite difference discretization can be found in Oristaglio and Hohmann (1984).<br />
3.4 Time Integration Techniques<br />
Krylov Subspace Methods Krylov subspace methods cover a big range of applications and are in<br />
use for several decades. What we want to focus on are Krylov subspace methods for the evaluation<br />
of matrix functions, like we can see in (13), where the function is the matrix exponential and the<br />
matrices come from the spatial discretization using finite elements. A nice side effect of the origin of<br />
those matrices is, that they are usually symmetric which many algorithms can benefit from.<br />
Given a square matrix A of size n × n, a vector b of length n and a suitable scalar function f, wecan<br />
write<br />
f(A) b = p(A) b. (14)<br />
with p a polynomial that Hermite-interpolates in the eigenvalues of A. We will be focusing on the<br />
exponential function and on rational Krylov subspaces, that are defined as follows<br />
with Km(A, b) = b,Ab,A 2 b,...,A m−1 b and<br />
Q(A, b) := qm−1(A) −1 Km(A, b) (15)<br />
qm−1(z) =<br />
m−1 <br />
j=1<br />
ξj=∞<br />
(z − ξj) . (16)<br />
Such subspaces are constructed, e. g., with the rational Arnoldi method to obtain an orthonormal<br />
basis of Qm from which approximations to f(A) b can be computed with several procedures. We have<br />
to choose the poles ξj of the rational Krylov subspace. Their number and choice highly impact the<br />
quality of the approximations that can be extracted from the subspace. Luckily, there is a heuristic to<br />
determine good poles for the approximation of the exponential function, given a certain parameter<br />
interval (in our case the times we want to advance to) and accuracy requirements. These, and many<br />
more things concerning rational Krylov subspaces, are tackled in Güttel (2010), which is also a good<br />
starting place to dig deeper into the literature.<br />
Looking more closely at (13) we see, that in order to apply a Krylov subspace method we need to<br />
remove the matrix M in front of the time derivative, e. g. by moving it to the right<br />
∂te = M −1 Ke. (17)<br />
The inverse of the matrix doesn’t have to be explicitly computed, but we have to solve a system of<br />
linear equations in every iteration step. This might be considered too expensive, so there is a technique<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
19
termed mass lumping which converts M into a diagonal matrix that can be easily inverted without<br />
distorting the solution too much.<br />
However, even if we don’t have to solve linear systems with the matrix M, wedohavetosolveshifted<br />
systems with the matrix K during the construction of the rational Krylov subspace. Therefore, an<br />
efficient solver is also an important ingredient. We will see in the numerical examples what options we<br />
have.<br />
Time Stepping On the other hand there are simple, yet relatively efficient explicit time stepping<br />
techniques. One of them is the well-known Du Fort-Frankel method which is implemented here to<br />
perform the intergration in time for the finite difference discretization.<br />
4 Numerical Examples<br />
All computations were carried out on a machine with four AMD Opteron 8380 quad cores, running<br />
at 2.5 GHz with a total of 128 GiB of RAM. The algorithms were implemented in pure MATLAB<br />
code running under MATLAB 2008b, with some finite element related tasks (nothing time-critical)<br />
performed by COMSOL 3.5a, both running in a 64-bit Linux environment. To generate reproducible<br />
timing we restricted ourselves to one computational thread that was explicitly pinned to one of the<br />
cores.<br />
4.1 Model<br />
The considered models are basically those from Oristaglio and Hohmann (1984), all based on graded<br />
tensor product grids. We extended the mesh slightly so as not to get a distortion from the boundaries<br />
for late times. The resulting grid has 236 × 88 cells, is 12800 m wide and 5255 m deep, with grid spacings<br />
ranging between 10 m and 240 m. To make the computations comparable with the finite difference code<br />
we decided to use exactly the same grid for the finite element code. We simply subdivided each cell<br />
into two triangles, however, the number of unknowns or degrees of freedom and their locations are<br />
identical. Technically, this is of course not necessary and a properly adapted unstructured grid would<br />
certainly yield even better results at lower cost.<br />
The initial field e0 at t0 =10 −6 s is due to two line sources at positions (x, z) = (−500 m, 0m) and<br />
(x, z) = (0 m, 0m), the former being negative while the latter is positive with a unit source current<br />
strength. It is computed analytically for a homogeneous half-space that corresponds to the background<br />
conductivity of the model.<br />
We have looked at two models with different conductivity structures. See Figure 2 to get an approximate<br />
idea of their location inside the grid. The background has a resistivity of 300 Ωm and the conductor<br />
(red) 0.3 Ωm.<br />
Γ0<br />
Ω<br />
z =0<br />
Figure 2: Schematic view of the computed models. The models are the homogeneous half-space and a conductor<br />
(red) inside a homogeneous half-space. The conductor is located at the position 290 m ≤ x ≤ 310 m,<br />
100 m ≤ z ≤ 400 m in the grid.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
20
4.2 Methods and Setup<br />
Independently of the method of choice we start at the time t0 =10 −6 s and integrate from there till<br />
we reach tn =10 −2 s. We seek to compute the configuration of the electric field at 10 logarithmically<br />
equidistant times per decade.<br />
For the finite difference code we mostly stick to the implementation by Oristaglio and Hohmann<br />
(1984), with the only option to perform the upward continuation in the traditional way, which is a<br />
significant expense per time step, or we can decide to precompute the matrix formulation of this<br />
upward continuation and to apply it in every time step.<br />
For the finite element method we have no choice regarding the exact boundary condition. However, we<br />
can choose between different variants of the rational Krylov method. For all of them we have a total<br />
of 25 poles, which should be close to optimal for this application. For the first 24 poles we alternate<br />
between 2.7826 · 10 5 and 0.0242 · 10 5 . The last pole is set to infinity.<br />
We have four implementations, that we included in this test. Two of them use mass lumping to reduce<br />
the mass matrix to a diagonal matrix. Furthermore, for each group we can either call the standard<br />
MATLAB solver or we can exploit the fact that we have multiple identical poles for which we have to<br />
solve with different right hand sides, but identical matrices. We do this by computing a sparse LU<br />
factorization (we use UMFPACK for this) for every unique pole and then leverage that factorization to<br />
speed up the solves during the construction of the rational Krylov subspace.<br />
4.3 Results<br />
We first look at the computation times that are pictured in Figure 3 and listed in Table 1. The times<br />
shown here are for the homogeneous half-space model, but aren’t significantly different for the other<br />
models since the grid and the minimal conductivity are identical. Thus, we can use these numbers to<br />
judge the acceleration we can get by using our method.<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
Figure 3: Computation times split into solve and setup/post-processing step, cf. also Table 1. On the left<br />
are the computation times for the default grid and on the right for the regularly refined version<br />
of this grid with approximately four times as many nodes and degrees of freedom. The following<br />
abbreviations were used: classical upward continuation (cUC), matrix upward continuation (mUC),<br />
rational Krylov (rK), mass lumping (ML), UMFPACK solver (UMF).<br />
As is clearly visible, all FE-based methods are significantly faster than their finite difference counterparts,<br />
sometimes even 22 times faster.<br />
Looking at the transients (cf. Figure 4) at some select points inside the mesh, we see that although<br />
these newer methods are so much faster, we don’t really sacrifice accuracy. In fact, in many cases the<br />
finite element solution is more accurate than the one obtained by finite differences.<br />
We conclude our numerical examples with showing some cross-sections (cf. Figure 5), that are nothing<br />
but a few snapshots of an animation that is just not suitable for this medium, but nevertheless gives a<br />
coarse idea of how the electric field propagates with time.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
21
Figure 4: Transients of the homogeneous half-space model with relative errors (left) and the model with a<br />
conductor (right). Negative field values are denoted by dashed lines while positive values are drawn<br />
with a solid line. We can see that we have a very good agreement between the FD and FE solution.<br />
When comparing against an analytical solution in the homogeneous half-space, we see that the relative<br />
errors are often lower than those of the FD solution, despite the lower numerical effort.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
22
Figure 5: Cross-sections of a homogeneous half-space model (left) and the model with a conductor (right).<br />
These cross-sections were plotted to give a visual idea of how the fields advance in time. The common<br />
initial configuration was omitted. We decided to plot the results of the FE simulation, but there was<br />
hardly any visible difference compared to the FD solution. On the right we can nicely see how the<br />
field gets captured inside the high conductivity structure and creates an anomaly compared to the<br />
homogeneous case.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
23
Default grid Refined grid<br />
Method Solve Other Solve Other<br />
FD, time stepping, classical upward continuation 31.7s < 0.1 s 171.8s < 0.1s<br />
FD, time stepping, matrix upward continuation 8.6s 0.9s 98.4s 2.7s<br />
FE, rational Krylov 3.6s 0.6s 19.4s 1.9s<br />
FE, rational Krylov, mass lumping 3.1s 0.6s 17.7s 1.9s<br />
FE, rational Krylov, UMFPACK 1.9s 0.6s 10.4s 1.9s<br />
FE, rational Krylov, mass lumping, UMFPACK 1.4s 0.6s 7.7s 1.9s<br />
Table 1: Computation times for the default grid and one, that was obtained from this with regular refinement.<br />
The column Solve denotes the time spent for performing the integration in time, while the column<br />
Other lists the time spent in setting up the boundary condition and the computation of field values<br />
from the solution vectors in case of the FE method. Other setup times are not shown. The advantage<br />
of FE-based methods can be clearly seen.<br />
5 Conclusions<br />
We were able to leverage several state-of-the-art techniques to create a combined efficient forward<br />
modeling code that is significantly faster than traditional codes. We have also seen that we don’t have<br />
to make sacrifices regarding the accuracy of those computations. This was already achieved by using<br />
the mesh from the finite difference discretization, which was helpful for comparison, but which also<br />
has many limitations due to its regular structure. An unstructured grid—properly adapted to the<br />
problem—could be used to further increase accuracy or speed.<br />
We are working on creating a similar framework for the three-dimensional time domain TEM problem.<br />
All of the methods and tools are already existing or have a straightforward extension to 3D. Given the<br />
results from 2D we expect an even greater speed-up when applying this to three dimensions.<br />
6 Acknowledgments<br />
This work was supported by the German Research Foundation (DFG) under signature Spi 356/9.<br />
References<br />
Goldman, Y., C. Hubans, S. Nicoletis, and S. Spitz (1986). A finite-element solution for the transient electromagnetic<br />
response of an arbitrary two-dimensional resistivity distribution. Geophysics 51(7): 1450–1461.<br />
Goldman, Y., P. Joly, and M. Kern (1989). The Electric Field in the Conductive Half Space as a Model in Mining and<br />
Petroleum Prospecting. Mathematical Methods in the Applied Sciences 11: 373–401.<br />
Güttel, S. (2010). Rational Krylov Methods for Operator Functions. PhD thesis. TU Bergakademie Freiberg.<br />
Oristaglio, M. L. and G. W. Hohmann (1984). Diffusion of Electromagnetic Fields into a Two-Dimensional Earth: A<br />
Finite-Difference Approach. Geophysics 49(7): 870–894.<br />
Wang, T. and G. W. Hohmann (1993). A Finite-Difference, Time-Domain Solution for Three-Dimensional Electromagnetic<br />
Modeling. Geophysics 58(6): 797–809.<br />
Yee, K. S. (1966). Numerical Solution of Initial Boundary Problems Involving Maxwell’s Equations in Isotropic Media.<br />
IEEE Trans. Ant. Propag. 14: 302–309.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
24
Abstract<br />
TEM with anomalous diffusion in fractal conductive media<br />
Tilman Hanstein<br />
KMS Technologies GmbH, Köln<br />
The transient electromagnetic response of an inductive loop source over a half-space with<br />
fractal characteristics is simulated. The conductivity of the ground has a spatial distribution,<br />
which is described by a roughness parameter. The roughness can be related to the fractal<br />
dimension and controls the transient decay of the signal. Asymptotic limits are observed for<br />
the early and late time behaviour.<br />
The work is based on a paper by Everett (2009), which has been reviewed and the aspect of<br />
asymptotic limits in time has been further investigated. Contrary to that paper, here a<br />
relationship between roughness and the electromagnetic decay at early and late time has been<br />
deduced. The transient decay for normal diffusion is t -5/2 and for anomalous diffusion the<br />
decay is slower. The new power law for anomalous diffusion has been proven by theoretical<br />
analysis and has been verified by numerical experiments.<br />
The numerical evaluation of the inverse Laplace transform with the method of the fast Hankel<br />
transform are excellent in numerical accuracy and the method with the Gaver-Stehfest<br />
algorithm is not sufficient to estimate the power law decay at late times.<br />
Anomalous Diffusion<br />
The concept of anomalous diffusion is a useful approach for the description of diffusion<br />
process and transport dynamics in complex systems. The fractional equations are derived<br />
asymptotically from basic random walk models and become a complementary tool for<br />
handling non-exponential relaxation patterns.<br />
For transient electromagnetic diffusion Everett starts this concept with a generalized Ohm’s<br />
law (Everett 2009, Weiss and Everett 2007)<br />
j <br />
E<br />
1 t t<br />
E j E <br />
0<br />
the parameter ~t - describes the generalized electrical conductivity and is appropriate for<br />
the anomalous diffusion coefficient. The Ohm’s law becomes a convolution between the<br />
generalized conductivity and the electric field E. The roughness parameter can vary 0 1.<br />
d<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
25
After applying Ampere‘s and Faraday‘s law to the generalized current density, we get the<br />
fractional diffusion equation<br />
<br />
E 0<br />
* E 0<br />
0Dt<br />
t<br />
2 1<br />
where now the fractional derivative or Riemann-Louiville operator is introduced (Metzler and<br />
Klafter 2000)<br />
t<br />
1<br />
1 E<br />
<br />
0 Dt<br />
E t <br />
d<br />
1<br />
t<br />
t <br />
0<br />
<br />
and is the Gamma function which serves as normalization constant.<br />
The fractional diffusion equation will be solved in the Laplace domain and the transformation<br />
of the operator yield to<br />
1<br />
1<br />
~<br />
D E<br />
t s E<br />
s<br />
L 0 t<br />
with the complex Laplace variable s=i (Abramowitz & Stegun 1964).<br />
The fractional diffusion equation becomes now a simple expression<br />
2 ~<br />
E s<br />
1<br />
0<br />
<br />
This differential equation can be solved straightforward with the standard methods as used for<br />
the electromagnetic 1-D formulation in a layered medium.<br />
~<br />
E<br />
s<br />
Electromagnetic responese of a loop over rough half-space<br />
Everett (2009) has used a separate horizontal loop configuration for transmitter and receiver<br />
as it is typically used for TEM. The transmitter loops usually a square loop can be represented<br />
by a circular loop whith equivalent area. For the separate loop configuration the time<br />
derivative of the vertical magnetic field is measured outside the transmitter loop.<br />
In frequency domain the vertical magnetic field is presentated as Hankel Integrals<br />
~<br />
h z<br />
s Ia<br />
J aJr <br />
0<br />
<br />
2<br />
<br />
2<br />
k <br />
2<br />
<br />
1<br />
0<br />
E<br />
t<br />
d<br />
.<br />
The integration is over the spatial wavenumber and I is the current in the transmiter, a the<br />
transmitter loop radius, r the distance to the receiver point, J0 and J1 are Bessel functions of<br />
order 0 and 1 and k the fractional wavenumber. The integral can be solved analytically for an<br />
infinitessimally magnetic dipole, therefore the limit of the first order Bessel function for small<br />
radius is taken into account (Abramowitz & Stegun 9.1.10)<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
26
a<br />
J1 a .<br />
a<br />
0<br />
2<br />
The response of a vertical magnetic dipole in frequency domain over a rough conductive<br />
media is<br />
m 1<br />
2<br />
u<br />
s 999u4uue ~ 3<br />
z<br />
3 2<br />
H<br />
2<br />
r<br />
with the fractional induction number u = k r = (s 1- 0 ) 1/2 r.<br />
Asymptotic limits in frequency and time domain<br />
u<br />
For normal diffusion with roughness equal zero the transient response can be also given in<br />
analytical expression but for a general fractional induction number with the roughness<br />
parameter the transformation to time domain has to be done by numerical techniques. Only<br />
the asymptotic limits for the high and low frequency limit can be calculated in closed form.<br />
For high frequency and early time we get<br />
~<br />
H<br />
h.<br />
f .<br />
z<br />
m 9<br />
m 9 1<br />
<br />
t<br />
2 2<br />
z<br />
3 2<br />
2<br />
r u<br />
2<br />
r r <br />
e.<br />
t.<br />
s <br />
H t<br />
The low frequency can be developed in a series<br />
n1n3 <br />
2<br />
~ m 1<br />
n<br />
H z s 1<br />
2<br />
3 <br />
2<br />
r 2 n4<br />
n!<br />
The first 3 terms are important for the late time behavior<br />
~<br />
H z<br />
m<br />
2<br />
r<br />
1 <br />
2 <br />
<br />
0<br />
2<br />
u <br />
<br />
2<br />
3<br />
s 1<br />
u u <br />
3<br />
1<br />
2<br />
4<br />
15<br />
<br />
<br />
<br />
.<br />
1 In time domain the first term is a -function without influence on late time decay. The second<br />
term is responsible for the late time behavior in a rough medium<br />
L<br />
1<br />
1<br />
s <br />
<br />
<br />
t<br />
<br />
<br />
<br />
<br />
t<br />
1<br />
2<br />
for 0<br />
for 0<br />
The third term describes the classical -5/2 decay response in a non-fractional medium and for<br />
a fractional medium this term decays faster than the previous second term.<br />
<br />
.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
27<br />
<br />
.
The late time response for the magnetic field can be summarized<br />
<br />
.<br />
H t l<br />
z<br />
t<br />
m<br />
<br />
2<br />
r<br />
3<br />
3 2<br />
5<br />
0<br />
2<br />
t<br />
10 <br />
<br />
<br />
0<br />
<br />
<br />
t<br />
<br />
2 1<br />
2<br />
for 0<br />
for 0<br />
The cause for the heavy tailed decay response in rough geological media can be explained by<br />
the additive second term which yield to a continuous transition from non-fractional to<br />
fractional diffusion as is shown in figure 1 and 2.<br />
Numerical inverse Laplace transform<br />
Numerical methods are applied for the transformation to time domain. Since the transient is a<br />
real and causal function, the inverse Fourier or Laplace transform can be calculated by a sinetransform.<br />
<br />
2<br />
h z<br />
z<br />
0<br />
t Im h ~ sint In this expression I have already considered that the current step in the transmitter decribed by<br />
1/i and the time derivative of the receiver coil by a multiplication with i cancel each other.<br />
Since for diffusion processes the kernel function is a smooth function, the technique of the<br />
Fast Hankel Transform can be applied. The sine function is experessed by Bessel function<br />
with fractional order ½<br />
2<br />
J 1 x sinx.<br />
2 x<br />
The fast Hankel transform is a well know technique for calculating the transient response, I<br />
used 250 filter coefficients, 15 /decade, calculated with the program by Christensen (1990).<br />
Everett (2009) has chosen the Gaver-Stehfest algorthim for his investigation. This method is<br />
also good for diffusion processes and has been succesfully applied. It is a favourized method<br />
in hydrology, because it needs only real arithmetic and the Laplace variable s is considered as<br />
a real variable. Knight and Raiche (1982) introduced this technique to the electromagnetic<br />
community. Stehfest (1970) published an Algol routine, which can be straight forward<br />
translated to other computer languages and Everett (2009) has tabled the coefficients for<br />
several total number of coefficients.<br />
d<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
28
Figure 1: Transient loop-loop response with early and late time approximation over halfspace<br />
with different roughness in conductivity, the offset is 100 m, = 0.1 S/m.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
29
The disadvantage of the Gaver-Stehfest algorithm is that the numerical accurracy cannot<br />
generally be increased by increasing the number of coeffcients. The accuracy is depending on<br />
the accurracy in number of digits of the kernel function and the number of digits of the<br />
maschine precision. Stehfest recommended 8 coefficients for single precision and 18<br />
coeffcients for double precision. Everett used 18 coefficients. In my experiments I have found<br />
out that 12 is an optimal number for electromagnetic application.<br />
Figure 2: Numerical evaluation of the exponent in the transient decay and theoretical<br />
asymptotic limit for different roughness<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
30
Results<br />
For the numerical experiment I used the same model as Everett, beside that here a vertical<br />
magnetic dipole instead of a horizontal circular loop is used. The base conductivity of the<br />
half-space is 0.1 S/m with combination of various roughness parameter has been applied.<br />
Figure 1 shows the transient as induced voltage due to an step response of the transmitter<br />
current for different roughness paramters. The large time range has been chosen to<br />
demonstrate the early and late time behavior in comparison with the asymptotic limits. For<br />
real measurements the time range will be much smaller. All transients - shown here - are<br />
calculated with the fast Hankel transform. As reference the transient of homogenous halfspace<br />
with normal diffusion is also shown with grey lines. For separate loop configuration the<br />
transient shows a sign change. The negative values are shown with dashed lines and positive<br />
with solid lines. The asymptotic limits agree very well with theoretical prediced limits at eary<br />
and late times shown here as straight lines. The early time behaviour is for practical pupose of<br />
minor relevance, because in real meausrements the early time is influenced by system<br />
response of the system as the ramp for the current step off. So that in the data the straight line<br />
will not be visible. But the late time behaviour will be visible if the late time data can be<br />
measured and the data quatlity is sufficient.<br />
To analyze how accurate the power law decay can be estimated from the calculated transients<br />
especialle for low roughness numbers, the transients are calculated up to extrem late times and<br />
the power law is determined by taking the numerical derivative d ln H(t) / d ln t. The result is<br />
shown in figure 2.<br />
Figure 3: Numerical evaluation of the exponent in the transient decay and theoretical<br />
asymptotic limit for different roughness calculated with Gaver-Stehfest alogorithm<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
31
The graphs show the exellent accuracy of the fast Hankel transform, even for small roughness<br />
numbers, and indicates the smooth transition, when the roughness approaches zero. The<br />
influence of the anomalous diffusion moves to later times for decreasing roughness.<br />
For very low roughness there will be a time range showing almost normal decay t -5/2 and then<br />
goint to the predicted power law decay at extrem late time – approaching infinity. Responsible<br />
for this behavior is the second term of the low frequency approximation which is added.<br />
The Gaver-Stehfest algorithm is not sufficient to estimate the power law decay. I tried<br />
different ways of programming, e.g to consider numerical accuracy the Laplace transform is<br />
done before the Hankel transform over the spatial wave number as recommended by Knight<br />
and Raiche (1982). The best results are shown in figure 3, achieved with 12 coefficients and<br />
using the analytical response for a vertical magnetic dipole in Lapace domain. Notice that the<br />
time range is shorter but still 3 decades more than in Everett’s paper.<br />
Reference:<br />
Abramowitz, M. & Stegun, I. A., 1966. Handbook of Mathematical Functions with Formulas,<br />
Graphs and Mathematical Tables. National Bureau of Standards, Applied Mathematics Series<br />
55.<br />
Christensen, N. B., 1990. Optimized Fast Hankel Transform Filters, Geophys. Prosp., 38,<br />
545-568.<br />
Everett, M., 2009. Transient electromagnetic response of a loop source over a rough<br />
geological medium, Geophys. J. Int., 177, 421-429.<br />
Knight, J. H. & Raiche, A. P., 1982. Transient electromagnetic calculation using Gaver-<br />
Stehfest inverse Laplace transform method, Geophysics, 47, 47-50.<br />
Metzler, R. & Klafter, J., 2000. The random walk’s guide to anomalous diffusion: a fractional<br />
dynamics approach, Phys. Rep.,339, 1-77.<br />
Stehfest, H., 1970. Numerical inversion of Laplace transforms, Comm. A.C.M.,13,47-49 (see<br />
also remark p.624).<br />
Weiss, C. J. & Everett, M. E., 2007. Anomalous diffusion of electromagnetic eddy currents in<br />
geological formations, J. geophys. Res. 112, B08102, doi:10.1029/2006JB004475.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
32
Analysis of seafloor marine EM data with respect to motion-induced noise<br />
K. M. Bhatt 1 , A. Hördt 1 and T. Hanstein 2<br />
1<br />
Inst. f. Geophysik u. Extraterrestrische Physik, TU Braunschweig<br />
2 KMS Technologies - KJT Enterprises Inc.<br />
Abstract<br />
Any appreciable movement of sea water induces an electromagnetic field, which acts as<br />
noise for marine controlled source electromagnetic (mCSEM) data. Noise in seafloor<br />
mCSEM data is considered small, but since the characteristic reservoir signal is also small,<br />
the understanding and possible removal of noise may be essential to increase the number<br />
of possible target reservoirs.<br />
A power spectral density plot of the time-series extracts the power of each frequency<br />
contribution, but the shortcoming is that the time information is lost. To get time<br />
information corresponding to each frequency, we plotted a spectrogram which shows how<br />
the spectral density of a signal varies with time. The collective use of these three plots, i.e.<br />
time series, power spectral density and spectrogram, helps to analyse the information<br />
concealed in the time-series. In a measured signal-free mCSEM data, we are able to<br />
identify various oceanic features like microseisms, swell etc, which play a significant role<br />
in inducing an electric field.<br />
Key words: mCSEM, Swell, Microseisms<br />
____________________________________________________________________________<br />
1. Introduction<br />
Marine controlled source electromagnetic (mCSEM) data is generally contaminated by<br />
some unwanted electromagnetic (EM) signals ambient at the seafloor. Understanding and<br />
possible removal of the noise is essential as technology is advancing from deep ocean to<br />
shallow ocean, where the noise is much more effective.<br />
In general, there could be two possible sources of oceanic noise production.<br />
Internal oceanic processes or any other external influences like ionospheric and<br />
magnetospheric current systems. Faraday (1832) reported that any appreciable<br />
movement of seafloor and sea water by induction generates electromagnetic signals with<br />
in the ocean. Motional induction got more attention after world war II, with the detailed<br />
oceanic induction study by many authors like Sanford (1971), Podney (1975), Chave and<br />
Cox (1982), Chave and Luther (1990). Similarity in all these studies is that they all<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
33
studied the theoretical aspect of motional induction occurring due to various movements<br />
in the ocean. Experimental evidence was provided by Webb and Cox (1986), who made a<br />
nice comprehensive study about the seafloor microseisms by observing pressure and<br />
electric field spectra at the seafloor.<br />
Wave progressions and oceanic processes are, usually, a function of the regional<br />
and local conditions. A variety of motional processes like swell, ripples, internal waves,<br />
microseisms, hum etc generate a range of EM noise at the seafloor, which varies from<br />
place to place. External, ionospheric and magnetospheric, current systems also<br />
contribute EM noise, but they are limited to the lower frequency range due to the<br />
conductive filtering of the ocean. Characteristics of the noise are poorly known till now,<br />
as only few studies are made. The present study is dedicated to the understanding of the<br />
noise characteristics at the sea floor.<br />
For the study, a marine EM data set recorded at the seafloor in the absence of a<br />
transmitter, is used. Recording is made for two horizontal components of electric fields<br />
i.e. Ex and Ey, which are perpendicular to each other. Visibly, Ex and Ey time series data<br />
appears different in their pattern and amplitudes despite the same oceanic environment.<br />
This suggests that a comprehensive study on the directional characteristics of the<br />
events/sources of noise could play a decisive role in identifying the noise sources.<br />
The oscillatory occurrence, due to a disturbance, can be comfortably visualised in<br />
frequency domain. The power spectral density (PSD) of each frequency contribution<br />
helps in marking the power of the individual sources. But the shortcoming of the PSD<br />
calculation is that if the source of the characteristic frequency is not known a priori, it is<br />
difficult to conclude about the source environment with the frequency information only.<br />
In this case, time information together with the PSD may help at least to distinguish<br />
between localised and ambient sources, which itself is of significant use. To overcome the<br />
PSD shortcoming to retain time information corresponding to each frequency,<br />
spectrograms are plotted, which show how the spectral density of a signal varies with<br />
time.<br />
In the signal-free mCSEM data, we have observed many features which are new<br />
for mCSEM studies. The spectrogram presents clear prints of oceanic features like<br />
microseisms and swell, which play a significant role in inducing an electric field at the<br />
seafloor.<br />
2. Marine Controlled Source Electromagnetism (mCSEM)<br />
A mCSEM data acquisition methodology is shown in Figure 1. In practice, an EM<br />
transmitter is towed close to the seafloor to maximize the coupling of electric and<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
34
magnetic fields with seafloor rocks. These fields are recorded by instruments deployed on<br />
the seafloor at some distance from the transmitter. Details about the data acquisition<br />
practice in both time and frequency domain of mCSEM is well documented in the article<br />
by Constable and Srnka, (2007).<br />
Figure 1: Schematic diagram showing field layout for mCSEM sounding.<br />
3. Sources of mCSEM noise<br />
The noise convolved in mCSEM data is capable of masking significant marine<br />
information. Broadly, two origins are expected as noise sources in mCSEM data:-<br />
a) External origin: The magnetospheric and ionospheric current system induces<br />
natural electric and magnetic (EM) fields in the conductive formations. The<br />
induced fields are signal for the magnetotelluric (MT) technique but noise for<br />
mCSEM data. The conductive ocean filters the higher frequencies. The skin depth<br />
() is given by<br />
This implies, f 76,000 / 2<br />
503<br />
1<br />
f<br />
for =3.3 S/m<br />
Therefore, at the ocean floor the external origin fields effectively contribute<br />
frequencies less than or equal to f = (76,000/ D 2 ), where D is the depth of ocean<br />
floor in meters and f is given in (1/s). As example for an ocean of depth 500 m, the<br />
external field will contribute frequencies less than 0.3 Hz at the sea floor. This<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
35
suggests that the frequencies higher than 0.3 Hz could be by internal oceanic<br />
origin.<br />
b) Internal origin: The dynamics of oceanic water with in the ambient geomagnetic<br />
field induces electric and magnetic field in the ocean. A frequency range of the<br />
fields is generally 0.3 < f (Hz) < 16 (Dahm et. al., 2006) excluding long period<br />
waves like tidal waves, tsunami, seiches and storm surge etc.<br />
Other than these noise sources, there are processes like earthquakes, volcanoes,<br />
manmade explosions etc occurring at the vibrant ocean floor, which add noise in<br />
mCSEM data.<br />
4. Motional induction<br />
The ocean is an electrically conducting fluid containing the ionic charge particles.<br />
Moving charge particles in the ambient geomagnetic field experience a deflective Lorentz<br />
force. If v is the velocity of charge particle q moving in the geomagnetic field B , then the<br />
Lorentz force is-<br />
<br />
q(<br />
v B)<br />
(1)<br />
F L<br />
The charge q experiences the deflecting force FL because of the action of an electric field<br />
which we call Lorentz electric field ( EL ),<br />
<br />
FL<br />
<br />
qEL<br />
(1-a)<br />
<br />
v B<br />
(2)<br />
E L<br />
The field EL generates a secondary electric field E , mainly by two processes: i) Galvanic<br />
process, ii) Inductive process (Bhatt et al, 2010). For a stationary frame of reference, the<br />
current density in the Ohms law is given by<br />
<br />
(<br />
E E ) (<br />
E v B)<br />
(3)<br />
J L<br />
Here, is the conductivity of the ocean, E is the current term generated by both a<br />
<br />
galvanic and an inductive process and (<br />
v B)<br />
is the source current term for the motional<br />
induction case. Under the quasi-static approximation, the set of Maxwell’s equation to be<br />
solved for the motional induction case is<br />
<br />
H (<br />
E v B0)<br />
<br />
(4-a,b)<br />
E <br />
H<br />
0<br />
t<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
36
Simplification of the source (<br />
v B)<br />
in terms of component Jx, Jy and Jz can be written as,<br />
J<br />
J<br />
J<br />
x<br />
y<br />
z<br />
(<br />
E<br />
<br />
(<br />
E<br />
(<br />
E<br />
x<br />
y<br />
z<br />
v<br />
v<br />
v<br />
y<br />
z<br />
x<br />
B<br />
B<br />
B<br />
z<br />
x<br />
y<br />
The non-divergent requirement condition of the current (i.e. . J 0<br />
<br />
) suggests that the<br />
current will close its loop with in the earth rather than closing in the ocean. Therefore,<br />
the horizontal electric currents (i.e. Jx and Jy) in the ocean are more effective than the<br />
vertical currents (i.e. Jz) as the horizontal extent of ocean is much larger than the vertical<br />
depth. In mCSEM, the data is recorded at the seafloor, where normally the vertical<br />
velocity component (i.e. vz) is very small. Therefore, terms vzBBy and vzBx B is negligible.<br />
Above argument simplifies (4),<br />
Jx<br />
(<br />
Ex<br />
v yBz<br />
)<br />
(6)<br />
J (<br />
E v B )<br />
Using (6), simplification of (4-a) gives<br />
E<br />
E<br />
y<br />
x<br />
y<br />
x<br />
x<br />
1<br />
<br />
( zH<br />
<br />
1<br />
<br />
( zH<br />
<br />
y<br />
x<br />
z<br />
v<br />
v<br />
v<br />
<br />
<br />
y<br />
x<br />
z<br />
x<br />
y<br />
B<br />
B<br />
B<br />
z<br />
z<br />
y<br />
z<br />
x<br />
)<br />
)<br />
)<br />
H ) v B<br />
y<br />
H ) v B<br />
Where the first term is the induced field and second term is the source term. Normally,<br />
the area used for mCSEM data acquisition is very small to observe horizontal variation in<br />
the vertical magnetic field Hz. Therefore, term y z H and xH z are negligible. Finally, we<br />
have a simplified equation,<br />
E<br />
E<br />
x<br />
y<br />
1<br />
<br />
zH<br />
<br />
1<br />
<br />
zH<br />
<br />
y<br />
x<br />
v B<br />
y<br />
v B<br />
Noticeably, the horizontal particle motion (i.e. vx and vy) is source for the induction of the<br />
horizontal electric field (i.e. Ex and Ey).<br />
5. The Data<br />
We have analysed horizontal component electric field (i.e. Ex and Ey) data recorded at<br />
500 m depth on the ocean floor. The recording is made in the absence of transmitter<br />
current to understand the oceanic background noise. Time series is shown in Figure 2. It<br />
is evident that the strength and pattern of Ex and Ey are different.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
37<br />
x<br />
z<br />
z<br />
<br />
<br />
<br />
<br />
<br />
<br />
x<br />
z<br />
z<br />
<br />
<br />
<br />
<br />
<br />
<br />
(5)<br />
(7)<br />
(8)
In equation (8), the first term is much smaller than the second term (i.e.<br />
( 1/<br />
)<br />
H v B ), suggesting x-and y- component of velocity principally controls the<br />
z<br />
y|<br />
x<br />
y|<br />
x<br />
z<br />
strength of the Ey and Ex component of the electric field respectively. Evidently, the<br />
strength of Ex is higher than Ey (Figure 2). This suggests that the wave velocity in the ydirection<br />
will be higher than the x-direction (i.e. vy > vx). Further, as normally the surface<br />
wave moves towards the coast. Accordingly, the velocity component pointing towards the<br />
coast has higher velocity than the other horizontal component. The present data is<br />
recorded with a setting that the y-component of the receiver points towards the cost and<br />
constructs an angle of approx. 55 with it. Thus, the data acquisition setting as well<br />
favours for vy > vx. In general, electric field measurements characterize an average<br />
motion over the length of the antenna. Therefore electric field components can be used to<br />
derive information about the average velocity of the movements.<br />
Electric field (nV/m)<br />
Electric field (nV/m)<br />
200<br />
100<br />
0<br />
-100<br />
-200<br />
100<br />
50<br />
0<br />
-50<br />
-100<br />
Time series for E x<br />
50 100 150 200 250 300<br />
Time (minute)<br />
Time series for E<br />
y<br />
50 100 150 200 250 300<br />
Time (minute)<br />
Figure 2: Five hour times-series of two horizontal components of electric fields. Ex (top) and Ey<br />
(bottom). The recording is made at the seafloor, 500 m below the sea surface.<br />
6. Power Spectral Density (PSD) and Spectrogram<br />
Power spectral density is calculated to study the power content of the frequencies. For<br />
the PSD calculation, steps followed are as follows:<br />
I. Inspection of time series (Figure 2): Inspection is done to recognize glitches or<br />
other outliers in the data that are not consistent with the rest of the time series.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
38
II. Detrending: Detrending ensures that the first Fourier coefficient does not<br />
dominate the PSD.<br />
III. Windowing: A Hanning window is applied to minimise the leakage of spectral<br />
density.<br />
IV. Calculation of discrete Fourier transform (DFT):<br />
V. Calculation of PSD: PSD is calculated by segment averaging method (Welch,<br />
1967).<br />
The PSD for Ex and Ey is shown in Figure 3. Over all spectral power is decreasing with<br />
increase in frequency. Four different slopes can be seen in the both PSD’s (Ex and Ey).<br />
Let the slope containing frequencies between 0.001 – 0.1 Hz, 0.1 - 2 Hz, 2 – 10 Hz and 10<br />
- 25 Hz represent respectively low, intermediate, sub-high and high frequency range.<br />
Evidently, in the low frequency range the power of the electric field rises sharply.<br />
Generally, external origin field mainly affects the low frequency range, as higher<br />
frequencies are filtered out by the ocean resulting in passage for low frequencies only<br />
(i.e. less than 0.3 Hz for a 500 m deep ocean) to the ocean bottom due to skin depth. The<br />
oceanic eddies (very long wavelength oceanic feature) are another low frequency source<br />
(Chave and Filloux, 1984). Here, these two sources are presumed for the sharp rise in<br />
power in the low frequency range (0.001 – 0.1 Hz). In the range of intermediate<br />
frequency, four peaks are evident, at 0.2, 0.3, 0.4 and 1 Hz, in the PSD representing that<br />
this frequency range is receptive for the various oceanic flows close to seafloor. The subhigh<br />
frequency range is nearly flat. A flat PSD, generally, corresponds to a field which<br />
contains equal power within a fixed bandwidth which resembles noise. In the high<br />
frequency range a sharp rise is evidenced which is presumed due to digitisation noise.<br />
The frequency and power information of a PSD is insufficient to characterise the<br />
source process corresponding to the frequencies. Time preservation may help in this<br />
subject. Together with the frequency and power, information about time may help in<br />
identifying and characterising the source nature of process. The localised time process<br />
may represent a source progressing in an interval of time while an ambient time process<br />
reflects a continuous process. For the purpose therefore spectrograms are plotted which<br />
preserve the time information together with frequency and power. The spectrogram is<br />
the discrete-time Fourier transform for a sequence, computed using a sliding window.<br />
For a spectrogram calculation, the time series is divided in to segments equal to the<br />
length of the hamming window. Each segment overlaps 50% of the samples with the<br />
adjacent segment and then PSD is calculated for a defined frequency length. Again and<br />
again PSD is calculated by sliding the window to build a spectrogram. The Spectrograms<br />
corresponding to time series shown in Figure 2 is shown in Figure 5. There are three<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
39
distinct anomalous features evident in spectrograms. One feature is corresponding to<br />
time approx. 165 min (like spike) and other two are at frequencies 1, 0.3 Hz. It is difficult<br />
to confidently propose a source nature for the spiky feature at approx. 165 min. This<br />
could be due to either by a regional earthquake or by a volcano like feature. In general,<br />
regional earthquakes are found in band frequencies of 0.1 to 1 Hz and here the spiky<br />
feature correspond the same range. The other two features at frequencies 1 and 0.3 Hz<br />
are respectively by the microseisms and swell, that will be discussed in the next section.<br />
E x (nV/m) 2 /Hz<br />
10 5<br />
10 4<br />
10 3<br />
10 2<br />
10 1<br />
10 0<br />
10 -1<br />
10 -2<br />
10 -2<br />
Power Spectral Density (E x )<br />
10 -1<br />
Frequency (Hz)<br />
0.2 & 0.4 Hz<br />
0.3 Hz<br />
10 0<br />
1 Hz<br />
10 1<br />
E y (nV/m) 2 /Hz<br />
10 5<br />
10 4<br />
10 3<br />
10 2<br />
10 1<br />
10 0<br />
10 -1<br />
10 -2<br />
10 -2<br />
Power Spectral Density (E y )<br />
0.2 & 0.4 Hz<br />
10 -1<br />
Frequency (Hz)<br />
Figure 3: Spectra of the two horizontal electric components Ex and Ey corresponding to time series<br />
shown in Figure 2. The graphs show the PSD vs. frequency. The anomalous peaks are visible at 0.2, 0.4 and<br />
1 Hz. There is also a weaker peak at 0.3 Hz. Four slopes can be observed in the spectra (pink dashed lines).<br />
7. Microseisms and swell<br />
Microseisms are like a soft earth tremor originating with in the ocean by non linear<br />
interaction of oceanic waves, which causes a continuing oscillation of the ocean floor. The<br />
broad frequency range for microseisms is between 0.05 to 1 Hz, which mainly depends<br />
on the ocean depth and oceanic conditions.<br />
Languet-Higgins (1950) proposed a mechanism for microseisms (0.05 -2.0 Hz) and<br />
showed that if two identical progressive waves travelling in opposite directions interact<br />
with each other, there is a second order pressure term effect which does not vanish with<br />
depth and can thus reach the deep ocean bottom (Figure 4). Consider two surface waves<br />
of frequencies f1 and f2 (f1f2), moving with approx. same velocity in opposite direction.<br />
Let the wave number of frequencies are respectively k1 and -k2, and then k1-k2.<br />
Interaction of these waves will leave behind a wave with very small wave number {i.e.<br />
k1+ (-k2) = diminutive} and very large wavelength. The large wavelength is capable of<br />
creating a pressure disturbance effectively at the ocean floor. The amplitude of the<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
40<br />
0.3 Hz<br />
1 Hz<br />
10 0<br />
10 1
pressure disturbance is proportional to the product of the interacting wave heights and<br />
the frequency. These pressure fluctuations in the water column might then excite<br />
Rayleigh waves in the solid earth and be observed as microseisms.<br />
In short, the favourable oceanic conditions for microseisms generation are:<br />
1. Shoreline geological setting creating grounds for nonlinear interaction of<br />
surface waves.<br />
2. Near shore reflection of high frequency surface waves and thereafter head<br />
on interaction.<br />
3. A fast moving storm creating a sequence of wave in different directions.<br />
4. High frequency wave interaction may generate whitecaps (a wind blown<br />
wave whose crest is broken and appears white), which leads in acoustic<br />
energy transmission to ocean bottom.<br />
The depth of the ocean is another important factor in microseisms generation. The wave<br />
generation in the ocean depends on the wind velocity, which become efficient when the<br />
wave velocity is close to the phase<br />
velocity of the ocean waves. Velocity of<br />
ocean wave (V) is, roughly speaking,<br />
(gD), where D is ocean depth and g is<br />
acceleration due to gravity. Therefore,<br />
for a 4 km deep ocean and a 500 m<br />
shallow ocean, the velocity V is<br />
Figure 4: Microseisms mechanism (from Elgar approximately 200 m/s and 70 m/s<br />
(2008)). Nonlinear interaction of short opposing respectively. A typical wind velocity<br />
waves leads to a long wavelength wave generation, rarely exceeds few tens of m/s. This<br />
which reaches the seafloor and generates<br />
suggests wind velocities are quite close<br />
i i<br />
to the oceanic velocities especially in the shallow oceans. Therefore the generation of<br />
microseisms is likely to be efficient in shallow oceans (Tanimoto, 2005).<br />
Swells are a kind of oceanic surface waves. They are often created by the<br />
breaking of storms thousands of kilometres away from the seashore. The distance allows<br />
the waves comprising the swells to become more stable, clean, and continuous as they<br />
travel toward the coast. For microseisms generation, in general, higher frequency waves<br />
are more efficient than the lower frequency waves like swells. Swells are more<br />
directional and therefore chances for non linear interaction is not as much as of higher<br />
frequency gravity waves (Webb and Cox, 1986), which are generated with in the ocean by<br />
the influence of gravity at the interface involving the density contrast. In Figure 3 and 5,<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
41
the anomalous feature close to 1 Hz is likely due to a microseism. A microseism is a time<br />
localised feature, the duration of which depends on the time of effective nonlinear wave<br />
interaction. It is evident from Figure 5 that close to 1 Hz, there are four time localised<br />
features (Ey component has more clear appearance than Ex), corresponding to time 8, 40,<br />
90 and 250 min approx. Further evidence in support of microseisms is the observed<br />
frequency (i.e. 1 Hz).<br />
The acquired electric field data was recorded in a shallow ocean (500 m depth).<br />
Normally, a shallow ocean may contribute the microseisms in high frequency range (i.e<br />
close to 1 Hz) as analogous to high frequency even a comparable wavelength (not<br />
necessary a huge wavelength) may reach ocean bottom to produce microseisms. The<br />
mechanism of microseisms generation suggests that maximum electric field power will be<br />
experienced in the component perpendicular to the direction of wave propagation. Here,<br />
the velocity component Vy > Vx as amplitude of Ex > Ey. The strong power of microseisms<br />
in the Ey component compared to the Ex component further supports to interpret 1 Hz<br />
frequency as microseisms.<br />
Figure 5: Spectrogram for Ex & Ey. Electric field power is colour coded in dB and displayed as function of<br />
time and frequency. Two anomalous features are visible, one at 1 Hz and the other covering a broad<br />
frequency range, having max. power at 0.2 & 0.4 Hz.<br />
The individual feature ambient in time at 0.3 Hz (Figure 3 & 5) is interpreted here<br />
as a swell. A swell consists of long-wavelength surface waves which are more stable in<br />
their directions and frequency than normal wind waves. Swells are dispersive in nature<br />
and their frequency is given by 2 gs tanh(sD), where g is acceleration of gravity; s =<br />
2/ is wave number, is the wavelength of the swell and D is the depth of ocean. A<br />
calculation suggests that a swell of frequency 0.3 Hz corresponds to a wavelength of 18<br />
m for 500 m deep oceanic water. The obtained wavelength is well in range for the<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
42
wavelengths of swells. As 0.3 Hz frequency is evident throughout the time in<br />
spectrogram (Figure 5), this is another strong support for 0.3 Hz to stand for a swell.<br />
7. Microseisms during mCSEM data recording<br />
Microseisms are a powerful feature to effectively generate EM signal at the ocean floor.<br />
They are quite capable of masking the mCSEM signals. We have an example comparing<br />
the power of microseism and mCSEM transmitter current. In Figure 5, we have seen the<br />
spectrograms of mCSEM the time-series recorded in the absence of transmitter current.<br />
On the other hand, Figure 6 represents spectrograms of the mCSEM data when<br />
transmitter was transmitting signals. Systematic decay in the power is evident in Figure<br />
6, representing transmitter is moving away from the receiver. A gathering of high power<br />
(yellow colour) patch is evident corresponding to frequency 0.1 Hz and time 200 min,<br />
may be representing a microseism. Clearly, the power of a microseism is significant<br />
enough to contaminate the mCSEM data. It is as well evident that the effect of swell is<br />
feeble in the presence of a transmitter current. For a larger transmitter receiver distance<br />
this effect may be significant.<br />
Microseisms Microseisms<br />
Figure 6: Presence of a microseism during mCSEM data recording. Electric field power is colour coded<br />
in dB and displayed as function of time and frequency. Corresponding to frequency 0.1 Hz and time 200 min<br />
(approx), a yellow coloured high power patch is evident, which may be is by a microseism. The other efficient<br />
high power patches in the spectrogram is by the transmitter current<br />
8. Conclusions<br />
By the analysis seafloor electric field data, we provide evidence for the observation of<br />
microseisms and swell in mCSEM data. Power spectral density (Figure 3) and<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
43
spectrogram (Figure 5) clearly present the evidence for the possible presence of<br />
microseism and swell corresponding to the frequencies 1 Hz and 0.3 Hz respectively.<br />
Swells are time ambient oceanic feature, which can be clearly marked in the<br />
spectrograms. The stability of 0.3 Hz in spectrograms indicates that oceanic conditions<br />
were quite stable during the recording of present mCSEM data.<br />
Figure 6 shows that the power of a microseism is significant even when the<br />
transmitter current is ‘on’. Overall, the time series analysis suggests that features like<br />
microseism and swell induce a significant electric field at the seafloor to contaminate the<br />
mCSEM data. Proper measures for removal of the noise are essential for a better<br />
interpretation of mCSEM data.<br />
Acknowledgement<br />
We thank KMS Technologies –KJT Enterprises Inc. for sponsoring the work.<br />
References<br />
1. Bhatt, K. M., A. Hördt, P.Weidelt, T. Hanstein, 2009. Motionally induced<br />
electromagnetic field within the ocean, 23. Kolloquium Electromagnetische<br />
Tiefenforschung (EMTF), Brandenburg, Germany, September-October, this issue.<br />
2. Chave, A.D., and C.S. Cox, 1982. Controlled electromagnetic sources for measuring<br />
electrical conductivity beneath the oceans, 1, forward problem and model study, J.<br />
Geophys. Res., 87, 5327-5338.<br />
3. Chave, A.D., and D.S. Luther, 1990. Low-frequency, motionally induced<br />
electromagnetic fields in the ocean, 1, theory, J. Geophys. Res., 95, 7185-7200.<br />
4. Chave, A.D., and J.H. Filloux, 1984. Electromagnetic induction fields in the deep<br />
ocean off California: oceanic and ionospheric sources, Geophys. J. R. astr. Soc., 77,<br />
143-171.<br />
5. Constable, Steven and Leonard J. Srnka, 2007. An introduction to marine controlledsource<br />
electromagnetic methods for hydrocarbon exploration, Geophysics, vol. 72, no.<br />
2, p. wa3–wa12.<br />
6. Crews, A., and J. Futterman, 1962. Geomagnetic micropulsation due to the motion of<br />
ocean waves, J. Geophys. Res., 67, 299-306.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
44
7. Dahm T, F. Tilmann, J. P. Morgan, 2006. Seismic broadband ocean-bottom data and<br />
noise observed with free-fall stations: experiences from long-term deployments in the<br />
North Atlantic and the Tyrrhenian Sea, Bull. Seismol. Soc. Am, 96, 647–664, doi:<br />
10.1785/0120040064.<br />
8. Elgar, S. ,2008. Ripples run deep, Nature, Vol 455, Page 888.<br />
9. Longuet-Higgins, M.S., E. Stern, H. Stommel, 1954. The electrical field induced by<br />
ocean currents and waves with application to the method of towed electrodes, Pap.<br />
Phys. Oceanog. Meteorol., 13(1),1-37.<br />
10. Longuet-Higgins, M.S., 1950. A theory of the origin of microseisms, Philos.Trans. R.<br />
Soc. London, Ser. A, 243, 1–35.<br />
11. Podney, W., 1975. Electromagnetic fields generated by ocean waves, J. Geophys. Res.,<br />
80, 2977-2990.<br />
12. Sanford, T.B., 1971. Motionally induced electric and magnetic fields in the sea, J.<br />
Geophys. Res., 76, 3476-3492.<br />
13. Tanimoto, T., 2005. The oceanic excitation hypothesis for the continuous oscillations<br />
of the Earth, Geophy. J. Int., 160, 276-288.<br />
14. Webb, S.S. and C.S. Cox, 1985. Observation and modelling of seafloor microseisms, J.<br />
Geophys. Res., 91, B-7, 7343-7358.<br />
15. Welch, P. D. 1967. The use of fast-Fourier transform for the estimation of power<br />
spectra: A short method based on time averaging over short, modified periodograms,<br />
IEEE Transactions of the Audio and Electroacoustics, AUlS, 70-13.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
45
Motionally Induced Electromagnetic Field within the Ocean<br />
K. M. Bhatt 1 , A. Hördt 1 , P.Weidelt 1,* and T. Hanstein 2<br />
1<br />
Inst. f. Geophysik u. Extraterrestrische Physik, TU Braunschweig<br />
2<br />
KMS Technologies - KJT Enterprises Inc<br />
Abstract<br />
The contribution of motionally induced electromagnetic (EM) fields at the seafloor is generally<br />
considered small, but since the characteristic reservoir signal in marine controlled source<br />
electromagnetic (mCSEM) data is also small, the inclusion of the motional induction<br />
contribution in modelling the signal will enhance the probability of reservoir detection. Here,<br />
we have studied the electromagnetic induction caused by ocean water flow with in earth’s<br />
magnetic field.<br />
When a charge particle moves with certain velocity in earth’s magnetic field, it<br />
experiences a Lorentz force. The action of Lorentz force generates a secondary electric field<br />
through galvanic and inductive processes. For the mathematical formulation, we considered<br />
Lorentz electric field as a source in the corresponding set of Maxwell’s equations. We solved<br />
these Maxwell's equations for a 1D model and velocity structure using two different Green’s<br />
function i.e. a two half space Green’s function and a layered Green’s function. The layered<br />
Green’s function is especially useful in studying the sensitivity of electric and magnetic field for<br />
different conductivity structures in the earth. Further, the signal variation with the conductivity<br />
of ocean, depth of ocean and wave velocity is studied to profoundly understand the effect of<br />
these parameters.<br />
________________________________________________________________________________<br />
* Deceased<br />
1. Introduction<br />
Oceanic water movements in the earth’s magnetic field induce electric and magnetic<br />
fields in the ocean. The induced fields are signal for oceanographic and seismological<br />
applications but noise for magnetotelluric (MT) and marine controlled source<br />
electromagnetic (mCSEM) applications. Oceanographers and seismologists use the fields<br />
respectively to study the velocity structure and the seismic background noise. The MT<br />
and mCSEM signals, which are generally used for lithospheric and hydrocarbon<br />
exploration studies, are contaminated by the induced field and act as noise. Our prime<br />
focus of motional induction study here is on mCSEM problems; nevertheless the results<br />
are also applicable for other scientific applications.<br />
Generally, motionally induced noise in seafloor mCSEM data is considered small,<br />
but since the characteristic reservoir signal is also small, the understanding and possible<br />
removal of noise may be essential to increase the number of possible target reservoirs.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
46
The electromagnetic induction investigation has a long history. Initially, by an<br />
observation of deflection in the galvanometer by stream waves Faraday (1832) concluded<br />
that flow of water will induce electric currents, which was afterwards measured<br />
experimentally by Young et al (1920). Later it was almost neglected and got re-attention<br />
after World War II. Longuet-Higgins et al., (1954) initiated the investigation of electric<br />
field induction by surface waves. Crews and Futterman (1962) investigated the magnetic<br />
field induction by oceanic movement. Further, Sanford, (1971) extended the theory of<br />
motional induction by considering three-dimensional water flows. A comprehensive study<br />
of the theory has been made by Podney (1975), who generalised and extended the<br />
previous work by removing the restrictive assumptions imposed on ocean-water velocity<br />
field. Chave (1983) generalised the EM induction process by considering driving electric<br />
<br />
field term (i.e. v B ) and further, Chave and Luther (1990) re-examined the motional<br />
induction problem.<br />
A primary purpose of this paper is to present a generalised simple illustrative<br />
theory for the problems of motional induction. For the purpose, we have formulated a set<br />
of Maxwell’s equation for our problem. These equations are simplified for electric and<br />
magnetic field components by considering a horizontally progressing ocean wave and<br />
then solved by using the Green’s function, with appropriate boundary conditions. Here,<br />
we solved the problem with two different Green’s functions. For a uniformly conductive<br />
earth, a two half space Green’s function is utilised and for a layered earth, a layered<br />
Green’s function is utilised. A two half space Green’s function doesn’t offer reflections<br />
because of homogeneous conductivity consideration and thus expresses only the case of<br />
downward progressive diffusive waves. The layered Green’s function includes both<br />
downward and upward propagating diffusive waves offered by the layered boundaries.<br />
2. Problem formulation in terms of Maxwell’s equation<br />
2.1 Basics<br />
An electrically conducting fluid like ocean consists of charged particles. The particles in<br />
the ambient geomagnetic field experience a deflective Lorentz force. If ‘ v ’is the velocity<br />
of charge particle ‘q’ moving in the geomagnetic field ‘ B0 ’, then the Lorentz force is:<br />
<br />
q(<br />
v B )<br />
(1)<br />
FL 0<br />
The charge ‘q’ experiences the deflecting force ‘ FL ’ because of the action of an electric<br />
field which we call as Lorentz electric field ( EL ),<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
47
Therefore, L 0<br />
<br />
FL<br />
<br />
qEL<br />
(1-a)<br />
<br />
E<br />
<br />
v B<br />
(2)<br />
The field ‘ EL ’ generates a secondary electric field ‘ E ’, mainly by two processes-<br />
(1) Galvanic process: At locations, where ‘ EL ’ has a component parallel to the<br />
conductivity gradient '<br />
'<br />
( is conductivity in S/m), space/surface charges are<br />
accumulated. The accumulated charges galvanically create a secondary field ‘ E ’,<br />
even if the wave velocity is constant.<br />
(2) Inductive process: The magnetic field ‘ HL <br />
’ is created by the current density<br />
<br />
’ i.e.<br />
‘ JL<br />
<br />
JL<br />
(<br />
v B0<br />
)<br />
<br />
H J<br />
L<br />
L<br />
when the velocity ‘ v ’ changes with time. This induces a secondary electric field E <br />
via the law of induction.<br />
We assume that all the hydrodynamics, including Coriolis force, is included in the given<br />
‘ v ’. We do not care about the sources of forces. Finally, for a stationary frame of<br />
reference, the current density in the Ohms law is given by<br />
<br />
( E E ) ( E v B )<br />
(3)<br />
J L<br />
0<br />
Here, ‘ ’ is the conductivity of the fluid, ‘ E ’ is the current term generated by both a<br />
galvanic and an inductive process and ‘<br />
<br />
( v B ) ’ is the source current term for motional<br />
0<br />
induction case. Under the quasi-static approximation, the set of Maxwell’s equation to be<br />
solved for the motional induction case is<br />
<br />
H (<br />
E v B0<br />
)<br />
<br />
(4- a, b)<br />
E <br />
H<br />
0<br />
t<br />
In an exact formulation, the ambient magnetic field is the sum of geomagnetic field ‘ B0 ’,<br />
the external magnetic field by ionospheric currents ‘ Bext <br />
’, the field generated by small<br />
local anomalies<br />
<br />
‘ B lano<br />
’ and the motionally induced field ‘ Bmi <br />
’ i.e.<br />
<br />
( B<br />
<br />
B<br />
<br />
B <br />
<br />
B ) . In general, the last three terms are orders of magnitude<br />
B 0 ext l ano mi<br />
smaller than the geomagnetic field B0 . Therefore, for computations, the local value of<br />
earth’s magnetic field would be a good choice.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
48
3. A halfspace model<br />
Let the north, east and positive downward directional depth in Cartesian coordinate<br />
system be represented by xˆ , yˆ and zˆ . Consider a two halfspace model with the boundary<br />
at the earth surface (i.e. z 0 ). Let the insulating air halfspace extend in the negative<br />
vertical upward direction while the earth halfspace extends in positive vertical downward<br />
direction from the boundary surface (Fig 1). The conductivity of the earth’s half space is<br />
=3.33 S/m, which contains an ocean and sediments below. In general, the conductivity<br />
of the ocean depends on dissolved ions content and their mobility, which is primarily a<br />
function of temperature and pressure. The conductivity in warm, top shallow water is<br />
normally high (5 S/m) compared to cold, deep water (2.5 S/m). For simplicity, oceans<br />
can be considered homogeneous as the range of conductivity difference (i.e. 2.5–5 S/m) is<br />
small. Consider with depth d=1000 m ocean moving with a velocity v in the x-direction<br />
(Fig 1). Below the ocean, a subsurface of stationary sediments (i.e. v = 0, motionless)<br />
exists.<br />
z=0<br />
+z<br />
z=d<br />
V x<br />
Ocean (v0) (v0) (v0)<br />
, , 0<br />
Sediments (v = 0)<br />
, , 0<br />
Air Half-Space<br />
x<br />
Earth Half-Space<br />
Fig 1. The two halfspace model. z = 0 is the<br />
boundary between earth and air halfspace. Depth is<br />
considered positive in the downward direction. The<br />
earth halfspace consists of two layers. The top layer<br />
represents the ocean having non-zero velocity in<br />
the x-direction. The depth of the ocean is d. Below<br />
the ocean, a second layer of static (i.e. velocity<br />
zero) sediments extends. The conductivity and<br />
magnetic permeability of earth halfspace is and<br />
0 respectively.<br />
<br />
<br />
Let B0 B0<br />
zˆ and v v x ( z,<br />
t)<br />
xˆ<br />
This will generate a horizontal electric field perpendicular to the velocity direction and a<br />
horizontal magnetic field perpendicular to the electric field, and thus we can write:<br />
<br />
<br />
E ( z,<br />
t)<br />
yˆ and H ( z,<br />
t)<br />
xˆ<br />
From eq. (4), we have<br />
E y<br />
2<br />
H x<br />
EBv Assume a harmonic time dependence for simplicity,<br />
~ it<br />
Then E ( z,<br />
t)<br />
E ( z)<br />
e , H<br />
y<br />
y<br />
z Ey 0<br />
t y 0 t x<br />
(5)<br />
x<br />
( z,<br />
t)<br />
x<br />
v<br />
x<br />
( z,<br />
t)<br />
~ it<br />
v ( z)<br />
e ,<br />
it<br />
H ( z)<br />
e<br />
~ <br />
and eq. (5) reduces to<br />
2~<br />
2 ~<br />
E k (z) E (z) - g(z)<br />
(6)<br />
z<br />
y<br />
y<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
49<br />
x
where,<br />
2<br />
g(z) k (z)<br />
~<br />
v x(z)<br />
B0<br />
is the source term, k(<br />
z)<br />
i0(<br />
z)<br />
represents the<br />
electromagnetic damping phenomenon. It is complex such that amplitude decay is<br />
associated with a phase shift with respect to the field at the surface. The Green’s function<br />
G( z | z0<br />
) corresponding to equation (6) is defined by<br />
2<br />
2<br />
G 0<br />
0<br />
0<br />
( z | z ) k (z)G(z | z ) - (z<br />
- z )<br />
(7)<br />
Crosswise multiplication with G and y E<br />
~<br />
and subtraction yields<br />
<br />
~<br />
G<br />
E<br />
~<br />
E <br />
~<br />
G)<br />
- g(z) G E (<br />
z z ) (8)<br />
z ( z y y z<br />
y 0<br />
The integration of (8) over depths z yields the solution<br />
d<br />
~<br />
E y(z0)<br />
g(z) G(z | z0<br />
) dz , for - z0<br />
(9)<br />
0<br />
i.e. a convolution of the source term with the Green’s function. The integration, which<br />
formally must be carried out from minus to plus infinity, reduces to 0 to d, because that is<br />
~<br />
y 0<br />
where the source is nonzero. For the horizontal electric field E (z ) calculation at any<br />
desired depth, knowledge of the Green’s function is required.<br />
3.1 Green’s function<br />
3.1.1. Half space<br />
The Green's function is commonly used to solve inhomogeneous boundary value<br />
problems. The solution by means of Green’s function gives a special advantage because<br />
of its reciprocity property, which states ‘relationship between a oscillating source and the<br />
resulting field at some point of observation is unchanged even if the observation and<br />
source points are interchanged’. Using the boundary conditions, the Green’s function is<br />
calculated for the halfspace. Calculation is as follows:<br />
For an arbitrary small , equation (7) follows,<br />
( z | z ) G(<br />
z | z ) 1<br />
(10)<br />
zG<br />
0 0 z 0 0<br />
The second boundary condition is that G( z | z0)<br />
is continuous at z z0<br />
i.e.<br />
G( z0<br />
| z0<br />
) G(<br />
z0<br />
| z0<br />
) 0<br />
(11)<br />
For the uniform halfspace,<br />
<br />
A e<br />
kz<br />
G ( z | z0<br />
) <br />
Be<br />
kz<br />
for z z0<br />
for z z0<br />
(12-a,b)<br />
A and B are determined from (10) and (11) which yields the Green’s function for the half<br />
space<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
50
3.1.2 Two halfspaces<br />
G(<br />
z | z<br />
0<br />
)<br />
1<br />
2k<br />
k|<br />
zz<br />
|<br />
0<br />
e<br />
(13)<br />
In order to consider the diffusive reflection and transmission effects due to boundaries,<br />
let us consider two uniform half spaces with k( z)<br />
k 0 in the air (i.e. z0). The Green’s function, for z0>0, can be written as<br />
G(<br />
z | z<br />
1 k | z z<br />
e<br />
) <br />
2k<br />
k z<br />
T e 0<br />
<br />
0<br />
|<br />
R e<br />
kz<br />
for z 0<br />
0 (14)<br />
for z 0<br />
where, T and R are respectively transmission and reflection coefficients. The continuity of<br />
G and zG<br />
at z 0 yields<br />
1 k | z z | k - k k(<br />
z z ) <br />
e<br />
0<br />
0 e<br />
0<br />
<br />
2k<br />
<br />
k k0<br />
G(<br />
z | z ) <br />
<br />
1<br />
<br />
k z kz<br />
e 0 0<br />
k k<br />
0<br />
for z<br />
0,<br />
z<br />
for z 0,<br />
z<br />
0<br />
0<br />
0 (15)<br />
In particular, for the insulating air halfspace k 0 0 , the two halfspace Green’s functions<br />
take the form<br />
1 k | z z | k(<br />
z z ) <br />
<br />
e<br />
0 e<br />
0<br />
<br />
G(<br />
z | z ) 2k<br />
1<br />
<br />
kz<br />
e 0<br />
<br />
k<br />
for z 0,<br />
z<br />
for z 0,<br />
z<br />
0<br />
0<br />
0<br />
0<br />
0<br />
0 (16)<br />
3.1.3. A simple model and field expression<br />
Equation (9) includes the source term g(z), which depends on conductivity and velocity,<br />
both are depth dependent. For simplicity, consider a uniform horizontal velocity i.e.<br />
v(z) v0<br />
, 0 z d . Let the depth dependent conductivity be zero in the air and<br />
3.<br />
33 S / m for the earth i.e.<br />
(z)<br />
<br />
<br />
, z 0<br />
0 , z 0<br />
The equation (9), (16) and (4), yields the horizontal electric and magnetic field expression<br />
for the full space.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
51<br />
0
Electric field in full space:<br />
~<br />
Ey<br />
( z0<br />
) v<br />
~<br />
Ey<br />
( z0<br />
) v<br />
~<br />
E ( z ) v<br />
Magnetic field in full space:<br />
H ~<br />
H ~<br />
H ~<br />
y<br />
x<br />
x<br />
x<br />
0<br />
( z<br />
( z<br />
( z<br />
0<br />
0<br />
0<br />
0<br />
0<br />
0<br />
) 0<br />
B ( 1 e<br />
B ( 1 e<br />
B<br />
0<br />
0<br />
0<br />
e<br />
kv0<br />
B<br />
) <br />
i<br />
kv0<br />
B<br />
) <br />
i<br />
kz<br />
0<br />
0<br />
0<br />
0<br />
kd<br />
kd<br />
)<br />
cosh( kz<br />
))<br />
, z<br />
0<br />
,0 z<br />
0 sinh( kd)<br />
, z d<br />
e<br />
e<br />
kd<br />
kz<br />
sinh( kz<br />
0<br />
0<br />
)<br />
, z<br />
0 sinh( kd)<br />
, z d<br />
0<br />
0<br />
0<br />
0<br />
,0 z<br />
0<br />
0<br />
0<br />
d<br />
d<br />
(17-a, b, c)<br />
(18-a, b, c)<br />
The graphical response of (17) and (18) is shown in Fig (2). The response is calculated for<br />
a halfspace model (Fig 1) of a conductivity of 3.33 S/m containing 1000 m thick layers of<br />
ocean and sediments. The magnetic permeability is kept constant for both halfspace and<br />
is equal to free space permeability i.e. 0 =410 -7 Vs/Am. The ocean has a homogeneous<br />
velocity of 10 cm/s in an ambient geomagnetic field of 510 -5 T. The responses are<br />
studied for frequencies 0.001, 0.01, 0.1, and 1 Hz. In general, the electric field becomes<br />
gradually weaker with depth (Fig 2). For 1 Hz, rather than weakening gradually, the<br />
electric field amplitude increases and becomes strongest with in the ocean (at 400 m<br />
depth). Further, the field Bx is zero at the surface and progressively becomes stronger<br />
with respect to depth. At the ocean bottom, it offers strongest amplitude. The zilch of Bx<br />
~<br />
at the earth surface is for the reason that in air halfspace field Ey<br />
is constant and<br />
~<br />
z y<br />
therefore E 0 in particular, which is BBx (eq. 4-b). Further, as far as frequency based<br />
variation are concerned, the smallest frequency Bx produces the strongest amplitude at<br />
the ocean bottom. The field Ey shows a contrary behaviour. Here the smallest frequency<br />
offers the weakest amplitude at the ocean floor. At large frequencies the field Ey can not<br />
reach to the ocean floor, because of the shallow skin depth, offers constant amplitude<br />
(Fig 3).<br />
The important result in Fig 2 & 3 is that the field Ey is smooth over the boundary<br />
between the ocean and sediment. On the other hand, the field Bx offers a sharp change in<br />
pattern there (at boundary). Note that the conductivity of the ocean and subsurface<br />
(sediments) are identical. They differ only in their velocity state (subsurface is static and<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
52
ocean is dynamic). Therefore the sensitivity of the field Bx at the boundary is interpreted<br />
as result of velocity change.<br />
Depth, (m)<br />
0<br />
-1000<br />
-2000<br />
0 1 2 3 4 5 6<br />
E (V/m)<br />
y<br />
0.001 Hz 0.01 Hz 0.1 Hz 1 Hz<br />
0<br />
Depth, (m)<br />
-1000<br />
-2000<br />
0 5 10<br />
B (nT)<br />
x<br />
15 20<br />
Fig 2. Variation of the horizontal comp. electric<br />
(Ey) and magnetic (Bx) field with respect to<br />
depth. Variation is shown for four frequencies viz.<br />
0.001, 0.01, 0.1 and 1 Hz. An ocean of conductivity<br />
=3.33 S/m extends from surface to 1000 m depth<br />
(i.e. 0 z 1000 m). Below the ocean (i.e. below<br />
1000 m depth) there is a sediment layer with the<br />
same conductivity as the ocean (i.e. =3.33 S/m).<br />
The horizontal line at 1000 m is to show the<br />
separation of these two boundaries.<br />
Depth, (m)<br />
0<br />
-1000<br />
-2000<br />
0 1 2 3 4 5 6<br />
E (V/m)<br />
y<br />
2 Hz 10 Hz 50 Hz<br />
Depth, (m)<br />
0<br />
-1000<br />
-2000<br />
0 0.5<br />
B (nT)<br />
x<br />
1 1.5<br />
Fig 3. Variation of the horizontal component<br />
electric (Ey) and magnetic (BBx) field with<br />
respect to depth. Variation is shown for three<br />
frequencies viz. 2, 10 and 50 Hz. Other information<br />
is the same as in Fig 2. It is evident thet at the ocean<br />
floor, the field Ey offers almost same amplitude for<br />
all the three frequencies. Bx is strongest and<br />
weakest respectively for the smallest and largest<br />
frequency.<br />
The field Ey (17-b) with in the ocean (i.e. 0 z0<br />
d ) consists of two terms. The first<br />
term is the kernel for the motional induction, while the second governs the depth<br />
dependent frequency based EM damping. From the expressions (17-b & 18-b), it is clear<br />
that the strength of both the fields Ey and BBx depends on the ocean depth (i.e. ocean<br />
-kd<br />
deepening) because of the factor e . Effect of the ocean deepening is studied for oceans<br />
of thickness (depth) varying from 1000 m to 9000 m. Observation depth is constant for all<br />
practical cases and is 1000 m (i.e. z0<br />
= 1000 m). Results are shown in Fig (4). Ey is<br />
strongest for the thickest ocean and weakens down gradually as ocean shallows up. The<br />
reverse is observed for the Bx case with weakest strength in the deepest ocean which<br />
progressively becomes stronger as the ocean gradually shallows up. The sensitivity for<br />
the deepening effect depends on frequency and therefore on skin depths. For that reason,<br />
the deepening effect is more effective at smaller frequencies. The skin depths, for a<br />
halfspace of 3.33 S/m, at frequencies 0.001, 0.01, 0.1, 1 and 10 Hz are approximately<br />
8664, 2739, 866, 273 and 86 m, respectively.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
53
The factor which may significantly influence the motional induction is the<br />
conductivity of the ocean. The results of the conductivity variation are shown in Fig 5.<br />
The field strength is observed at 1000 m depth for conductivities varying from 3 to 5 S/m.<br />
Observations are made for five different frequencies (i.e. 0.001, 0.01, 0.1, 1 and 10 Hz).<br />
The results suggest that the field Bx is more sensitive to conductivity variation than the<br />
field Ey. Further, lower frequencies are more responsive to conductivity variation than<br />
high ones because of the thicker skin-depth.<br />
Various Ocean Depths (m)<br />
9000<br />
8000<br />
7000<br />
6000<br />
5000<br />
4000<br />
3000<br />
2000<br />
1000<br />
10 0<br />
0.001 Hz 0.01 Hz 0.1 Hz 1 Hz 10 Hz<br />
E y (V/m)<br />
Various Ocean Depths (m)<br />
9000<br />
8000<br />
7000<br />
6000<br />
5000<br />
4000<br />
3000<br />
2000<br />
1000<br />
10 -5<br />
B x (nT)<br />
Fig 4. Deepening effect of an ocean: Conductivity<br />
of ocean kept constant (i.e. 3.33 S/m). For 1000 to<br />
9000 m deep ocean the field strengths are calculated<br />
at frequencies 0.001, 0.01, 0.1, 1 & 10 Hz. The y- and<br />
x-axis represents ocean depths and field strength<br />
respectively. The observation depth is constant and is<br />
1000 m. The magnetic component for 1 and 10 Hz is<br />
truncated for depths greater than 5000 and 2000 m<br />
respectively to save the Figure from masking and<br />
cluttering.<br />
10 0<br />
Conductivity (S/m)<br />
5<br />
4.8<br />
4.6<br />
4.4<br />
4.2<br />
4<br />
3.8<br />
3.6<br />
3.4<br />
3.2<br />
0.001 Hz 0.01 Hz 0.1 Hz 1 Hz 10 Hz<br />
3<br />
0.5 1 1.5 2 2.5 3<br />
E (V/m)<br />
y<br />
Conductivity (S/m)<br />
5<br />
4.8<br />
4.6<br />
4.4<br />
4.2<br />
4<br />
3.8<br />
3.6<br />
3.4<br />
3.2<br />
3<br />
0 5 10 15<br />
B (nT)<br />
x<br />
20 25 30<br />
Fig 5. Conductivity variation effect: The<br />
conductivity of the ocean is varied from 3 to 5 S/m<br />
and the effect is studies for frequencies 0.001,<br />
0.01, 0.1, 1 & 10 Hz. The horizontal component<br />
electric and magnetic field strength, observed at<br />
1000 m ocean depth is shown w.r.t. conductivities<br />
for the chosen frequencies. Electric and magnetic<br />
components both are sensitive to the conductivity<br />
of the ocean. Sensitivity increases with the<br />
decreasing frequency.<br />
Equation (17) and (18) indicate that the field BBx is more sensitive to conductivity<br />
variation in vertical direction, as conductivity is effectively convolved, than the field Ey.<br />
But it would not be appropriate to discuss conductivity variation in vertical direction<br />
using a halfspace model. Therefore, we will be back on this issue with a justified layered<br />
model and a layered Green’s function.<br />
Moreover, the equation (17) and (18) implies that the increase in the wave<br />
velocity v0 will cause a constant shift in the Ey<br />
and Bx field. Three experiments are<br />
conducted for wave velocities of 10, 1 and 0.1 cm/s to study the effects. The results are<br />
shown in the Fig 5. A log scale is used for x-axis plotting for clarity reasons. As at the<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
54
surface Bx=0, therefore the log scale plot starts from 100 m depth as 100 m is the 2 nd<br />
discretized layer of the ocean. The selection of velocities is such that the 1 st selection<br />
exceeds the 2 nd by factor of 10 and same is true for 3 rd and 2 nd . Evidently, response for<br />
the field Ey and Bx also exceeds by a static shift of factors of 10. The graph clearly<br />
illustrates that the velocity is an important parameter for practical simulation, any bias in<br />
velocity leads to the same bias in EM field.<br />
D epth, (m )<br />
0<br />
-1000<br />
10 -2<br />
-2000<br />
10 -1<br />
E y (V/m)<br />
0.1 m/s 0.01 m/s 0.001 m/s<br />
10 0<br />
10 1<br />
D epth, (m )<br />
0<br />
-1000<br />
10 -2<br />
-2000<br />
10 0<br />
B x (nT)<br />
Fig 6. Wave Velocity effect. Semi-log plot of<br />
horizontal electric and magnetic field component<br />
illustrating the effect of velocity change. The<br />
strongest velocity generates the strongest EM field.<br />
The Bx plots starts from 100 m depth as the strength<br />
of Bx at z=0 is zero and at z=100 m the next layer<br />
starts.<br />
4. Layered model<br />
4.1. Layered Green’s function<br />
10 2<br />
Depth (m )<br />
-0<br />
-1.000<br />
-2.000<br />
Let us consider a layered Green’s function defined as<br />
f=0.001 Hz f=0.01 Hz f=0.1 Hz f=1 Hz<br />
-3.100<br />
0 2 4 6<br />
Ey (Volt/meter)<br />
Depth (m )<br />
-0<br />
-1.000<br />
-2.000<br />
Ocean<br />
Sediments<br />
Reservoir<br />
Sediments<br />
-3.100<br />
0 5 10<br />
Bx (nT)<br />
15 20<br />
Fig 7. Variation of the horizontal component<br />
electric (Ey) and magnetic (Bx) field with<br />
respect to depth for a layered model. Variation is<br />
shown for four frequencies viz. 0.001, 0.01, 0.1 and<br />
1 Hz. An ocean of conductivity =3.33 S/m extends<br />
from the surface to 1000 m depth (i.e. 0 z 1000<br />
m). Below the ocean there, is a 1000 m thick<br />
sediments layer with conductivity =1 S/m. A 100<br />
m thick reservoir of conductivity 0.01 S/m is<br />
embedded at 2000 m depth. The horizontal line<br />
marks the boundary of different formation.<br />
1 k | z z | k z k z<br />
G( z | z )<br />
<br />
e 1 0 <br />
R e 1 <br />
R e 1<br />
0 <br />
<br />
0<br />
d ; 0 z d,<br />
0 z0<br />
d<br />
2k<br />
<br />
<br />
<br />
<br />
1<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
55
where, k1 i0<br />
is the electromagnetic damping in the ocean, R 0 describes the field<br />
1<br />
diffusing downward by reflection from the air-earth interface and R d describes the field<br />
diffusing upward by reflection from the seafloor (i.e. z d ). In this particular case the<br />
constants R 0 and R d are determined from boundary conditions:-<br />
a)<br />
( 0 | z ) 0 , Since ( z | z ) is constant in the air halfspace<br />
zG<br />
0<br />
G 0<br />
b) ( d | z ) b ( d)<br />
G(<br />
d | z ) 0 , where, b2 ( d)<br />
is determined via continuous<br />
transfer function.<br />
They yield,<br />
R<br />
0<br />
with<br />
zG<br />
0 0 2<br />
0<br />
1 R ce<br />
<br />
1 R e<br />
R<br />
c<br />
2k<br />
( d z<br />
)<br />
c<br />
0b<br />
<br />
b<br />
0<br />
2k<br />
d<br />
2<br />
2<br />
1<br />
1<br />
1 1<br />
k<br />
k<br />
0<br />
1<br />
1<br />
e<br />
k<br />
z<br />
1 0<br />
4.2 Layered model and Response<br />
;<br />
R<br />
d<br />
R c ( 1 e<br />
<br />
1 R e<br />
c<br />
2k<br />
z<br />
1 0<br />
2k<br />
d<br />
1 1<br />
)<br />
e<br />
k<br />
( 2d<br />
z<br />
)<br />
EM field responses calculated for a layered model are shown in Fig (6). The parameters<br />
of the model are tabled in table 1. The magnetic permeability of the each layer is equal to<br />
free space permeability (0). The response is computed for four frequencies viz. 0.001,<br />
0.01, 0.1 and 1 Hz. It is evident that the field Ey varies smoothly over layer boundaries,<br />
indicating its sensitivity to the vertically averaged conductivity rather than conductivity<br />
variation at layered boundaries. However, the field Bx senses each layer and offers a<br />
change in field strength at each layer boundary. Evidently, the magnetic field and its<br />
sensitivity for conductivity variation is frequency dependent. At a small frequency the<br />
field is strong and vice versa.<br />
Table 1: Layered Model<br />
Layers<br />
Thickness<br />
(m)<br />
Conductivity<br />
(S/m)<br />
Ocean 1000 3.33<br />
Sediments 1000 1<br />
Reservoir 100 0.01<br />
Below 1000 1<br />
1<br />
Other Parameters<br />
Wave Velocity=0.1 m/s<br />
Ext. Magnetic Field=510 -5 T<br />
Magnetic Permeability of each layer<br />
1<br />
=410 -7<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
56<br />
0
5. Discussion and Conclusion<br />
No doubt, the actual earth is 3D and therefore the electromagnetic response of the earth<br />
will always be more complex than the response of any simplified model. However, a<br />
simplified model, roughly representing actual earth layers can be used to validate results.<br />
The choice of the model dimension (i.e. 1D, 2D or 3D) sometimes has a large impact on<br />
the result. In particular, for mCSEM surveys, generally 2D data acquisition methodology<br />
is followed and therefore a 2D/3D modelling may lead to a better interpretation.<br />
However, 1D model formulations are simple and give good insight into the physics of the<br />
problems.<br />
Field strength depends on their components (depending on the physical process<br />
involved), whose production mainly depends on the involvement of the source<br />
components. For example, a 1D model in motional induction study results in Ey and Bx<br />
(two) components when a 1D source (i.e. wave-velocity in x-direction) is considered.<br />
However, a 2D source (i.e. wave-velocity in x-direction and a wave wavelength in ydirection)<br />
consideration results in Ey, Bx and Ez (three) components. This suggests, even a<br />
1D model, with a proper source formulation offers significant insight to the problems.<br />
Further, the simplicity of a 1D model is an important advantage. In this paper we<br />
thoroughly looked for the sources of the motional induction in the ocean, which is<br />
incorporated in 1D model for theoretical development. The horizontal (x-direction) wave<br />
motion and a vertical (z-direction) geomagnetic field consideration lead to the excitation<br />
of the field components ‘Ey’ and ‘Bx’. These fields illustrate some of important results and<br />
effects of the ocean dynamics in varying oceanic conditions.<br />
In general, for a uniform halfspace, there is gradual reduction and increase<br />
respectively in the strength of Ey and Bx. w.r.t. ocean depth (Fig 2). This conclusion is<br />
valid for frequencies with skin depth ‘’ greater than the ocean depth ‘d’. For the case<br />
when < d, Ey may show maximum amplitude somewhere in the ocean, rather than at<br />
the ocean surface. On the other hand in the ocean the field Bx always offers maximum<br />
strength at the floor and minimum (i.e Bx = 0) at the surface. Since Bx vanishes at the<br />
surface, the mCSEM surface measurement may offer significant noise-free signals there,<br />
if MT field is avoided. At the ocean floor as Bx always has maximum strength therefore it<br />
is necessary to correct mCSEM data for motionally induced field. It is evident from Fig<br />
(2), even in a halfspace where the conductivity of ocean and subsurface is same, Bx<br />
senses the boundary of the ocean and subsurface (i.e. sediments) though Ey does not see<br />
it. Still, an important question left to answer is ‘which field is/are sensitive to layer<br />
demarcation, either Ey or Bx or both?’ This issue will be conferred below later, in the<br />
paragraph with layered earth discussion.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
57
The fields Bx and Ey both are sensitive (Fig 5) to the conductivity variation of the<br />
oceans. The observations with several oceans differing in their conductivity suggests that<br />
at frequencies with < d, in general, the variation of ocean conductivity is negligible but<br />
at frequencies with > d, the more conductive ocean increases the strength of Ey and Bx<br />
field.<br />
The average depth of Atlantic, Pacific and Indian Ocean is respectively 3926 m,<br />
4282 m and 3963 m. Moreover, from place to place the oceanic depths are quite diverse<br />
and therefore the significance of deepening of the ocean is studied (Fig 4). For a fixed<br />
observation depth in ocean, the gradual increase in the strength of Ey and decrease in the<br />
strength of Bx is the consequence of deepening of the ocean. The component Ey and Bx<br />
show strongest and weakest strength respectively for the deepest ocean and vice-versa.<br />
For frequencies for which the skin depth is smaller than the ocean depth (i.e. < d), the<br />
further deepening of the ocean (greater than skin depth) is ineffective and therefore the<br />
field value may saturate with further deepening.<br />
Another important parameter is the ocean wave velocity. The study suggests (Fig<br />
5) that any bias in velocity value may cause same bias in field strength. The change in<br />
velocity does not modify the pattern with depth, but causes a static shift in the field.<br />
Expression for layered Green’s function has allowed us to study the EM field<br />
variation for different layers with in the earth. The layered model involves both resistive<br />
and conductive layers. Clearly, electric field Ey does not see the layered boundaries (Fig<br />
6) although the magnetic field clearly sees it by offering a change in the slope of Bx at the<br />
boundaries.<br />
Acknowledgement<br />
We thank KMS Technologies - KJT Enterprises for sponsoring the work.<br />
References<br />
1. Chave, A.D., and C.S. Cox, Controlled electromagnetic sources for measuring<br />
electrical conductivity beneath the oceans, 1, forward problem and model study, J.<br />
Geophys. Res., 87, 5327-5338, 1983.<br />
2. Chave, A.D., and D.S. Luther, Low-frequency, motionally induced electromagnetic<br />
fields in the ocean, 1, theory, J. Geophys. Res., 95, 7185-7200, 1990.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
58
3. Crews, A., and J . Futterman, Geomagnetic micropulsation due to the motion of ocean<br />
waves, J. Geophys. Res., 67, 299-306, 1962.<br />
4. Faraday, M., Bakerian Lecture-Experimental researches in electricity, Phil. Trans.<br />
Roy. Soc. London, Part 1, 163-177, 1832.<br />
5. Longuet-Higgins, M.S., E. Stern, H. Stommel, The electrical field induced by ocean<br />
currents and waves with application to the method of towed electrodes, Pap. Phys.<br />
Oceanog. Meteorol., 13(1),1-37, 1954.<br />
6. Podney, W., Electromagnetic fields generated by ocean waves, J. Geophys. Res., 80,<br />
2977-2990, 1975.<br />
7. Sanford, T.B., Motionally induced electric and magnetic fields in the sea, J. Geophys.<br />
Res., 76, 3476-3492, 1971.<br />
8. Young, F.B., H. Gerrard and W. Jevons, On electrical disturbances due to tides and<br />
waves, Phil. Mag. Ser. 6, 40,149-159.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
59
Three-dimensional finite element simulation of<br />
magnetotelluric fields incorporating digital elevation models<br />
S. Kütter, A. Franke-Börner, R.-U. Börner, K. Spitzer<br />
Institut für Geophysik, TU Bergakademie Freiberg, Gustav-Zeuner-Str. 12, 09599 Freiberg<br />
1 Introduction<br />
Pronounced topography may have an important impact on the magnetotelluric (MT) response and its<br />
neglect may lead to severe misinterpretations. Common simulation techniques based on rectangular<br />
grids are generally not well suited to deal with arbitrary geometry. We therefore use vector finite<br />
element (FE) schemes formulated on unstructured grids to cope with realistic topography and/or<br />
bathymetry. As an example, we have chosen the volcanic island of Stromboli which is located in the<br />
Mediterranean Sea off the west coast of Italy. Stromboli is an extreme electric environment with very<br />
conductive sea water surrounding the steep topography of the resistive volcanic edifice. Moreover, the<br />
varying bathymetry and the topography of the other islands of the Liparian archipelago cause distinct<br />
effects on the magnetotelluric response. Due to the large conductivity contrast and these extraordinary<br />
features it can be expected that the magnetotelluric apparent resistivities as well as the phases show a<br />
very complex behaviour. The objective of this work is to incorporate digital elevation models of this area<br />
into our numerical MT simulations allowing for a realistic look at the electromagnetic (EM) induction<br />
phenomena in such a complicated environment. On top of its challenge for numerical simulation<br />
methods this volcano has always fascinated geoscientists (Fig. 1) and future interdisciplinary studies<br />
aim at investigating the inner structure and the processes that lead to the continuous mild eruptions<br />
recorded over the last 2 000 years (the so called ’Strombolian activity’).<br />
Figure 1: Map of the Aeolian Islands (left, Wikipedia (2010)) and Stromboli (right, SwissEduc (2010))<br />
2 Physical and numerical basics<br />
Based on Maxwell’s equations and assuming a harmonic time dependency e iωt of the incoming plane<br />
wave, the equation of induction in terms of the vector potential A reads as<br />
∇×μ −1 (∇×A)+(iωσ − ω 2 ε)A = 0, (1)<br />
where μ, σ, ε, ω and i are the magnetic permeability, the electric conductivity, the permittivity, the<br />
angular frequency and the imaginary unit, respectively. The magnetic field H and the electric field E<br />
are obtained by<br />
H = μ −1 (∇×A) and E = −iωA −∇V . (2)<br />
The scalar potential V was eliminated from eq. (1) by the gauge condition ψ = −iV/ω and substituting<br />
A for A −∇ψ which leaves H and E unchanged.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
60
To calculate A in a bounded domain Ω ⊂ R 3 , electric and magnetic insulation are required for all outer<br />
boundaries parallel (Γ ||) and perpendicular (Γ⊥) to the current flow, respectively:<br />
n × H = 0 on Γ ||,<br />
n × A = 0 on Γ⊥.<br />
Depending on whether the condition of electric insulation is applied to the boundaries parallel to the<br />
x- ory-direction, xy- oryx-polarisation data are obtained, respectively. At the top (Γtop) and bottom<br />
boundaries (Γbottom) we impose boundary values in terms of the magnetic field induced in a horizontally<br />
layered half-space (Wait (1953)):<br />
H⊥ =1Am −1<br />
on Γtop,<br />
H⊥ = Hj(z), j = x, y on Γbottom.<br />
For the construction of the full 3-D MT impedance tensor, both xy- andyx-polarization data are<br />
required. Arranging the x- andy-components of the electric and magnetic fields obtained for xy- and<br />
yx-polarization such that<br />
<br />
Ex(xy) Ex(yx)<br />
E =<br />
(5)<br />
H =<br />
Ey(xy) Ey(yx)<br />
Hx(xy) Hx(yx)<br />
Hy(xy) Hy(yx)<br />
<br />
, (6)<br />
the impedance tensor Z can the be obtained by Z = EH −1 . The off-diagonal elements of the impedance<br />
tensor are therefore<br />
Zxy = Ex(yx)Hx(xy) − Ex(xy)Hx(yx)<br />
Hx(xy)Hy(yx) − Hx(yx)Hy(xy)<br />
(3)<br />
(4)<br />
, Zyx = Ey(xy)Hy(yx) − Ey(yx)Hy(xy)<br />
. (7)<br />
Hx(xy)Hy(yx) − Hx(yx)Hy(xy)<br />
For any period T =2π/ω, the apparent resistivity ρa and phase φ can then be calculated by<br />
ρxy = 1<br />
ωμ |Zxy| 2 , ρyx = 1 2<br />
|Zyx|<br />
ωμ<br />
and φxy = arg(Zxy), φyx = arg(Zyx). (8)<br />
Inthefollowing,weusethetermsZxy mode or Zyx mode to distinguish xy- andyx-polarization<br />
from quantities derived by a combination of both (Nam et al. (2007)). An important property of<br />
time-harmonic EM field is the skin depth δ ≈ 503 ρ/f, whereδ, the resistivity ρ and the frequency<br />
f are given in [m], [Ω m] and [Hz], respectively. is determined, whereas δ, the resistivity ρ and the<br />
frequency f are given in [m], [Ω m] and [Hz], respectively.<br />
To solve the boundary value problem (1) - (4), the finite element method (FEM) is applied on unstructured<br />
tetrahedral meshes (see Börner (2010); Schwarzbach (2009)). Tetrahedral meshes are well<br />
suited for the spatial discretization of arbitrary 3-D geometries which occur when digital elevation<br />
models have to be incorporated into numerical simulations. In the 3-D simulations presented here,<br />
curl-conforming Nédélec elements with second-order basis functions are employed. In all experiments the<br />
FE discretization has been carried out using the Electromagnetics Module of the COMSOL Multiphysics<br />
package.<br />
3 Description of the digital elevation model<br />
The area used in the simulations extends from 38.4° to 39.2° N and 14.7° to 15.7° E. At 39° N, the<br />
distance between two degrees of latitude is 110.95 km, whereas the distance between two degrees of<br />
longitude is 86.51 km. Hence, the model area has a length of 86.51 km in the east-west (x) and 88.76 km<br />
in the north-south (y) direction.<br />
We have used two sets of digital terrain data. The ETOPO1 data set is a 1 arc-minute model and<br />
provides elevation values for both land and sea. It is available online at the National Geophysical Data<br />
Center (http://www.ngdc.noaa.gov) and gives a good approximation for the regional bathymetry. Since,<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
61
however, its spatial data density is not sufficient to describe Stromboli’s topography, a second data set<br />
has been used. These data are available from the Shuttle Radar Topography Mission (SRTM), a project<br />
to obtain high-resolution topographic data (http://srtm.csi.cgiar.org/). The horizontal resolution of<br />
the two data sets are listed in the following table. The digital elevation model for the area considered<br />
here has been obtained by blending both ETOPO1 and SRTM data sets.<br />
resolution x-direction resolution y-direction<br />
SRTM 0.0721 km 0.0925 km<br />
ETOPO1 1.4419 km 1.8491 km<br />
Interpolating the bathymetry data and fitting the two data sets along the coastlines leads to a digital<br />
elevation model as depicted in Fig. 2.<br />
φ [ o ]<br />
39.1<br />
39<br />
38.9<br />
38.8<br />
38.7<br />
38.6<br />
38.5<br />
38.4<br />
14.8 14.9 15 15.1 15.2<br />
λ [<br />
15.3 15.4 15.5 15.6 15.7<br />
o ]<br />
Figure 2: Data points obtained from ETOPO1 (coarse grid) and SRTM (fine grid resembling the islands) for<br />
bathymetry and topography (left), digital elevation model obtained by smooth interpolation onto the<br />
SRTM grid (right). Stromboli island is located at the centre of both images.<br />
4 Numerical studies<br />
The geometry of the volcanic island of Stromboli is incorporated into the simulation models in three<br />
different levels of increasing complexity. This approach enables us to better differentiate which part of<br />
the model has an influence on the electromagnetic fields and how distinct this influence is. The first<br />
model uses a frustum as the volcanic edifice embedded in a layered background consisting of an air<br />
layer, a sea layer and a substratum. In the second model the frustum is replaced with the topography<br />
of Stromboli. The edges of this topographic surface have been adjusted to the sea-floor. The third,<br />
most complex model uses both topography and bathymetry data.<br />
For the numerical computations three different computer architectures have been used:<br />
computer name architecture RAM<br />
klio 4Intel(R)Xeon(R)@3GHz 16 GB<br />
erato 4 Quad-Core AMD Opteron @ 2.5 GHz 128 GB<br />
RM 56 nodes based on Intel Xeon (E5450 @ 3 GHz) 16 GB per 8-core blade<br />
For each of the three models, apparent resistivities ρa and phases φ were calculated along profiles<br />
running along the sea-floor and over the slopes of the volcano for a period of T =10 3 s. Furthermore,<br />
sounding curves were determined for selected sites on the sea-floor and on the volcano to demonstrate<br />
the dependency of ρa and φ on frequency.<br />
4.1 Frustum model<br />
The geometry and appropriate conductivities of the frustum model as well as a map of the locations<br />
of the sounding sites are depicted in Fig. 3. Since there are large conductivity contrasts between the<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
62
frustum, the sea and the air layer, a non-trivial behaviour of the electromagnetic fields is to be expected.<br />
The electric currents are mainly induced in the conductive sea layer and are compressed above the<br />
slopes. Moreover, the currents tend to concentrate below the sea surface due to the skin effect.<br />
Due to horizontal conductivity contrasts, electric charges accumulate at the interfaces between the<br />
volcano and the sea layer. Their surface density is proportional to −E ·∇σ. Therefore, an additional<br />
electric field is generated which leads to an increase of the total electric field towards the top of the<br />
volcano (around x = ±0.5 km in the profile, Fig. 4) and to a decrease towards the sea-floor (around<br />
x = ±5.6 km in the profile). This static shift effect results in discontinuities of the apparent resistivity<br />
at the base points of the frustum on the sea-floor, at the coastline points and at the top of the volcano.<br />
Unlike the apparent resistivity, the phase has a more continuous behaviour and appears to be a<br />
rather robust parameter even for pronounced topography. 40 km away from the center of the volcano<br />
(x, y = ± 40 km), both phase and apparent resistivity reach their values expected for the half-space<br />
(45°, 100 Ω m).<br />
-20<br />
[km] 0<br />
2<br />
air: 10 −14 Sm −1<br />
volcano: 0.02 Sm −1<br />
half-space: 0.01 Sm −1<br />
sea: 5 Sm −1<br />
32<br />
-40 -5.6 0<br />
[km]<br />
5.6 40<br />
<br />
<br />
<br />
<br />
Figure 3: Vertical cross-section through the frustum model (left) and plan view of the model domain (right)<br />
showing the locations of the sounding curves (T, N40, E40).<br />
To verify our results, two independent numerical codes have been used to provide reference datasets<br />
for the frustum model. More precisely, data have been obtained using a finite element package (FEP)<br />
by Schwarzbach (2009), and a finite difference package (FD) by Mackie et al. (1994).<br />
A comparison of the apparent resistivities with those obtained by FEP reveals that they slightly differ<br />
near the boundaries of the model and at the lower part of the volcano slopes (Fig. 4, left), where ρa is<br />
lower for FEP. Moreover, the peaks at the base of the frustum, at the coastline points and at the edges<br />
of the plateau are even more pronounced for FEP. Here, we assume an influence of the different spatial<br />
discretization of the model. The curves of the phase are nearly congruent for both FE codes.<br />
Apparent resistivities and phases obtained by FD (Fig. 4, right) show similar features in general.<br />
However, there are significant differences along the slopes of the volcano. We attribute this to the<br />
fact that in the FD code considered here electric field components are located perpendicular to cell<br />
faces of a rectangular cartesian grid. Problems in interpolating the discrete electric field components at<br />
the desired positions along the volcano slopes are thus to be expected. Finite element techniques on<br />
unstructured grids clearly show their superiority in this respect.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
63
[km]<br />
ρ a [Ω m]<br />
φ [ o ]<br />
−2<br />
0<br />
2<br />
4<br />
10 2<br />
10 1<br />
10 0<br />
50<br />
COMSOL<br />
FEP<br />
0<br />
−40 −30 −20 −10 0<br />
x [km]<br />
10 20 30 40<br />
Figure 4: Comparison of the profile curves of the frustum model, T =10 3 s.<br />
ρ a [Ω m]<br />
φ [ o ]<br />
10 2<br />
10 1<br />
10 0<br />
50<br />
COMSOL<br />
FD<br />
0<br />
−40 −30 −20 −10 0<br />
[km]<br />
10 20 30 40<br />
Fig. 5 depicts three sounding curves for the sites on top (point T in Fig. 3) and 40 km away from the<br />
volcano (E40, N40 in Fig. 3).<br />
Sounding curves for the volcano site T are displayed for periods between 10 −3 and 10 3 s (Fig. 5, left).<br />
For short periods, where the skin depth is small (δ ≈ 355 m for 50 Ω m and 10 −2 s), the resistivity of the<br />
volcano is reproduced and the phase reaches values around 45°. For periods between 10 −1 sand10 2 s,<br />
the influence of the conductive sea layer is dominant and therefore, the apparent resistivity decreases<br />
whereas the phase increases. For longer periods the resistive half-space dominates the curves (δ ≈ 16 km<br />
for 100 Ω m and 10 3 s). The sounding curves for the two sites on the sea-floor are displayed for a subset<br />
of the full period range (between 10 1 and 10 3 s) because of the strong attenuation within the very<br />
conductive sea water (Fig. 5). High frequencies generally require a much finer spatial discretization in<br />
regions where the skin depth is small. However, the computational experiments have been carried out<br />
on a mesh well suited for mid to low frequencies. Results for high frequencies were not satisfactory for<br />
the coarse meshes used here, and thus have not been included in the comparison.<br />
The long period ends of the sounding curves nicely approach the values of the lower half-space, i.e.<br />
100 Ωm for ρa and about 45° for φ. For short periods there are deviations from these values, which are<br />
larger for the points on the profiles oriented perpendicular to the component of the driving electric<br />
field, i.e. for point E40 at y =40kmforρxy and the point at x =40kmforρyx (cross markers in Fig. 5,<br />
cf. Fig. 3, right). As mentioned above, these deviations are due to the insufficient spatial discretization<br />
for short periods in regions of strong attenuation.<br />
ρ a [Ω m]<br />
φ [ o ]<br />
10 2<br />
10 1<br />
10 0<br />
100<br />
50<br />
10 −3<br />
0<br />
10 −2<br />
10 −1<br />
10 0<br />
T [s]<br />
10 1<br />
xy−polarisation<br />
yx−polarisarion<br />
10 2<br />
10 3<br />
ρ a [Ω m]<br />
10 3<br />
10 2<br />
10 1<br />
10 0<br />
100<br />
50<br />
10 1<br />
0<br />
10 2<br />
T [s]<br />
xy−pol. (x)<br />
xy−pol. (y)<br />
yx−pol. (x)<br />
yx−pol. (y)<br />
Figure 5: Sounding curves for the volcano site T (left) and the sea-floor sites E40 and N40 (right) for the<br />
frustum model. Here, xy-pol. and yx-pol. refer to ρxy and ρyx, resp.<br />
The following table summarizes important computational parameters for the three different codes,<br />
such as polynomial degree of the basis functions (BFD) for the FE method and degrees of freedom<br />
(DOF). The overall CPU time required to obtain the result for one frequency was in the order of two<br />
minutes. The total time required to compute all field components for one polarization sums up to<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
64<br />
φ [ o ]<br />
10 3
about two hours for all 49 frequencies for the FE codes and 30 minutes for the FD code. Note that the<br />
3-D impedance requires the calculation of both polarizations (see eq. (7)).<br />
computer CPUs BFD elements DOFs time [min]<br />
COMSOL Multiphysics <br />
xy-pol. klio 4 quadratic 82 661 523 580 119,56<br />
FEP<br />
xy-pol. erato 8 quadratic 57 666 434 456 119,07<br />
FD code<br />
FD code 131 x 131 x 154 grid cells + 15 air layers ≈ 30<br />
4.2 Stromboli-Topography model<br />
In the second level of complexity, a realistic topography of Stromboli is incorporated. Since this<br />
topography is smoothly adjusted to the flat sea-floor, the edifice is larger in this model than in reality.<br />
Furthermore, the conductivity of the volcano has been reduced to 0.01 Sm −1 .<br />
Figs 6 and 7 illustrate the model with the positions of the profiles and the locations for which sounding<br />
curves have been calculated. Sounding curves are shown for sites a, N and E only.<br />
Figure 6: Flat sea-floor model including the real topography of Stromboli island. Red and blue lines indicate<br />
the x- and y-profiles, respectively.<br />
Figure 7: Locations for which sounding curves are computed are denoted by a, b, c, d on the island (left) and<br />
N,E,S,W on the sea-floor (right).<br />
Fig. 8 displays the apparent resistivities ρxy and phase Φxy for T =10 3 s along two profiles aligned with<br />
the x-axis (x-profile) and the y-axis (y-profile). The apparent resitivity for the x-profile is associated<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
65
with the xy-impedance tensor elements (Zxy mode) while the apparent resistivity for the y-profile<br />
is derived from the yx-elements (Zyx mode). The curves for both Zxy and Zyx modes show similar<br />
features. This can be explained by the similar topography along the two different profiles. Nevertheless,<br />
the extent of the volcano’s edifice in y-direction is larger than in x-direction. Hence, the corresponding<br />
minima and maxima of the curves are shifted relative to each other.<br />
Due to the smoother transition of the topography from the sea-floor to the volcano slopes the peaks at<br />
the base points of the edifice are not as pronounced as for the frustum model. Since there is no volcano<br />
plateau in the x-profile, there is just one minimum at x ≈ 44 km for ρxy.<br />
As for the frustum model, the sounding curves for the volcano (e.g. for site a in Fig. 9, left) reflect<br />
the layering of the earth with regard to the electrical conductivity whereas long periods correspond<br />
to large skin depths. Since the same conductivity is assigned to the volcano and the half-space the<br />
sounding curves reach values around 100 Ω m for short periods.<br />
Due to the asymmetry of the model, offsets occur between the curves derived from Zxy and Zyx that<br />
are higher for ρa than for φ. Hence, it can be assumed that the phase is less affected by these two<br />
aspects.<br />
The sounding curves for the sea-floor are displayed in Fig. 9 (right). Since the edifice is larger than the<br />
one represented by the frustum the long period apparent resitvities deviate more significantly from the<br />
undisturbed half-space values.<br />
z [km]<br />
ρ a [Ω m]<br />
φ [°]<br />
−5<br />
0<br />
5<br />
10 4<br />
10 2<br />
10 0<br />
40<br />
20<br />
xy−pol. x profile<br />
yx−pol. y profile<br />
0 10 20 30 40 50 60 70 80<br />
[km]<br />
Figure 8: Topography (top), apparent resistivity (center), phase (bottom), derived from Zxy and Zyx (xy- and<br />
yx-pol., resp.) along the x- andthey-profile for T =10 3 s.<br />
The differences in the short period range may again be attributed to the insufficient spatial discretization.<br />
ρ a [Ω m]<br />
φ [°]<br />
10 3<br />
10 2<br />
10 1<br />
10 0<br />
10 −1<br />
100<br />
50<br />
10 −3<br />
0<br />
10 −2<br />
10 −1<br />
10 0<br />
T [s]<br />
10 1<br />
10 2<br />
xy−pol. a<br />
yx−pol. a<br />
10 3<br />
ρ a [Ω m]<br />
φ [°]<br />
10 3<br />
10 2<br />
10 1<br />
80<br />
60<br />
40<br />
10 1<br />
20<br />
10 2<br />
T [s]<br />
xy−pol. N<br />
yx−pol. N<br />
xy−pol. E<br />
yx−pol. E<br />
Figure 9: Sounding curves for the island site a (left) and for the sea-floor sites N and E (right). xy-pol. and<br />
yx-pol. refer to Zxy and Zyx, resp.<br />
The incorporation of the detailed Strombolian topography leads to a massively increasing number of<br />
mesh points. In order to keep computational costs low the meshes for computing the full sounding<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
66<br />
10 3
curves were coarsened.<br />
Computer CPUs BFD No. of elements DOFs Time [min]<br />
profile curves<br />
xy-pol. erato 8 quadratic 443 563 2 823 488 114.78<br />
sounding curves<br />
xy-pol. erato 8 quadratic 164 662 1 048 116 ≈ 721<br />
4.3 Bathymetry-Topography model<br />
In the third level of complexity, the topography of the island is incorporated as well as the regional<br />
bathymetry including the other Liparian islands. This allows to analyse the effect on the electromagnetic<br />
fields not only due to the volcano itself but also due to the variable thickness of the sea layer.<br />
Figs 10 and 11 illustrate the model and the locations of the profiles and points for which sounding<br />
curves have been calculated. As an example, apparent resistivities and phases for a period of 1000 s<br />
are shown for the xN- and xS-profiles as well as the full sounding curves for points a, c, N and S.<br />
Figure 10: Model including the regional bathymetry and the topography of Stromboli and the Liparian Islands.<br />
Figure 11: Left: Elevation map inferred from digital terrain data showing the locations of seafloor sounding<br />
points N, E, S, W and the profile lines xN, xS in the EW-direction and yE, yW in the NS-direction.<br />
Right: Coast line of Stromboli and locations of sounding points a, b, c, d.<br />
The xN and xS profiles running East-West in parallel (Fig. 12) are close to each other. Hence, the<br />
features of the respective apparent resistivity and phase curves are similar. As for the previous models,<br />
peaks at the coastline and at the top of the volcano can be observed. However, serious perturbations<br />
are provoked by the regional bathymetry. Induced currents are vertically compressed in the shallow<br />
parts of the ocean yielding higher apparent resistivities. Vice versa, apparent resistivites are decreased<br />
in deep water areas. The soundings at site c (Fig. 13, right) again show the influence of the volcano,<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
67
z [km]<br />
ρ a [Ω m]<br />
φ [°]<br />
−5<br />
0<br />
5<br />
10 4<br />
10 2<br />
10 0<br />
200<br />
100<br />
0<br />
xy−pol. xN profile<br />
xy−pol. xS profile<br />
0 20 40<br />
x [km]<br />
60 80<br />
Figure 12: Topography (top), apparent resistivity (center) and phase (bottom) for T = 10 3 satprofilesxS and<br />
xS. xy-pol. and yx-pol. refer to Zxy and Zyx, resp.<br />
ρ a [Ω m]<br />
φ [°]<br />
10 3<br />
10 2<br />
10 1<br />
10 0<br />
400<br />
200<br />
10 −3<br />
0<br />
10 −2<br />
10 −1<br />
10 0<br />
T [s]<br />
10 1<br />
10 2<br />
xy−pol. a<br />
yx−pol. a<br />
10 3<br />
Figure 13: Sounding curves for the island sites a (left) and c (right). xy-pol. and yx-pol. refer to Zxy and Zyx,<br />
resp.<br />
the sea layer and the half-space for the appropriate period ranges. However, at site a, which is located<br />
directly at the coastline, the sea layer affects the curves already for shorter periods, especially for the<br />
Zyx mode. Still, for periods longer than T =10 −1 s the shapes of the curves are similar to those at<br />
point c (although shifted to higher ρa-values).<br />
All curves derived from Zxy and Zyx show an offset which can be attributed to the asymmetric elevation<br />
data used in the MT simulations.<br />
For the two sea-floor sites N and S the topography effect varies for the two cases considered (Fig. 14).<br />
Site N (Fig. 14, left) is located at greatest distance from the islands. Therefore, the curves are similar to<br />
those obtained using the simpler models and for longer periods they are almost assume the corresponding<br />
half-space values.<br />
The sounding curves at site S (Fig. 14, right) seem to be affected by two different facts. For ρxy the<br />
graben-like structure in this area (cf. Fig. 11, left) leads to lower apparent resistivities. On the other<br />
hand, the more resistive islands affect the ρyx-curves.<br />
We note that the out-of-quadrant phases in Fig. 14 occur mainly at the high frequencies, which indicates<br />
that the mesh used for the numerical computations is still not fine enough to obtain results with the<br />
desired accuracy.<br />
The following table summarizes the mesh properties and CPU times associated with the bathymetrytopography<br />
model.<br />
ρ a [Ω m]<br />
10 3<br />
10 2<br />
10 1<br />
10 0<br />
100<br />
50<br />
10 −3<br />
0<br />
10 −2<br />
10 −1<br />
10 0<br />
T [s]<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
68<br />
φ [°]<br />
10 1<br />
10 2<br />
xy−pol. c<br />
yx−pol. c<br />
10 3
ρ a [Ω m]<br />
φ [°]<br />
10 4<br />
10 3<br />
10 2<br />
10 1<br />
400<br />
200<br />
0<br />
10 1<br />
−200<br />
10 2<br />
T [s]<br />
xy−pol. N<br />
yx−pol. N<br />
10 3<br />
Figure 14: Sounding curves for the sea-floor sites N (left) and S (right). xy-pol. and yx-pol. refer to Zxy and<br />
Zyx, resp.<br />
ρ a [Ω m]<br />
φ [°]<br />
10 4<br />
10 3<br />
10 2<br />
10 1<br />
400<br />
200<br />
10 1<br />
0<br />
10 2<br />
T [s]<br />
xy−pol. S<br />
yx−pol. S<br />
Computer CPUs BFD No. of elements DOFs Time [min]<br />
profile curves<br />
xy-pol. erato 16 quadratic 546 759 3 479 830 44.41<br />
sounding curves<br />
xy-pol. erato 16 quadratic 312 721 1 987 514 ≈ 1103<br />
5 Conclusions<br />
In this work, distortion effects on MT data caused by topography and bathymetry are examined for<br />
the area around Stromboli. The geometry of Stromboli has been incorporated into FE simulations<br />
applying three different levels of increasing complexity.<br />
In the first model, a frustum representing the volcano has been embedded in a layered background.<br />
This model yields a non-trivial apparent resistivity profile curve which is mainly dominated by peaks at<br />
the base points of the frustum, at the coastline points and at the edges of the volcano plateau. These<br />
are caused by static shift effects and the high conductivity contrast between the volcano and the sea<br />
layer. The results of the frustum model have been verified by a finite element code by Schwarzbach<br />
(2009) and a finite difference code by Mackie et al. (1994).<br />
For the second model of Stromboli digital elevation data of the island have been included in the model<br />
geometry. Compared to the frustum model, the apparent resistivity and phase curves along the profiles<br />
and the sounding curves for the volcano show similar features.<br />
Finally, digital bathymetry data have been added to the third, most complex Stromboli model to<br />
take into account realistic topography and bathymetry features of the region. The apparent resistivity<br />
still shows the characteristic peaks at the coastline points for most of the profiles. The effects of<br />
the bathymetry and topography of the surrounding islands are superposed, which results in very<br />
complicated profile and sounding curves.<br />
These results clearly point out that topography and bathymetry can heavily distort MT data and thus<br />
need to be considered for an accurate numerical interpretation of MT measurements.<br />
Finally, we emphasize that FE methods on unstructured grids provide a sophisticated tool for treating<br />
problems associated with complex geometry.<br />
Acknowledgements<br />
We would like to thank Randall L. Mackie for providing the results for the FD simulations. Furthermore,<br />
we are thankful to Christoph Schwarzbach for providing his FE code and for his useful help. We are<br />
grateful to the German Research Foundation DFG for funding our numerical research work (Spi 356-9).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
69<br />
10 3
References<br />
Börner, R.-U. (2010). Numerical Modelling in Geo-Electromagnetics: Advances and Challenges. Surveys in Geophysics<br />
31(2): 225–245.<br />
Mackie, R. L., J. T. Smith, and T. R. Madden (1994). Three-dimensional electromagnetic modeling using finite difference<br />
equations: The magnetotelluric example. Radio Science 29(4): 923–935.<br />
Nam, M. J., H. J. Kim, Y. Song, T. J. Lee, J.-S. Son, and J. H. Suh (2007). 3D magnetotelluric modelling including<br />
surface topography. Geophysical Prospecting (55): 277–287.<br />
Schwarzbach, C. (2009). Stability of Finite Element Solutions to Maxwell’s Equations in Frequency Domain. PhDthesis.<br />
TU Bergakademie Freiberg.<br />
SwissEduc (2010). SwissEduc: Stromboli Online - Die Insel. url: http://www.swisseduc.ch/stromboli/volcano/geogr/<br />
aerial-de.html.<br />
Wait, J. R. (1953). Propagation of radio waves over a stratified ground. Geophysics 18(2): 416–422.<br />
Wikipedia (2010). Liparische Inseln. url: http://de.wikipedia.org/wiki/Liparische_Inseln.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
70
Establishing Controlled Source MT at <strong>GFZ</strong><br />
M. Becken 1,2 , R. Streich 1,2 , O. Ritter 2<br />
1 Potsdam University, Department of Geosciences, Karl-Liebknecht-Strasse 24, 14476 Potsdam, Germany<br />
2 Helmholtz Centre Potsdam <strong>GFZ</strong> German Research Centre for Geosciences, Telegrafenberg, 14473 Potsdam, Germany<br />
Summary<br />
The multidisciplinary GeoEn project (the Brandenburg pilot project in the BMBF program<br />
„Spitzenforschung und Innovation in den Neuen Ländern“) integrates research in the fields of<br />
geothermal energy, carbon dioxide capture, transport and storage (CCTS) as well as the exploration of<br />
unconventional gas reserves (shale gas). The electrical conductivity is one key parameter to characterize<br />
reservoirs and to monitor changes due to circulation or injection of fluids at reservoir depth. Bulk<br />
electrical conductivity is highly sensitive to fluids within interconnected pores, and therefore EM<br />
techniques are powerful tools for exploring and monitoring geothermal reservoirs, CO2 storage sites and<br />
shale gas reservoirs. In the framework of the GeoEn project, we establish Controlled Source MT (CSMT)<br />
at <strong>GFZ</strong> Potsdam. We aim at combining active and passive MT to image the electrical conductivity<br />
structure within the Earth, ultimately in 3D.<br />
For CSMT, we intend to use grounded electrodes to inject a frequency-dependent current into the Earth<br />
and measure the induced electric and magnetic fields at near-field to far-field distances. We will use<br />
novel transmitter systems from Metronix (Braunschweig) which are presently under development.<br />
Standard MT receivers will be utilized to measure the induced electric and magnetic fields. In the scope<br />
of the project, we will (i) assemble the transmitter system and the source dipole, (ii) optimize and design<br />
CSMT field procedures for geothermal exploration, carbon dioxide reservoir characterization and shale<br />
gas exploration (Streich et al., 2010a), (iii) develop and implement time-series processing, (iv) develop<br />
and implement 1D modeling (Streich and Becken, 2009) and inversion software and 3D modeling codes<br />
(Streich, 2009; Streich et al., 2010b).<br />
In this contribution, we describe a test of long steel electrodes as the current electrodes of the source<br />
dipole, and we examine the resolution power of CSMT using 1D inversion of synthetic data.<br />
Current electrodes of CSMT source dipole<br />
Low grounding resistances of the current electrodes are crucial for injecting strong currents into the<br />
subsurface. The Metronix 22 kVA transmitter generates currents of max. 40 A (560 V), which can,<br />
however, only be achieved if the total resistance of the dipole (1-km long cable and grounding<br />
electrodes) is less than 14 . We use a thick cable that has a resistance of 2 /km. Accordingly, to<br />
achieve maximum currents, the electrode grounding resistances should be less than 12 . At high<br />
frequencies, the maximum current will be further limited by the inductance of the cable.<br />
The grounding resistance depends primarily on the surface of the electrode that is in contact with the<br />
ground, and the surrounding resistivity within the Earth. For a homogenous earth, the grounding<br />
resistance R of a steel rod is<br />
R<br />
4L<br />
ln<br />
2 L<br />
a<br />
1<br />
, (1)<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
71
Figure 1: Electrode test. We measured the ground resistance of a steel electrode (diameter 32 mm) vs. depth of<br />
the rod at two different locations, which may represent typical conditions in northern Germany. Ground resistance<br />
measurements were taken against two reference electrodes at ~30 m distance to the left and right. DC Wenner<br />
spreads were measured at the same locations (left panels). The panels to the right show the measured resistances<br />
(red and blue lines), compared to theoretical predictions (dashed black lines) based on the resistivity-depth<br />
profiles at the electrode locations shown in the rightmost plots. The sudden drop in resistance at
greater depths, we penetrate conductive glacial till at test location 1 (upper panel in Figure 1) and a<br />
groundwater layer at test location 2 (lower panel). The measured grounding resistance values are<br />
roughly consistent with predictions (dashed lines) based on the geometry and the subsurface resistivity<br />
distribution. For resistance predictions, we approximated the resistance R from a parallel circuit of steel<br />
rod segments of length L embedded into a medium with piecewise constant resistivity i .<br />
Approximate i values were extracted from the 2D DC resistivity-depth sections. Let the individual rod<br />
pieces have resistance<br />
R<br />
i<br />
i 4 L<br />
ln<br />
2 L a<br />
then the resistance of the entire rod to a depth L is given by<br />
1<br />
R<br />
1<br />
i i R<br />
1<br />
, (3)<br />
. (4)<br />
Equations 3 and 4 were used to estimate grounding resistances from the 2D resistivity models (Figure 1).<br />
This test shows that long steel rods may be suitable electrodes for CSMT source dipoles. At one test<br />
location with favorable geology (glacial till), we achieved a ground resistance of ~10 with electrodes<br />
penetrating 7.5 m deep; the second test location, a sandy soil with a freshwater layer at ~3 m depth,<br />
may require deeper electrodes. Our predictions suggest that the ground resistance of the steel rod may<br />
drop at this location to values of ~10 at ~15 m depth. The simple model used to predict grounding<br />
resistance has proven to be useful for practical purposes. Before installing electrodes, it may be<br />
advisable to investigate potential locations with DC resistivity soundings in order to predict expectable<br />
resistances and required electrode lengths.<br />
1D CSMT Inversion<br />
Vertical currents, galvanically injected into the subsurface with a grounded dipole source, exhibit<br />
sensitivity to buried resistors at depth. This makes the CSMT technique suitable for imaging resistive<br />
layers, whereas both passive MT and CSMT are sensitive to conductive layers. The effect of anomalously<br />
resistive subsurface structures on surface CSMT data is typically smaller in land applications than in<br />
deep-water marine applications. Land applications of frequency-domain CSMT suffer from energy<br />
travelling through the air even more than marine applications, which obscures the response from<br />
deeper targets. Nevertheless, forward modeling studies suggest that in many cases, the e.m. response<br />
of thin resistive layers is above noise level (see Streich and Becken, this issue).<br />
A practical question is to what extent the resistivity models giving rise to these anomalies can be<br />
recovered from measured data. We ran 1D Occam-type inversions to investigate the resolution power of<br />
CSMT data. To infer the resistivity structure from 1D inversion, we can exploit the spatial decay of CSMT<br />
fields with increasing distance to the source, and the frequency dependency of the fields. In land-based<br />
applications, we will typically have only few source locations combined with a relatively large number of<br />
receivers (say, 100), and potentially long transmitting times that allow us to cover a broad frequency<br />
range. In contrast, marine applications use a source towed continuously over a number of receivers,<br />
effectively yielding many source point locations, but only short transmitting times and thus a narrow<br />
frequency band for a given source location. Hence, land applications will primarily utilize the frequency<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
73
Figure 2: 1D Occam inversion of synthetic CSMT data. The data are the real and imaginary parts of the inline<br />
electric field, contaminated with white noise with 1% of the signal amplitude. In addition, white noise with a<br />
standard deviation of 10 -9 V/m was added. a) Inversion of the spatial decay of the field for a single frequency (1<br />
Hz), using an unconstrained Occam inversion and using an Occam variant where the deviation from an a priori<br />
model is penalized outside of the depth of interest. b) Inversion of the frequency spectrum recorded at 5 km<br />
distance from the transmitter.<br />
Figure 3: Comparison of single and joint 1D inversion models for passive MT and CSMT data. In (a), the left panel<br />
shows passive MT data (apparent resistivity and phase) for the true model and MT inversion result displayed in the<br />
right panel. In (b), CSMT data (real and imaginary part of the inline electric field) are displayed for the true<br />
resistivity model and CSMT inversion result. (c) shows the resistivity model resulting from joint MT and CSMT<br />
inversion. In this example, the MT data add information only at greater depths, where the sensitivity of CSMT<br />
sounding is low. In practice, we anticipate that MT data will help construct a (2D/3D) regional background model,<br />
and CSMT data will help refine the model at reservoir scale (1D/3D).<br />
dependency of the fields, whereas marine applications will primarily exploit spatial variations of the<br />
fields.<br />
We have implemented a 1D CSMT inversion based on Weidelt’s formulation of the forward problem<br />
(Weidelt, 2007), and analytically derived expressions for the Jacobian matrix required in the inversion.<br />
Finite source dipoles (Streich and Becken, this issue) have been incorporated into the inversion;<br />
however, in the present form of the inversion scheme, the source and receivers are placed at the<br />
surface. Here, we only consider horizontal electric point dipole sources. The inversion is similar to<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
74
Occam inversion (Constable et al., 1987); however, Occam’s second phase (i.e., the determination of the<br />
regularization parameter that yields optimal trade-off between data residuals and model norm) has not<br />
been implemented. Instead, we fix the regularization parameter for one entire inversion run.<br />
Using this inversion scheme, we found that both the inversion of multi-offset data at a single frequency<br />
and multi-frequency data at a single receiver location have the power to yield comparable inversion<br />
models, if the frequency range and the distance to the transmitter are appropriate for imaging the<br />
target depth (Figure 2). The left panels in Figure 2a and b depict the inline electric field as a function of<br />
distance from the transmitter for the frequency f = 1 Hz and as a function of frequency at 5 km distance,<br />
respectively, for a model containing a resistive layer embedded into a homogeneous half-space.<br />
Random noise of 1% of the field amplitudes and a noise floor of 10 -9 V/m were added to the data prior to<br />
inversion. The central panel in Figure 2a shows the inversion model obtained from smooth inversion<br />
using a standard Laplacian penalty on the logarithmic model resistivities (red line). A similar result (not<br />
shown) was obtained from inversion of the multi-frequency electric field data shown in Figure 2b. In<br />
both cases, the recovered 1D model contains a resistive zone that is smeared over a wide depth range,<br />
because regularization penalizes thin model layers.<br />
For monitoring applications (e.g., CO2 injection), the depth of interest is known. A focused inversion that<br />
searches for deviations from an a priori model primarily at reservoir depth may help improve the model.<br />
We have therefore incorporated a term into the Occam inversion that allows us to weight the penalty<br />
depending on the difference of every inversion model parameter to an a priori model (Key, 2009). The<br />
right panels in Figure 2a and b show the result of this variant of regularized inversion. The hashed areas<br />
correspond to depth levels where the inversion model is constrained to deviate minimally from the a<br />
priori model (which is the true model in this case), whereas the rest of the model was permitted to vary<br />
freely (in the sense of a Laplacian regularization). This approach clearly helps focusing the resistive<br />
anomaly at the true depth and may thus be adequate for a focused model search.<br />
The penetration depth of CSMT data is limited by the lowest frequencies used and the greatest offsets<br />
providing reasonable signal quality. Combing CSMT with passive MT may prove helpful to expand the<br />
model scale. Furthermore, both techniques exhibit different sensitivities to resistive and conductive<br />
structures (and to deviations from 1D media) and thus provide complementary information. We have<br />
therefore implemented a joint 1D inversion for both data types. In Figure 3, we compare the 1D<br />
inversion models for a multi-layered structure obtained from MT data and CSMT data alone and from<br />
joint CSMT and MT inversion. Here, the CSMT data are the frequency-dependent inline electric field<br />
data, measured at 5 km offset from the transmitter. The individual inversion models show that (i) both<br />
techniques resolve shallow conductive layers, (ii) the CSMT data resolve a resistive layer at 1 km depth,<br />
and (iii) the MT data resolve the conductivity of the basal layer that is too deep to be sensed with CSMT.<br />
All of these model features are revealed by joint inversion, suggesting increased resolution capabilities<br />
from combined applications of both techniques.<br />
Acknowledgements<br />
This work is funded by the German Federal Ministry of Education and Research (BMBF) within the<br />
framework of the GeoEn project. S. Costabel, TU Berlin, made the DC measurements and provided the<br />
2D inversion models.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
75
References:<br />
Constable, S. C., Parker, R. L. and Constable, C. G., 1987, Occam's inversion: A practical algorithm for<br />
generating smooth models from electromagnetic sounding data, Geophysics 75(3), 289-300.<br />
Key, K., 2009, 1D inversion of multicomponent, multifrequency marine CSEM data: Methodology and<br />
synthetic studies for resolving thin resistive layers, Geophysics 74(2), F9-F20.<br />
Streich, R., 2009, 3D finite-difference frequency-domain modeling of controlled-source electromagnetic<br />
data: direct solution and optimization for high accuracy, Geophysics 74(5), F95-F105.<br />
Streich, R., Becken, M. and Ritter, O., 2010a, Imaging of CO2 storage sites, geothermal reservoirs, and<br />
gas shales using controlled-source magnetotellurics: modeling studies, Chemie der Erde, submitted.<br />
Streich, R. and Becken, M., 2010, EM fields generated by finite-length wire sources in 1D media:<br />
comparison with point dipole solutions, Protokoll zum Kolloquium “Elektromagnetische Tiefenforschung”,<br />
Seddiner See, 28.09.-02.10.2009.<br />
Streich, R., Schwarzbach, C., Becken, M. and Spitzer, K., 2010b, Controlled-source electromagnetic<br />
modelling studies: utility of auxiliary potentials for low-frequency stabilization, EAGE 70 th Conference<br />
and Exhibition, Barcelona, Spain, submitted.<br />
Weidelt, P., 2007, Guided waves in marine CSEM: Geophysical Journal International, 171, 153–176.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
76
2D-SIP-Modellierung mit anisotropen<br />
Widerständen<br />
Johannes Kenkel a∗ , Andreas Hördt a , Andreas Kemna b<br />
a Institut für Geophysik und extraterrestrische Physik, TU Braunschweig<br />
b Fachbereich Angewandte Geophysik, Universität Bonn<br />
1 Einleitung<br />
Die Modellierung von Messdaten der spektralen induzierten Polarisation (SIP) geschieht<br />
üblicherweise unter der Annahme makroskopisch, also in der Größenordnung einer Zelle<br />
der Modellrechnung, isotroper komplexer Widerstände im Untergrund. Für diese Art der<br />
Modellierung von Daten der SIP wurde von A. Kemna das Finite-Elemente-Programm<br />
CRMod entwickelt ([1], [2]). Mit der Annahme isotroper Widerstände lassen sich viele<br />
Messdaten hinreichend erklären. Nguyen et al. ([3]) zeigten jedoch für den reellwertigen<br />
gleichstromgeoelektrischen Fall, dass, wenn der Untergrund aus vielen abwechselnd gelagerten<br />
Schichten besteht, die eine geringe räumliche Ausdehnung gegenüber der bei der<br />
Inversion benutzen Gitterzellen haben, die Annahme isotroper Leitfähigkeiten zu Fehlinterpretationen<br />
führen kann. Hier kann die Erweiterung auf anisotrope Widerstände<br />
Abhilfe schaffen, weil diese auf makroskopischer Skala, also Zellenebene, Schichtlagerungen<br />
beschreiben können. Das Auftreten von Anisotropieeffekten auch bei SIP wurde<br />
von Winchen et al. ([4]) demonstriert. Das Programm CRMod wurde zum Zweck der<br />
Modellierung anisotroper komplexer Widerstände erweitert. In diesem Artikel sollen anhand<br />
von verschiedenen modellierten Beispielen verschiedene Effekte eines Untergrund<br />
mit anisotropen komplexen Widerständen gezeigt werden.<br />
2 Methodik<br />
2.1 Spektrale Induzierte Polarisation (SIP)<br />
Die spektrale induzierte Polarisation (SIP) ist ein elektrisches Verfahren, das wie die Geoelektrik<br />
aus dem Messen von Potentialverläufen künstlicher elektrischer Quellen Rückschlüsse<br />
auf die Leitfähigkeitsverteilung im Untergrund zieht.<br />
∗ Adresse: Mendelssohnstraße 3, D-38106 Braunschweig, E-Mail: j.kenkel@tu-bs.de<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
77
Das Aufnehmen der Messwerte geschieht in einem wiederum der Geoelektrik sehr<br />
ähnlichen Aufbau, jedoch werden zusätzlich Phasenbeziehungen zwischen Sende- und<br />
Empfangssignal aufgezeichnet. Als Quellsignal werden Wechselströme unterschiedlicher<br />
Frequenzen gewählt. Die Messdaten sind dann Spannungswerte und deren Phasenverschiebungen<br />
zum Quellsignal. Aus diesen Daten und den Positionen der Elektroden ergeben<br />
sich mit den Geometriefaktoren die Messwerte als scheinbare spezifische Widerstände<br />
(Amplituden) und deren Phasen. Diese lassen sich zur Abschätzung der Eigenschaften<br />
des Profils als Pseudosektion auftragen.<br />
Das physikalische Problem der spektralen induzierten Polarisation und der Geoelektrik<br />
lässt sich für den Fall anisotroper komplexer Widerstände in x-, y- und z-Richtung in<br />
Form der Poisson-Gleichung<br />
∂x(σx∂xφ)+∂y(σy∂yφ)+∂z(σz∂zφ)+Iδ(x − xs)δ(y − ys)δ(z − zs) =0 (1)<br />
mit dem Potential φ = φ(x, y, z), dem Strom I und der Diracschen Delta-Funktion δ<br />
schreiben. Die Anisotropie beschränke sich dabei auf die Diagonalelemente des Leitfähigkeitstensors<br />
⎛<br />
σx<br />
σ = ⎝ 0<br />
0<br />
σy<br />
⎞<br />
0<br />
0 ⎠ . (2)<br />
0 0 σz<br />
Die Stromeinspeisung findet an den Koordinaten (xs,ys,zs) statt.<br />
2.2 Modellierung<br />
Für die Modellierung wird das oben aufgeführte Problem der Potentialverteilung durch<br />
einen 2D-Finite-Elemente-Algorithmus gelöst (vgl. [2]). Die Ergebnisse sind Potentialverläufe<br />
und zugehörige Phasen an jedem Gitterpunkt. Aus diesen Informationen und<br />
den Messelektrodenpositionen lassen sich damit die zum Modell gehörigen synthetischen<br />
Messdaten - Amplituden und Phasen der scheinbaren spezifischen Widerstände - berechnen<br />
und in Pseudosektionen auftragen. Das hier benutzte Finite-Elemente-Programm<br />
CRMod ist von einer bereits bestehenden Version für die Modellierung isotroper spezifischer<br />
Widerstände und Phasen ([2]) um die Möglichkeit der Modellierung von Anisotropie<br />
erweitert worden.<br />
Die Poisson-Gleichung 1 wird unter der Annahme eines konstanten Leitfähigkeitstensors<br />
in y-Richtung durch die Fourier-Transformation in den Wellenzahlbereich mit<br />
∞<br />
F (k) = f(y)e<br />
−∞<br />
iky dy<br />
∞<br />
∞<br />
= f(y) cos(ky)dy + i f(y) sin(ky)dy<br />
−∞<br />
und der Wellenzahl k transformiert. Bei Reduktion auf Funktionen, die in y-Richtung<br />
zum Nullpunkt symmetrisch sind, fällt das zum Nullpunkt y = 0 antisymmetrische<br />
Teilintegral weg, so dass sich die Fourier-Kosinus-Transformation<br />
∞<br />
F (k) = f(y) cos(ky)dy (3)<br />
−∞<br />
−∞<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
78
ergibt. Das Ergebnis der Transformation der Gleichung 1 in den Wellenzahlbereich entlang<br />
der y-Achse ist dann<br />
∂x(σx∂x ˜ φ) − σyk 2 ˜ φ + ∂z(σz ˜ φ)+ I<br />
2 δ(x − x0)δ(z − z0) =0 (4)<br />
mit ˜ φ = ˜ φ(x, k, z). Diese Gleichung wird von dem hier verwendeten Finite-Elemente-<br />
Programm auf einem zweidimensionalen Gitter in x- und z-Richtung gelöst. Aus den<br />
erhaltenen Werten für ˜ φ, bzw. nach Fourier-Kosinus-Rücktransformation φ lassen sich<br />
mit den Geometriefaktoren der Elektrodenkonfiguration die synthetischen Messwerte für<br />
Amplitude und Phase der scheinbaren spezifischen Widerstände angeben.<br />
2.3 Bewertung der Genauigkeit<br />
Das Programm muss auf seine Gültigkeit und Genauigkeit in Bezug sowohl auf analytisch<br />
berechenbare Probleme als auch auf verschiedene, mit dem bisherigen Programm zum<br />
Modellieren isotroper spezifischer Widerstandsverteilungen berechnete Modelle geprüft<br />
werden. Für einen homogenen Halbraum mit spezifischem Widerstand ρh und Phase<br />
φh erwartet man eine Pseudosektion mit dem konstanten Wert ρ = ρh des scheinbaren<br />
spezifischen Widerstands und φ = φh für die scheinbare Phase. Die Messwerte werden<br />
mit einer simulierten Dipol-Dipol-Anordnung mit Elektrodenabstand 1 m berechnet. Die<br />
Abweichungen der Modellierungsergebnisse zum analytischen Ergebnis betragen weniger<br />
als 10 % in den scheinbaren spezifischen Widerständen und weniger als 1 % in den scheinbaren<br />
Phasen. Ein homogener Halbraum mit anisotropen spezifischen Widerständen in<br />
Horizontal- und Vertikalrichtung, ρhorizontal und ρvertikal, weist nach analytischen Berechnungen<br />
einen scheinbaren spezifischen Widerstand von<br />
ρ = √ ρhori. · ρvert.<br />
(nach z.B. [5, S. 95]) auf. In Abb. 1 ist ein Modellierungsergebnis für diesen Fall dargestellt.<br />
Der spezifische Widerstand des Halbraums beträgt 100 Ωm in horizontaler Richtung<br />
(x- und y-Richtung) und 25 Ωm in z-Richtung. Dies entspricht einem scheinbaren<br />
spezifischen Widerstand von 50 Ωm in der Pseudosektion. Die Phase des Halbraums<br />
beträgt gleichmäßig (homogen und isotrop) −5 mrad. Die Abweichung des Modellierungsergebnisses<br />
zu diesem analytischen Wert beträgt an keiner Stelle mehr als 3 %<br />
sowohl für den scheinbaren spezifischen Widerstand als auch für die scheinbare Phase.<br />
3 Modellparameter<br />
Unter der Annahme einer horizontalen Wechsellagerung unterschiedlich leitfähiger Schichten<br />
wurde ein Modell mit isotropen spezifischen Widerständen erstellt und mit einem<br />
entsprechenden Modell mit anisotropen spezifischen Widerständen verglichen. Zur<br />
Abschätzung der Stärke der Anomalien wird ein Hintergrundmodell mit homogenem und<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
79<br />
(5)
(a) (b)<br />
Abbildung 1: Dipol-Dipol-Pseudosektion von spezifischem Widerstand und Phase eines<br />
homogenen Halbraums mit den anisotropen spezifischen Widerständen<br />
(Amplituden) 100 Ωm in x- und y-Richtung und 25 Ωm in z-Richtung und<br />
der isotropen Phase −5 mrad.<br />
isotropem spezifischen Widerstand angegeben. Abb. 2 zeigt die Diskretisierung des Modellraumes<br />
in 233 Zellen in x-Richtung und 114 Zellen in z-Richtung mit variablem Gitterlinienabstand<br />
in den Außenbereichen und konstantem Gitterlinienabstand von 0, 5m<br />
m in einem Bereich von 0 m bis 100 m in x-Richtung.<br />
Abbildung 2: Gittermodell des Profils mit 233x114 Gitterzellen und einem Gitterlinienabstand<br />
von 0, 5 m im linearen Bereich.<br />
Für das Hintergrundmodell wurde wie im vorherigen Abschnitt ein homogener und<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
80
isotroper spezifischer Widerstand mit Amplitude 100 Ωm und Phase −5 mrad gewählt.<br />
3.1 Wechsellagerungen mit unterschiedlichen spezifischen Widerständen<br />
Im ersten Beispiel wird eine Wechsellagerung von unterschiedlich resistiven, isotropen<br />
Schichten (100 Ωm und 1 Ωm) in einem Hintergrund mit spezifischem Widerstand von<br />
100 Ωm betrachtet. Die Phase des spezifischen Widerstands beträgt überall −5 mrad.<br />
Eine Skizze der Anordnung findet sich in Abb. 3(a). Die Schichtung beginnt 1 m unter<br />
(a) (b)<br />
Abbildung 3: Schema des Untergrundmodells: a) vertikale Schichtung mit den spezifischen<br />
Widerständen 100 Ωm in hellgrau und 1 Ωm in dunkelgrau. b) Vertikale<br />
Schichtung mit spezifischem Widerstand 1, 98 Ωm in x- und y- sowie<br />
50, 5 Ωm in z-Richtung. Der spezifische Hintergrundwiderstand beträgt<br />
100Ωm in x-, y- und z-Richtung. Die Phase beträgt überall −5 mrad.<br />
der Oberfläche und reicht bis zu einer Tiefe von 7, 5 m. Die Ausdehnung in x-Richtung<br />
beträgt 40 m.<br />
Die gleiche Schichtung wird in Abb. 3(b) durch einen Block mit anisotropem spezifischen<br />
Widerstand beschrieben. Die anisotropen spezifischen Widerstandswerte berechnen<br />
sich nach<br />
ρ = R A<br />
l<br />
⇔ R = ρ l<br />
(6)<br />
A<br />
mit dem Widerstand R eines Blocks der Länge l und der Querschnittsfläche A. Bei<br />
Reihenschaltung von zwei Schichten mit den Widerständen R1 und R2 mit je der Querschnittsfläche<br />
A und der Länge l ergibt sich aus dem Gesamtwiderstand R, der die<br />
Querschnittsfläche A und die Länge 2 · l hat:<br />
RSerie = R1 + R2<br />
2 · l l l<br />
⇔ ρSerie = ρ1 + ρ2<br />
A A A<br />
⇔ ρSerie = ρ1 + ρ2<br />
.<br />
2<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
81<br />
(7)
Bei Parallelschaltung von zwei Schichten gleicher Querschnittsfläche A und gleicher<br />
Länge l und den Widerständen R1 und R2 ergibt sich der Gesamtwiderstand R, der<br />
die Querschnittsfläche 2 · A und die Länge l hat:<br />
⇔<br />
2 · A<br />
l<br />
1<br />
RSerie<br />
1<br />
ρSerie<br />
⇔ ρP arallel =2·<br />
= 1<br />
+<br />
R1<br />
1<br />
R2<br />
= A<br />
l ρ1 + A<br />
l ρ2<br />
1<br />
1 1 + ρ1 ρ2<br />
Mit diesen Beziehungen folgt für die äquivalenten anisotropen spezifischen Widerstände<br />
50, 5Ωminx-und1, 98 Ωm in z-Richtung. In y-Richtung beträgt der spezifische Widerstand<br />
wie in x-Richtung 1, 98 Ωm. Die Phasen der spezifischen Widerstände betragen<br />
homogen und isotrop −5 mrad. Die Modellierungsergebnisse sind als Pseudosektionen<br />
einer Dipol-Dipol-Anordnung in den Abb. 4(a) bis 5(b) jeweils für Betrag und Phase des<br />
spezifischen Widerstands dargestellt.<br />
(a) (b)<br />
Abbildung 4: Pseudosektion des Betrags des scheinbaren spezifischen Widerstands. a)<br />
Vertikale Wechsellagerung isotroper Schichten mit 1 Ωm und 100 Ωm. b)<br />
Anisotrope Beschreibung der Wechsellagerung durch spez. Widerstände<br />
50, 5 Ωm in x- und 1, 98 Ωm in y- und z-Richtung. Die Phase beträgt in<br />
beiden Modellen überall −5 mrad.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
82<br />
.<br />
(8)
(a) (b)<br />
Abbildung 5: Pseudosektion der scheinbaren Phase. a) Vertikale Wechsellagerung isotroper<br />
Schichten mit 1 Ωm und 100 Ωm. b) Anisotrope Beschreibung der<br />
Wechsellagerung durch spez. Widerstände 50, 5Ωminx-und1, 98 Ωm in<br />
y- und z-Richtung. Die Phase beträgt in beiden Modellen überall −5 mrad.<br />
Beide Modelle - isotropes mit Schichten und anisotropes mit Block - zeigen nahezu<br />
gleiche Ergebnisse in den Pseudosektionen. Sehr deutlich ist die durch die Schichtlagerung<br />
bzw. durch den Block anisotroper spezifischer Widerstände erzeugte Anomalie<br />
sichtbar. Der maximale Wert für den scheinbaren spezifischen Widerstand ρa ist etwa<br />
100 Ωm, was dem Hintergrundmodell entspricht. Der minimale Wert für ρa beträgt etwa<br />
10 0,5 Ωm = 3 Ωm. Der erwartete minimale scheinbare spezifische Widerstand beträgt im<br />
Fall des anisotropen Modells gemäß Gleichung 5 etwa 10 Ωm. Dieser Wert wird im Bereich<br />
des Blocks und darunter erreicht und sogar unterschritten. Der bis in 7, 5 m Tiefe<br />
ausgedehnte Block führt einen ”Schatten” mit, der in Form eines nach unten gerichteten<br />
Dreiecks bis in ca. 20 m Tiefe reicht. An beiden Flanken des Dreiecks zeigen schräge<br />
”Schatten” nach außen, die ebenfalls einen relativ niedrigen spezifischen Widerstand von<br />
etwa 10 Ωm aufweisen. Die scheinbaren Phasen betragen gleichmäßig −5 mrad, entsprechend<br />
den Vorgaben der beiden Modelle.<br />
3.2 Wechsellagerungen mit unterschiedlichen Phasen des spezifischen<br />
Widerstandes<br />
In diesem Abschnitt soll die Schichtlagerung mit Schichten unterschiedlicher Phasen betrachtet<br />
werden. Dazu wird das Schichtmodell aus Abb. 3(a) für Schichten mit isotropen<br />
Phasen und das Blockmodell aus Abb. 3(b) als entsprechendes Modell mit anisotropen<br />
Phasen betrachtet. Die spezifischen Widerstände ρ sind in allen Schichten gleich<br />
und entsprechend auch im anisotropen Fall in allen Richtungen gleich. Die Reihenschaltung<br />
von Elementen mit unterschiedlichen Phasen - also komplexen Impedanzen<br />
X = R(cos φ1 + i sin φ1) - ergibt sich mit Realteil ℜ(X) und Imaginärteil ℑ(X) zu<br />
XSerie = X1 + X2<br />
= ℜ(X1 + X2)+iℑ(X1 + X2)<br />
=(Rcos φ1 + R cos φ2)+i(R sin φ1 + R sin φ2).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
83<br />
(9)
Für kleine Winkel von φ1 und φ2 kann man die Kosinus-Terme über eine Taylor-Entwicklung<br />
in erster Ordnung um 0 mit<br />
cos φ ≈ 1 (10)<br />
und die Sinus-Terme mit<br />
annähern. Es ergibt sich dann für die Reihenschaltung<br />
sin φ ≈ φ (11)<br />
XSerie ≈ 2 · R + iR(φ1 + φ2) (12)<br />
Diese Beziehung lässt sich zur Berechnung des komplexen spezifischen Widerstandes<br />
heranziehen. Zwei Elemente mit den Impedanzen X1 und X2 und je der Querschnittsfläche<br />
A und der Länge l ergeben in Reihenschaltung eine Impedanz X mit der gleichen<br />
Querschnittsfläche und der doppelten Länge 2 · l:<br />
ρ = X A<br />
l<br />
A<br />
= XSerie<br />
2l<br />
≈ (R + iR φ1 + φ2<br />
)<br />
2<br />
A<br />
l<br />
≈ R(cos φ1 + φ2<br />
+ i sin<br />
2<br />
φ1 + φ2<br />
)<br />
2<br />
A<br />
l .<br />
Die Reihenschaltung zweier gleich mächtiger Schichten mit gleichem spezifischen Widerstand<br />
und unterschiedlicher Phase bedeutet folglich näherungsweise eine Mittelung der<br />
Phasen, also<br />
φSerie ≈ φ1 + φ2<br />
. (14)<br />
2<br />
Die Parallelschaltung unterschiedlicher Phasen bei gleichem Betrag des spezifischen Widerstands<br />
ergibt sich mit den obigen Näherungen und φ ≪ 1zu<br />
XP arallel =<br />
≈ R<br />
1<br />
1 1 + X1 X2<br />
1<br />
1+iφ1<br />
1<br />
+ 1<br />
1+iφ2<br />
≈ R( 1 i<br />
+<br />
2 2 (φ1 + φ2<br />
)).<br />
2<br />
Werden zwei Elemente mit gleichem spezifischen Widerstand und unterschiedlicher Phase<br />
und je der Querschnittsfläche A und der Länge l parallel geschaltet, ergibt sich die<br />
Impedanz XP arallel mit der doppelten Querschnittsfläche 2 · A und der gleichen Länge.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
84<br />
(13)<br />
(15)
Der komplexe spezifische Widerstand ergibt sich damit zu<br />
ρ = X A<br />
l<br />
= XP arallel<br />
≈ R( 1 i<br />
+<br />
2<br />
2A<br />
l<br />
2 (φ1 + φ2<br />
2<br />
)) 2A<br />
l<br />
= R(1 + i( φ1 + φ2<br />
))<br />
2<br />
A<br />
l<br />
≈ R(cos φ1 + φ2<br />
2<br />
+ i sin φ1 + φ2<br />
)<br />
2<br />
A<br />
l ,<br />
und für die Phasen wie schon bei der Reihenschaltung näherungsweise:<br />
(16)<br />
φP arallel ≈ φ1 + φ2<br />
. (17)<br />
2<br />
Dieses Ergebnis macht klar, dass bei konstantem spezifischen Widerstand eine Wechsellagerung<br />
von Schichten unterschiedlicher Phasen wohl in ihrer Stärke im Mittel, nicht<br />
jedoch in ihrer Richtung erkannt werden kann. Nur in Verbindung mit unterschiedlich<br />
spezifischen Widerständen der einzelnen Schichten können auch die Phasenwerte der<br />
abwechselnden Schichten unterschieden werden.<br />
Ein Schichtmodell mit den Phasen der einzelnen Schichten von −15 mrad bzw. −5 mrad<br />
(Hintergrund) hat demnach die anisotropen Phasen −10 Ωm in x-,y- und z-Richtung.<br />
Die Ergebnisse sind in den Abb. 6(a) bis 7(b) als scheinbare spezifische Widerstände und<br />
scheinbare Phasen in Pseudosektionen dargestellt.<br />
(a) (b)<br />
Abbildung 6: Dipol-Dipol-Pseudosektion der Amplitude des scheinbaren spezifischen<br />
Widerstands. a) Vertikale Wechsellagerung isotroper Schichten mit<br />
−15 mrad und −5 mrad. b) Anisotrope Beschreibung der Wechsellagerung<br />
durch Phasen −10 mrad in x-, y- und z-Richtung<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
85
(a) (b)<br />
Abbildung 7: Dipol-Dipol-Pseudosektion der scheinbaren Phase. a) Vertikale Wechsellagerung<br />
isotroper Schichten mit −15 mrad und −5 mrad. b) Anisotrope<br />
Beschreibung der Wechsellagerung durch Phasen −10 mrad in x-, y- und<br />
z-Richtung<br />
Wie erwartet zeigt sich in den Pseudosektionen des scheinbaren spezifischen Widerstandes<br />
keine Variation mit der eingestellten Phase. Dadurch wird der in Gl. 14 und 17<br />
aufgestellte Zusammenhang bekräftigt. Die Pseudosektionen der Phasen weisen wie die<br />
Pseudosektionen im Abschnitt 3.1 starke ”Schatten” unterhalb der unteren Störkörpergrenze<br />
auf. Auch die seitlichen Flanken sind sichtbar.<br />
4 Schlussfolgerungen<br />
Es ist mit dem auf anisotrope spezifische Widerstände und Phasen erweiterten Programm<br />
CRMod nun möglich, Effekte von Anisotropie, z.B. bei Wechsellagerungen von Sedimenten,<br />
zu berücksichtigen. Die vorliegenden Ergebnisse zeigen in den möglichen Vergleichen<br />
zu analytischen Berechnungen die Gültigkeit des Programms. Es wurde hierbei insbesondere<br />
Wert auf die Vorschrift zur Berechnung des spezifischen Widerstands im Fall eines<br />
homogenen Halbraums mit anisotropen spezifischen Widerständen (vgl. Gleichung 5)<br />
gelegt. Im Zusammenhang mit der Wechsellagerung von Phasen wurde mittels der Gleichungen<br />
14 und 17 gezeigt, dass horizontale und vertikale Wechsellagerungen gleicher<br />
Mächtigkeit und gleichen spezifischen Widerstands nicht unterschieden werden können.<br />
Im Zusammenhang mit dieser Arbeit ist eine Messkampagne geplant, die auf stark<br />
anisotropem Untergrund stattfinden soll. Die Messdaten sollen mit entsprechenden Modellen<br />
mit anisotropen spezifischen Widerständen des vorliegenden Modellierungsprogramms<br />
erklärt werden. Zusätzlich zu dieser Anwendung besteht die Möglichkeit, auch<br />
das auf dem Modellierungsprogramm CRMod basierende Inversionsprogramm CRTomo<br />
([2]) auf Anisotropie zu erweitern.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
86
5 Danksagungen<br />
Diese Arbeiten werden im Rahmen des Forschungsverbunds Geothermie und Hochleistungsbohrtechnik<br />
(GEBO) vom Niedersächsischen Ministerium für Wissenschaft und<br />
Kultur und von Baker Hughes/ Celle gefördert<br />
Literatur<br />
[1] A. Kemna, ”Tomographic Inversion of Complex Resistivity - Theroy and Application”,<br />
Berichte des Instituts für Geophysik der Ruhr-Universität Bochum, Reihe A, Nr.<br />
56, Der Andere Verlag, 2000, ISBN-13: 978-3934366923.<br />
[2] A. Kemna, ”Tomographische Inversion des spezifischen Widerstandes in der Geoelektrik”,<br />
Diplomarbeit, Institut für Geophysik und Meteorologie der Universität zu<br />
Köln, 1995.<br />
[3] F. Nguyen, S. Garambois, D. Chardon, D. Hermitte, O. Bellier, D. Jongmans, ”Subsurface<br />
electrical imaging of anisotropic formations affected by a slow active reverse<br />
fault, Provence, France”, Journal of Applied Geophysics 62, Seiten 338 - 353, 2007.<br />
[4] T. Winchen, A. Kemna, H. Vereecken und J.A. Huisman, ”Characterization of bimodal<br />
facies distributions using effective anisotropic complex resistvity: A 2D numerical<br />
study based on Cole-Cole models”, Geophysics 74, Seiten A19 - A22, 2009.<br />
[5] K. Knödel, H. Krummel, G. Lange, ”Geophysik”, 2. überarbeitete Auflage, Springer-<br />
Verlag, Berlin, 2005, ISBN 3-540-22275-8.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
87
From forward modelling of MT phases over 90 ◦ towards<br />
2D anisotropic inversion<br />
X. Chen 1,2 , U. Weckmann 1 , K. Tietze 1,3<br />
1 Helmholtz Centre Potsdam - German Research Center for Geosciences,Germany<br />
2 University of Potsdam, Institute of Geosciences, Germany<br />
3Free University of Berlin, Institute of Geological Sciences, Germany<br />
xiaoming@gfz-potsdam.de, uweck@gfz-potsdam.de, ktietze@gfz-potsdam.de<br />
Abstract<br />
Within the framework of the German - South African geo-scientific research initiative Inkaba<br />
yeAfrica several magnetotelluric (MT) field experiments were conducted along the Agulhas-<br />
Karoo Transect in South Africa. This transect crosses several continental collision zones between<br />
the Cape Fold Belt, the Namaqua Natal Mobile Belt and Kaapvaal Craton. Along the profile<br />
we can identify areas (> 10km) with phases over 90 ◦ . This phenomenon usually occurs in presence<br />
of electrical anisotropy. Due to the dense site spacing we are able to observe this behaviour<br />
consistently at several sites.<br />
In this presentation we focus on the profile section between Prince Albert and Mosselbay. With<br />
isotropic 2D inversion we are able to explain most features in the MT data but not the abnormal<br />
phase behavior. With several anisotropic forward modelling studies we have tested the influence<br />
of anisotropy parameters on the MT responses. In a first step we use simple 2D models with<br />
embedded zones of electrical anisotropy to get a basic understanding of anisotropic responses. In<br />
a second step isotropic 2D inversion results serve as background models in which we included<br />
anisotropic zones, e.g. to fit the abnormal phase curves. These resolution tests are necessary and<br />
important for the future development of a 2D inversion with spatially constraint anisotropy.<br />
1 Introduction<br />
For a 2D geoelectric model with a finite system of homogeneous, but generaly anisotropic blocks the<br />
electrical conductivity is a tensor instead of a scalar quantity. It is symmetric and positive-definite.<br />
Due to its symmetry the conductivity tensor σ within each layer of model can be diagonalized and expressed<br />
by three principal conductivities σ1, σ2, σ3 and a rotation matrix, which can be decomposed<br />
into three elementary Euler’s rotations αS, αD, αL respectively.<br />
⎛<br />
⎞<br />
σ =<br />
⎝<br />
σxx σxy σxz<br />
σyx σyy σyz<br />
σzx σzy σzz<br />
⎠<br />
⎛<br />
= Rz(−αS)Rx(−αD)Rz(−αL) ⎝<br />
σ1 0 0<br />
0 σ2 0<br />
0 0 σ3<br />
⎞<br />
⎠Rz(αL)Rx(αD)Rz(αS)<br />
where Rx and Rz are elementary rotation matrices around the coordinate axis x and z, respectively.The<br />
angles αS, αD, αL are typically called anisotropy strike, dip and slant, respectively.<br />
In this work we investigate the feasibility of use of 2D forward anisotropy modelling to simulate<br />
phases over 90 ◦ which we observed in field data from the MT survey in South Africa. Using the 2D<br />
anisotropy forward modelling algorithm of Pek and Verner (1997) we vary the model parameters in<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
88
order to gain a basic understanding of the MT transfer function in presence of anisotropy. First we<br />
will give a brief introduction of the survey area and then we will show the results of 2D isotropic<br />
inversion and 2D anisotropic forward modelling.<br />
2 Survey area<br />
Within the framework of the German - South African geo-scientific research initiative Inkaba yeAfrica<br />
four magnetotelluric (MT) field experiments were conducted along the Agulhas-Karoo Transect in<br />
South Africa (Weckmann et al., 2007, Stankiewicz et al., 2008). These transects cross several continental<br />
collision zones and their respective units, such as the Cape Fold Belt, the Namaqua Natal<br />
Mobile Belt and the Kaapvaal Craton. The MT profile on which we focus in this paper is located in<br />
CFB<br />
NNMB<br />
MT2 profile<br />
SCCB<br />
BMA<br />
MT<br />
NVR<br />
WRR<br />
Offshore seismics<br />
Kaapvaal<br />
Craton<br />
Karoo<br />
Agulhas-Falkland Fracture Zone<br />
Agulhas<br />
Plateau<br />
Transkei<br />
Basin<br />
Figure 1: Left: Map of the Agulhas-Karoo Geoscience Transect with MT Profiles (red lines) across<br />
Cape Fold Belt , the Namaqua Natal Mobile Belt and the Kaapvaal Craton. Right: The profile MT2<br />
contains 46 broad band MT sites and 8 broad band / long period MT sites. It extends 120 km from<br />
Prince Albert in the North to Mosselbay in the South and covers the entire Cape Fold Belt.<br />
the Cape Fold Belt. In total, 54 MT sites were deployed along this 120 km long profile MT 2 (Fig. 1).<br />
It extends from Prince Albert in the North to Mosselbay in the South and covers the entire Cape Fold<br />
Belt (CFB), its inliers, the Oudtshoorn and the Kaaimans Basins, the Swartberg and the Outeniequa<br />
Mountain ranges and several major thrusts and faults.<br />
3 Data and isotropic 2D inversion<br />
Along the profile MT2 we acquired 5-component MT data at all stations in a period range from<br />
0.001s to 1000s using GPS synchronized S.P.A.M. MkIII (Ritter et al., 1998) and CASTLE broadband<br />
instrument. Metronix MFS06/06 induction coil magnetometer and non-polarizable Ag/AgCl telluric<br />
electrodes were used to record natural magnetic and electric field variations. The data were processed<br />
with the EMERALD software package (Ritter et al., 1998) using both robust single site and remote<br />
reference techniques. At some sites, for which a suitable reference site was not available and the<br />
data were affected by cultural noise (extensive farming), we applied the frequency domain selection<br />
scheme after Weckmann et al. (2005) to improve data quality.<br />
Mozambique Ridge<br />
SCCB<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
89
Figure 2: Pseudo-sections of TE and TM mode apparent resistivity and phase (x direction pointing<br />
west).We can observe phase values over 90 degrees in the middle of the profile in both TE and TM<br />
component (marked with black ellipses in the upper panel). Two exemplary sites (site 121 and site<br />
116) are displayed as apparent resistivity, phase and induction arrows (Wiese convention) over period.<br />
The red arrows in the pseudo-sections show the location of the selected sites. We can identify that the<br />
phases, (especially in the TE component) leave the first quadrant at a minimum period of 10 s, but<br />
typically at longer periods of 100-1000 s.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
90
After data processing, we applied different strike and dimensionality analyses, e.g. the ellipticity<br />
analysis after Becken & Burkhardt (2004), obtaining an electromagnetic strike direction of −90 ◦ , i.e.<br />
East-West direction. This is in general compatible with the tectonic grain of the CFB. Furthermore,<br />
most of the data seem to be compatible with a 2D interpretation approach. However, in a limited area<br />
we observe impedance phases exceeding 90 degrees at long periods. They typically appear in both,<br />
TE and TM mode, and persit in either component even if we rotate the coordinate system (Fig. 2).<br />
Figure 3: TE and TM mode data used for isotropic 2D inversion. Prior to inversion they were rotated<br />
into a common coordinate system according to the geoelectric strike direction of ≈ 90 ◦ (obtained by<br />
strike and phase tensor analysis). Data with phases over 90 ◦ are discarded from the data base.<br />
For an isotropic 2D inversion approach we use RLM2D (Rodi & Mackie, 2001, implemented in<br />
the WinGlink software package). Before starting a 2D inversion we have to make sure that large phase<br />
values over 90 ◦ are excluded as they cannot be fitted by a 2D inversion approach. Figure 3 shows<br />
the data which were finally used for 2D inversion. Figure 4 displays our preferred 2D conductivity<br />
image of the upper 35 km together with a section of the geological map after Hälbich et al. (1993).<br />
The inversion was started from a homogeneous half-space of 100Ωm with τ = 10 using TE and<br />
TM component, with intermediate inclusion of the vertical magnetic transfer function. Within the<br />
framework of this work, we refrain from interpreting conductivity anomalies in a geological context,<br />
but focus on the area where phases over 90 ◦ occur. One of the most prominent conductivity anomalies<br />
is a triangular shaped, highly conductive structure in the middle of the profile. In this area we also<br />
observe phases leaving the quadrant. This conductivity anomaly is required by the data which is<br />
compatible with a 2D interpretation approach; however, we believe that the phases over 90 ◦ are caused<br />
by some electrically anisotropic structures in this area (in Fig. 4 between black dashed lines). We<br />
should also note that the deep part of this area only possesses a very limited resolution (in Fig. 4<br />
marked with semitransparent mask) because most of the data which relates to this part are excluded<br />
in order to satisfy the 2D isotropic procedure (see Fig. 3). Similar phase behaviour was explained<br />
with crustal anisotropy by Weckmann et al. (2003) and Heise & Pous (2003), where data within old<br />
continental collision zones in Namibia and on the Iberian Peninsula, respectively. In both cases the<br />
data could be fit by using a shallow and a deeper electrically anisotropic zone with different anisotropy<br />
strike.<br />
4 Anisotropic forward modelling<br />
Isotropic 2D inversion is adequate to explain the data in most parts along the profile but it is very<br />
unsatisfactory not being able to include phases greater than 90 degrees and thus neglect a substantial<br />
amount of data. In order to develop a 2D inversion with spatially constraint anisotropy, resolution<br />
tests and synthetic modelling studies are necessary. In a first step towards the constraint anisotropic<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
91
Figure 4: 2D inversion model together with a section from the geological map after Hälbich et al.<br />
(1993). The area where we observe phases over 90 ◦ is located where the inversion puts a triangular<br />
shaped high conductivity anomaly (area between black dashed lines). Zones of limited resolution are<br />
also in this area and marked with a semitransparent mask.<br />
inversion we included zones with electrical anisotropy in a 2D isotropic model and calculated its<br />
forward responses which aim to understand the anisotropy effect.<br />
4.1 Lateral extension<br />
Pek & Verner (1997) have suggested that a combination of two azimuthal anisotropies with anisotropy<br />
strikes perpendicular to each other could produce phases that exceed 90 ◦ . Based on the suggestion<br />
we study at first a combination of two different anisotropic blocks. The initial model (Fig. 5, left)<br />
consists of a 300m isotropic surface layer with a resistivity of 30Ωm and an anisotropic block starting<br />
at a depth of 300m embedded in a medium of 100Ωm. The principal resistivities of the block<br />
are σ1/σ2/σ3 = 50/0.5/50Ωm and the anisotropy strike αS is 120 ◦ . The block is underlain by an<br />
isotropic layer with a resistivity of 15Ωm. Beneath the isotropic layer a second anisotropic block<br />
with σ1/σ2/σ3 = 30/0.3/30Ωm and αS = 30 ◦ (perpendicular to αS of the first block) is embedded in<br />
an isotropic half-space with 100Ωm. The second, deeper anisotropic block has a lateral extension of<br />
15km in the first (Fig. 5, left upper panel), and 80km in the second (Fig. 5, left lower panel) model,<br />
respectively.<br />
The forward responses are displayed in figure 5 (right) as apparent resistivities and phases in xy<br />
and yx component, respectively. Comparing the responses of both models we see that the phases of<br />
yx component (Φyx) for the second model (Fig. 5, left lower panel) at sites above the first anisotropy<br />
block leave the quadrant at a period of ≈ 100s (Fig. 5, right lower panel), while they are smaller than<br />
90 ◦ (Fig. 5, right upper panel) for the first model (Fig. 5, left upper panel) with the narrower deep<br />
anisotropic block.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
92
Figure 5: Models (left) and their forward responses (right). The models differ only in lateral extension<br />
of anisotropy block. The lateral extension of the second block is about 15km in the first model (left<br />
upper panel) and 80km (left lower panel) in the second model. The major difference in the responses<br />
appears in the yx component. Phases over 90 ◦ occur if the lateral extension of the deeper anisotropy<br />
block is greater than it of the shallower anisotropy block.<br />
4.2 Depth<br />
In a second step, we used the same anisotropy parameters as described above. The initial model<br />
consists of a 300m isotropic surface layer and a isotropic half-space with resisitivity of 100Ωm. Two<br />
anisotropy blocks are embedded in the half-space. The first block started at depth of 300m and shaped<br />
as a trapezoid. In this attempt we vary the depth of the second block. In the first model (Fig. 6, left<br />
upper panel) the second block started at a depth of 1.8km and in the second model (Fig. 6, left lower<br />
panel) started at a depth of 6.8km.<br />
In the forward responses (Fig. 6, right) we see that the phases Φyx for the second model become<br />
larger than 90 ◦ at periods > 10s at sites above the first block, while they for the first model are<br />
all smaller than 90 ◦ . In this attempt the second anisotropy block has always sufficiently greater<br />
lateral extension than the first anisotropy block, but we only see the phase anomalies by model with<br />
anisotropy block in adequate depth. Besides the lateral extension of the anisotropy block, also the<br />
depth is one of those key conditions under which phase anomalies appear.<br />
4.3 Rotation angle<br />
In a third test we use a similar model as described above for step two. The model contains a 300m<br />
isotropic surface layer and an isotropic half-space with resisitivity of 100Ωm. Two anisotropic blocks<br />
are embedded in the half-space. The first block starts directly beneath the surface layer and the<br />
second block in a depth of 6km. They have the principal resistivities σ1/σ2/σ3 = 50/0.5/50Ωm and<br />
σ1/σ2/σ3 = 30/0.3/30Ωm, respectively. In this attempt we vary the anisotropy strike angle αS1 and<br />
αS2 for both blocks.<br />
The forward response for the model with αS1 = αS2 = 0 ◦ is displayed in the right upper panel<br />
of figure 7. The phases of the yx component approach 90 ◦ for long periods. For αS1 = 60 ◦ and<br />
αS2 = 120 ◦ (Fig. 7, left lower panel) the yx component phases become larger than 90 ◦ for periods<br />
> 10s, while they stay below 90 ◦ for αS1 = 90 ◦ and αS2 = 120 ◦ (Fig. 7, right lower panel). Comparing<br />
the three examples, we see that model with different combination of strike angle in both shallow and<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
93
Figure 6: Models (left) and their forward responses (right). The models differ by depth of the second<br />
anisotropy block. In the first model (left upper panel) the second block starts in a depth of 1.8km<br />
and in the second model (left lower panel) the block starts in a depth of 6.8km. The phases of yx<br />
component for the model with the deeper anisotropic block become larger than 90 ◦ at periods > 10s<br />
at sites above the first block. For the model with the shallower block all phases are smaller than 90 ◦ .<br />
αS1 and αS2<br />
αS1 = 0 ◦ and αS2 = 0 ◦<br />
αS1 = 60 ◦ and αS2 = 120 ◦ αS1 = 90 ◦ and αS2 = 120 ◦<br />
Figure 7: Model (left upper panel) and its forward responses with different anisotropy strike angles.<br />
The model contains the same surface layer, background medium and resistivities for both anisotropic<br />
blocks as the models in fig. 6. We vary the strike angles αS1 and αS2 for the shallow and the deep<br />
block (left upper panel). The forward responses are displayed as apparent resistivities and phases in<br />
xy and yx component.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
94
deep blocks produce forward responses with pronounced difference. This can be observed not only in<br />
yx component (ρa(yx) and Φyx ) but also in xy component (ρa(xy) and Φxy ). A possible explanation is<br />
that a change of strike angle forces the current flow to change its preferred direction from the shallow<br />
to the deep block. According to this fact we may conclude that the anisotropy strike angle influences<br />
forward responses significantly and phase greater than 90 ◦ will appear if the strike angle of both<br />
blocks differ by an adequate amount.<br />
5 Conclusions<br />
Phases greater than 90 ◦ can only be modelled by a combination of a deep and a shallow anisotropic<br />
block. We varied several model parameters to study the changes in forward responses. For our models<br />
we can conclude that phases out of the first quadrant occur when: (i) the anisotropy ratio (σmax/σmin)<br />
is high (for both blocks); (ii) the deep anisotropic block has a much larger lateral extension than the<br />
shallow block; (iii) the deep anisotropic block is located in greater depth; (iv) the angles of anisotropy<br />
of both blocks differ by a considerable amount. In our models the difference should be at least 45 ◦ so<br />
that the preferred direction of current flow changes significantly from the shallow to the deep block.<br />
References<br />
Becken, M., & Burkhardt, H. (2004). An ellipticity criterion in magnetotelluric tensor analysis.<br />
Geophys. J. Int., 159, 69–82.<br />
Eisel, M., & Haak, V. (1999). Macro-anisotropy os the electrical conductivity of the crust: a magnetotelluric<br />
study from the German continental Deep Drilling site (KTB). Geophys. J. Int., 136, 109–<br />
122.<br />
Hälbich, I. (1993). Cape Fold Belt – Agulhas Bank Transect across Gondwana suture, southern<br />
Africa. Geosci. Transect, 9, 18 pp. AGU, Washington, D. C.<br />
Heise, W., & Pous, J. (2003). Anomalous phases exceeding 90 ◦ in magnetotellurics: anisotropic<br />
model studies and a field example. Geophys. J. Int., 155, 308–318.<br />
Kellett, R. L., Mareschal, M., & Kurtz, R. D. (1992). A model of lower crustal electrical anisotropy<br />
for the Pontiac Subprovince of the Canadian Shield. Geophys. J. Int., 111, 141–150.<br />
Mareschal, M., Kellett, R. L., Kurtz, R. D., Ludden, J. N., Ji, S., & Bailey, R. C. (1995). Archaean<br />
cratonic roots, mantle shear zones and deep electrical anisotropy. Nature, 375, 134–137.<br />
O’Brien, D. P., & Morrison, H. F. (1967). Electromagnetic fields in an N-layer anisotropic half space.<br />
Geophysics, 32, 668–677.<br />
Pek, J., & Verner, T. (1997). Finite-difference modelling of magneto-telluric fields in two-dimensional<br />
anisotropic media. Geophys. J. Int., 128, 505–521.<br />
Reddy, I. K., & Rankin, D. (1975). Magnetotelluric response of lateerally inhomogeneous and<br />
anisotropic media. Geophysics, 40, 1035–1045.<br />
Ritter, O., Junge, A., & Dawes, G. J. K. (1998). New equipment and processing for magnetotelluric<br />
remote reference observations. Geophys. J. Int., 132, 535–548.<br />
Rodi, W., & Mackie, R. L. (2001). Nonlinear conjugate gradients algorithm for 2-d magnetotelluric<br />
inversion. Geophysics, 66, 174–187.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
95
Stankiewicz, J., Ryberg, T., Schulzw, A., Lindeque, A., Weber, M., & de Wit, M. (2007). Initial<br />
results from wide-angle seismic refraction lines in the southern Cape. South African Journal of<br />
Geology, 110, 407–418.<br />
Weckmann, U., Jung, A., Branch, T., & Ritter, O. (2007). Comparison of electrical conductivity<br />
structures and 2D magnetic modelling along two profiles crossing the Beattie Magnetic Anomaly,<br />
South Africa. South African Journal of Geology, 110, 449–464.<br />
Weckmann, U., Magunia, A., & Ritter, O. (2005). Effective noise separation for magnetic single site<br />
data processing using a frequency domain selection scheme. Geophys. J. Int., 161, 635–652.<br />
Weckmann, U., Ritter, O., & Haak, V. (2003). A magnetotelluric study of the damara belt in namibia<br />
– 2. MT phases over 90 ◦ reveal the internal structure of the Waterberg Fault/Omaruru Lineament.<br />
Phys. Earth Planet. Inter., 138(2), 91–112.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
96
Substitute models for static shift in 2D<br />
Kristina Tietze 1,2 , Oliver Ritter 1,2 , Ute Weckmann 1,3<br />
1 Helmholtz Centre Potsdam - German Research Centre For Geosciences <strong>GFZ</strong>, Department 2, Section 2.2, Telegrafenberg, 14473 Potsdam<br />
2 Free University Berlin, Department of Earth Sciences, Malteserstr. 74-100, 12249 Berlin<br />
3 University of Potsdam, Department of Geosciences, Karl-Liebknecht-Str. 24, 14476 Potsdam<br />
Contact: kristina.tietze@gfz-potsdam.de<br />
Introduction<br />
MT practitioners often down-weight apparent resistivity TE mode data prior to 2D inversion to avoid<br />
problems with static shift, obviously assuming that static shift of the TM mode is handled automatically<br />
by the inversion. Static shift is caused by conservation of charges at local conductivity discontinuities<br />
which are small with respect to the inductive scale length.<br />
Here we present a class of shallow conductivity anomalies which can produce significant up- or<br />
downwards shifted TM mode apparent resistivity curves (static shift in the TE mode cannot be<br />
simulated with 2D modeling). We examine how this static shift is reproduced by 2D inversion and show<br />
that the results are strongly influenced by grid design and regularization. We conclude that modern 2D<br />
inversion packages are not optimized to handle static shift. Our results also indicate that it is no good<br />
reason to assume that static shift cancels out on average.<br />
2D substitute structures for static shift<br />
As 2D models only allow for conductivity changes in one of the horizontal directions, static shift can only<br />
be accounted for in one component (TM) and re-produced by forward modeling.<br />
Fig. 1 (right): Starting from a 1D layered halfspace<br />
(LH) model (Fig. 1a) 2D inhomogeneities<br />
were added to the resistivity model just<br />
below/at the surface (Figs 1b-g, left panel). TE<br />
and TM model responses were calculated for<br />
frequencies between 10 3 Hz and 10 -3 Hz directly<br />
above the center of the inhomogeneities. The<br />
2D forward computations are compared to the<br />
1D LH results indicated by crosses (Figs 1b-g,<br />
right panel).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
97
Fig. 1b (left): A 15 m wide 10 m conductor with a<br />
thickness varying from 1 m to 5 m causes downward shift<br />
of the TM resistivity curves. The shift increases in<br />
dependence of the block thickness, i.e. the extent of the<br />
horizontal resistivity contrast. The amount of static shift<br />
which can be generated by extending the conductive inset<br />
downwards is limited. At some point the structure is<br />
becoming inductively effective.<br />
Fig. 1d (left): Enhancing the conductivity contrast by<br />
padding the conductor with resistive (10 3 m) cells TM<br />
downward shift is increased. Structures of this type can be<br />
observed in inversion models below static shift affected<br />
sites when static shift is not taken into account during<br />
inversion (cf. pane view in Fig. 2c (2)).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
98<br />
Fig. 1c (left): For a fixed thickness of<br />
3 m the lateral dimensions of the 10<br />
m-block were altered between 15<br />
m and 65 m. With increasing block<br />
width, the static shift effect becomes<br />
smaller as the lateral resistivity<br />
contrast which is responsible for the<br />
distortion of the electrical field<br />
moves away from the site location.<br />
Fig. 1e (left): Placing a highly<br />
resistive block (10 4 m) with a width<br />
of 15 m and varying thicknesses<br />
between 5 m and 110 m beneath the<br />
site results in upward shift of the TM<br />
mode. Static shift increases with<br />
increasing vertical block dimensions.
Fig. 1f (left): For a fixed thickness of<br />
55m the lateral extent of the 10 4 mblock<br />
was altered between 15 m and 65<br />
m. As seen for the 10 m-block, the<br />
static shift effect becomes smaller with<br />
increasing block width as the resistivity<br />
contrast distorting the electrical field<br />
moves away from the site location.<br />
Fig. 1g (left): Padding the resistive block<br />
with a thin column of conductive (10<br />
m) cells increases the upward shift of<br />
the TM resistivity curve slightly. The<br />
horizontal padding cells must be very<br />
small horizontally to prevent inductive<br />
effects.<br />
As expected, the TE resistivity curves and the phase curves of both, TM and TE, modes are mostly<br />
unaffected by these very small scale structures.<br />
Fig. 1h (left): Distortion of the TM<br />
apparent resistivities is due to static<br />
shift as the skin depth exceeds 15 km<br />
for =10 m and f=0.01 Hz. The<br />
downward shifting effect above the<br />
good conductor is significantly stronger<br />
than the upward shift above the<br />
resistive block, although resistivity<br />
contrasts of the two blocks to the<br />
background are the same. Static shift<br />
values along this profile do not sum up<br />
to zero. The figures above indicate that<br />
structures causing upwards shift have to<br />
have a larger vertical extent to produce<br />
the same amount of shift.<br />
So, it is at least questionable if a zero sum/average assumption for static shift values, which is often<br />
applied in inversion schemes, is appropriate.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
99
2D Inversion<br />
Fig. 2a (left): A 2D resistivity model was<br />
created by adding a low and a high<br />
resistive block to the layered half-space<br />
model (cf. Fig. 1a). Data were calculated<br />
for 50 sites along a 50km long profile with<br />
1 km site spacing adding 2.5% Gaussian<br />
noise.<br />
Fig. 2b (above): Right: Synthetic shift values c were generated obeying a modified logarithmic Gammadistribution:<br />
log10(c) = -(0.5,2)+0.5 and shift=c. The maximum of the modified distribution is at 10 0 ,<br />
the expected value is 10 -0.5 . Left: Shift was applied randomly to 2/3 of the sites, independently for TE<br />
and TM. The left panel shows the applied shift values for each site along the profile.<br />
Fig. 2c (below): Inversion results for un-shifted (1) and shifted (2)-(4) data for 100 iterations starting<br />
from a 100m homogeneous half space, applying a uniform smoothing with =10. Error floors were 2%<br />
(0.6°) for the phases and 5% and 500% for TM and TE apparent resistivities, respectively.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
100
Fig. 2c (above, continued): The model grid was created setting the column widths to 0.5 times of the<br />
minimum skin depth pmin, for (1) and (2), and was refined to 0.1*pmin for (3) and (4) (cf. pane views).<br />
(2) & (3): As smoothing (regularization) is working against sharp resistivity contrasts, the inversion<br />
introduces complex and wide-stretched structures below the sites affected by static shift. The refined<br />
grid in (3) reduces smoothing lengths for uniform regularization. As a consequence, introduced<br />
compensatory structures appear simpler or diminish.<br />
(4) Inversion result after introducing a so-called “tear zone” (feature of WinGLink) at site 114 (red<br />
outline). Tear zone inversion does not penalize sharp conductivity contrasts at the outline of the tear<br />
zone. Now, the inversion can properly model static shift. The model can account for TM static shift<br />
within this zone while obtaining surrounding resistivity values close to the original ones.<br />
Fig. 2d (right) shows un-shifted data,<br />
shifted data as symbols and the inversion<br />
results (cf. Fig. 2c) for sites 114 and 133 as<br />
lines. For periods longer than 10 -2 s shifted<br />
data are fitted well by all inversions. For<br />
shorter periods, the inversion struggles to<br />
fit the data, especially at site 114 where<br />
TM resistivities were shifted downwards by<br />
two decades. TM phase responses below<br />
10 -2 s deviating from forward data clearly<br />
show the inductive effects of the near<br />
surface structures. Static shift effects<br />
decrease with the refined grid (3) and<br />
introduction of a tear zone (4).<br />
Summary<br />
Static shift for the TM mode can easily be produced by a range of simple structures at surface being<br />
introduced to the regional resistivity models as substitutes for natural structures causing static shift in<br />
field data.<br />
To account for TM static shift within 2D inversion, model grids have to be very fine in the vicinity of sites<br />
with column widths and thicknesses much smaller than the induction volume of the highest frequency<br />
to be analyzed.<br />
Smoothing routinely applied in minimum structure inversions does not allow for the sharp conductivity<br />
contrasts required to produce sensible static shift. It would therefore be desirable to be able to apply<br />
different smoothing factors to a static shift compensating top layer and the remaining model.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
101
Die Übergangsimpedanz einer kapazitiv angekoppelten Elektrode<br />
Andreas Hördt, Peter Weidelt, Anita Przyklenk<br />
Institut für Geophysik und extraterrestrische Physik, TU Braunschweig<br />
Vorwort von A. Hördt<br />
Die in diesem Artikel vorgestellte Theorie beruht auf einer unveröffentlichen Arbeit von Prof.<br />
Peter Weidelt, der am 1.7.2009 unerwartet verstorben ist. Die Arbeit ist ein typisches Beispiel<br />
für seine außerordentliche Hilfsbereitschaft, für die er unter Kollegen in aller Welt bekannt<br />
war. Im Jahr 2007 hatte ich ihn um Unterstützung bei der Berechnung der<br />
Übergangsimpedanz einer kapazitiven Elektrode gebeten. Einige Wochen später hatte er die<br />
vollständige Lösung hergeleitet und in ein Programm umgesetzt. Er selbst hat die Aufgabe als<br />
nicht allzu schwierig oder wichtig empfunden, und hätte die Ergebnisse vermutlich nie<br />
veröffentlicht.<br />
1 Zusammenfassung<br />
Es wird eine Möglichkeit vorgestellt, die Impedanz einer Kreisscheibe zu berechnen, die an<br />
einen Halbraum mit endlicher Leitfähigkeit angekoppelt ist. Basierend auf den<br />
Maxwellgleichungen wird eine Integralgleichung für die Ladungsdichte q auf der<br />
Kreisscheibe aufgestellt. Mit der Randbedingung, dass das Potential auf der Scheibe konstant<br />
ist, kann die Ladungsdichte gender zu angehobener Platte stetig aund damit die Gesamtladung<br />
bestimmt werden. Es zeigt sich, dass der Übergang von aufliels Funktion des Abstandes<br />
verläuft. Zudem nimmt die Impedanz als Funktion des Abstandes zu, d.h. die Ankopplung<br />
einer kapazitiv angekoppelten Elektrode kann niemals geringer sein, als die einer<br />
aufliegenden mit gleicher Fläche. Je nach Modellparametern kann die Impedanz jedoch<br />
innerhalb weniger Nanometer über Größenordnungen variieren, so dass es dennoch günstig<br />
sein kann, die Elektrode zu isolieren, um starke Schwankungen der Impedanz zu vermeiden.<br />
2 Einleitung<br />
Bei geoelektrischen Messungen über sehr schlecht leitendem Untergrund kann es von Vorteil<br />
sein, den Strom kapazitiv anzukoppeln (Kuras et al., 2006). Dabei wird ein hochfrequentes<br />
Wechselfeld an eine Elektrode angelegt, die keinen direkten Kontakt mit dem Untergrund hat.<br />
Eine genaue Berechnung der Übergangsimpedanz einer solchen Elektrode ist notwendig, um<br />
zu beurteilen, unter welchen Bedingungen eine kapazitive Ankopplung einer galvanischen<br />
überlegen ist. Für die praktische Umsetzung ist die Frage von Bedeutung, ob es günstig ist,<br />
eine kapazitive Elektrode zu isolieren und einen galvanischen Kontakt auszuschließen. Die<br />
Standardformeln für die Ankopplung kapazitiver Elektroden sind allerdings nur für hohe<br />
Leitfähigkeiten des Untergrundes gültig. Hördt (2007) hat analytische Gleichungen für eine<br />
Kugelelektrode im sphärisch geschichteten Vollraum hergeleitet und diskutiert. Hier wird nun<br />
eine Lösung für eine zylindrische Scheibe über einem Halbraum hergeleitet.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
102
3 Modellbeschreibung<br />
Abbildung 1 illustriert das zu lösende Problem. Zu berechnen ist, welcher Strom I bei einer<br />
vorgegebenen Spannung U zwischen zwei zylindrischen Scheiben über einem homogenen<br />
Halbraum fließt.<br />
Abbildung 1: Skizze der kapazitiv angekoppelten Stromquelle: Zwei zylindrische Scheiben<br />
sind kapazitiv an einen homogenen Halbraum mit Leitfähigkeit und Permittivität <br />
angekoppelt, es fließt ein Strom I bei vorgegebener Spannung U.<br />
Wenn der Untergrund ein idealer Leiter ist, gilt bei kleinem Abstand zwischen Elektrode und<br />
Untergrund für die komplexe Übergangsimpedanz (Smythe, 1968):<br />
1<br />
Z (1)<br />
iC<br />
wobei die Kreisfrequenz ist und C die Kapazität, mit<br />
A<br />
0<br />
C (2)<br />
d<br />
Dabei ist A die Fläche der Elektrode, d der Abstand zum Untergrund, und 0 die<br />
Dielektrizitätszahl des Vakuums zwischen den Elektroden. diese Formeln stellen allerdings<br />
eine Näherung für einen idealen Leiter dar. Je nach Frequenz und Leitfähigkeit gilt die<br />
Näherung nicht mehr, und die elektrischen Parameter des Untergrundes müssen explizit<br />
berücksichtigt werden. Dies erfordert eine vollständige Lösung der Maxwellgleichungen.<br />
Die geometrischen Parameter des Modelles sind in Abbildung 2 illustriert. Die ideal leitende<br />
Kreisscheibe mit Radius a befindet sich im Abstand d über dem Untergrund. Der Mittelpunkt<br />
der Scheibe ist der Ursprung des zylindrischen Koordinatensystemes. Es wird angenommen,<br />
dass alle zeitlichen Variationen harmonisch sind, die zeitliche Abhängigkeit sich also durch<br />
e it darstellen lässt. Die Leitfähigkeit des Untergrundes lässt sich dann als komplexe Größe<br />
darstellen: 1 1 i<br />
1,<br />
wobei 1 die Gleichstromleitfähigkeit ist. Das Medium zwischen<br />
Untergrund und Scheibe hat die rein imaginäre Leitfähigkeit 0 i<br />
0 .<br />
Wenn an die Scheibe ein Potential U angelegt wird, stellt sich eine radialsymmetrische<br />
Ladungsverteilung q(r) ein. Die Gesamtladung Q erhält man dann aus<br />
a<br />
<br />
Q 2 qds<br />
s<br />
(3)<br />
0<br />
U<br />
~<br />
, <br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
103<br />
I
Abbildung 2: Parameter des Modelles der ideal leitenden Kreisscheibe über homogenem<br />
Halbraum.<br />
Es muss also die Flächenladungsdichte q(r) gefunden werden, aus der dann Q und mit<br />
Q<br />
C (4)<br />
U<br />
die komplexe Kapazität, bzw. mit Gl. (1) die komplexe Impedanz der Kreisscheibe berechnet<br />
werden kann.<br />
Aus der Forderung, dass die Leitfähigkeit auf der Scheibe unendlich ist, ergibt sich, dass das<br />
elektrische Feld auf der Scheibe verschwindet. Die Lösungsstrategie besteht darin, eine<br />
Gleichung für das elektrische Feld im gesamten Raum als Funktion von q(r) zu berechnen<br />
und dann q(r) so zu bestimmen, dass die Radialkomponente Er auf der Scheibe verschwindet.<br />
4 Theorie<br />
4.1 Grundgleichungen<br />
Um das elektrische Feld zu berechnen, müssen die Maxwellgleichungen in<br />
Zylinderkoordinaten gelöst werden. Aus dem Amperegesetz folgt unter der Berücksichtigung<br />
der harmonischen Zeitabhängigkeit:<br />
<br />
E<br />
E iEErot H<br />
(5)<br />
t<br />
mit dem elektrischen Feld E und dem Magnetfeld H. Die Leitfähigkeit und die<br />
Dielektrizitätszahl hängen von z ab. Aus Symmetriegründen besitzt das elektrische Feld nur<br />
eine Vertikal- und eine Radialkomponente, die nur von z und r abhängen. Das Magnetfeld<br />
besitzt nur eine von r und z abhängige -Komponente.<br />
Damit folgt aus Gl. (5):<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
104
H<br />
<br />
Er <br />
(6a,b)<br />
z<br />
1 r<br />
H <br />
E z <br />
r r<br />
Aus dem Induktionsgesetz:<br />
B<br />
rotE<br />
t<br />
folgt:<br />
(7)<br />
E<br />
z E<br />
r<br />
i<br />
0 H <br />
(8)<br />
r<br />
z<br />
<br />
Durch Einsetzen von (6a) und (6b) in (8) erhält man eine entkoppelte Gleichung für H:<br />
1 <br />
rHzzH i<br />
H <br />
1<br />
r r<br />
0<br />
(9)<br />
r <br />
<br />
<br />
wobei die Schreibweise r angewandt wurde.<br />
r<br />
Der erste Term in Gl. 9 legt eine Trennung der Variablen mit Hilfe der Besselfunktion 1.<br />
Ordnung mit der Wellenzahl u nahe:<br />
r, zfz,<br />
uJur<br />
H 1<br />
(10)<br />
denn die Besselfunktion erfüllt die Bessel’sche Differentialgleichung:<br />
ur r J1<br />
r<br />
2<br />
r J1<br />
1<br />
2<br />
r<br />
so dass mit (10) folgt:<br />
2<br />
ur J uru J ur 2<br />
rHuH 1<br />
1<br />
r r r<br />
1<br />
<br />
(12)<br />
<br />
Bei dieser Strategie besteht eine Analogie zur Fouriertransformation, bei welcher z.B. eine<br />
zweifache zeitliche Ableitung nach Transformation in den Frequenzbereich mit der Funktion<br />
e it zu einer Multiplikation mit – 2 führt. Entsprechend führt hier der Differentialoperator im<br />
1. Term in Gl. (9) zu einer Multiplikation mit –u 2 . Die Lösung von (9) lässt sich folglich<br />
darstellen als:<br />
z, uJurdu<br />
H r,<br />
z)<br />
f 1<br />
<br />
( <br />
(11)<br />
(13)<br />
0<br />
Die Wahl der Besselfunktion 1. Ordnung ergibt sich daraus, dass aus Symmetriegründen H<br />
bei r=0 verschwinden muss, was von J1 erfüllt wird, aber nicht von J0. Aus dem<br />
Amperegesetz (6a und 6b) ergeben sich damit direkt die Gleichungen für die Komponenten<br />
des elektrischen Feldes:<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
105
z, u<br />
1<br />
f<br />
Er ( r,<br />
z)<br />
J1<br />
z<br />
1<br />
E z ( r,<br />
z)<br />
u<br />
f 0<br />
<br />
<br />
0<br />
<br />
0<br />
urdu z, uJurdu<br />
Zur Bestimmung von f(z,u) wird zunächst Gl. (13) herangezogen. Durch Einsetzen in (9) sieht<br />
man unter Benutzung von (12), dass die Funktion f(z,u) die Gleichung<br />
(14)<br />
(15)<br />
1 2<br />
z z f f<br />
<br />
<br />
erfüllen muss, mit<br />
(16)<br />
2<br />
2<br />
i 0<br />
u<br />
(17)<br />
4.2 Bestimmung von f(z,u)<br />
Gleichung (14) muss unter der Berücksichtigung der Randbedingungen an den<br />
Schichtgrenzen und der Quellbedingung gelöst werden. Die Quelle wird durch einen Sprung<br />
der Vertikalkomponente des elektrischen Feldes an der Scheibe beschrieben:<br />
r <br />
E z<br />
wobei <br />
<br />
q<br />
r<br />
<br />
<br />
(18)<br />
<br />
0<br />
die Differenz der Werte unmittelbar oberhalb und unterhalb der Scheibe, also den<br />
Sprung des Funktionswertes, kennzeichnet.<br />
<br />
<br />
<br />
Ez0Ez0<br />
E (18a)<br />
<br />
Um aus (18) eine Bedingung für f(r,z) abzuleiten, transformien wir sie in den<br />
Wellenzahlbereich mittels Gl. (15), aus der folgt:<br />
<br />
1<br />
<br />
<br />
0<br />
<br />
<br />
( r,<br />
z)<br />
ufz,<br />
u<br />
J ur E z<br />
<br />
0<br />
r<br />
q<br />
, 0 r a<br />
0 <br />
du <br />
0,<br />
r a<br />
<br />
<br />
Die Ladungsdichte q(r) läßt sich mittels der Hankeltransformation in den Wellenzahlbereich<br />
transformieren:<br />
q~ <br />
<br />
0<br />
0<br />
u uqrJurdr<br />
(19)<br />
(20)<br />
Mit q u ~ als der Ladungsdichte im Wellenzahlbereich.<br />
Dementsprechend ist q(r) ist darstellbar als:<br />
q<br />
<br />
<br />
0<br />
0<br />
r uq~<br />
u J urdu (21)<br />
Durch Einsetzen von (21) in (19) wird klar, dass<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
106
0<br />
f z, u <br />
q~<br />
u (22)<br />
0<br />
d.h.die im Raumbereich formulierte Randbedingung (18) transformiert sich gemäß (22) in den<br />
Wellenzahlbereich.<br />
Nach der Formulierung der Randbedingungen wird nun die Differenzialgleichung für f gelöst.<br />
Aus der Form von Gleichung (16) ergibt sich direkt, dass die Lösungen Funktionen der Form<br />
z<br />
e <br />
sind, wobei berücksichtigt werden muss, dass im unteren Halbraum und in der Luft<br />
unterschiedlich sind. Daraus ergibt sich folgender Ansatz für die Funktion f(z,u):<br />
f<br />
z, u<br />
wobei<br />
0 Ae<br />
<br />
0z<br />
<br />
Ae<br />
<br />
<br />
Ce<br />
<br />
z<br />
<br />
<br />
Be<br />
<br />
z<br />
1<br />
Be<br />
,<br />
<br />
z<br />
<br />
z<br />
0<br />
,<br />
0<br />
,<br />
0 z d<br />
z d<br />
z 0<br />
(23a,b,c)<br />
2<br />
2<br />
2<br />
2<br />
0 i0 0<br />
u und 1 i 0<br />
1<br />
u<br />
(24a,b)<br />
die Parameter in Luft und im unteren Halbraum sind.<br />
Die Vorzeichen der Exponenten in den Ansatzfunktionen in (23) ergeben sich jeweils aus der<br />
Forderung, dass f(z,u) im Unendlichen sowohl für z0 verschwinden muss.<br />
Die Randbedingungen an Schichtgrenzen ergeben sich aus den Stetigkeitsbedingungen für die<br />
elektromagnetischen Felder. Da die Horizontalkomponenten von E und H an Schichtgrenzen<br />
1 f<br />
stetig sind, folgt mittels (13) und (14), dass f und stetig sein müssen.<br />
z<br />
Damit und mit der Sprungbedingung für f (Gl. (22)) lassen sich nun die Konstanten A,B und C<br />
in Gl. (23) berechnen. Beispielsweise folgt aus (23a,b), dass<br />
f<br />
z 0 <br />
_ B A<br />
z0 A B<br />
f<br />
<br />
und damit für den Sprung von f bei z=0:<br />
<br />
<br />
<br />
fz, u<br />
f z0fz02A (25a,b)<br />
(26)<br />
und mit (22) erhält man hieraus für A:<br />
0<br />
A q~<br />
u (27)<br />
2 0<br />
Die Bestimmung von B und C aus den Stetigkeitsbedingungen erfolgt analog, ist aber<br />
komplexer und wird hier nicht im Einzelnen ausgeführt. Man erhält folgendes Endergebnis<br />
für f:<br />
f<br />
<br />
2<br />
0 z, u<br />
q~<br />
u 0<br />
<br />
<br />
e<br />
<br />
<br />
e<br />
<br />
T<br />
<br />
0 z<br />
0 z 2<br />
d <br />
Rue<br />
0 z<br />
0 z2d Rue<br />
,<br />
1<br />
zd <br />
0d<br />
ue e ,<br />
mit Reflektions-und Tranmissionsfaktor<br />
,<br />
z 0<br />
0 z d<br />
z d<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
107<br />
(28a,b,c)
R<br />
T<br />
u u <br />
<br />
0<br />
u11u0 u11u0 2<br />
0u1<br />
u11u0 0 (29)<br />
(30)<br />
<br />
0<br />
Mit (28-30) und (13-15) lassen sich dann alle Feldkomponenten berechnen, wobei für die<br />
Bestimmung der Ladungsdichte nur die Gleichung für Er im Bereich zwischen Halbraum und<br />
Scheibe benötigt wird. Im Folgenden wird die Näherung 0=u verwendet. Im Vakuum ist<br />
nämlich<br />
2<br />
2 2 2<br />
2 2<br />
0 u <br />
0<br />
0 u u<br />
(24c)<br />
2<br />
c<br />
Die relevanten Wellenzahlen u ergeben sich aus der Dimension des Messystemes, welches bei<br />
den hier berücksichtigten Frequenzen in jedem Fall klein ist gegen die elektromagnetische<br />
Wellenänge im Vakuum. Es ist also u c und damit der 1. Term auf der rechten Seite<br />
vernachlässigbar.<br />
Die Gleichung für Er ergibt sich aus (28b) und (14), wobei die Ableitung von f nach z einen<br />
Faktor u unter dem Integral ergibt:<br />
E<br />
u<br />
z<br />
uz2d<br />
ueRueJurdu 0 z d<br />
<br />
1<br />
r ( r,<br />
z)<br />
uq~<br />
1<br />
2<br />
0 0<br />
4.3 Bestimmung der Ladungsdichteverteilung<br />
Zur Bestimmung der Ladungsdichteverteilung wird wir oben erläutert gefordert, dass die<br />
Radialkomponente des elektrischen Feldes auf der Scheibe verschwindet. Benötigt wird also<br />
das Feld bei z=0:<br />
E<br />
1<br />
<br />
rz0uq~ u1Ru , <br />
2u<br />
d<br />
e Jurdu r 1<br />
2<br />
0 0<br />
Die Idee ist nun, eine Ladungsverteilung q(r) zu suchen, so dass Er überall verschwindet.<br />
Damit lässt sich Gleichung (32) im Prinzip in ein Gleichungssystem überführen. Setzt man<br />
die Transformation der Ladungsdichte (Gl. 20) in (32) ein, so erhält man:<br />
1<br />
2<br />
a<br />
<br />
0 0 0<br />
sq<br />
2u<br />
d<br />
s J s u1R<br />
u e Jurduds 0<br />
0<br />
1<br />
Dabei wurde die neue Variable s eingeführt, die ebenso wie r von 0 bis a läuft, allerdings<br />
verschieden von r sein muss, da ja das elektrische Feld bei r von der gesamten<br />
Ladungsdichteverteilung abhängt. Definiert man nun die Funktion:<br />
F<br />
<br />
sr u 1<br />
Ru<br />
, <br />
0<br />
2u<br />
d<br />
e JusJurdu 0<br />
1<br />
(31)<br />
(32)<br />
(33)<br />
(34)<br />
so kann man Gl. (33) schreiben als:<br />
1<br />
2<br />
0<br />
a<br />
<br />
0<br />
s q<br />
s Fs,<br />
r<br />
ds 0<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
108<br />
(35)
Die Funktion F(s,r) lässt sich nach Gl. (34) für jedes s und r direkt durch Integration über u<br />
berechnen. Die Scheibe wird also geeignet diskretisiert, und mit den diskreten Werten für s<br />
und r erhält man eine Matrix F . Gleichung 35 wird dann zur linearen Gleichung<br />
F q 0<br />
(36)<br />
in der die Einträge von q die unbekannte Ladungsdichteverteilung darstellen. Durch<br />
Gleichung (36) ist q allerdings noch nicht vollständig bestimmt, da sie die triviale Lösung<br />
q 0 besitzt.<br />
Um eine endliche Ladungsdichte zu erhalten, benötigt man eine Gleichung für das Potential<br />
der Scheibe. Eine Möglichkeit besteht darin, ein aus 2 identischen Scheiben bestehendes<br />
System zu betrachten (Abb. 3).<br />
Abbildung 3: Geometrie des aus zwei identischen Scheiben aufgebauten Sendesytemes. Der<br />
Abstand der Mittelpunkte ist L, der Radius der Scheiben a.<br />
Die Integration von Er entlang der Verbindungslinie der beiden Scheibenmittelpunkte ergibt<br />
die angelegte Spannung U. Da Er=0 auf den Scheiben, gilt:<br />
LaLa<br />
Er0ELr, 0dr<br />
2 E r, <br />
r<br />
r <br />
U , 0 dr<br />
(37)<br />
a<br />
a<br />
r<br />
Er(r,0) wurde in Gl. (32) berechnet und lässt sich direkt einsetzen. Die Integration über r kann<br />
man ausführen, und man erhält:<br />
U 2<br />
1<br />
<br />
<br />
1<br />
<br />
<br />
La<br />
<br />
<br />
0 0<br />
<br />
<br />
0 0<br />
<br />
a<br />
uq~<br />
q~<br />
1<br />
2<br />
0 0<br />
uq~<br />
u1Ru 2u<br />
d<br />
e Jur u1Ru L<br />
a<br />
2u<br />
d<br />
e J1ur 2u<br />
d<br />
e J uaJuLa u1Ru <br />
<br />
a<br />
<br />
drdu<br />
<br />
du dr<br />
du<br />
0<br />
1<br />
0<br />
Um eine explizite Diskretisierung definieren zu können, wird die Ladungsdichte wieder<br />
mittels Gl. (20) im Raumbereich aufgeschrieben. Fasst man alle Terme mit u zusammen, so<br />
erhält man:<br />
1<br />
U <br />
<br />
1<br />
<br />
<br />
mit<br />
a<br />
<br />
0 0<br />
a<br />
<br />
0 0 0<br />
sq<br />
s J usds 1<br />
R<br />
u<br />
0<br />
sq<br />
s Gds<br />
s<br />
2u<br />
d<br />
e J uaJuLa <br />
0<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
109<br />
0<br />
du<br />
(38)<br />
(39)
G<br />
<br />
<br />
0<br />
2u<br />
d<br />
e JusJuaJuLa s R<br />
u<br />
0<br />
du<br />
1 (40)<br />
0<br />
0<br />
Gleichungen (39) und (40) werden wiederum durch Diskretisierung der Scheibe in eine<br />
Gleichung der Form<br />
q G U<br />
(41)<br />
umgesetzt.<br />
Gleichung (41) ist die inhomogene Bedingung, die nötig ist, um die triviale Lösung q 0<br />
auszuschließen und eine vom Potential abbhängige Ladungsdichte zu erhalten. Man wählt<br />
also ein beliebiges U und erhält aus der Lösung von (36) unter der Bedingung (41) eine<br />
Ladungsdichteverteilung q(r). Die Gesamtladung erhält man mit Gl. (3), und die Kapazität<br />
aus Gl. (4).<br />
Für ein aus zwei Scheiben bestehendes System gilt die Lösung nur in hinreichendem Abstand<br />
der Scheiben zueinander, da die gegenseitige Beeinflussung der Ladung auf den beiden<br />
Scheiben hier nicht berücksichtigt wurde. Eine quantitative Abschätzung dieses Effektes<br />
wurde bisher nicht vorgenommen. Betrachtungen mit halbkugelförmigen Elektroden im<br />
Gleichstromfall legen nahe, dass die vernachlässigten Terme von der Ordnung (a/L) 4 sind, so<br />
dass die Näherung schon bei geringen Elektrodenabständen gut funktioniert.<br />
Die berechnete Kapazität C der Elektrode ist komplex, da der Reflektionsfaktor (Gl. 29)<br />
komplex ist. Aus ihr lässt sich mit<br />
1<br />
Z <br />
iC<br />
die komplexe Impedanz der Elektrode, bzw. deren Betrag und Phase berechnen.<br />
5 Ergebnisse<br />
Um die Genauigkeit der Lösung zu bewerten, wurden die Ergebnisse mit einer analytischen<br />
Lösung für eine Kugelelektrode verglichen, die von Hördt (2007) vorgestellt wurde.<br />
Abbildung 4 zeigt die Geometrie und die Parametrisierung. Für die mittlere Schale mit Radius<br />
r1 wird hier Vakuum gewählt werden, mit 1=0 und 1=0 (Permitticität des Vakuums).<br />
Abbildung 4: Geometrie der Kugelelektrode. Eine ideal leitende Elektrode mit Potential V<br />
und Radius r0 ist in eine Kugelschale mit Radius r1 eingebettet.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
110
Die Impedanz der Elektrode ergibt sich dann zu:<br />
* *<br />
2 2 r1<br />
1<br />
* *<br />
1<br />
<br />
1 1 r0<br />
Z<br />
* *<br />
4<br />
r0<br />
1 2 r1<br />
*<br />
1 r0<br />
wobei<br />
*<br />
i<br />
<br />
<br />
<br />
<br />
<br />
<br />
In Abbildung 5 wird der Betrag der Impedanz der Scheibenelektrode mit der einer<br />
Kugelelektrode verglichen. Die Radien wurden so gewählt, dass die Oberfläche der Kugel<br />
gleich der einseitigen Fläche der Scheibe ist. Bei der Kugelelektrode wurde für die mittlere<br />
Schale die Werte für Vakuum eingesetzt, der Abstand entspricht der Dicke der mittleren<br />
Schale.<br />
Für mittlere Abstände stimmen die Impedanzen der Kugelelektrode und der<br />
Scheibenelektrode überein, bei großen und kleinen Abständen sind die Impedanzen<br />
unterschiedlich, was auch so zu erwarten ist. Der Grenzwert großer Abstände entspricht der<br />
Impedanz der Elektrode im Vakuum. Die Impedanzen beider Elektroden in Abb. 5<br />
konvergieren gegen den jeweiligen Grenzwert.<br />
Abbildung 5: Impedanzbetrag der kapazitiven Elektrode gegen Abstand der Elektrode zum<br />
Medium für verschiedene Leitfähigkeiten des Mediums. Durchgezogene Linie:<br />
Kugelelektrode mit Radius 0.05m. Gestrichelte Linie: Scheibenelektrode mit Radius 0.1m.<br />
Die Frequenz ist f=10 kHz, die relative Permittivität des Untergrundes ist in beiden Fällen<br />
r=4.<br />
Für die Kugel mit Radius r0 beträgt dieser (Smythe, 1968):<br />
(42)<br />
(43)<br />
1<br />
Z <br />
4<br />
i<br />
0 r0<br />
(44)<br />
und für eine Scheibe mit Radius a:<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
111
1<br />
Z (45)<br />
8i<br />
0 a<br />
Wenn a=2r0, wie für die Rechnung aus Abb. 5 der Fall, entspricht die Gesamtfläche der<br />
Kugel der einseitigen Fläche der Scheibe. Dann ist die Vakuumimpedanz der Scheibe etwas<br />
kleiner, weil im Vakuum beide Seiten der Scheibe gleich wirksam sind.<br />
Für kleine Abstände ist der Vergleich von der Leitfähigkeit des Untergrundes abhängig. Bei<br />
moderaten Leitfähigkeiten (rote und grüne Kurve) konvergiert der Impedanzbetrag gegen den<br />
Grenzwert der galvanischen Ankopplung, also den Fall, dass die Scheibe aufliegt, bzw. die<br />
Kugel direkten Kontakt zum leitenden Medium hat. Der Grenzwert ist für die Kugel<br />
1<br />
Z (46)<br />
4<br />
r0<br />
und für die kreisförmige Scheibe<br />
1<br />
Z (47)<br />
4<br />
a<br />
wobei die Gleichstromleitfähigkeit ist.<br />
In diesem Fall liefert die Scheibe bei gleicher Oberfläche (mit a=2r0) die etwas größere<br />
Impedanz. Eine Gleichheit ist hier nicht zu erwarten, da nicht nur die Fläche wichtig ist,<br />
sondern auch die Stromverteilung im Medium. Bei sehr hohen Leitfähigkeiten in der<br />
Größenordnung von Metallen (blaue Kurve) wird der Grenzwert auch für extrem kleine<br />
Abstände noch nicht erreicht. Bei =0 (schwarze Kurve) sind nur die Verschiebungsströme<br />
(mit r=4) zur Ankopplung wirksam. Die Scheibe hat eine etwas höhere Impedanz verglichen<br />
mit der Kugel.<br />
Insgesamt verhalten sich die Kurven wie erwartet. Im mittleren Abstandsbereich ist nur die<br />
Ankopplungsfläche relevant und die Impedanzen von Kugel und Scheibe stimmen überein.<br />
Bei sehr großen und sehr kleinen Abständen werden jeweils die analytischen Grenzwerte<br />
erreicht, die für Kugel und Scheibe unterschiedlich sind.<br />
Der Verlauf der Impedanz in Abbildung 5 hat zwei wichtige Konsequenzen. Zum Einen<br />
nimmt die Impedanz stetig als Funktion der Höhe zu. Dies schließt den Grenzwert einer<br />
galvanisch gekoppelten Elektrode ein. Das bedeutet, dass eine kapazitiv gekoppelte Elektrode<br />
niemals eine geringere Impedanz haben kann, als eine galvanisch gekoppelte, aufliegende.<br />
Zum Anderen ist der Übergang von der aufliegenden zur abgehobenen Elektrode stetig. Dies<br />
könnte der Intuition widersprechen; man würde evtl. einen Sprung der Impedanz zwischen<br />
aufliegender und abgehobener Impedanz erwarten.<br />
Um das beobachtete Verhalten zu verstehen, ist in Abb. 6 die Impedanz in Real-und<br />
Imaginärteil aufgespalten dargestellt, mit anderem Abstandsbereich. Der Realteil (Mitte) der<br />
Impedanz ist für sehr große und sehr kleine Leitfähigkeiten verschwindend gering bzw.<br />
identisch null und wird nicht mit dargestellt. Bei moderaten Leitfähigkeiten (grüne und rote<br />
Kurve) wird die Impedanz durch den Realteil dominiert. Bei extrem niedriger und bei extrem<br />
hoher Leitfähigkeit überwiegt der Imaginärteil.<br />
Dieses Verhalten lässt sich verstehen, wenn man die Gesamtimpedanz näherungsweise als<br />
Reihenschaltung zwischen einem kapazitiven Anteil und einem Erdungswiderstand betrachtet.<br />
Der kapazitive Anteil Zc entspricht der Überwindung des Raumes zwischen Boden und Luft<br />
und ist rein imaginär, der Erdungswiderstand Ze ist komplex und entspricht dem<br />
Ankopplungswiderstand der aufliegenden Platte. Man kann daher näherungsweise schreiben:<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
112
d<br />
1<br />
Z Z c Z e <br />
(48)<br />
2<br />
i<br />
0<br />
a 4ai01<br />
Zunächst einmal erklärt das Modell, warum der Übergang zwischen aufliegender und<br />
abgehobener Platte stetig ist und kein Sprung auftritt. Wenn der Abstand d0, beginnt der 2.<br />
Term zu dominieren und „übernimmt“ stetig die Ankopplung. Allerdings ist in der Abbildung<br />
zu erkennen, dass der Übergang bei extrem kleinen Abständen erfolgt, die sogar unterhalb<br />
eines Atomdurchmessers liegen können. In der Praxis wird der stetige Übergang in der Regel<br />
also nicht zu sehen sein, sondern als Sprung erscheinen.<br />
Abbildung 6: Betrag der Impedanz (oben), Realteil (mitte) und Imaginärteil (unten) für die<br />
Scheibe für dasselbe Modell wie in Abb. 5. Der Realteil der Impedanz (mitte) für =0 ist<br />
identlisch null und nicht darstellbar, der Realteil für die sehr gut leitende Platte ist extrem<br />
klein und wird nicht dargestellt.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
113
Das Modell aus Gl. (48) erklärt auch die Zerlegung der Impedanz in Real-und Imaginärteil.<br />
Bei sehr kleiner Leitfähigkeit und kleinem Abstand dominiert der Imaginärteil der Impedanz,<br />
bei moderater und hoher Leitfähigkeit dominiert der Realteil. Bei sehr hoher Leitfähigkeit<br />
wird der Abstand, in dem der 1.Term vernachlässigbar ist, in der Abbildung nicht erreicht und<br />
es dominiert immer noch der rein imaginäre 1. Term die Gesamtimpedanz.<br />
6 Schlussfolgerungen<br />
Es wurde ein Programm entwickelt, um die Impedanz einer kreisförmigen Scheibe über einem<br />
homogenen Halbraum zu berechnen. Das Verfahren beruht auf einer vollständigen Lösung<br />
der Maxwellgleichungen. Es verwendet eine Näherung, bei der angemommen wird, dass der<br />
Radius der Scheibe klein ist gegen die elektromagnetische Wellenlänge. Die Näherung<br />
verliert bei sehr hohen Frequenzen ihre Gültigkeit, ist jedoch beispielsweise für einen Radius<br />
0.1m und eine Frequenz von 1 MHz vollkommen unproblematisch. Das Programm wurde<br />
durch Vergleich mit einer analytischen Lösung für eine Kugelelektrode und durch analytische<br />
Grenzwerte für große und kleine Abstände verifizert.<br />
Der Übergang zur Impedanz einer aufliegenden Elektrode erfolgt stetig als Funktion des<br />
Abstandes zwischen Elektrode und Halbraum. Die Impedanz der aufliegenden Scheibe ist<br />
immer kleiner als die der angehobenen. Dies bedeutet, dass die Übergangsimpedanz nicht<br />
durch eine kapazitive Elektrode gegenüber einer gleich gebauten, galvanischen verringert<br />
werden kann. Der galvanische Anteil der Impedanz ist auch bei kapazitiver Ankopplung mit<br />
enthalten.<br />
In der Praxis kann es dennoch sinnvoll sein, kapazitive Elektroden statt galvanische zu<br />
verwenden, also den galvanischen Kontakt durch Isolierung oder Anheben zu verhindern. Bei<br />
Aufliegenden Elektroden ist die Impedanz sehr schwer zu kontrollieren, da die Auflagefläche<br />
stark variieren kann. Dies kann dazu führen, dass die Impedanz um Größenordnungen<br />
schwankt, da sie sehr stark als Funktion des Abstandes variiert und die aufliegenden Teile die<br />
Ankopplung dominieren können. Eine Isolierung stabilisiert die Impedanz und verhindert<br />
starke Schwankungen, was für die Messung von Vorteil sein kann.<br />
7 Referenzen<br />
Hördt, A., 2007, Contact impedance of grounded and capacitive electrodes, in: Ritter, O.,<br />
Brasse, H., Protokoll zum 22. Kolloquium „Elektromagnetische Tiefenforschung“, 164-173.<br />
Kuras, O., Beamish, D., Meldrum, P.I., and Ogilvy, R.D., 2006, Fundamentals of the<br />
capacitive resistivity technique. Geophysics, 71, G135-G152.<br />
Smythe, W., 1968. Static and dynamic electricity, McGRaw-Hill.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
114
H2O <br />
<br />
<br />
<br />
<br />
<br />
<br />
H2O <br />
<br />
H2O <br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
H2O <br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
H2O <br />
<br />
<br />
<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
115
5mm <br />
<br />
<br />
<br />
<br />
<br />
<br />
±0.5mm <br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
116
EG EB<br />
EN EB <br />
dx =50m<br />
<br />
By<br />
<br />
<br />
<br />
<br />
<br />
(EG)<br />
(EN) <br />
<br />
<br />
(EB) <br />
<br />
<br />
EGel <br />
Enorm <br />
<br />
(BY ) <br />
<br />
EGel Enorm BY <br />
<br />
<br />
<br />
16 2/3Hz 50Hz <br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
0 − 250Hz <br />
<br />
<br />
<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
117
EGel Enorm EnormEGel <br />
EGel/BY <br />
Enorm/BY EGel Enorm<br />
<br />
0 − 250Hz <br />
EGel Enorm <br />
<br />
Txy = <br />
< |Y | ><br />
X, Y EGel, Enorm, BY <br />
EGel = TxyEnorm <br />
Enorm = TxyEGel <br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
118
204Hz<br />
8Hz <br />
<br />
<br />
EGel = TxyBy Enorm =<br />
TxyBy <br />
ρa <br />
EGel/BY Enorm/BY <br />
<br />
204Hz 1Hz <br />
<br />
<br />
<br />
H2O<br />
<br />
<br />
<br />
<br />
<br />
• <br />
• <br />
<br />
<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
119
Attitude Algorithm Utilised in Mobile Geophysical<br />
Measuring Systems<br />
Johannes B. Stoll, Celle<br />
Christopher Virgil, Institut für Geophysik und extraterrestrische Physik, TU-Braunschweig<br />
1 Introduction<br />
Accurate real-time tracking of orientation or attitude of rigid bodies has wide applications in robotics,<br />
aerospace, on-land and underwater vehicles, automotive industry, and virtual reality. A quite recent<br />
application in geophysics is the analysis of the attitude effect of geophysical sensors such as coil<br />
arrays or magnetometers attached to mobile measuring platforms e.g. helicopters and fixed wing<br />
airplanes and the correction with respect to the Earth’s reference coordinate frame (Reid etal, 2003,<br />
Yin and Fraser, 2004). Another field of application is wireline logging with a borehole tool. While<br />
measuring, the tool spins about its vertical body axis due to the torsion force of the logging cable,<br />
which results in a disorientation of geophysical sensors.<br />
In the borehole industry, it is essential to accurately monitor and guide the direction of the drill bit. It is<br />
also necessary for an oil rig to log the location of its boreholes at a regular frequency such that the oil<br />
rig can be properly monitored. To determine the location of a drill bit in a borehole, it is necessary to<br />
know the position and the attitude, which includes the vertical orientation and the North direction.<br />
In prior art systems, a magnetometer is used to determine the magnetic field direction from which the<br />
direction of North is approximated. Triple component magnetometers were widely used to sense the<br />
tool rotation during a log run. However orientation via magnetometers is subject to restrictions. The<br />
employing of magnetometers in boreholes drilled into strongly magnetized formations such as oceanic<br />
crust, volcanoes and ore deposits, make it very difficult to utilize the Earth’s magnetic field as a<br />
reference (Steveling et al, 2003). First, these systems must make corrections for magnetic<br />
interference and use of magnetic materials for the drill pipe. Second, systems that rely only on<br />
magnetometers to determine North can suffer accuracy degradation due to the Earth's magnetic field<br />
variations. Third, a magnetometer alone cannot give an unequivocal measurement of a set of sensor<br />
attitude. Measurements made with such a sensor define the angle between the Earth’s magnetic field<br />
and a particular axis of the sensor. However, this axis can lie anywhere on the surface of a cone of<br />
semi-angle equal to that angle about the magnetic vector. Hence, an additional measurement is<br />
required to determine attitude with respect to another fixed reference frame.<br />
In order to determine the direction of a spinning borehole tool another type of sensor technology<br />
must be utilised to provide orientation that is independent from the Earth’s magnetic field as a<br />
reference (Galliot et al, 2004, Stoll and Leven, 2002).<br />
The problem to be solved consists of a way to sense the attitude of a borehole tool, which are free to<br />
rotate about any direction. A good compromise between accuracy and sensor size are the fibre optical<br />
gyros. These devices measure the turning rate about the sensor axis in the tool frame. In this paper an<br />
algorithm is presented that approaches the attitude determination based on this discrete data.<br />
2 Principle of Fibre Optical Gyro (FOG)<br />
Gyroscopes are used in various applications to sense the angular rate of turn about some defined<br />
axis. The most basic and the original form of gyroscopes make use of the inertial properties of a<br />
wheel, or rotor, spinning at high speed. The wheel tends to maintain the direction of its spinning axis in<br />
space due to conservation of the angular momentum vector. These devices are susceptible to<br />
damage from shock and vibration, exhibit cross-axis acceleration sensitivity and, for the lower cost<br />
versions, have reliability problems.<br />
Another type is the vibratory gyroscope. The basic principle of operation of such sensors is that the<br />
vibratory motion of part of the instrument creates an oscillatory linear velocity. If the sensor is rotated<br />
about an axis orthogonal to this velocity, a Coriolis acceleration is induced. The acceleration modifies<br />
the motion of the vibrating element which is an indication of the magnitude of the applied rotation.<br />
However, this type of sensor tends to produce biases in the region of 1°/s.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
120
Optical gyroscopes use an interferometer to sense angular motion. An optical gyroscope, laser or<br />
fibre, measures the interference pattern generated by two light beams, traveling in opposite directions<br />
within a mirrored ring or fibre loop, in order to detect very small changes in motion. This type of<br />
gyroscope can be subdivided into fibre optical gyro (FOG), ring laser gyro (RLG), and ring resonator<br />
gyro (PARR).<br />
In order to keep track on the orientation of the geophysical sensors the utilisation of fibre optical<br />
gyroscopes are now common in navigation and replace prior art systems like mechanical gyros.<br />
Important attributes of this new technology are no moving parts, high reliability, stable performance<br />
and low costs. Many of its components are based on proven technology from the fiber optical<br />
telecommunications industry. Using optical gyros, inertial navigation is accomplished by integrating the<br />
output of a set of angular rate sensors to compute the attitude.<br />
In figure 1 a FOG manufactured by LITEF is shown exemplary. The gyro operates at an optical<br />
wavelength of 820 nanometers, with a 110 meter coil of elliptical-core polarization maintaining fiber.<br />
The low coherence reduces unwanted interference between waves reflected from the fusion splices<br />
used to join the components.<br />
Fig. 1: Example of the FORS -family (LITEF, Freiburg) (here FORS-6U)<br />
3 Attitude Computation<br />
Inertial orientation tracking of borehole tools is based upon the same methods and algorithm as<br />
those used for aircrafts, ships, and missiles. There is a large quantity of technical literature that<br />
describes the fundamentals of inertial navigation technology in great detail, e.g., Grewal et al. (2001),<br />
v. Hinüber (1993), Savage (1998a, b), Stovall (1997). If angular rates of a tool are measured<br />
constantly with depth, the orientation of the sensors of the tool with respect to inertial space can be<br />
determined by applying a suitable coordinate transformation.<br />
A transformation from one coordinate frame to another can be carried out as three successive<br />
rotations about different axes. The transformation matrix is given by the product of these three<br />
separate transformations as follows:<br />
C n<br />
b<br />
C C<br />
C<br />
(1)<br />
3<br />
Various mathematical representations can be used to define the attitude of a body with respect to a<br />
coordinate reference frame. One of the most common ways of parameterizing the transformation<br />
matrix is by use of direction cosine matrix (DCM). Other methods of describing rotations are the use of<br />
Euler Angles and Quaternions.<br />
The DCM is a 3x3 matrix, the columns of which represent unit vectors in body axes projected along<br />
n<br />
the reference axes. Matrix Cb<br />
describes the transformation from coordinate frame “b” to the frame “n”<br />
and is written here in component form as follows:<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
121<br />
2<br />
1
c<br />
<br />
c<br />
<br />
c<br />
31<br />
c<br />
32<br />
33<br />
<br />
<br />
<br />
<br />
<br />
11 12 13<br />
C 21 c22<br />
c23<br />
n<br />
b (2)<br />
c<br />
The element in the i th row and the j th column represent the cosine of the angle between the i th axis of<br />
the reference frame “n” and the j th axis of the body frame “b”.<br />
When designing a navigation system it is necessary to relate the information from the sensors to a<br />
navigation coordinate. The coordinate choice for borehole applications is a geographic or navigation<br />
frame “n” with axes {N, E, D}, (North, East and Down). The inertial measurement unit is mounted on<br />
the borehole tool constituting a new frame. This is called the body frame “b”, and has axes {X, Y, Z}.<br />
This frame will be in rotation with respect to the geographic frame. The velocity of this rotation is<br />
measured by three orthogonal gyros. The transformation matrix that relates “b” and “n” coordinate<br />
frames propagates with time in accordance with the following equation:<br />
0 <br />
z y<br />
<br />
<br />
<br />
n n b<br />
C b<br />
b Cb<br />
nb<br />
, where nb<br />
z<br />
0 <br />
x (3)<br />
<br />
<br />
<br />
y<br />
x<br />
0 <br />
b<br />
n<br />
nb is the skew symmetric matrix formed from the elements of the vector b x y z , which<br />
th th<br />
represents the turn rate of the body between the i -frame and the (i+1) -frame as measured by the<br />
gyroscopes in the body frame. In real time applications the integration is implemented with the<br />
following approximation<br />
t<br />
C<br />
n<br />
b<br />
c<br />
c<br />
n t t<br />
C t A<br />
t<br />
(4)<br />
b<br />
T , <br />
,<br />
A t<br />
A t<br />
where is the DCM which relates the b-frame at time t to the b-frame at time t . For small<br />
angle rotations, may be written as follows:<br />
<br />
where I is the 3×3 identity matrix and<br />
0<br />
<br />
<br />
<br />
<br />
<br />
t I A (5)<br />
<br />
0<br />
<br />
<br />
<br />
<br />
0 <br />
<br />
in which , and are small rotation angles through which the body-frame has rotated over the<br />
time interval t about its yaw, pitch and roll axes respectively. If the limit t approaches zero, small<br />
angle approximations are valid and the order of the rotation becomes unimportant.<br />
Hence, the minimum sampling time of the gyros is important for obtaining the transformation matrix<br />
with reasonable accuracy. This interval will be a function of the severity of manoeuvres expected from<br />
the borehole tool. In many applications the rotation velocity expected is usually less than 25<br />
degrees/sec. With a sampling time of 100 Hz the maximum angle variation will be less than 0.25<br />
degree satisfying the small angle approximation. Rotations with faster dynamics will require smaller<br />
sampling time to compute the transformation matrix appropriately.<br />
n<br />
In order to update the DCM Cb<br />
, it is necessary to solve a matrix differential equation (3). Over a<br />
single computer cycle, from time tk<br />
to tk+1, the solution of the equation may be written as:<br />
k 1<br />
(6)<br />
C C exp dt (7)<br />
k 1<br />
k<br />
Provided that the orientation of the turn rate vector remains fixed in space over the update interval,<br />
we may define:<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
122<br />
t<br />
t<br />
k
and<br />
t<br />
k 1<br />
t<br />
k<br />
dt <br />
(8)<br />
CkAk C C exp <br />
k 1<br />
k<br />
(9)<br />
Where Ck represents the direction cosine matrix which relates body to reference axes at the k th<br />
computer cycle, and Ak the direction cosine matrix which transforms a vector from body co-ordinates<br />
at the k th computer cycle to body co-ordinates at the (k+1) th computer cycle.<br />
The variable is an angle vector with direction and magnitude such that a rotation of the body frame<br />
about through an angle equal to the magnitude of will rotate the body frame from its orientation at<br />
computer cycle k to its position at the computer cycle k+1. The components of are denoted by x,<br />
y, and z and its magnitude given by:<br />
and<br />
<br />
<br />
(10)<br />
2<br />
x<br />
2<br />
y<br />
2<br />
z<br />
0 <br />
z y <br />
<br />
<br />
z 0 <br />
x (11)<br />
<br />
<br />
<br />
y x 0 <br />
Expanding the matrix exponential function in a power series gives:<br />
where<br />
Thus we may write<br />
A<br />
k<br />
2<br />
3<br />
4<br />
<br />
Ak I <br />
(12)<br />
2!<br />
3!<br />
4!<br />
2<br />
2 2 <br />
y <br />
x<br />
<br />
x<br />
y<br />
z<br />
and which may be written as follows:<br />
<br />
<br />
z<br />
<br />
<br />
2 2 <br />
x<br />
x<br />
y<br />
<br />
<br />
3 2 <br />
<br />
2 2<br />
4<br />
<br />
<br />
y<br />
z<br />
z<br />
<br />
<br />
<br />
x<br />
<br />
<br />
<br />
2 2<br />
<br />
(14)<br />
(15)<br />
I <br />
2!<br />
<br />
2 4 <br />
I <br />
1<br />
...<br />
<br />
3!<br />
5!<br />
<br />
<br />
<br />
<br />
2 2<br />
3!<br />
2<br />
<br />
1<br />
<br />
2!<br />
2<br />
<br />
4!<br />
x<br />
y<br />
z<br />
z<br />
4!<br />
4<br />
<br />
6!<br />
y<br />
<br />
(13)<br />
2 <br />
...<br />
<br />
<br />
1cos 2 <br />
sin <br />
Ak I <br />
(17)<br />
2<br />
<br />
Equation (17) is input in equation (9) that updates the frame at time t to the frame at time t+t. To<br />
compute the attitude of the tool, first the measured rotation rates are input in equation 17. Then,<br />
beginning with C0 = I the rotation matrices Ck+1 for each computer cycle k+1 are successively<br />
computed by equation (9). The tool attitude r in the external frame at the k th cycle results to<br />
n<br />
k<br />
2<br />
...<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
123<br />
<br />
<br />
<br />
(16)
n<br />
with the starting orientation .<br />
r 0<br />
<br />
<br />
(18)<br />
n<br />
n<br />
rk Ck<br />
r0<br />
4 Attitude Computation allied to the Göttingen Borehole Magnetometer (GBM)<br />
The Göttingen Borehole Magnetometer (GBM) contains a triple of FORS-36M sensors, which are<br />
attached rigidly to the borehole tool. The rotation axes of the sensors are aligned with the body axes of<br />
the tool, and thereby sense the rotation rate about the vertical axis and the tilting movement about the<br />
two horizontal axes. Angular rates are output with a sample rate of 1 second. The salient<br />
specifications of the FORS-36m are the maximum input rotation rate of ±720°/s that allows sensing a<br />
maximum rotation rate of 2 revolutions per second with a resolution of 9·10 -5 °. Moreover, the<br />
dimensions of the sensor (53mm x 58mm x 19mm) are small enough to fit to the inner diameter of<br />
65mm of the GBM. This is important to reduce the overall diameter of the tool and make it applicable<br />
in narrower boreholes. The bias drift is 6/h (1) and is reduced to
Steveling, E., J.B. Stoll, and M. Leven, 2003. Paleomagnetic age dating from magnetisation and<br />
inclination studies of the Mauna Kea/Hawaii, Geochem.Geophys. Geosystems<br />
Stoll, J.B., and M. Leven, 2002. Results of the oriented magnetic logging at Detroit Seamount. In:<br />
Tarduno, J.A., Duncan, R.A., Scholl, D.W., Proc. ODP, Initial Reports, Vol. 197. Available from<br />
http://www-odp.tamu.edu/publications/197_IR/197ir.htm.<br />
Stovall, S.H., 1997. Basic Inertial Navigation, Naval Air Warefare Center Weapons Division,<br />
http://www.fas.org/spp/military/program/nav/basicnav.pdf<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
125
Joint Interpretation of Magnetotelluric and Seismic Models for Exploration of<br />
the Gross Schoenebeck Geothermal Site<br />
G. Muñoz, K. Bauer, I. Moeck, O. Ritter<br />
<strong>GFZ</strong> <strong>Deutsche</strong>s GeoForschungsZentrum<br />
Introduction<br />
Due to the non-uniqueness of the inverse problem, the interpretation of geophysical models in<br />
terms of geological units is not always straightforward. It is common to use a combination of<br />
different geophysical methods to obtain the distribution of independent physical properties<br />
over the area of interest in order to discriminate between the different lithologies or geologic<br />
units. This kind of studies is usually limited to qualitative comparisons of the different<br />
models, which may – or may not –support a relation between the parameters in certain areas.<br />
Quantitative approaches are in general based on empirical relations between physical<br />
parameters, which often are not of universal applicability.<br />
The magnetotelluric (MT) and seismic methods resolve the physical parameters electric<br />
resistivity (ρ) and (seismic) velocity (Vp, Vs) respectively with similar spatial resolution and<br />
are often used in combination to derive earth models. By looking at both resistivity and<br />
velocity simultaneously, we can keep the strengths of both methods while avoiding their<br />
weaknesses. The problem with a joint interpretation is that there is no unique universal law<br />
linking electrical and acoustic properties. While electrical resistivity in deep sedimentary<br />
basins is mostly sensitive to the pore geometry and contents , seismic velocity is mostly<br />
imaging rock matrix properties. However, with a statistical analysis of the distributions of<br />
both resistivity and velocity, we can find certain areas of the models space where a particular<br />
relation between the physical parameters holds locally, thus allowing us to characterize this<br />
region as a particular lithology. In the present work, we use a statistical analysis, as described<br />
by Bedrosian et al. (2007) in order to correlate two independently obtained models of the<br />
Groß Schönebeck geothermal test site in the Northeast German Basin.<br />
Methodology description<br />
The methodology used in the present paper was described by Bedrosian et al. (2007) and is<br />
based on a probabilistic approach developed by Bosch (1999), in a sense that diverse<br />
geophysical parameters are represented as a probability density function (pdf) in the joint<br />
parameter space. The coincident velocity and resistivity models are first interpolated onto a<br />
common grid. Therefore, a joint parameter space is built, where each point in the modelled<br />
area is associated with a velocity – resistivity pair. By plotting one parameter against the other<br />
in a cross-plot and including the error estimates we can then construct a joint pdf in the<br />
parameter space. The areas of enhanced probability can be identified with classes represented<br />
by a certain range of values in both resistivity and velocity. By mapping back these classes<br />
onto the spatial domain they can be related to certain lithologies and/or geological units.<br />
In the present work, the resistivity model was interpolated onto the seismic mesh, given that it<br />
is uniformly spaced and finer than the magnetotelluric mesh. An inverse distance weighted<br />
interpolation scheme was used, which forms estimates from a weighted average of many<br />
samples found within a pre-defined area around the point, with decreasing weights with<br />
distance.<br />
Each element of this distribution can be interpreted as the outcome of a process defined by a<br />
probability density function (pdf). Assuming normal error distribution and independence of<br />
the data, the joint pdf is expressed as the sum of the individual pdfs (pdfi) for each data point,<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
126
according to the following expression, with the errors of the resistivity and velocity (δlog(ρi)<br />
and δVp,i) estimated from the sensitivity matrix and the hit count distribution respectively:<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
The different classes are identified as zones of enhanced probability in the joint pdf crossplot.<br />
Assuming that the geological units are characterized by uniform physical properties<br />
normally distributed, each class is defined by a mean point and a covariance matrix<br />
(representing the ~60 per cent confidence interval ellipse for the peak) in the joint parameter<br />
space.<br />
Geophysical models<br />
The Groß Schönebeck low enthalpy geothermal site, with the well doublet GrSk 3/90 and<br />
GrSk 4/05, is located in the Northeast German Basin (NEGB). MT data was collected along a<br />
40 km-long profile centred on the well doublet. The profile consists of 55 stations with a site<br />
spacing of 400 m in the central part (close to the borehole) of the profile, increasing to 800 m<br />
towards both profile ends. The period range of the observations was 0,001 to 1000 s. This<br />
profile is spatially coincident with the seismic tomography profile and most of the stations<br />
were located at the same places as the seismic shot points. At all sites, we recorded horizontal<br />
electric and magnetic field components and the vertical magnetic field.<br />
The resistivity model for the MT profile (Figure 1a) shows a shallow conductive layer<br />
extending from the surface down to depths of about 4 km, with an antiform-type shape below<br />
the central part of the profile. At a depth range of 4-5 km two conductive bodies are found,<br />
separated by a region of moderate conductivity. According to the seismic tomography, which<br />
shows high velocity values for depths greater than 4 km, a resistive basement was introduced<br />
a priori in the resistivity model (Muñoz et al., 2010).<br />
A 40 km long seismic profile was measured coincident with the MT experiment (Figure 1b).<br />
The objective was to derive a regional 2-D seismic model, which can be combined with the<br />
electrical conductivity model from the MT data analysis to study the potential reservoir layers<br />
and overlying sediments. The experimental setup was designed to provide data suitable for<br />
refraction tomography. 45 explosion shots were fired from 20 m deep boreholes with charge<br />
sizes of 30 kg. The shot spacing was 800 m on average. The recording instrumentation<br />
consisted of 4.5 Hz 3-component geophones. These were deployed as a 40 km long receiver<br />
spread with spacings of 200 m. Each shot was recorded by all receivers.<br />
The velocity model (Figure 1b) can be divided into three major sequences: The upper section<br />
(depth range 0-2 km) is characterized by low velocities (2-3.5 km/s) and a strong increase of<br />
velocity with depth. The section between 2 and 4 km depth shows velocities between 4 and<br />
4.5 km/s and is characterized by a strong topography on top which is related with salt<br />
mobility. The third, deepest section is bounded on top by a subhorizontal interface at 4.2 km<br />
depth and reveals velocities of more than 5 km/s (Bauer et al., 2010).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
127
a)<br />
b)<br />
Figure 1: Electrical resistivity model obtained from inversion of the magnetotelluric data<br />
using a priori information from the seismic velocity model for the deeper part of the model (><br />
5 km) (a) and seismic velocity model obtained from (Vp) travel time inversion (b).<br />
Joint analysis<br />
In the cross-plot of the probability density function (Figure 2a) we can identify five more or<br />
less clear peaks, or areas of enhanced probability with respect to the neighbouring region. The<br />
coloured ellipses in Figure 2a represent the ~60 per cent confidence intervals of the Gaussian<br />
peaks best fitting the pdf. The clusters were mapped back to the cross section providing a<br />
depth distribution of the classes along the profile (Fig. 2c). In order to interpret the nature of<br />
these litho-types, the model is superimposed on the stratigraphy derived from pre-existing<br />
reflection seismic data and borehole information (Moeck et al., 2008).<br />
Class 1, the shallowest, is characterised by low velocity (1.8 – 2.7 km/s) and moderate<br />
resistivity (5 – 70 Ωm) and comprises of unconsolidated sediments. Class 2, with higher<br />
velocities (2.7 – 3.9 km/s) and lower resistivities (0.5 – 3.5 Ωm) encompasses weak or soft<br />
rocks with high porosity, which are more conductive because they provide storage for a<br />
greater volumes of fluids. Class 3 coincides with successions of Middle Triassic to Lower<br />
Permian. Significantly these successions represent harder brittle rock of limestone and<br />
sandstone as indicated by increasingly higher velocities (4 – 5 km/s) and resistivities (2 – 15<br />
Ωm). This class includes also thick salt rock layers (Zechstein) which, however, yield no<br />
significant variation of resistivity or velocity within the class. Class 4, the deepest one, is<br />
characterized by the highest velocity and resistivity values (4.7 – 5.5 km/s and around 3000 -<br />
30000 Ωm). It represents the basin floor, comprising volcanic rock, quartzite and slate. This<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
128
class reflects the fact that the magnetotelluric model used seismic a priori information to<br />
introduce a resistive basin floor. Obviously this causes the very high correlation between<br />
velocity and resistivity of class 4 in Fig. 2b. Class 5 is characterized by high velocities (4.7 –<br />
5.5 km/s) but extremely low resistivities (0.1 – 0.7 Ωm).<br />
Figure 2: Cross-plot of the probability density function in gray-scale plan view (a) and threedimensional<br />
view (b). Spatial distribution of classes (c). Colours correspond with the ellipses<br />
in (a) defining the class boundaries.<br />
It is remarkable that Class 5 is restricted to salt lows where presumably anhydrites of Upper<br />
Permian age remain after salt movement. These anhydrites have a brittle behaviour and are<br />
expected to be highly fractured. In this case, an estimation of the resistivity by using Archie’s<br />
Law (Archie, 1942) for fracture-controlled porosity and assuming a formation fluid salinity of<br />
260 g/l (Giese et al., 2001), and a reasonable range of porosities and temperatures (15% and<br />
130ºC) the modelled resistivities of 0.1 – 0.7 Ωm can be explained. High velocities can be<br />
explained by the high density of anhydrite (2.9 g/cm 3 ). The classes are summarized in the<br />
Table 1, below.<br />
Table 1: Resistivity and P velocity of the classes from figure 9. Also included are lithology<br />
and stratigraphic units.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
129
Conclusions<br />
MT and seismic data were used to derive independent 2-D models of the electrical<br />
conductivity and the seismic P velocity around the geothermal research well GrSk 3/90. The<br />
resulting models were combined in a statistical analysis to determine correlating features in<br />
both models. The classification method used in this analysis revealed 5 distinct litho-types<br />
which show up as separate clusters in the underlying geophysical parameter space of Pa and<br />
VP.<br />
This study demonstrates the concert of MT, seismic and structural geologic models. Therefore<br />
the combination of different geophysical methods (MT and seismic) combined with structural<br />
geological information revealed information which was previously unknown and which could<br />
not be determined with the individual methods. Clearly, this approach is applicable in other<br />
areas too and could represent a promising new approach for geothermal exploration.<br />
Acknowledgements<br />
This work was funded within the 6 th Framework Program of the European Union (I-GET<br />
Project, Contract nº 518378). The instruments for the geophysical experiments were provided<br />
by the Geophysical Instrument Pool Potsdam (GIPP). We wish to thank Paul Bedrosian for<br />
making his statistical analysis code available to us.<br />
References<br />
Archie, G. [1942] The electrical resistivity log as an aid in determining some reservoir<br />
characteristics. Trans. Am. Inst. Min. Metall. Pet. Eng., 146, 54–62.<br />
Bauer, K., Moeck, I., Norden, B., Schulze, A., Weber, M. [2010] . Tomographic P velocity<br />
and gradient structure across the geothermal site Gross Schoenebeck (NE German Basin):<br />
Relationship to lithology, salt tectonics, and thermal regime. J. Geophys. Res., Submitted.<br />
Bedrosian, P.A., Maercklin, N., Weckmann, U., Bartov, Y., Ryberg, T., Ritter, O. [2007]<br />
Lithology-derived structure classification from the joint interpretation of magnetotelluric and<br />
seismic models. Geophysical Journal International, 170 170, 170<br />
737-748.<br />
Bosch, M. [1999] Lithologic tomography: from plural geophysical data to lithology<br />
estimation. Journal of Geophysical Research, 104, 749-766.<br />
Giese, L., Seibt A., Wiersberg T., Zimmer M., Erzinger J., Niedermann S. and Pekdeger A.<br />
[2001] Geochemistry of the formation fluids, in: 7. Report der Geothermie Projekte, In situ-<br />
Geothermielabor Groß Schönebeck 2000/2001 Bohrarbeiten, Bohrlochmessungen, Hydraulik,<br />
Formationsfluide, Tonminerale. GeoForschunsZentrum Potsdam.<br />
Moeck, I., Schandelmeier, H., Holl, H.G. [2008] The stress regime in Rotliegend reservoir<br />
reservoir of the Northeast German Basin. International Journal of Earth Sciences (Geol.<br />
Rundsch.), doi:10.1007/s00531-008-0316-1.<br />
Muñoz. G., Ritter, O., Moeck, I. [2010] A target-oriented magnetotelluric inversion scheme<br />
for characterizing the low enthalpy Groß Schönebeck geothermal reservoir. Geophysical<br />
Journal International, submitted.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
130
Magnetotelluric measurements to explore for deeper structures of the Tendaho<br />
geothermal field, Afar, NE Ethiopia<br />
Ulrich Kalberkamp<br />
Bundesanstalt für Geowissenschaften und Rohstoffe (BGR), Hannover, Germany<br />
Email: ulrich.kalberkamp@bgr.de<br />
Introduction<br />
In regions with high heat flow, like at volcanically active plate margins, high total<br />
thermodynamic energy is accumulated in the so called high enthalpy resources. The East<br />
African Rift System is one of the privileged areas to yield numerous sites for potential<br />
geothermal energy extraction including power generation. Along the Ethiopian part of the rift<br />
high temperature geothermal resources are associated with zones of quaternary tectonic and<br />
magmatic activity. Including the Afar depression at least 120 independent geothermal systems<br />
have been identified since the 1970s (UNEP 1973) about 24 of them are judged to have high<br />
enthalpy potential. To explore for these resources up to a pre feasibility stage a range of<br />
geoscientific methods is used in a defined sequence starting with regional reviews and remote<br />
sensing followed by geologic, hydrologic, geochemical and geophysical surveys.<br />
The applied geophysical methods usually comprise temperature measurements (gradient<br />
boreholes), seismology, magnetics and resistivity methods, including Magnetotellurics (MT).<br />
Geothermal surface manifestations like hot springs, fumaroles, geysers and the associated<br />
geological and geochemical settings are indicating the presence of a geothermal reservoir.<br />
Particularly resistivity methods<br />
may be used for delineating the<br />
lateral and depth extensions of<br />
such potential reservoirs. The<br />
MT method is frequently used<br />
for this purpose since it easily<br />
covers the necessary<br />
exploration depth down to<br />
approximately 10 km.<br />
In the following paper MT data<br />
and interpretation is presented,<br />
showing the already known<br />
shallow reservoir of the<br />
Tendaho geothermal field and<br />
its so far unknown deep<br />
structure, possibly feeding the<br />
shallow reservoir.<br />
Figure 1: Survey area at the SE end of the Hararo and Dabbahu<br />
magmatic segments. TGD = Tendaho-Goba’ad Discontinuity<br />
(after Ebinger et al. 2008).<br />
Survey area and tectonic<br />
setting<br />
The Tendaho geothermal field<br />
is located in the Afar<br />
depression (NE Ethiopia), at or<br />
very close to the assumed triple<br />
junction formed by the Red Sea,<br />
Gulf of Aden and East African<br />
rift arms. The extension of the Manda-Hararo axial rift zone in its south-easterly strike<br />
direction ends up at the Tendaho geothermal system (figure 1). The Red Sea and Aden rifts<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
131
are characterised by oceanic crust. The Afar triple junction is therefore a zone where thinned<br />
continental crust of the Main Ethiopian rift joins with crust of oceanic character (Barberi et al.<br />
1972).<br />
Within this area of active extensional tectonics several geothermal manifestations can be<br />
observed. Quaternary faulting and volcanism formed chains of fissural flows, basaltic cones<br />
and stratovolcanoes, usually accompanied by shallow seismicity and positive gravity<br />
anomalies. These active magmatic segments are comparable to slow spreading mid oceanic<br />
ridge segments and are thought to form new igneous crust by dyke like intrusions as<br />
demonstrated in the recent Dabbahu rift events (since 2005 up to now) just 80 km NW of the<br />
survey area. A previously mapped magmatic segment within the Manda-Hararo rift zone had<br />
been ruptured. Earthquake recordings indicated a 60 km long and about 8 m wide dyke<br />
intrusion (Ebinger et al. 2008, figures 8 & 9).<br />
Tendaho geothermal field<br />
The Tendaho geothermal field has been<br />
investigated in detail since the 1980s.<br />
Geological, geochemical as well as magnetic,<br />
seismological and resistivity data (DC<br />
soundings) were acquired by an extensive<br />
Ethiopian-Italian cooperation (Aquater, 1996).<br />
Based on this results, six exploratory wells had<br />
been drilled, three of them hitting a production<br />
horizon at approx. 300 m depth yielding steam<br />
temperatures above 250 °C (figure 2). Additional<br />
gravity and magnetic data had been acquired<br />
recently by Lemma and Hailu (2006) to delineate<br />
fractured zones which could serve as potential<br />
pathways for hydrothermal<br />
fluid flow from a deeper<br />
reservoir.<br />
To further investigate the<br />
proposed deep reservoir<br />
and/or heat source, deep<br />
reaching resistivity<br />
methods had to be applied.<br />
Due to the generally low<br />
resistivities the penetration<br />
depths of the DC<br />
soundings were limited to a<br />
few hundred meters. To<br />
extend this depth of<br />
exploration the MT method<br />
has been applied by BGR<br />
in collaboration with the<br />
Geological Survey of<br />
Ethiopia (GSE) during a<br />
field survey in 2007 (figure<br />
3). To reach the desired<br />
exploration depth of about 5<br />
km a frequency range from<br />
Figure 2: Production test of well TD5 (at well<br />
head: T >250 °C, p > 18 bar). The well is located<br />
inside the MT survey area.<br />
L1<br />
L2<br />
Kurub<br />
Figure 3: Survey area Tendaho geothermal field (red frame) projected onto<br />
satellite images (google earth). Red triangles = MT stations, blue squares =<br />
TEM stations. Line nos. L1, L2, L3 and L97 refer to MT lines in SE-NW<br />
bearing. TD5 = productive exploratory well. Yellow line = main road to<br />
Djibouti.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
132<br />
TD5<br />
<br />
L97<br />
L3
10 kHz to 0.01 Hz (100 s) has been acquired.<br />
MT resistivity signature<br />
Due to alluvial and lacustrine infill of the Tendaho graben apparent resistivity curves show<br />
generally values below 10 Ohm*m, even at high frequencies. A typical sounding curve is<br />
shown in figure 4.<br />
The minimum is<br />
reached at approximately<br />
4 Hz with<br />
apparent resistivities<br />
just below 1 Ohm*m.<br />
With decreasing<br />
frequency the<br />
apparent resistivity<br />
rises again. Source<br />
signal was generally<br />
quite low, especially<br />
within the dead band<br />
region ranging from<br />
approximately 1 to<br />
0.1 Hz (1 to 10 sec.).<br />
Figure 4: Typical sounding curve. Apparent resistivities (top) and phases Nonetheless 16 hrs<br />
(bottom). Due to low signal strength in the dead band from about 1 to 0.1 Hz of recording time<br />
unstable estimation especially of phase values, resulting in increased error bars.<br />
proved to be<br />
sufficient to yield<br />
acceptable estimates<br />
of the impedance<br />
tensor, using mainly<br />
single site processing.<br />
This apparent resistivity pattern, which is also reflected in the resistivity models (see below),<br />
is indicative for high temperature geothermal reservoirs (figure 5, see also e.g. Kalberkamp<br />
2007) where the resistivity low is interpreted as the clay cap while the increasing resistivities<br />
below the clay cap point to the core of the reservoir and represent a possible drilling target.<br />
In the 2d inverted resistivity sections, exemplarily<br />
shown for profile L1 in figure 6, this resistivity<br />
pattern can be seen quite clearly. Although<br />
resistivities are generally low, they show signatures<br />
typical for hydrothermal alteration halos (see e.g.<br />
Johnston et al. 1992, Kalberkamp 2007) and (partly)<br />
molten magma intrusions at depth below 4 km.<br />
Taking the Dabbahu rift events into account it seems<br />
to be likely that the heat source for the geothermal<br />
reservoir is formed by magma intrusions along dyke<br />
like fracture zones as it is suggested by our<br />
interpretation of the MT data. Areas with high<br />
resistivities (>300 Ohm*m) may be associated with<br />
basalts from the Afar Stratoid Series, constituting the<br />
borders of the Tendaho graben structure.<br />
Figure 5: Schema of a generalised<br />
geothermal system. The smectite cap<br />
formed exhibits resistivities in the<br />
range of 2 Ohm*m, the mixed layer<br />
around 10 Ohm*m (modified after<br />
Johnston et al. 1992).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
133
Figure 6: 2 d inverted resistivity section (L1). Lateral extension is 27 km; vertical depth covers 8 km.<br />
Interpretation and recommendations<br />
As has been shown for profile L1 exemplarily the magnetotelluric soundings show in their 2dimensional<br />
inverted resistivity sections regions with resistivities as low as 2 Ohm*m,<br />
especially in near surface layers along profiles L1, 2, and 3 as well as at greater depth (below<br />
5 km) in profiles P1, 3, and 97. When using the 2d MT sections to compile resistivity maps<br />
for constant elevations each, thus based on the inverted resistivities, we get a general view of<br />
the lateral resistivity structure as presented in figure 7 for selected elevations.<br />
Figure 7: Resistivity maps of survey area as covered by red frame in figure 3, resistivity range 2 (red) – 2048<br />
(blue) Ohm*m. Left: at 200 masl (150 m below surface). Shallow low resistive layer (red) due to sedimentary<br />
infill. The sediments are up to 1 km thick. Centre: at 1000 mbsl (1350 m below surface). Slight increase of<br />
resistivity (> 10 Ohm*m) possibly due to mixed layer clays and advancement towards deeper reservoir (blue<br />
circle). Right: at 9000 mbsl (9350 m below surface). Resistivity drop below 2 Ohm*m along NW-SE<br />
trending feature (partly molten magma dyke?). This may form the deep heat source feeding the shallow<br />
geothermal reservoir.<br />
At greater depth of around 7 km a<br />
resistivity anomaly below 2 Ohm*m<br />
appears elongating in NW-SE direction.<br />
This direction coincides with the strike<br />
direction of the Tendaho graben and rift<br />
structures further to the NW, up to the<br />
Dabbahu rift and Boina vent, where lava<br />
ascended along a fault system, almost<br />
reaching the surface (figure 8). Therefore<br />
it seems likely, that the low resistive<br />
structures at greater depth are caused by<br />
lava filled fracture zones. (figure 7 right).<br />
Ebinger et al. (2008) have presented a<br />
working model of the Dabbahu magmatic<br />
Figure 8: Volcanic vent, created during 2006 eruption<br />
event (Photo by J. Rowland).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
134
Figure 9: Working model of the Dabbahu magmatic system as<br />
presented by Ebinger et al. (2008). Pink ellipses = shallow magma<br />
chambers, connected to assumed lower crust/upper mantle feeding<br />
zones.<br />
system (figure 9) based<br />
mainly on seismological,<br />
radar interferometric and<br />
structural data. They suggest<br />
that lower crustal/upper<br />
mantle source zones are<br />
feeding the basaltic dyke<br />
and shallow magma<br />
chambers at a few kilometres<br />
depth.<br />
Magnetotelluric measurements<br />
by the Afar Rift<br />
Consortium / University of<br />
Edinburgh (Scotland, U.K.)<br />
are ongoing in that area and<br />
results thereof may help to<br />
refine the interpretation of<br />
the Tendaho geothermal<br />
field data.<br />
With the current data therefore the following preliminary conclusions may be drawn:<br />
The deep heat source feeding the currently drilled reservoir at 300 m depth is most likely<br />
(partly) molten magma along NW-SE trending dykes or faults. In the Tendaho geothermal<br />
field these structures may be as shallow as 4 km depth.<br />
A deep reservoir may be expected below 1300 m.<br />
An up flow zone could be present at the S end of the survey area indicated by low<br />
resistivities throughout the acquired depth range.<br />
To identify the proposed up flow zone at the S end of the current survey area it is<br />
recommended to apply additional TEM soundings which could enhance the resolution at<br />
shallow depth.<br />
Additional MT sounding profiles are recommended further towards the Hararo and Dabbahu<br />
magmatic segments NW of the survey area. Since rift events including ascending magma are<br />
evident further to the NW it may be assumed that high temperature reservoirs could be<br />
reached at comparatively shallow depth there. First MT results from the research work within<br />
the Afar Rift Consortium do support this assumption (Desissa et al., 2009).<br />
Also additional MT soundings beyond the Awash river up to the geothermal manifestations of<br />
Alalobad in the south would be helpful to establish either the extension of one large reservoir<br />
or the existence of a separate second geothermal system.<br />
Acknowledgements<br />
The MT survey in Ethiopia has been carried out as part of a cooperation project between the<br />
Geological Survey of Ethiopia (GSE) and BGR as part of the GEOTHERM Programme. We<br />
gratefully acknowledge the cooperation with Mohammednur Desissa, Yohannes Lemma and<br />
the Geothermal Working Group within the GSE. We also appreciate helpful communications<br />
and cooperation with Kathy Whaler and the Afar Rift Consortium.<br />
GEOTHERM (www.bgr.de/geotherm/) is a technical cooperation programme to promote the<br />
use of geothermal energy in partner countries, implemented by the BGR on behalf of the<br />
German Federal Ministry for Economic Cooperation and Development (BMZ) under contract<br />
no. 2002.2061.6.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
135
References<br />
Aquater; 1996. Tendaho geothermal project, final report. Ministry of Mines of Ethiopia and<br />
Ministry of Foreign Affairs of Italy, San Lorenzo in Campo, Italy (unpublished).<br />
Barberi, F., Tazieff, H. & Varet, J., 1972. Volcanism in the Afar depression: its tectonic and<br />
magmatic significance. Tectonophysics, 15, 19-29.<br />
Desissa; M., Whaler, K., Hautot, S., Dawes, G., Fisseha, S., Johnson, N., 2009. A<br />
Magnetotelluric study of continental lithosphere in the final stages of break-up: Afar,<br />
Ethiopia. Abstract I.06-1158, 11th IAGA Scientific Assembly, Sopron, Hungary.<br />
Ebinger, C.J., Keir, D., Ayele, A., Calais, E., Wright, T.J., Belachew, M., Hammond, J.O.S.,<br />
Campbell, E. & Buck, W.R., 2008. Capturing magma intrusion and faulting processes<br />
during continental rupture: seismicity of the Dabbahu (Afar) rift, Geophys. J. Int.<br />
Johnston, J.M., Pellerin, L. & Hohmann, G.W. (1992): Evaluation of Electromagnetic<br />
Methods for Geothermal Reservoir Detection, Geothermal Resources Council<br />
Transactions, 16, 241-245.<br />
Kalberkamp, U., 2007. Exploration of geothermal high enthalpy resources using<br />
Magnetotellurics – an Example from Chile, in: Ritter, O., Brasse, H. (eds.): Protokoll<br />
22. Kolloquium Elektromagnetische Tiefenforschung, Hotel Maxiky. Dín, Czech<br />
Republic, 1.-5. Oktober 2007, DGG, 194-198.<br />
Lemma, Y., Hailu, A., 2006. Gravity and magnetics survey at the Tendaho geothermal field.<br />
GSE, Addis Ababa, Ethiopia, 23pp (unpublished),.<br />
UNEP (ed.), 1973. Ethiopia: Investigation of geothermal resources for power development –<br />
Geology, geochemistry and hydrology of hot springs of the East African Rift System<br />
within Ethiopia, Technical Report DP/SF/UN/116, United Nations, New York.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
136
Electromagnetic Monitoring of CO2 Storage in Deep Saline<br />
Aquifers - Numerical Simulations and Laboratory<br />
Experiments<br />
J. H. Börner, V. Herdegen, R.-U. Börner, K. Spitzer<br />
Institut für Geophysik, TU Bergakademie Freiberg, Gustav-Zeuner-Str. 12, 09599 Freiberg<br />
1 Introduction<br />
The knowledge of petrophysical parameters and their contrasts is crucial to reliably monitoring CO2<br />
storage processes. The electrical conductivity appears to be a sensitive indicator for a resistive gaseous<br />
or supercritical CO2 phase replacing a conductive pore fluid in a porous medium like a saline sandstone<br />
aquifer. However, detailed knowledge on the influence of supercritical CO2 on the electrical resistivity<br />
of a formation is not sufficiently available yet.<br />
Therefore, we have carried out laboratory experiments to predict the contrast in electrical resistivity due<br />
to the presence of CO2 in an initially water-saturated sand sample resembling the petrophysical situation<br />
typical for a reservoir. Furthermore, the expected parameter contrasts were estimated according to<br />
empirical equations and numerical simulations (Fig. 1).<br />
On a middle- to long-term perspective, we aim at developing an electromagnetic monitoring technique<br />
using a borehole transient electromagnetic sensor which offers a unique opportunity to generate<br />
enhanced sensitivity at depth with respect to detecting migrating CO2 in a reservoir. This work is<br />
therefore integrated into national CCS research programs with an interdisciplinary variety of partners.<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
Figure 1: Workflow of the numerical simulation steps necessary for the transformation of water saturation into<br />
electrical resistivity.<br />
2 Theory<br />
Two-phase flow is governed by two equations simultaneously enforcing continuity of the water (w) and<br />
the CO2 (co2) phase flow (Busch et al. 1993). Both equations are linked by retention curves Se(pc) and<br />
relative permeabilities kr(S), such that<br />
∇·d w [− κkw r<br />
ηw (∇pw + d w g∇D)] = −Φ ∂(dwS w e )<br />
+ d<br />
∂t<br />
w w0<br />
∇·d co2 [− κkco2 r<br />
ηco2 (∇pco2 + d co2 g∇D)] = −Φ ∂(dco2Sco2 e )<br />
+ d<br />
∂t<br />
co2 w0<br />
with d as density, κ as intrinsic permeability, η as viscosity, Φ as porosity and g as gravitational acceleration.<br />
The link between both phases has been established by experimental data or parameterization,<br />
23 . Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
137<br />
(1)
e.g., by van Genuchten (1980) or Mualem (1976) (with effective saturation Se, capillary pressure height<br />
Hc and parameters α, n,m,L):<br />
1<br />
S w e =<br />
(1 + |αHc| n ) m<br />
k w r =(S w e ) L ·<br />
<br />
1 −<br />
<br />
1 − (S w e ) 1/m m 2<br />
S co2<br />
e =1− S w e (2)<br />
k co2<br />
r<br />
=(1− S w e ) L ·<br />
<br />
1 − (S w e ) 1/m 2m<br />
The resulting distribution of water saturation S(x, y, z, t) can be transformed into electrical formation<br />
resistivity ρ using empirical resistivity models, e.g., Archie’s law (Archie 1942):<br />
ρ = a1 · Φ −a2 −a3 1<br />
· S · (4)<br />
σw<br />
with σw denoting pore water conductivity. Archie’s empirical parameters a1, a2 and a3 strongly depend<br />
on rock formation characteristics and can be combined with porosity and water saturation yielding the<br />
formation factor F , such that<br />
ρ = F · ρw. (5)<br />
For sands and sandstones with considerable clay content Waxman and Smits (1968) expanded Archie’s<br />
law to account for both electrolytic conductance and interfacial conductance:<br />
ρ =<br />
F ∗<br />
S n · (σw + B·Qv<br />
S )−1 (6)<br />
with a unique formation factor F ∗ , the counterion mobility B describing the weak dependance of the<br />
interface conductivity on pore water salinity and the shalyness parameter Qv being the cation exchange<br />
capacity normalized to the pore volume.<br />
3 Numerical simulation studies<br />
The process of CO2 sequestration as well as the indication of a successful storage by observation of<br />
changes in electrical resistivity has been simulated following the steps indicated by the workflow shown<br />
in Fig. 1.<br />
Figure 2: CO2 saturation after 6 months of constant injection of 1000 m 3 per day into a 25 m thick reservoir.<br />
The injection well is located at the geometric center. These results were obtained using the FD<br />
simulation code Mod2PhaseThermo.<br />
23 . Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
138<br />
(3)
23 . Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
139
23 . Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
140
Final experiments have demonstrated that an increase in resistivity may also be achieved under<br />
supercritical conditions. However, the sample was not homogeneously infiltrated by the CO2 and<br />
preferential flow paths were built up during the experiment. The pore water has been pressed out only<br />
partially, leaving behind tube-shaped flow channels and a very heterogeneous water distribution with<br />
average residual water contents of more than 70 %.<br />
pressure height in m<br />
500<br />
400<br />
300<br />
200<br />
100<br />
van Genuchten<br />
measured data<br />
0<br />
0 0.2 0.4 0.6 0.8 1<br />
water saturation<br />
Figure 6: Relation between CO2 pressure and water saturation derived from data obtained by the initial<br />
pressure build-up for an experiment with a projected maximum pressure of 50 bar. The reference<br />
data have been calculated according to van Genuchten (1980) with α = 0.02 and n =3.9 (cf. eqs.<br />
(2)-(3)).<br />
The complications are due to the density of supercritical CO2 which is large compared to the density<br />
of CO2 in its gaseous state. Consequently, the current experimental assembly causes the high density<br />
to result in low flow velocities within the measuring cell. In addition to the flow channels, this effect<br />
prevents an effective replacement of the pore water by supercritical CO2.<br />
The experiments have shown that theoretical pressure-saturation relations can generally be verified in<br />
practice using our set-up (Fig. 6). All experimental data sufficiently agree with Archie’s law. However,<br />
we have to carry out further investigations on the effects of CO2 dissolving in the pore water. There<br />
are indications that this has an important impact on the pore water resistivity depending on pressure,<br />
temperature, and brine salinity.<br />
6 Conclusions<br />
Numerical simulation studies and laboratory experiments show that the electrical resistivity of a<br />
fluid-saturated porous medium is highly sensitive to the presence of CO2. Geo-electromagnetic methods<br />
are therefore considered as a promising approach for monitoring CO2 storage.<br />
Still, the laboratory set-up has to be improved further to provide reliable results at high pressures.<br />
Subject to these prerequisites, well-founded simulations of electromagnetic monitoring scenarios can be<br />
carried out.<br />
Further laboratory experiments will also aim at quantifying the influence of dissolved CO2 on pore<br />
water and formation resistivity. A feasibility study could show whether electromagnetic methods are<br />
able to monitor, e.g., an expanding plume of CO2-rich formation water during CO2 injection. Finally,<br />
the applicability of the laboratory assembly for measuring reliable retention curves of unconsolidated<br />
sedimentary rocks will be further tested in the near future.<br />
23 . Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
141
References<br />
Archie, G. E. (1942). The electrical resistivity log as an aid in determining some reservoir characteristics. Trans. Americ.<br />
Inst. Mineral. Met. (146): 54–62.<br />
Busch, K.-F., L. Luckner, and K. Tiemer (1993). Geohydraulik. Edited by G. Matthess. 3rd edition. Volume 3. Lehrbuch<br />
der Hydrogeologie. Gebrüder Borntraeger. isbn: 3-443-01004-0.<br />
Häfner, F. and S. Boy (2009). Mod2PhaseThermo Nutzermanual. IBeWa.<br />
Mualem, Y. (1976). A new Model for predicting the hydraulic condyctivity of unsaturated porous media. Water Resour.<br />
Res. 12: 283–291.<br />
Schön, J. (1996). Physical Properties of Rocks - Fundamentals and Principles of Petrophysics. 1st edition. Volume 18.<br />
Handbook of Geophysical Exploration. Section I, Seismic Exploration. Pergamon.<br />
van Genuchten, M. (1980). A Closed-form Equation for Predicting the Hydraulic Conductivity of Unsaturated Soils. Soil<br />
Science Society of America Journal 44(5): 892–898.<br />
Waxman, M. H. and L. J. M. Smits (1968). Electrical conductivities in oil-bearing shaly sands. Society of Petroleum<br />
Engineers Journal 8: 107–122.<br />
23 . Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
142
Erkundung eines Aquifers unter dem Mittelmeer<br />
vor der israelischen Küste mit LOTEM.<br />
K. Lippert 1 , B. Tezkan, R. Bergers, M. Gurk, M. v. Papen, P. Yogeshwar<br />
Institut für Geophysik und Meteorologie, Universität zu Köln<br />
Abstract<br />
Im Rahmen dieses BMBF-geförderten Projektes 2 kommt die Long-Offset Transient<br />
Elektromagnetik (LOTEM) Methode zum ersten Mal in mariner Umgebung zur Erkundung<br />
von Grundwasseraquiferen zum Einsatz. Hauptziel des Projektes ist die Detektion<br />
der Süßwassergrenze unter dem Mittelmeer an der Küste Israels. Zu diesem Zweck wurden<br />
Hardware-Modifikationen durchgeführt und bereits in einer ersten Testmessung in Israel<br />
erprobt. Des weiteren werden die Interpretationen der ersten Testmessung sowie die Modellierungsergebnisse<br />
des Küstenabschnitts vorgestellt, welche zur Planung der Hauptmessung<br />
wichtig sind.<br />
Einleitung<br />
Die Bedeutung von Offshore-Süßwasserleitern<br />
für das Grundwassermanagment nahm in den<br />
letzten Jahren stark zu. Das Vorhandensein<br />
dieser Aquifere, die sich bis zu mehreren<br />
Kilometern unter dem Meeresboden erstrecken<br />
können, wurde in der Fachliteratur beschrieben:<br />
z.B. an der Küste von Guyana [Arad, 1983]<br />
oder vor den Niederlanden [Groen et al.,<br />
2005]. Der israelische Küstenaquifer ist eine<br />
der Hauptgrundwasserresourcen des Landes<br />
(vgl. Fig. 1). Aufgrund der intensiven<br />
Nutzung verschlechtert sich allerdings die<br />
Wasserqualität zumehmend. Die Gründe hierfür<br />
liegen sowohl in der anthropogenen Verschmutzung<br />
vom Land aus, als auch im Eindringen<br />
von Salzwasser.<br />
1 E-mail: lippert@geo.uni-koeln.de<br />
2 BMBF-Förderkennzeichen: 02WT0987<br />
Figure 1: Die großen Aquifere der Region.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
143
Geologie und Fragestellung<br />
Der Küstenaquifer besteht hauptsächlich aus<br />
kalkhaltigem Sandstein und Sanden und ist<br />
selbst noch in vier Sub-Aquifere unterteilt<br />
(vgl. Fig.2). Die oberen zwei grundwasserführenden<br />
Schichten werden vom Land<br />
aus anthropogen und vom Meer aus durch<br />
Salzwassereindringen verunreinigt. Bisher<br />
wurde davon ausgegangen, daß die unteren Sub-<br />
Aquifere stellenweise vom Meerwasser getrennt<br />
sind und dort keine Salzwasserintrusion stattfindet.<br />
Neuere Untersuchungen ([Kafri, U.,<br />
Goldman, M., 2006],[Yechieli et. al, 2009])<br />
hingegen deuten jedoch auf das Fortsetzen des<br />
Aquifers unter dem Meer hin. So zeigen<br />
z.B. TDEM-Messungen auf Land, nahe der<br />
Küste, einen schlechteren elektrischen Leiter<br />
10 (Ωm) eingebetet zwischen zwei guten elektrischen<br />
Leitern (∼ 2 Ωm). Diese Schicht,<br />
in der Tiefe der unteren Sub-Aquifere, wird<br />
als Grundwasserführenden Schicht interpretiert<br />
(vgl. Fig.3). Die Fragestellung an die LOTEM-<br />
Methode ist nun 1. die Prüfung der Existenz<br />
des unteren Sub-Aquifers in etwa 100m Tiefe<br />
und 2. dessen Ausbreitung unter dem Mittelmeer.<br />
Figure 2: Möglicher Verlauf des unteren Subaquifers an der Küste [Yechieli et. al, 2009]: a)<br />
Der untere Subaquifer (CD) hat keine Verbindung zum Meer, b): Das Salzwasser dringt in den<br />
unteren Aquifer ein.<br />
Figure 3: Onshore SHOTEM-Ergebnisse [Kafri, U., Goldman, M., 2006].<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
144
Hardwaremodifikationen<br />
Das Kölner LOTEM-Equipement ist für<br />
Landmessungen konzipiert. Deswegen waren<br />
einige Hardware-Modifikationen für die Offshoremessung<br />
nötig. Als Sender konnte aus<br />
logistischen Gründen nur der kleinere Kölner<br />
Sender 3 verwendet werden. Da dieser bei<br />
verwendeten Stromstärken von ∼ 9A zuviel<br />
Wärme entwickelt, war es nötig eine zusätzliche<br />
Wasserkühlung einzubauen. Für die Einspeisung<br />
des Sendesignals wurden neue Elektroden<br />
entwickelt (vgl. Fig.4), deren Form,<br />
durch Maximierung des angekoppelten Wasservolumes,<br />
elektrochemische Vorgänge minimiert.<br />
Zusätzlich kann das Sendesignal Stromgeregelt<br />
werden, um einen konstanten Stromverlauf zu<br />
erreichen. In Fig.5 sind Sendesignale gezeigt.<br />
Links wird mit konstanter Spannung gesendet.<br />
Bei sich verändernder Ankopplung ergibt sich<br />
kein konstanter Stromverlauf. Deswegen wurde<br />
der Sendestrom nachträglich auf einen konstan-<br />
Figure 5: Aufgezeichnete Sendesignale bei<br />
einem Test in Wilhelmshaven.<br />
ten Wert heruntergeregelt (Fig.5 rechts).<br />
Die verwendeten Magnetfeldsensoren 4 wurden<br />
in Druck- und Salzwasserbeständige Behälter<br />
verpackt (Fig.6).<br />
Figure 4: Neuentwickelte Elektroden für die<br />
Einspeisung des Sendesignals.<br />
Figure 6: Seewasserfestes Gehäuse für die<br />
Magnetfeld-Sensoren.<br />
3 NT-20 von Zonge Engineering & Research Organization, Inc., Tucson AZ, USA<br />
4 TEM/3-Spulen von Zonge Engineering & Research Organization, Inc., Tucson AZ, USA<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
145
Ergebnisse Messung 2008<br />
Anfang November 2008 wurden die ersten Messung<br />
vor Ort durchgeführt. Hierbei wurden verschiedene<br />
Messkonfigurationen gestestet:<br />
• BroadSide: Sender an der Küstenlinie,<br />
Empfänger an Land. Offset 200m. vgl.<br />
Fig.7: Tx2 und Rx3.<br />
• BroadSide: Sender an der Küstenlinie,<br />
Empfänger im Wasser. Offset 300m. vgl.<br />
Fig.7: Tx2 und Rx2.<br />
• BroadSide: Sender im Wasser,<br />
Empfänger an Land. Offset 500m. vgl.<br />
Fig.7: Tx1 und Rx1.<br />
Figure 7: Setups der Messung 2008.<br />
Das Setup “Sender an der Küstenlinie,<br />
Empfänger an Land” ist das einzige, welches<br />
annehmbar 1D interpretiert werden konnte.<br />
Als Beispiel ist hier die dHz/dt - Komponente<br />
gezeigt. Das Endmodell einer Marquardt-<br />
Inversion ist in Fig.9, die zugehörige Datenanpassung<br />
in Fig.8 zu sehen. Das Startmodell<br />
wurde anhand der Vorinformationen und<br />
dem Ergebnis einer Occam-Inversion gewählt 5 .<br />
Figure 8: Datenanpassung des “besten” Modells<br />
(rot) in Fig. 9.<br />
Figure 9: 1D-Inversionsergebnis mit<br />
Äquivalenzmodellen.<br />
Das Ergebnis ist gut mit den SHOTEM-<br />
Ergebnissen (Fig. 3) vergleichbar. Der gesuchte<br />
Aquifer ist deutlich in einer Tiefe ab ca. 90m<br />
zu erkennen. Sowohl die Schichtdicke mit ca.<br />
100m als auch viele<br />
Äquivalenzmodelle mit<br />
einem Widerstand von ∼ 10 Ωm sind vergleichbar<br />
und glaubwürdig.<br />
Die anderen beiden Messkonfigurationen<br />
5Vgl. M.v.Papen, B.Tezkan: “On the analysis of LOTEM time series from Israel and preliminary 1D<br />
inversion of data.”, Poster EMTF 2009.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
146
können nicht als eindimensionales Problem<br />
angesehen und somit auch nicht 1D ausgewertet<br />
werden. Die gemessenen Transienten<br />
wurden mit synthetischen Daten verglichen.<br />
Das 2D Leitfähigkeitsmodell ist in Fig. 11<br />
dargestellt. Mit diesem synthetischen Modell<br />
war eine qualitative Anpassung möglich. Exemplarisch<br />
ist in Fig.10 die Datenanpassung für<br />
die Ex-Komponente gezeigt.<br />
Figure 10: Setup Tx2 - Rx3 (vgl.Fig.7): Anpassung<br />
der Messdaten der Ex-Komponente mit<br />
modellierten Daten.<br />
Vorab-Modellierungen<br />
Für die Messung im November 2009 sind die<br />
Empfängerlokationen von Bedeutung. Eine<br />
Frage hierbei ist u.a. ob das Target auch mit<br />
einem Sender auf dem Meer und Empfängern<br />
auf dem Land auflösbar ist. Diese Konfiguration<br />
ist logistisch relativ einfach durchführbar.<br />
Es wurde also ein 2D-Küstenmodell erstellt,<br />
mit welchem synthetische Daten mittels<br />
SLDMEM3T [Druskin, Knizhnermann, 1988]<br />
erzeugt wurde. Als Datenbeispiel sind in Fig.12<br />
die Ex-Komponenten dargestellt. Hierfür werden<br />
im Modell (Fig.11) Daten mit Targetlayer<br />
und einmal ohne Targetlayer unter dem Meeresboden<br />
verglichen. “Ohne Targetlayer” bedeutet<br />
in diesem Fall, daß der Targetlayer 300m im<br />
Landesinneren endet (nicht dargestellt). Die<br />
darüberliegende Schicht (1.2 Ωm, gelb) wurde<br />
nach unten erweitert.<br />
Modelliert wurden verschiedene Senderpositionen<br />
in unterschiedlichen Entfernungen zur<br />
Küstenlinie. Im Beispiel beträgt die Entfernung<br />
1500m. Betrachtet werden nun 2<br />
Empängerpositionen: einmal 275m von der<br />
Küstenlinie entfernt auf dem Meeresboden und<br />
einmal 95m im Landesinneren. Modelliert<br />
wurde die BroadSide Konfiguration (Senderrichtung<br />
und Profilrichtung senkrecht zueinander)<br />
und die InLine Konfiguration (Senderrichtung<br />
und Profilrichtung auf einer Linie). Die<br />
System-Antwort wurde für diese Modellierungen<br />
nicht berücksichtigt.<br />
Figure 11: Verwendetes Küstenmodell. Das<br />
gesuchte Target ist die 15 Ωm-Schicht (rot).<br />
Blau = Wasser.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
147
Sowohl bei der InLine als auch bei der Broad-<br />
Side Konfiguration ist in dieser Komponente<br />
der Targetlayer klar auflösbar: die Transienten<br />
unterscheiden sich deutlich in den Amplituden<br />
von den Transienten ohne Targetlayer. Bei<br />
der BroadSide Konfiguration verschiebt sich der<br />
auftretende Vorzeichenwechsel durch das Target<br />
zu späten Zeiten. Ein Vorzeichenwechsel ist<br />
bei der Auswertung immer schwer interpretierbar.<br />
Aufgrund der Unterscheidbarkeit der Transienten<br />
sind Empfänger auf dem Land für diese<br />
Fragestellung sinnvoll.<br />
Figure 12: Modellierte Transienten der Ex-<br />
Komponente an den Positionen, die in Fig.11<br />
als Kreise markiert sind.<br />
Zusammenfassung und Ausblick<br />
Aufgrund der Ergebnisse und der Erfahrungen<br />
der ersten Messung (2008) und der Modellierungen<br />
sind die Setups für die erste Hauptmessung<br />
im November 2009 festgelegt worden. Die Messung<br />
ist inzwischen durchgeführt worden. Es<br />
wurden die Konfigurationen InLine und Broad-<br />
Side gemessen. Der Sender befand sich dabei<br />
immer auf dem Meer in verschiedenen Abständen<br />
zur Küste. Pro Senderposition gab es mehrere<br />
Empfängerpositionen, sowohl auf dem Wasser<br />
als auch auf Land.<br />
Literatur<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
148<br />
Arad, A.: “ A summary of artesian coastal<br />
basin of Guyana.”, J.Hydrol. 63:299-313, 1983.<br />
Druskin, V.L., Knizhnermann, L.A.: ”A<br />
spektral semi-discrete method for the numerical<br />
solution of 3d nonstationary problems in electrical<br />
prospecting.”, Phys. Solid Earth, 24:641-<br />
648, 1988.<br />
Groen J., Groen M., Post V., Kooi H.,<br />
Mendizabal I. and Groot S.M.: ” Seaward<br />
continuation of groundwater flow systems along<br />
the coast of the Netherlands.”, GSA Salt Lake<br />
Ann.Meet. Abstract Paper No.211-9., 2005.<br />
Kafri U. and Goldman M.: “ Are the lower<br />
subaquifers of the Mediterranean coastal aquifer<br />
blocked to seawater intrusion? Results of a TDEM<br />
(time domain electromagnetic) study.”, Isr.J.Earth<br />
Sci., 2006.<br />
Yechieli et. al: “ The inter-relationship between<br />
coastal sub-aquifers and the Mediterranean<br />
Sea, deduced from radioactive isotopes analysis.”<br />
Hydrogeology Journal Vol 17 Num 2 /<br />
2009.
On the analysis of LOTEM time series from Israel and the preliminary 1D inversion of<br />
data<br />
M. von Papen, B. Tezkan<br />
Institute of Geophysics and Meteorology, University of Cologne, Germany<br />
Abbildung 1: Overview of measurement setups (A,B,C from left to<br />
right)<br />
1. Data Processing of TEM time series from the<br />
first Israel survey<br />
Abbildung 2: Power spectrum of Ex at station A/B before (blue<br />
asterisks) and after application of different filters<br />
The measurement took place at the coast near Ashdod,<br />
Israel (fig.1), where the task is the detection of fresh<br />
groundwater bodies within the mediterranean submarine<br />
aquifers [3]. For that matter a 380 m long dipole emitting<br />
in a 50% duty cycle was used, located on the shoreline for<br />
stations A and C and at -300m offshore on the seabottom<br />
for station B. Receiver stations recorded four components<br />
of the EM field (Ex,Ey, ˙ Hy, ˙ Hz) with Summit receivers. Receivers<br />
at station C were set up 5mbeneath the water on<br />
the sea bottom.<br />
• The measured transients consist of 4096 data points<br />
(DP) with a sample rate of 1/16 ms and 256 DP as<br />
onset. Switch time of Tx is t0 = 1500ms.<br />
Abbildung 3: Parts of transient [5]<br />
Noise measurements showed a dominant 50 Hz noise<br />
together with multiples (fig.2). The average noise<br />
power for each component is: PN,Ex = 0.2V 2 ,PN,Ey =<br />
0.02V 2 ,P N, ˙ Hy = 4.8V 2 ,P N, ˙ Hz = 0.2V 2 (measured on<br />
land). Due to relative short offsets of 200-500 m between<br />
receiver and transmitter the signal is - except for Ey component<br />
- clearly visible in the raw transients.<br />
Abbildung 4: Stacked transients of only on-switches and only offswitches<br />
2. Application of a segmented lockin filter<br />
First step of the processing was to filter out the 50 Hz<br />
noise. This was done using a segmented lockin filter (LOF)<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
149
Abbildung 5: Quantiles of normal distribution for stations A,B,C<br />
from top to bottom<br />
[1]. It fits harmonics in segments of 60 ms to the data and<br />
subtracts them. Amplitude and phase are computed with<br />
Fourier series expansion. Regarding the high dynamic of<br />
the transient at the time of the switch (e.g. switch on in<br />
part B of fig.3) the fitting is being done in parts A and<br />
C only [5]. In addition a 3rd degree polynomial is fitted<br />
in part C to simulate the descent of the transient. Other<br />
filters like Delay Lower Nyquist frequency (DLN), Delay<br />
filter or (not segmented) Lockin-Filter [1, 5] remove the<br />
noise less efficient for this dataset (fig.2).<br />
• Application of the filter reduced the noise power, computed<br />
at the end of the time series, to PN,Ex =<br />
0.2nV 2 ,P N, ˙ Hy =0.2V 2 and P N, ˙ Hz =1.4nV 2 .<br />
• The DC bias is removed by levelling the data in the<br />
onset after filtering.<br />
Due to the short onset and the resulting error in determining<br />
the DC bias the inverted conductivities will exhibit<br />
a slight shift [4]. Higher accuracy can be achieved by averaging<br />
over a complete transmitter period.<br />
3. Analysis of data distribution<br />
The data measured with 50% duty cycle consists of two<br />
different sets of transients. One set is created when switching<br />
the current off and another when switching on. Because<br />
the given transmitter system shows a much shorter<br />
ramp time for off-switches, these transients are generally<br />
less broad and have a higher peak than those created by<br />
switch-on (fig.4). Cluster analysis [2] in the blue area of<br />
fig.4 resulted in two equally populated groups identified as<br />
on and off switches by cross checking the Ex component.<br />
The distribution of the data in terms of probability can<br />
be visualized with quantiles of gauss distribution (fig.5).<br />
The mean values of all transients in a certain time segment<br />
are sorted by ascending value and are then plotted<br />
against an axis, which is scaled in such a way that normal<br />
distributed data will show as a line of slope 1. This has<br />
been done for each component over a 62 ms window at the<br />
end of each switch-off time series, which have been filtered<br />
and levelled beforehand.<br />
The top graph shows that the distribution of Hz (and<br />
- to a smaller extent - Ex) at station A is dominated by<br />
a step. However, cluster analysis could not sort out these<br />
time series. The other stations show a much better result<br />
with normal distributed data for all components.<br />
4. Stacking and 1D Occam inversion<br />
After the data is analysed and if it needs no further<br />
editing (like correction of the right switch moment) it is<br />
ready to be stacked and smoothed.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
150<br />
• The data was stacked selectively omitting the highest<br />
and lowest 5% of the values per DP.
Abbildung 6: Normalized stacked transients sorted by component<br />
Abbildung 7: Normalized system responses measured broadside<br />
Abbildung 8: Result of Occam inversion of data at station A for 1st<br />
and 2nd order roughness<br />
• Afterwards it was smoothed with a dynamic Hanning<br />
window function.<br />
Fig.6 shows the normalized transients of the stations<br />
sorted by components. Field setup is according to fig. 1.<br />
Ex shows a noticeable “bump” in early times at receiver<br />
station C, which could be a 2D effect of the water-land<br />
interface. More on interpretation of this data can be found<br />
in Lippert et al., 2010 .<br />
System responses measured on land are shown in fig.7<br />
and are used for convolution in forward calculation. Fig.8<br />
shows the Occam inversion results for station A. All components<br />
show a good conductor at about 80 m depth between<br />
two more resistive layers at approximately 10 m and<br />
150 m depth, which represent the sought fresh water bodies.<br />
Hz ˙ could resolve a second good conductor, which is<br />
not visible in ˙ Hy and only indicated in Ex. These results<br />
are in good agreement with the SHOTEM measurements<br />
of this area.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
151
5. Future research<br />
Some system responses measured broadside include negative<br />
values. This is supposed to be a geometric phenomenom<br />
related to the ’air wave’. Future measurements of<br />
the system response will therefore be conducted in the laboratory.<br />
For the geometries of stations B and C there are at the<br />
moment no inversion codes available, therefore they will be<br />
interpreted using 2D forward modelling. Joint inversions<br />
using different LOTEM components will be realized.<br />
6. References<br />
Literatur<br />
[1] T. Hanstein. Digitale Optimalfilter für LOTEM Daten. Protokoll<br />
über das 16. Kolloquium EMTF, pages 320–328, 1996.<br />
[2] S. Helwig. Clusteranalyse als Tool zur Selektion und Verarbeitung<br />
elektromagnetischer Daten. Protokoll über das 16. Kolloquium<br />
EMTF, pages 156–161, 1996.<br />
[3] K. Lippert et al. Erkundung eines Aquifers unter dem Mittelmeer<br />
vor der israelischen Küste mit LOTEM. Protokoll über das 23.<br />
Kolloquium EMTF, 2010.<br />
[4] A. Osman. Interpretation der LOTEM-Daten in näherer Umgebung<br />
von den Bohrungen des KTB. Diploma Thesis, Institute<br />
for Geophysics, University of Cologne, 1995.<br />
[5] C. Scholl. Die Periodizität von Sendesignalen bei LOTEM. Diploma<br />
Thesis, Institute for Geophysics, University of Cologne,<br />
2001.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
152
Grundwasserkontamination bei Roorkee/Indien:<br />
2D Joint Inversion von Radiomagnetotellurik und Gleichstromgeoelektrik<br />
Daten<br />
P. Yogeshwar 1 , B. Tezkan 1 , M. Israil 2<br />
1: Institut für Geophysik und Meteorologie, Universität zu Köln, Email: yogeshwar@geo.uni-koeln.de<br />
2: Earth Science Departement, Indian Institute of Technology, Roorkee<br />
In der landwirtschaftlich geprägten Region um Roorkee in Nordindien ist der Einsatz von Düngemittel<br />
sowie die Bewässerung der Felder mit Abwasser gängige Praxis, wobei das Abwasser durch<br />
offene, teilweise unbefestigte Kanäle transportiert wird. Die umliegende Landbevölkerung bezieht<br />
ihr Frischwasser oftmals aus Brunnen, die aus einem oberflächennahen Aquifer gespeist werden.<br />
Eine Gefährdung dieser oberflächennahen Aquifersysteme durch den praktizierten Umgang mit<br />
Abwasser und Düngemittel ist nicht auszuschließen.<br />
Im Rahmen der hier durchgeführten Untersuchungen wurden ein landwirtschaftlich genutztes und<br />
stark mit Abwasser bewässertes Gebiet nahe einer Mülldeponie und ein als kontaminationsfrei<br />
angenommenes Referenzprofil mit den Methoden der Gleichstromgeolelektrik (DC) und der Radiomagnetotellurik<br />
(RMT) vermessen.<br />
Die einzelnen Profile wurden neben Benutzung von gängiger 1D und 2D Auswertesoftware mittels<br />
eines neu entwickelten 2D Joint Inversionsprogrammes invertiert und interpretiert.<br />
Verglichen mit den Ergebnissen des Referenzprofils zeigen die Ergebnisse auf dem kontaminierten<br />
Gebiet nahe der Deponie deutlich erhöhte Leitfähigkeitswerte bis zu einer Tiefe von ca. 15 m. Der<br />
Tiefenbereich von 5 m bis ungefähr 15 m konnte mit dem oberflächennahen Aquifer identifiziert<br />
werden. Darüber hinaus ergibt sich eine Abnahme der Leitfähigkeitswerte mit dem Abstand von<br />
der Mülldeponie und den Abwasserauslässen.<br />
Durch die Anwendung einer 2D Joint Inversion konnten die Datensätze beider Methoden durch<br />
einheitliche Untergrundmodelle angepasst werden.<br />
Einleitung<br />
Roorkee befindet sich in den Vorläufern des Himalaja<br />
am rechten Ufer des Solani Flusses im Distrikt Haridwar<br />
des nordindischen Staates Uttharakand (Geographische<br />
Breite: 29 ◦ 50 ′ bis 29 ◦ 56 ′ N, Geographische<br />
Länge: 77 ◦ 48 ′ bis 77 ◦ 56 ′ E). Der Solani ist ein Nebenfluss<br />
des Ganges. Der Ganges ist in dem Gebiet<br />
kanalisiert und teilt Roorkee in „Old“- und „New“-<br />
Roorkee. Die Stadt hat ca. 100000 Einwohner, mit<br />
einer jedoch wachsenden Bevölkerung. Die Region ist<br />
stark landwirtschaftlich geprägt, wobei sich dort vermehrt<br />
Industrie entwickelt.<br />
Die Untersuchungen fanden in Zusammenarbeit<br />
mit dem Indian Institute of Technology in Roorkee/Nordindien<br />
(IIT-Roorkee) im Rahmen eines<br />
deutsch-indischen Partnerprojektes statt. Das Hauptanliegen<br />
dieses Projektes besteht in der Anwendung<br />
geophysikalischer Methoden zur Abschätzung der<br />
Gefährdung der Aquifersysteme der Region um Roorkee.<br />
In der landwirtschaftlich geprägten Region um Roorkee<br />
ist der Einsatz von Düngemittel sowie die Bewässerung<br />
der Felder mit Abwasser gängige Praxis,<br />
wobei das Abwasser durch offene, teilweise nicht befestigte,<br />
Kanäle transportiert wird. Gleichzeitig bezieht<br />
die umliegende Landbevölkerung ihr Frischwasser<br />
oftmals aus Brunnen, die aus oberflächennahen<br />
Aquiferen gespeist werden. Eine Gefährdung dieser<br />
oberflächennahen Aquifersysteme durch den praktizierten<br />
Umgang mit Abwasser und Düngemittel ist<br />
nicht auszuschließen.<br />
Ein weiteres Risiko geht von Mülldeponien aus, die<br />
zu Grundwasserkontamination in deren unmittelbaren<br />
Umgebung führen können und deren schädlicher<br />
Einfluss auf die Aquifersysteme der Region ebenfalls<br />
nicht auszuschließen ist. Die Grundwasserproblematik<br />
wird, durch den Frischwasserbedarf der zunehmenden<br />
Bevölkerung und der wachsenden Industrie,<br />
verschärft.<br />
<strong>Geophysikalische</strong> Erkundungsmethoden haben sich<br />
in ihrer Anwendung auf hydrogeologische Fragestellungen,<br />
sowie zur Untersuchung von Altlasten mehrfach<br />
bewährt ([Tezkan, 1999], [Recher, 2002] und [Seher,<br />
2005]). Insbesondere eignet sich die Kombination<br />
von DC und RMT, da sich beide Methoden in ihrem<br />
Erkundungstiefenbereich und ihrer Sensitivität<br />
gegenüber leitfähigen Strukturen ergänzen.<br />
Im Rahmen des Projektes wurde das umliegende Gebiet<br />
einer Mülldeponie nahe dem Dorf Saliyar mit diesen<br />
Methoden profilweise vermessen. Seit 1975 wird<br />
dieses Gebiet zur landwirtschaftlichen Nutzung an<br />
Kleinbauern verpachtet und intensiv mit Abwasser<br />
bewässert, welches von Roorkee durch Rohrleitungen<br />
auf die Felder transportiert wird.<br />
In Abbildung 1 ist eine Übersichtskarte der Region<br />
dargestellt. Das Messgebiet bei Saliyar befindet sich<br />
ca. 6 km nordwestlich von Roorkee auf der rechten<br />
Uferseite des Solani. Das ungeklärte Abwasser der<br />
Stadt wird von der Mahingram Pumpstation über<br />
eine Kanalleitung zur Mülldeponie geleitet. Das vermessene<br />
Referenzgebiet befindet sich auf der linken<br />
Uferseite des Solani und ist ca 10 km von Saliyar entfernt.<br />
Das ganze Gelände hat ein leichtes Gefälle in<br />
Richtung Südosten.<br />
Der Einfluss von Abwassernutzung auf die Grund-<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
153
googlemaps<br />
Abbildung 1: Übersicht der Region um Roorkee. Das Messgebiet Saliyar ist rot, das Referenzgebiet grün<br />
umrandet. Weitere wichtige Plätze sind farblich hervorgehoben und beschriftet. Brunnen, die überwacht und<br />
untersucht wurden sind als weiße Kringel mit roten Punkten dargestellt (Rechte Abbildung: [Singhal et al.,<br />
2003]).<br />
wasserqualität wurde bisher überwiegend unter Berücksichtigung<br />
von hydrochemischen Daten untersucht.<br />
Die chemischen und bakteriellen Analysen des<br />
Grundwassers zeigen bisher nur erhöhte Werte für die<br />
Proben aus der unmittelbarem Umgebung von Saliyar,<br />
wo sich die Mülldeponie befindet. Singhal et al.<br />
[2003] schließen eine Gefährdung der Aquifersysteme<br />
weiter flussabwärts in Richtung Roorkee auf Grund<br />
der angenommenen Grundwasserfließrichtung. Sudha<br />
et al. [2010] haben bereits in den Gebieten nördlich<br />
von Roorkee, in Saliyar und in Khanjapur, eine Vielzahl<br />
von DC-Profilen und „Time Domain Electromagnetic“<br />
(TEM) Stationen vermessen, um den Einfluss<br />
des Abwassers auf das oberflächennahe Aquifer zu detektieren<br />
und die Ausmaße des kontaminierten Bereichs<br />
zu bestimmen.<br />
Ziel dieser Arbeit ist es ebenfalls, den Einfluss der<br />
Kontamination auf die Leitfähigkeitsverteilung der<br />
oberen 20 − 25 m zu bestimmen. Hierzu wurden die<br />
auf dem kontaminierten Gebiet vermessenen Profile<br />
mit dem Referenzprofil verglichen, auf dem keine<br />
Kontamination erwartet wird. Durch die Vielzahl der<br />
vermessenen Profile soll auch eine Aussage über die<br />
laterale Verbreitung des kontaminierten Bereichs, sowie<br />
die Ausbreitung von Kontaminationsfahnen getroffen<br />
werden.<br />
Zusätzlich wurden die Datensätze beider Methoden<br />
mittels des 2D Joint Inversionsprogrammes<br />
RMTDC2D von Candansayar und Tezkan [2008] einzeln<br />
und gemeinsam invertiert. Es soll dabei untersucht<br />
werden, inwiefern die Anwendung der 2D<br />
Joint Inversion verlässlichere und einheitlichere Untergrundmodelle<br />
zu liefern vermag.<br />
Geologie des Messgebiet<br />
Der indische Subkontinent lässt sich geologisch in drei<br />
Bereiche unterteilen: den Himalaja im Norden, die<br />
Gangesebene und die indische Halbinsel im Süden.<br />
Die Gangesebene stellt dabei das Entwässerungsbecken<br />
der großen indischen Flüsse Ganges, Indus,<br />
Brahmaputra und Yamuna dar. Das Messgebiet befindet<br />
sich im nördlichen Teil der Gangesebene (Abbildung<br />
1). Da es sich um Schwemmland des Ganges<br />
und des Solani handelt, weist der Boden überwiegend<br />
ungehärtete alluviale Sedimente jüngeren Alters<br />
mit einer einfachen eindimensionalen Schichtung auf.<br />
Diese Sedimente bestehen aus abwechselnden Schichten<br />
von Ton, Sand, Kunkur und in manchen Gebieten<br />
Kies, wobei Kunkur ein nodulares Kalziumkarbonat<br />
ist, welches sich häufig in Semi-ariden Regionen<br />
bildet. Lithologische Diagramme bei [Singhal et al.,<br />
Bohrlochinformation DC Ergebnisse<br />
Lithologie z (m) ρ (Ωm) z (m)<br />
Sandiger Lehm 0-4 >100 0-7<br />
Sand 4-32 40-100 7-38<br />
Ton + Kies 32-45 30-50 38-45<br />
Sand 45-105 — —<br />
Tabelle 1: Bohrlochinformationen verglichen mit den<br />
DC Ergebnissen in Sherpur von Sudha et al. [2010].<br />
2003] zeigen eine oberflächennahe Schicht aus sehr<br />
feinem bis sandigem Lehm von 3 − 6 m Mächtigkeit.<br />
Die darunterliegende Sandschicht in 3 − 27 m Tiefe<br />
bezeichnet den ersten unbegrenzten, oberflächennahen<br />
Aquifer. Darunter befindet sich eine nichtpermeable<br />
Ton-Kies-Schicht von ca. 13 m Mächtigkeit<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
154
googlemaps<br />
Abbildung 2: Die vermessenen Profile sind rot dargestellt und tragen die Bezeichnung der Profilnummer<br />
(PR1 bis PR13) sowie den Zusatz RMT oder RMT-DC, je nachdem ob nur RMT oder zusätzlich DC gemessen<br />
wurde. (Ab-)Wasserkanäle sind blau und Hochspannungsleitungen schwarz gestrichelt dargestellt. Der<br />
Hauptabwasserkanal verläuft von 1 nach 2. Auslässe für das Abwasser befinden sich entlang des Hauptabwasserkanals<br />
ca. alle 50m. An den Orten S1, S2 und S3 wurden Wasserproben entnommen.<br />
und trennt den oberen Grundwasserleiter von dem<br />
tieferen, der überwiegend aus Sand und Kies besteht.<br />
In der Tabelle 1 sind die DC-Inversionsergebnisse von<br />
Sudha et al. [2010] mit den Bohrlochdaten verglichen.<br />
Die Bohrlochdaten wurden von einem Litholog auf<br />
dem Referenzprofil bei Sherpur abgeleitet und zeigen<br />
gute Übereinstimmung mit den DC-Ergebnissen.<br />
Messung bei Roorkee<br />
Insgesamt wurden 13 RMT und 8 DC Profile bei einer<br />
zweiwöchigen Kampagne im Februar 2009 vermessen<br />
(Abbildung 2). Zu dieser Jahreszeit war das Messgebiet<br />
nur teilweise kultiviert und daher gut zugänglich.<br />
Die Profile sind, ausgehend von der Mülldeponie<br />
am südlichen Bildrand bis zum Flussbett des Solani,<br />
überwiegend parallel angeordnet. Der Abstand der<br />
Profile betrug ca. 50 m, je nach verfügbarem Platz.<br />
Damit wurde eine ca. 200 × 600 m 2 große Fläche abgedeckt.<br />
Das ganze Gebiet nordwestlich vom Hauptabwasserkanal<br />
hat ein leichtes Gefälle in Richtung<br />
Nordosten zum Solani hin. Die Mülldeponie im unteren<br />
Bildteil umfasst eine Fläche von 170 × 100 m 2 .<br />
Eine Messung direkt auf der Mülldeponie war nicht<br />
möglich, deswegen wurden Profil 8 und Profil 10 seitlich<br />
positioniert um einen etwaigen Einfluss der Mülldeponie<br />
durch z.B. Sickerwasser zu ermitteln.<br />
Der Hauptabwasserkanal verläuft von der Mülldepo-<br />
nie nordöstlich in Richtung des Solani. Ungefähr alle<br />
50 m befinden sich Wasserauslässe, um die Felder<br />
zu fluten. Der kleinere offene Kanal (hellblau dargestellt)<br />
im linken Bildbereich verläuft parallel zum<br />
großen und wurde von den Profilen gekreuzt. Der Bereich<br />
oberhalb von Profil 6 war stark bewachsen und<br />
konnte nicht vermessen werden. Das oberste Profil<br />
(Profil 7) verläuft im trockenen Flussbett des Solani<br />
und ist ca. 600 m von der Kontaminationsquelle entfernt,<br />
weswegen hier eine Abnahme der Kontamination<br />
zu erwarten ist. Der Bereich zwischen Profil 3 und<br />
Profil 6 war stark bewässert, konnte aber dennoch<br />
vermessen werden. Profil 1, welches im Folgenden explizit<br />
vorgestellt wird, wurde am ersten Tag mit einer<br />
Länge von 710 m und einem Stationsabstand von<br />
10 − 50 m vermessen, um den Einfluss der Kontamination<br />
mit der Entfernung von der Quelle abzuschätzen.<br />
Die RMT-Messung wurde mit dem RMT-F<br />
Gerät der Universität zu Köln durchgeführt. Es arbeitet<br />
im erweiterten Frequenzbereich von 10 kHz bis<br />
1 MHz und ermöglicht dadurch ein verbessertes Auflösungsvermögen<br />
oberflächennaher Schichten [Wiebe,<br />
2007]. Der Messpunktabstand für die RMT-Stationen<br />
betrug bei allen anderen Profilen 10 m und die Anzahl<br />
in der Regel 21 Stationen. Gemessen wurden beide<br />
horizontalen Komponenten des elektrischen und<br />
des magnetischen Feldes zur Bestimmung der spezifischen<br />
Widerstands- und Phasenwerte der TE- und<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
155
(a) (b)<br />
(c) (d)<br />
Abbildung 3: Inversionsmodelle des Referenzprofils 11 bei Sherpur der Daten der Wennerauslage (a), der<br />
Daten der TM-Mode (b), der der TE-Mode (c), sowie beider Datensätze (TE+TM) gemeinsam invertiert<br />
(d). Die RMT Stationen sind als schwarze Dreiecke dargestellt. Die 0.2 Isolinie des DOI ist als schwarz<br />
gestrichelte Linie geplottet und stellt die maximale Erkundungstiefe dar.<br />
der TM-Mode. Für die DC wurde eine ABEM Terrameter<br />
SAS 1000 benutzt. Als Messauslage wurde<br />
eine Wenner und eine Schlumberger „Short-Long“-<br />
Auslage gewählt. Durch den geringeren Elektrodenabstand<br />
im mittleren Bereich der Auslage ist eine<br />
oberflächennah bessere Auflösung gewährleistet.<br />
Eine geologische Streichrichtung zur Ausrichtung der<br />
Profile lässt sich anhand der Kontaminationsausbreitung<br />
nur ungenügend festlegen. Zu erwarten wäre<br />
eine Ausbreitung der Kontamination von den Abwasserauslässen<br />
Richtung Nordwesten oder auch eine<br />
Verbreitung der Kontamination von der Mülldeponie<br />
ausgehend.<br />
Demzufolge und in Anbetracht der nahezu eindimensionalen<br />
Struktur des Untergrundes wurden die Profile<br />
anhand der verfügbaren Sender für die RMT und<br />
der Verfügbarkeit von Platz ausgerichtet.<br />
Inversionsergebnisse Referenzprofils 11 bei<br />
Sherpur<br />
In den Abbildungen 3 sind die Inversionsmodelle der<br />
einzelnen 2D Inversionen für die Wennerauslage und<br />
für die RMT Daten mit DC2DINVRES [Günther,<br />
2004] bzw. RUND2_NLCG2_FAST (RUND2) [Rodi<br />
und Mackie, 2001] dargestellt. Die nicht-sensitiven<br />
Bereiche sind in den Inversionsergebnissen ausgeblendet<br />
und der „Depth Of Investigation“-Index (DOI)<br />
nach Oldenburg und Li [1998] ist zur Abschätzung der<br />
maximalen Erkundungstiefe in den RMT-Modellen<br />
dargestellt.<br />
Das DC Inversionsergebnis zeigt wie erwartet eine<br />
hochohmige Deckschicht mit einem spezifischen Widerstand<br />
von ρ>200 Ωm und einer Mächtigkeit von<br />
5 − 7 m, welche als sandige Lehmschicht identifiziert<br />
werden kann. Im Bereich von 0 − 60 m war das Profil<br />
sehr trocken, da dieser Bereich nicht als landwirtschaftliche<br />
Nutzfläche diente. Der Bereich von 60-<br />
120 m war kultiviert und wurde wahrscheinlich regelmäßig<br />
künstlich bewässert, wodurch sich der erniedrigte<br />
Widerstand der Deckschicht bis zu einer<br />
Tiefe von ungefähr 2 m erklären lässt. Der Bereich<br />
danach zeigt wieder eine durchgehende hochohmige<br />
Deckschicht, obwohl dieser ebenfalls kultiviert war.<br />
Entsprechend den lithologischen Informationen aus<br />
Tabelle 1 und den DC Ergebnissen von Sudha et al.<br />
[2010] zeigt das Inversionsmodell unter der Deckschicht<br />
eine leitfähige Schicht mit einem spezifischen<br />
Widerstand von 30−50 Ωm, welches das oberflächennahe<br />
Aquifer, also eine wassergesättigte Sandschicht,<br />
darstellt und als die aufzulösende Zielstruktur aufgefasst<br />
wird.<br />
Die Inversionsmodelle der Daten der TM-Mode, der<br />
TE- Mode und der Joint Inversion beider Datensätze<br />
zeigen qualitativ eine identische Leitfähigkeitsstruktur.<br />
Auch die Abnahme des Widerstands im mittleren<br />
Teil des Profils ist identisch zum DC Inversionsmodell.<br />
Die hochohmige Deckschicht wird allerdings mit einem<br />
spezifischen Widerstand von 100 − 200 Ωm rekonstruiert,<br />
welcher verglichen mit dem DC-Ergebnis<br />
zu gering ist. Die darunterliegende Zielstruktur wird<br />
wiederum mit einem spezifischen Widerstand von<br />
20 − 40 Ωm aufgelöst und ist vergleichbar mit dem<br />
DC-Ergebnis. Zusätzlich wurde ein zweites RMT Profil<br />
in ca. 200 m Entfernung von Profil 11 vermessen.<br />
Die geologische Struktur ist hier identisch und der<br />
spezifische Widerstand der Zielstruktur wird ebenfalls<br />
mit einem Wert von 20 − 40 Ωm gut aufgelöst<br />
Yogeshwar [2010].<br />
Eine Erklärung für den verminderten Widerstand der<br />
Deckschicht ist die geringe Sensitivität der RMT für<br />
hochohmige Bereiche. Desweiteren könnte ein statischer<br />
Versatz der ρa-Daten den geringen Widerstand<br />
der Deckschicht erklären. Oberflächennahe Inhomogenitäten<br />
in der Nähe der Empfängerelektroden kön-<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
156
nen z.B. einen zusätzlichen Versatz auf die ρa-Daten<br />
bewirken.<br />
Die Unterkante des Aquifers wird in den Inversionsmodellen<br />
nicht aufgelöst. Zum einen hängt das<br />
mit der maximalen Erkundungstiefe zusammen, zum<br />
anderen ergibt sich wahrscheinlich kein entscheidender<br />
Leitfähigkeitskontrast zwischen dem oberflächennahen<br />
Aquifer und der darunterliegenden nichtpermeablen<br />
Ton-Kies-Schicht, deren spezifischer Widerstand<br />
ebenfalls ungefähr 40 − 50 Ωm beträgt.<br />
Um die maximale Erkundungstiefe abzuschätzen<br />
wurde der DOI nach Oldenburg und Li [1998] berechnet.<br />
Für die Wennerauslage ergibt sich bei [Yogeshwar,<br />
2010] eine maximale Erkundungstiefe von 20 m<br />
bis 25 m für den mittleren Bereich des Profils.<br />
Für die RMT-Modelle ist der DOI als schwarz gestrichelte<br />
Linie in den Modellen dargestellt. Für das<br />
Inversionsmodell der Daten der TM-Mode ist die Erkundungstiefe<br />
ungefähr 15 − 20 m, wobei im Bereich<br />
von x = 100 − 200 m der DOI, auf Grund der erhöhten<br />
Leitfähigkeit, etwas abnimmt. Für die beiden<br />
anderen RMT-Inversionsmodelle in Abbildung 3 ergibt<br />
sich eine ähnliche Erkundungstiefe von ca. 15 m.<br />
Inversionsergebnisse des kontaminierten Profils 1<br />
bei Saliyar<br />
Das Profil 1 befindet sich in ca 100 m Entfernung<br />
zur Mülldeponie und wurde am ersten Messtag vermessen<br />
(Abbildung 2). Es war nicht bewässert und<br />
kaum kultiviert. Benutzt wurde eine Wenner Long-<br />
Auslage mit einer Länge von 120 m, im Gegensatz zu<br />
den anderen DC Profilen, die ausschließlich eine Profillänge<br />
von 200 m haben. Das RMT-Profil wurde bis<br />
zu einer Entfernung von 710 m vom Hauptabwasserkanal<br />
(Abbildung 2) fortgesetzt um den Einfluss der<br />
Kontamination mit dem Abstand zur Quelle zu bestimmen,<br />
wobei der Stationsabstand im Bereich von<br />
x =0− 200 m zehn Meter betrug und dann bis zu<br />
50 m vergrößert wurde. Ausgewertet wurden nur die<br />
Daten der TE-Mode, da die der TM-Mode viele Senderlücken<br />
und teilweise negative Phasenwerte aufwiesen.<br />
In den Abbildungen 4(a) und 4(b) sind die Inversionsmodelle<br />
der einzelnen 2D Inversionen für die Wennerauslage<br />
und für die Daten der TE- Mode dargestellt.<br />
Das DC-Modell zeigt eine einfache, eindimensionale<br />
geologische Schichtung, wobei die Deckschicht im<br />
Vergleich zum Referenzprofil in Abbildung 3 wesentlich<br />
weniger mächtig ist und einen geringeren Widerstand<br />
von ca. 30 Ωm aufweist. Eine mögliche Ursache<br />
für die insgesamt weniger hochohmige Deckschicht<br />
mit ca. 2-3 m Mächtigkeit ist wahrscheinlich die sonst<br />
häufige Bewässerung mit Abwasser. Im rechten Bereich<br />
nahe des Hauptabwasserkanals befindet sich ein<br />
hochohmiger Bereich der Deckschicht, welcher von einer<br />
Erdaufschüttung herrühren könnte, und der in<br />
dem RMT Inversionsmodell identisch rekonstruiert<br />
ist. Darunter folgt das kontaminierte Aquifer, bzw.,<br />
verglichen mit dem Ergebnis für das Referenzprofil,<br />
ein Bereich stark erhöhter Leitfähigkeit bis zu<br />
einer Tiefe von ungefähr 12 m. Dies ist im RMT-<br />
(a)<br />
(b)<br />
Abbildung 4: Inversionsergebnis des Profils 1 bei Saliyar<br />
der Daten der Wennerauslage (a) und der Daten<br />
der TE-Mode (b). Die RMT Stationen sind als<br />
schwarze Dreiecke dargestellt, der DOI als schwarz<br />
gestrichelte Linie.<br />
Profil ebenfalls identisch wiedergegeben. Der spezifische<br />
Widerstand fällt im DC-Modell mit 6 − 15 Ωm<br />
etwas höher aus als der des RMT-Modells. Die Anpassung<br />
beider Datensätze ist mit einem RMS < 2.5<br />
gut.<br />
Die maximale Erkundungstiefe für das DC-Modell ergibt<br />
sich mit dem DOI-Index bei Yogeshwar [2010]<br />
zu 15 − 20 m im mittleren Bereich des Profils. Die<br />
schwarze Linie in Abbildung 4 stellt die maximale<br />
Erkundungstiefe von 10 − 15 m für das RMT-Modell<br />
dar. Mit dieser Erkundungstiefe für die RMT wird<br />
die Unterkante des kontaminierten Bereichs nur mit<br />
der DC aufgelöst.<br />
Auf den vermessenen Profilen in Saliyar wurde mit<br />
dem Programm EMUPLUS [Wiebe, 2007] eine 1D<br />
RMT-DC und eine 1D RMT-DC Joint Inversion im<br />
jeweiligen Mittelpunkt des Profils durchgeführt um<br />
die Leitfähigkeit des kontaminierten Aquifers genauer<br />
zu bestimmen. Die 1D Marquardt Inversionen für<br />
Profil 1 (Abbildung 5) entsprechen sehr gut den 2D<br />
Ergebnissen im Mittelpunkt des Profils. Das Modell<br />
der Geoelektrik zeigt einen Dreischichtfall mit gut<br />
aufgelösten Schichtgrenzen und spezifischen Widerständen,<br />
außer für den der obersten Schicht. Die äquivalenten<br />
Modelle in Abbildung 5 unterscheiden sich<br />
also kaum. Ebenso zeigt das Ergebnis für die RMT-<br />
Daten nahezu identische äquivalente Modelle, was auf<br />
eine gute Auflösung beider Schichten schließen lässt.<br />
Beide Inversionsergebnisse stimmen gut überein, wobei<br />
die RMT die Unterkante des Aquifers nicht auflösen<br />
kann.<br />
Das Modell der 1D Joint Inversion passt die Datensätze<br />
beider Methoden mit einem RMS von 1.65 gut<br />
an und löst die Unterkante der zweiten Schicht durch<br />
Hinzunahme der DC Daten auf. Der Widerstand des<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
157
z (m)<br />
0<br />
5<br />
10<br />
15<br />
Profile 1, Schlumberger, VES at x=60m<br />
equi. models<br />
RMS=1.39<br />
20<br />
0 10 20 30 40 50 60 70<br />
ρ (Ωm)<br />
(a)<br />
z (m)<br />
0<br />
5<br />
10<br />
15<br />
Profile 1, te, sounding at x=60m<br />
equi. models<br />
RMS=1.75<br />
20<br />
0 10 20 30 40 50 60 70<br />
ρ (Ωm)<br />
(b)<br />
z (m)<br />
Profile 1, x=60m, Joint Inversion RMT/DC<br />
0<br />
5<br />
10<br />
15<br />
RMS all =1.65,<br />
RMS dc =1.4,<br />
RMS rmt =2.1, cf=0.82<br />
20<br />
0 10 20 30 40 50 60 70<br />
ρ (Ωm)<br />
Abbildung 5: 1D Marquardt Inversionsmodelle für die Sondierungskurve der Schlumbergerauslage (a) und<br />
der Daten der TE-Mode (b), sowie der Joint Inversion beider Datensätze (c) bei x =60m in der Mitte des<br />
Profils 1 bei Saliyar. Aufgetragen ist der spezifische Widerstand gegen die Tiefe. Äquivalente Modelle sind<br />
grün und der Mittelwert der äquivalenten Modelle sind rot dargestellt.<br />
kontaminierten Bereichs ist mit ρ ≈ 10 Ωm ziemlich<br />
genau bestimmt. Da weitaus mehr DC-Daten vorhanden<br />
sind, stellt das Inversionsergebnis zwar einen Informationsgewinn<br />
für die RMT dar, entspricht aber<br />
ziemlich genau dem 1D Modell der DC Inversion.<br />
Nur oberflächennah liefern die hochfrequenten RMT-<br />
Daten Zusatzinformation zur Verbesserung der Auflösung<br />
der obersten Schicht.<br />
2d Joint Inversion<br />
In der Abbildung 6 sind die Inversionsmodelle der<br />
Wenner Long Auslage, der Daten der TE-Mode und<br />
das Ergebnis der 2D Joint Inversion beider Datensätze<br />
mit RMTDC2D dargestellt. Die Leitfähigkeitsstruktur<br />
der Endmodelle ist gut vergleichbar mit den<br />
Inversionsmodellen der einzelnen 2D Inversionen mit<br />
DC2DINVRES und RUND2.<br />
Die Modelle der einzelnen Inversionen sind bis zu einer<br />
Tiefe von ungefähr 10 m gut vergleichbar, wobei<br />
die Anpassung der RMT-Daten mit einem RMS von<br />
1.9 besser ist als die der DC-Daten. Der kontaminierte<br />
Bereich wird im RMT Inversionsmodell etwas leitfähiger<br />
rekonstruiert. Eine Aussage über die Unterkante<br />
des Aquifers lässt sich auf Grund der mangelnden<br />
Eindringtiefe der RMT nicht treffen, wohingegen<br />
die Unterkante im DC-Modell aufgelöst wird. Die 2D<br />
Joint Inversion in Abbildung 6 passt beide Datensätze<br />
mit einem RMS von 5.0 zufriedenstellend an. Die<br />
Unterkante und der Widerstand des Aquifers sind gut<br />
aufgelöst.<br />
Die mit dem mittleren Tiefenbereich korrespondierenden<br />
DC-Daten in Abbildung 7(b) sind über die<br />
ganze Profillänge relativ gleichmäßig,<br />
z (m)<br />
z (m)<br />
−5<br />
0<br />
5<br />
10<br />
15<br />
(c)<br />
Profile 1/Saliyar, Wenner long, ρ 0 =9 Ω m, It.=9 Alpha=5 RMS=3.0816<br />
20<br />
0 10 20 30 40 50<br />
x (m)<br />
60 70 80 90 100<br />
5<br />
−5<br />
0<br />
5<br />
10<br />
15<br />
(a)<br />
Profile 1/Saliyar, TE, ρ 0 =9 Ω m, It.=30 Alpha=5 RMS=1.9128<br />
20<br />
0 10 20 30 40 50<br />
x (m)<br />
60 70 80 90 100<br />
5<br />
(b)<br />
Profile 1/Saliyar, TE/Wenner long, ρ =9 Ω m, It.=12 Alpha=10 RMS=5.0686<br />
0<br />
−5<br />
100<br />
z (m)<br />
0<br />
5<br />
10<br />
15<br />
20<br />
0 10 20 30 40 50<br />
x (m)<br />
60 70 80 90 100<br />
5<br />
(c)<br />
Abbildung 6: Inversionsergebnis der Wenner Long-<br />
Auslage (a), der TE-Mode (b) und der 2D Joint Inversion<br />
(c) des Profils 1. Die RMT-Stationen sind als<br />
schwarze Dreiecke dargestellt.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
158<br />
100<br />
65<br />
42<br />
28<br />
18<br />
12<br />
8<br />
65<br />
42<br />
28<br />
18<br />
12<br />
8<br />
ρ (Ω m)<br />
100<br />
65<br />
42<br />
28<br />
18<br />
12<br />
8<br />
ρ (Ω m)<br />
ρ (Ω m)
ρ a (Ωm)<br />
ρ a (Ωm)<br />
100<br />
80<br />
60<br />
40<br />
20<br />
0<br />
100<br />
80<br />
60<br />
40<br />
20<br />
0<br />
45 °<br />
Profile 1, TE−mode,RMS=6.52, f=18000 Hz<br />
ρ a,mess<br />
ρ a,calc<br />
φ mess<br />
φ calc<br />
20<br />
0 20 40 60<br />
x (m)<br />
80 100 120<br />
Profile 1, TE−mode,RMS=6.52, f=666000 Hz<br />
ρ<br />
a,mess<br />
ρ<br />
a,calc φ<br />
mess<br />
φ<br />
calc<br />
80<br />
45 °<br />
0 20 40 60<br />
x (m)<br />
(a)<br />
80 100 120<br />
80<br />
70<br />
60<br />
50<br />
40<br />
30<br />
70<br />
60<br />
50<br />
40<br />
30<br />
20<br />
φ (°)<br />
φ (°)<br />
n−level<br />
n−level<br />
n−level<br />
1<br />
3<br />
6<br />
8<br />
10<br />
12<br />
1<br />
3<br />
6<br />
8<br />
10<br />
12<br />
1<br />
3<br />
6<br />
8<br />
10<br />
12<br />
Meas. Pseudo−section:Wenner<br />
Calc. Pseudo−section: Wenner<br />
0 20 40 60<br />
x (m)<br />
(b)<br />
80 100 120<br />
ρ a (Ω m)<br />
50<br />
40<br />
30<br />
20<br />
10<br />
ρ a (Ω m)<br />
50<br />
40<br />
30<br />
20<br />
10<br />
%Misfit Section: Wenner (RMS=4.0584) %MISFIT<br />
10<br />
Abbildung 7: Daten und Datenanpassung der Daten der TE-Mode (a) und der Wenner Long Auslage (b) für<br />
das Ergebnis der 2D Joint Inversion mit RMTDC2D. Die gemessenen ρa und ϕ Werte sind als schwarze,<br />
bzw. rote Punkte und die Modellantwort als schwarze, bzw. rote Kreuze für f =18kHz, 666 kHz gegen die<br />
Station aufgetragen. Die DC-Daten und die Modellantwort sind als Pseudosektionen dargestellt, wobei die<br />
untere Abbildung (b) die relative Differenz beider zeigt.<br />
wohingegen die RMT Daten der 18 kHz Frequenz in<br />
Abbildung 7(a) und das Inversionsmodell der einzelnen<br />
2D RMT Inversion in Abbildung 6(b) auch Bereiche<br />
erhöhter Leitfähigkeit zeigen. Bei der 2D Joint<br />
Inversion werden die RMT-Daten daher nicht so gut<br />
wie die DC-Daten angepasst. Die Modellantwort der<br />
RMT ist durch die Regularisierung und dem Einfluss<br />
der insgesamt gleichmäßigeren DC-Daten sehr glatt.<br />
Der RMS für die RMT-Daten ergibt sich zu 6.5. Die<br />
Hinzunahme der RMT-Daten führt zu einer besseren<br />
Bestimmung der Deckschicht vor allem im etwas<br />
hochohmigeren Bereich von x =80− 100 m. Die DC-<br />
Daten in diesem Bereich sind nicht so gut angepasst,<br />
wie in der Darstellung der relativen Differenz der Daten<br />
und der Modellantwort in Abbildung 7(b) erkennbar<br />
ist. Eine Verringerung des Glättungsparameters<br />
passt diese zwar besser an, führt allerdings zu Inversionsartefakten<br />
für die tieferen Bereiche.<br />
Qualitativ zeigen die Auflösungsmatrizen der 2D<br />
Joint Inversion bei Yogeshwar [2010] eine verbesserte<br />
Auflösung im Vergleich zur RMT-Einzelinversion,<br />
wobei die Randbereiche im Vergleich zur DC-<br />
Einzelinversion besser aufgelöst sind, da die DC dort<br />
keine Tiefeninformation liefert. Die maximale Erkundungstiefe<br />
wird durch die DC-Daten bestimmt, weswegen<br />
keine erhöhte Sensitivität der tieferen Strukturen<br />
durch Hinzunahme der RMT-Daten erreicht wird.<br />
Abschätzung der Kontaminationsverbreitung<br />
Um die lateralen Ausmaße des kontaminierten Bereichs<br />
abzuschätzen, wurden insgesamt 9 parallele<br />
RMT und 5 parallele DC Profile auf einem ca 200 ×<br />
600 m 2 großen Gebiet vermessen (Abbildung 2). Die<br />
pr6<br />
pr5<br />
pr4<br />
pr3<br />
pr2<br />
pr1<br />
pr9<br />
pr10<br />
pr7<br />
Waste<br />
Solani<br />
Kanal<br />
0<br />
−10<br />
Abbildung 8: Ansicht aller RMT-Profile von der<br />
Mülldeponie bis zum Solani. Der Hauptabwasserkanal<br />
verläuft von der Mülldeponie in Richtung des Solani<br />
und ist rot dargestellt.<br />
RMT-Profile sind in Abbildung 8 als Flächenschnitte<br />
der xz-Ebene dargestellt. Die Profile verlaufen parallel<br />
zwischen der Mülldeponie und dem Solani. Die<br />
RMT-Modelle liefern auf Grund der teilweise mäßigen<br />
Datenqualität nicht so glatte und gleichmäßige<br />
Modelle wie die DC-Modelle bei Yogeshwar [2010].<br />
In der Darstellungen 8 ist eine leichte Abnahme der<br />
Leitfähigkeit zum Solani hin zu beobachten. Das Profil<br />
7 am Solani in Abbildung 8 zeigt insgesamt hochohmigere<br />
Bereiche im Inversionsmodell für die RMT.<br />
Die Inversionsergebnisse der DC-Daten sind bei Yo-<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
159
z (m)<br />
0<br />
5<br />
10<br />
15<br />
20<br />
Profile 1/3/5/11, Joint Inversion model<br />
1D Joint Pr11<br />
1D Joint Pr1<br />
1D Joint Pr3<br />
1D Joint Pr5<br />
reference site ρ=50 Ωm<br />
10 100<br />
(a)<br />
ρ (Ωm)<br />
z (m)<br />
0<br />
5<br />
10<br />
15<br />
20<br />
Profile 7/8/9/13/11, Joint Inversion model<br />
1D Joint Pr11<br />
1D Joint Pr7<br />
1D Joint Pr8<br />
1D Joint Pr9<br />
1D Joint Pr13<br />
reference site ρ=50 Ωm<br />
10 100<br />
(b)<br />
ρ (Ωm)<br />
z (m)<br />
Profile 2/4/6/10/11, RMT Inversion modells<br />
0<br />
5<br />
10<br />
15<br />
20<br />
1D Joint Pr11<br />
1D RMT Pr2<br />
1D RMT Pr4<br />
1D RMT Pr6<br />
1D RMT Pr10<br />
reference site ρ=50 Ωm<br />
10 100<br />
Abbildung 10: 1D Marquardt Joint Inversionsmodelle der DC- und RMT-Daten im jeweiligen Mittelpunkt<br />
der Auslage für Profil 1,3,5 (a) und Profil 7,8,9,13 (b), sowie einzelne RMT 1D Inversionsmodelle für Profil<br />
2,4,6,10 (c). Das 1D Joint Inversionsmodell des Referenzprofils (Profil 11) ist rot dargestellt.<br />
geshwar [2010] mit denen der RMT sehr gut vergleichbar,<br />
wobei die Unterkante des Aquifers jedoch mit der<br />
DC aufgelöst wird. Um eine genauere Vorstellung<br />
z (m)<br />
z (m)<br />
Relative difference between model result of pr11 and pr1: Wenner ρ 01 =120 Ωm, ρ 02 =15 Ωm<br />
−5<br />
0<br />
5<br />
10<br />
15<br />
0.81<br />
0.64<br />
0.46<br />
0.29<br />
0.11<br />
−0.06<br />
20<br />
−0.24<br />
−0.41<br />
25<br />
−0.59<br />
30<br />
0 20 40 60<br />
x (m)<br />
80 100<br />
−0.76<br />
120<br />
−5<br />
0<br />
5<br />
10<br />
15<br />
20<br />
25<br />
(a)<br />
Relative difference between model results of pr11 and pr1: ρ 01 =50 Ωm, ρ 02 =10 Ωm<br />
30<br />
0 20 40 60 80 100 120<br />
x (m)<br />
(b)<br />
0.99<br />
0.79<br />
0.63<br />
0.47<br />
0.31<br />
−0.51<br />
relative difference<br />
relative difference<br />
0.14<br />
−0.02<br />
−0.18<br />
−0.34<br />
Abbildung 9: Dargestellt ist die relative Differenz<br />
der RMT- (a) und der DC-Inversionsmodelle (a) des<br />
Referenzprofils 11 und des kontaminierten Profils 1<br />
(Relative Differenz > 0 ⇒ ρref >ρpr1).<br />
des Leitfähigkeitsunterschiedes zwischen dem Inversionsmodell<br />
des Referenzprofils und den Profilen auf<br />
dem kontaminierten Bereich zu erhalten, wurde für<br />
(c)<br />
ρ (Ωm)<br />
jedes Profil die relative Differenz beider Modelle berechnet.<br />
Die relative Differenz des Profils 1 mit dem<br />
Referenzprofil 11 in Abbildung 9(a) zeigt für die Geoelektrik<br />
über die gesamte Auslagenlänge einen wesentlich<br />
geringeren spezifischen Widerstand. In einer<br />
Tiefe von ca. 15 m befindet sich die Unterkante des<br />
kontaminierten Aquifers und die Widerstände werden<br />
vergleichbar, bzw. der des Profils 1 wird größer. Für<br />
die RMT zeigt die Abbildung 9(b) ein identisches Ergebnis.<br />
Auch hier ist der spezifische Widerstand des<br />
Inversionsmodells für das Referenzprofil weitaus höher.<br />
Bei beiden Ergebnisses zeigen nur kleine oberflächennahe<br />
Bereiche der Deckschicht ein anderes Verhalten.<br />
In der Darstellung der relativen Differenz bei<br />
Yogeshwar [2010] zeigen die vermessenen Profile auf<br />
dem kontaminierten Gebiet in Saliyar im Vergleich<br />
mit dem Referenzprofil ausschließlich erhöhte Leitfähigkeitswerte<br />
bis zu einer Tiefe von ca. 15 − 20 m, bis<br />
auf das Profil 7, welches sich am Ufer des Solani befand<br />
und am weitesten von der Kontaminationsquelle<br />
entfernt war.<br />
In Abbildung 10(a) und 10(b) sind die 1D Joint Inversionsmodelle<br />
der Daten aller Profilmittelpunkte in<br />
Saliyar mit dem des Referenzprofils verglichen. Insgesamt<br />
zeigen alle Profile unterhalb einer etwas hochohmigeren<br />
Deckschicht einen spezifischen Widerstand<br />
von ρ ≈ 10 Ωm für das kontaminierte Aquifer, wiederum<br />
gefolgt von einer Schicht mit einen erhöhten spezifischen<br />
Widerstand. Das in grün geplottete Modell im<br />
rechten Teil der Abbildung 10 entstammt dem Pro-<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
160
y (m)<br />
y (m)<br />
600<br />
300<br />
250<br />
200<br />
150<br />
100<br />
50<br />
0<br />
600<br />
300<br />
250<br />
200<br />
150<br />
100<br />
50<br />
0<br />
xy−planeview of RMT profiles at the depth of z=1m<br />
0 50 100<br />
x (m)<br />
150 200<br />
(a)<br />
xy−planeview of RMT profiles at the depth of z=10m<br />
0 50 100<br />
x (m)<br />
150 200<br />
(c)<br />
y (m)<br />
y (m)<br />
600<br />
300<br />
250<br />
200<br />
150<br />
100<br />
50<br />
0<br />
600<br />
300<br />
250<br />
200<br />
150<br />
100<br />
50<br />
0<br />
xy−planeview of RMT profiles at the depth of z=5m<br />
0 50 100<br />
x (m)<br />
150 200<br />
(b)<br />
xy−planeview of RMT profiles at the depth of z=15m<br />
pr7<br />
pr6<br />
pr5<br />
pr4<br />
pr3<br />
pr2<br />
pr1<br />
pr9<br />
pr10<br />
canal<br />
waste<br />
0 50 100<br />
x (m)<br />
150 200<br />
2<br />
Abbildung 11: xy-Schnitte der RMT Inversionsmodelle von RUND2 für verschiedene Tiefen: 1 m (a), 5 m<br />
(b),10 m (c) und 15 m (d). Die Kreuze bezeichnen die RMT-Stationen.<br />
fil 7 am Solani und zeigt eine etwas andere Struktur<br />
mit höheren Widerständen, wie schon in Abbildung<br />
8 ersichtlich war. Die Modelle der 1D RMT Inversionen<br />
in Abbildung 10(c) zeigen ebenfalls geringere<br />
spezifische Widerstände für das Aquifer. Das Ergebnis<br />
von Profil 4 in Abbildung 5(c) zeigt den Beginn<br />
einer hochohmige Schicht in ca. 8 m Tiefe, die auch im<br />
2D Modell der TM-Mode zu finden ist, aber nicht in<br />
dem der TE-Mode, und auf mangelnde Datenqualität<br />
oder die Streichrichtung zurückzuführen ist [Yogeshwar,<br />
2010]. Die Ergebnisse der einzelnen 2D Inversionsmodelle<br />
wurden für beide Methoden linear auf ein<br />
dreidimensionales Gitter interpoliert. Um eine Aussage<br />
über die laterale Verbreitung der Kontamination<br />
mit der Tiefe zu treffen, sind in Abbildung 11, am<br />
Beispiel der RMT, Flächenschnitte in der xy-Ebene<br />
für vier verschiedene Tiefen, z=1 m, 5 m, 10 m und<br />
1 5m, dargestellt.<br />
In einer Tiefe von einem Meter erkennt einen sehr<br />
hochohmigen Bereich direkt in der Umgebung der<br />
(d)<br />
ρ(Ωm)<br />
300<br />
172<br />
Mülldeponie. Das gesamte Profil 10, Profil 9 und die<br />
ersten 50 m direkt am Kanal von Profil 1 zeigen diese<br />
hochohmige Deckschicht. Dies könnte von einer Erdaufschüttung<br />
im Zusammenhang mit der Mülldeponie<br />
herrühren. Genaueres ist jedoch nicht bekannt.<br />
Für z = 5 m zeigt die Flächendarstellung spezifischen<br />
Widerstände um die 10 Ωm im Bereich von y =0m<br />
bis 300 m. Die Leitfähigkeit ist in der Nähe der Mülldeponie<br />
erhöht und nimmt dann in Richtung des Solani<br />
etwas ab. Das Profil 7 am Solani ist wie bereits<br />
gesagt über den gesamten Tiefenbereich hochohmiger<br />
als die anderen Profile.<br />
In einer Tiefe von 10 m und 15 m lässt die Kontamination<br />
etwas nach und es zeichnet sich die Unterkante<br />
des Aquifers ab, wobei der Bereich nahe der Mülldeponie<br />
im Vergleich zum Rest leitfähiger ist. Dies lässt<br />
wieder auf einen Einfluss der Mülldeponie schließen,<br />
vor allem da die Leitfähigkeit der Profile in der unmittelbaren<br />
Nähe der Deponie bis zu einer größeren<br />
Tiefe erhöht ist. So ist für Profil 9 in 15 m keine Un-<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
161<br />
99<br />
56<br />
32<br />
19<br />
11<br />
6<br />
3
terkante des kontaminierten Aquifers erkennbar. Die<br />
DC-Ergebnisse bei Yogeshwar [2010] bestätigen die<br />
der RMT, wobei die Unterkante des kontaminierten<br />
Bereichs sich in 15 m wesentlich deutlicher bei den<br />
DC-Ergebnissen abzeichnet.<br />
Eine solche flächenhafte Darstellung ist auf Grund<br />
der Interpolation zwischen den Profilen kritisch zu<br />
bewerten. Zwischen dem letzten Profil und dem Ufer<br />
des Solani lässt sich keine verlässliche Aussage treffen,<br />
da sich die Entfernung auf ca. 300 m beläuft. Problematisch<br />
sind auch Inversionsartefakte, die bei der<br />
Interpolation zwischen den Profilen überakzentuiert<br />
werden. Trotzdem eignet sich die Flächendarstellung<br />
zur Visualisierung der lateralen Leitfähigkeitsverteilung<br />
und insbesondere, um ein räumliches Bild des<br />
Untergrundes zu gewinnen.<br />
Diskussion und Ausblick<br />
Die Inversionsmodelle, der bei der Mülldeponie aufgenommenen<br />
Datensätze, haben in der Regel einander<br />
ähnliche Strukturen. Unterhalb einer etwas hochohmigeren<br />
Deckschicht zeigen diese, im Vergleich zu<br />
dem Inversionsmodell des Referenzprofils, Werte des<br />
spezifischen Widerstands kleiner als 10 Ωm bis zu einer<br />
Tiefe von ungefähr 15 m. Dieser Bereich konnte<br />
mit dem oberflächennahen Grundwasserleiter identifiziert<br />
werden und der im Vergleich geringere Wert<br />
des spezifischen Widerstands kann auf den Einfluss<br />
des Abwassers zurückgeführt werden.<br />
Aus der flächenhaften Darstellung der Inversionsmodelle<br />
konnte die laterale Verbreitung der Kontamination<br />
abgeleitet werden. Erkennbar ist eine Zunahme<br />
des spezifischen Widerstands mit der Entfernung<br />
von der Mülldeponie, so dass diese vermutlich auch<br />
als Kontaminationsquelle angenommen werden kann.<br />
Auch reichen die Werte verringerter spezifischer Widerstände<br />
nahe der Deponie vergleichsweise tief.<br />
Darüber hinaus sind die spezifischen Widerstände der<br />
RMT-Inversionsmodelle in der Nähe des Hauptabwasserkanals,<br />
wo sich die Wasserauslässe befinden,<br />
ebenfalls verringert. In ungefähr 600 m Entfernung<br />
von der Mülldeponie im Flussbett des Solani sind die<br />
Werte des spezifischen Widerstands wiederum deutlich<br />
höher und vergleichbar mit denen des Referenzprofils.<br />
Insgesamt lassen sich die Mülldeponie und das Abwasser<br />
als Kontaminationsquellen bestätigen und es<br />
kann eine Systematik der Kontaminationsverbreitung<br />
mit dem Abstand von den Quellen abgeleitet werden.<br />
Das Programm RMTDC2D [Candansayar und Tezkan,<br />
2008] lieferte vergleichbare Inversionsmodelle<br />
wie die einzelnen Inversionen der Datensätze mit<br />
DC2DINVRES [Günther, 2004] und RUND2 [Rodi<br />
und Mackie, 2001]. Insbesondere die Inversionsmodelle<br />
der einzelnen Inversionen mit RMTDC2D waren<br />
den Inversionsmodellen der anderen Programmen<br />
sehr ähnlich.<br />
Bei Inversionsmodellen, die in ihrer Struktur starke<br />
Unterschiede zeigten, war die Joint Inversion schwierig,<br />
da beide Datensätze nicht gleichzeitig anzupassen<br />
waren. Trotzdem konnte mit der Jointinversion<br />
ein einheitliches Modell beider Datensätze gefunden<br />
werden, was eine nützliche Hilfe bei der Interpretation<br />
der Ergebnisse darstellt.<br />
Vor allem konnte die Unterkante des kontaminierten<br />
Bereichs durch Hinzunahme der DC-Daten aufgelöst<br />
werden und gleichzeitig lieferten die RMT-Daten<br />
Zusatzinformation an den Rändern der Profile, wo<br />
mit der DC keine Tiefenaussage zu treffen möglich<br />
war. Die Joint Inversion führte unter diesem Gesichtspunkt<br />
zu verbesserten Endmodellen.<br />
Problematisch gestaltete sich allerdings die Auswirkung<br />
des Regularisierungsparameters. Bei einem zu<br />
geringen Wert ergaben sich stark überstrukturierte<br />
Inversionsmodelle und Inversionsartefakte.<br />
Schwierig war auch die Bewertung der Inversionsmodelle<br />
anhand der Auflösungsmatrizen. Die Auflösungsmatrizen<br />
der Joint Inversion waren mit denen<br />
der einzelnen Inversionen von RMTDC2D nicht<br />
vergleichbar und eine quantitative Aussage über das<br />
verbesserte Auflösungsvermögen der Joint Inversion<br />
konnte daher nicht getroffen werden.<br />
Im Rahmen des DFG-DST Projektes, welches diese<br />
Arbeit ermöglicht hat, ist eine weitere Messung auf<br />
dem selben Gebiet vorgesehen. Hierbei soll die „Transient<br />
Electromagnetic“ Methode (TEM) zum Einsatz<br />
kommen. Mit dieser Methode besteht die Möglichkeit<br />
tiefer liegende Strukturen aufzulösen. Wünschenswert<br />
wäre mit dieser Methode den Einfluss der<br />
Kontamination auf den tiefer liegenden Grundwasserleiter<br />
zu untersuchen, da sich die hier vorliegende<br />
Arbeit auf die Erkundung des oberen Grundwasserleiter<br />
konzentriert hat.<br />
Des Weiteren eignen sich die verwendeten Methoden<br />
gut für die Erkundung von kontaminierten Böden<br />
und könnten gerade in Indien vermehrt zum Einsatz<br />
kommen um damit zur Verbesserung der Grundwasserversorgung<br />
beizutragen.<br />
Interessant wäre auch die Einbeziehung der geophysikalischen<br />
Ergebnisse in eine hydrogeologische Interpretation,<br />
um die Auswirkung der Kontamination auf<br />
größeren räumlichen und auch zeitlichen Skalen abzuschätzen.<br />
Die 2D Joint Inversion stellt ein zusätzliches Hilfsmittel<br />
zur Interpretation von RMT- und DC-<br />
Inversionsmodellen dar. Die einzeln erhaltenen Inversionsmodelle<br />
lassen sich durch die Hinzunahme der<br />
Joint Inversionsmodelle leichter qualitativ und quantitativ<br />
bewerten. Eine weitere Anwendung dieses Programmes<br />
ist daher wünschenswert.<br />
Literatur<br />
Candansayar, M. und B. Tezkan, Two-dimensional<br />
joint inversion of radiomagnetotelluric and direct current<br />
resistivity data, Geophysical Prospecting, 56, 737–<br />
749, 2008.<br />
Günther, T., Inversion Methods and Resolution Analysis<br />
for the 2D/3D Reconstruction of Resistivity Structures<br />
from DC Measurements, Dissertation, Technischen<br />
Universität Bergakademie Freiberg, 2004.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
162
Oldenburg, D. W. und Y. Li, Estimating the depth<br />
of investigation in DC resistivity and IP surveys, Geophysics,<br />
64, 403–416, 1998.<br />
Recher, S., Dreidimesionale Erkundung von Altlasten<br />
mit Radiomagnetotellurik - Vergleiche mit geophysikalischen,<br />
geochemischen und geologischen Analysen an<br />
Bodenproben aus Rammkernsondierungen, Dissertation,<br />
Universität zu Köln, Institut für Geophysik und<br />
Meteorologie, 2002.<br />
Rodi, W. und R. L. Mackie, Nonlinear conjugate gradients<br />
algorithm for 2D magnetotelluric inversion, Geophysics,<br />
66, (1), 174–187, 2001.<br />
Seher, T., Untersuchung von Feuchtbiotopen in Ostfriesland:<br />
Gefährdungsabschätzung mit Multielektroden-<br />
Geoelektrik und Radiomagnetotellurik, Diplomarbeit,<br />
Universität zu Köln, Institut für Geophysik und Meteorologie,<br />
2005.<br />
Singhal, D., T. Roy, H. Joshi und A. Seth, Evaluation<br />
of Groundwater Pollution in Roorkee Toen, Uttaranchal,<br />
Journal Geological Society of India, 62, 465–<br />
477, 2003.<br />
Sudha, B.Tezkan, M.Israil, D. Singhal und J.Rai,<br />
Geoelektrical mapping of aquifer contamination: a case<br />
study from Roorkee, India, Near Surface Geophysics, 8,<br />
33–42, 2010.<br />
Tezkan, B., A review of environmental applications of<br />
quasi-stationary electromagnetic techniques, Surveys<br />
in Geophysics, 20, 279–308, 1999.<br />
Wiebe, H., 1D-Joint-Inversion von Geoelektrik und Radiomagnetotellurik,<br />
Diplomarbeit, Universität zu Köln,<br />
Institut für Geophysik und Meteorologie, 2007.<br />
Yogeshwar, P., Grundwasserkontamination bei Roorkee/Indien:<br />
2D Joint Inversion von Radiomagnetotellurik<br />
und Gleichstromgeoelektrik Daten, Diplomarbeit,<br />
Universität zu Köln, Institut für Geophysik und Meteorologie,<br />
2010.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
163
Site Effect Assessment in the Mygdonian Basin (EUROSEISTEST area, Northern Greece)<br />
using RMT and TEM Soundings<br />
Widodo 1,2) , Marcus Gurk 2) , Bülent Tezkan 2)<br />
1) Adhi Tama Institute of Technology Surabaya (ITATS), Indonesia<br />
2) Institute of Geophysics and Meteorology, University of Cologne, Germany<br />
Abstract<br />
During the project “Euroseistest Volvi-Thessaloniki”, a strong-motion test site<br />
(EUROSEISTEST) for Engineering Seismology was installed in the Mygdonian Basin between the<br />
two lakes Volvi and Lagada ca. 45 km northeast of Thessaloniki (Northern Greece). The basin itself<br />
is a neotectonic graben structure (5 km wide) with increased seismic activity along distinct normal<br />
fault patterns. Fluvioterrestrial and lacustrien sediments (approximately 350-400 m thick) are<br />
overlying the basement consisting of gneiss with schist. To improve the seismic wave propagation<br />
model it is vital to know about site effects, e.g. the geotectonic properties of the area such as the topof-basement,<br />
vertical tectonic boundaries (faults and basement fracturation) and the geothermal<br />
regime. Therefore, we carried out near surface EM studies to understand the distribution of the<br />
active faulting and the top of basement structure of this particular area.<br />
The RMT (Radiomagnetotelluric) and TEM (Transient electromagnetic) measurements were<br />
carried out on three profiles and thirty sounding. The inverted RMT and TEM data show generally a<br />
four layer model. The layers are indicated as metamorphic and sediment rocks, which are in detail:<br />
marly silty sand with gravel (>> 100 Ωm), marly silty sand with clay (50 - 100 Ωm), sandy clay<br />
(30 – 50 Ωm) and silty sand (10 - 30 Ωm). Due to the high resistivity of the top layer, the skin depths<br />
of the RMT soundings are around 35 m. The TEM data gives detail information of the lower structure<br />
down to a depth of 200 m. According to our analysis, a normal fault next to the Euroseistest could be<br />
located having a strike direction of N 60 E. The joint interpretation of RMT and TEM data proves to<br />
be an effective tool to investigate complex geology structures.<br />
Introduction<br />
This study refers to the Thessaloniki area, which has recently been affected by the 1978<br />
destructive earthquake sequence [Papazachos and Papazachou, 1997]. It has been well established<br />
that the strong ground motion of such a seismic event causes irregularly distributed modification to<br />
the local geology [Tranos and Mountrakis, 1998], [Raptakis, et. al., 1999]. Different geophysical<br />
methods have been applied in this area [Thanassoulas, et.al.,1987], [Raptakis,et.al.,1999],<br />
[Savvaidis,et.al.,1999], [Tranos, et.al, 2003], [Gurk, et.al., 2007], however detail information of fault<br />
structures has not been verified so far. Ambient noise measurements from the area east of the<br />
Euroseistest experiment give strong implication for a complex 3-D tectonic setting. Therefore we<br />
carried out near surface EM studies to understand the distribution of the active faulting and the top of<br />
basement structure of this particular area.<br />
Joint TEM and RMT inversion has been successfully applied to geological and engineering<br />
problems in the past [Tezkan et.al., 1995], [Schwinn, 1999], [Steuer, 2002]. The RMT method has<br />
low penetration depth and therefore gives information of the surface layers, whereas the TEM<br />
method gives detail information of the lower structure of the investigated area. Hence, a joint<br />
interpretation of RMT and TEM data will produce a good resolution of resistivities and thicknesses, in<br />
shallow and in deeper parts of the subsurface.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
164
Test Area and Geological Setting<br />
The investigated sites are located between the Lagada and the Volvi Lake in Northern Greece. The<br />
area is now covered with alluvial deposits as shown in (Fig.1). The local geological mapping in this<br />
area was done by the Geology Institute the University of Cologne [Maith, 2009]. The Mygdonian<br />
system corresponds to various types of sediments that were deposited in a quaternary graben<br />
structure. Due to its hydrothermal activity, the top was intensively covered with travertine/tufa<br />
deposits, now mostly removed by erosion [Jongmans, et.al. 1998]. The local geological map of the<br />
area is shown in Fig.1. We find four major units: metamorphic basement formed by schist and<br />
gneisses in the deeper depths, lower terrace deposit was deposited on the top of the basement<br />
metamorphic, lacustrine and deltaic sediments including conglomerate, gravel and sand. Fans<br />
comprise of soil and silt are located between lower terrace deposit and the Holocene deposit is<br />
composed of sand, silt and clay which are deposited on the top of these sediments. To obtain detail<br />
information of the geology in this area, TEM and RMT data were observed on all various types of the<br />
sediments (Fig.1).<br />
Lower<br />
Terrace Dep.<br />
Fans<br />
Holocene<br />
Deposit<br />
Profile Direction of RMT<br />
Profile 1 of RMT<br />
Gneiss-Schist<br />
Fig. 1. Site Location, geology of the study area, location of RMT station ( ), TEM station ( ) and<br />
location of boreholes ( ) are also displayed in the figure GTEM-1 at the location of TEM-1<br />
sounding.<br />
RMT Measurements and Interpretation<br />
GTEM-1<br />
S-1<br />
Profile 3 of RMT<br />
Profile 2 of RMT<br />
The RMT measurements were carried out on three profiles as indicated in Figure 1. Parameter of<br />
the survey design is listed in Table 1. The RMT of profile 1 is along 1600 m with direction N 0 S,<br />
whereas the direction of profile 2 is located N60 E, which are parallel with transmitters. The direction<br />
of profile 3 in this area is N60W with length of 1400 m, which are perpendicular with transmitter. Due<br />
to partial inaccessibility of the area the profiles could not be set up in a straight line. The RMT-F<br />
system consists out of two magnetic sensors (induction coils, 30 cm length), a preamplifier for the<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
165
electrical channels, two electrical antennae and a recorder. It is a four channel instrument (Ex, Ey,<br />
Hx, Hy) with the capability of estimating the full impedance tensor [Tezkan and Saraev, 2008].<br />
Table1. Parameter of RMT survey design<br />
RMT Length Site distance [m] Orientation Transmitter<br />
Profile [m]<br />
azimuth<br />
Profile 1 1600 25 N0oS N60oE Profile 2 1000 25 N60oE N60oE Profile 3 1400 25 N60oW N60oE A special feature is the way how the electric field is sampled. Instead of grounded dipoles,<br />
the device uses symmetrical electrical dipoles (e.g. two arms of 20 m length) that are capacitively<br />
coupled to the ground. The system records time series in two bands: D2 band (10-100 kHz) and D4<br />
band (100 kHz—1 MHz). The coherency level is important to prevent signals with high noise level,<br />
generally we used for this study a coherency value of 0.8.<br />
N S<br />
Silty Sand<br />
Marly Salty Sand<br />
Fault structure<br />
(????)<br />
Sandy clay<br />
Silty Sand<br />
Fig 2. The RMT 2-D inversion model of profile 2 indicates metamorphic rock (marly salty<br />
sand) in a depth of 0 -5 m. Stations 1 – 16 and stations 25 – 38 show a very low resistivity (
Fault Structure<br />
(???)<br />
Fig.3. The Fault structure direction (black arrow) derived from 2-D RMT inversions<br />
ρ (Ωm)<br />
Fig.4. Correlation between the 1-D inversion of the TEM data (GTEM1) at the reference site and<br />
borehole S-1 information<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
167
In order to calibrate the geophysical measurements with the lithology, TEM data at our<br />
reference site was correlated with borehole data in S-1 (Fig.1). This analysis allows identifying<br />
characteristic strata based on the electrical conductivity. The modeling of RMT data was done by<br />
using 2-D inversion techniques (Fig. 2 and Fig.3). The 2-D inversion was performed with the 2-D<br />
Mackie code [Mackie, et.al, 1997], the 1-D inversion was done with the Marquardt [Cerv and Pek,<br />
1979] and Occam algorithm [Constable, 1978]. Prior to any 2-D modeling, the penetration depth<br />
have been estimated using the ρ* (z*) transformation [Schmucker, 1979]. Penetration depths are<br />
found to be around 35 m.<br />
As an example, we explain the RMT model of profile 2 in detail. This model (Fig.2) shows<br />
high resistivities on the top layer (more than 100 Ω m) which correlates to the marly silty with clay of<br />
the borehole data (Fig.4). For stations 1 -16 and stations 25-38 there is a low resistive structure<br />
beneath the surface layer. Between stations 17-24 there is a high resistive structure with the same<br />
resistivities as the surface layer. This region is interpreted as a fault structure (Fig.2), which is filled<br />
with sedimentary rock (sandy clay).<br />
The data quality and the consistency between ρa and φ could be checked in the apparent<br />
resistivity curves, which are shown in Fig.5. RMT station 8 on profile 2 show that the resistivity<br />
values decrease from highest resistivity (> 80 Ωm) to low resistivity (30 Ωm) with phase values less<br />
than 50 °. The sounding curves of station 8 and 13 (figure 5) and station 16 (figure 5) on profile 2<br />
shows that the phases exceeds value more than 45° indicating a low resistive (sediment) structure<br />
at larger depths, meanwhile the good conductive metamorphic series is indicated by the phase<br />
values lower than 45. Fig. 6 shows a comparison between measured and calculated data at<br />
selected frequencies (79 kHz, 387 kHz and 784 kHz). Measured and synthetic data fit well together<br />
keeping in mind the geological complexity and inhomogeneous in this survey area, however some<br />
misfits which are associated as 3-D effect.<br />
The direction of the fault can be constructed using the 2-D inversion models of profile 1-3 as<br />
shown in Fig.3. The direction of the fault structure is found to be N60E.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
168
N60E<br />
N60E<br />
N60E<br />
Fig.5. Comparison of measured and calculated RMT data of stations 8, 13 and 16 on profile 2<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
169
Fig.6. Comparison of measured and calculated RMT data at frequencies 79, 387 and 784 kHz<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
170
TEM soundings<br />
In order to overcome the resolution problem of TEM soundings for near surface structures,<br />
we carried out RMT soundings at each TEM location to use this additional information in the<br />
following joint inversion. The distance between the TEM stations is depending on the accessibility of<br />
the area. Thirty loop-loop measurements (Tx: 50 m x 50 m and Rx: 10 m x 10 m) were performed<br />
during this campaign. TEM data were obtained using the Zonge NT 20 transmitter and the Zonge<br />
GDP32 receiver.<br />
c<br />
a b<br />
Time (s)<br />
Time (s)<br />
0.01<br />
0.0001<br />
1e-06<br />
1e-08<br />
1e-10<br />
e<br />
Fig.7. The segmentation recording scheme for the in-loop Zonge manufactured TEM system;<br />
low-gain Nano TEM (a), high-gain Nano TEM (b), automatic-gain Zero TEM (c), combination of alltransient<br />
(d), TEM sounding of station 1 was processed with deconvulation and inversion (e).<br />
The circle indicate unreliable decay curves for very late times that we address to be an A/D<br />
conversion problem of the Zonge device.<br />
d<br />
Time (s)<br />
Time (s)<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
171
The system allows to measure TEM data in two distinct modes, according to different<br />
investigation depths: Nano TEM with low and high gain and Zero TEM measurements (Fig.7). Both<br />
modes give unreliable decay curves for very late times that we address to be an A/D conversion<br />
problem of the Zonge device. To cope with this problem we simply changed polarity of the receiver<br />
loop. Consequently a time consuming data acquisition procedure was required.<br />
The first measurement for shallow investigation was done in NANO TEM low gain mode.<br />
This mode uses low currents (about 0.3 A) which is switched of quickly. Moreover, the ramp time<br />
were taken from 0.3 μs to 2.5 ms with manual low-gain settings. Whereas, Nano TEM high-gain uses<br />
a higher transmitter current (about 3 A), which gives a higher penetration depth with a saturation for<br />
earlier time. Nano TEM low-gain and high-gain uses 12 Volt power supply. To obtain higher<br />
penetration depths, Zero TEM measurements were carried out from 31 μs to 6 ms time window with<br />
accompanying automatic-gain settings. It uses a relatively high transmitter current (10 A) at 24 Volts<br />
and 50-55 μs turn-off ramp time.<br />
The penetration depth of TEM methods depends on the time after the transmitters current is<br />
switched off [Parasnis, 1986]. The diffusion process of the transient electromagnetic induction field<br />
can be visualized using the smoke ring concept [Nabighian, 1979]. Due to the low conductivity<br />
(metamorphic rock), the maximal penetration depth (δT) of TEM sounding is approximately 200-300<br />
meter. The next step is a deconvolution of the data that was done by the EADEC algorithm (Lange,<br />
2002) followed by a 1-D inversion with EMUPLUS (e.g. Scholl, 2005) (Fig.7 (e)). The 1-D<br />
interpretation of the TEM data also shows the lateral boundary of the fault structure at stations 12- 17<br />
on profile 1 (Fig.8).<br />
Fig.8. One-dimensional TEM model on profile 1<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
172
Joint Interpretation of RMT and TEM data<br />
A joint interpretation can be used to get a better resolution of the model parameter and this<br />
will enhance the possibility to detect the active fault structure. Generally we conducted RMT<br />
soundings at each TEM site location so that we can estimate a jointly inverted model at each TEM<br />
location (Fig.9).<br />
a).<br />
b).<br />
c).<br />
Table2. The importance parameter of station 10<br />
RMT Imp TEM Imp Joint Imp<br />
ρ1[Ωm] 322.46 0.99 367.98 0.57 247.99 0.99<br />
ρ2[Ωm] 158.17 0.99 367.98 0.88 344.23 0.27<br />
h1[m] 5.56 0.97 9.75 0.97 9.15 0.84<br />
RMS[%] 4.90 3.10 3.20<br />
Table 3. Model resistivies obtained from the<br />
RMT and TEM data for the selected stations<br />
Station RMT TEM Joint<br />
10 322.46 367.98 344.23<br />
11 332.20 273.45 303.06<br />
12 450.73 372.20 616.36<br />
13 562.36 271.88 531.33<br />
14 363.93 420.46 865.49<br />
17 199.91 260.49 208.14<br />
26 4.47 23.48 26.93<br />
25 28.71 15.83 32.98<br />
Fig. 9. (a) One-dimensional model section derived by RMT data (b). One-dimensional model<br />
section of TEM data (c) Joint interpretation of TEM and RMT data. The comparison between<br />
TEM, RMT and Joint on station 10, 11,12,14,17 and 25 are shown with ellipses.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
173
The 2-D model of the RMT data (Fig.2) gives information about the top structure of this area,<br />
but the deeper structures and the fault system cannot be resolved in detail solely based on RMT<br />
soundings. On the other hand, TEM shows a good resolution of deeper structure. The joint<br />
interpretation/inversion process was done using the Marquardt algorithm. The interpretation of the 1-<br />
D RMT model between station 13 – 25 resolves of the top layer (Fig 9 (a)) and the TEM 1-D model<br />
(Fig.9b) shows the bottom of boundary layers of fault structure (Fig.8) in this area more distinctly,<br />
however the top layers of TEM stations 13-25 could not be resolved well because the top layer has<br />
high resistivity value (> 80 Ω m) addressed as metamorphic rock from our reference borehole.<br />
Joint inversion can increase the number of model parameter that includes some of which the<br />
methods cannot resolve separately (Fig.9c). The importance parameters were calculated for the<br />
model resistivities and thickness [Jupp and Vozoff, 1975] [Tezkan et.al., 1995], [Schwinn, 1999],<br />
which are plotted in Table 2. It shows the parameter (ρ1, ρ2, h1) as a result of the individual<br />
single/joint inversions. Values between 0 (unimportant) and 1 (important) are calculated. From this<br />
analysis we conclude that the resistivity values and thicknesses of the fault structure (Fig.9) is well<br />
resolved at station 10, 11,12,14,17 and 25; however, at station 13, it is hardly resolved (importance<br />
0.65). The resistivity value at station 26 is weakly resolved with importance parameter 0.12. The<br />
model resistivities of the second layer (between 5-20 m) obtained from 1-D models of TEM and RMT<br />
data are in good agreement with the joint inversion models (Table 3).<br />
Conclusion<br />
The Inversion of RMT and TEM data indicates a normal fault structure with a strike direction<br />
of approximately N 60 E. The RMT and TEM models generally show four layers that can be<br />
separated according to their resistivity values. We address them as to be metamorphic and sediment<br />
rocks, which are marly silty sand with gravel (>> 100 Ω m), marly silty sand with clay (50 - 100 Ω m),<br />
sandy clay (30 – 50 Ω m) and silty sand (10-30 Ω m). The RMT data was interpreted using 2-D<br />
inversions technique that results in a good fitting between observed and calculated data. Due to the<br />
high resistivity of the top layer, the skin depths of the RMT soundings are around 35 m. The TEM<br />
data gives detail information of the lower structure down to a depth of 200 m.<br />
This study was financed by the Marie Curie project: IGSEA – Integrated Nonseismic<br />
Geophysical Studies to Assess the Site Effect of the EUROSEISTEST Area in Northern Greece –<br />
PERG03-GA-2008-230915 {REF RTD REG/T.2 (2008)D/596232}.<br />
References:<br />
Constable C.Steven, Parker L.R.,Constable G.C., Occam’s Inversion: A practical algorithm for generating<br />
smooth models from electromagnetic sounding data, Geophysics vol.52 no.3 P280-300, March, 1978.<br />
Cerv, V. and Pek, J., Solution of one-dimensional magnetotelluric problem. Stud. Geophys. Geod., 23:349,<br />
1979.<br />
Gurk.M, Savvaidis A.S., Bastani M., Tufa Deposit in the Mygdonian Basin (Northern Greece) studied with<br />
RMT /CSTAMT, VLF & Self-Potential, EMTF Kolloqium, June, 2007.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
174
Jongmans, D., Patilakis,K., Demanet, D., Raptakis, D., Hor-rent, C., Tsokas, G., Lontzetidis, K., Reipl, J.,<br />
EURO-SEISTEST: Determination of The geological Stucture of The Volvi Basin and Validation of The Basin<br />
Respone. Bull. Seismoll.Soc. Am. 88, 473 – 487, 1998.<br />
Jupp, D.L.B. and K. Vozoff, stable iterative methods for the inversion of geophysical data,<br />
Geophys.J.R.astr.soc., 42,957-976, 1975.<br />
Lange J., Joint Inversion von central Loop TEM und Long TEM Transient am Beispiel von Messdaten aus<br />
Israel , Diplomarbeit,Universität zu Köln, Institut fur Geophysik und Meteorologie (2003), 16,31,51, 2002.<br />
Maith, I., Erläuterungen zur Kartierung der kristallinen Randbereiche der paläo-mesozoischen Abfolge des<br />
quartären-Makedonischen Massivs und der quartären Beckenfüllung im Bereich von Stivos (Mygdonisches<br />
Becken, NE-Greichenland), Diplomkartierung, der Mathematisch-Naturwissenschaftlichen Fakultät der<br />
Universität zu Köln, 2009.<br />
Mackie R., Rieven S., Rodi W., User Manual and Software Documentation for two-dimensional of<br />
magnetotelluric data, Cambridge Massachuesetts, USA, 1997.<br />
Nabighian, M.N., Quasy-static transient response of a conducting half space: An approximate representation.<br />
Geophysichs, 44:1700-1705, 1979.<br />
Papazachos,B.C., Papazachou, C., The Earth quake of Greece, Ziti Publications , Thessaloniki, 1997.<br />
Parasnis, D. S., Principle of applied geophysics (4 th ed.): Chapman and Hall, London, 402p, 1986.<br />
Raptakis D., F.J. Chavez-Garcia, Makra K., Pitilakis K., Site efecst at Eurositetest – I. Determination of the<br />
valley structure and confrontation of observations with 1-D analysis, Soil dynamic and earthquake<br />
Engineering, 19 1 -22, 1999.<br />
Savvaididis , A., Pedersen, L.B., Tsokas, G.N.,Dawes, G.J., Structure of the Mygdonian Basin (N.Greece)<br />
inferred from MT and gravity data, Tectonophysics, 317, 171-1886, 2000.<br />
Scholl C., The influence of multidimensional structures on the interpretation of LOTEM data with onedimensional<br />
models and the application to data from Israel, Inaugural Dissertation, Institute Geophysik und<br />
Meteorologie Uni Zu Koln, 2005.<br />
Schwinn W., 1-D Joint Inversion Radiomagnetotellurik (RMT) und Transientelektromagnetik Daten (TEM):eine<br />
Anwendung zur Grundwasser exploration in Grundfor, Danemark, Diplomarbeit, Institute Geophysik und<br />
Meteorologie Uni Zu Koln, Juli, 1999.<br />
Steuer, A., Kombinierte Auswertung von Messungen mit Transient-Elektromagnetik und Radio Magnetotellurik<br />
zur Grundwassererkundung im Bechen von Quarzazate (Maroko), Diplomarbeit, Institute Geophysik und<br />
Meteorologie Uni Zu Koln, June, 2002.<br />
Schmucker, U., Erdmagnetische Variationen und die elektrische Leitfahigkeit in tieferen Schichten der Erde.<br />
Sitzungsber. Mitt. Braunschw Wiss. Ges. Sonderh., 4:45-102, 1979.<br />
Tranos, M.D., Mountrakis D.M., Neotectonic joints of the northern Greece; their significance on the<br />
understanding of the active deformation. Bulletin of Geological Soceitey of Greece 32,209-219, 1998.<br />
Thanassoulas, C., Tselentis, G-A., Traganos, G., A preliminary resistivity investigation (VES) of the Lagada<br />
hot springs area in northern Greece, Gheothermics, Vol. 16, NO. 3, pp. 227-238, 1978.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
175
Tranos D.M., Papadimitriou E.E., Kilias A.A.,Thessaloniki-Gerakarou Fault Zone (T<strong>GFZ</strong>): the western<br />
extension of the 1978 Thessaloniki earthquake fault (Northern Greece) and seismic hazard assessment,<br />
Journal of structural geology, 25 2109-2123, 2003.<br />
Tezkan, B., M. Goldman, S. Greinwald, A.Hordt, I.Muller, F.M. Neubauer and H.G. Zacher, A joint application<br />
of radio magnetotellurics and transient electromagnetic to the investigation of a waste deposit in Cologne<br />
(Germany), Applied Geophysics, 34, 199-212,1995.<br />
Tezkan, B., and Saraev, A., A new broadbrand Radiomagnetotelluric instrument: Application to near surface<br />
investigation, Near surface Geophysichs, 6.243-250, 2008.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
176
Die deutsche Nordseeküste im Fokus von aeroelektromagnetischen<br />
Untersuchungen<br />
Teilgebiete Langeoog mit Wattenmeer und Elbemündung<br />
Gerlinde Schaumann 1 , Annika Steuer 2 , Bernhard Siemon 2 , Helga Wiederhold 1 & Franz Binot 1<br />
1. Einleitung<br />
1 Leibniz-Institut für Angewandte Geophysik (LIAG), Hannover<br />
2 Bundesanstalt für Geowissenschaften und Rohstoffe (BGR), Hannover<br />
gerlinde.schaumann@liag-hannover.de<br />
Hubschrauberelektromagnetische (HEM) Untersuchungen bieten ein großes Potential für die flächendeckende<br />
Kartierung der Sedimente der ersten hundert Meter des Untergrundes (Siemon et al., 2009).<br />
Sie sind für hydrogeologische Fragestellungen von großer Bedeutung, da mit Hilfe des spezifischen<br />
Widerstands die Verteilung sandiger und tonhaltiger Sedimente im Untergrund sowie Versalzungszonen<br />
und Süßwasserbereiche ermittelt werden können.<br />
In den Jahren 2008 und 2009 wurden in Kooperation von LIAG und BGR im Rahmen des LIAG-<br />
Projektes zur „Flächenhaften Befliegung“ und des „D-AERO“-Projektes der BGR gemeinsam aerogeophysikalische<br />
Erkundungen zu Salz-/Süßwasserfragestellungen in insgesamt fünf Messgebieten<br />
im Bereich der deutschen Nordseeküste durchgeführt (Wiederhold et al. 2008, Steuer et al. 2009).<br />
Das Messgebiet Langeoog umfaßt die Ostfriesischen Inseln Langeoog und Spiekeroog und das angrenzende<br />
Wattenmeer, das Messgebiet Glückstadt umfaßt den Bereich der Elbemündung nordwestlich<br />
von Hamburg (Abb. 1).<br />
Abbildung 1: Lage der Messgebiete Langeoog und Glückstadt (nordwestlich von Hamburg).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
177
Für die Insel Langeoog waren die Lage und die Ausdehnung der Süßwasserlinse zu ermitteln. Im küstennahen<br />
Bereich erhoffte man im Wattenmeer mutmaßliche Süßwasseraustritte zu finden. Im Bereich<br />
der Elbemündung sollten die grundwasserführenden Schichten und mögliche Versalzungszonen<br />
kartiert werden.<br />
Die Daten dienen als Grundlage für die Planung und Arbeit in vielfältigen ökonomischen und ökologischen<br />
Bereichen, wie z.B. Raumplanungen und der Entwicklung von Wassernutzungs- und Wasserschutzkonzepten.<br />
Sie werden über das Fachinformationssystem (FIS) Geophysik des LIAG nutzbar<br />
sein.<br />
Das eingesetzte Messsystem wird bei Steuer et al. (2010) beschrieben. Die Karten des scheinbaren<br />
spezifischen Widerstands a und der Schwerpunktstiefe z* (Siemon, 2009) für die verschiedenen<br />
Messfrequenzen geben einen ersten Überblick über die Leitfähigkeitsstrukturen in den Messgebieten.<br />
Die Erkundungstiefe nimmt dabei mit abnehmender Frequenz und Leitfähigkeit zu. Die HEM-<br />
Messungen liefern die Verteilung der elektrischen Leitfähigkeit bis maximal 150 m Tiefe. Dabei können<br />
Salzwasser und Süßwasser oder ton- und sandhaltige Sedimente unterschieden werden. Hier dargestellt<br />
sind erste Ergebnisse in Form von Karten auf dem Bearbeitungsstand nach der Grundprozessierung,<br />
bei der Korrekturfaktoren und Filtereinstellungen für jeden Flug gleich angesetzt werden. Die<br />
darauf folgende Feinprozessierung behandelt jeden Flug und jede Frequenz individuell. Die Prozessierung<br />
der Daten beinhaltet: Filterung, Sprungkorrektur, Nullniveaukorrektur, Feinjustierung der Kalibrierfaktoren<br />
und Berechnung der Halbraumparameter.<br />
2. Messgebiet Langeoog<br />
Langeoog und Spiekeroog sind zwei der Ostfriesischen Inseln im Wattenmeer der Nordsee, die sich<br />
wenige Kilometer vor der Küste befinden. Sie bestehen aus quartären Sedimenten wie Sanden, Tonen<br />
und Schluffen (Abb. 3). Durch Versickerung der Niederschläge in den Dünengürteln werden dort<br />
Grundwasserreservoire mit Süßwasser aufgefüllt, die die Versorgung der Inseln mit Frischwasser sicherstellen.<br />
Weil Süßwasser ein geringeres spezifisches Gewicht als das versalzte Grundwasser hat,<br />
schwimmt es als „Linse“ auf dem Salzwasser. Verändert sich der Wasserstand des dichteren Salzwassers,<br />
ändert sich das Druckniveau und entsprechend auch Stand und Ausdehnung der Süßwasserlinse.<br />
Durch Stürme können Salzwassereinbrüche bis an den Rand des inneren Dünengürtels der<br />
Inseln gelangen und dort versickern. Solche Ereignisse, aber auch der erhöhte Wasserbedarf für den<br />
Tourismus in den Sommermonaten gefährden das Süßwasserreservoir. Noch erfolgt die gesamte<br />
Trinkwasserversorgung mit inseleigenem Grundwasser, welches in den Wintermonaten allein durch<br />
Regenfälle wieder aufgefüllt wird. Unterhalb wasserundurchlässiger Bodenschichten kann Süßwasser<br />
darüber hinaus vom Festland aus unterschiedlich weit in das Wattenmeer vordringen, um dann unter<br />
gewissen Bedingungen an die Oberfläche zu treten. Die Befliegungen wurden bei Niedrigwasser<br />
durchgeführt, um vergleichbare Bedingungen für jeden Flug zu haben und insbesondere auch, um<br />
den Untergrund des Wattenmeeres ohne die bei Flut vorhandene Meerwasserschicht besser erkun-<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
178
den zu können. Aufgrund der Berücksichtigung von Naturschutzauflagen konnte die aerogeophysikalische<br />
Erkundung nur in den Wintermonaten durchgeführt werden. Das Messgebiet umfasst die gesamte<br />
Insel Langeoog, den westlichen Teil der Insel Spiekeroog und das Wattenmeer zwischen Langeoog<br />
und dem Festland (Abb. 2). Einen Überblick über die Messkampagne gibt Tabelle 1.<br />
Abbildung 2: Insel Spiekeroog mit Wattenmeer während der Befliegung im März 2009 und westlicher Teil von<br />
Langeoog mit der Dünenlandschaft, unter der sich Süßwasserlinsen verbergen (Fotos: W. Voß).<br />
Tabelle 1:<br />
Größe des Messgebietes: 259 km²<br />
Gesamtprofillänge: 1200 km<br />
Linienabstand: 67 Mess-Linien mit 250 m (N-S), 7 Kontroll-Linien mit 2000 m (W-O)<br />
Messzeitraum: Februar und November 2008, Februar und März 2009; nur bei Niedrigwasser<br />
Abbildung 3: Geologische Karte des Messgebietes aus dem FIS Geophysik des LIAG.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
179
Abbildung 4 zeigt die vorläufigen Karten des scheinbaren spezifischen Widerstands (mit a bis maxi-<br />
mal 2000 m) der verschiedenen Messfrequenzen (133400 Hz, 41520 Hz, 8390 Hz, 5345 Hz, 1823 Hz und<br />
387 Hz) für das Messgebiet. Jeder Frequenz wird dabei zu den einzelnen Messpunkten eine Schwerpunktstiefe<br />
z* zugeordnet, die in Abbildung 5 dargestellt ist. Für die höchsten Frequenzen sind dies<br />
einige wenige Meter unter der Erdoberfläche, die tiefsten Frequenzen erreichen eine Erkundungstiefe<br />
bis etwa 100 m. Diese Erkundungstiefen sind neben der Frequenz auch von den spezifischen Widerständen<br />
der Untergrundstrukturen abhängig und können somit von Messpunkt zu Messpunkt variieren.<br />
Die Abfolge einer Auswahl von Karten des scheinbaren spezifischen Widerstands in einer 3D-<br />
Darstellung gibt daher eine Abbildung der Leitfähigkeitsstrukturen mit Bezug zur Tiefe nur bedingt<br />
wieder (Abb. 7). Die tiefste Frequenz bildet die Basis der Süßwasserlinse ab.<br />
Deutlich sind die den Süßwasserlinsen zugeordneten Strukturen auf der Insel und die mit salzhaltigem<br />
Nordseewasser gefüllten Priele zu erkennen. Im Bereich vor der Festlandküste kann man Gebiete mit<br />
für Süßwasser typischen Werten der scheinbaren Widerstände erkennen. Hier ist zu klären, ob dies<br />
Austritten von Süßwasser im Watt entspricht.<br />
Abbildung 4: Vorläufige Karten des scheinbaren spezifischen Widerstands a für die Messfrequenzen<br />
133400 Hz, 41520 Hz, 8390 Hz, 5345 Hz, 1823 Hz und 387 Hz (a bis maximal 2000 m).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
180
Abbildung 5: Vorläufige Karten der Schwerpunktstiefe z* für die Messfrequenzen133400 Hz, 41520 Hz,<br />
8390 Hz, 5345 Hz, 1823 Hz und 387 Hz.<br />
Benutzt man einen anderen Farbkeil für die Darstellung der scheinbaren spezifischen Widerstände<br />
(mit a bis maximal 200 m), kann man beispielsweise die Priele noch viel deutlicher erkennen, siehe<br />
Abbildung 6. Auch die mutmaßlichen Austritte von Süßwasser im Bereich vor der Festlandküste sind<br />
wesentlich besser aufgelöst.<br />
Abbildung 6: Vorläufige Karten des scheinbaren spezifischen Widerstands a für drei ausgewählte Frequenzen<br />
(41520 Hz, 8390 Hz und 387 Hz), dargestellt mit einem anderen Farbkeil (a bis maximal<br />
200 m).<br />
In Abbildung 8 wird die topographische Karte mittels des digitalen Höhenmodells (digital elevation model,<br />
DEM) aus dem FIS Geophysik des LIAG gezeigt. Dabei wurden die Daten des Höhenmodells auf<br />
die Fluglinien übertragen. Die topographische Karte kann auch aus GPS- und Laseraltimeterdaten des<br />
Hubschraubermesssystems abgeleitet werden.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
181
Abbildung 7: 3D-Darstellung der Karten des scheinbaren spezifischen Widerstands a für die Frequenzen<br />
41520 Hz, 8350 Hz und 387 Hz. Bild aus Google-Earth, September 2009.<br />
Abbildung 8: Topographische Karte mittels des digitalen Höhenmodells (digital elevation model, DEM) aus<br />
dem FIS Geophysik des LIAG. Dabei wurden die Daten des Höhenmodells auf die Fluglinien<br />
übertragen.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
182
3. Messgebiet Glückstadt<br />
Das Messgebiet Glückstadt umfasst große Teile des Elbemündungsbereichs (Abb. 9). Teile der Ergebnisse<br />
der HEM-Befliegung werden in das Projekt KLIMZUG-NORD (www.klimzug-nord.de) einfließen,<br />
bei dem unter anderem die Auswirkungen des Klimawandels auf das Ästuar der Elbe erforscht<br />
werden. Es ist davon auszugehen, dass sich der Elbwasserstand in diesem Gebiet erhöht, was eine<br />
Intrusion von Brackwasser in den Süßwasseraquifer nach sich ziehen würde. Mit hydraulischen Strömungsmodellen<br />
können solche Fragestellungen untersucht werden. Bereits vorhandene geologische<br />
Informationen sollen mit Hilfe von elektrischen Leitfähigkeitsmodellen der HEM flächenhaft zu einem<br />
geologischen Strukturmodell ergänzt werden. Dieses wird mit hydraulischen Parametern belegt und<br />
dient damit als Grundlage für eine hydraulische Modellierung.<br />
10 km<br />
Abbildung 9: Flugplan für das Messgebiet Glückstadt, welches große Teile des Elbemündungsbereichs<br />
umfasst.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
183
Einen Überblick über den Umfang der Messkampagne gibt Tabelle 2.<br />
Tabelle 2:<br />
Größe des Messgebietes: 1949 km²<br />
Gesamtprofillänge: 8500 km<br />
Linienabstand: 209 Mess-Linien mit 250 m (W-O), 19 Kontroll-Linien mit 2500 m (N-S)<br />
Messzeitraum: Juli und Oktober 2008, März-Mai 2009<br />
Aus den GPS- und den Laseraltimeterdaten des Hubschraubermesssystems kann ein digitales Höhenmodell<br />
abgeleitet werden (Abb. 10). Hier wird besonders der Verlauf der Geest (Höhenzug bestehend<br />
aus pleistozänen Sandablagerungen von Moränen) und der Marsch (flaches holozänes<br />
Schwemmland) deutlich.<br />
10 km<br />
Abbildung 10: Digitales Höhenmodell des Elbemündungsgebietes, abgeleitet aus GPS- und Laseraltimeterdaten<br />
des Hubschraubermesssystems.<br />
Erste Ergebnisse in Form von vorläufigen Karten des scheinbaren spezifischen Widerstands und der<br />
Schwerpunktstiefe sind in Abbildung 11 dargestellt. Die Datenlücken sind teilweise auf Hochspannungsleitungen<br />
und entlang der Elbe auf Radarstationen bzw. Sperrzonen um die Atomkraftwerke<br />
herum zurückzuführen.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
184
Elbe<br />
3 2<br />
1<br />
3<br />
4<br />
41 520 Hz<br />
6<br />
8 390 Hz<br />
1 823 Hz<br />
387 Hz<br />
5<br />
2<br />
a<br />
[m]<br />
10 km<br />
a<br />
[m]<br />
a<br />
[m]<br />
a<br />
[m]<br />
a<br />
[m]<br />
41 520 Hz<br />
8 390 Hz<br />
1 823 Hz<br />
Abbildung 11: Vorläufige Karten des scheinbaren spezifischen Widerstands a und der Schwerpunktstiefe z*<br />
für vier ausgewählte Messfrequenzen.<br />
387 Hz<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
185<br />
z*<br />
[m]<br />
z*<br />
[m]<br />
z*<br />
[m]<br />
z*<br />
[m]<br />
10 km<br />
z*<br />
[m]
Mit Hilfe der geologische Karte vom Mündungsbereich der Elbe (Abb. 12a) können z.B. in der Karte<br />
des scheinbaren spezifischen Widerstands zu der Frequenz 41 kHz (Abb. 12b) folgende Leitfähigkeitsstrukturen<br />
identifiziert werden: 1. Brackwasser der Elbe, 2. Perimarine Ablagerungen: Ton, schluffig,<br />
3. Brackische Ablagerungen: Ton bis Schluff, feinsandig, 4. Moor: Torfe, 5. Rotschlammdeponie:<br />
Eisen- und Titanoxide, 6. Glazifluviatile Ablagerungen: Sand und Kies.<br />
Elbe<br />
4. Zusammenfassung und Ausblick<br />
3<br />
2<br />
3<br />
4<br />
a) b)<br />
2<br />
41 520 Hz<br />
Abbildung 12: a) Geologische Karte und b) Karte des scheinbaren spezifischen Widerstands der Frequenz<br />
41 kHz. Im Mündungsbereich der Elbe können folgende Leitfähigkeitsstrukturen identifiziert<br />
werden: 1. Brackwasser der Elbe, 2. Perimarine Ablagerungen: Ton, schluffig, 3. Brackische<br />
Ablagerungen: Ton bis Schluff, feinsandig, 4. Moor: Torfe, 5. Rotschlammdeponie: Eisen-<br />
und Titanoxide, 6. Glazifluviatile Ablagerungen: Sand und Kies.<br />
In beiden Messgebieten wurde die HEM zur Erkundung von Grundwasserstrukturen erfolgreich eingesetzt.<br />
Die hier gezeigten Karten dokumentieren die vorläufigen Ergebnisse. Nach Abschluss der Prozessierung<br />
werden die HEM-Daten in Schichtmodelle des spezifischen Widerstands invertiert.<br />
Auf den Ostfriesischen Inseln Langeoog und Spiekeroog konnten die Lage und Ausdehnung der<br />
Süßwasserlinsen erfasst werden. Daneben zeigten sich insbesondere in den Daten der tiefsten Messfrequenz<br />
Bereiche veränderter scheinbarer spezifischer Widerstände im Wattenmeer vor der Festlandküste,<br />
die die Vermutung nahe legen, dass hier Süßwasser austritt.<br />
Im Gebiet Glückstadt werden als Fragestellung mögliche Intrusionen von Brackwasser in den Süßwasseraquifer<br />
untersucht. Die bereits vorhandenen geologischen Informationen werden mit Hilfe der<br />
HEM-Modelle flächenhaft zu einem geologischen Strukturmodell ergänzt. Nach einer Belegung mit<br />
hydraulischen Parametern kann es damit als Grundlage für eine Interpretation mit Strömungsmodellen<br />
dienen.<br />
Elbe<br />
3 2<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
186<br />
1<br />
3<br />
4<br />
6<br />
5<br />
2<br />
a<br />
[Ohm*m]<br />
a<br />
[m]<br />
10 km
Danksagung<br />
Die hier vorgestellten Teilgebiete des BGR/LIAG-Projektes wurden in Zusammenarbeit mit dem Institut<br />
für Chemie und Biologie des Meeres (ICBM), der Universität Oldenburg sowie der Behörde für<br />
Stadtentwicklung und Umwelt Hamburg durchgeführt.<br />
Referenzen<br />
Siemon, B. [2009] Electromagnetic methods – frequency domain: Airborne techniques, In: Kirsch, R.<br />
(Ed.), Groundwater Geophysics – A Tool for Hydrogeology. 2nd Edition, Springer-Verlag, Berlin,<br />
Heidelberg, 155-170.<br />
Siemon, B., Christiansen, A.V. & Auken, E. [2009] A review of helicopter-borne electromagnetic<br />
methods for groundwater exploration. Near Surface Geophysics, 7, 629-646.<br />
Steuer, A., Siemon, B., Schaumann, G., Wiederhold, H., Meyer, U., Pielawa, J., Binot, F. & Kühne, K.<br />
[2009] The German North Sea Coast in Focus of Airborne Geophysical Investigations, AGU Fall<br />
Meeting 2009, San Francisco, USA.<br />
Steuer, A., Siemon, B. & Grinat, M. [2010] The German North Sea Coast in Focus of Airborne<br />
Electromagnetic Investigations: The Freshwater Lenses of Borkum, EMTF, this volume.<br />
Wiederhold, H., Binot, F., Kühne, K., Meyer, U., Siemon, B. & Steuer, A. [2008] Airborne geophysical<br />
investigation of the German North Sea Coastal Area, 20th Salt Water Intrusion Meeting 2008, Naples,<br />
USA.<br />
http://www.liag-hannover.de/forschungsschwerpunkte/grundwassersysteme-hydrogeophysik/salz-<br />
suesswassersysteme/flaechenhafte-befliegung.html<br />
http://www.bgr.bund.de/cln_101/nn_328750/DE/Themen/GG__Geophysik/Aerogeophysik/Projektbeitr<br />
aege/D-AERO/deutschlandweite__aerogeophysik__befliegung__D-AERO.html<br />
http://www.fis-geophysik.de/<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
187
The German North Sea Coast in Focus of Airborne Electromagnetic<br />
Investigations: The Freshwater Lenses of Borkum<br />
Annika Steuer 1 , Bernhard Siemon 1 and Michael Grinat 2<br />
1<br />
Federal Institute for Geosciences and Natural Resources (BGR), Hannover, Germany<br />
2<br />
Leibniz Institute for Applied Geophysics (LIAG), Hannover, Germany<br />
annika.steuer@bgr.de<br />
Abstract<br />
Helicopter-borne electromagnetic (HEM) measurements were conducted over the North Sea island of<br />
Borkum to determine the size of the freshwater lenses. Additionally, geoelectrical measurements were<br />
carried out at locations where data of 13–17 years old Schlumberger soundings exist. HEM<br />
successfully revealed the lateral extension as well as the thickness of the freshwater lenses of the<br />
island of Borkum. At several locations the depth of the freshwater/saltwater boundary determined by<br />
HEM was confirmed by the results of the Schlumberger soundings. No significant changes of the<br />
freshwater/saltwater boundary were found at the locations where old and new Schlumberger<br />
soundings exist. Regarding the results of direct-push and borehole measurements, the HEM related<br />
resistivity structure could be determined more precisely and additional to the freshwater/saltwater<br />
boundary thin clay layers were revealed. Based on the HEM inversion results the volume of the<br />
freshwater resource was estimated.<br />
North Sea<br />
Germany<br />
Figure 1: The location of the island of Borkum in the North Sea<br />
and a Google-Earth map with the flight-lines.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
188
Introduction<br />
Since 2008, the German North Sea Coast has been in the focus of airborne geophysical investigations<br />
carried out by the Federal Institute for Geosciences and Natural Resources (BGR) and the Leibniz<br />
Institute for Applied Geophysics (LIAG). Especially electromagnetics is the most versatile of the<br />
airborne geophysical methods and widely applied in hydrogeological investigations as the<br />
measurements respond to both lithologic and water-chemistry variations. The applications comprise<br />
geologic mapping and aquifer structure, delineation of soil and groundwater salinization, saltwater<br />
intrusion into coastal aquifers etc.<br />
Here, we present the HEM results for the North Sea island of Borkum, which was the first of the five<br />
HEM project areas investigated in 2008-2009, and compare them with DC Schlumberger soundings,<br />
lithological logs and direct-push measurements. These data will be used for a hydraulic modelling of<br />
the freshwater lenses of the island of Borkum in the CLIWAT project (http://cliwat.eu, EU Interreg IVB<br />
North Sea Region). Additionally we show an estimation of the volume of the freshwater lenses<br />
calculated by Siemon et al. (2009b).<br />
Other HEM results for the Wadden Sea, the North Sea islands of Langeoog and Spiekeroog, and the<br />
Elbe estuary are shown by Schaumann et al. (2010) in this volume. Wiederhold et al. (2009) presented<br />
the results of two SkyTEM surveys at the North Sea island of Föhr and the saltdome Bad Segeberg,<br />
which were conducted in 2008 by SkyTEM Aps by order of LIAG.<br />
HEM Survey<br />
The Borkum survey covers an area of 88 km² and was flown within two days in March 2008. The flightline<br />
spacing was 250 m and the tie-line spacing was 500 m, totalling to about 412 line-kilometres<br />
(Figure 1). The BGR helicopter-borne geophysical system simultaneously uses frequency-domain EM,<br />
magnetics and radiometrics (Figure 2).<br />
The HEM system, a RESOLVE bird manufactured by Fugro Airborne Surveys, operates at six<br />
frequencies ranging from 386 Hz to 133 kHz.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
189
Figure 2: BGR helicopter-borne geophysical system:<br />
Electromagnetics:<br />
RESOLVE (Fugro Airborne Surveys)<br />
Frequencies: 386, 1 823, 5 495, 8 338,<br />
41 430 and 133 300 Hz<br />
Coil separations: ~8 m<br />
Coil orientation: 5× horizontal coplanar<br />
1× vertical coaxial<br />
Investigation depth: 1–70 m (at Borkum)<br />
Magnetics: Cs magnetometer<br />
Radiometrics: 256 channel gamma-ray<br />
spectrometer<br />
Navigation/Positioning:<br />
GPS / radar and laser altimeter<br />
Bird altitude: ~ 34 m<br />
Flight speed: ~140 km/h<br />
Sampling distance: ~4 m<br />
The investigation depth increases with decreasing frequency. The EM fields of the lowest frequency<br />
penetrate the freshwater lenses nearly completely as seen in the apparent resistivity maps (Figure 3).<br />
Here, higher resistivities are represented by green and blue colours (freshwater) and lower resistivities<br />
by red colours (saltwater). The apparent resistivity maps, which are based on the half-space<br />
approximation, provide a first insight into the subsurface conductivity distribution.<br />
133 kHz 41 kHz 1.8 kHz 0.4 kHz<br />
Figure 3: Apparent resistivity maps: The investigation depth increases with decreasing frequency. The<br />
freshwater lenses – represented by higher resistivities (green & blue) – are nearly completely<br />
penetrated by the EM fields of lowest frequency.<br />
The HEM data were also inverted using Marquardt-Levenberg 1-D inversion technique based on a<br />
layered half-space model (Siemon et al., 2009a). Starting models required were derived automatically<br />
from the apparent resistivities vs. centroid depth sounding curves. Resistivity maps at selected depths<br />
and stitched together vertical resistivity sections were derived from these 1-D inversion models. The<br />
lateral extensions as well as the thicknesses of the freshwater lenses were mapped successfully with<br />
HEM, as shown in the resistivity maps and the vertical resistivity sections of Figure 4.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
190
Figure 4: Resistivity maps at 5 and 40 m bsl and vertical resistivity sections along the tie lines T13.9<br />
and T6.9.<br />
The first layer represents the resistive dry dune sands above the groundwater table (dark blue) and<br />
the bottom layer the conductive saltwater-filled sediments (red) below the freshwater lenses (blue).<br />
The two layers in-between are associated with sediments filled with fresh or brackish water. The<br />
freshwater/saltwater boundary was detected down to about 60 m depth.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
191
DC survey: Comparison of measurements 16 years apart<br />
In 2008, 36 Schlumberger soundings were<br />
carried out at sites already measured<br />
between 1991 and 1995 (Worzyk, 1995a<br />
and 1995b).<br />
All the sounding curves at Borkum have a<br />
Q-type shape with three or more layers (e.g.<br />
Figure 5). Higher resistivities are due to dry<br />
and freshwater filled sediments, whereas the<br />
lower resistivities at the end of the sounding<br />
curves are due to saltwater-filled sediments<br />
underlying the freshwater lenses.<br />
apparent resistivity [m]<br />
10000<br />
1000<br />
1 10 100 1000<br />
current electrode separation AB / 2 [m]<br />
In 1992, the freshwater filled sediments<br />
were associated with resistivities between<br />
Figure 5: DC sounding curve of site GTS 60.<br />
70 and 110 m in the water catchment area<br />
Ostland; these resistivities were almost reproduced in 2008. The underlying saltwater-filled sediments<br />
showed resistivities below 10 m, but these values were not determined exactly. Taking equivalent<br />
models into consideration the thickness of the freshwater lens can be determined with an accuracy of<br />
about 6–8 m.<br />
No significant changes of the freshwater/saltwater boundary were found at the locations where old and<br />
new Schlumberger soundings exist.<br />
100<br />
10<br />
GTS 60 (5/1992): Data + model (-)<br />
GTS 60 (10/2008): Data + model (-)<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
192
Combined analysis<br />
At several locations the depth of the freshwater/saltwater boundary determined by HEM was<br />
confirmed by the results of the Schlumberger soundings (e.g., Figure 6).<br />
a)<br />
c)<br />
BR40<br />
T 7.9<br />
NE SW<br />
HEM T 7.9<br />
GTS 60<br />
BR 40<br />
HF510S3L5 D2 SF075 200<br />
The elevation map in Figure 6b shows the location of the Schlumberger sounding GTS 60 (blue circle),<br />
the drilling BR 40 (orange circle) and a part of flight-line T 7.9 (red line). The flight-line crosses a dune<br />
and the Schlumberger sounding is located about 80 m apart from the flight-line in a dune valley. The<br />
vertical resistivity section of this part in Figure 6c shows that the saltwater/freshwater boundary was<br />
determined by the HEM 1-D inversions at about 60 m depth. This is confirmed by the 1-D inversion<br />
result of the Schlumberger sounding of 2008 which has not changed significantly since 1992 (Figure<br />
6d).<br />
Regarding the results of direct-push (Winter, 2008) and borehole measurements, the HEM related<br />
resistivity structure could be determined more precisely. A more detailed 6-layer approach revealed<br />
additionally to the freshwater/saltwater boundary thin conductive layers at about 10 m depth, not seen<br />
in the 4-layer models of the standard inversion procedure (Figure 7b). Partly, they were identified as<br />
clay layers by lithological logs; and also the result of a direct-push measurement shows an increasing<br />
conductivity at 10 m depth (Figure 7c). The depth of the freshwater/saltwater boundary remains nearly<br />
constant and has a layer with little higher resistivity above, which could be interpreted as transition<br />
b)<br />
d)<br />
1<br />
2<br />
3<br />
4<br />
5<br />
6<br />
7<br />
0<br />
8<br />
depth [m]<br />
HEM<br />
GTS 60 (10/2008) GTS 60 (5/1992) BR 40<br />
4560 2578 m<br />
Br unnen 40<br />
2181 m<br />
Feinsand,<br />
z.T. schluffig<br />
76<br />
Filter I<br />
69<br />
1.05<br />
92<br />
1.1<br />
87<br />
1.1*<br />
M itt elsand<br />
Mittelsand,<br />
fein- bis<br />
grobsandig<br />
Geschiebelehm,<br />
Schluff, tonig<br />
Filter II<br />
ET<br />
Figure 6: a) Resistivity map (5 mbsl) with the location of a part of flight-line T7.9 (red line), the lithological log<br />
BR40 (orange circle) and the DC site GTS 60 (blue circle). b) Digital elevation model revealed<br />
from laser scan data of NLWKN. c) Vertical resistivity section T7.9 as result of a 5-layer 1-D<br />
inversion (every 5 th model is shown). d) Comparison of 1-D inversion results of HEM and DC<br />
reveals the freshwater/saltwater boundary at a similar depth of about 60 mbsl.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
193<br />
BR40<br />
T 7.9
zone of brackish water. The data misfit q [%] of the 6-layer inversion is smaller than the misfit of the 4layer<br />
inversion, especially in the south-eastern part of the section.<br />
a)<br />
b)<br />
c)<br />
NW HEM L16.1<br />
SE<br />
DP 06<br />
DirectPush 06<br />
GTS 8<br />
m/min mS/m Ohmm<br />
10-12 m<br />
L16.1<br />
WWBO 1026 BORK II 811<br />
H_F5_10_S3_L4_D2_SF100_<br />
0<br />
10<br />
20<br />
30<br />
40<br />
50<br />
60<br />
70<br />
78<br />
37<br />
2.0<br />
80<br />
depth [m]<br />
HEM GTS 8<br />
9890<br />
80<br />
18<br />
103<br />
9.3<br />
2.0<br />
9883<br />
NW HEM L16.1<br />
SE<br />
Figure 7: a) Apparent resistivity map (41 kHz) with the location of the vertical resistivity section<br />
L16.1, the lithological logs and the direct-push measurement. b) Comparison of HEM 1-D<br />
inversion results (4- and 6-layer models) along L16.1 with c) lithological logs (WWBO 1016 and<br />
BORK II 811) (LBEG), direct-push results (DP 06) (Winter, 2008) and DC 1-D inversion results<br />
(GTS8).<br />
118<br />
61<br />
2.3<br />
DP 06<br />
1035 m<br />
GTS 8<br />
WWBO 1026<br />
WWBO 1026 BORK II 811<br />
> 8.9 m clay<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
194<br />
H_F5_10_S3_L6_D2_SF100_200<br />
BORK II 811<br />
6.3-9.5 m clay<br />
9.5-12 m peat
Estimation of the aquifer thickness<br />
The aquifer thickness was estimated using Archie’s law with the following approach: The resistivity of<br />
freshwater is assumed to be higher than 5.6 m (180 mS/m) and thereby the resistivity of freshwater<br />
saturated sediment with 33% porosity and a formation factor of 4 should be above 23 m (Siemon et<br />
al., 2009b). Based on that, the accumulated layer thickness with resistivities of 30–500 m is defined<br />
as aquifer thickness and is shown in Figure 8. From that, the freshwater volume was estimated to 175<br />
x 10 6 m³.<br />
Conclusions<br />
Figure 8: Aquifer thickness derived from DC and HEM.<br />
(1991/92)<br />
(based on 4-layer model)<br />
At several locations the depth of the freshwater/saltwater boundary as derived by the HEM inversion<br />
was confirmed by Schlumberger soundings. The DC measurements were carried out at the same sites<br />
as 16 years before, no significant changes of the freshwater/saltwater boundary were found.<br />
Regarding direct-push and borehole data, the HEM inversion provided more information about<br />
resistivity structure. A 6-layer approach revealed additionally to the freshwater/saltwater structure thin<br />
clay layers.<br />
Based on the HEM inversion results the volume of the freshwater resource was estimated.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
195
Acknowledgments<br />
The HEM survey was funded by LIAG/BGR project “Airborne Geophysical Investigation of the German<br />
North Sea Coastal Area” (Wiederhold et al., 2008):<br />
http://www.liag-hannover.de/forschungsschwerpunkte/grundwassersysteme-<br />
hydrogeophysik/salz-suesswassersysteme/flaechenhafte-befliegung.html<br />
http://www.bgr.bund.de/cln_101/nn_328750/DE/Themen/GG__Geophysik/Aerogeophysik/Proj<br />
ektbeitraege/D-AERO/deutschlandweite__aerogeophysik__befliegung__D-AERO.html<br />
The lithological logs were taken from the LBEG Map Server (http://www.geophysics-database.de/) of<br />
the Niedersächsisches Landesamt für Bergbau, Energie und Geologie (LBEG).<br />
The direct-push results were provided by the Ingenieurbüro für Hydrogeologie, Sedimentologie und<br />
Wasserwirtschaft (HSW) (Winter, 2008).<br />
The digital elevation model was provided by the Niedersächsischer Landesbetrieb für<br />
Wasserwirtschaft, Küsten- und Naturschutz (NLWKN), Forschungsstelle Küste, An der Mühle 5, 26548<br />
Norderney.<br />
References<br />
Schaumann, G., Steuer, A., Siemon, B., Wiederhold, H. & Binot, F. [2010] Die deutsche Nordseeküste<br />
im Fokus von aeroelektromagnetischen Untersuchungen, Teilgebiete Langeoog mit Wattenmeer und<br />
Elbemündung, EMTF, this volume.<br />
Siemon B. [2006] Electromagnetic methods – frequency domain: Airborne techniques, In: Kirsch, R.<br />
(Ed.), Groundwater Geophysics – A Tool for Hydrogeology. Springer-Verlag, Berlin, Heidelberg, 155-<br />
170.<br />
Siemon B., Auken E. & Christiansen A.V. [2009a] Laterally constrained inversion of frequency-domain<br />
helicopter-borne electromagnetic data, Journal of Applied Geophysics, 67, 259-268.<br />
Siemon, B., Christiansen, A.V. & Auken, E. [2009b] A review of helicopter-borne electromagnetic<br />
methods for groundwater exploration. Near Surface Geophysics, 7, 629-646.<br />
Wiederhold, H., Binot, F., Kühne, K., Meyer, U., Siemon, B. & Steuer, A. [2008] Airborne geophysical<br />
investigation of the German North Sea Coastal Area, 20th Salt Water Intrusion Meeting 2008, Naples,<br />
USA.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
196
Wiederhold, H., Kirsch, R., Schaumann, G., Scheer, W. & Steuer, A. [2009] Airborne geophysical<br />
investigations for hydrogeological purposes in Northern Germany. Extended Abstracts Book of the<br />
15th European Meeting of Environmental and Engineering Geophysics – Near Surface 2009, Dublin,<br />
Ireland, 7.-9.9.2009.<br />
Winter, S. [2008] Erkundung und Bewertung von Grundwasserbelastungen im Bereich der<br />
Süßwasserlinsen Borkum, Teilprojekt Direct-Push-Sondierungen, Ergebnisbericht, HSW, Leer.<br />
Worzyk P. [1995a] Geoelektrische Tiefensondierungen auf der Insel Borkum – Untersuchungen zur<br />
Veränderung der Süßwasserlinse im Ostland. NLfB Archives no 111751, Hannover.<br />
Worzyk P. [1995b] Geoelektrische Tiefensondierungen auf der Insel Borkum – Untersuchungen zur<br />
Struktur der Süßwasserlinse im Einzugsbereich des Gewinnungsgebietes Waterdelle. NLfB Archives<br />
no 114088, Hannover.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
197
Some notes on bathymetric effects in marine<br />
magnetotellurics, motivated by an amphibious<br />
experiment at the South Chilean margin<br />
Gerhard Kapinos, Heinrich Brasse 1<br />
Freie Universität Berlin, Fachrichtung Geophysik, Malteserstr. 74-100, 12249 Berlin,<br />
Germany<br />
1 Introduction<br />
Whilst the onshore magnetotelluric method is established as a useful method for imaging<br />
electrical conductivity structures in the deep earth’s interior, electromagnetic investigation<br />
in marine environment has attracted until recently much less attention, although<br />
the oceans cover the most part of the Earth’s surface. Beside of logistical challenges,<br />
high technical demands on the instruments and involved higher costs, physical conditions<br />
in marine environments make the acquisition and interpretation of offshore data<br />
by far more difficult than onshore data. The broad period range of usable signals utilized<br />
in the terrestrial MT is substantially limited on the seafloor. The ionospheric and<br />
magnetospheric primary source field recorded on the seafloor is at short periods strongly<br />
attenuated by the covering highly conductive ocean layer and at long periods contaminated<br />
by motionally induced secondary electromagnetic fields. Additionally, attention<br />
has to be payed to the shape of the ocean bottom, which can massively distort the<br />
electromagnetic fields particularly in coast region.<br />
2 Decay of electromagnetic fields in the ocean<br />
The decay of the electromagnetic amplitude in a high conductive homogeneous half space<br />
like ocean water (ρ =0.3Ωm) and a less conductive medium like earth (ρ = 100Ωm)<br />
shows Fig. 1. In a homogeneous half space without boundary in vertical direction the<br />
field behaviour in both media can be estimated according to the simple formula used for<br />
the calculation of the skin depth<br />
F = F0e −kz , (1)<br />
1 kapinosg@geophysik.fu-berlin.de, h.brasse@geophysik.fu-berlin.de<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
198
z/δ<br />
0<br />
1<br />
2<br />
3<br />
4<br />
Re<br />
Im<br />
magnitude<br />
land<br />
ocean<br />
5<br />
-0.5 0 0.5 1<br />
E(z)/E(0)<br />
Figure 1: Decay of the of magnitude of electromagnetic field F (green) and its real<br />
(red) and imaginary part (blue) as function of depth in a highly conductive (i.e. ocean<br />
water, dashed line) and a less conductive homogeneous half space (solid line). The<br />
decay is scaled to the surface value F0.<br />
where k = √ iμσω. At a depth of 4000 m, which corresponds to the average ocean<br />
depth, the amplitude of a field propagating in the ocean at a period of 10 s is attenuated<br />
hundredfold, i.e. the field on the ocean bottom would be approximately 1% of the surface<br />
value.<br />
This simple relationship becomes more complicated when taking into account real<br />
conditions with a highly conductive ocean layer which is limited downward by a relatively<br />
resistive basement. Analog to the N-layered half space the solution has to be upgraded<br />
by a further term according to the reflection on the ocean bottom at depth h. The<br />
solution for the x-component of the electric field in the first and second layer is<br />
E1,x = E11e −k1z + E12e k1z<br />
E2,x = E22e −k2z<br />
where E11 is the amplitude of the downgoing and E12 of the upgoing wave in the first<br />
medium and E22 the amplitude of the downgoing wave in the second medium. For the<br />
y−component of the magnetic field follows from Faraday’s law in a 1-D case<br />
B1,y = − 1 ∂E1x<br />
iω ∂z<br />
B2,y = − 1 ∂E2x<br />
iω ∂z<br />
= k1<br />
iω<br />
<br />
E11e −k1z k1z<br />
− E12e<br />
(2)<br />
(3)<br />
(4)<br />
= k2<br />
iω E22e −k2z . (5)<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
199
A similar relation can be derived for the pair Ey and Bx. Using the continuity of the<br />
fields at z = h resulting from the boundary conditions<br />
and introducing the admittances<br />
ζ1 = k1<br />
iω =<br />
E1,x(h) =E2,x(h), B1,y(h) =B2,y(h) (6)<br />
<br />
iωμσ1<br />
iω =<br />
<br />
μσ1<br />
iω<br />
and ζ2 = k2<br />
iω =<br />
<br />
μσ2<br />
iω<br />
the ratio of amplitudes of the downgoing and upgoing fields in the ocean is<br />
where ζ12 = ζ1−ζ2<br />
ζ1+ζ2<br />
E12<br />
E11<br />
(7)<br />
= ζ1 − ζ2<br />
e<br />
ζ1 + ζ2<br />
−2k1h<br />
= ζ12e −2k1h<br />
, (8)<br />
is the reflection coefficient. Its value is estimated according to eq. 7 by<br />
the conductivity contrast at the boundary between layers. For instance, in case of sharp<br />
contrast like ocean water (σ1 =3.2 S/m) and basement (σ2 =0.001 S/m) ζ12 =0.965.<br />
Note, that using as solution in eq. 2 and 3 the magnetic field instead of electric field, the<br />
reflection coefficient has to be expressed by reciprocal impedances instead of admittance<br />
and yields: η12 = η1−η2<br />
η1+η2 ,withη1,2 = k1<br />
μσ1,2 =<br />
iω<br />
μσ1,2 .<br />
Combining the equations 8 and 2 the electric field at depth z ≤ h becomes<br />
E1,x(z) =<br />
−k1z<br />
E11 e + ζ12e −2k1h k1z<br />
e <br />
and at the surface z =0<br />
<br />
E1,x(0) = E11 1+ζ12e −2k1h . (10)<br />
The ratio of the equations 9 and 10 gives the attenuation of the electric field in the ocean<br />
for 0 ≤ z ≤ h normalized in terms of the total field at the surface<br />
|VE| = E1,x(z)<br />
E1,x(0) = e−k1z + ζ12e−2k1h 1+ζ12e−2k1h −2k1(h−z)<br />
1+ζ12e<br />
= e−k1z<br />
1+ζ12e−2k1h This applies accordingly for the ratio of magnetic fields<br />
B1,y(z) =<br />
−k1z<br />
B11ζ1 e − ζ12e −2k1h k1z<br />
e <br />
B1,y(0) =<br />
−2k1h<br />
B11ζ1 1 − ζ12e<br />
(12)<br />
(13)<br />
|VB| = B1,y(z)<br />
B1,y(0) = e−k1z − ζ12e−2k1h 1 − ζ12e−2k1h −2k1(h−z)<br />
1 − ζ12e<br />
=<br />
1 − ζ12e−2k1h e−k1z . (14)<br />
The behaviour of the electric and magnetic fields as function of depth in a 5 km deep<br />
ocean for two selected and representative periods (100 s and 10000 s) and realistic conductivities<br />
in the ocean (σ1 = 3.2 S/m) and basement (σ1 = 0.001 S/m) is illustrated in<br />
Fig. 2(left). The fields are normalized to their surface magnitude according to relations<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
200<br />
(9)<br />
(11)
z [km]<br />
0<br />
1<br />
2<br />
3<br />
4<br />
5<br />
3200<br />
B<br />
E<br />
E,B (hs)<br />
T=100s<br />
T=10000s<br />
0 0.2 0.4 0.6 0.8 1<br />
|V | , |V |<br />
B E<br />
T=10000s<br />
B<br />
E<br />
E,B (hs)<br />
=3.2<br />
3200<br />
0 0.2 0.4 0.6 0.8 1<br />
|V B | , |V E |<br />
Figure 2: The decay of electromagnetic fields in an ocean shows a deviation from simple<br />
half space attenuation due to boundary conditions on the seafloor and depends beside<br />
period length and ocean depth also on resistivity contrast between ocean water and<br />
basement. Attenuation of electric (blue) and magnetic (red) fields for two periods<br />
T1 = 100 s (solid line) and T1 = 10000 s (dashed line) in a 5 km deep ocean with a sea<br />
water resistivity of 0.3Ωm and seafloor resistivity of 1000 Ωm(right). Attenuation of the<br />
fields at a period of 10000 s for two different conductivities of the basement (0.001 S/m<br />
and 1 S/m) (left). See also (Brasse, 2009).<br />
11 and 14. They show different decay in dependence of period. At short periods (solid<br />
line) both the electric (blue) and magnetic (red) fields experience strong attenuation and<br />
the curve shape is determined by exponential decay. Large differences in attenuation<br />
are observed at long periods (dashed line). While the electric field penetrates the ocean<br />
layer from surface to the seafloor nearly unchanged, the magnetic field is attenuated<br />
about fifteen-fold and reaches on the ocean bottom just about 7% of its surface value.<br />
However, both fields would behave identically in an infinitely extended ocean, that would<br />
be regarded as a homogeneous half space.<br />
At first sight of the equations 11 and 14 one would think that the behaviour of the<br />
fields depends rather on the ocean thickness than on period. However, actually both<br />
is true. The decay of the fields is determined by both period T and thickness h of the<br />
ocean layer. That becomes evident considering the exponential and crucial term in the<br />
equations as function of skin depth according to the relations of skin depth δ =<br />
and complex wave number k =(1+i) ωσμ<br />
2 :<br />
2<br />
ωσμ<br />
e −2k1h = e − 2h<br />
δ 1 (i+1) . (15)<br />
The quantity h<br />
δ1 is the electrical layer thickness and via δ ∼ √ T also directly associated<br />
with the period length of the penetrating wave (McNeill and Labson, 1986). The fields<br />
behave identically if the exponent remains constant, regardless if high frequency fields<br />
propagate in a shallow ocean or low frequency fields in a deep ocean. The behaviour<br />
of the fields replicates, similar to a self-similarity principle by stimulating the exponent<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
201
h , i.e. by appropriate and simultaneous increasing or decreasing the frequency and<br />
δ1<br />
the ocean thickness and thus creating either electrical thick h >> 1 or thin conditions<br />
δ1<br />
h > 1andbothe−2k1h<br />
δ1<br />
and e−2k1(h−z) → 0 (Brasse, 2009). Thus the effect of the basement on the fields is only<br />
marginal and the behaviour in a thick electrical layer is governed by exponential decay<br />
E1,x(z)<br />
E1,x(0)<br />
≈ e−k1z<br />
respectively B1,y(z)<br />
B1,y(0) ≈ e−k1z . (16)<br />
The decay of both fields becomes more alike the more the ocean depth or the frequency<br />
increases until finally approaching the decay curve of a homogeneous half space.<br />
For low frequencies or for an electrically thin ocean, where the skin depth is significantly<br />
greater than ocean depth, h
would imply that the currents only flow in the ocean, inducing a secondary magnetic<br />
field which removes entirely the opposite primary field in agreement with the above<br />
mentioned result. In real conditions with high but not infinite basement resistivity<br />
the currents diffuse also to the seafloor, so that the the secondary magnetic field of<br />
the diminished currents in the ocean can in fact reduce but not completely cancel the<br />
primary magnetic field. Thus the ratio B1,y(z)<br />
B1,y(z)<br />
doesn’t become zero rather 0 < < 1<br />
B1,y(0) B1,y(0)<br />
depending on ζ12 and electrical conditions of the ocean layer. In other words the reflection<br />
coefficient regularizes the conditions for distribution of electrical currents in sea bottom<br />
and estimates in this manner the damping behaviour of the fields (Chave et al., 1991).<br />
The decay of the electric field in the ocean is against intuition, and negligible compared<br />
to the magnetic field. In the theoretical case with infinite conductivity contrast, (ζ12 =1)<br />
the ocean represents for long periods an electrical thin layer that the electric field passes<br />
almost intact. In real condition mentioned above ζ12 is still high and the low frequency<br />
field decays gently, in terms of an electrical thin layer. That would change assuming<br />
a layer of highly conductive sediments on the ocean bottom. Since the layer becomes<br />
thicker, the electric field decays more strongly, the magnetic field in contrast much less<br />
since a part of currents now flows in the sediments. Thus both fields would approximate<br />
the curve of attenuation in a homogeneous half space, what for short periods corresponds<br />
to exponential decay Fig. 2(right).<br />
Similar effects would be observed considering the fields at a fixed point z1, within the<br />
ocean (0
It is also instructive to consider the electric and magnetic fields in the ocean as function<br />
of periods. Fig. 3 illustrates graphically the ratio |VB| = B1,y(z)<br />
B1,y(0) and |VE| = E1,x(z)<br />
E1,x(0)<br />
for various thicknesses of the ocean layer. It becomes clear immediately that in the<br />
real ocean environment attenuation of the fields differs fundamentally over a total period<br />
range. While the electric field remains nearly unchanged in electric thin layer<br />
( h
ocean depends on pressure, salinity and temperature. While the effect of pressure and<br />
salinity is marginally, the temperature variations in ocean change resistivity of oceanic<br />
water from 0.16 Ωm at 30 ◦ C to 0.33 Ωm at 1.0 ◦ C. The estimated resistivity values about<br />
0.3Ωm corresponds very well to literature values (Filloux, 1987)<br />
3 Offshore magnetotelluric and magnetic transfer<br />
functions in presence of bathymetry<br />
The discussions about magnetotellurics on the seafloor are often dominated by attenuation<br />
of electromagnetic fields by high conductive ocean and secondary induction by<br />
sea tides. Fewer studies refer to distorting effects on electromagnetic fields caused by<br />
changes in seafloor bathymetry in presence of overlaying high conductive environment<br />
(Constable et al., 2009).<br />
For a model incorporating a homogeneous half space with resistivity of 100Ωm and<br />
conductive ocean with bathymetry that is observed at the South Chilean continental<br />
margin (Fig. 5, top), magnetotelluric transfer function were calculated with the algorithm<br />
of Mackie et al. (1994). At least two clear findings can be derived from the model<br />
0<br />
2<br />
4<br />
6<br />
10 0<br />
10 1<br />
10 2<br />
10 3<br />
104 100 10 1<br />
10 2<br />
10 3<br />
10 4<br />
ocean<br />
ρ= 0.3Ωm ρ=100Ωm TM<br />
TM<br />
0 50 100 150 200 250 300 350 400<br />
distance [km]<br />
0 50 100 150 200 250 300 350 400<br />
distance [km]<br />
Figure 5: Responses of a synthetic amphibious model shown on the top. Middle: resistivity<br />
of TM and TE mode. Bottom: phase of TM and TE mode. Dashed line separates<br />
the profile into offshore and onshore part.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
205<br />
TE<br />
TE<br />
4<br />
3<br />
2<br />
1<br />
0<br />
90<br />
75<br />
60<br />
45<br />
30<br />
15<br />
0
esponses (Fig. 5, middle and bottom). Firstly the ocean has rather a marginal effect<br />
on the onshore magnetotelluric transfer function (right of the dashed line). Actually<br />
only in TM mode of onshore stations close to the ocean slightly decreasing phases and<br />
enhanced apparent resistivity on the edge between offshore- and onshore profile marked<br />
by a dashed line, can be observed. Secondly the bathymetry produces dramatic anomalous<br />
effects in the ocean, which can be observed in both modes as well as in phases and<br />
resistivities. The resistivity in TE mode rises and falls dramatically producing cusps in<br />
the image and phases even exceed −180 ◦ and 180 ◦ .<br />
Similar anomalous features can be observed in the magnetic transfer functions (Fig.<br />
6). Slight changes in the shape of the seafloor, adumbrated by a dotted line, are overdrawn<br />
reproduced by dramatic variations in the magnitude of the induction arrows (note<br />
the smooth bathymetry change between stations A09 and A13 and their huge impact on<br />
the magnitude of the induction arrows). On the other hand, the ocean effect on onshore<br />
station is limited to long-periods and to near-coastal region, again.<br />
10 0<br />
10 1<br />
10 2<br />
10 3<br />
10 4<br />
tipper M r<br />
0 50 100 150 200 250 300 350 400<br />
distance [km]<br />
tipper Mag<br />
0 50 100 150 200 250 300 350 400<br />
distance [km]<br />
Figure 6: Real part (left) and the magnitude (right) of the tipper calculated for synthetic<br />
and amphibious model shown on the top of figure 5. The dotted line indicates<br />
the shape of the bathymetry of the model.<br />
Comparing these synthetic responses with the real induction vectors observed at the<br />
only station at the continental slope, ob7, shows that the bathymetry indeed affects the<br />
measured data, as shown in Fig. 7. The huge induction vectors (over 1) at middle<br />
and long periods correspond roughly with magnitude values calculated from the simple<br />
amphibious model. For short periods (until 100 s) a comparison is impossible due to<br />
Re<br />
1.0<br />
0.5<br />
0.0<br />
10 2<br />
ob7<br />
10 3<br />
Figure 7: Real part of the observed induction vectors at station ob7, deployed on the<br />
continental margin off Southern Chile. Induction vectors observed land-side are presented<br />
by Brasse (2009).<br />
10 4<br />
10 5<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
206<br />
1.2<br />
1.0<br />
0.8<br />
0.6<br />
0.4<br />
0.2<br />
0.0
unavailable data (note the different period range of the images).<br />
The anomalous responses can be elucidated considering electric and magnetic fields in<br />
presence of changing seafloor shape. An enhanced concentration of electrical currents at<br />
the continental slope produces anomalous strong vertical magnetic field, that is known<br />
also on the land-side as coast effect (e.g., Fischer, 1979). Due to the high sea water<br />
conductivity the induced electric currents concentrate primarily in the ocean avoiding<br />
the much more resistive subsurface and generate a secondary magnetic field which compensates<br />
the primary opposite field, so that the magnetic field diminish considerably or<br />
disappears completely in case of insulating substrate as mentioned in 2 and shown in<br />
Fig. 2. At the continental margin, where the seafloor shallows towards land, the density<br />
of electric currents flowing in the ocean along the coast (i.e. in the TE mode) and above<br />
(instead below) measurement points increase inducing an anomalous and opposite magnetic<br />
field. This opposite, secondary field becomes predominant on the ocean bottom,<br />
where the primary field is strongly damped whereas the electric field is by the coast effect<br />
only marginally affected, and causes jumps in the phase (Weidelt, 1994). Moreover the<br />
resistivities in TE mode rise steeply generating upward cusps and the tipper gets very<br />
large as presented in Fig. 6 and 5.<br />
The periods at which the effects occur depend on position of the probe in relation to<br />
the slope and corresponds to lateral distance to which a transfer function is sensitive<br />
to bathymetry. However, at high frequencies the field is generally unaffected by the<br />
bathymetry and the results are MT-like because the fields still behave like plane waves.<br />
At low frequencies, the currents are below the ocean. At the middle frequencies, the<br />
currents are in the ocean and this causes the responses looking like those due to infinite<br />
line currents right above the station. Note that if the MT responses were recorded<br />
on the top of the sea, the responses would be perfectly fine as expected in an usual<br />
2-D case, since all the currents pass below the sounding surface instead above like in<br />
the presented marine model. This behavior is overcome or reduced by introducing a<br />
very conductive layer below the ocean bottom (e.g., sediments), which has to be at least<br />
several kilometers thick. However, the assumption of an oceanic crust being entirely well<br />
conductive seems unrealistic; this is even more the case for the continental crust below<br />
the slope: every aquiclude would prohibit the intrusion of sea water into deeper layers.<br />
Another class of models makes the effect described above disappear too: the association<br />
of the upper part of the downgoing plate with a good conductor, in accordance with<br />
standard models of subduction.<br />
4 Summary<br />
Offshore magnetotellurics is an useful method for exploration of seafloor and oceancontinent<br />
subduction zones, where fluids and melts are known to control the subduction<br />
process. A combined on- and offshore magnetotelluric transect across subduction zone<br />
is assumed to be able to resolve high conductive structures associated with dehydration<br />
processes, water migration and melts building.<br />
However special conditions on the ocean floor not allow to interpret and to deal<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
207
with the marine records in the same manner as with onshore data. The presence of<br />
bathymetry distorts the magnetic and magnetotelluric transfer functions, particularly<br />
in the TE mode. Even a gently changing seafloor shape generates an enhanced concentrations<br />
of electric currents flowing above instead below measurement station and<br />
inducing on the ocean bottom a predominant anomalous opposite magnetic field. This<br />
results in phases exceeding the quadrant and cusps in the apparent resistivity.<br />
The high conductive sea water causes a strong attenuation of electric and magnetic<br />
field at short periods. Towards long periods the decay differs clearly for both fields. The<br />
electric field penetrates the ocean layer from surface to the seafloor, counterintuitive,<br />
nearly unchanged, while the magnetic field experiences a strong decay and reaches the<br />
ocean bottom just with a fraction of its surface value. Moreover the decay depends<br />
strongly from the resistivity contrast between ocean and seafloor. Reducing the resistivity<br />
of the basement the electrical thickness of layer increases and both fields approximate<br />
a field decay like in a homogeneous half space.<br />
References<br />
Brasse, H., G. Kapinos, Y. Li, L. Mütschard, W. Soyer, D. Eydam (2009): Structural<br />
electrical anisotropy in the crust at the South-Central Chilean continental margin as<br />
inferred from geomagnetic transfer functions, Phys. Earth Planet. Inter., 173, 7-16.<br />
Brasse, H. (2009): Methods of geoelectric and electromagnetic deep sounding, Lecture<br />
Notes, Free University of Berlin.<br />
Chave, A.D., S.C. Constable, and R.N. Edwards (1991): Electrical Exploration Methods<br />
for the Seafloor, in: Electromagnetic Methods in Applied Geophysics, Vol. 2 (Ed. M.N.<br />
Nabighian), Soc. Expl. Geophys., Tulsa, 931-966.<br />
Constable, S.C., K. Key, and, L. Lewis (2009): Mapping offshore sedimentary structure<br />
using electromagnetic methods and terrain effects in marine magnetotelluric data,<br />
Geophys. J. Int., 176, 431-442.<br />
Filloux, J.H. (1987):Instrumentation and experimental methods for oceanic studies, in:<br />
Geomagnetism, (Eds. J.A. Jacob), Academic Press,1987,143–248.<br />
Fischer, G. (1979): Electromagnetic induction effects on the ocean coast, Proc. IEEE,<br />
67, 1050-1060<br />
McNeill, J.D., and V. Labson (1986), Geological mapping using VLF radio fields. Electromagnetic<br />
Methods in Applied Geophysics. Volume 2, Appliction, Parts A and B,<br />
(Ed. M.N. Nabighian), SEG, Tulsa, 521-639.<br />
Mackie, R., J. Smith, and T.R. Madden (1994), Three-dimensional modeling using finite<br />
difference equations: The magnetotelluric example, Radio Science, 29, 923-935.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
208
Weidelt, P. (1994), Phasenbeziehungen für die B-Polarisation, in: Protokoll über das<br />
Kolloquium ”Elektromagnetische Tiefenforschung” (Eds. K. Bahr and A. Junge),<br />
Höchst, Odenwald, 60-65.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
209
A permanent array of magnetotelluric stations located at the<br />
South American subduction zone in Northern Chile.<br />
Introduction<br />
Dirk Brändlein 1,2 , Oliver Ritter 1,2 , Ute Weckmann 1,3<br />
1 <strong>GFZ</strong> German Research Centre for Geosciences, Potsdam, Germany<br />
2 Freie Universität Berlin, Fachrichtung Geophysik, Berlin, Germany<br />
3 University of Potsdam, Institute of Geosciences, Potsdam, Germany<br />
Monitoring the dynamic behavior of an active deep<br />
subduction system is focus of the Integrated Plate Boundary<br />
Observatory Chile (IPOC), a permanent array of combined<br />
geophysical and geodetic stations in Northern Chile which is<br />
operated since 2006 by the <strong>GFZ</strong> German Research Centre for<br />
Geosciences. Magnetotelluric (MT) data is gathered at seven<br />
out of a total of eleven observation sites.<br />
The MT set up consists of three component long period<br />
fluxgate magnetometers (GeoMagnet) and non-polarizing<br />
Ag/AgCl electrodes to measure all three components of the<br />
magnetic field and both horizontal components of the<br />
electric field. The signals of the electromagnetic fields are<br />
continuously sampled at a rate of 20 Hz and at four sites<br />
transferred via satellite link to the <strong>GFZ</strong> in Germany. The<br />
objective of the project is to monitor and analyze<br />
electromagnetic data to decipher possible changes in the<br />
subsurface resistivity distribution, e.g. as a consequence of<br />
large scale fluid relocation.<br />
Here, we present vertical magnetic transfer functions as time<br />
series over a time span of more than two years for the<br />
period range from 10 -1 to 10 4 seconds 1 . These vertical<br />
magnetic transfer functions are sensitive to lateral changes<br />
of electric conductivity in the subsurface. Some components<br />
of these transfer functions show frequency dependent<br />
variations with a periodicity of roughly one year. These<br />
effects are observed at all sites of the array.<br />
Figure 1: The IPOC-MT array in<br />
Northern Chile showing sites PB01<br />
to PB07 (red symbols). Blue stars<br />
indicate major earthquakes.<br />
Due to the extreme dry ground of the Atacama desert<br />
continuous monitoring the electric field is difficult. Contact<br />
resistances are on the order of MΩ and electrolyte is leaking. Different types of electrodes are<br />
currently being tested.<br />
1 Periods shorter than 10 seconds are not considered since this is the shortest period fluxgate magnetometers are able to resolve.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
210
Periodic variations in the MT monitoring data<br />
To identify changes in the subsurface we examine time series of vertical magnetic transfer functions<br />
Tx(ω) and Ty(ω):<br />
Bz(ω) = Tx(ω) Bx(ω) + Ty(ω) By(ω).<br />
These quantities are sensitive to lateral changes of electric conductivity in the subsurface.<br />
The value of each day and component of the transfer function time series is obtained by processing<br />
magnetic field time series of 3 days (using the day before and after). Subsequently the 3 day time<br />
window is forwarded by one day and the procedure is repeated until the entire time series is<br />
processed. The geomagnetic transfer functions were obtained using the robust processing described<br />
in Ritter el al. (1998), Weckmann et al. (2005), and Krings (2007).<br />
Figure 2: Time series showing the differences between daily values of Ty and the median of the<br />
entire data section of 730 days at each period for site PB03. For monitoring purposes it is<br />
desirable to amplify variations in the vertical magnetic transfer function components. Therefore,<br />
the median is calculated for each period and component and subtracted from the value of each<br />
day. The median of each frequency band is plotted on the right hand side. The median is more<br />
robust against outliers than the arithmetic average. Vertical red lines indicate major earthquakes<br />
(see Fig. 1), vertical blue lines indicate turn of the year. Interestingly, the ocean effect, which<br />
would be indicated by a large positive value of the median of Re[Ty], is rather small at site PB03.<br />
The time series of the real and imaginary parts of the Ty (east-west) component at PB03 show a<br />
remarkable frequency dependent feature (Fig. 2). At periods between approximately 40 and 1000 s<br />
the observed Re[Ty] values show a continuous variation of the amplitudes with a periodicity of<br />
roughly one year. The amplitudes vary in the range ± 0.1 and reach maximum values in the austral<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
211
winter before the winter solstice. In the austral summer, the minimum occurs before the summer<br />
solstice. Similar effects can be observed consistently at all sites of the array (Fig. 3).<br />
Figure 3: Time series of real (blue) and imaginary (red) parts of vertical magnetic transfer<br />
functions Tx and Ty of around 600 days at a period of 1000 seconds of sites PB01, PB02, PB03<br />
and PB05.<br />
Possible causes for the periodic variations of Ty<br />
Possible causes for this periodic variation of the Ty component could be source field<br />
inhomogeneities. The sources for the MT-method are the natural variations of the earth's magnetic<br />
field which are assumed to be far away from the observer so that electromagnetic waves penetrate<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
212<br />
Figure 4: Time series<br />
of the magnetic auto<br />
spectra of PB03,<br />
extracted from the 3day-average<br />
processing described<br />
in section 2. The<br />
ByBy * spectra show<br />
the same variations<br />
with a periodicity of<br />
one year as Ty.
the subsurface as plane waves.<br />
To examine if the variations in the time series of vertical magnetic transfer functions are caused by<br />
seasonal alterations of source fields we plotted the time series of magnetic auto spectra of PB03 (Fig.<br />
7) for the same time window as in Figure 2. These magnetic auto spectra show a clear correlation<br />
between the By-auto spectrum (east-west direction) and the Ty data (Fig. 8). This would support the<br />
idea that the results are influenced by source field effects.<br />
Sq variations and equatorial electrojet<br />
There exist several effects on the geomagnetic field<br />
which have a wide spectrum of variations<br />
(Onwumechili, 1997). Annual variations of the<br />
geomagnetic solar quiet daily variation (Sq) show the<br />
same periodicity as the measured variations of the<br />
magnetic auto spectra at PB03. The geomagnetic Sq<br />
field has a variation with a dominant periodicity of 1<br />
day. Solar radiation generates ionized molecules in<br />
parts of the ionosphere (around 80 to 300 km height)<br />
which produces variable charged particles and,<br />
consequently, conducting air. Additionally, the solar<br />
radiation causes thermo-tidal winds which move the<br />
conducting ionosphere through the geomagnetic field.<br />
The result is a system of electric currents depending<br />
on the position of the sun. Horizontal and vertical<br />
components of Sq tend to be predominantly semiannual<br />
in the zone of Equatorial Electrojet (EEJ) but<br />
predominantly annual at other latitudes (Onwumechili,<br />
1997).<br />
The IPOC MT array (21 - 23° south) in<br />
Northern Chile is located near the<br />
tropic of Capricorn (23° 26' south). This<br />
means, it is influenced by the<br />
Equatorial Electrojet (EEJ), a varying<br />
electric current flowing eastward in the<br />
ionosphere at a height of 100 - 130 km.<br />
Over South America the EEJ is warped<br />
to the south because it follows the<br />
magnetic dip equator (Fig. 8). The<br />
amplitudes are aligned to the north<br />
and to the south by return currents<br />
(Lühr et al.,2004) with peak values at<br />
latitudes some 5° away from the<br />
magnetic dip equator.<br />
Figure 5: Exemplary electric currents in<br />
the ionosphere cause Sq variations in<br />
the northern summer.<br />
http://geomag.usgs.gov/images/ionosp<br />
heric_current.jpg (10.02.2010)<br />
Figure 6: Electrojet current densities inferred from 2600<br />
passes of the CHAMP satellite over the magnetic<br />
equator between 11:00 and 13:00 local time.<br />
http://www.geomag.us/info/equatorial_electrojet.html<br />
(10.02.2010)<br />
Annual Sq variations are most likely the cause of the variations of the Y-component of the vertical<br />
magnetic transfer functions at the MT-array in Northern Chile.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
213
Non-periodic variations in the MT monitoring data<br />
No periodic long term variations can be identified in the Tx component of the vertical magnetic<br />
transfer function time series (Fig. 7). A continuous variation at a period of approximately 20 s can be<br />
observed in the Re[Tx] component. It has a high amplitude variation between -0.2 and +0.2.<br />
Figure 7: Time series of differences between daily values of Tx and median of the entire data<br />
section at each period at site PB03. Vertical red lines indicate major earthquakes, vertical blue<br />
lines indicate turn of the year. Values of median versus period are plotted on the right hand side.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
214<br />
Figure 8: Time series of<br />
differences between daily<br />
values of vertical magnetic<br />
transfer functions and<br />
median at site PB03 versus<br />
period. Length and period<br />
range correspond to the<br />
black box in figure 6.<br />
Vertical red lines indicate<br />
major earthquakes (Fig. 1),<br />
vertical blue lines indicate<br />
turn of the year.
Figure 9: Time series of differences between daily values of<br />
vertical magnetic transfer functions and median at site PB05<br />
versus period. Length and period range correspond to the black<br />
box of Fig. 6. Vertical red lines indicate major earthquakes (Fig.<br />
1), vertical blue lines indicate turn of the year.<br />
Note that the curve of the<br />
median, which represents the<br />
average transfer function,<br />
varies smoothly and<br />
consistently in this period<br />
range.<br />
A remarkable feature can be<br />
observed between two major<br />
earthquakes in December<br />
2007 (Fig. 7, black box). The<br />
area of interest shows a<br />
sudden rise in amplitudes up<br />
to ±0.25 (Fig. 8), which<br />
corresponds to 100% of the<br />
median value of Tx (Fig. 7,<br />
right hand side). A similar but<br />
less pronounced effect is<br />
observed at site PB05 (Fig. 9).<br />
More modeling is necessary<br />
to quantify the scale and location of structural changes of the electrical conductivity in the<br />
subsurface which could explain these non-periodic variations in the vertical magnetic transfer<br />
function time series. Calculation and modeling of inter station transfer function time series should<br />
also give more detailed information about changes in the conductivity structure of the underground.<br />
Acknowledgements<br />
We acknowledge funding from the <strong>GFZ</strong> German Research Centre for Geosciences. For continuous<br />
logistic help in Chile we want to thank Prof. Guillermo Chong and the people from the Universidad<br />
Catolica del Norte in Antofagasta. We are grateful to Günter Asch for his cooperation and help to<br />
install and maintain the technical infrastructure in the Atacama desert. We are very thankful for<br />
support in the field by Kristina Tietze and Thomas Krings.<br />
References<br />
Krings, T., 2007. The influence of Robust Statistics, Remote Reference, and Horizontal Magnetic Transfer<br />
Functions on data processing in Magnetotellurics. Diploma Thesis, WWU Münster ― <strong>GFZ</strong> Potsdam.<br />
Lühr, H., S. Maus, and M. Rother, 2004. Noon-time equatorial electrojet: Its spatial features as determined by<br />
the CHAMP satellite, Journal of Geophysical Research, 109, A01306, doi:10.1029/2002JA009656.<br />
Onwumechili, C. A., 1997. The Equatorial Electrojet, Gordon and Breach, Newark, N. J.<br />
Ritter, O., Junge, A. and Dawes, G., 1998. New equipment and processing for magnetotelluric remote<br />
reference observations. Geophysical Journal International, 132, 535-548.<br />
Weckmann, U., Magunia, A. and Ritter, O., 2005. Effective noise separation for magnetotelluric single site<br />
data processing using a frequency domain selection scheme. Geophysical Journal International, 161,<br />
635-652.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
215
Discussion on backarc mantle melting in the central Andean<br />
subduction zone, based on results of magnetotelluric studies<br />
D. Eydam ∗ & H. Brasse<br />
Free University of Berlin, Malteserstr. 74-100, 12249 Berlin<br />
Abstract<br />
Long-period magnetotelluric measurements in the Bolivian orocline revealed a large<br />
conductor in the backarc upper mantle of the central Andean subduction zone at 18◦S. Data analysis and interpretation by 2-D inversion has already been described by Brasse<br />
& Eydam (2008a) who interpreted this conductor as an image of partial melt triggered<br />
by fluid influx from the subducting slab.<br />
Remarkable are its high conductivities ranging up to above 1 S/m and implying high<br />
melting rates of more than 6vol%. However, regarding the distribution of intermediate<br />
depths seismicity which marks dehydration reactions in the subducting slab, one has<br />
to note the large distance of more than 60 km between fluid source and the assumed<br />
fluid-triggered partial melting of mantle peridotite. How do actual concepts of subduction<br />
zone dynamics fit to this resistivity image?<br />
There are at least two approaches which are both followed by in the subsequent discussion:<br />
first is to study the mobility of deep fluids in subduction zones, second is to<br />
study the mobility of the slab itself.<br />
Introduction<br />
Ocean-continent subduction zones are earth regions of major raw material recyclings.<br />
Hot magmas erupt nearby where cold oceanic lithosphere subducts. This apparently<br />
contradictory observation is explainable by consumption of deep slab-derived fluids for<br />
flux melting of the peridotitic mantle in the asthenospheric wedge:<br />
Hydrous minerals in the oceanic lithosphere are progressively dehydrating during subduction.<br />
The hereby released water may trigger partial melting of the overlying asthenospheric<br />
wedge by lowering mantle solidus generally at depths below 80 km. Melts<br />
and fluids ascent with some intermediate storage at density contrasts at the Moho<br />
or intra-crustal boundaries and give rise to arc and backarc volcanism. Due to their<br />
∗ now at <strong>GFZ</strong> Potsdam, Telegraphenberg, 14437 Potsdam<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
216
Eydam & Brasse: Discussion on backarc mantle melting<br />
enhanced conductivities, saline fluids and (hydrous) melt should principally be detectable<br />
by deep electromagnetic sounding methods, as was demonstrated, e.g., in the<br />
Cascadia/Juan de Fuca subduction zones (Kurtz et al., 1986; Wannamaker et al., 1989;<br />
Varentsov et al., 1996).<br />
Hydrous phases in the oceanic crust relevant for sub-arc water release are amphibole<br />
and lawsonite. Amphibole is abundant in the basaltic oceanic crust and decays only<br />
slightly pressure-dependent at 65-80 km depths, marking the transition from blueschist<br />
to eclogite facies. In contrast, lawsonite can be stable up to high pressures of 8.4 GPa<br />
(z ≈ 250 km) and dehydrates when temperatures exceed 700-800◦C. Mantle water release<br />
is mainly controlled by serpentinite dehydration between 500-700◦C. Serpentinite<br />
can be stable up to high pressures of 6 GPa where it transforms through a waterconserving<br />
reaction to the hydrous DHMS (Dense Hydrous Magnesium Silicat) -phase<br />
A.<br />
The amount of released water thus strongly depends on the thermal structure as well<br />
as on subduction geometry. Young and slowly subducting slabs may dehydrate completely<br />
whereas old and rapidly subducting lithospheres may import large amounts of<br />
mantle water into the deeper mantle, making them crucial for the global water cycle.<br />
The definitive amount of deep fluids in subduction zones is still uncertain because<br />
of the dependency of dehydrations from slab heating via mantle convection which in<br />
turn, strongly depends on water content in the wedge, besides other mostly poorly<br />
constrained parameters.<br />
Geological setting<br />
The central Andean subduction zone<br />
Along 7000 km length the oceanic Nazca plate subducts beneath the South American<br />
continental margin where the Andes evolved as a Neogene volcanic chain embedded in<br />
a compressive backarc tectonic setting. Holocene volcanism is seperated in four active<br />
segments, the austral, southern, central and northern volcanic zone (fig. 1).<br />
Characteristic for the whole subduction system are low slab dips of less than 30◦ and inactive volcanism in regions where aseismic ridges subduct. Those hot oceanic<br />
lithospheres probably flatten at main depths of sub-arc water release due to a retard<br />
of dehydration and thus density increase of the slab which may lead to ’horizontal subduction’<br />
over several hundreds of kilometres (Gutscher et al., 2000). Therefore, slab<br />
pull and buoyancy forces should be quite equilibrated probably making the slab mobile<br />
in geological time scales.<br />
The Bolivian orocline (13-28◦S) encompasses the central volcanic zone (CVZ) where<br />
the oldest part of the Nazca plate (55 Ma) subducts obliquely with an angle of N77◦E and with a current velocity of 6.5 cm/a (Klotz et al., 2006). Convergence rate has<br />
continuously slowed down from high 15 cm/a since the breakup of the oceanic Farallon<br />
plate in Nazca and Cocos plate 26 Ma ago (Somoza, 1998).<br />
During the highly compressive Miocene regime, stresses are mainly relaxed by extreme<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
217
Eydam & Brasse: Discussion on backarc mantle melting<br />
10°S<br />
20°S<br />
30°S<br />
40°S<br />
Nazca Rise<br />
Iquique Rise<br />
Juan-Fernández Ridge<br />
Chile Rise<br />
7.8 cm/a<br />
SVZ<br />
Easte<br />
Altiplano<br />
CVZ<br />
Puna<br />
80°W 75°W 70°W 65°W<br />
Figure 1: South American subduction zone with structural units of the central volcanic zone (CVZ)<br />
backarc and foreland. Hachures: Altiplano-Puna transition between 22-24 ◦ S; SB, SP - thick-skinned<br />
folding and thrusting in the Santa-Bárbara system and Sierras Pampeanas; P - Pica-Gap of Holocene<br />
volcanism (19.2-20.7 ◦ S) where Iquique Rise subducts (Wörner et al., 2000); SVZ - southern volcanic<br />
zone. Convergence rate after Somoza (1998).<br />
crustal shortening, thickening and uplift of the entire Andean arc and backarc region<br />
(e.g., Allmendinger et al., 1997; Scheuber et al., 1994; Elger et al., 2005). As a consequence<br />
the Altiplano-Puna high plateau developed, encompassing the highest volcanoes<br />
of the world (e.g., 6500 m for Sajama volcano) in the Western Cordillera to the west,<br />
rough mountain ranges in the Eastern Cordillera at the eastern boundary and the intramontanous<br />
drained Altiplano and Puna basins with thick Miocene stratas in-between.<br />
The up to 1800 km long and 350 km wide Altiplano-Puna plateau is the worldwide<br />
greatest high plateau in a subduction context. Crustal thickness is enormous (up to<br />
75 km; e.g., Wigger et al., 1994) and decreases beyond the plateau extension to normal<br />
values for continental lithosphere of 35-40 km. Likewise volcanism in the CVZ is highly<br />
crustally contaminated against more island-arc denoted lavas in the SVZ.<br />
Altiplano and Puna form an unique tectonomorphic entity within the orogen although<br />
their morphology differs. While the Altiplano in the north is rather flat and tectonically<br />
quiet, the Puna in the south is pervaded by many deep reaching, active horst- and<br />
graben structures. Average elevations are around 3700 m in the Altiplano and 4500 m<br />
in the Puna.<br />
Actual compressive tectonics are mainly restricted to the Andean foreland, where<br />
westvergent thin-skinned thrusting in the Subandean to the north change to thickskinned<br />
thrusting in the Santa Bárbara system and Sierras Pampeanas to the south<br />
(Allmendinger et al., 1997). At Altiplano latitudes, the comparatively fast westward<br />
subduction of the Brazilian Shield enhances the compressive backarc regime.<br />
P<br />
rn<br />
Cordi<br />
llera<br />
SP<br />
Suba<br />
SB<br />
nde<br />
an<br />
Chaco<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
218
Eydam & Brasse: Discussion on backarc mantle melting<br />
50 km<br />
75 km<br />
Pacific Ocean<br />
100 km<br />
125 km<br />
Argentina<br />
100 km<br />
CC<br />
LV<br />
150 km<br />
175 km<br />
125 km<br />
PC<br />
200 km<br />
Western Cordillera<br />
San Andrés fault<br />
P S<br />
225 km<br />
Coniri<br />
fault<br />
system<br />
Corque Basin<br />
Altiplano<br />
Rio Desaguadero<br />
Chuquichambi<br />
Eastern Cordillera<br />
Laurani fault<br />
Figure 2: Survey area with locations of the 30 MT sites in the Bolivian orocline. Profile trends roughly<br />
perpendicular to the strike of main morphological units and to the contour lines of the Wadati-Benioff<br />
zone (dotted lines). CC denotes Coastal Cordillera, LV Longitudinal Valley and PC Precordillera; P,<br />
S are Parinacota, Sajama volcanoes; mN is magnetic north.<br />
The survey area<br />
The magnetotelluric profile transverses the central Bolivian orocline from the Coastal<br />
Cordillera in northermost Chile, to the active volcanic arc and the Altiplano in central<br />
Bolivia and ends in the Eastern Cordillera. The profile follows a general trend of N48◦S which is roughly perpendicular to main structural units as well as to contour lines of<br />
the Wadati-Benioff zone (fig. 2).<br />
Main morphological units in the forearc are paralleling the trench and correspond to an<br />
ancient magmatic arc which shifted eastward through time, recordable since the Jurassic<br />
in the Coastal Cordillera, afterwards in the Longitudinal Valley, in the Precordillera<br />
and until the Neogene in the Western Cordillera (Scheuber et al., 1994). The exposure<br />
of the forearc crust to periods of extensive and long-lived magmatism gave rise to the<br />
formation of deep, long and wide fault zones (sub-)paralleling the trench among which<br />
some are still active like the Precordillera Fault System at 19.2-21◦S. Here, large copper<br />
deposits evidence strong fluid circulation and mineralisation after the eastward shift of<br />
the arc 32 Ma ago.<br />
In the closer survey area, the forearc is incised by deep, E-W running valleys which<br />
open to the ocean and may have served as Miocene drainage systems. Some of them<br />
may be associated with active faults (Wörner et al., 2000). A system of west vergent,<br />
steeply dipping thrust faults in the eastern Precordillera, the West-Vergent-Thrust-<br />
System, marks the transition to the high plateau and is thought to be the location of<br />
major plateau uplift (Muñoz & Charrier, 1996).<br />
The smooth monoclinal western slope of the high plateau is locally shaped by huge<br />
land slides, e.g., the Oxaya collaps 12 Ma ago (Wörner et al., 2002). In the West-<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
219<br />
thrust<br />
275 km<br />
250 km<br />
mN N
Eydam & Brasse: Discussion on backarc mantle melting<br />
ern Cordillera, Plio-Pleistocene to recent, andesitic to rhyodacitic stratovolcanoes are<br />
build on thick layers of ignimbrites which are widespread throughout the entire forearc<br />
and are dated between 22-19 Ma (Oxaya Ignimbrites) and to 2.7 Ma (Lauca-Pérez<br />
Ignimbrite) (Wörner et al., 2000).<br />
The Altiplano can be separated into the volcanic Mauri region to the west and in huge<br />
sedimentary basins easterly. The San-Andrés fault marks the boundary in-between and<br />
limits the Arequipa terrane to the east, a component of the Arequipa-Antofalla massif<br />
which comprises up to 2 Ga old rocks and underpins the volcanic arc probably along<br />
the whole plateau (see in James & Sacks, 1999). The most prominent feature in the<br />
Altiplano is the 80 km wide Corque basin, a deep asymmetric sedimentary basin with<br />
thick and folded tertiary strata, exceeding 10 km thickness (Hérail et al., 1997) and<br />
evidencing orogeny dynamics during Miocene’s pronounced compressive regime.<br />
To the east the basin is controlled by the Chuquichambi thrust system. The Coniri-<br />
Laurani fault system marks the transition to the Eastern Cordillera, the rough eastern<br />
flank of the plateau. Farther to the east the Interandean Zone and Main Andean Thrust<br />
separate the Altiplano plateau from the actual deformation front in the Subandean,<br />
with thin-skinned folding and thrusting holding the record of shortening rates within<br />
the study area (Allmendinger et al., 1997).<br />
After the Oligocene eastward shift of the magmatic arc, Neogene volcanism first relived<br />
in the Eastern Cordillera 27 Ma ago and afterwards encompassed the whole plateau region.<br />
Neogene volcanism of the Altiplano plateau is composed of three main units:<br />
Pliocene to recent stratovolcanoes are erupting in the main arc and Miocene to recent<br />
backarc calderas and mafic mongenetic pulses are distributed in the arc and backarc.<br />
Intensive ignimbritic volcanism erupted 10-4 Ma ago at the Altiplano-Puna transition<br />
between 21-24◦S and formed one of the greatest ignimbrite provinces of the earth, the<br />
Altiplano-Puna-Volcanic-Complex. Recent ignimbrite eruptions are older than 1 Ma.<br />
The high water, phenocrystal and silicate content signify long storage in huge crustal<br />
magma chambers with pronounced fluidal circulation.<br />
Holocene volcanic activity is less than in the Puna where the lithosphere probably is<br />
50 km thinner (Whitman et al., 1996). Volcanism in the closer survey area is categorised<br />
from dormant (Parinacota) to solfataras state with intensive fumarolic activity<br />
at Guallatiri volcano. The profile volcano Parinacota may possess a stable magma<br />
chamber with only minor magmatic input.<br />
Glance on data analysis and 2-D inversion<br />
We measured horizontal components of the electromagnetic field as well as the vertical<br />
magnetic field for long periods between 10 and 20 000 s. Data analysis and its interpretation<br />
by 2-D inversion has already been described by Brasse & Eydam (2008a) and<br />
will not be replicated in this article.<br />
Some important results are just mentioned: We could not resolve subduction-related<br />
features near the slab, like fluid curtains or fluid and melt pathways, due to significant<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
220
Re<br />
Period: 993 s<br />
Eydam & Brasse: Discussion on backarc mantle melting<br />
Figure 3: Real part induction arrows in<br />
Wiese convention for one chosen period<br />
demonstrating the remarkable good alignement<br />
of arrows in profile direction. Arrows<br />
in the forearc are deflected; they are pointing<br />
parallel to the coast.<br />
3-D effects in the forearc. This can easily be seen in the behavior of real part induction<br />
arrows pointing parallel to the coast (fig. 3) which is also observed in other coastal regions<br />
in Chile so that one might call them ’Chile arrows’. Brasse et al. (2008b) showed<br />
that they can be explained by anisotropy in the forearc crust. Seven forearc station<br />
data were thus discarded from further interpretation by 2-D inversion.<br />
Induction arrows as well as the analysis of impedance data, which reveals phasesensitive<br />
skew values below 0.3 for plateau sites, suggest fairly good electrical 2-D<br />
approximation for the plateau region. We observe a remarkable good alignement of<br />
middle to long period real part induction arrows in profile direction, as seen in figure 3.<br />
Correspondingly, electrical strike derived by impedance data is fairly perpendicular<br />
to the profile. Real arrows disappear in the central Altiplano where resistivities are<br />
signifcantly reduced for all sites and rotation angles and follow a well marked downward<br />
trend for longer periods indicating assumably well conducting features at greater<br />
depths.<br />
2-D inversions were performed with the inversion program of Rodi & Mackie (2001).<br />
Accounting for the consistency of induction arrows in the Altiplano, tipper weight was<br />
set high by assigning a small (absolute) error floor of 0.02. Error floors for resistivities<br />
and phases were set to 20% and 5% in order to overcome the static shift problem.<br />
Reliable models are achieved after numerous experiments inculding tests of starting<br />
models, specifying the regularisation term and further resolution tests (see Eydam,<br />
2008).<br />
Discussion on backarc mantle melting<br />
The final model, shown in figure 4, is obtained by jointly inverting tippers, TE and<br />
TM mode apparent resistivities and phases. Static effects were accounted for by using<br />
the program internal adjustment. Fair regularisation parameters lie around 20. Model<br />
RMS is 1.80, with larger misfits at the electrically less 2-D plateau borders (see Brasse<br />
& Eydam, 2008a).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
221
depth (km)<br />
Eydam & Brasse: Discussion on backarc mantle melting<br />
SW NE<br />
Moho<br />
600°C<br />
W. Cordillera Altiplano<br />
E. Cordillera<br />
fluids<br />
&<br />
melt<br />
A1 A2 A3<br />
C1<br />
B C2<br />
MASH - zone<br />
D<br />
mantle convection<br />
6-7vol% melt<br />
WB-zone<br />
San-Andrés<br />
Fault<br />
lithospheric<br />
mantle layer?<br />
not<br />
resolved<br />
distance (km)<br />
Coniri-Laurani<br />
Fault<br />
fluids & melt?<br />
Figure 4: Final resistivity model with suspectable fluid & melt dynamics - Ai: Corque and minor<br />
basins, B: Arequipa block(?), Ci: structures beneath the Eastern Cordillera, D: backarc mantle magma<br />
reservoir plus rise of fluids and melts. Circles indicate location of mayor earthquakes (M > 4.5, Engdahl<br />
& Villaseñor, 2002). WB denotes Wadati-Benioff zone. MASH-zone stands for melting, assimilation,<br />
storage and homogenisation zone of arc magmas. Below: RMS for the inverted station data.<br />
The large upper mantle conductor D is interpreted as an image of partial melts triggered<br />
by fluid influx from the slab. Conductivities range up to above 1 S/m and are<br />
definitely required by data down to 115 km depth, particularly to fit the downward<br />
trends in long period app. resistivities; they are consistent with data down to slab<br />
depths. This implies high melting rates of more than 6vol% (Eydam, 2008).<br />
Lithospheric mantle material should stabilise magma and trap hot fluids which otherwise<br />
would rapidly trigger wide scale crustal melting like this is the case in the southern<br />
Altiplano (e.g., ANCORP Working Group, 2003). This is consistent with seismic data<br />
from Dorbath & Granet (1996) who resolve lithospheric characteristics just beneath<br />
the Altiplano Moho.<br />
The source region of subduction related magmatism is offset from the arc by almost<br />
100 km, implying fluid and melt storage near the Moho (in so called MASH-zones) and<br />
non-vertical rise of only few fluidal and molten material towards the volcanoes. This<br />
accords to low volcanic productivity in the closer survey area, while lavas are bearing<br />
highly modified mantle signatures (Wörner et al., 1992). Magma and fluid lateral motion<br />
should be enforced by internal convections.<br />
This ’scenario’ implies a widely hydrated crust and upper mantle beneath the high<br />
plateau which is consistent with other geophysical anomalies, like high heat flow densities<br />
(e.g., Springer & Förster, 1998), neagtive Bouguer anomalies (Tassara et al., 2006),<br />
slow seismic velocities in a thick and highly absorbing crust (e.g., Beck & Zandt, 2002)<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
222<br />
?<br />
C3<br />
<br />
4<br />
3<br />
lg m 2<br />
1<br />
0
component of<br />
oceanic plate<br />
sediments<br />
crust<br />
mantle<br />
total<br />
Eydam & Brasse: Discussion on backarc mantle melting<br />
water import<br />
per m<br />
Nazca plate [kg]<br />
2<br />
1 700<br />
170 000<br />
250 000<br />
1 000 000<br />
421 700<br />
1 171 700<br />
in percentage of<br />
imported water<br />
20<br />
55<br />
60<br />
water release in 80-180 km depth<br />
per subducted<br />
meter [kg]<br />
340<br />
93 500<br />
150 000<br />
600 000<br />
243 840<br />
693 840<br />
per Ma [10 kg] 2) 6<br />
32<br />
8 883<br />
14 250<br />
57 000<br />
23 165<br />
65 915<br />
3)<br />
in neogene cycle,<br />
6<br />
WC = VA [10 kg]<br />
832<br />
230 958<br />
1) 1) 1) 1)<br />
2) 2) 2) 2)<br />
370 500<br />
1 482 000<br />
602 290<br />
1 713 790<br />
Table 1: Water budget for main dehydration depths in the central Andean subduction zone calculated<br />
by using 1) a standardised mantle hydration model after Rüpke et al. (2004), 2) a mantle hydration<br />
model deduced from local data after Ranero & Sallarès (2004) and 3) by using averaged subduction<br />
rates after Somoza (1998). WC - Western Cordillera, VA - main volcanic arc.<br />
and a strong mantle signature of geothermal fluids (Hoke et al., 1994).<br />
Estimation of arc and backarc water release<br />
A critical mind might state that the ability of water import is probably not sufficient<br />
to nurture such wide scale melting of mantle material. Table 1 shows the water balance<br />
for slab dewatering at main dehydration depths between 80-180 km, deduced from intermediate<br />
depths microseismicity published in David (2007) and seen in figure 7.<br />
For the calculation we used hydration estimates after Rüpke et al. (2004) modified by local<br />
data (fig. 5). Mantle hydration is poorly definable and depends on depth and density<br />
of bending related faults as well as on the thermal structure of the oceanic lithosphere.<br />
Therefore estimates based on regional seismic and gravimetric data evaluated by Ranero<br />
& Sallarès (2004) are four times higher than those assumed by Rüpke et al. (2004) who<br />
standardised hydration levels with regards to the age of the lithosphere. Water balances<br />
are presented for both models.<br />
The presented water releases are based on results of chemo-thermo-mechanical modellings<br />
performed by Rüpke et al. (2004) who solved for slab dehydration during sub-<br />
moho<br />
fault<br />
pelagic<br />
sediments<br />
sheeted dykes<br />
& pillow lavas<br />
1)<br />
100 m 7.3wt% H2O gabbro intrusions 4 km ~ dry<br />
serpentinized<br />
peridotite<br />
1 km 2.7wt% H O<br />
2km 1wt%HO<br />
2<br />
2)<br />
~23 km<br />
2<br />
2.5wt% H O<br />
Figure 5: Oceanic lithosphere with estimated water content in sediments, crust and mantle after<br />
Rüpke et al. (2004) modified by local data: 1) the thickness of sedimentary deposits after von Huene<br />
et al. (1999) and 2) mantle hydration after Ranero & Sallarès (2004).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
223<br />
2<br />
2)
Eydam & Brasse: Discussion on backarc mantle melting<br />
duction regarding latent heat consumption and neglecting shear heating1 (fig. 6).<br />
Thereafter, one column of one square metre Nazca lithosphere releases 700 t water at<br />
main dehydration depths (80-180 km) which sums up to 1.7 Gt water regarding the entire<br />
Neogene cycle of Nazca plate subduction starting 26 Ma ago (Sempere et al., 1990).<br />
Using the molar volume of water at 120 km depths, this mass fills a cross-section of<br />
more than 1500 km2 which finally lies in the dimension of the mantle magma reservoir.<br />
Phases of oroclinal bending might have been important for the water budget, because<br />
bending related faults should have opened parallel as well as perpendicular to the trench<br />
so that large amounts of water could have been fixed in the mantle.<br />
Slab-derived water in the mantle wedge should thus be sufficient to trigger a wide<br />
scale melting of mantle material, well noticing the presumably prior hydration of the<br />
Western Cordillera as pre-Neogene backarc. If the residual water, retained in the<br />
oceanic lithosphere, is released as well, the cross-section will expand for additional<br />
170 km 2 in case of complete crustal dehydration resp. 880 km 2 for complete mantle<br />
depth [km]<br />
retained water [%]<br />
100<br />
200<br />
100<br />
50<br />
T(°C)<br />
1300<br />
1100<br />
900<br />
700<br />
500<br />
300<br />
100<br />
100 200 300 400<br />
distance [km]<br />
crust<br />
sediments<br />
mantle<br />
50 100 150 200 250<br />
depth [km]<br />
Figure 6: Dehydration of oceanic lithosphere<br />
during subduction, presented above) as main<br />
depths’ water release of sediments, crust and<br />
mantle and below) as percentage of retained water<br />
respectively; after Rüpke et al. (2004).<br />
a water-conserving reaction to mantle phase A (Eydam, 2008).<br />
dehydration which points out the role<br />
of oceanic mantle for subduction zone<br />
dynamics.<br />
The double seismic layer shown in figure<br />
7 allows allocating the events to stability<br />
limits of mayor hydrous minerals in the<br />
oceanic lithosphere which can be used to<br />
predict slab temperatures at depths.<br />
Crustal seismicity stops around 140 km<br />
depth, whereas seismicity in the oceanic<br />
mantle is ongoing until the stability limit<br />
of serpentine in 180 km. Lawsonite is the<br />
only hydrous mineral in the crust which is<br />
stable down to 140 km depth. Lawsonite<br />
decay starts at the hot interplate boundary<br />
at shallower dephts and proceeds during<br />
subduction to deeper slab layers where temperatures<br />
at 140 km depth should exceed<br />
730◦C. Serpentine as the main hydrous<br />
mantle phase should still be present at<br />
180 km depth where it transforms through<br />
These estimates correspond to a fairly warm subduction system although the Nazca<br />
plate is quite old (55 Ma). Shear heating at the interplate boundary may elevate<br />
temperatures in the oceanic crust, because lubricates like sediments are sparse in the<br />
1 Modifying the mantle hydration model to higher estimates after Ranero & Sallarès (2004) will change<br />
the modelling results to even larger water releases than calculated in table 1, considering that dehydrations<br />
are mostly exothermic reactions.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
224
Eydam & Brasse: Discussion on backarc mantle melting<br />
central Andean subduction zone as a consequence of its arid climate (Lamb & Davis,<br />
2003). Thus more water than listed in table 1 may be released into the mantle wedge.<br />
Dehydrations finish below 200 km depth, suggesting that slab-derived fluids supply<br />
partial melting of mantle peridotite more than 60 km laterally away from their source<br />
region. Might this be possible?<br />
Discussion on fluid mobility<br />
There are several models describing fluid mobility in subduction zones like fluid channeling<br />
along cracks opened by dehydration induced hydrofracturing, lateral transports<br />
in the coupled mantle wedge, porous flows and thermally induced convections in the<br />
asthenospheric wedge.<br />
At considered depths hydrofractured faults shouldn’t reach far into the mantle and<br />
a coupled mantle wegde should be unrealistically thick to explain the requested<br />
transportation lengths. Porous flows start when interstitial fluids or melts interconnect<br />
depending on temperature, fluid salinity and melt viscosity. Deep slab-derived fluids<br />
are highly saline (Scambelluri & Philippot, 2001) and temperatures below 140 km<br />
depth definitively exceed 620◦C (see above), the minimum temperature of pure water<br />
interconnection in an olivine matrix (Mibe et al., 1999).<br />
Water enhances mantle convection via lowering the peridotitic solidus in the mantle<br />
wedge. This effect may be more pronounced than expected so far, considering the<br />
very slow reaction kinetics of olivine melting in laboratory experiments (Grove et al.,<br />
2006). The correction of peridotitic solidus to much lower temperatures, e.g. to 860◦C at 100 km depth, implies that first water-enriched melts may be generated just above<br />
the slab. The melt fraction should be minor due to fast segregation of the low-viscous<br />
melts into lower and hotter mantle regions.<br />
Additionally, water import in the central Bolivian orocline might have been temporarily<br />
elevated in phases of its oroclinal bending leading to a significant reduction of<br />
mantle viscosity and enabling effective material transports via thermal convections,<br />
like it is modelled for some general case studies by Gerya et al. (2006). Saline<br />
fluids and melts may be transported far into the hotter backarc mantle where melt<br />
fractions rise and segregation rates decrease and fluids and melts become detectable<br />
by magnetotellurics.<br />
Models of mantle convection deal with numerous low constrainted parameters, like<br />
the degree of mantle hydration, shear heating at the interplate boundary, latent<br />
heat produced by mineral reactions or advective heat transports by fluids and melts.<br />
Therefore mantle rheology is only poorly definable and viscosity values may differ over<br />
several magnitudes.<br />
Water release at intermediate depths plays thus an important role in subduction zone<br />
dynamics. Currie & Hyndman (2006) even postulate that hot backarc mantles are a<br />
fundamental characteristic of an ocean-continent subduction zone. And indeed, other<br />
magnetotelluric images from the central Andean backarc mantle seem to concur with<br />
this: Schwarz & Krüger (1997) modelled 0.02 S/m conductive mantle features at the<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
225
Eydam & Brasse: Discussion on backarc mantle melting<br />
Profil<br />
E4<br />
E5<br />
Depth [km]<br />
Depth [km]<br />
0<br />
100<br />
Amp<br />
Ser<br />
600°C<br />
< 40°<br />
Law<br />
730°C<br />
E4<br />
200<br />
0 200 400<br />
0<br />
100<br />
E5<br />
200<br />
0 200<br />
Distance [km]<br />
400<br />
Figure 7: Local microseismic events (M > 1) after David (2007) with map of projection lines E4, E5<br />
and MT profile. Above: Events allocated to decay temperatures of hydrous minerals, this are in the<br />
crust: Amp - amphibole and Law - lawsonite and in the oceanic mantle: Ser - serpentine. Note the<br />
double seismic layer.<br />
transition to the Puna around 23 ◦ S and northwards at 20-21 ◦ S upper mantle material<br />
seems to be quite conductive as well with values over 0.05 S/m (Lezaeta, 2001), though<br />
electromagnetic fields are severely attenuated by a crustal high conductivity zone here<br />
(Brasse et al., 2002).<br />
Discussion on the mobility of the slab itself<br />
The Nazca plate subducts along the whole margin with small dip angles between 25-30 ◦<br />
or even horizontally in greater depths at some segments (so called ’flat subduction’)<br />
where volcanism above is inactive.<br />
Depth [km]<br />
0<br />
100<br />
Nazca Plate<br />
A (25-20 Ma)<br />
Flat to normal slab transition (25-20 Ma)<br />
uplifting<br />
volcanic front<br />
migration<br />
hydrated lithosphere<br />
200<br />
0 200 400<br />
Distance [km]<br />
Eastern<br />
Cordillera<br />
melt<br />
asthenospheric<br />
influx & slab -<br />
steepening<br />
600<br />
Figure 8: ’Mobile slab model’ for the central Andes after<br />
James & Sacks (1999). Due to retarded dehydration at<br />
intermediate depths, slab flattens and dewaters sparsely,<br />
insufficient to melt but to hydrate the base of the overlying<br />
lithosphere. Slab compacts and steepens again when<br />
hot asthenospheric material influx and triggers effective<br />
dehydrations; magmatism and volcanism above develop.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
226<br />
Thus, slab pull and buoyancy<br />
forces are fairly equilibrated and<br />
the slab may react sensible to density<br />
changes due to retarded dehydrations;<br />
it gets mobile in geological<br />
time scales.<br />
Such ’chemically controlled slab<br />
mobility’ was already discussed by<br />
James & Sacks (1999) and applied<br />
on central Andean history by supposing<br />
a period of flat subduction<br />
35-25 Ma ago, basically due to<br />
absent lava dates for this period<br />
(fig. 8).<br />
Remarkable is the slight slab steepening<br />
in the central orocline (fig. 7).
Depth [km]<br />
Depth [km]<br />
Depth [km]<br />
0<br />
100<br />
A<br />
Eydam & Brasse: Discussion on backarc mantle melting<br />
Nazca Plate<br />
Western<br />
Cordillera<br />
melt<br />
200<br />
oroclinal bending (16-10 Ma)<br />
0 200 400<br />
0<br />
100<br />
B<br />
formation of backarc<br />
melt reservoire (< 10-6 Ma)<br />
huge mantle<br />
water release<br />
200<br />
0 200 400<br />
0<br />
100<br />
C<br />
Altipano<br />
pulse of mafic<br />
volcanism<br />
melt thermal weakening<br />
melt<br />
fluid-melt-path<br />
melt<br />
200<br />
isolation of backarc<br />
melt reservoire (< 6 Ma)<br />
slab steepening<br />
0 200<br />
Distance [km]<br />
400<br />
Figure 9: Attempt to reconstruct Neogene Nazca<br />
plate subduction in the Bolivian orocline:<br />
A) Increased hydration of oceanic Nazca plate<br />
mantle due to significant oroclinal bending.<br />
B) 2-4 Ma later, highly hydrated mantle reaches<br />
dehydration depths of serpentine (100-180 km) where<br />
the released fluids trigger high-grade partial melting<br />
in the backarc. Backarc melts rise vertically as well as<br />
non-vertically to the arc via internal convections.<br />
C) Slab steepens during dehydration. The voluminous<br />
backarc reservoir persists but is maintained by<br />
less fluids. Volcanic productivity decreases.<br />
Tracing the regression line of shallow Wadati-Benioff events (z < 100 km) to greater<br />
depths, this line cuts the mantle conductor at dehydration depths of serpentine<br />
(z = 180 km), the most important hydrous mantle mineral which may bound huge<br />
amounts of water. Therefore backarc peridotitic melting may have been triggered<br />
just above the slab by enormous fluid influx whereby the slab was compacting and<br />
steepening slowly to its actual position (fig. 9). Today, the magma reservoir should be<br />
maintained by much less fluidal material and probably cools down.<br />
This scenario implies a prior period of extensive oceanic mantle hydration which might<br />
have occured during phases of oroclinal bending. Bending dates are hard to fix, probably<br />
block rotations in the closer survey area give a hint, they were dated at 12 Ma<br />
(Wörner et al., 2000).<br />
Using an average for ancient subduction rates of 10 cm/a, evaluated by Somoza (1998),<br />
and a constant slab dip of 25◦ , those huge amounts of fixed water will have started<br />
to deliberate after 2.4 Ma of subduction (z ≈ 100 km) and will have reached depths<br />
of high-grade partial backarc melting (z ≈ 180 km) further 1.9 Ma later. Surficial expression<br />
of high-grade mantle melting are Pliocene to recent pulses of mafic andesites<br />
erupting in the Bolivian Altiplano backarc (Davidson & de Silva, 1994). Furthermore<br />
backarc mantle magma might have been transported via internal convections to the<br />
sub-arc region and have also erupted at the main volcanic arc. However, the basal input<br />
into the deep magma system of the recent strato-volcanoes is only minor (Wörner<br />
et al., 2000).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
227
Eydam & Brasse: Discussion on backarc mantle melting<br />
Conclusion<br />
The huge backarc mantle conductor imaged by 2-D inversion of magnetotelluric data<br />
from the central Bolivian orocline inspired reflections about fluid and melt dynamics<br />
in the central Andean subduction zone. Among the two approaches considered here<br />
to explain the remarkable offset between intermediate depths mircoseismic events and<br />
the imaged magma reservoir, sub-arc and -backarc water release plays a key role for<br />
subduction zone dynamics.<br />
In the central Bolivian orocline huge amounts of water might have been released at<br />
main slab dehydration depths some million years after oroclinal bending occurred where<br />
faults should have opened not only parallel but also perpendicular to the trench allowing<br />
deep water infiltration and fixation in the oceanic mantle. This implies on the one hand<br />
that mantle viscosity is definitely lowered and leads to significant material transports<br />
due to thermal convections, whereby fluids and melts may be widespreaded in the<br />
arc and backarc mantle. On the other hand, enhanced slab dehydration and slab<br />
compacting may have been followed by slab steepening and the imaged mantle melting<br />
process isn’t supplied by much fluidal material anymore and therefore probably cools<br />
down.<br />
References<br />
Allmendinger, R.W., Jordan, T.E., Kay, S.M., & Isacks, B.L. 1997. The evolution of<br />
the Altiplano-Puna plateau of the Central Andes. Annual Review of Earth and Planetary<br />
Sciences, 25, 139–174.<br />
ANCORP Working Group. 2003. Seismic imaging of a convergent continental margin<br />
and plateau in the Central Andes (ANCORP’96). Journal of Geophysical Research, 108<br />
(B7), 2328, doi:10.1029/2002JB001771.<br />
Beck, S.L., & Zandt, G. 2002. The nature of orogenic crust in the Central Andes. Journal<br />
of Geophysical Research, 107(B10), doi:10.1029/2000JB000124.<br />
Brasse, H., & Eydam, D. 2008a. Electrical conductivity beneath the Bolivian Orocline and<br />
its relation to subduction processes at the South American continental margin. Journal of<br />
Geophysical Research, 113(B07109), doi:10.1029/2007JB005142.<br />
Brasse, H., Lezaeta, P., Rath, V., Schwalenberg, K., Soyer, W., & Haak, V.<br />
2002. The Bolivian Altiplano conductivity anomaly. Journal of Geophysical Research,<br />
107(B5), doi:10.1029/2001JB000391.<br />
Brasse, H., Kapinos, G., Li, Y., Mütschard, L., Soyer, W., & Eydam, D. 2008b.<br />
Structural electrical anisotropy in the crust at the South-Central Chilean continental margin<br />
as inferred from geomagnetic transfer functions. Physics of the Earth and Planetary<br />
Interiors, 173, 7–16.<br />
Currie, C.A., & Hyndman, R.D. 2006. The thermal structure of subduction zone back<br />
arcs. Journal of Geophysical Research, 111, doi:10.1029/2005JB004024.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
228
Eydam & Brasse: Discussion on backarc mantle melting<br />
David, C. 2007. Comportement actuel de l’avant arc et de l’arc du coude de Arica dans<br />
l’orogénèse des Andes Centrales. PhD thesis at university Toulouse III.<br />
Davidson, J.P., & de Silva, S.L. 1994. Late Cenozoic magmatism of the Bolivian Altiplano.<br />
Contributions to Mineralogy and Petrology, 119 (4), 387–408.<br />
Dorbath, C., & Granet, M. 1996. Local earthquake tomography of the Altiplano and<br />
the Eastern Cordillera of northern Bolivia. Tectonophysics, 259, 117–136.<br />
Elger, K., Oncken, O., & Glodny, J. 2005. Plateaustyle accumulation of deformation:<br />
Southern Altiplano. Tectonics, 24, doi:10.1029/2004TC001675.<br />
Engdahl, E.R., & Villaseñor, A. 2002. Global Seismicity: 1900-1999. In: Int. Handbook<br />
of Earthquake and Engineering Seismology, Part A (eds. Lee, W.H.K., Kanamori, H.,<br />
Jennings, P.C. & Kisslinger, C.). Academic Press, 665–690.<br />
Eydam, D. 2008. Magnetotellurisches Abbild von Fluid- und Schmelzprozessen in Kruste und<br />
Mantel der zentralen Anden. Diploma thesis at Freie Universität Berlin.<br />
Gerya, T.V., Connolly, J.A.D., Yuen, D.A., Gorczyk, W., & Capel, A.M. 2006.<br />
Seismic implication of mantle wedge plumes. Earth and Planetary Science Letters, 156,<br />
59–74.<br />
Grove, T. L., Chatterjee, N., Parman, S. W., & Médard, E. 2006. The influence of<br />
water on mantle wedge melting. Earth and Planetary Science Letters, 249, 74–89.<br />
Gutscher, M.-A., Spakman, W., Bijwaard, H., & Engdahl, E. R. 2000. Geodynamics<br />
of flat subduction: Seismicity and tomographic constraints from the Andean margin.<br />
Tectonics, 19(5), 814–833.<br />
Hoke, L., Hilton, D., Lamb, S., Hammerschmidt, K., & Friedrichsen, H. 1994. 3 He<br />
evidence for a wide zone of active mantle melting beneath the Central Andes. Earth and<br />
Planetary Science Letters, 128, 341–355.<br />
Hérail, G., Rochat, P., Baby, P., Aranibar, O., Lavenu, A., & Masclez, G.<br />
1997. El Altiplano Norte de Bolivia, evolución geológica terciaria. In: El Altiplano: ciencia<br />
y conciencia en los Andes, Actas 2. Simposio Internacional Estudios Altiplánicos, Arica<br />
1993. R. Charrier, P. Aceituno, M. Castro, A. Llanos and L.A. Raggi (eds). Universidad<br />
de Chile: Santiago. 33-44.<br />
James, D.E., & Sacks, I.S. 1999. Cenozoic formation of the Central Andes: A geophysical<br />
perspective. Society of Economic Geology, Special Publication, 7, 1–26.<br />
Klotz, J., Abolghasem, A., Khazaradze, G., Heinze, B., Vietor, T., Hackney,<br />
R., Bataille, K., Maturana, R., Viramonte, J., & Perdomo, R. 2006. Longterm<br />
signals in the present-day deformation field of the central and southern Andes and<br />
constraints on the viscosity of the earths upper mantle. In: The Andes: Active Subduction<br />
Orogeny. O. Oncken et al. (Eds.). Berlin: Springer. 65–89.<br />
Kurtz, R.D., Delaurier, J.M., & Gupta, J.C. 1986. A magnetotelluric sounding across<br />
Vancouver Island detects the subducting Juan de Fuca plate. Nature, 321, 596–599.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
229
Eydam & Brasse: Discussion on backarc mantle melting<br />
Lamb, S., & Davis, P. 2003. Cenozoic climate change as a possible cause for the rise of the<br />
Andes. Nature, 425, 792–797.<br />
Lezaeta, P. 2001. Distortion analysis and 3-D modeling of magnetotelluric data in the<br />
Southern Central Andes. PhD thesis at Freie Universität Berlin.<br />
Mibe, K., Fujii, T., & Yasuda, A. 1999. Control of the location of the volcanic front in<br />
island arcs by aquous fluid connectivity in the mantle wedge. Nature, 401, 259–262.<br />
Muñoz, N., & Charrier, R. 1996. Uplift of the western border of the Altiplano on a<br />
west-vergent thrust system, Northern Chile. Journal of South American Earth Sciences,<br />
9, 171–181.<br />
Ranero, C.R., & Sallarès, V. 2004. Geophysical evidence for alteration of the crust<br />
and mantle of the Nazca Plate during bending at the North Chile trench. Geology, 32(2),<br />
549–552.<br />
Rodi, W., & Mackie, R.L. 2001. Nonlinear conjugate gradients algorithm for 2-D magnetotelluric<br />
inversions. Geophysics, 66, 174–187.<br />
Rüpke, L.H., Morgan, J.P., Hort, M., & Connolly, J.A.D. 2004. Serpentine and the<br />
subduction zone water cycle. Earth and Planetary Science Letters, 223, 17–34.<br />
Scambelluri, M., & Philippot, P. 2001. Deep fluids in subduction zones. Lithos, 55,<br />
213–227.<br />
Scheuber, E., Bogdanic, T., Jensen, A., & Reutter, K.J. 1994. Tectonic development<br />
of the North Chilean Andes in relation to plate convergence and magmatism since the<br />
Jurassic. In: Tectonics of the Southern Central Andes. K.J. Reutter, E. Scheuber and P.J.<br />
Wigger (eds). Berlin: Springer. 121–139.<br />
Schwarz, G., & Krüger, D. 1997. Resistivity cross section through the southern Central<br />
Andes as inferred from magnetotelluric and geomagnetic deep soundings. Journal of<br />
Geophysical Research, 102, 11957–11978.<br />
Sempere, T., Hérail, G., Oller, J., & Bonhomme, J. 1990. Late Oligocene - early<br />
Miocene major tectonic crisis and related basins in Bolivia. Geology, 18, 946–949.<br />
Somoza, R. 1998. Updated Nazca (Farallon)-South America relative motions during the<br />
last 40 My: Implications for the mountain building in the central Andean region. Journal<br />
of South American Earth Sciences, 11(3), 211–215.<br />
Springer, M.H., & Förster, A. 1998. Heatflow density across the Central Andean subduction<br />
zone. Tectonophysics, 291, 123–139.<br />
Tassara, A., Götze,H.J.,Schmidt,S.,&Hackney,R.2006. Three-dimensional density<br />
model of the Nazca plate and the Andean continental margin. Journal of Geophysical<br />
Research, 111(B6), doi:10.1029/2005JB003976.<br />
Varentsov, I.M., Golubev, N.G., Gordienko, V.V., & Sokolova, E.Y. 1996. Study<br />
of deep geoelectrical structure along EMSLAB Lincoln-Line. Phys. Solid Earth, 32, 375–<br />
393.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
230
Eydam & Brasse: Discussion on backarc mantle melting<br />
von Huene, R., Weinrebe, W., & Heeren, F. 1999. Subduction erosion along the North<br />
Chile margin. Journal of Geodynamics, 27, 345–358.<br />
Wannamaker, P.E., Booker, J.R., Jones, A.G., Chave, A.D., Filloux, J.H., Waff,<br />
H.S., & Law, L.K. 1989. Resistivity cross-section through the Juan de Fuca subduction<br />
system and its tectonic implications. Journal of Geophysical Research, 94, 14127–14144.<br />
Whitman, D., Isacks, B.L., & Kay, S.M. 1996. Lithospheric structure and along-strike<br />
segmentation of the Central Andean Plateau: seismic Q, magmatism, flexure, topography<br />
and tectonics. Tectonophysics, 259, 29–40.<br />
Wigger, P.J., Schmitz, M., Araneda, M., Asch, G., Baldzuhn, S., Giese, P.,<br />
Heinsohn, W.D., Martinez, E., Ricaldi, E., Röwer, P., & Viramonte, J. 1994.<br />
Variation in the crustal structure of the Southern Central Andes deduced from seismic<br />
refraction investigations. In: Tectonics of the Southern Central Andes. K. J. Reutter and<br />
E. Scheuber and P. J. Wigger (eds). Berlin: Springer. 23–48.<br />
Wörner, G., Uhlig, D., Kohler, I., & Seyfried, H. 2002. Evolution of the West<br />
Andean Escarpment at 18 ◦ S (N. Chile) during the last 25 Ma: Uplift, erosion and collapse<br />
through time. Tectonophysics, 345, 183–198.<br />
Wörner, G., López-Escobar, L., Moorbath, S., Horn, S., Entenmann, J., &<br />
Davidson, J.D. 1992. Variaciones geoquímicas, locales y regionales, en el arco volcánico<br />
Andino del norte de Chile (17 ◦ 30’S-22 ◦ 00’S). Revista Geológica de Chile, 19(1), 37–56.<br />
Wörner, G., Hammerschmidt, K., Henjes-Kunst, F., Lezaun, J., & Wilke, H.<br />
2000. Geochronology ( 40 Ar/ 39 Ar, K-Ar and He-exposure ages) of Cenozoic magmatic<br />
rocks from Northern Chile (18-22 ◦ S): Implications for magmatism and tectonic evolution<br />
of the Central Andes. Revista Geológica de Chile, 27, 205–240.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
231
Thin Sheet Conductance Models from<br />
Geomagnetic Induction Data:<br />
Application to Induction Anomalies at the Transition from<br />
the Bohemian Massif to the West Carpathians<br />
VclavČerv a (vcv@ig.cas.cz), Světlana Kov čikov a , Michel Menvielle b and Josef Pek a<br />
a Inst. Geophys., Acad. Sci. Czech Rep., v.v.i., Prague, Czech Rep.<br />
b CEETP CNRS, Saint Maur des Fosses, France<br />
Abstract<br />
Thin sheet approximation of Earth’s conductive structures made it possible to quantitatively estimate effects of<br />
lateral conductivity variations in the Earth long before full 3-D electromagnetic modeling was practicable. The<br />
thin sheet approach is still useful when induction data with limited vertical resolution are to be interpreted on a<br />
surface. It especially refers to collections of long period induction arrows across large areas and geological units.<br />
For this purpose, inverse procedures, both linearized and stochatic, for a conductance distribution in a thin sheet have<br />
been suggested recently. We present a stochastic Monte Carlo inversion of geomagnetic induction data based on<br />
the bayesian formulation of the inverse problem. An example of the inversion of practical induction data from the<br />
transition zone between the Bohemian Massif and the West Carpathians suggests that an SW-NE anomalous induction<br />
zone observed above the eastern slopes of the Bohemian Massif admits explanation in terms of a phantom effect due<br />
to the superposition of fields of the strong SE-NW Carpathian conductivity anomaly to the east with NW-SE to W-E<br />
trending conductivity zones to the west that conform with the fault pattern of the eastern Bohemian Massif.<br />
1 Introduction<br />
Regional geoelectrical information on the contact zone between the Variscean Bohemian Massif and the Alpine Western<br />
Carpathians is mainly available from long-period geomagnetic transfer functions. Magnetotelluric data and broadband<br />
electromagnetic induction experiments are scarce in the region, mainly because of high level of the civilization<br />
noise all over the area, and also because of the lack of instrumentation in the past. Only recently, magnetotelluric<br />
experiments have been carried out in some subareas of the region, mainly for commercial targets (e.g., Voz r, 2005).<br />
Long-period geomagnetic induction data covering a period band of about one decade in the range of thousands<br />
of seconds cannot provide detailed geoelectrical information on structures beneath the region of interest. They are,<br />
however, suitable characteristics to be used to model large-scale horizontal conductivity distribution in the Earth’s<br />
crust or lithosphere, and thus to indicate regional lateral conductivity anomalies that are responsible for the observed<br />
induction pattern over the area under study. With data at periods of the order of thousands of seconds, with penetration<br />
depths starting at several tens of km for standard conditions of the continental lithosphere, a quasi 3-D thin sheet<br />
approximation of crustal conductivity structures is a reasonable induction model (e.g., Vasseur and Weidelt, 1977).<br />
The thin sheet approximation largely reduces the computational demands of the modeling procedure as compared with<br />
a full 3-D treatment and, moreover, bypasses some intrinsic difficulties of dealing with the geomagnetic induction data<br />
alone, especially their low sensitivity with respect to the normal layered background of the model.<br />
Recently, two methods of inversion of geomagnetic induction data for conductances in a thin sheet have been<br />
developed, one based on the non-linear conjugate gradient technique (Wang and Lilley, 2002) and the other on a<br />
Bayesian approach with Markov chain Monte Carlo (MCMC) method used for a stochastic sampling process (Grandis,<br />
2002). In this contribution, we are using the latter approach to analyze conductance distributions which are compatible<br />
with the geomagnetic induction data in the West Carpathians region and at its transition to the Bohemian Massif on<br />
the territories of the Czech Republic, Slovakia and Poland.<br />
More specifically, the data base of this study are induction response data at 150 field stations that were published<br />
earlier (Praus and Pěčov , 1991) covering the Bohemian Massif (BM), the Brunovistulicum (BV), and the West<br />
Carpathian region (WCP) and re-analyzed in (Pěčov and Praus, 1996). Spatial distribution of both the in-phase and<br />
the out-of phase induction vectors, contour maps of individual transfer functions (TF) and contour maps of anomalous<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
232
Latitude (deg)<br />
54<br />
53<br />
52<br />
51<br />
50<br />
49<br />
48<br />
47<br />
Mid-G<br />
Saxo-Thuringian<br />
Rheno-Hercynian<br />
Moldanubian region<br />
North-German-Polish Caledonides<br />
ine<br />
erma n C rystall R<br />
ise<br />
BM<br />
it of Alpi atio<br />
Alpine Front<br />
ne deform<br />
o r li<br />
ute<br />
outer limn<br />
mit of Varisceandefor tion<br />
East European<br />
Platform<br />
Holy Cross<br />
Mts<br />
8 10 12 14 16 18 20 22<br />
Longitude (deg)<br />
T<br />
eisseyre rnquist Line<br />
Vienna<br />
Basin<br />
ma<br />
-To<br />
Midmountains<br />
Pannonian<br />
Basin<br />
Bohemian Massif (BM ) with uni ts:<br />
Moldanubicum<br />
Teplá-Barrandian region<br />
Krušné hory-Thuringian<br />
region<br />
Bohemian Cretaceous<br />
basin<br />
West-Sudetic region<br />
Brunovistulicum (BV)<br />
including Silesicum, Moravicum<br />
and Brno Massif<br />
West Carpathians (WCP) with units:<br />
Tertiary basins<br />
(Carpathian foredeep,<br />
Pannonian basin)<br />
Outer (Flysch) Carpathians<br />
Klippen Belt<br />
Central (Inner) Carpathians<br />
Neovolcanites<br />
Figure 1: Geological scheme of the region of Central Europe relevant to our modeling study. The dashed-line rectangle<br />
shows the region in which the inversion for a laterally variable conductance distribution in a thin sheet is carried out.<br />
vertical field that were generated from the TF’s by the hypothetical field of different polarizations and systems of internal<br />
anomalous currents indicate the existence of two zones of anomalous induction at the eastern margin of the BM<br />
and near the boundary of the Carpathian plate (Kov čikov et al., 1997). The previous analyses, and the recent one<br />
performed by a new approach to imaging the induction data of the Wiese vectors at almost 1800 localities covering<br />
mainly the Central European area (Wybraniec et al., 1999) suggest that these anomalies might be connected with the<br />
North-German-Polish anomalous zone, representing an important part of the Trans-European Suture Zone (TESZ).<br />
The analysis of certain models of electrical conductivity distribution is performed to fit the anomalous features of<br />
the induction response data over the Central European area, specifically zones of anomalous induction in the eastern<br />
margin of the BM, across the entire block of the BV and near the margin of the Carpathian tectonic plate.<br />
The structure of the article is as follows: In Section 2, we give a short overview of the principal geological units of<br />
the region under study. Section 3 briefly summarizes the geoelectrical features of the region previously inferred from<br />
the geomagnetic induction data in the region. In Section 4, we formulate the thin sheet model used in the inversion.<br />
The principles of the Bayesian inversion of the geomagnetic induction data for the conductance distribution in the<br />
thin sheet are presented in Section 5. Section 6 then summarizes outputs of the Bayesian inversion for the induction<br />
data over the BM/BV/WCP region with indications on possible correlations of the conductance model with regional<br />
geological structures.<br />
2 Geological context of the induction studies<br />
Fundamental elements of the central European geological structure that are relevant to our modeling experiment are<br />
schematically displayed in Fig. 1 together with the major geological units over the Czech and Slovak Republics covered<br />
by induction response data involved in the modeling process.<br />
(i) The Tornquist-Teisseyre tectonic zone (TTZ) constitutes one substantial part of a number of fault zones and<br />
sutures found within the Trans-European Suture Zone (TESZ) that represents the most important geological boundary<br />
in Europe separating mobile Phanerozoic western terranes (Meso-Europe) from the Precambrian east European Craton.<br />
It is as clearly defined in the deep lithosphere as in the upper crust, Moho depths increase across this zone from 30 km<br />
beneath Variscan Europe to 45 km beneath the East European Craton. In contrast to the relatively cold eastern craton<br />
relatively high heat flow characterises Western Europe.<br />
(ii) The Bohemian Massif (BM) represents the easternmost consolidated block of the Variscan branch of the European<br />
Hercynides (Meso-Europe) that builds up Bohemia and the western part of Moravia. The major elements of<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
233
the BM are several SW-NE trending zones separated usually by deep-seated faults, which are reflected in the results<br />
of the deep seismic profiling, in the gravity, geomagnetic and heat flow maps (Suk et al.,1984). The fault structures<br />
are essentially parallel with the boundaries of individual Hercynian zones. The intersection of these zones with the<br />
second order system of NW-SE trending faults is responsible for the complicated block structure of the BM.<br />
(iii) The Carpathians belong to the young Tertiary Alpine-orogenic belt and they constitute the NE branch of the<br />
Alpides. Their boundary with the Eastern Alps runs along the Danube Valley. The northern boundary with the East<br />
European Platform is defined by the erosive margin of the flysch nappes. Our model covers the West Carpathians<br />
(WCP) and we distinguish here the inner (central) Carpathians, the Klippen zone, the outer (Flysch) Carpathians and<br />
the Carpathian foredeep (Fig. 1, dashed-line rectangle).<br />
(iv) The Brno unit, termed recently the Brunovistulicum (BV) (Dudek, 1980; Suk, 1995) lies in between those<br />
previously mentioned structural elements. The BV unit is assumed to be an independent geological structure forming<br />
the Precambrian (Cadomian) basement (Palaeo-Europe) of both the eastern part of the Hercynides (Variscides) of the<br />
BM and the Alpides, i.e. of the WCP in Moravia. Recent geophysical and geological data have shown that the BV<br />
and the whole Moravian block occupy a highly independent position and form a separate geological unit belonging<br />
probably to the Fenno-Sarmatian Platform (Dudek, 1980).<br />
3 Geoelectrical data and features of the region<br />
3.1 The data<br />
The experimental data were collected in 1970’s and 1980’s in a series of field experiments organized in co-operation<br />
with Czech, Slovak and Polish colleagues. The field measurements consisted in analogue recordings of magnetic<br />
transient variations at alltogether 150 temporary field sites over an area of about 500 × 250 km 2 . They were analysed<br />
in terms of induction arrows (Wiese, 1962). At each station, in-phase and out-of phase induction arrows have been<br />
estimated for periods in the range of 1200 to 5840 s. They correspond to the real and imaginary components of singlestation<br />
transfer functions between the horizontal and vertical components of the transient magnetic field at the station.<br />
A sample of the real and imaginary induction arrows at a particular period of T = 3860 s is presented in Fig. 2.<br />
3.2 Major conductivity anomalies<br />
As the magnetic Z-component is known to be highly sensitive to laterally inhomogeneous distribution of the internal<br />
electrical conductivity, maps of these vectors provide us with a view of the changing anomalous behaviour of the<br />
Z-variation as function of frequency and location. Reversals of the arrows distinguish zones of anomalous induction,<br />
which often mark important geological features such as contacts between blocks with different geological histories of<br />
development, zones of past and recent tectonic activities, collision zones and etc. A qualitative analysis of the maps<br />
of induction arrows led to evidence two major conductivity anomalies in the area under study. Quantitative modeling<br />
allowed to characterize the conductivity distribution in the lithosphere that accounts for the observed induction arrows.<br />
The Carpathian anomaly, WCA The induction vectors across the WCP region show a clear perpendicular orientation<br />
with respect to a general trend of the anomalous induction zone localized along the external margin of the<br />
Carpathians Mts. chain (see WCA in Fig. 2. They show almost a perfect 180 o reversals of their azimuths above the<br />
anomalous induction zone. In the WCP, modules of the induction vectors situated to the south from the zero line are<br />
by about 25-50% larger than the corresponding induction vectors located to the north of the anomaly. The Carpathian<br />
14˚ 15˚ 16˚ 17˚ 18˚ 19˚ 20˚ 21˚ 22˚ 23˚<br />
51˚ 51˚<br />
Real arrow 0.5<br />
50˚ 50˚<br />
49˚<br />
WCA<br />
49˚<br />
EBMA<br />
48˚ 48˚<br />
14˚<br />
15˚<br />
16˚<br />
?<br />
17˚<br />
18˚<br />
19˚<br />
20˚<br />
21˚<br />
22˚<br />
23˚<br />
14˚ 15˚ 16˚ 17˚ 18˚ 19˚ 20˚ 21˚ 22˚ 23˚<br />
51˚ 51˚<br />
Imag arrow 0.5<br />
50˚ 50˚<br />
49˚ 49˚<br />
48˚ 48˚<br />
Figure 2: Sample of experimental real and imaginary induction arrows in the BM/BV/WCP region for the period of<br />
3860 s, with two main regional conductivity anomalies indicated, the West Carpathian conductity anomaly (WCA)<br />
and the conductivity anomaly on the eastern margin of the Bohemian Massif (EBMA).<br />
14˚<br />
15˚<br />
16˚<br />
17˚<br />
18˚<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
234<br />
?<br />
19˚<br />
20˚<br />
21˚<br />
22˚<br />
23˚
geoelectrical anomaly is constrained by the presence of high-conductivity rock series at depth of 10-25 km in different<br />
segments of the orogen. Different geological models explaining the presence of highly conductive rocks or solutions<br />
as sources of the anomalies have been suggested. The role of altered and/or fractured rocks, saturated with hot mineral<br />
waters, as well as the proximity of anomaly sources to boundaries of different crustal blocks, including those between<br />
the Carpathians and adjacent platforms, have been taken into account ( d m and Posp il, 1984; Jankowski et al.,<br />
1984; 1991; 2008; Hvo dara and Voz r, 2004). Other authors discuss possible connection between the Carpathian<br />
anomaly and the presence of metamorphosed coal bearing Carboniferous strata beneath the orogen or relation to the<br />
graphitized rocks occurring close to the boundaries of crustal blocks (Glover and Vine, 1992; Glover and Vine, 1995;<br />
´Zytko, 1997).<br />
The EBMA anomaly The induction vectors on the eastern margin of the BM are distinguished by the vectors<br />
oriented predominantly parallel to the general SW-NE trend of the anomalous zone. Also the reversal of the azimuths<br />
at the anomalous zone is rather poorly developed, only sudden changes of azimuths are observed at individual profiles.<br />
These facts are clear indications of a 3-D character of the conductivity distribution. The EBMA has been generally<br />
attributed to processes in the the subduction of the eastern margin of the BM beneath the WCP plate, but precise<br />
induction processes for generating the peculiar 3-D features of this anomaly are still unknown (e.g. Kov čikov et al.,<br />
2005). Speculating on the physical/geological sources of the anomaly may thus be premature.<br />
3.3 Results from numerical modeling<br />
From the equivalent current systems (Červ et al., 1997) it was concluded that the source depth of the induction<br />
anomalies can be about 18 km in the WCP region and about 10 km in the EBM/BV. These estimates are suggesting the<br />
source of the anomalies at shallower depth than those obtained previously by separating the magnetic field variations<br />
into internal and external parts (Pěčov and Praus, 1996) and applying the line current approximation (Jankowski et<br />
al., 1985).<br />
2-D models for induction vectors along the profiles crossing the WCP are summarized in (Jankowski et al., 1985).<br />
The models featured anomalous bodies with a cross-section× conductivity parameter of the order of 10 7 to 10 8 Sm,<br />
with the top of the bodies at depths beneath 12-15 km. In (Jankowski et al., 1991) the 2-D modeling was used for<br />
simultaneous modeling of the induction vectors and apparent resistivities, collected in a series of sites in the Polish<br />
section of the Carpathian Foredeep. In the latter models, the source of the WCA was situated at shallower depths, less<br />
than about 10 km, and hypothesized to be related to deep sediments of the Carpathian Foredeep. The 2-D inversion on<br />
the Carpathian data was firstly used in (Červ and Pek, 1981). The geoelectrical structural model along the DSS profile<br />
No VI, crossing the BM, BV and WCP, based on the MT and MV results was presented in (Červ et al., 1984).<br />
3.4 Depth of the asthenosphere<br />
The electrical asthenosphere, if present, is an additional structural feature that can affect the induction data especially<br />
at long periods. In the model derived from the P-wave residuals for the Bohemian Massif the depth of the lithosphereasthenosphere<br />
transition zone are between 90 and 140 km (Babu ka et al., 1988). From MT sounding a layer of<br />
increased electrical conductivity attributable to the asthenosphere was interpreted at depth between 100 and 150 km<br />
(Červ et al., 1984).<br />
The most likely depth of the conducting layer in the upper mantle in the Pannonian Basin region are between<br />
60 and 85 km in the central part of the depression. The depths seem to increase towards the flanks of the basin to<br />
about 100 km ( d m, 1976). In some parts (R ba-Ro nava tectonic line) the thickness of the litosphere reduces to<br />
even less than 60km ( d m, 1988). The regions of lithosphere thinning penetrate from the Pannonian Basin into<br />
inner parts of the WCP in several promontories. In the thinned part of the lithosphere in the West Carpathians the<br />
lithosphere-asthenosphere transition zone is at the depth 90-120 km (Červ et al., 1984).<br />
A gradual increase of the lithosphere thickness to 140-180 km occurs in the Outer Carpathians and father towards<br />
the margin of the East European Platform (Praus et al., 1990). In the Alps the depth of the asthenosphere varies from<br />
100 to 200 km (Praus et al., 1990).<br />
4 A possible thin sheet model<br />
In the subsequent inversion for a laterally non-uniform conductance, we will use a thin sheet model formally defined<br />
as follows:<br />
Let us consider a model consisting in a heterogeneous thin sheet at the surface of or embedded in a 1-D medium,<br />
hereafter called normal model. The normal model is defined by the conductivities σn(r) at any point r. In the thin<br />
sheet, the actual conductivity σ(r) differs from the normal one in a domain of interest Ω. InsideΩ, σa(r) denotes the<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
235
difference σ(r) − σn(r): σa(r) is the anomalous conductivity that is zero outside Ω. Solutions of the forward and<br />
inverse problems have already been published for such a model. They are briefly recalled in the next section.<br />
With regard to the experimental geomagnetic TF’s available, several aspects of the thin sheet model must be<br />
considered. First, the validity of the thin sheet hypothesis must be assessed. We assume that the anomalous electrical<br />
targets are localized mainly in the crust. Penetration depths for typical continental crustal sections should not be less<br />
than about 50 km in the period range of thousands of seconds (corresponds to the resistivity of 10 Ωm and period<br />
of 1000 s). Thus, the anomalous zone of 20-30 km below the Earth’s surface can be reasonably (though, may be,<br />
not completely unquestionably for highly conductive anomalous zones) approximated by a thin sheet. According to<br />
Bruton (1994), the size of the square tiles a used to discretize the anomalous subdomain Ω of the sheet should meet<br />
the condition aωμSmax ≪ 1. For the most extreme parameters in our models, say Tmin = 1200 sandSmax =<br />
5000 siemens, the size of the tiles should be much less than about 30 km. We typically use tiles with a =20to 25 km<br />
in our models, which meets the Bruton’s condition for most of the model situations.<br />
The second aspect is the depth of the thin sheet. As we are not able to employ a multiple sheet model in the<br />
inversion at present, we use a single thin sheet that integrates all the conductivity anomalies from the surface down to<br />
the crust. Thus, the sheet is situated on the surface of the Earth in our experiments. This may result in increased misfit,<br />
especially in imaginary induction arrows, if the anomalous current flow deep in the crust.<br />
The third aspect is the embedding normal structure. Due to large differences of the depth vs. resistivity sections<br />
of the two main region of our study, the BM and the WCP with the adjacent Pannonian Basin, we cannot suggest<br />
any single 1-D normal model down to the asthenospheric depths for the whole region. Therefore, we simplified the<br />
normal model into a two-layer structure with a poor conductor resistivity of several hundreds of Ωm up to 1000 Ωm)<br />
underlaid by a more conductive asthenosphere with resistivity within the range of 100-500 Ωm. The effect of the<br />
topography of the asthenospheric layer was checked independently by a 3-D modeling experiment with seismic data<br />
(Praus et al., 1990) taken to approximate the top of the asthenosphere. The checks showed that the effect of the<br />
asthenosphere is negligible in induction arrows for periods of the order of thousands of seconds unless the resistivity<br />
of the asthenosphere is less than about 10 Ωm.<br />
5 Bayesian Monte Carlo Markov Chain thin sheet inversion<br />
We present in this section the Bayesian Monte Carlo Markov Chain (MCMC) method we used to solve the inverse<br />
problem. For the sake of self-completeness, let us first briefly recall basic notions concerning Bayesian inversion and<br />
Markov chains behaviour. The readers are referred to Roussignol et al. (1993), Menvielle and Roussignol (1995), and<br />
Grandis et al. (1999; 2002), and references therein for more details.<br />
5.1 The Bayesian approach<br />
Let the a priori knowledge be the information available on the model before processing the data, and the a posteriori<br />
knowledge the information available after processing the data. A priori and a posteriori distributions account in a<br />
probabilistic way for the a priori and a posteriori knowledge respectively. In the Bayesian context, solving the inverse<br />
problem thus comes down to determining the a posteriori knowledge by updating the a priori knowledge with the<br />
information gleaned from the data (Box et Tiao, 1973; Berger, 1985; Press et al., 1989).<br />
Solving the inverse problem first requires the direct problem to be solved. Let F be the direct problem function<br />
which enables computation of the observations d for a model m. Assuming that the error δd is only related to the<br />
data acquisition, it becomes<br />
d = F (m)+δd<br />
The a priori knowledge is given by a probability distribution function (pdf ) P0(M = m | M ∈M) defined on<br />
the set of possible models M. The a posteriori probability for the parameter vector M to take the value m given the<br />
observations d is given by Bayes’ formula (Bayes, 1763; also see, e.g., Bolstad, 2004, for a more modern treatment).<br />
Noting P (M = m | D = d, M ∈M) the conditional probability of M given D, and assuming that the error δd is<br />
gaussian with standard deviation τ, it becomes<br />
<br />
<br />
||F (m) − d)||2<br />
exp − P0(m)<br />
P (M = m|D = d, M ∈M)=<br />
<br />
m∈M<br />
exp<br />
<br />
−<br />
2τ 2<br />
||F (m) − d)||2<br />
2τ 2<br />
<br />
P0(m)<br />
The value d of the random vector D corresponds to the observed data. Since d is not modified during the inversion<br />
process, we will from now on omit D in the expression of the probability distributions and denote the a posteriori pdf<br />
P (M = m|D = d, M ∈M) simply by Π(m).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
236<br />
(1)
The normalization constant, which appears in the denominator of eq. (1) is a sum over the set of possible models,<br />
the dimension of which is very large: for instance, it is equal to M × L for models with M parameters that can each<br />
take on L different values. It is actually very difficult, if not impossible to estimate directly. We choose to use a Monte<br />
Carlo Markov Chain (MCMC) simulation method to achieve this estimation.<br />
5.2 Markov chains<br />
In order to estimate the pdf Π(m), let us define a Markov Chain on the set M of possible models that have Π(m)<br />
as invariant probability. The marginal pdf ’s of Π(m) will be estimated by empirical averages from simulations of the<br />
Markov Chain.<br />
Consider a system which can take a certain number of states and evolves at random with time. At a given time,<br />
the state of the system can be described as a random variable M, the values of which, m, belong to M. The system is<br />
a Markovian process if, at any time, its future evolution depends only on its present state. It means that a Markovian<br />
system depends only on its past through its present state. Markov chains are a particular class of Markovian processes<br />
such that (i) the set M is finite or countable, and (ii) the successive times for the evolution are denoted by integers.<br />
Aseries{M(n),n =0, 1, 2,...,N} of random variables with values in a finite or countable space is then a Markov<br />
chain if the state of the system at the time n +1, M(n +1), depends only on its past through the state of the system<br />
at time n, M(n). The behaviour of a Markov chain is defined by the set of its transition probabilities,<br />
Pn(m, m ′ )=P (M(n +1)=m ′ | M(n) =m) ,<br />
where m and m ′ belong to M. If (i) the transition probabilities do not depend on time n, (ii) M is finite, and (iii) each<br />
possible state can be reached from any other, the chain is a homogeneous ergodic aperiodic Markov chain, and there<br />
exists one, and only one probability density function Π on M which is invariant for the Markov chain. When n<br />
increases towards infinity, the behavior of ergodic Markov chains is such that the average fraction of time at which the<br />
chain is at a state m (m ∈M) tends towards the invariant probability of m,i.e.Π(m).<br />
Methods for Bayesian inversion with Markov chains has been proposed for the 1-D magnetotelluric (Grandis et<br />
al., 1999) and DC problems (Schott et al., 1999) with the a priori of smooth variation of resistivity with depth, and for<br />
the thin sheet approximation (Roussignol et al. 1993; Grandis, 1994; Grandis et al., 2002). The present analysis of<br />
crustal conductivity over the transition from the Bohemian Massif to the West Carpathians relies on the inversion of<br />
induction arrows using the latter method, which is briefly described in the following sections.<br />
5.3 The thin sheet forward problem<br />
Let us consider a model consisting of a heterogeneous thin sheet at the surface of or embedded in a 1-D medium,<br />
specified in detail in the first two paragraphs of Section 4 above. Let E(r,ω,σ) and H(r,ω,σ) be the time Fourier<br />
transforms of the electric and magnetic field at point r and circular frequency ω, ω =2π/T, T period, for the actual<br />
conductivity structure σ. Using the Green kernel method, we can rewrite basic equations of electromagnetism as<br />
(Weidelt, 1975)<br />
<br />
E(r,ω,σ) = En(r,ω) − iωμ σa(r<br />
Ω<br />
′ ) G(r, r ′ ,ω) E(r ′ ,ω,σ(r ′ )) d 3 r ′ , (2)<br />
<br />
H(r,ω,σ) = Hn(r,ω)+ σa(r ′ )curl(G(r, r ′ ,ω)) E(r ′ ,ω,σ(r ′ ) d 3 r ′ , (3)<br />
Ω<br />
at any point r and any frequency ω. The quantity μ is the magnetic permeability of the vacuum in our models.<br />
G(r, r ′ ,ω) is a 3 × 3 complex matrix, named Green kernel, and represents the Fourier transform at frequency ω of the<br />
electric field created at point r by a unit dipole δ(r ′ ) located at point r ′ (Morse and Feshbach, 1953), curl(G(r, r ′ ,ω))<br />
is a 3 × 3 complex matrix obtained by taking the curl at point r of the field of the column vectors of the matrix<br />
G(r, r ′ ,ω).<br />
A numerical solution of the forward problem has been first proposed by Vasseur and Weidelt (1977) for a superficial<br />
thin sheet. The Vasseur and Weidelt’s (1977) approach can be extended to a thin sheet at any depth below the surface.<br />
The code was made operational by Tarits (1989). These solutions are based upon a digitization of Ω in K small square<br />
cells Pk, k = 1,...,K, in which conductivities, Green kernels, electric and magnetic fields are nearly constant.<br />
Let S(Pk) and Sn(Pk) be the actual and normal integrated conductivities (i.e. conductances) of the cell Pk, and<br />
Sak = S(Pk) − Sn(Pk), k =1,...,K, the anomalous conductance of Pk. Then basic equations become, using the<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
237
same notations for the sake of simplicity,<br />
E(r,ω,S) = En(r,ω) − iωμ <br />
H(r,ω,S) = Hn(r,ω)+ <br />
k<br />
k<br />
Sak G(r,Pk,ω) E(Pk,ω,S(r ′ )) |Pk|, (4)<br />
Sak curl(G(r,Pk,ω)) E(Pk,ω,S(r ′ )) |Pk| (5)<br />
at any point r and any frequency ω.<br />
Let us consider the linear system made up of equation (4) written at each point Pk of Ω, k =1,...,K.Given<br />
the distribution of anomalous conductances Sak, k =1,...,K, the quantities E(Pk,ω,S(r ′ )) are solutions of this<br />
system. For each frequency, the system has 6 × K equations and 6 × K real unknowns in the general case. Solving<br />
this system allows us to compute the electric field E(r,ω,S(r)) at any point r and frequency ω. The magnetic field<br />
H(r,ω,S(r)) is then derived using (5) written at any point r and frequency ω.<br />
5.4 The inverse problem<br />
The structure of the embedding 1-D model, i.e., Sn(r), the depth of the heterogeneous thin sheet, the domain Ω, and<br />
the cells Pk are first defined, and remain fixed during the inversion. Their values are to be deduced from previous<br />
investigations in the studied area (geology, other geophysical methods, ...) and from laboratory measurements of<br />
rock conductivity. The conductances Sk, k =1,...,K, are the parameters, and solving the inverse problem consists<br />
in estimating their a posteriori pdf given the data d =(dij), i =1,...,I, j =1,...,J, and the a priori pdf P0.<br />
The data d are actually estimates of a function T [ E(r,ω,S(r)), H(r,ω,S(r)) ] at points ri, i =1,...,I,ofthe<br />
surface and frequencies ωj, j =1,...,J, plus an experimental noise δd,<br />
dij = T [ E(ri,ωj,S(r)), H(ri,ωj,S(r)) ] + δdij. (6)<br />
The δdij, and therefore the dij are assumed to be independent gaussian random variables with zero mean value and<br />
standard deviation τ.<br />
In situations for which no particular information is available, we choose as a priori distribution the product of a<br />
uniform pdf on each layer. In this case P0 is a uniform distribution over all the possible models and where the different<br />
parameters are independent. P0 is digitised over a set of conductance values, hereafter called possible conductance<br />
values, the choice of which depends on the a priori knowledge of the considered medium. There is no constraint on<br />
this choice, but it is clear that the larger this number the better the determination of the a posteriori distribution, but<br />
the greater the computer time. The possible conductance values may depend on the cell. We will only consider here<br />
situations with the same number L of possible values Sl, l =1,...,L, for each cell. The a posteriori distribution will<br />
accordingly be expressed as the a posteriori probability of these possible conductance values.<br />
The a posteriori pdf of the parameters S is estimated by means of a Markov chain. Let an algorithm that considers<br />
each cell Pk successively and updates the value of the parameter for this cell with a transition probability equal to the<br />
conditional probability distribution of Sl, l =1,...,L, given the actual values of the parameters for the other cells,<br />
the data d and the a priori pdf P0(S). The sequence of images thus obtained after each scanning of the whole set of<br />
K cells is a homogeneous Markov chain, because the probability of an image depends only on the previous image,<br />
and does not change with step n. This process is a Markov chain on M called the Gibbs sampler (Robert, 1996).<br />
It is ergodic, and its invariant probability Π is the a posteriori probability is the conditional probability Π(S) of the<br />
parameters given the data d and the a priori pdf P0(S) (Roussignol et al., 1993; Roussignol and Menvielle, 1995;<br />
Robert, 1996; Grandis et al., 1999; 2002).<br />
In the present case, the required L × K solutions of the forward problem that are necessary for computing the<br />
transition probability of the Markov chain (in the present case, the K conditional pdf ’s corresponding to the whole set<br />
of cells) requires a very long CPU time. In order to limit the CPU time, the forward problem is not fully solved at each<br />
scanning, and the electric field is estimated by the Markov chain. The Markov chain is then a two-dimensional one,<br />
which estimates a quantity (the conductance) distributed over a finite set of real values and another one (the electric<br />
field) distributed over a continuous set of real vectors.<br />
The asymptotic behavior of such a Markov chain has first been studied empirically by Jouanne (1991), then Grandis<br />
(1994) for situations representatives of those encountered in geophysics. Using synthetic models, Grandis (1994)<br />
evidenced that the empirical average of the transition pdf of this Markov chain tends towards an invariant pdf that corresponds<br />
to the a posteriori pdf of the parameters, provided the estimate of the electric field at each scanning is precise<br />
enough for the transition probabilities to be reasonably estimated. Later on, Touijar (1994) gave the first demonstration<br />
that such a chain converges under given conditions. The physical interpretation of Touijar’s mathematical results<br />
suggests that the empirical results established by Jouanne (1991) and Grandis (1994) are likely to be valid for most of<br />
geophysical situations (Menvielle and Roussignol, 1995).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
238
6 The inversion<br />
6.1 The description of the model and data<br />
In accord with the above discussion, we chose the following particular parameters for the thin sheet model used in the<br />
MCMC inversion of the long-period geomagnetic induction data over the BM/BV/WCP region:<br />
(i) The anomalous domain of the superficial thin sheet is divided into 21(N)×29(E) square cells, each 25×25 km 2<br />
in size, covering thus a region of about 500(N)×750(E) km 2 . This gives in total 609 variable parameters (cell conductances)<br />
for the inversion.<br />
(ii) The normal model was composed of a two-layer background model with the first layer of 100 km/1000 Ωmand<br />
a uniform basement of 300 Ωm. The basement was used to simulate the asthenospheric layer. The topography of the<br />
asthenosphere was neglected. The layered model was overlain by a uniform infinite thin sheet with the conductance<br />
of 300 siemens.<br />
(iii) For the inversion purposes, the 150 experimental geomagnetic TF’s available were interpolated into the centers<br />
of the sheet cells. As reliable data variances were not available for the experimental TF’s, we assumed the data to be<br />
contamined with Gaussian errors with the standard deviation of 0.03.<br />
(iv) For the MCMC run, data for only one period, T = 3860 s, were used to keep the computation times in<br />
reasonable limits. Fit of the data and model arrows for further periods available were only checked by direct modeling<br />
tests.<br />
6.2 The MCMC procedure, convergence, output models, post-processing<br />
For the MCMC procedure (Gibbs sampler), the conductance of each of the 609 variable cells could take on one of<br />
12 predefined values from the interval of 3 to 5000 siemens. The Markov chains were typically running for 1000 cycles<br />
of the Gibbs sampler, which took about 10 hours of the CPU time on a standard PC station (Intel Xeon 3 GHz, 1 MB<br />
RAM). Repeated MCMC runs were carried out to simulate parallel chains and check the convergence of the MCMC<br />
process. First 500 cycles of each of the MCMC chains were discarded as a burn-in period during which the chain<br />
approaches a vicinity of the solution and stabilizes. In our runs, the chains did not change substantially after the burnin<br />
period, and simulated parallel chains converged to similar conductance distributions for the same cells. From this<br />
rough diagnostics, we can conclude that the chains converged and can provide reasonable estimates of the means of<br />
the cell conductances within the selected model class M. As the number of MCMC iteration cycles is relatively small<br />
in our experiments, it is not realistic to obtain reasonably accurate estimates of higher moments of the aposteriori pdf ’s<br />
for the cell conductances. In fact, only highly qualitative conclusions may be made as regards the uncertainties of the<br />
variables.<br />
The MCMC procedure gives a series of models which should be probabilistically distributed in accord with the<br />
experimental data. We obtain a whole histogram of conductance values for each cell, which shows how much likely<br />
are the particular conductance values in the cell considered. After long enough iteration process, the histograms<br />
approximate the marginal aposteriori probabilities of the discrete conductance values in the individual sheet cells. We<br />
can estimate the mean values of the cell conductances by integrating over the histogram, as well as the most probable<br />
cell conductances (maximum aposteriori probabability, or MAP estimates) by taking the most frequent value from the<br />
histogram.<br />
Fig. 3 displays sheet models over the BM/BV/WCP region set together from the mean values of the conductance<br />
in the individual sheet cells and from the MAP estimates of the cell conductances. These models do not represent<br />
the average or MAP models from the chain, which would both require integration of the multi-dimensional empirical<br />
posterior pdf over the whole parameter space. The presented models are put together from estimates based on individual<br />
marginal pdf ’s for the individual sheet cells. It is clear that these models may differ considerably from each<br />
other, especially in regions where the cell conductances are poorly constrained by the data. Nevertheless, the models<br />
in Fig. 3 represent simple integrated information on the conductance structure projected from a 3-D probabilistic (histogram)<br />
image of the cell conductances provided by the whole underlying Markov chain. Fit of the model data from<br />
the ‘average’ model in Fig. 3 (left) and the experimental induction arrows is shown in Fig. 4 for the inverted period<br />
of 3860 s. Model vs. experiment fit for longer periods (not shown) is very similar, larger discrepancies are observed<br />
for the period T = 1200 s with model induction arrows substantially smaller than the experimental data. This is most<br />
likely due to a misspecified depth of the sheet in our model (superficial sheet).<br />
Qualitatively, we can also estimate the uncertainty of the conductance by simply observing the ‘flatness’ or ‘peakiness’<br />
of the histograms for the cell conductances. In Fig. 5 (left), we show the ‘average’ model from Fig. 3 (left)<br />
modified by keeping in it only those cells for which 90% of the cell conductance values in the MCMC histograms fall<br />
within an interval of a width of 2/3 of a decade. All remaining cells, with flatter conductance histograms, have been<br />
discarded. The cells in Fig. 5 (left) thus indicate regions of the model in which the conductance is reasonably well constrained<br />
by the data. More precise, quantitative uncertainty estimation would be perhaps possible from longer MCMC<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
239
chains which would allow us to integrate directly for variance-covariance functions estimates for the parameters.<br />
Because of high computational demands, thin sheet models for the MCMC inversion are mostly designed with<br />
only a coarse tiling. In the model post-processing stage, spatial smoothing often results in a more realistic image of<br />
the conductance distribution. The smoothed image has to be checked by an additional direct modeling test. We show<br />
a smoothed version of the ‘average’ conductance model from Fig. 3 (left) in the right-hand side panel of Fig. 5.<br />
Figure 3: Left: Conductance model over the BM/BV/WCP region designed from average cell conductances obtained<br />
form the stabilized part of the Markov chain. Right: Conductance model designed from maximum aposteriori conductances<br />
in the individual sheet cells. The color scale corresponds to the logarithm of the conductances. Period of<br />
inverted data was 3860 s.<br />
14˚ 15˚ 16˚ 17˚ 18˚ 19˚ 20˚ 21˚ 22˚ 23˚<br />
51˚ 51˚<br />
50˚ 50˚<br />
49˚ 49˚<br />
48˚ 48˚<br />
14˚<br />
15˚<br />
16˚<br />
17˚<br />
18˚<br />
19˚<br />
20˚<br />
21˚<br />
22˚<br />
23˚<br />
14˚ 15˚ 16˚ 17˚ 18˚ 19˚ 20˚ 21˚ 22˚ 23˚<br />
51˚ 51˚<br />
50˚ 50˚<br />
49˚ 49˚<br />
48˚ 48˚<br />
Figure 4: Fit of the real (left) and imaginary (right) induction arrows for the period of 3860 s. White arrors show the<br />
input data obtained by interpolating the experimental TF’s into the centers of the sheet cells. Orange-yellow arrows<br />
were generated by the thin sheet model designed from average cell conductivities from the MCMC chain (model to<br />
the left in Fig. 3).<br />
7 Tectonic correlation<br />
Interpretation of narow-band long-period geomagnetic data by the MCMC sampling can only provide a large-scale<br />
image of the conductance distribution which conforms with the observed TF’s. The conductance pattern restored from<br />
the induction arrows collected above the eastern slopes of the BM and its transition to the WCP gives a geologically<br />
plausible image of the region. The Carpathian conductivity anomaly is reconstructed in detail, suggesting possible<br />
weakening/local interruption at the crossing with the Central Slovakia fault zone (near 19.2 o E in Fig. 5 right) where<br />
other geophysical fields and geological indicators show discontinuous features on a presumably strike-slip fault (e.g.,<br />
Kov č and H t, 1993). Though in coarse mesh, the high conductance anomaly well fits the position of a low resistivity<br />
layer from the magnetotelluric data collected in the esternmost sector of the model, close 22 o E (Voz r, 2005).<br />
The induction anomaly at the eastern margin of the Bohemian Massif seems to have a more complex explanation<br />
than that of a quasi-linear conductor along the SW-NE zone of a rapid change of direction of the induction arrows<br />
(see Fig. 2). Two features may be observed in the model in Fig. 5 (right), which were already discussed earlier by<br />
Kov čikov et al. (2005). First, it is a conductive zone in the NE of the BM, which may generate induction arrows<br />
directed towards the SW, as clearly observed all across the area to the west of the EBMA. Second, it is an alteration<br />
of conductive and resistive E-W zones in the eastern part of the BM. Though still not clear how much of this effect<br />
14˚<br />
15˚<br />
16˚<br />
17˚<br />
18˚<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
240<br />
19˚<br />
20˚<br />
21˚<br />
22˚<br />
23˚
14˚ 15˚ 16˚ 17˚ 18˚ 19˚ 20˚ 21˚ 22˚ 23˚<br />
51˚ 51˚<br />
50˚ 50˚<br />
49˚ 49˚<br />
48˚ 48˚<br />
14˚<br />
15˚<br />
16˚<br />
17˚<br />
18˚<br />
19˚<br />
20˚<br />
21˚<br />
22˚<br />
23˚<br />
14˚ 15˚ 16˚ 17˚ 18˚ 19˚ 20˚ 21˚ 22˚ 23˚<br />
51˚ 51˚<br />
50˚ 50˚<br />
49˚ 49˚<br />
48˚ 48˚<br />
14˚<br />
15˚<br />
16˚<br />
17˚<br />
18˚<br />
19˚<br />
20˚<br />
21˚<br />
22˚<br />
23˚<br />
4<br />
3<br />
2<br />
1<br />
0<br />
log10(S)<br />
Figure 5: Left: Model from Fig 3 (left) with only those cells of the original ‘average’ model shown that have 90% of<br />
all conductance values within 2/3 of a decade. Cells with flat histograms are eliminated. Right: Spatially smoothed<br />
‘average’ conductance model. The dashed lines indicate main regional fault zones.<br />
may be caused by the profile arrangement of the data, this quasi-anisotropic domain is actually required by the data. It<br />
helps in feeding the induction process all across the Moravia region and in keeping the relatively large modules of the<br />
induction arrows in that region. If these induction sources could be verified, the regional induction in the eastern part<br />
of the BM would be mainly driven by the NW-SE to W-E pattern of tectonic zones (Elbe fault zone, Sudety faults,<br />
Poˇr č -Hronov fault zone, Odra fault zone, etc.) rather than by the SW-NE zones which conform with the tectonics of<br />
the transition to the WCP.<br />
8 Conclusion<br />
From the experience accumulated with the MCMC inversion we can conclude that for moderately sized thin sheet<br />
models (tens of cells in each direction), the MCMC inverse procedure is practicable with standard computer facilities.<br />
Speeding-up the forward solutions by using a 2-D Markov chain derived from the integral equation numerical approach<br />
by Vasseur and Weidelt (1977) is essential for the algorithm.<br />
Though computationally still demanding, the advantage of the MCMC is that it provides a probabilistic output<br />
for the inverse solution. Parameters can be assessed with respect to their values and uncertainty ranges, though true<br />
uncertainties (variance-covariance matrices) are hard to obtain in problems with demanding direct solutions. Effective<br />
visualization of complete probabilistic outputs from the MCMC algorithm may still be one of its weak points.<br />
The conductance pattern restored by the MCMC sampling from the induction arrows collected above the eastern<br />
slopes of the BM and its transition to the WCP gives a geologically plausible image of the region. The Carpathian<br />
conductivity anomaly is reconstructed in detail. The induction anomaly suggested at the eastern margin of the BM<br />
admits an alternative explanation in terms of a NW-SE to W-E conductance patterns, hypothetically coinciding with<br />
the fault pattern of the eastern Bohemian Massif. A detailed verification of the structural hypotheses in this region is,<br />
however, hardly possible by employing the long period geomagnetic induction data alone.<br />
Acknowledgements<br />
Financial assistance of the Czech Sci. Found., contract No. 205/07/0292, and of the Grant Agency Acad. Sci. Czech<br />
Rep., contract No. IAA300120703, is highly acknowledged.<br />
References<br />
d m, A., 1976. Results of deep electromagnetic investigations. In: d m (Editor), Geoelectric and Geothermal<br />
Studies, KAPG Geophysical Monograph, Akad miai Kiad , Budapest, 547-559.<br />
d m, A. and Posp il, L., 1984. Crustal conductivity anomalies in the Carpathian region. Acta Geodaet. Geophys.<br />
Montanist. Hung., 19, 19-34.<br />
d m, A., Landy, K. and Nagy, Z., 1988. New evidence for the special distribution of the electric conductivity in the<br />
Earth’s crust and upper mantle in the Pannonian basin as a ‘hot spot’, Tectonophysics, 164, 361–368.<br />
Babu ka, V., Plomerov , J. and Pajdu k, P., 1988. Lithosphere-asthenosphere in Central Europe: Models derived<br />
from P-residuals, in Proceedings of the Fourth Workshop on the European Geotraverse (EGT) project: The Upper<br />
Mantle, G. Nolet and B. Dost (Eds.), ESF, Strasbourg, 37-48.<br />
Berger, J. O., 1985. Statistical Decision Theory and Bayesian Analysis, Springer-Verlag, New York.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
241
Bolstad, W. M., 2004. Introduction to Bayesian Statistics, John Wiley & Sons, 2004.<br />
Box, G. and Tiao, G., 1973. Bayesian Inference in Statistical Analysis, Addison-Wesley, Reading.<br />
Bruton P., 1994. Testing Vasseur and Weidelt’s thin sheet algorithm, in Proceedings of the 15th Colloquium ‘Electromagnetic<br />
Depth Research’, Bahr, K. and Junge, A. (Eds.), DGG, pp. 54-59.<br />
Červ, V. and Pek, J., 1981, Numerical solution of the two-dimensional inverse geomagnetic induction problem. Studia<br />
geoph. et geod., 25, 69-80.<br />
Červ, V., Pek, J. and Praus, O., 1984. Model of geoelectrical anomalies in Czechoslovakia. J. Geophys., 55, 161-168.<br />
Červ, V., Pek, J. and Praus, O., 1987.Numerical modelling of geoelectrical structures in Czechoslovakia. Phys. Earth<br />
Planet. Inter., 45, 170-178.<br />
Červ, V., Pek, J. and Praus, O., 1990. Limitations of the unimodal thin sheet modelling of near-surface inhomogeneities<br />
in Czechoslovakia, Phys. Earth Planet. Inter., 60, 100-114.<br />
Červ, V., Kov čikov , S., Pek, J., Pěčov , J. and Praus, O., 1997. Model of Electrical Conductivity Distribution across<br />
Central Europe, J. Geomag. Geoelectr., 49, 1585-1600.<br />
Dudek, A., The crystalline basement block of the Outer Carpathians in Moravia: Brunovistulicum, Ropr. Cs. Akad.<br />
Ved, R. mat. a pr. ved, Praha, 90, 1980, 1-85.<br />
Glover, P.W.J. and Vine, F.J., 1992. Electrical conductivity of carbon-bearing granulite at raised temperatures and<br />
pressures, Nature 360, 723-726.<br />
Glover, P.W.J. and Vine, F.J., 1995. Beyond KTB-Electrical conductivity of the deep continental crust, Surveys in<br />
Geophysics, 16, 5-36.<br />
Grandis, H., 1994. Imagerie electromagnetique Bayesienne par la simulation d’une chaine de Markov, Doctorat<br />
d’Universit , Universit Paris VII.<br />
Grandis, H., Menvielle, M. and Roussignol, M., 1999. Bayesian inversion with Markov chains—-I. The magnetotelluric<br />
one-dimensional case, Geophys. J. Int., 138, 757–768.<br />
Grandis, H., Menvielle, M. and Roussignol, M., 2002. Thin-sheet electromagnetic inversion modeling using Monte<br />
Carlo Markov Chain (MCMC) algorithm, Earth Planets Space, 54, 511–521.<br />
Hvo dara, M. and Voz r, J., 2004. Laboratory and geophysical implications for explanation of the nature of the<br />
Carpathian Conductivity Anomaly. Acta Geophysica Polonica, Vol. 52, No. 4, pp. 497-508.<br />
Jankowski, J., Petr, V., Pěčov , J. and Praus , O., 1984. Geoelectric anomaly in the Czechoslovak-Polish section of the<br />
Carpathians on the basis of geomagnetic and magnetotelluric soundings, Acta Geodaet., Geophys. Montanist. Hung.,<br />
19, 81-91.<br />
Jankowski, J., Tarlowski, Z., Praus, O., Pěčov , J. and Petr, V., 1985. The results of Deep Geomagnetic Soundings in<br />
the West Carpathians, Geophys. J. Royal Astron. Soc., 80, 561-574.<br />
Jankowski, J., Pawliszyn, J., J wiak, W. and Ernst, T., 1991. Synthesis of electric conductivity surveys performed<br />
on the Polish part of the Carpathians with geomagnetic and magnetotelluric sounding methods, Publs. Inst. Geophys.<br />
Pol. Ac. Sci., A-19(236),183-214.<br />
Jankowski, J., J ´zwiak, W. and Voz r, J., 2008. Arguments for ionic nature of the Carpathian electric conductivity<br />
anomaly, Acta Geophys., 56, 455-465.<br />
Jouanne, V., 1991. Application des techniques statistiques Bayesiennes a l’inversion de donn es lectromagn tiques,<br />
Doctorat d’Universit , Universit Paris VII.<br />
Kov č, P. and H k, J., 1993. The Central Slovak Fault System-=field evidence of a strike slip. Geol. Carpath., 44,<br />
133-159.<br />
Kov čikov , Červ, V., Praus, O., 1997. Modelling of geomagnetic transfer functions on the eastern margin of the<br />
Bohemian Massif, Annales Geophysicae, 15, Suppl. I, C 159.<br />
Kov čikov , S., Červ, V. and Praus, O., 2005. Modelling of the conductance distribution at the eastern margin of the<br />
European Hercynides, Studia geophys. geod., 49, 403-421.<br />
Menvielle, M. and Roussignol, M., 1995. Imagerie lectromagn tique bayesienne. In: Munschy, M., Sauter, D.,<br />
Schlich, R. (Eds.), Comit National Fran ais de G od sie et G ophysique, Rapport Quadriennal 1991–1994, Paris,<br />
France, pp. 177–190.<br />
Pěčov , J. and Praus, O., 1996. Anomalous induction zones in the Czech republic in relation to large scale European<br />
anomalies, Studia geoph. et geod., 40, 50-76.<br />
Praus, O., Pěčov , J., 1991. Anomalous geomagnetic fields of internal origin in Czechoslovakia, Studia geoph. et<br />
geod., 35, 81-89.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
242
Praus, O., Pěčov , J., Petr, V., Babu ka, V. and Plomerov , J., 1990. Magnetotelluric and seismological determination<br />
of the lithosphere-asthenosphere transition in Central Europe, Phys. Earth Planet. Inter., 60, 212-228.<br />
Press, S.J., Flannery, B.P., Teukolsky, S.A. and Vetterling, W.T., 1989. Numerical Recipes: the Art of Scientific<br />
Computing, Cambridge University Press, Cambridge.<br />
Robert, C., 1996. M thodes de Monte Carlo par Chaines de Markov, Economica, Paris.<br />
Roussignol, M., Jouanne, V.,Menvielle, M. and Tarits, P., 1993. Bayesian electromagnetic imaging, in Computer<br />
Intensive Methods, pp. 85–97, Hardle, W. and Siman, L. (Eds.), Physical Verlag.<br />
Schott, J.-J., Roussignol, M., Menvielle, M. and Nomenjahanary, F. R., 1999. Bayesian inversion with Markov<br />
chains—-II. The one-dimensional DC multilayer case, Geophys. J. Int., 138, 769–783.<br />
Suk, M., Bl kovsk , Buday, M., Chlup č, I., Cicha, I., Dudek, A., Dvoˇr k, J., Eli , M., Holub, V., Ibrmajer, J.,<br />
Kodym, O., Kukal, Z., Malkovsk , M., Menc k, E., M ller, V., Tyr ček, J., Vejnar, Z., Zeman, A., 1984. Geological<br />
history of the Czech Socialist Republic, Publ. of stˇredn stav geologick , 17-86.<br />
Suk, M., 1995. Regional geological division of the Bohemian Massif, Exploration Geophysics, Remote Sensing and<br />
Environment, II. 2., 25-30<br />
Touijar D., 1994. Stochastic algorithms and application to Bayesian electromagnetic imaging, PhD Thesis, Travaux<br />
Universitaires–Thčse nouveau doctorat, Universit de Lille 1, Villeneuve-d’Ascq, France, 104 pp.<br />
Vasseur, G. and Weidelt, P., 1977. Bimodal EM induction in non-uniform thin sheets with an application to the<br />
Northern Pyrenean Anomaly, Geophys. J. R. astr. Soc., 51, 669-690.<br />
Voz r, J., 2005. Magnetotelluric Soundings on the Territory of Slovakia, PhD Thesis, GFU SAV Bratislava, 2005, 108<br />
pp. (in Slovak).<br />
Wang, L. J. and Lilley, F. E. M., 2002. Inversion of magnetometer array data by thin-sheet modelling, Geophys. J.<br />
Int., 137, 128-138.<br />
Weidelt, P., 1975, Electromagnetic induction in three-dimensional structures, J. Geophys., 41, 85-109.<br />
Wiese, H., 1962. Geomagnetische Tiefentellurik Teil II: Die Streichrichtung der Untergrundstrukturendes elektrischen<br />
Widerstandes, erschlossen aus geomagnetischen Variationen, Geofisica pura e applicata, 52, 83-103.<br />
Wybraniec, S., Jankowski, J., Ernst, T., Pěčov , J., Praus, O., 1999. A new method for presentation of induction vector<br />
distribution in Europe, Acta geophys. polonica, XLVII, 323-334.<br />
´Zytko, K., 1997, Electrical conductivity anomaly of the Northern Carpathians and the deep structure of the orogen,<br />
Annales Societatis Geologorum Poloniae, 67, 25-43.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
243
Subsurface Conductivity Obtained from DC Railway<br />
Signal Propagation with a Dipole Model<br />
Anne Neska<br />
Institute of Geophysics PAS, Ul. Ks. Janusza 64, 01-452 Warszawa, Poland, email: anne@igf.edu.pl<br />
Abstract<br />
In the presented study, an attempt is made to model the propagation behavior of signals<br />
emitted by a Polish DC railway line with a number of grounded horizontal electric dipoles on<br />
the surface of a homogeneous half-space. The signals were measured on a magnetotelluric<br />
profile perpendicular to the railway line mainly in its near-field and transition zone, and they<br />
were separated from the natural electromagnetic variations by means of a reference site.<br />
Fitting the DC signals to the dipole model yields the conductivity of the homogeneous halfspace.<br />
Its value for the vertical and for the perpendicular horizontal magnetic component is<br />
confirmed by the usual 2D MT model obtained from the profile.<br />
Introduction<br />
The electrified part of the Polish railway network (fig. 1) is run by DC current. This makes<br />
magnetotelluric (MT) studies in this country difficult. On the other hand, the “disturbing” (in<br />
terms of MT) signals emitted by this system propagate according to a dipole model (Oettinger et<br />
al. 2001 and citations therein) which contains, like MT, the electrical conductivity of the<br />
subsurface as a parameter. So, after designing and performing a MT field experiment<br />
comprehending railway signals in an appropriate way, it should be possible to obtain information<br />
about the subsurface conductivity by both the dipole model and the MT approach. By this idea<br />
the present study has been motivated.<br />
The methodology of this study lies somewhere in-between two other domains encountered in<br />
geophysics. One is certainly the electromagnetic methods of the controlled-source near-field<br />
branch. The difference to our case is that the source is more under control there, but in return not<br />
free of experimental effort. The second related domain is techniques for modeling the<br />
propagation behavior of disturbances from DC railways and cables, applied especially to<br />
investigate their influence on magnetic measurements run at observatories (e.g. Pirjola et al.<br />
2007, Lowes 2009, Maule et al. 2009). However, the induction character of this propagation<br />
including phenomena like frequency-dependent damping, phase shift, and the electric<br />
conductivity as a propagation parameter is almost completely omitted there in contrast to the<br />
approach presented here.<br />
In the following, there will be considered the data processing including the separation of the<br />
natural electromagnetic variations from the railway signals and the estimation of the transfer<br />
functions between different stations. The dipole model will be described and the data be inverted<br />
according to it. The conductivity result will be compared to the output of a usual 2D MT model.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
244
Fig. 1 The electrified part of the Polish railway network (red lines) with a profile of MT sites for<br />
investigation of one line (zoomed-in part). Green squares denote positions of reference stations.<br />
Measurement and data processing<br />
As shown in fig. 1, a profile of six MT sites has been installed perpendicular to an isolated railway<br />
line. The distances of sites to the line were 0.8, 7, 16, 25, 50, and 75 km. Two remote reference<br />
sites were set up at relatively noise-free places, and the whole array was running synchronously.<br />
The obtained data were processed with codes by Neska 2006, but not only for MT purposes.<br />
By means of its correlation to the reference sites, the natural part of the electromagnetic<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
245
signal could be removed from the measured data using a method by Larsen et al. 1996 (fig. 2). In<br />
this way a dataset for analysis of the railway signals was established (cf. fig. 3). It has to be emphasized<br />
that it can be problematic to work with such “deduced” data, since distortions introduced<br />
to it during the separation, e.g. due to uncorrelated noise in the reference site, can bias the<br />
subsequent results. It proved useful to separate the data of the station closest to the railway which<br />
plays a special role in the following with one reference site and the rest with the second one.<br />
Fig. 2 Separation of time series into a “natural” and a “railway” part (site s02, see fig. 1).<br />
Fig. 3 The propagation/decay of railway signals (component Bx) over the profile. Note that a rest of<br />
natural variations remained in the time series in spite of separation.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
246
All following considerations refer to “railway data” separated from natural signals in the way<br />
indicated above.<br />
The formulas for the dipole response are given in frequency domain, so our calculation has to<br />
take place there. Furthermore, in these formulas there occur factors like current I and dipole<br />
length L (see table 1), i.e. technical parameters of railway traction which we cannot measure<br />
immediately. So it appeared reasonable to remove these factors by normalizing each field<br />
component to that of the railway-nearest site which represents the strongest, most pronounced, or<br />
least attenuated railway signal, respectively.<br />
To perform this normalization in a convenient way, the algorithm for calculation of the interstation<br />
transfer function or Horizontal Magnetic Tensor (HMT) was used, which played an<br />
essential role during the separation already. The site nearest to the railway takes the function of<br />
the reference, i.e. its data are input channels, and the data of the remaining sites are the output<br />
channels. So Bx and By data of “local” and “reference” sites lead to normalized values for both<br />
horizontal magnetic components, Ex and Ey to those for the electric components (even if the<br />
analogy to the HMT is overstretched here since it refers to magnetic data only, the formalism is<br />
the same), and Bz and some other, uncorrelated channel (e.g. of one of the real, off-profile<br />
references, just to fit the requirement of the HMT formalism that there must be two input<br />
channels) to the normalized value for the vertical magnetic component. This description remains<br />
a bit vague because it cannot be recommended for imitation for the following reason:<br />
For this study, it has been noticed too late that this approach is not quite correct. The problem is<br />
that the HMT formalism bases on bivariate statistics (i.e. there are two independent variables or<br />
input channels, respectively), whereas the railway provides only one independent source<br />
polarization (pers. comm. K. Nowożyński). A similar problem occurs in Controlled-Source MT,<br />
where only one scalar transfer function instead of the usual 2x2 impedance tensor can be<br />
obtained if only one transmitter is used (pers. comm. M. Becken). So it has to be expected that<br />
our normalized values are influenced in an unfavorable way. This impact is expected to be very<br />
small in the case of Bz, since the warranted uncorrelatedness of the second input channel should<br />
make the difference between the univariate and the bivariate result vanish. This will be<br />
confirmed in the following section (cf. fig. 4: better result for Bz). However, the direction of the<br />
railway line under investigation is almost East-West (fig. 1), so its signal has a polarization<br />
almost coinciding with one of the directions we measure and calculate in with the MT or HMT<br />
approach. So there is hope that the error introduced due to this wrong statistic approach is<br />
somehow bearable if the diagonal elements of the “HMT” are used.<br />
Finally, the normalized values are displayed over period and distance in 3D-plots (red lines in<br />
fig. 4). The decay of amplitudes with distance as well as their period-dependency (stronger<br />
damping at low periods) becomes clearly visible for Bx and Bz components. Bx has partly<br />
unsystematic and, especially at long periods, rather high values that make the impression to be<br />
distortions, maybe due to incomplete separation or the wrong statistic approach mentioned.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
247
Hence the possibility to include data errors to the business was provided, since they have an<br />
important meaning as weights during the inversion.<br />
The dipole model and its application<br />
The railway line was simulated by a chain of horizontal grounded electric dipoles following its<br />
(not constant) bearing within a distance of some dozens of km around the profile. The distance<br />
between single dipole centers was 250 m. The solutions for the electromagnetic field components<br />
for the quasi-static approximation are taken from Zonge & Hughes 1987 and hold for one dipole<br />
on the surface of a homogeneous half-space (see table 1). These values are transformed from<br />
cylindrical to Cartesian coordinates to match the measuring system of MT and summarized over<br />
all dipoles subsequently. Then the values for each component and station were divided by the<br />
value of the corresponding component of the station closest to the railway. Since we look only<br />
for one model parameter (i.e. the conductivity σ), the inversion consists just of testing the whole<br />
model space and selecting the value with minimum RMS. The relevant software related to the<br />
dipole model was created by K. Nowożyński.<br />
The components behaving best in our study were Bx and Bz. The inversion of Bx alone gave a<br />
resistivity of 5 Ωm, that of Bz alone 14 Ωm, and the joint inversion 10 Ωm. The model response<br />
for the latter is shown in fig. 4 as blue lines.<br />
Table 1<br />
Solution for horizontal grounded electric dipole on HH surface<br />
according to Zonge & Hughes, 1987<br />
Where I/K_0/1 - modified Bessel functions and<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
248
Fig. 4 Railway signal data (red) and dipole model response for a 10 Ωm homogeneous half-space (blue)<br />
over period and distance. Upper part for Bx, lower part for Bz component amplitude. The values on the<br />
vertical axis are normalized to that of site s01 nearest to the railway line (cf. fig. 1).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
249
Fig. 5 2D MT model (TE and TM mode) of the profile on fig. 1 obtained with the REBOCC code<br />
(Siripunvaraporn & Egbert 2000). The shallow cells beneath the northernmost sites have resistivity values<br />
between 9 and 80 Ωm.<br />
Discussion and conclusions<br />
The MT model yields resistivities of 9-80 Ωm beneath the northernmost sites close to the railway<br />
(fig. 5). This is consistent with the dipole model result, so there is evidence that the dipole<br />
approach is not unreasonable.<br />
Nevertheless, there are unexplained features in this model and one has to be critical with it. Not<br />
shown here is the behavior of phases and of electric components, where a similarity between data<br />
and model response is much harder to find for reasons still unclear.<br />
However, taking into account first, the similarity of data and model response (fig. 4), and second,<br />
the compatibility of MT and dipole modeling results in spite of some quite principal problems in<br />
this implementation (i.e. input data biased during separation, errors due to application of<br />
bivariate instead of univariate statistics, limitation to the oversimplified case of a homogeneous<br />
half-space), there can be stated that the dipole model is not only a valid approach to model the<br />
propagation of DC railway signals, but even a relatively robust one.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
250
Acknowledgements<br />
This work would not exist without the help of Paweł Czubak, Tomasz Ernst, Jerzy Jankowski, Waldemar<br />
Jóźwiak, Krzysztof Kucharski, Janusz Marianiuk, Mariusz Neska, Krzysztof Nowożyński, Jan Reda,<br />
Michał Sawicki, and Józef Skowroński.<br />
The project is funded by the Ministry of Science and Higher Education of the Republic of Poland (grant<br />
number N N307 2496 33).<br />
Literature<br />
Larsen, J. C., R. L. Mackie, A. Manzella, A. Fiordelisi, and S. Rieven. Robust smooth magnetotelluric<br />
transfer functions. Geophysical Journal International, 124:801819, 1996.<br />
Lowes, F. J., DC railways and the magnetic fields they produce - the geomagnetic context, Earth Planets<br />
Space, 61, i–xv, 2009<br />
Maule, C.F., P. Thejll, A. Neska, J. Matzka, L.W. Pedersen, and A. Nilsson, Analyzing and correcting for<br />
contaminating magnetic fields at the Brorfelde geomagnetic observatory due to high voltage DC power<br />
lines, Earth Planets Space, 61, 1233–1241, 2009<br />
Neska, A., Remote Reference versus Signal-Noise Separation: A least-square based comparison between<br />
magnetotelluric processing techniques, PhD thesis, Freie Universität Berlin, Fachbereich Geowissenschaften,<br />
available at http://www.diss.fu-berlin.de/2006/349, 2006.<br />
Oettinger, G., V. Haak, and J.C. Larsen. Noise reduction in magnetotelluric time-series with a new signalnoise<br />
separation method and its applications to a field experiment in the Saxonian Granit Massif.<br />
Geophysical Journal International, 146:659669, 2001.<br />
Pirjola, R., L. Newitt, D. Boteler, L.a Trichtchenko, P. Fernberg, L. McKee, D. Danskin, and G. J. van<br />
Beek, Modelling the disturbance caused by a dc-electrified railway to geomagnetic measurements, Earth<br />
Planets Space, 59, 943–949, 2007.<br />
Siripunvaraporn, W., and G. Egbert, An efficient data-subspace inversion for two-dimensional magnetotelluric<br />
data, Geophysics, 65, 791-803, 2000.<br />
Zonge, K.L. and L.J. Hughes, Controlled source audio-frequency magnetotellurics, In: M.N. Nabighian<br />
(editor), Electromagnetic Methods in Applied Geophysics, Volume 2, Application, Parts A and B, pp.<br />
713809, SEG, Tulsa, 1987.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
251
Magnetotelluric investigation of the<br />
Sorgenfrei-Tornquist Zone and the NE German Basin<br />
Anja Schäfer 1∗ , Heinrich Brasse 1 , Norbert Hoffmann 2<br />
and EMTESZ working group<br />
1 Freie Universität Berlin, Fachrichtung Geophysik, Malteserstr. 74-100, 12249 Berlin, Germany<br />
2 Wilhelm-Külz-Straße 70, 14532 Stahnsdorf, Germany<br />
Abstract<br />
The Sorgenfrei-Tornquist Zone (STZ) is the northwestern branch of the Trans-European Suture<br />
Zone (TESZ). The TESZ runs along more than 2 000 km through the North Sea, South Scandinavia,<br />
the Baltic Sea, and Poland into the Black Sea. It divides old Precambrian lithosphere in<br />
the northeast from younger, Caledonian and Variscan one in the southwest. Data, dimensionality<br />
analysis and 2D models of a long-period magnetotelluric (LMT) profile crossing this prominent<br />
tectonic border in South Scandinavia are presented and a possible interpretation of conductivity<br />
anomalies below Rügen Island and south of Strelasund Basin is given. Furthermore these models<br />
map the saline aquifer of the Northeast German Basin. The ”North German Conductivity<br />
Anomaly” is perhaps mainly due to the effect of the basin edges.<br />
Introduction<br />
The EMTESZ project (Electromagnetic Study of the Trans-European Suture Zone) was a multinational<br />
research project to study the electrical conductivity of the Trans-European Suture Zone<br />
(TESZ), which is one of the largest tectonic boundaries in Europe. It separates the East European<br />
Platform from the Paleozoic mobile belt of central and western Europe and is traced from the<br />
Black Sea through Poland and Southern Scandinavia into the North Sea (Gee (1996); Pharaoh<br />
(1999)). In the northeast the margin is marked by the lineaments of the Sorgenfrei-Tornquist<br />
Zone (STZ) and the Teisseyere-Tornquist Zone (TTZ, which runs in a SE-NW direction through<br />
Poland and the Baltic Sea, fig. 1). Several seismic refraction and reflection experiments have been<br />
carried out (BABEL and BASIN 9601 in Northeast Germany, POLONAISE and the LT surveys<br />
in Poland). Major results include a sedimentary thickness in the Northeast German Basin and<br />
below the TTZ of 4 to more than 11 km. Also sharp lateral boundaries, a Moho from 32-35 km<br />
in the SW to 40-45 km in the NE and a reflector below the STZ/TTZ at depth of 50-55 km are<br />
known. Most of the earlier EMTESZ measurements were carried out in 2003 to 2005 (e.g., Brasse<br />
et al. (2006), Ernst et al. (2008)). Profiles MVB and MVS were conducted in 2006 and 2009.<br />
The long period magnetotelluric profile shown here is one of these EMTESZ profiles. The direction<br />
of this profile was chosen because of earlier measurements by the Bundesanstalt für Geowissenschaften<br />
und Rohstoffe (BGR) in the 1990ies. Due to tipper data along this profile and of site<br />
MAT deployed in 2005 on Rügen Island (MT working group Free University of Berlin) a direction<br />
∗ e-mail: schaefer@geophysik.fu-berlin.de<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
252
OST<br />
DAB<br />
BAK<br />
MIR<br />
SWI<br />
NIE<br />
KOC<br />
SCZ WIA<br />
DRE<br />
PIE<br />
GAT<br />
WIE<br />
ZAR<br />
ZAL<br />
GRA<br />
KUS<br />
KNY<br />
PRZ<br />
KUZ<br />
SIE<br />
SLA<br />
B010<br />
ZOL ZOL<br />
B020<br />
POM<br />
GLE<br />
B030<br />
A010 END<br />
REC<br />
B040<br />
A020<br />
RAD<br />
B050<br />
A030 CRI<br />
LUB<br />
KUK<br />
B060<br />
TEE<br />
LYS<br />
B070<br />
A040 LOE<br />
MOC<br />
LES<br />
LAW<br />
B080<br />
A050 OLE<br />
CHO<br />
WAR<br />
B090<br />
A060<br />
LAK<br />
B100<br />
PAS<br />
MOS<br />
A070<br />
STO<br />
SAR<br />
B110<br />
A080<br />
B120<br />
NKU BLE<br />
GOR<br />
B130<br />
VEV<br />
A090<br />
FRI<br />
B140<br />
A100<br />
HOB FAL<br />
B150<br />
A110<br />
MAR MAD<br />
B160<br />
B170<br />
AGO KUN<br />
B180<br />
A120<br />
B190<br />
DAM<br />
BRI<br />
B200<br />
A130<br />
GLI<br />
B210<br />
A140<br />
B220<br />
BUS<br />
A150<br />
TOR<br />
B230 A160<br />
B240<br />
HBO<br />
A170<br />
B250 A180<br />
B260<br />
B270<br />
B280<br />
56°N<br />
55°N<br />
54°N<br />
53°N<br />
52°N<br />
VDF<br />
TEF<br />
Rostock<br />
AAF<br />
STZ<br />
BGR-B<br />
KIS<br />
ROH<br />
TET<br />
WIL<br />
BRE<br />
DOL<br />
Magdeburg<br />
Trelleborg<br />
CDF<br />
IBI<br />
DAR<br />
LEL<br />
WEN<br />
ROT<br />
JAB<br />
MVS<br />
NGK<br />
KRU<br />
PAP<br />
k31<br />
MAT<br />
PAR<br />
TAN<br />
BAR<br />
Germany<br />
BGR-A<br />
LOV<br />
TOM<br />
BOO<br />
k25<br />
Berlin<br />
KOP<br />
NYT<br />
HOR<br />
AND<br />
Rügen<br />
KRI<br />
BER<br />
REG<br />
BLU<br />
MVB<br />
BS1<br />
DZI<br />
FEL<br />
BS3<br />
Odra R.<br />
Bornholm<br />
Baltic Sea<br />
Poland<br />
0 50<br />
11°E 12°E 13°E 14°E 15°E 16°E 17°E 18°E 19°E<br />
BOR<br />
Szczecin<br />
Kostrzyn<br />
Frankfurt<br />
CYB<br />
NIN<br />
Koszalin<br />
KAR<br />
WLO<br />
ROT<br />
TAC<br />
BLA<br />
MAS<br />
BS2<br />
ZAS<br />
JAN<br />
Poznan<br />
LT-7<br />
VDF<br />
km<br />
study area<br />
Figure 1: Measured profiles of the EMTESZ project and profile B from the Bundesanstalt für Geowissenschaften<br />
und Rohstoffe (BGR-B). Location of MT sites in Poland, Germany, Sweden and Bornholm<br />
(Denmark); STZ-Sorgenfrei-Tornquist-Zone; TTZ-Tornquist-Teiseyre-Zone; TEF-Trans European Fault;<br />
CDF-Caledonian Deformation Front; VDF-Variscan Deformation Front; AAF-Amorica Avalonia Fault.<br />
perpendicular to the assumed strike-direction of the Sorgenfrei-Tornquist Zone (see fig. 1) was<br />
chosen. It runs over about 370 km with a site spacing from 6 to 12 km.<br />
Data examples<br />
Stable transfer functions result from remote reference analysis according to Egbert & Booker<br />
(1986). The remote data is from the observatories of Belsk and Niemegk and simultaneous measuring<br />
sites. In figure 2 three representative data examples for sites of profile MVS are shown.<br />
Site ROT, which is located in the southern part of the profile, shows the characteristic curves for<br />
the North German Basin. Apparent resistivities at short periods are very low until 100 s (1 to<br />
3 Ωm), pointing to a very good conductor near the surface. With increasing periods app. resistivity<br />
is growing, accompanied by a splitting of impedance components Zxy and Zyx, reflecting<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
253<br />
STZ<br />
P2<br />
Gdansk<br />
P08<br />
TTZ<br />
TTZ<br />
Wisła R.<br />
HLP<br />
E
ρ a [Ωm]<br />
10 4<br />
10 3<br />
10 2<br />
10 1 10 1<br />
rot<br />
Zxy<br />
Zyx<br />
10 0<br />
10 1 10 2 10 2 10 3 10 4 10 5<br />
10<br />
90<br />
0<br />
10 1 10 2 10 2 10 3 10 4 10 5<br />
10<br />
90<br />
0<br />
10 1 10 2 10 2 10 3 10 4 10 5<br />
90<br />
φ ο<br />
Re<br />
Im<br />
45<br />
10 4<br />
10 3<br />
10 2<br />
10 1 10 1<br />
45<br />
boo<br />
0<br />
10 1 10 2 10 2 10 3 10 4 10 5<br />
1.0<br />
0.5<br />
0.0<br />
-0.5<br />
-1.0<br />
10 1 10 2 10 2 10 3 10 4 10 5<br />
↑Ν<br />
1.0<br />
0.5<br />
0.0<br />
-0.5<br />
-1.0<br />
10 1 10 2 10 2 10 3 10 4 10 5<br />
0<br />
10<br />
T [s]<br />
1 10 2 10 2 10 3 10 4 10 5<br />
1.0<br />
0.5<br />
0.0<br />
-0.5<br />
-1.0<br />
10 1 10 2 10 2 10 3 10 4 10 5<br />
↑Ν<br />
1.0<br />
0.5<br />
0.0<br />
-0.5<br />
-1.0<br />
10 1 10 2 10 2 10 3 10 4 10 5<br />
0<br />
10<br />
T [s]<br />
1 10 2 10 2 10 3 10 4 10 5<br />
1.0<br />
0.5<br />
0.0<br />
-0.5<br />
-1.0<br />
10 1 10 2 10 2 10 3 10 4 10 5<br />
↑Ν<br />
1.0<br />
0.5<br />
0.0<br />
-0.5<br />
-1.0<br />
10 1 10 2 10 2 10 3 10 4 10 5<br />
T [s]<br />
Figure 2: Examples of transfer functions for sites of the profile (ρa-apperent resistivity, φ-Phase and Re-<br />
Real part, Im- imaginary part of induction arrows as function of period). Site ROT as an example for the<br />
southern part of the profile shows low resitivity in short periods, which shows the thick sedimentary covers<br />
of North German Basin (ρa at low periods around 1-3 Ωm), for longer periods it shows higher resitivity.<br />
BOO is the site at the most southern part of the swedish part and shows the margin of the sedimentary<br />
basin. NYT is an example for the northern part of the profile in Southscandinavia. There the profile<br />
reaches the crystalline basement (ρa has higher values even for low periods).<br />
multidimensional structures at greater depths. Induction arrows are very small. For sites in the<br />
southern part of the profile a change of direction for induction arrows is observed. Site BOO is<br />
located at the edge of the East European Craton. Here real parts of induction arrows are very<br />
large (up to an absolute value greater than 0.8) and are perpendicular to the strike direction of the<br />
Sorgenfrei-Tornquist Zone. They point to a very good conductor in southeastern direction. Apparent<br />
resistivities for this site are even for short periods much higher than in the North German<br />
part of the profile and increase rapidly for large periods, which points to the resistive basement of<br />
the East European Craton. NYT is an example for the northernmost part of the profile and shows<br />
high apparent resistivity even for the lowest periods (about 1 000 Ωm), reflecting the crystalline<br />
basement of Baltica.<br />
Magnetic transfer functions<br />
Induction arrows at long periods increase from south to the north, hinting at well-conductive<br />
structures in the south, i.e., below the Baltic Sea. The apparent resistivities ρa in the north of<br />
the profile are relatively high and get lower at the southern sites, which agrees with the geological<br />
structure. The high resistivities in the north are caused by the crystalline basement and the low<br />
resistivities in the south by the sedimentary cover.<br />
Figure 3 shows a map of the real part of induction arrows at a period 1 820 s for the EMTESZ<br />
profiles LT-7, MVB and the discussed profile MVS, according to Wiese-Convention (Wiese, 1962).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
254<br />
Zxy<br />
Zyx<br />
10 4<br />
10 3<br />
10 2<br />
10 1 10 1<br />
45<br />
nyt<br />
Zxy<br />
Zyx
For profile MVS arrows are mainly perpendicular to the NW-SE striking direction. With values<br />
greater than 0.8, the unusually large induction arrows in the northern part of the profile indicate<br />
a good conductor below Rügen Island. Real induction arrows at the southernmost station of<br />
Sweden, the sites in the Baltic Sea and at the northernmost point of Rügen Island are largest.<br />
The ”flip-around” of induction arrows in Northern Germany and Poland was already recognized<br />
in the initial days of electromagnetic deep sounding. The underlying cause – a large conductivity<br />
anomaly at depth – was termed the North German-Polish Conductivity Anomaly (e.g.,<br />
Schmucker (1959); Untiedt (1970); Jankowski (1967)). Until today, however, the depth extent<br />
of this anomaly is controversially discussed and models range from an upper mantle high<br />
conductivity zone to the simple effect of the basin edges. The magnitude of the inductive effect<br />
at the basin margins and their rapid decrease seem to favor the second explanation, which will<br />
also become evident from 2-D inversion (see later).<br />
56°N<br />
55°N<br />
54°N<br />
53°N<br />
52°N<br />
12°E 13°E 14°E 15°E 16°E 17°E 18°E 19°E<br />
km<br />
0 50<br />
1820 sec<br />
Figure 3: Induction arrows for profiles MVS, MVB and LT-7 (real part) at a period of 1 820 s. In South<br />
Sweden and north of Rügen Island induction arrows are very large. They obviously result from the edge<br />
of the Baltic Shield (STZ). The change of direction of the real parts in the center of the NE German and<br />
Polish Basins shows the effect of the so-called ”North German-Polish Conductivity Anomaly”. It can be<br />
seen in all profiles. Profile MVS shows this effect (sign reversal of real part) approximately at the location<br />
of the Stralsund-Anklam-Fault (SAF). In the northern part of profile MVS the induction arrows are very<br />
large. The highest absolute value with more then 0.8 is reached in the southern part of Sweden and 0.7 in<br />
the northernmost part of Rügen Island.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
255
Dimensionality analysis<br />
Strike angles were calculated for single and multi sites after Smith (1995), which allows to sort<br />
out bad data with a weighting matrix (see figure 4). The strike angle for profile MVS has a strong<br />
variation especially in the northern part; the average for the profile is around N67 ◦ W.<br />
WEST<br />
NORTH<br />
SOUTH<br />
EAST<br />
Figure 4: Average strike angle after Smith (1995) for profile MVS: N67 ◦ W.<br />
Figure 5 shows the phase tensor (Caldwell et al., 2004) for each site of the profile for two<br />
different periods. The orientation of the ellipses point to the strike direction of the conductive<br />
structures. The northwest-pointing direction of the main axis of the ellipse corresponds with the<br />
calculated strike direction.<br />
a) b)<br />
56˚<br />
55˚<br />
54˚<br />
53˚<br />
PAP<br />
IBI<br />
DAR<br />
LEL<br />
WEN<br />
ROT<br />
JAB<br />
KIS<br />
ROH<br />
TET<br />
WIL<br />
BRE<br />
DOL<br />
MAT<br />
PAR<br />
TAN<br />
BAR<br />
KRU<br />
KOP<br />
NYT<br />
HOR<br />
BER<br />
REG<br />
LOV<br />
TOM<br />
BOO<br />
KRI<br />
0 50<br />
12˚ 13˚ 14˚ 15˚ 12˚ 13˚ 14˚ 15˚<br />
-16 -12 -8 -4 0 4 8 12 16<br />
β [ o ]<br />
PAP<br />
IBI<br />
DAR<br />
LEL<br />
WEN<br />
ROT<br />
JAB<br />
KIS<br />
ROH<br />
TET<br />
WIL<br />
BRE<br />
DOL<br />
MAT<br />
PAR<br />
TAN<br />
BAR<br />
KRU<br />
KOP<br />
NYT<br />
HOR<br />
BER<br />
REG<br />
LOV<br />
TOM<br />
BOO<br />
KRI<br />
0 50<br />
Figure 5: Phase tensor plot for periods of 992 s and 1 820 s calculated after Caldwell et al. (2004).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
256
Skew angle β is small throughout the study area with the exception of the northernmost sites in<br />
Sweden. Summarizing, it suffices to carry out a 2-D modeling of the data.<br />
2-D modeling<br />
All presented 2-D inversion models were calculated with the CG-FD inversion algorithm of Rodi<br />
&Mackie(2001). First the data were rotated by -67 ◦ . A homogeneous halfspace of 100 Ωm<br />
and a fine grid for the start model was chosen. The smoothing factor was set to τ=15, which is<br />
the best trade-off between model roughness and data misfit (Hansen & O’Leary, 1993). For<br />
the inversion of the 2D-models in fig. 6 all components (TE-, TM-mode and tipper) were used.<br />
In these models all resolved structures can be seen; these models have the best fitting model<br />
response to the data (see examples in fig. 7). For a better overview of the modeled features, the<br />
profile intersecting geological structures in the respective locations are marked. Furthermore, the<br />
location of the basement is shown. Figure 6a shows the result of inversion of TE-, TM-mode and<br />
tipper data after 200 iterations. The resulting RMS is 2.27. The model shown in figure 6b also<br />
results from inversion of all components (TE-, TM-mode and tipper), furthermore a horizontal<br />
weighting function was used (β=1). The resulting RMS of this inversion after 200 iterations is<br />
with 2.67 worse then the RMS of the model without weighting function, but the model is more<br />
consistent with geological assumptions (Hoffmann & Franke, 2008).<br />
a)<br />
b)<br />
z [km]<br />
z [km]<br />
0<br />
20<br />
40<br />
60<br />
80<br />
0<br />
20<br />
40<br />
60<br />
80<br />
S<br />
VDF<br />
dol<br />
bre<br />
wil<br />
tet<br />
roh<br />
kis<br />
jab<br />
rot<br />
wen<br />
lel<br />
dar<br />
ibi<br />
pap<br />
kru<br />
OA<br />
OA<br />
A<br />
A<br />
East-Avalonia<br />
Paleozoic Platform<br />
SAF CDF<br />
Strelasund<br />
-Basin<br />
D<br />
D<br />
bar<br />
Wiek<br />
trough<br />
tan<br />
par<br />
mat<br />
k31<br />
C<br />
Baltica<br />
Baltica<br />
k25<br />
boo<br />
STZ<br />
tom<br />
lov<br />
reg<br />
and<br />
ber<br />
hor<br />
nyt<br />
kop<br />
Baltica<br />
0 50 100 150 200 250 300 350<br />
distance [km]<br />
C<br />
Baltica<br />
Baltica<br />
East European Craton<br />
Figure 6: 2-D models for profile MVS inverted by applying Rodi and Mackie’s algorithm starting from<br />
a homogeneous halfspace. a) For this model all components were inverted (TE-, TM-mode and tipper<br />
data). The resulting RMS after 200 iterations was 2.27. b) For inversion of this model all components<br />
were used (TE-, TM-mode and Tipper). A horizontal weighting function was used. Resulting RMS after<br />
200 iterations was 2.67. For both models the smoothing factor was set to τ=15. Letters sign resolved<br />
conductivity structures: A-saline aquifer in Northeast Germany, Baltica-basement of the Baltic Continent,<br />
OA-basement, associated with the basement of the East-Avalonian plate. Also the assumed location of<br />
plates and fault zones is marked.<br />
In both models one can see the underlying plate of Baltica in the north as a poor conductor, the<br />
sediments in Northeast Germany as a very good conductor (structure A) and two conductivity<br />
anomalies below Rügen Island (structure C) and south of Stralsund (structure D) in depths from<br />
8 to 30 km. Figure 7 shows examples of model response of the inversion shown in figure 6a.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
257<br />
kri<br />
N<br />
4<br />
3<br />
log ρ [Ωm]<br />
2<br />
1<br />
0
Rho App. (ohm.m)<br />
Phase (deg)<br />
Tip Mag<br />
10 4<br />
10 3<br />
10 2<br />
10 1<br />
10 0<br />
90<br />
60<br />
30<br />
0<br />
1,0<br />
,5<br />
,0<br />
dol<br />
10 1<br />
TE measured<br />
TE model response<br />
rms= 1.2261<br />
10 4<br />
10 3<br />
10 2<br />
10 1<br />
10 0<br />
90<br />
60<br />
30<br />
0<br />
1,0<br />
,5<br />
lel rms= 1.2746<br />
10 2<br />
10 3<br />
10 4<br />
10 1<br />
10 2<br />
10 3<br />
,0<br />
10 4 10 1<br />
10 2<br />
10 3<br />
10 4<br />
,0<br />
4<br />
10<br />
Period(sec)<br />
1<br />
10 2<br />
,0<br />
Period(sec) Period(sec)<br />
Period(sec)<br />
TM measured<br />
TM model response<br />
Hz measured<br />
Hz model response<br />
10 4<br />
10 3<br />
10 2<br />
10 1<br />
10 0<br />
90<br />
60<br />
30<br />
0<br />
1,0<br />
,5<br />
mat rms= 1.6917<br />
10 4<br />
10 3<br />
10 2<br />
10 1<br />
10 0<br />
90<br />
60<br />
30<br />
0<br />
1,0<br />
,5<br />
boo rms= 2.1131<br />
Figure 7: Examples of model responses for the 2-D inversion result as shown in fig. 6a.<br />
Geological Interpretation<br />
Structure A seen in the models of fig. 6 is characteristic for the North German Basin; it represents<br />
the saline aquifer and is already known from previous measurements. This structure is resolved in<br />
all inversions of this profile and could be seen in all components (TE-, TM-mode and tipper) and<br />
with its vertical elongation and extension to depths of 3 to 4 km it is consistent with simulations<br />
of Magri et al. (2007) and models of the BGR-profiles in Hoffmann et al. (1997). This<br />
structure is resolved as a surface conductor in all other sections of the EMTESZ project (e.g.,<br />
Brasse et al. (2006), Ernstetal.(2008)) and in the models of Houpt (2008) as well. It is<br />
caused by huge deposits of Zechstein salt, which is dissolved by deeper ground water. Through<br />
thermal and hydraulic transport mechanisms it reaches in some cases up to the surface (Magri<br />
et al., 2005). In some parts of the North German Basin salinities up to 350 g/l are observed<br />
(Hoth et al., 1997). The conductivity of such layers depends on the salinity of the fluid, the<br />
size and connection of the pore spaces. High salt contents and pore sizes of sediment layers in<br />
North Germany may cause conductivities well above 1 S/m.<br />
Structure ”Baltica” is also resolved in all inversions of profile MVS - a very poor conductor in the<br />
north of the profile. It represents the crystalline Precambrian basement of Baltica. This basement<br />
of Baltica is with an age of 1.9 billion years one of the oldest rocks in Europe. Its high resistivity<br />
is due to the strong compression of rocks at depth with almost no interconnected fluid inclusions<br />
left.<br />
Extension and value of the conductivity of the good conductor below Rügen Island (structure<br />
C) can also be regarded as assured due to its shallow depth and the sensitivity tests carried out<br />
(Schäfer, 2010). Its boundary to the north is indicated by the size of the induction arrows at<br />
MAT, the northernmost station on Rügen Island. Structure C is located at about 8 to 20 km<br />
depth. This good conductor is depicted in all inversions as a structure separated from the surface<br />
conductor (structure A).<br />
Also structure D, with a depth of 10 to 30 km is resolved in all models of the joint inversion of TE-,<br />
TM-mode and tipper, and also proofed by sensitivity tests as a separate structure from the surface<br />
conductor. Conductors in these depths were explained with the occurrence of highly-carbonated<br />
paleozoic black shales, the so-called Scandinavian alum shales (e.g., Hoffmann et al. (1998);<br />
Hengesbach (2006)).<br />
These alum shales crop out at the Andrarum quarry in the southern Swedish province of Scania,<br />
near MT sites AND and REG, where they have been mined since centuries. In order to test<br />
the hypothesis of high conductivity, several DC-geoelectric array measurements were conducted<br />
in this area during a student field trip of the Free University of Berlin (Field Report Scania,<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
258<br />
10 3<br />
10 4
2006). Fig. 8 displays an example. It clearly shows the black shale (blue colors) as an intermediate<br />
conductor only, with resistivities in the range of 100 to 150 Ωm. Nevertheless, this does not exclude<br />
high conductivities of deeply buried shales; at or near the surface, the relatively high resistivity<br />
may be due to strong weathering, destroying the conductive paths, which are otherwise continuous<br />
in deeper layers.<br />
Figure 8: Geoelectric section near the Andrarum black shale quarry (South Sweden). Note the reverse<br />
color scale.<br />
From laboratory studies and drilling black or alum shale at greater depths and without weathering<br />
it is obtained with conductivities of less than 1 Ωm (Duba et al., 1988). Black shale has been<br />
investigated by drilling in the well of G14 (north of Rügen Island) and Rügen 5, where it reaches<br />
thicknesses of 50 to 70 m. In the well Pröttlin I further south, black shales were found, too, but<br />
could not be fully intersected. In addition, studies demonstrated that the necessary pore size for<br />
an electrolytic conduction mechanism not exists in deeper sediment layers (Jödicke, 1991). This<br />
would exclude the interpretation of structure D as a deep sedimentary trough.<br />
Thus, the conductive ”layer” in the model of figure 6b can be seen as a reference to alum shale occurrence.<br />
However, this layer can be even more conductive because of hydrothermal waters, which<br />
are often rising in geological fault zones. Under high pressure and special temperature conditions<br />
in highly-carbonated black shales also graphite-structures may form, according to estimates from<br />
Duba et al. (1988) and Raab et al. (1998). With a layer of mostly continuous black shale<br />
or graphite surface, a relatively low-friction thrust faulting with little deformation of the plate<br />
boundaries in the collision of Eastern Avalonia and Baltica at the time of the Ordovician could<br />
be explained.<br />
Furthermore, Raab et al. (1998) could demonstrate in laboratory experiments that pyrite inclusions<br />
(which often occur in black shale layers) may convert to much more conductive pyrrhotite<br />
at temperature and pressure conditions as encountered in the middle crust. Pyrrhotite usually<br />
occurs in dendritic form and increases conductivity even through small volumes.<br />
Model experiments and sensitivity tests show that a 70 m thick alum shale layer at such depths is<br />
not sufficient, even if modeled with bulk resistivities as low 0.1 Ωm. Another possible explanation<br />
is the occurrence of intrusive bodies, which rose in the time of Rotliegend (Upper Carboniferous<br />
to Middle Permian). Their contact aureoles usually consist of highly conductive material (such as<br />
ilmenite or even graphite), and can extend over larger areas. So the high conductive structures C<br />
and D might be explained by a combination of highly conductive layers of black shales and such<br />
dykes. Motivation for this hypothesis emerged from the well Greifswald I, where intrusive granite<br />
bodies from the Rotliegend with apophyses could be found. This granite body is also known as<br />
Südrügen Pluton (Hoth et al., 1993) and is possibly extending to the south of Rügen Island.<br />
In fact, the combination of these two interpretations, seems to be the best explanation of the high<br />
conductivities of structures C and D at depth. Thus, it is very likely a combination of graphitized<br />
black shale and intrusive bodies (e.g., Südrügen Pluton) with the accompanying conductive material<br />
and the associated apophyses. This interpretation would explain the very good conductivity<br />
in such depths and the resolution of the structures as an almost vertical body in the mid crust.<br />
Unfortunately no better resolution of these structures is possible due to the strong shielding effect<br />
of the saline aquifer.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
259
The good conductor of model 6b ends in the south of the Stralsund Anklam Fault (SAF). The trend<br />
of the SAF, which is also indicated by magnetic transfer functions and their graphic presentation<br />
in fig. 3 fits very well to the results of Houpt (2008) and Ernst et al. (2008), which were<br />
obtained further east. According to these models and the induction arrows shown in fig. 3, the<br />
SAF appears to be a fundamental crustal boundary of Northeastern Europe which can be traced<br />
to the southeast of Szczecin in Poland.<br />
Figure 9 shows a comparison with an interpretation of the BASIN 9601 project and 2 prolonging<br />
profiles in the Baltic Sea from McCann & Krawczyk (2001). There we observe a good correlation<br />
with underlying Baltica and the Zechstein base (strong reflector), and the well-conducting<br />
sediments in the upper layers of the Northeast German Basin.<br />
z [km]<br />
0<br />
20<br />
40<br />
60<br />
80<br />
S VDF<br />
SAF CDF<br />
dol<br />
bre<br />
wil<br />
tet<br />
roh<br />
kis<br />
jab<br />
A<br />
OA<br />
Moho<br />
Baltica<br />
Baltica<br />
0 50 100 150 200 250 300 350<br />
distance [km]<br />
STZ<br />
Grimmen High Baltic Sea<br />
G14<br />
ber<br />
hor<br />
nyt<br />
kop<br />
Figure 9: 2-D models for profile MVS with an overlayed interpretation of the seismic reflection/refraction<br />
profile BASIN 9601 and 2 prolonging profiles in the Baltic Sea from McCann & Krawczyk (2001).<br />
References<br />
Brasse, H., Červ, V., Ernst, T., Jó´zwiak, W., Pedersen, L., Varentsov, I. & Pomerania<br />
working group, E. (2006). Probing the electrical conductivity structure of the Trans-<br />
European Suture Zone. EOS Trans. AGU, 87, 29, doi:10.1029/2006EO290002.<br />
Caldwell, T., Bibby, H.M. & Brown, C. (2004). The magnetotelluric phase tensor. Geophys.<br />
J. Int., 158, 457–469.<br />
Duba, A., Huenges, E., Nover, G., Will, G. & Jödicke, H. (1988). Impedance of black<br />
shale from Münsterland 1 borehole: An anomalously good conductor? Geophys. J., 94, 413–419.<br />
Egbert,G.D.&Booker,J.R.(1986). Robust estimation of geomagnetic transfer functions.<br />
Geophysical Journal of the Royal Astronomical Society, 87, 173–194.<br />
Ernst, T., Brasse, H., Červ, V., Hoffmann, N., Jankowski, J., Jó´zwiak, W., Kreutzmann,<br />
A., Neska, A., Palshin, N., Pedersen, L., Smirnov, M., Sokolova, E. &<br />
Varentsov, I. (2008). Electromagnetic images of the deep structure of the Trans-European<br />
Suture Zone beneath Polish Pomerania. Geophys. Res. Lett., 35, doi:10.1029/2007GL034610.<br />
Field Report Scania (2006). Investigation of near-surface structures of the Sorgenfrei-Tornquist-<br />
Zone (STZ) at the margin of Baltica in Scania (Sweden). Geophysical Excursion to Scania<br />
(Sweden)-Field Report, FU Berlin, Fachrichtung Geophysik, Berlin.<br />
Gee, D. G. & Zeyen, H. (1996). Europrobe 1996-lithosphere dynamics. origin and evolution of<br />
continents. EUROPROBE Secretariat, Uppsala University, EUROPROBE 1996 (eds.), pp.138.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
260<br />
kri<br />
N<br />
4<br />
3<br />
log ρ [Ωm]<br />
2<br />
1<br />
0
Hansen,P.C.&O’Leary,D.P.(1993). The use of the L-curve in the regularization of discrete<br />
ill-posed problems. SIAM J. Sci. Comput., 14, 1487–1503.<br />
Hengesbach, L. (2006). Magnetotellurische Studien im Nordwestdeutschen Becken: Ein Beitrag<br />
zur paleaogeographischen Entwicklung des Unterkarbons. Dissertation, Universität Münster.<br />
Hoffmann, N. & Franke, D. (2008). The Avalonia-Baltica suture in NE Germany - new constraints<br />
and alternative interpretations. Abschlußbericht, Bundesanstalt für Geowissenschaften<br />
und Rohstoffe, Hannover.<br />
Hoffmann, N., Fluche, B., Jödicke, H., Jording, A., Kallaus, G., Müller, W. & Pasternak,<br />
G. (1997). Erforschung des tieferen Untergrundes der Nordwestdeutschen Senke - ein<br />
Beitrag der Magnetotellurik zur Untersuchung des präwestfalen Muttergesteinpotentials. Abschlußbericht,<br />
Bundesanstalt für Geowissenschaften und Rohstoffe, Hannover.<br />
Hoffmann, N., Jödicke, H., Fluche, B., Jording, A. & Müller, W. (1998). Modellvorstellungen<br />
zur Verbreitung potentieller präwestfalischer Erdgas-Muttergesteine in Norddeutschland<br />
– Ergebnisse neuer magnetotellurischer Messungen. Z. angew. Geol., 44, 140–158.<br />
Hoth, K., Rusbühlt, J., Zagora, K., Beer, H. & Hartmann, O. (1993). Die tiefen Bohrungen<br />
im Zentralabschnitt der Mitteleuropäischen Senke. Dokumentation für den Zeitabschnitt<br />
1962 bis 1990. In: Schriftenr. Geowiss. Berlin, 7–145.<br />
Hoth, P., Seibt, A., Kellner, T. & Huenges, E. (1997). Geowissenschaftliche Bewertungsgrundlagen<br />
zur Nutzung hydrogeothermaler Ressourcen in Norddeutschland. <strong>GFZ</strong> Potsdam - Sc.<br />
Techn. Report, 15.<br />
Houpt, L. (2008). Neue elektromagnetische Untersuchungen zur norddeutschen Leifähigkeitsanomalie.<br />
Diploma thesis, Fachrichtung Geophysik, FU Berlin.<br />
Jankowski, J. (1967). The Marginal Structures of the East European Platform in Poland on<br />
Basis of Data on Geomagnetic Field Variations. Polish Scientific Publishers, 93-102.<br />
Jödicke, H. (1991). Zonen hoher elektrischer Krustenleitfäigkeit im Rhenoherzynikum und<br />
seinem nördlichen Vorland. Dissertation, Universität Münster, Münster, Hamburg.<br />
Magri, F., Bayer, U., Jahnke, C., Clausnitzer, V., Diersch, H. J., Fuhrmann, J.,<br />
Möller, P., Pekdeger, A., Tesmer, M. & Voigt, H. J. (2005). Fluid-dynamics driving<br />
saline water in the North East German Basin. Int. J. Earth Sci., 94, 1056–1069.<br />
Magri,F.,Bayer,U.,Tesmer,M.,Möller, P. & Pekdeger, A. (2007). Salinization problems<br />
in the NEGB: results from thermohaline simulations. Int. J. Earth Sci., 97, 1075–1085.<br />
McCann, T. & Krawczyk, C. M. (2001). The Trans-European Fault: a critical reassessment.<br />
Geol. Mag., 138 (1), 19–29.<br />
Pharaoh, T. C. (1999). Paleozoic terranes and their lithospheric boundaries within the Trans-<br />
European Sutur Zone (TESZ): a review. Tectonophysics, 314, 17–41.<br />
Raab, S., Hoth, P., Huenges, E. & Müller, H. J. (1998). Role of sulfur and carbon in the<br />
electrical conductivity of the middle crust. J. Geophys. Res., 103, 9681–9689.<br />
Rodi,W.&Mackie,R.L.(2001). Nonlinear conjugate gradients algorithm for 2-D magnetotelluric<br />
inversions. Geophysics, 66, 174–187.<br />
Schäfer, A. (2010). Magnetotellurische Untersuchung der Transeuropäischen Suturzone auf einer<br />
Traverse von Südschweden nach Nordostdeutschland. Diploma thesis, Fachrichtung Geophysik,<br />
FU Berlin.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
261
Schmucker, U. (1959). Erdmagnetische Tiefensondierung in Deutschland 1957 – 59; Magnetogramme<br />
und erste Auswertung. Abh. Akad. Wiss., Göttingen. Math.–Phys. Kl. Heft 5.<br />
Smith, J. T. (1995). Understanding telluric distortion matrices. Geophysical Journal International,<br />
122, 219–226.<br />
Untiedt, J. (1970). Conductivity anomalies in Central and Southern Europe. J. Geomag. Geoel.,<br />
22, 131–149.<br />
Wiese, H. (1962). Geomagnetische Tiefentellurik Teil II: Die Streichrichtung der Untergrundstrukturen<br />
des elektrischen Widerstandes, erschlossen aus geomagnetischen Variationen. Geofis.<br />
Pura e Appl., 52, 83–103.<br />
Acknowledgement<br />
Other members of the EMTESZ Working Group include T. Ernst, V. Červ, J. Jankowski, W.<br />
Jó´zwiak, A. Kreutzmann, A. Neska, N. Palshin, L. Pedersen, M. Smirnov, E. Sokolova and I.<br />
Varentsov. We thank Belsk and Niemegk Observatories for supplying data and German Science<br />
Foundation (DFG) for funding.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
262
Magnetotelluric data from the Tien Shan and Pamir<br />
continental collision zones, Central Asia<br />
P. Sass*, O. Ritter*, A. Rybin**, G. Muñoz*, V. Batalev** and M. Gil*<br />
*Helmholtz Centre Potsdam <strong>GFZ</strong>, German Research Centre for Geosciences<br />
**Research Station of the Russian Academy of Sciences, Bishkek, Kyrgyzstan<br />
1 Intoduction<br />
We present magnetotelluic (MT) data obtained within the framework of the multidisciplinary<br />
Tien Shan - Pamir Geodynamic Program (TIPAGE). The dynamics of the<br />
Tien Shan and Pamir orogenic belts are dominated by the collision of the Indian and<br />
Eurasian continental plates. With the geophysical components, we intend to image the<br />
deepest active intra-continental subduction zones on Earth (the N-dipping Hindu Kush<br />
and the S-dipping Pamir zones) and to establish how the highest strain over the shortest<br />
distance that is manifested in the India-Asia collision zone is accommodated structurally.<br />
The Tien Shan-Pamirs mountain knot forms the northwestern corner of the India-Asia<br />
Figure 1: Central and eastern Asia orogens. TIPAGE and INDEPTH surveys are<br />
marked on the map. TIPAGE investigation area lies inside the blue circle.<br />
collision zone and the Pamir-Tibet Plateau (Fig. 1). Note that N - S shortening in<br />
the Pamir and western Tien Shan is absorbed in less than 50% of the distance than<br />
further east. Basins denote areas little affected by intra-continental shortening. The<br />
Pamir excels over the adjacent Tibet by including the most active areas of intermediatedepth<br />
seismicity in the world and by far the most active one not associated with oceanic<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
263
subduction, three coevally active, thick-skinned, high-strain contractional belts and four<br />
distinct intra-continental magmatic belts.<br />
2 Measurements<br />
The MT data were recorded in summer 2008 at 80 stations in the Pamir mountain<br />
ranges in Tajikistan and in summer 2009 at 98 stations in the Pamir and the southern<br />
Tien Shan in Kyrgyzstan and Tajikistan(Fig. 2). A typical spacing was approximately<br />
2 km between BB-only sites and 14 km for the combined BB+LMT sites. The stations<br />
form an approximately 340 km long profile from Osh in Kyrgyzstan via Sarytash, the<br />
Kyrgyz-Tajik border, Karakul and Murgab to Zorkul in southern Tajikistan.<br />
72˚<br />
73˚<br />
74˚<br />
75˚<br />
41˚ 41˚<br />
Uzbekistan<br />
Osh<br />
Kyrgyzstan<br />
40˚ 40˚<br />
39˚<br />
Kara-Kul<br />
Lake 186L<br />
39˚<br />
Tajikistan<br />
Sary-Tash<br />
Afganistan<br />
China<br />
7000<br />
Sarez<br />
Lake<br />
6000<br />
Murghab<br />
38˚<br />
5000<br />
38˚<br />
m<br />
134<br />
116L<br />
102L<br />
088L<br />
074L<br />
km<br />
356<br />
354L<br />
340L<br />
326L<br />
312L<br />
298L<br />
270L<br />
256L<br />
242L<br />
228L<br />
214L<br />
284L<br />
4000<br />
3000<br />
2000<br />
1000<br />
37˚<br />
0 50<br />
0<br />
37˚<br />
72˚<br />
73˚<br />
74˚<br />
75˚<br />
200L<br />
172L<br />
158L<br />
144L<br />
130L<br />
060L<br />
046L<br />
000<br />
032L<br />
018L<br />
004L<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
264<br />
Figure 2: Topographic map with<br />
stations and profile of the TIPAGE<br />
survey. Numbers of stations with<br />
broad band and long period measurements<br />
(L) are marked. The station<br />
alignement follows the a single<br />
road in that remote area with rough<br />
terrain.
3 Preliminary Data Analysis<br />
Data examples<br />
Some representative graphs of apparent resistivities and phases are shown in Fig. 3.<br />
The curves were obtained after a recording time of three days, using robust single site<br />
processing. Varying shapes of the curves indicate considerable changes in the underlying<br />
conductivity structure. The distances between the shown stations are approximately<br />
50 km. For station locations, see Fig. 2. The overall data quality is superb, especially<br />
in the very remote southern parts of the profile.<br />
Figure 3: Exemplary apparent resistivity and phase curves of four broad-band stations<br />
from the Pamir. For station locations, see Fig. 2.<br />
Pseudosections<br />
Apparent resistivities and impendance phases of all TIPAGE stations are plotted as<br />
pseudo sections in Fig. 4. In the southern part of the profile, the TE and TM modes<br />
apparent resistivities show values below 10 Ωm at higher periodes and high phase values<br />
reaching 90 ◦ . There are more low-resistivity features further north at middle to high<br />
periods, which are more present in the TE mode. This hints at a deep conducting<br />
structure in the southern profile part and a complex distribution of conducting features<br />
in the northern segment of the profile.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
265
Figure 4: Pseudosections of apparent resistivity and phase data, in TE and TM modes.<br />
In the left part of each pseudosection the southern profile stations are plotted, in the<br />
right part the northern stations. For stations location, see Fig. 2.<br />
Induction arrows<br />
The presence or absense of lateral variation in conductivity can be inferred from induction<br />
arrow maps. In Fig. 5 induction arrows are plotted using the Wiese convention for<br />
periods of 1, 32, 128, and 1024 s. The arrow distribution indicates several ’reversals’,<br />
which suggest the presence of elongated good conductors below the profile. Alltogether<br />
the induction arrow plots exhibit a complex distribution of the subsurface conductivity<br />
structure.<br />
Strike analysis<br />
Regional strike analysis (Becken & Burkhard, 2003) of sites recorded in the Pamir and<br />
Tien Shan is displayed in Fig. 6. The analysis of surface sensitive higher frequencies (Fig.<br />
6 left) shows a E - W (N - S) strike direction. This is consistent with the predominant<br />
E - W distribution of geological structures. The lower frequencies (Fig. 6 right) reveal a<br />
pronounced geoelectric strike of approximately −17 ◦ to −20 ◦ . This strike direction may<br />
be in agreement with structures of the India-Asian collision zone.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
266
Figure 5: Maps of induction vectors (Wiese convention) for periods of 1, 32, 128, and<br />
1024 s. Each arrow starts at one station location, compare Fig. 2.<br />
Figure 6: Regional strike analysis (Becken & Burkhard, 2003) of all sites recorded in<br />
the years 2008 and 2009. Left for the period range 0.001 s - 10 s, right for the period<br />
range 10 s - 10000 s.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
267
Figure 7: The Kyrgyz/Russian - Tajik - German MT team of 2008 (left image) and<br />
2009 (right image). The Research station of the Russian Academy of Sciences in Bishkek<br />
provided most of the field logistics, including 4WD vans and 4WD trucks.<br />
Figure 8: The amazing hospitality of local people during the field campaigns was a<br />
memorable experience.<br />
4 Acknowledgments<br />
The MT field team: Employees and students of the international Research Center,<br />
Russian Academy of Sciences, Bishkek and from Germany: D. Brändlein, X. Chen,<br />
T. Krings, A. Nube, M. Schüler, K. Tietze, C. Twardzik and G. Willkommen.<br />
Dr. V. E. Minaev, Dr. N. Radjabov, Dr. I. Oimahmadov, Prof. A. R. Faiziev: Institute<br />
of Geology, Academy of Sciences of the Republic of Tajikistan, Dushanbe.<br />
Dr. B. Moldobekov, Prof. H. Echtler, Dr. A. Mikolaichuk: Central Asian Institute for<br />
Applied Geosciences, Bishkek.<br />
Dr. S. K. Negmatullaev: PMP International / Seismic Monitoring Network in Tajikistan,<br />
Dushanbe.<br />
All other German colleagues participating in the TIPAGE collaborative research project.<br />
We very gratefully acknowledge substantial funding which we received from the <strong>GFZ</strong><br />
and the DFG. The magnetotelluric instruments were provided by the <strong>GFZ</strong> Geophysical<br />
Instrument Pool (GIPP).<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
268
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
269
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
270<br />
16 2/ 3
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
271<br />
E x<br />
E y<br />
= Z xx<br />
Z yx<br />
Z xy<br />
Z yy<br />
B x<br />
B y
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
272
Z B =0.5· (Z xy − Z yx )<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
273
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
274
·<br />
·<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
275
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
276
In our session "What I always wanted to know ..." we had the question<br />
How would you choose the period for LOTEM?<br />
Carsten Scholl continued thinking about it even after the colloquium and send an email with some<br />
additional thoughts. Here is his email:<br />
Well, in the discussion I said that it is not necessary to wait until the signal is decayed into the noise<br />
level before switching again, which is totally true. Actually, this would be to some extent a circular<br />
argument, because the noise level depends on the number of stacks, and the number of stacks<br />
depends on the period.<br />
So, one should check before with synthetic models, which time range is required to resolve a certain<br />
feature. Note, the time range to RESOLVE a feature, is not the same (and typically significantly larger)<br />
than the time range up to the point where deviations between the “target” and “no target” curves<br />
appear. I recommend doing 1D inversions with reasonable noise estimates for resolution studies… .<br />
Further, it is advisable to increase this time, because the true background resistivity might deviate<br />
from the synthetic und thus delay the features representing the target.<br />
The time saved by using a higher period and thus collecting the required number of stacks faster can<br />
be used by measuring at more sites.<br />
Often, in academic LOTEM campaigns, the target depth is not well defined but you’d like to get as<br />
deep as possible. Further, setting up LOTEM stations is tedious and the number of stations is limited<br />
anyways because of limited equipment. Often, the time spent on deploying and packing up the<br />
sensors exceeds the actual measurement time. So, in this case, there is not really the option to<br />
measure more sites on the same day (well, I always envision a roll-along scheme, where you start to<br />
measure at one station while building up the next site. When the final site is set up, the first site is<br />
redeployed, while the TX is still running…).<br />
In this case, I recommend to take some generous period the first day. In the evening the data for the<br />
day should be processed to see what the latest usable time is (this should be done with either a<br />
switch-off E-field or a magnetic component). For the subsequent days, the period should be set to<br />
something slightly longer (say, a factor of 2) longer than this time.<br />
Another aspect of this question: You should check, whether your time domain forward solver actually<br />
can handle higher switching times by taking into account previous transients. Otherwise, it might be<br />
better to stick to a more conservative period as well.<br />
23. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung,<br />
Heimvolkshochschule am Seddiner See, 28 September - 2 Oktober 2009<br />
277