Biogeography, diversity and environmental relationships of shelf and deep-sea benthic Amphipoda around Iceland

Amphipoda data

We compiled data on occurrences and abundance of 355 amphipod species for 532 localities from the literature. The following expeditions and respective data sources were used: extensive literature search, data from BIOICE and IceAGE expeditions. The assembled dataset was highly heterogenous regarding sampling effort and method, time, location and date of the different expeditions. Many only listed one or two species, in particular the historic data from e.g., Boeck (1861), Hansen (1887) and Stephensen (1933, 1938, 1942, 1944a, 1944b) only providing occurrence data. However, other had high abundances of individuals (max: 2,709) and high species richness (max: 72). Due to this high heterogeneity, we aggregated the data at a coarser spatial resolution.

A common approach is to construct a coarse rectangular grid in which species occurrences are joined. We constructed a hexagonal grid using QGIS (QGIS Development Team, 2019) with a horizontal diameter of 1° per grid cell. Within each grid cell, the occurrence and abundance information were pooled, so that a grid cell contained information from multiple localities but species were not double counted, yet the sum of the abundances per species could be calculated. Hexagonal grids have several advantages over rectangular grids, e.g., symmetric neighbourhood relations or reduced edge effects (Birch, Oom & Beecham, 2007). For our study the most compelling reason to favour a hexagonal grid was the match of the polygons to the coastlines of Iceland and Greenland. Hexagonal grids provided a much better fit to this jagged pattern with an appropriate size, whereas rectangular grid cells would have to be much smaller and would then be too small for the purposes of our sampling. Given the case that many of our samples were near the coast, the hexagonal design clearly improved our sampling design.

Environmental layers

We extracted twelve variables from the Bio-Oracle 2.0 database (Assis et al., 2018) using the sdmpredictors package (Bosch, Tyberghein & De Clerck, 2018). Variables were chosen to represent major environmental deep-sea gradients (Table 1). All variables, except minimum depth, represented long-term maximum values modelled at minimum depths on a raster with 7 km² resolution per cell. In order to use the parameters on the same spatial scale as the species data, we aggregated the raster data to the scale of the hexagonal grid cells by calculating the mean raster value for each grid cell. Hexagons then represented the summed species abundances and averaged environmental data.

We analysed the environmental data for multicollinearity on the level of the hexagons. We calculated a Pearson correlation matrix (AppS1) for all environmental layers and removed all layers with a Pearson’s r above 0.75. As expected, we found strong correlation between parameters of the same information type, i.e., Chl-a and primary productivity or all nutrient related parameters. Finally, we retained the following parameters: depth, pH, phytobiomass, velocity, dissolved oxygen, dissolved iron, salinity and temperature. We kept salinity although it was correlated with temperature as it is an important parameter to structuring deep-sea communities around Iceland (e.g., Weisshappel & Svavarsson, 1998).

Environmental cluster analysis

We hypothesized that deep-sea regions with similar environmental conditions would have a similar benthic fauna. Hence, we clustered the hexagonal grid cells based on the reduced set of the averaged environmental layers into a small set of environmentally homogenous regions. We used the mclust package (Scrucca et al., 2016) to conduct model-based hierarchical clustering using finite Gaussian Mixtures. The clustering algorithm compares 14 differently shaped types of Gaussian covariance structures representing different kinds of elliptical shapes ordered by an increasing complexity. The different models are compared using the Bayesian Information Criterion (Burnham & Anderson, 2002) choosing the model with the lowest complexity. Based on the plot of the different BIC models for possible cluster sizes from 2 to 10 (S3), we identified the optimal cluster as that one with highest regionalization capacity, i.e., having a low number of clusters but already touching the plateau of the curve, signalling little differences in the model fit. We further confirmed the optimal number of clusters using a bootstrapped sequential likelihood ratio test (Scrucca et al., 2016) by comparing an increasing number of cluster sizes. Finally, we calculated mean, standard deviation, minimum and maxima for each parameter and cluster combination. This was done to allow an interpretation of the environmental conditions representing the clusters.

Taxonomic data

To interpret the overlap between clusters in terms of species composition, we first performed a constrained analysis of principal coordinates (CAP) (Anderson & Willis, 2003) with presence absence information and the Jaccard distance measure. CAP is an ordination technique, that allows to visualize similarities in sites based on species composition and environmental correlates. The ordination diagram was visually inspected by plotting the sites encircled by hulls on the first two axes. We further calculated the ANOSIM statistic on presence/absence transformed species data. ANOSIM is a non-parametric method to measure the community-wise overlap between different clusters (Clarke, 1993). It yields a statistic called R that is in the range from 0 to 1 with values of R below 0.5 indicating strong overlap. The statistic is tested for significance using a permuted p-value (n = 9,999). R-values above 0.75 indicate largely non-overlapping clusters with strongly different species composition. Both analyses were performed using the vegan package (Oksanen et al., 2019).

To identify characteristic species for each cluster, we identified all species being positively associated with one specific cluster or combinations of clusters using the multipatt function of the indicspecies package (Cáceres & Legendre, 2009). We used the group-size corrected Indicator Value (IndVal.g) as a measure of association. The null hypothesis tested is that the association of a species is not higher in a specific cluster than in the other clusters. This function calculates a p-value based on 9,999 permutations, which is not corrected for multiple testing. However, as we are not interested in the number of indicator species, but in whether a species has a high association to a cluster or not, the p-values do not have to be adjusted (De Cáceres, Legendre & Moretti, 2010). After the analysis, species with high association values were extracted as lists for each cluster combination. The resulting species-cluster relationship was compared with literature and information from the World Register of Marine Species (WoRMS) database (Horton et al., 2021).

Diversity

We aimed to compare amphipod diversity between the different clusters. However, due to different numbers of samples (n = 136), i.e., hexagonal cells, that contained the species data, clusters were not directly comparable in terms of diversity. Hence, we conducted a combined rarefaction-extrapolation analysis based on Hill numbers (Chao, Chiu & Jost, 2014). The concept of comparing species diversity using Hill numbers stems from the fact that most diversity indices are measures of entropy, such as Shannon or Simpson and do not translate directly into a diversity measure although often applied in such a way (Jost, 2006). Yet three well known measures of diversity i.e., species richness, Shannon and Simpson diversity (Shannon & Weaver, 1949; Simpson, 1949) can be generalized by a formula derived by Hill (Jost, 2006; Chao, Chiu & Jost, 2014) which orders the indices along an order of q, i.e., q = 0, 1, 2 translating to richness, Shannon and Simpson, respectively. This order reflects an increasing importance of the evenness component of diversity, while the richness component becomes less effective. This means that for richness, there is no effect of abundance on the diversity measure, while for the Simpson index, rare species only have little effect on the estimated diversity values. Hence, the Simpson index is often thought to be the most robust index, when number of individuals strongly differ, as is the case here. The diversity information is transformed into a common measure of diversity, the effective number of species, which is the number of species having equal abundances that would be required to reach e.g., the Shannon entropy value of the sample. This measure allows comparisons of all three different indices having the same unit, the effective number of species. We performed the analysis using the iNEXT package (Hsieh, Ma & Chao, 2016) based on the summed abundance vectors per species and cluster.

When studying deep-sea organisms, the most important indirect environmental gradient is depth in meters. In order to evaluate the diversity pattern related to depth we studied the original data from the different stations (n = 532) and expeditions. First, we calculated a Poisson Generalized Linear Model (GLM) to quantify the relationship between the number of species per station and depth in meters. Then we split the depth gradient in 100-m intervals to study the trend of the maximum number of species across the depth gradient.

General

The total number of amphipod individuals analyzed is n = 71,108. The assembled dataset contained 355 species from 141 genera and 44 families (Tab. 2). From these, 101 species were only be identified to the genus level, where species were given a numerical code. The original number of stations from the expeditions (n = 532) were reduced to a set of 136 one-degree wide hexagonal cells in order to reduce the heterogeneity in the dataset. These hexagonal cells were clustered according to their environmental conditions. The entire dataset is available via Peer J supplement as well as Pangaea (GfBIO) https://doi.pangaea.de/10.1594/PANGAEA.931959 (Lörz et al., 2021).

Environmental clusters

The mclust algorithm identified six clusters to be the optimal configuration according to BIC and the likelihood ratio tests. However, when aggregating the species data to six clusters, this would result in clusters with disproportionally large differences in samples per cluster. Hence, we reduced the final number of clusters to four (Fig. 1). As the clustering is hierarchical, and the four-cluster solution is not much worse in terms of BIC we were confident that this aggregation is more informative with regard to the species than the six-cluster solution which would have split the northern and southern clusters into separate regions for the specific basins (the six cluster map is shown in the Supplementary Data). The four-cluster solution also provides a good overview of the large-scale spatial pattern. There is a “Coastal” cluster (n = 34 cells) which is always close to the coastline and is characterized by shallow depth, high amounts of dissolved iron and phytobiomass and warm, oxygen-rich waters with a high current speed (Fig. 2). The second cluster resembles the GIFR (n = 55), which spreads from west to east and separates the northern and southern basis. In many points it is similar to the coastal cluster but is deeper and with less dissolved iron, oxygen, and phytobiomass. The other two clusters are called “Deep South” (n = 19) and “Deep North” (n = 28) as they represent the deep-sea regions around Iceland. They differ strongly from the first two clusters by having very low values for many parameters. “Deep North” differs from “Deep South” by being much colder, with almost no current velocity. Further, “Deep North” has a much higher amount of dissolved oxygen and pH. The lowest depths of around 3,400 m are observed in the Aegir ridge. These four clusters thus characterize the environmental conditions around Iceland on a regional spatial scale.

Constrained ordination

We conducted a constrained ordination to verify the amount of variation explained in the species data by the environmental information contained at the level of the hexagonal cells. The constrained axes of the ordination explained 11% of the total variation, while 89% is explained by the 357 unconstrained axes. According to a permutation test of the marginal effects of each environmental variable carried out using the anova.cca function of the vegan package, the most important environmental variables were temperature (F = 2.34, p < 0.001), depth (F = 2.123, p < 0.001), and salinity (F = 2.01, p < 0.001).

The four different clusters strongly overlapped in ordination space (Fig. 3A). The ANOSIM-R value of 0.197 signals considerable similarity in species composition between the clusters. All clusters overlap in the centre of the diagram; their large spread indicates strong heterogeneity. The deep-sea clusters overlapped less than the coastal and GIFR-cluster. In general, the first constrained axis represented the depth gradient, which was in contrast to all other variables. Salinity, temperature and pH characterized the second constrained axes, with pH being in contrast to temperature and salinity (Fig. 3A). The species pattern clumped near the centroid of the ordination diagram (Fig. 3B) indicating that many species are found in intermediate environmental conditions. Fewer species have a clear centroid in deeper waters, instead many species favour higher temperatures and an above average salinity. Large variation appears in the direction of pH and dissolved iron, as indicated by the strong scatter of species centroids (Fig. 3B).

Indicator species analysis

To characterize the different clusters with regard to faithful species, i.e., so-called indicator species we conducted a multipattern indicator species analysis. We compared 15 different combinations with an increasing number of clusters. From 355 species, we identified 56 to have a strong association to one or more clusters. Fourty-three species were associated to one cluster only, while twelve and one species were associated to two and three clusters, respectively (Table 3). Only two species were found for the GIFR cluster, but more species from GIFR appear in combination with other clusters.

Three of the clusters, the Deep North, the Deep South and the Coastal have indicator species belonging to the genus Rhachotropis. While different species of a genus might be specialized on different diets, all Rhachotropis species are very good swimmers (Lörz, 2010). The Deep South cluster has four Rhachotropis as indicator species. While the GIFR cluster only had two endobenthic species, belonging to the family Ampeliscidae, which are not considered strong swimmers (Peart, 2018), the combined GIFR and coastal cluster indicate Rhachotropis aculeata (Lepechin, 1780) as an indicator—a species that is known to have a circum-Arctic distribution (Lörz et al., 2018). Caprella microtuberculata G. O. Sars, 1879 and Aeginella spinosa Boeck, 1861 are indicator species of the combined coastal and GIFR cluster. These two species belong to the amphipod group Caprellidae, skeleton or ghost shrimps, which are known for their clinging lifestyle. The indicator species with the highest values, over 0.5, are Cleippides quadricuspis Heller, 1875 from the Deep North, Eusirus holmi Hansen, 1887 from the combined Coastal and Deep North cluster and Rhachotropis thordisae Thurston, 1980 from the Deep South cluster—these three species are all large amphipods of several cm body length and known as predators (Lörz et al., 2018).

Diversity

The number of aggregated hexagonal cells differed for each cluster, hence we had to apply a rarefaction and extrapolation analysis to make the three diversity measures comparable. The rarefaction of the summed abundances revealed that the two clusters “coastal” and “GIFR” have about twice the number of species than the deep-sea clusters (Fig. 4A). This even holds when only the lowest comparable value of approximate 10,000 individuals is considered. Although there were so many individuals per cluster, the curves do not level off, indicating that still more sampling would be required to reach a plateau in species richness. The Shannon diversity (Fig. 4B) considers the richness-abundance component of diversity. The “coastal” and “GIFR” clusters are at the same level of 60 effective species; the deep-sea clusters again have a much lower diversity, i.e., almost three times lower. All curves reach a plateau, indicating that there is little more diversity to expect when abundances are considered. Hence, only rare species might be added by future sampling. Considering the Simpson diversity (Fig. 4C), i.e., when no rare species but only dominant species have an influence on the diversity measure, then the “coastal” cluster becomes the most diverse cluster while the “GIFR” is only half as diverse as the coastal cluster.

The richness pattern across the depth gradient showed high variation at depths above 1,500 m with richness values up to 79 species per station (Fig. 5A). Most of the stations recorded rather few species i.e., up to 10 species with an average of 20 species at the shallowest parts (18 m) and an estimated richess of eight species at the lowest depths. The trend for the maximum number of species aggregated per 100-m interval showed an unimodal pattern with a peak at depths around 500 m and a much lower richness at depths lower than 1,000 m (Fig. 5B). These figures support the finding that the Coastal and GIFR clusters are much more diverse than the deep-sea clusters (Fig. 5C).

Environmental and historical imprints on amphipod distributions

Distributional groupings given in the present study corresponded to earlier findings, in which distinctive boundaries between a northern and a southern deep-sea fauna were inferred, while the composition of the shallow-water fauna (<500 m) around Iceland was very similar (Weisshappel & Svavarsson, 1998; Weisshappel, 2000; Bett, 2001; Weisshappel, 2001). Unsurprisingly, the spatial distribution of amphipods appeared to be most strongly influenced by bathymetry, salinity and seafloor temperature. The latter two were interconnected and indicative of particular water masses (Puerta et al., 2020).

The presence of the GIFR is known as an effective barrier to disrupt the dispersal of benthic organisms between the North Atlantic and the Nordic seas (Weisshappel & Svavarsson, 1998; Brix & Svavarsson, 2010; Schnurr et al., 2018). With a saddle depth averaging 600 m in the Strait of Denmark and 480 m between Iceland and the Faroe Islands and a maximum depth of c. 840 m, the depth increases towards the abyssal basins on each side of the ridge exceeding 3,000 m. Depth, or rather ecological and environmental variables that change with depth, such as hydrostatic pressure, temperature, food availability, or competition, have been demonstrated to have a large impact on species distributions (Rex & Etter, 2010; Brown & Thatje, 2011; Tittensor et al., 2011). In contrast, there are several examples of amphipod species, mostly within the more motile scavenger and predator guilds, with large depth distributions and thus at least the intrinsic capability to overcome topographical barriers (Lacey et al., 2018; Lörz, Jażdżewska & Brandt, 2018; Weston et al., 2021).

The GIFR also marks the transition between different bodies of water, and hence the effects of depth and water mass properties are intertwined. Generally, physical and chemical water mass attributes such as temperature, salinity, pH, organic matter, and dissolved oxygen play critical roles in structuring benthic communities including microbes, fish, crustaceans, corals, and sponges (Koslow, 1993; Weisshappel & Svavarsson, 1998; Brix & Svavarsson, 2010; Schnurr et al., 2018; Puerta et al., 2020; Roberts et al., 2021). Reasons for this involve physiological tolerances of larvae, juveniles and adults towards certain environmental conditions, dispersal constraints invoked by density differences or current shear, as well as enhanced nutrient input linked to hydrography (Puerta et al., 2020; Roberts et al., 2021).

Obviously, cold sub-zero temperatures in the Nordic sea basins restrict species distributions, as only few species are pre-adapted to such low temperatures while withstanding high hydrostatic pressures (Svavarsson, Stromberg & Brattegard, 1993; Brown & Thatje, 2011). This is supported by the fact that many amphipod species in our study prefer moderate conditions, at least in terms of temperature. Initially, however, species originating from the North Atlantic had to overcome the GIFR and enter the Nordic seas against the overflow water from the Denmark Strait and Faroe Bank Channel (Yasuhara et al., 2008), the latter being limited to species with broad bathymetric distributions or eurytherm “shallow”-water taxa. The presence of the GIFR is thereby inevitably linked to the opening of the North-east Atlantic about 55 Mya, representing a barrier between the Nordic seas and North Atlantic ever since (Hjartarson, Erlendsson & Blischke, 2017). Alternatively, species from the North Pacific had to cross the Bering Strait sill, and experience subsequent trans-Arctic migration (Hardy et al., 2011). While the shelf fauna represents a mixture of North Pacific, North Atlantic and to a lesser extent endemic Arctic fauna (e.g., Svavarsson, Stromberg & Brattegard, 1993; Hardy et al., 2011), large parts of the contemporary deep-sea fauna of the Arctic and Nordic seas likely originate from the North Atlantic (e.g., Bluhm et al., 2011 and citations therein; Svavarsson, Stromberg & Brattegard, 1993).

In our indicator analysis, species were identified based on their predominant affiliation to certain oceanographic conditions. Identifying areas of endemicity, Arfianti & Costello (2020) defined our study area as part of a larger region that comprised North American boreal, Arctic and North Pacific areas. Our results, however, are consistent with the view that the deep-sea fauna of the Nordic seas appears to originate from shelf genera or less pronounced deep-sea taxa that were able to cross the GIFR (Dahl, 1979; Just, 1980; Svavarsson, Stromberg & Brattegard, 1993). The study by Arfianti & Costello (2020) contained data for the entire Arctic and sub-Arctic regions, encompassing both shelf and deep-sea areas, with the first reportedly representing a mixture of Atlantic, Arctic and Pacific elements (see above). Contrasting distribution patterns in hyperbenthic Eusiridae and Calliopiidae represent good examples to illustrate the barrier effect of the ridge; the family Eusiridae, which is more prevalent in deep water, has only a few species north of the GIFR, which is in contrast to the shallow water family Calliopiidae, whose species diversity is higher in the north (Weisshappel, 2000; Weisshappel, 2001). Overall, Svavarsson, Stromberg & Brattegard (1993) describe the deep-sea fauna of the Arctic and Nordic seas as very young, probably less than 100,000 yrs. old, due to the presence of the ridge and the adverse conditions prevailing in the northern regions (“topographic and environmental filtering”). Accordingly, little time remained for speciation and formation of endemic species (Svavarsson, Stromberg & Brattegard, 1993).

Our coastal amphipod assemblage, as well as the one associated with the GIFR, consisted of indicator species with broad North Atlantic distributions. Over the past millennia the biogeography of northern latitudes had been shaped by recurring glacial cycles (Darby, Polyak & Bauch, 2006). During the last glacial maximum (ending about 6,000 yrs ago; Darby, Polyak & Bauch, 2006) Arctic shelves were largely covered by grounded ice sheets forcing the fauna towards more southerly (North Atlantic) ice-free areas or deeper waters (Dunton, 1992; Darby, Polyak & Bauch, 2006). The latter may have become the ancestors of today’s Nordic deep-sea fauna (Nesis, 1984). While evidence exists that at least parts of the shelf had remained ice-free and thus served as glacial refugia, notably here Iceland and the Faroe Islands (Maggs et al., 2008; Hardy et al., 2011), most species must have recolonized the previously ice-covered areas rather swiftly. Given the close overlap of coastal and GIFR fauna in our study, the ridge could have provided a potential shallow-water link for brooding taxa that has promoted the recolonization from suitable ice-free habitats.

Diversity trends

The comparison of the diversity between the environmental clusters showed that the diversity of the shallow clusters (coastal and GIFR) was higher than that of the deep clusters north and south of the ridge. While species richness had the highest number of effective species (Fig. 4A), its sole use is usually not encouraged as it is heavily affected by sample size and shows high sensitivity in recording rare species (Jost, 2006). There were some profound differences between Hill numbers—species richness, Shannon, and Simpson diversity—likely because each of these indices scales rarity differently (Chao, Chiu & Jost, 2014; Roswell, Dushoff & Winfree, 2021; Figs. 4B, 4C). The fact that none of the richness-based rarefaction curves has stabilized yet, could therefore be an artifact; many species have only been found once, either because they could not be identified to species level or because only a small number of individuals were sampled during the historical missions. The Simpson index, on the other hand, is considered as being most robust when sampling effort differs strongly between samples, since it largely reflects patterns in the most common species (Jost, 2006). Shannon diversity can be seen as a intermediate measure in terms of its responses to sample size and rarity (Roswell, Dushoff & Winfree, 2021). Overall, though, all estimates applied have their merits and pitfalls, and typically using all three indices provides the best representation of the diversity in a given area (Roswell, Dushoff & Winfree, 2021). Nevertheless, a consistent pattern of a higher diversity in the shallows—relative to the deep clusters—was evident in all three indices. In the same way, analysis of the entire data set showed an unimodal pattern, with richness peaking at around 500 m, and then a sharp decline in richness with increasing depth (Fig. 5). Compared to other studies that often show a peak between 2,000–3,000 m (cf. Rex & Etter, 2010 and citations therein), maximum richness in amphipods of the Nordic Seas seems to be much shallower and to resemble patterns in isopods from the same area (Brix et al., 2018, but see Svavarsson, 1997). However, it should be noted here that differences in sampling intensity between grid cells and depth were a confounding factor in our study and the results therefore will have to be reassessed with additional future sampling.

Combined historical and ecological explanations have been utilized to interpret the overall low diversity of the Nordic basins compared to the other deep-sea regions (Svavarsson, Stromberg & Brattegard, 1993; Bluhm et al., 2011). In general, it is believed that variation in energy supply (temperature and productivity) affect deep-sea diversity (e.g., Woolley et al., 2016; Yasuhara & Danovaro, 2016; Jöst et al., 2019). However, cold temperatures per se do not seem to have a negative impact on diversity, since benthic communities at sub-zero temperatures in the Southern Ocean abyss appear to be extraordinarily rich (Brandt et al., 2007), but when coupled with the very low productivity and geographical isolation of the Nordic basins, the diversity of invertebrates is relatively low (Svavarsson, Stromberg & Brattegard, 1993; Egilsdottir, McGinty & Gudmundsson, 2019; Jöst et al., 2019). In addition, antagonistic effects of high hydrostatic pressure and low temperatures that prevail in the deep Nordic Sea basins could explain the low diversity there (Brown & Thatje, 2011, 2014).

Notably, the diversity of the “Deep South” cluster in our study was as low as that of the Deep North, which contrasts with the perception of an impoverished Nordic deep-sea fauna (Bouchet & Warén, 1979; Dahl, 1979; Rex et al., 1993; Svavarsson, 1997; Weisshappel & Svavarsson, 1998; Jöst et al., 2019). Although amphipods are typically less well presented in the deep sea (e.g., when compared to isopods; Lörz, Kaiser & Bowden, 2013), their ‘deficiency’ in Nordic waters was established earlier. For example, Dahl (1979) found that gammaridean species in the Norwegian Sea is a mere 20% of that in the North Atlantic. Yet, it is not clear whether this is a valid conclusion, since pure richness comparisons are very susceptible to differences in sample sizes and sample effort (see discussion above). In addition, different taxa north and south of the ridge can have different diversity patterns resulting e.g., from their different evolutionary histories, lifestyles (brooding vs. broadcaster) or physiological scope. This becomes very evident in isopods, a sister group of the amphipods, where the diversity of the deep North Atlantic exceeds that of the Nordic seas (Svavarsson, 1997).

Although not strictly comparable, but in line with our results, Egilsdottir, McGinty & Gudmundsson (2019), found local deep-sea diversity of bivalve and gastropod molluscs north and south of the GIFR to be equally low. They attributed this to specific oceanographic conditions prevailing at the deep southern stations. In addition, changes in environmental conditions in the course of past glacial maxima in the northern North Atlantic and in the North Sea were associated with cyclical changes of low (glacial) and relatively increased (interglacial) diversity (Cronin & Raymo, 1997; Yasuhara et al., 2014). The related environmental consequences of these climatic changes, in particular variation in bottom-water temperature, seasonality and meltwater runoff, evidently had a strong impact on deep-sea diversity, with recent deep-sea fauna still in the process of recovering from these events (Rex et al., 1993; Cronin & Raymo, 1997; Wilson, 1998; Yasuhara et al., 2008; Yasuhara et al., 2014; but see Jöst et al., 2019 and citations therein).

Compared to the deep-sea cluster, the diversity of the shallower coastal and GIFR clusters was considerably higher (Fig. 4A). This is in stark contrast to an allegedly poor amphipod fauna, for example when compared to the South polar region (Arfianti & Costello, 2020). Although a direct comparison with other regions at complementary depth is still pending, it is already clear that the shelf and upper slope amphipod fauna on the border between the North Atlantic and North Sea, consisting of more than 300 effective species, is not depleted (Fig. 4A). In comparison, De Broyer & Jazdzewska (2014) counted ~560 amphipod species for the entire Antarctic region (south of the Polar front), which is considered to have a significantly higher amphipod diversity relative to high northern latitudes (Arfianti & Costello, 2020). In addition, through the application of molecular techniques, but also additional sampling, especially of the deeper and less frequently explored areas, more species are likely to be discovered for the northern region (Bluhm et al., 2011; Jażdżewska et al., 2018; Lörz, Jażdżewska & Brandt, 2018; Schwentner & Lörz, 2020). We admit the comparison is slightly misleading, as cryptic species are discovered across all environments at similar rates (Pfenninger & Schwenk, 2007), plus different geological histories, oceanographic settings, and the size of the Arctic vs. Antarctica, among other things represent additional confounders. We thus believe that the diversity of the northern regions should not be underestimated and presumably occupies globally at least a middle ranking.