Introduction

Primate encounter rates often vary profoundly between seasons (Branch, 1983; van Schaik & Brockman, 2005). This is because seasonal differences in primate activity levels (Erkert & Kappeler, 2004), feeding ecology (Hanya & Chapman, 2013; Poulsen et al., 2001), ranging (Reyna-Hurtado et al., 2018), and social behavior (Heldstab et al., 2021) can affect our ability to observe and detect wild primates at different times of the year. Whilst primate encounter rate data can be collected for the whole year to calculate the most reliable population density estimates, this often is not possible due to the remote location of many primate populations and the limited financial resources available to conduct such studies (Cowlishaw & Dunbar, 2021; Turner & Corlett, 1996). Because accurate and reliable data on primate population densities are required for their conservation, primate studies need to be carefully planned to avoid periods when primate observation and detection rates are likely to be lower due to certain behaviors.

Torpor, the act of entering into a state of physical inactivity, is one such adaptation that impacts seasonal detection rates in some primate species (Miller et al., 2018). Seasonal torpor is a rare and extreme adaptation among primates (Lymann et al., 1982; Schülke & Ostner, 2007), and this is primarily because most species of primates live in the tropics where seasonal variations in temperature are less profound (Lovegrove, 2003). With the exception of the small-bodied tamarins (Moynihan, 1970), all documented cases of primate torpor occur in nocturnal strepsirrhine species that have a body mass of less than 600 g (Dausmann & Warnecke, 2016). Many primate taxa that undergo torpor employ only daily torpor in times of low food and water availability (Dausmann & Warnecke, 2016; Perret & Aujard, 2001). This is a well-known trait in mouse lemurs (genus Microcebus) (Kobbe et al., 2011; Lovegrove et al., 2014; Ortmann et al., 1997; Schmid, 2000), but it also occurs in the Javan slow loris (Nycticebus javanicus), the African lesser bushbaby (Galago moholi), the hairy-eared dwarf lemur (Allocebus trichotis), and possibly Coquerel’s giant mouse lemur (Mirza coquereli) (Dausmann, 2014; Dausmann & Warnecke, 2016; Nowack et al., 2010). However, some strepsirrhines, such as the dwarf lemurs of the Cheirogaleus genus (Dausmann & Warnecke, 2016), the pygmy slow loris (N. pygmaeus) (Ruf et al., 2015), Goodman’s mouse lemur (M. lehilahytsara) (Andriambeloson et al., 2020), the reddish-gray mouse lemur (M. griseorufus) (Kobbe et al., 2011), and the brown mouse lemur (M. rufus) (Atsalis, 1999) enter torpor for prolonged periods of several days or months (hibernation) if they are able to store enough energy reserves. Studying such species is therefore difficult, because they are inactive and undetectable for long periods of the year.

The dwarf lemurs (genus Cheirogaleus) are a group of nine, small, omnivorous, primate species from Madagascar (Groves, 2016; McLain et al., 2017) and are the sole group of primates that enter obligate hibernation (Dausmann, 2008; Dausmann & Blanco, 2016; Wright & Martin, 1995). To do this, Cheirogaleus spp. spend much of their active periods feeding and building up reserves of fat in order to survive periods of food and water shortages during the dry season (Dausmann & Blanco, 2016). Cheirogaleus are distributed throughout Madagascar and are present in a range of habitat types, including dry forest (e.g., the fat-tailed dwarf lemur (C. medius): Dausmann et al., 2005), humid forest (e.g., the greater dwarf lemur (C. major), the Montagne d'Ambre dwarf lemur (C. andysabini): Lahann, 2007; Dausmann, 2013; Lei et al., 2015), central highland and montane forest (e.g., Crossleyi’s dwarf lemur (C. crossleyi), Sibreei’s dwarf lemur (C. sibreei): Blanco et al., 2008; Blanco & Godfrey, 2014), and transitional forests that have characteristics of both humid and dry forests (e.g., C. crossleyi, Sheth’s dwarf lemur (C. shethi): Hending et al., 2017a). They also have been observed in agroecosystems (Hending et al., 2018; Webber et al., 2020). Habitat structure and resource availability vary considerably among these habitat types, and differences in encounter rates, behavior, ecology, and hibernation length have been described for different Cheirogaleus species from different forest types and between different bioclimatic zones (Dausmann & Blanco, 2016). Information on Cheirogaleus is generally lacking, and our knowledge of species-specific encounter rates and hibernation duration is generally restricted to just a few studies from only one or two study locations (e.g., C. major in eastern Madagascar: Lahann, 2007; Blanco et al., 2018). Because some species of Cheirogaleus have large geographic distributions in multiple habitat types and climatic zones (Frasier et al., 2016), we therefore have no knowledge of when many Cheirogaleus populations enter and emerge from hibernation. As Cheirogaleus encounter rates are certain to vary seasonally due to hibernation periods, this limits our ability to study Cheirogaleus behavioural ecology and constrains the period in which we can reliably carry out population assessments.

With a seasonally fluctuating mean body size of 150–270 g, Cheirogaleus medius is one of the smallest Cheirogaleus (Fietz & Ganzhorn, 1999), and, like other Cheirogaleus, it has the ability to store fat reserves as adipose tissue in its tail. This adaptation allows C. medius to hibernate for periods of up to 8 months, the longest for any primate (Dausmann & Blanco, 2016). C. medius has a very large, discontinuous geographic distribution, and it inhabits the dry forests of the west of Madagascar and the transitional forest of the Sambirano Domain in the northwest. The hibernation duration of this species has thus far been studied in only dry and littoral forest habitat (Dausmann & Blanco, 2016; Dausmann et al., 2005; Lahann & Dausmann, 2011), and differences in hibernation length between these two habitat types have been documented, suggesting that this species has flexible life-history traits in relation to its environment (Lahann & Dausmann, 2011).

In this study, we investigated how encounter rates of C. medius vary among the wet and dry season in transitional forests of the Sahamalaza-Iles Radama National Park, northwest Madagascar, to gain insight into the hibernation schedule of this species. We also aimed to improve survey methods and study-planning for C. medius to obtain reliable density estimates. As the hibernation schedule of this species is likely to affect detection rates, and therefore population density estimates, we predicted that detection rates would vary significantly between the wet and dry seasons. As hibernation duration differences have already been observed between dry and littoral forest due to climatic differences (Dausmann & Blanco, 2016; Lahann & Dausmann, 2011), we also predicted that C. medius hibernation would be different in transitional forest in comparison to the other two habitat types.

Methods

Ethical Note

All research complied with UK Home Office policies when working with animals and all research adhered to the legal requirements of Madagascar. Research in the Sahamalaza-Iles Radama National Park was permitted by Madagascar National Parks (Permit numbers 245/19 and 124/22 – MEEF/SG/DGGE/DAPRNE/SCBE.Re). We consulted the Code of Best Practices for Field Primatology when planning all methods undertaken in this study.

Data Availability

The datasets used during this study are available from the corresponding author on reasonable request.

Conflict of Interest

The authors declare that they have no conflict of interest.

Study Site

Sahamalaza-Iles Radama National Park (hereafter referred to as SIRNP) is located in North West Madagascar on the Sahamalaza peninsula, between 13°52’S – 14°37’S and 45°38’E – 48°04’E (Rode et al., 2013). SIRNP is a UNESCO Biosphere Reserve and gained full National Park status in 2007 (Schwitzer et al., 2007), yet despite this protection, it has undergone heavy deforestation in recent years (Seiler et al., 2014). SIRNP now has only two remaining forests: the 1,169 ha Anabohazo continuous forest block and the highly fragmented 1,020 ha Ankarafa forest (Volampeno, 2009). Whilst a few smaller, highly degraded forest fragments do remain, scattered throughout the National Park, much of the area is now characterised by disturbed, anthropogenic savannah, grassland, and some remnants of scrub and gallery forest (Volampeno et al., 2011). We conducted our study in both the forests of Ankarafa (14°22’S, 47°45E) and Anabohazo (14°19’S, 47°54’E) (Fig. 1).

Fig. 1
figure 1

The Sahamalaza-Iles Radama National Park, Madagascar, and the location of Ankarafa forest (A) and Anabohazo forest (B). Figure created in ArcMap, with a scale of 1:7,000,000 for Madagascar and a scale of 1:350,000 for the zoomed panel. The two red lines surrounding protected areas represent the protected area (inner) and a 3-km buffer zone (outer).

Full size image

SIRNP is located in the Sambirano Domain, an area of Madagascar that is characterized by subhumid transitional forest; the forests of SIRNP therefore encompass a mixture of deciduous and evergreen tree species, many of which are endemic to the Sambirano Domain (Du Puy & Moat, 1996; Koechlin, 1972). The climate of SIRNP is highly seasonal, with a separate hotter wet season (November–April) and a cooler dry season (May–October) (Mandl et al., 2018). Mean temperature range in SIRNP is 20.6–32.0 °C, with recorded extreme temperatures of 13.2–39.1 °C (Hending et al., 2017b; Volampeno et al., 2011) and mean annual precipitation rates of approximately 1,600 mm (Schwitzer et al., 2007).

Distance Sampling

Between June 9, 2018 and May 13, 2022, we conducted line-transect distance sampling along several transects, a method regularly used for arboreal primates (Buckland et al., 2010), to assess the encounter rates of Cheirogaleus medius in SIRNP. We established a total of 60 line transects in the forests of Ankarafa (N = 40, mean length = 1,021 m, range = 525–1,525 m) and Anabohazo (N = 20, mean length = 1,124 m, range = 500–1,500 m). Due to the large number of transects, we did not walk all transects each month. The Ankarafa transects were in 40 separate forest fragments (one transect per forest fragment). We began each transect walk at approximately 19:00, when C. medius would be actively feeding (Fietz & Ganzhorn, 1999; Müller, 1999). During each transect walk, two researchers and one guide walked along the transect at a steady and continuous pace of approximately 1 km/h. The same three participants conducted all transects to avoid observational bias (Buckland et al., 2001). We used high-lumen handheld torches (EC20, Nitecore, Guangzhou, China) and headlamps (Tikkina and Tikka + , Petzl, Crolles, France) to spot any C. medius individuals that were present along the transect based on eye-shine. For any sighted C. medius individuals, we recorded their geographic position using a handheld GPS (eTrex 30, Garmin, Olathe, KS), the date and time of the observation, and we measured the perpendicular distance from the transect line to the tree in which the lemur was positioned using a handheld laser measure (80 m, XCSource, San Francisco, CA). Despite the high density of the vegetation at the study site, we were able to distinguish C. medius from the sympatric Sahamalaza sportive lemur (Lepilemur sahamalaza) by their pelage colouration and size differences (Mittermeier et al., 2008). Furthermore, we could distinguish C. medius from the sympatric northern giant mouse lemur (Mirza zaza) and the Sambirano mouse lemur (Microcebus sambiranensis) by their size and movement; locomotion of C. medius is slower than that of M. zaza (Müller & Thalmann, 2002; Rode-Margono et al., 2016) and C. medius is larger than M. sambiranensis (Mittermeier et al., 2008). Although our study took place over almost 4 years, we did not conduct any transect surveys during December due to the inaccessibility of the site around this time.

Encounter Rates and Population Density Estimates

We calculated monthly mean encounter rates for this study, by dividing the number of observed individuals for each transect by the total transect distance (N/km). We then calculated mean C. medius encounter rate for each transect across all repeat walks, for both forests individually and for SIRNP as whole. We also calculated mean C. medius encounter rates for each month across transects. We submitted our C. medius count data, transect length values and perpendicular distance values (Supplementary File 1) to the Distance 7.5 software (Thomas et al., 2010) to quantify transect survey effort (estimated strip width: ESW—the area under the distance function from its left-truncation limit to its right-truncation limit) and to calculate C. medius population size and density, with lower and upper 95% confidence limits, for SIRNP. We used the Buckland method to calculate our population densities (Buckland et al., 2001), as it has been used successfully in other lemur population studies (Axel & Maurer, 2011; Quémére et al., 2010) and because it accounts for the decreasing probability of observing our study species as its perpendicular distance from the transect increases. We used instance detection functions, and we truncated our datasets with a 5% threshold to remove very distant observations. We also used Distance to estimate the population size for C. medius with lower and upper 95% confidence limits using our population density values and the area of each surveyed forest. Forest area was determined before this study with GPS waypoint data obtained during perimeter walks; forest area was later calculated using ArcGIS (Version 10.6.1, Esri, Redlands, CA).

Extraction of Temperature and Precipitation Variables

We used R Studio (R.Studio, Inc., Boston, MA) packages “raster” (Hijmans, 2019) and “rgdal” (Bivand et al., 2019) to download average monthly temperature and precipitation variable data for SIRNP for our study period (2018–2022) directly from the WorldClim global climate database (Fick & Hijmans, 2017). Specifically, we downloaded mean monthly data for average temperature, minimum temperature, maximum temperature, and precipitation. When specifying the parameters of our variable extraction, we used the GPS point corresponding to the midpoint of our study area (latitude =  − 14.3881, longitude = 47.8540), and we used the highest possible raster resolution (1-km2 tiles) to obtain the most accurate and localized data possible (Fig. 2).

Fig. 2
figure 2

Precipitation and temperature of the Sahamalaza-Iles Radama National Park, Madagascar, for February 2018–May 2022. Precipitation represents the total rainfall for each month. Figure created using data downloaded from the Worldclim database (Fick & Hijmans, 2017).

Full size image

Data Analysis

We performed all statistical analyses in R Studio with an α-level of 0.05. Our monthly encounter rate data for each transect was of normal distribution (Shapiro–Wilk normality test: w = 0.884, P = 0.116), so we compared how C. medius encounter rates vary between months by using a chi-square test. To investigate the relationship between C. medius encounter rate and temperature and precipitation, we first subjected the temperature and precipitation variables (independent variables) to tests of residual normality and homoscedasticity with the “olsrr” package (Hebbali, 2018). We investigated the relationship between C. medius activity levels and temperature and precipitation with individual Spearman’s rank correlation tests. To compare C. medius population density and size between the wet season (November–April) and dry season (May–October), we created a separate dataset of mean encounter rates for only the transects that we surveyed during both seasons (we removed transects surveyed in only the wet or dry season and excluded them from this dataset). The mean encounter rates dataset did not meet the assumptions of parametric analyses (Shapiro–Wilk normality test: w = 0.776, P = 0.005), and so we compared mean encounter rates between the two sampling periods (wet season transects N = 84, dry season transects N = 160) with a paired Wilcoxon test (one mean encounter rate value per transect for both the wet and dry season).

Results

Encounter Rates

We performed a total of 264 transect surveys across our 60 established transect lines (mean = 4.40 walks per transect, range = 1–10), with a total survey distance of 275.30 km. During our study, we observed a total of 130 Cheirogaleus medius individuals, a mean of 0.49 individuals per transect walk (0.47 per km). The encounter rates of C. medius were significantly uneven among the months that were surveyed (chi-square test: χ2 = 504.91, df = 207, P < 0.001; Fig. 3). Specifically, we observed the lowest encounter rates of C. medius from May to August.

Fig. 3
figure 3

Mean monthly encounter rates, with standard error (error bars), of Cheirogaleus medius in the Sahamalaza-Iles Radama National Park, Madagascar, during the study period June 2018–May 2022, with sample sizes of C. medius individuals (N) observed for each month.

Full size image

The mean encounter rate of C. medius was significantly positively correlated with monthly mean temperature (S (test statistic) = 69.658, ρ (correlation coefficient) = 0.683, P = 0.020) and mean maximum temperature (S = 50.000, ρ = 0.773, P = 0.008). However, C. medius encounter rate did not correlate significantly with monthly mean minimum temperature (S = 135.030, ρ = 0.386, P = 0.241) and precipitation (S = 148.000, ρ = 0.327, P = 0.327; Fig. 4).

Fig. 4
figure 4

Data and Spearman’s correlation test results of mean monthly temperature and precipitation against Cheirogaleus medius encounter rate, with 95% confidence intervals, in the Sahamalaza-Iles Radama National Park, Madagascar, June 2018–May 2022.

Full size image

Population Density and Size Estimates

The population density and size estimates for C. medius were profoundly different between the data collected for the wet and dry seasons (Fig. 3; Table I). Our statistical analysis of transect mean encounter rates also show a significant difference between the two periods (paired Wilcoxon test: V = 66, P = 0.034), with encounter rate being higher during the wet than dry season.

Table I The encounter rate, estimated population densities (mean number of individuals per kilometre squared (lower and upper 95% confidence limits)), and abundance (estimated mean population sizes (lower and upper 95% confidence limits)) of Cheirogaleus medius in two surveyed forests of the Sahamalaza-Iles Radama National Park, Madagascar, and for the protected area as a whole, June 2018–May 2022. Values are presented for the dataset as a whole, and for the wet (November–April) and dry (May–October) seasons separately
Full size table

Discussion

Seasonal Differences in Encounter Rates

Although we observed Cheirogaleus medius individuals throughout the whole study period, our encounter rates were significantly higher in the wet season than in the dry season (Fig. 3). These findings are certainly due to lower activity levels of C. medius during the dry season, as many individuals would have been hibernating, and these observations are important for the planning and organisation of future studies into this species at this site. As has been suggested for the populations of C. medius in dry forest habitat (Dausmann et al., 2005), C. medius activity may be lower during the dry season due to lower water and food availability and lower temperatures; we also observed C. medius less frequently during times of the year with low precipitation (although this difference was not significant) and lower temperatures in this study, as demonstrated by our correlation analyses (Fig. 4). However, our C. medius encounter rates were high in September and October (Fig. 3) when rainfall was still very low at our study site (Fig. 2). C. medius encounter rate therefore may not be fully dependent on climatic factors. The lack of significant correlation also may be related to inaccurate information on weather from the WorldClim datasets, because often the number of weather stations from which this information is collected are very limited in tropical regions. Species-specific behavioural ecology and the habitat-related and environmental causes for the onset and offset of hibernation are likely to provide a fuller explanation of why C. medius encounter rates vary between months and between seasons. As these factors often are site-specific and species-specific (Dausmann & Blanco, 2016), further research into C. medius ecology and natural history in SIRNP is required to provide a more comprehensive answer to our research question.

Insights into Cheirogaleus medius Hibernation

Encounter rates of C. medius were very low during the period May–August (Fig. 3), and these observations may indicate the period when C. medius enters hibernation in SIRNP. This period coincides with the lowest monthly rates of precipitation and the lowest annual temperatures (Fig. 2), which is supported by our Spearman’s correlation analyses of temperature (Fig. 4). This potential hibernation period of four months is much shorter than that of C. medius in dry forest habitat (6–8 months: Dausmann & Blanco, 2016) but similar to that of the littoral forest habitat medius-like (Thomas’s dwarf lemur, C. thomasi) populations (4–5 months: Lahann & Dausmann, 2011). As decreasing temperatures, low water availability, and food scarcity influence the hibernation patterns of C. medius (Dausmann et al., 2005), these differences among habitat types are expected as the transitional forests of North West Madagascar have a higher annual rainfall than dry forest habitat (Donque, 1972). Although we observed six C. medius individuals during the period May–August, these lemurs may have been juveniles that were not able to accumulate the necessary fat reserves for longer hibernation (Blanco & Rahalinarivo, 2010). As Cheirogaleus have flexible life history traits (Lahann & Dausmann, 2011), it is likely that the climate of SIRNP has influenced the timing of when the resident C. medius populations enter hibernation.

C. medius is considered to have the highest hibernation duration of any of the Cheirogaleus species (Dausmann & Blanco, 2016). However, our encounter rate data for SIRNP suggest that some populations of C. medius may hibernate for shorter periods than the rainforest species C. crossleyi (3.5–5.5 months) and C. sibreei (4.5–7 months) (Blanco & Godfrey, 2014). These differences in hibernation duration may be due to the actual cause of the hibernation itself; hibernation expression in sympatric C. crossleyi and C. sibreei for example is interspecific due to species-specific ecology (Blanco & Rahalinarivo, 2010; Dausmann, 2013). This further supports the theory of flexible life history traits in relation to hibernation among Cheirogaleus (Dausmann & Blanco, 2016).

Population Density and Size

The population density of C. medius in SIRNP (Table 1) was similar to that reported in other studies of this species from the dry forest (20–200 individuals/km2: Hladik et al., 1980; Fietz, 1999). C. medius population density also varied between the two forest sites that we surveyed in SIRNP (Table 1). Cheirogaleus may be sensitive to habitat degradation and edge-effects (Lehman et al., 2006; but see Hending, 2021), and the population density of C. medius in SIRNP may be limited by the ongoing forest degradation that is typical of the region (Seiler et al., 2010); this also may explain the differences in density between Anabohazo (a continuous forest) and Ankarafa (a highly-degraded and fragmented forest). Furthermore, C. medius lives sympatrically with the northern giant mouse lemur (Mirza zaza), the Sambirano mouse lemur (Microcebus sambiranensis), and the Sahamalaza sportive lemur (Lepilemur sahamalaza) in SIRNP (Hending et al., 2017c, 2022a). Although niche separation exists between these species, all four of these taxa require secure sleeping sites. Well-insulated tree holes are particularly sought after by these species, because they provide cover from predators and insulation from the cold (Hending et al., 2017c; Rode et al., 2013; Schmid, 1998; Seiler et al., 2013). Tree holes are particularly important for C. medius, and the high competition for these sleeping sites also may limit C. medius numbers in the forests of SIRNP, which has been observed in other cheirogaleid lemurs (Lutermann et al., 2010).

Although all Cheirogaleus are obligate hibernators (Dausmann & Blanco, 2016), many surveys of their population structure and density have occurred partly or entirely in Madagascar’s dry season, a period during which they are likely hibernating (Herrera et al., 2011; Lehman et al., 2006; Müller, 1998; Müller et al., 2000). We found significant differences in our encounter rate of C. medius between the wet and dry seasons. From these findings, it is clear that results of surveys that included either wholly or partly the hibernation period do not represent accurate population density values and are serious underestimates of total population size. Such underestimates would be detrimental for accurate IUCN Red List assessments for Cheirogaleus, considering that their forest habitat is highly threatened by deforestation and unmitigated climate change (Hending et al., 2022b; Vieilledent et al., 2018). Our results underpin the importance of a sound understanding of the activity periods of Cheirogaleus populations at specific sites and in specific habitats so that population surveys can be carried out at the correct time to calculate accurate and reliable population assessments.

Summary and Future Directions

This study provides an overview of seasonal activity level, hibernation and population density of C. medius in a transitional forest habitat. However, we still know nothing of the behavioural ecology of this species in transitional forest and how it compares to that of the C. medius populations in dry forest. To better understand C. medius ecology and hibernation and how it varies among habitat types and bioclimatic regions, the thermoregulatory physiology and hibernaculum choice of C. medius should be investigated for the populations in SIRNP, because these determine Cheirogaleus ecology and hibernation expression (Dausmann & Blanco, 2016). Furthermore, a more detailed knowledge of C. medius feeding ecology is needed and its implications for hyperphagy (the prehibernation “fattening” period) need to be explored (Fietz & Ganzhorn, 1999). Generally, the ecology and natural history of C. medius is still little known despite the large geographic distribution of this species. Such datasets may provide answers as to how and why lemurs have evolved a survival strategy as extreme as obligatory hibernation.