Ecological niche models (ENMs) are well-suited for studies of demography, taxonomy, and species conservation, because they can be utilised to assess how well-protected species are throughout their ranges and their susceptibility to anthropogenic pressures (Elith et al., 2006; Thorn et al., 2009; Urbina-Cardona & Loyola, 2008). ENMs can provide valuable information about the potential geographic distributions of species (Peterson & Soberón, 2012) and information on niche separation among sympatric species and between closely related taxa (Zimmerman, 2006) and integrated taxonomy (Padial & De La Riva, 2010). It is therefore useful for species delineation (Raxworthy et al., 2007; Rissler & Apodaca, 2007). Furthermore, ENMs can reveal information about the potential geographic ranges of understudied species (Guisan & Zimmermann, 2000) and thus facilitate their conservation and future-study (Ferrer-Sánchez & Rodríguez-Estrella, 2016). ENMs are particularly relevant tools to estimate the distributions of elusive animals that are hard to observe and study in the wild (Coudrat & Nekaris, 2013). Cryptic species are such an example of this, because their similar morphologies make them difficult to visually identify (Sattler et al., 2007). Unfortunately, the geographic distributions of many of these newly described species are unknown (Rissler & Apodaca, 2007). It also is difficult to determine the presence and absence of cryptic species at specific localities (Black, 2020), because they often are small, elusive, and inconspicuous (Gibson et al., 2007; Sutherland, 2000). This dearth of knowledge can be detrimental to their conservation, as many cryptic species are already threatened with extinction (Black, 2020; Morais et al., 2013).

The dwarf lemurs (genus Cheirogaleus) are a group of small-bodied, nocturnal, and cryptic primates endemic to Madagascar (Mittermeier et al., 2008). Like many other nocturnal lemurs (Groves, 2001; Yoder et al., 2000), the dwarf lemur genus has undergone taxonomic expansion in recent years, despite opposition from Groeneveld et al., (2009, 2011) who provided genetic and morphological evidence against this taxonomic splitting. Nine species of dwarf lemur are now formally described and are recognized by the IUCN (Frasier et al., 2016; IUCN, 2020a; Lei et al., 2015; McLain et al., 2017). Although dwarf lemurs are arboreal and require forest habitat for their survival (Lehman et al., 2016), some species, such as C. medius and C. major, are capable of surviving in degraded and anthropogenically disturbed areas (Hending, 2021a). Unlike nocturnal lemurs of the Microcebus, Lepilemur, and Phaner genera, whose species-specific distributions often are restricted by biogeographical and hydrological features (Craul et al., 2007; Hending et al., 2020; Olivieri et al., 2007; Wilmé et al., 2006), many dwarf lemur species are more geographically widespread (IUCN, 2020b; Lei et al., 2014), making their ranges difficult to predict. Studies of Cheirogaleus biogeography are made more challenging by the hibernation that they undergo for up to 7 months annually (Dausmann & Blanco, 2016), making them undetectable for large periods of the year. Despite this, many new dwarf lemur populations have been discovered recently (Gardner & Jasper, 2015; Hending et al., 2017), but morphological similarities among species make these new populations difficult to identify (Groeneveld et al., 2009).

Sympatry among multiple dwarf lemur species at numerous localities, such as C. crossleyi with C. sibreei and C. major with C. thomasi, further confuses our understanding of their geographic distributions (Blanco et al., 2009; Groeneveld et al., 2010) and raises questions of their habitat requirements and niche overlap (Hapke et al., 2005; Herrera et al., 2016; Lahann, 2007). Whereas the full ranges of the dwarf lemurs remain little known, it is already apparent that their forest habitat is being cleared at an alarmingly high rate, and much of what remains is now fragmented (Vieilledent et al., 2018). Eight of the nine dwarf lemur species are now listed as threatened on the IUCN Red List, whereas C. grovesi is listed as Data Deficient (IUCN, 2020a). An informed understanding of Cheirogaleus ecological niches, the area of suitable habitat that is available to them, and their potential geographic distributions is needed to conserve and manage their remaining populations and alleviate the threats that they face (Schwitzer et al., 2013).

We investigated the biogeography of Madagascar’s dwarf lemurs with ENMs. The specific objectives of this study were:

  1. 1.

    To compare ecological niche-overlap among the dwarf lemurs. As dwarf lemur sympatry sometimes occurs (Blanco et al., 2009; Herrera et al., 2016; Lahann, 2007), we hypothesized that little niche overlap would occur between species to maintain reproductive isolation.

  2. 2.

    To assess the total area of suitable forest habitat (realized niche—the area in which each species can survive) of each species. Some dwarf lemurs are regarded as adaptable, generalist primates (Schäffler & Kappeler, 2014), and we therefore hypothesized that their ENMs would encompass a large area of suitable habitat. We also predicted that the realized niche area (suitable habitat) would be lower in area than the total occupiable area (fundamental niche) identified in the ENMs due to widescale deforestation and forest fragmentation of Madagascar (Harper et al., 2007; Vieilledent et al., 2018).

  3. 3.

    To assess how much of each species’ realized niche is currently protected, and to determine how this habitat is affected by humans. We hypothesized that a large percentage of the realized niches would be located within protected areas due to the substantial efforts of conservationists to protect Madagascar’s remaining biota (Goodman et al., 2018). We also predicted that much of the realized niches would be at significant risk of anthropogenic disturbance, and that habitat suitability would correlate negatively with disturbance intensity.



We certify that all work undertaken in this manuscript complied with the protocols of our institutions and the UK Home Office, the legal requirements of the UK, and the UK Animals (Scientific Procedures) Act 1986 Amendment Regulations (SI 2012/3039).

Data Availability

The datasets generated during and/or analysed during the current study are available from the corresponding author on request.

Competing Interests

All authors state that they have no conflict of interest.

Occurrence Data

We collected known Cheirogaleus species’ occurrence data from both published and unpublished sources. To find occurrence data within the published literature, we searched the full volume-catalogue of the primatology journals, American Journal of Primatology, International Journal of Primatology, Primates, Folia Primatologica, Primate Conservation, and Lemur News, as well as the Madagascar-specific journals, Malagasy Nature and Madagascar Conservation and Development. This method ensured that we did not miss any studies from local or regional journals (e.g., Malagasy Nature). In addition to these primate-specific journals, we used Google Scholar to search for published articles in nonprimate-specific journals, dissertations, and edited book volumes. We used the keywords “population,” “distribution,” “presence,” “dwarf lemur, and “Cheirogaleus” in our literature search, in addition to the common and scientific name keywords for all nine Cheirogaleus species. Additionally, we supplemented these published datasets with occurrence records from unpublished sources (e.g., unpublished reports, conference posters, personal observations) listed in the most recent IUCN Red List assessments (Blanco et al., 2020a, b, c, d, e, f; Ganzhorn et al., 2020; Sgarlata et al., 2020a, b).

We included all occurrence points as separate data points in our database. For each point, we listed the geographic coordinate of the location (to as many decimal points listed in the source), the corresponding Cheirogaleus species, and their conservation status. Many records in our database were from studies conducted before recent species descriptions. We thus updated the species names to reflect the latest taxonomy using the geographic location of the respective point and species-specific range information available in the most recent Red List assessments (IUCN, 2020a, b) (Fig. 1); this range information is informed by genetic studies that have confirmed the identity of Cheirogaleus populations (Frasier et al., 2016; Lei et al., 2015; McLain et al., 2017). Although Groeneveld et al., (2009, 2011) argue for taxonomic deflation of the dwarf lemurs, we use the Cheirogaleus taxonomy as per the IUCN Red List due to the conservation biogeography theme of this paper (IUCN, 2020a). However, the IUCN assessments do not include the full range of some Cheirogaleus species (e.g., C. medius), so we used the wider published literature to identify species occurrences that fell outside of the IUCN polygons (Fig. 1). We excluded any points where we could not reliably confirm the identity of the species using this method (i.e., Cheirogaleus sp.), and we checked the validity of the point coordinates in the literary sources using ArcGIS. Our database included occurrence points from studies that occurred 1964–2020; due to ongoing deforestation in Madagascar (Vieilledent et al., 2018), we checked to ensure that all occurrence points were still forested using current high-resolution (< 1 m/pixel), cloud-free, satellite imagery in Google Earth Pro (version 7.3.3, Google LLC, Mountain View, CA).

Fig. 1
figure 1

Geographic ranges of the dwarf lemurs (genus Cheirogaleus), as described by the IUCN Red List (IUCN, 2020b), and our occurrence point dataset (triangles). The IUCN ranges do not include all localities at which dwarf lemurs are known to occur, as indicated by the occurrence points that we extracted from literary sources.

Environmental Variables

We considered 21 environmental variables for the construction of our ENMs (Table I). The specific variables were chosen based on their relevance to the ecology and natural history of the study-species, and because they have been frequently used to model distributions of primates in previous studies (Chetan et al., 2014; Thorn et al., 2009). Of our 21 chosen environmental variables, 19 were bioclimatic variables related to precipitation and temperature (Hijmans et al., 2005; Booth et al., 2014), and two were vegetation-cover related variables: Normalised Difference Vegetation Index (NDVI, a proxy of plant productivity) and Leaf Area Index (LAI, a proxy of tree cover density). We downloaded global geoTiff raster layers of all BIOCLIM variables from the WorldClim database (Hijmans et al., 2005, 1-km2 resolution). For the NDVI and LAI variables, we downloaded monthly geoTiff layers (250- × 250-m resolution) from January 2000 until January 2020, and then stacked these layers in R Studio (R Core Team, 2017) using the packages “raster” (Hijmans, 2017), “sp” (Bivand et al., 2013), and “rgdal” (Bivand et al., 2019) to create mean NDVI and LAI layers. Next, we resampled all layers to a resolution of 1 km2. In order to draw pseudo-absences from a species-appropriate area when constructing our ENMs, we then clipped the Madagascar-wide raster layers for each species to an area to which each species can realistically disperse (Barve et al., 2011) (Appendix 1). These areas were chosen based on topographic features (including the central highlands that separate Madagascar’s western regions from the eastern regions), major rivers (Goodman et al., 2018), and forest cover (Vieilledent et al., 2018). We chose the extent of these species-specific input layers based on the occurrence points available for each species, the extent of occurrence of each species, and known geographic dispersal barriers (Blanco et al., 2020a, b, c, d, e, f; Ganzhorn et al., 2020; Sgarlata et al., 2020a, b).

Table I Environmental variables that were considered for ecological niche modelling of the dwarf lemurs (genus Cheirogaleus)

It is recommended to remove correlated variables before running ENMs (Merow et al., 2013). To do this, we generated 2,000 random geographic points for Madagascar, and we extracted the corresponding values for each environmental layer. We then analysed the pairwise correlations between each variable for these 2,000 points by using Pearson correlation tests (Appendix 2, as in Hending, 2021b). When pairs of variables had correlation coefficients ≥ 0.70, we retained one variable from the pairwise analysis for creating our ENMs (Cobos et al., 2019). This approach reduced the total number of environmental variables from 21 to nine (Table I). The retained variables are likely drivers of dwarf lemur ecology and physiology, and proxies for the different habitats of Madagascar (Kamilar & Muldoon, 2010; Kamilar et al., 2016). The retained climatic variables represent temperature seasonality (Bio2 and Bio3), water availability (Bio12), and forest presence and quality (NDVI, and LAI), along with proxies of potential instances of extreme heat (Bio5), frost (Bio6), and drought (Bio17) (Blair et al., 2013), which are relevant for lemur infant mortality (Wright, 1999) and seasonal hibernation in dwarf lemurs (Dausmann & Blanco, 2016).

Ecological Niche Modelling and Model Validation

We used the maximum entropy algorithm in Maxent version 3.4.1 (Phillips et al., 2006) to construct ENMs of potential geographic distribution and habitat suitability for the dwarf lemurs. We opted to use the Maxent software because it is now a widely used method to identify species’ ecological niches (Elith et al., 2006; Merow et al., 2013), and it only requires presence points (it does not need absence points to run a model). This is advantageous for this study, as dwarf lemurs are small, nocturnal, cryptic, and they undergo long periods of seasonal hibernation (Dausmann & Blanco, 2016), making it difficult to confirm their absence from a location, similarly to mouse lemurs (Kamilar et al., 2016).

Because we could only find a high number (N > 30) of occurrence points for three of the Cheirogaleus species, we did not run ENMs for the other six species. Whilst models for these species may be useful for exploratory analysis (Wisz et al., 2008), they may not be reliable enough to draw any robust conclusions from (Riley et al., 2019). Before creating the ENMs, we used the “ENMeval” package (Muscarella et al., 2014) in R Studio to create species-specific bias files for inclusion in the Maxent algorithm. These bias files enabled us to control for spatial sampling bias in the occurrence points, by projecting the raster stack of environmental variables and occurrence points with two-dimensional Kernel density estimation (Kramer-Schadt et al., 2013); this was not possible to do via occurrence point thinning due to the small number of occurrence points for the thinly distributed species (Boria et al., 2014). We also used the “ENMeval” package to evaluate the optimal Maxent model features and regularization multiplier parameters (model selection) to use for each species (Table II), using 10,000, generated, geographic, background points and a tenfold cross-validation technique. We selected the model features and parameters to use based on the Akaike Information Criterion (AIC).

Table II Total number of occurrence points, feature parameters (L = Linear, Q = Quadratic, P = Product, T = Threshold, H = Hinge), and regularization multipliers used, the MTP threshold, and AUC and binomial test of omission results for fourfold ENM models of the dwarf lemur species (genus Cheirogaleus) of Madagascar

For all Maxent models, we set the maximum number of iterations to 500, the convergence threshold to 0.001, and the number of background points to 10,000 (as in Nazeri et al., 2012), and we fine-tuned the species-specific models with different features and multiplier parameters (Table II). We used a fourfold cross-validation procedure to run our models (as in Blair et al., 2013; Kamilar et al., 2016; Pearson et al., 2007, 2011). We used the area under response curve (AUC) method to evaluate the performance of our ENMs (Merow et al., 2013). In summary, higher AUC values equate to a better-performing model, where an AUC value equal to 1 indicates perfect model performance and an AUC value < 0.5 indicates that the model does not have the ability to perform reliably (Phillips, 2006). However, we also used binomial tests of omission to validate the performance of the ENMs, because AUC does not penalize for overfitting (Pearson et al., 2007).

Niche Overlap

We calculated the pair-wise niche overlap between dwarf lemur species in environmental space based on our ENMs by using Schoener’s D (Warren et al., 2008) and Hellinger’s I (Schoener, 1968) indices in the “ENMtools” package (Warren & Dinnage, 2021). This method calculates the difference in standardized suitability score for each cell used in the ENMs between species and assigns a score between 0 (no niche overlap) and 1 (complete niche overlap/identical niches); pair-wise scores > 0.8 often indicate significant niche overlap between species (Warren et al., 2008). We then conducted background tests of whether the pair-species niche overlaps were equivalent to each other. Background tests are suitable to compare niche overlap among allopatric and sister species (Warren et al., 2010), and these tests pool all occurrence data for each species-pair and then randomly assigns localities to two new samples of equal size to the two original species samples. We specified for this procedure to produce 99 additional replicates of random species-pairs with values of niche overlap. We then compared our original species-pair niche overlap values to the null distribution of the pseudo-replicate niche overlap values. If our original pair-species overlap values were within the bottom 5% of the null distribution, then we determined that the ENMs of the two species were not equivalent and significantly more diverged than expected considering the habitat available to each species. This is comparable to a one-sided test with an alpha level of 0.05.

Suitable Habitat

Maxent ENMs contain a habitat suitability output ranging from 0 (not suitable habitat) to 1 (completely suitable habitat). To interpret and visualize the remaining suitable niche areas of each species, we set a binary threshold for each model to distinguish presence from absence (Liu et al., 2005; Pearson et al., 2007). Assigning arbitrary thresholds for each species would lack any ecological basis (Liu et al., 2005), so we calculated minimum training presence thresholds (MTP) based on our ENMs, The MTP threshold integrates the prevalence of model-building data; it is conservative and robust regarding a species’ likely distribution (i.e., it avoids overfitting), and it has the lowest false-positive and false-negative error rates of all MaxEnt-generated thresholds (Liu et al., 2005). We used these binary ENMs to calculate the total area suitable for each species in ArcMap.

The habitat suitability of the ENMs were predicted by using a range of climatic and vegetation cover-related environmental variables. ENMs more-heavily influenced by nonvegetation cover-related variables (e.g., precipitation, temperature) can often result in the inclusion of areas outside of their natural habitat types (Hartel et al., 2010). Because all lemurs depend on forest habitat for their survival (Schwitzer et al., 2013), we downloaded the most recent forest cover raster data for Madagascar (resolution 30 m2, Vieilledent et al., 2018; Fig. 2A), and we clipped the binary ENMs to exclude areas outside of forest cover. We then measured the total area of the clipped ENMs to calculate the realized niche area for each Cheirogaleus species.

Fig. 2
figure 2

Remaining forest cover (A); distribution of protected areas (B); Human Footprint (HFP) map (C)) of anthropogenic activity throughout Madagascar. Cooler, blue colours indicate higher levels of anthropogenic activity in C. Figures were created using raster layers of forest cover (Vieilledent et al., 2018), protected areas (UNEP-WCMC, 2020), and HFP (Venter et al., 2016), with a scale of 1:6,500,000 for Madagascar.

Conservation and Anthropogenic Risk Assessment

We downloaded a raster layer of Madagascar’s protected area network (UNEP-WCMC, 2020) to assess whether the realized niche areas of each dwarf lemur species are protected. This layer is composed of all protected areas in Madagascar, including 29 National Parks, 24 Special and Strict Nature Reserves, and 26 Classified Protected Areas (Fig. 2B). We overlaid this protected area layer with the realized niche of each species in ArcGIS, and we calculated the total area that was within Madagascar’s protected area network.

We also assessed the risk of anthropogenic threats on each realized niche area by using the Human Footprint (HFP) raster (Venter et al., 2016), a commonly used proxy of anthropogenic disturbance (Campera et al., 2020; Di Marco et al., 2018). The HFP layer accounts for areas of anthropogenic activity, such as roads, settlements, agricultural areas, and buildings, and maps them based on cumulative intensity on a scale of 0 (no anthropogenic activity) to 50 (intense anthropogenic activity) at a resolution of 1 km2 (Fig. 2C). In ArcGIS, we filtered out all features within the HFP layer with a < 10 score. We then converted the continuous raster to a binary layer containing all features with a score of 10–50; this new layer still contained all roads, settlements, agricultural areas, and pastures. To assess the risk of anthropogenic threats to Cheirogaleus within their realized niches, we applied 5-km and 10-km buffers to all HFP features. We then calculated the area of each species’ realized niche that fell within these buffers to determine the areas at high risk (< 5 km), medium risk (5–10 km), and low risk (> 10 km) from anthropogenic disturbance features. Cheirogaleus spp. are hunted throughout Madagascar (Herrera et al., 2018), and therefore we selected these distances to classify anthropogenic risk based on the distance that people travel from settlements and roads to hunt and gather wood and other resources from the forest (Peres & Lake, 2002; Thorn et al., 2009).

Finally, we extracted the HFP and habitat suitability scores from 100 random 1-km2 pixels of the unfiltered HFP and ENM raster layers clipped to the realized niche area of each species. We chose 100 random pixels rather than using every pixel, because the sample size would have varied significantly among species. We then performed linear regression analyses on the HFP and habitat suitability scores to investigate whether anthropogenic disturbance affects dwarf lemur habitat suitability. Shapiro–Wilk tests confirmed that the HFP and habitat score datasets met the assumptions of normal distribution, homoscedasticity, and linearity for linear regression. We then bootstrapped each analysis to 10,000 replicates with bias-corrected and accelerated confidence intervals (BCAs) in the R package “boot” (Canty & Ripley, 2020). An alpha level of 0.05 was used for all analyses described in this article.


ENM Performance and Niche Overlap

Our Maxent ENMs (Appendix 3) performed well for the three Cheirogaleus species, with mean AUCs for each species ranging from 0.801 to 0.917 (Table II). All three ENMs also had statistically significant results for all model fold omission tests (Table II). The ENM percentage contributions of the nine environmental variables varied greatly between species (Table III). The species-pair Schoener’s D overlap indices ranged from 0.127 (between C. crossleyi and C. medius) to 0.653 (between C. crossleyi and C. major), and Hellinger’s I overlap indices ranged from 0.323 (between C. crossleyi and C. medius) to 0.798 (between C. crossleyi and C. major). These results indicate significant environmental niche separation and among pair-species (all P < 0.01; Table IV).

Table III Percentage contribution of each environmental layer to the overall Maxent ecological niche models of three dwarf lemur species of Madagascar
Table IV Schoener’s D (top right) and Hellinger’s I (bottom left) environmental niche overlap estimates between species pairs

Cheirogaleus Realized Niches

With the MTP threshold applied, the fundamental niche area of the three analysed Cheirogaleus species ranged from 92,127 km2 for C. crossleyi to 336,494 km2 for C. medius (Table V). The realized niches (forest area) of the three species were lower in comparison, ranging from 28,889 km2 for C. crossleyi to 41,934 km2 for C. medius (Fig. 3; Table V).

Table V Suitable habitat of three dwarf lemur species of Madagascar, as predicted by Maxent ecological niche models with minimum training presence thresholds applied (MTP), the area of this range that is forested (realized niche), the percentage of realized niche located within Madagascar’s protected area network, and the area of the realized niches within high-, medium-, and low-risk areas (respectively located within < 5, 5–10, and > 10 km of anthropogenic features, including roads, settlements, and agricultural areas)
Fig. 3
figure 3

Realized niches of three dwarf lemur species of Madagascar, based on species ecological niche models (Appendix 3) and Madagascar’s remaining forest habitat (Vieilledent et al., 2018). C. crossleyi (A); C. major (B); C. medius (C). Scale = 1:7,000,000 for Madagascar.

Conservation and Anthropogenic Threats

Of the realized niche areas of the analysed dwarf lemurs, the percentage that was within Madagascar’s protected area network varied between species, ranging from 47.8% (20,065 km2) for C. medius to 66.0% (25,266 km2) for C. major (Table V; Fig. 4). The percentage of the realized niche area that was less than 5 km from an HFP feature, and therefore at high risk of anthropogenic disturbance, also varied between species, ranging from 19.9% (5,744 km2) for C. crossleyi to 40.5% (16,999 km2) for C. medius (Table V). This also was the case for the percentage of the realized niche area in a medium-risk area (35.2% (10,157 km2) for C. crossleyi to 36.6% (15,334 km2) for C. medius) and percentage of the realized niche in a low-risk area (22.9% (9,601 km2) for C. medius to 44.9% (12,988 km2) for C. crossleyi) (Fig. 4).

Fig. 4
figure 4

A Percentage of the realized niche areas of three dwarf lemur species situated in Madagascar’s protected area network, and (B) percentage of the realized niches within areas of high (< 5 km), medium (5–10 km), and low (> 10 km) risk of anthropogenic disturbance

The habitat suitability scores for two analysed species (C. major and C. medius) had no significant relationship with anthropogenic disturbance, as indicated by confidence intervals overlapping with zero (Table VI). C. crossleyi had a significantly negative habitat score relationship with HFP (Table VI).

Table VI Results of linear-regression analyses investigating the influence of HFP on the habitat suitability score of a subset of 100 random 1-km2 points within the realized niche areas of three dwarf lemurs of Madagascar


Ecological Niche Models

Our ENMs define the ecological niches of three dwarf lemur species. The AUC values and significance levels of the binomial tests of omission suggest that the ENMs of these species are useful for predicting their distributions (Table II). The performance of these ENMs mirror the success of other studies that have used climatic and vegetation cover-related variables to model the niches of cryptic species (Hending, 2021b; Kamilar et al., 2016; Rissler & Apodaca, 2007; Sattler et al., 2007; Schüßler et al., 2021, 2023; Thorn et al., 2009).

Cheirogaleus Niche Determinants and Species Niche Overlap

Our results suggest that the geographic distributions of some dwarf lemur species are more-strongly influenced by climatic variables. For example, the distribution of C. crossleyi appears most-strongly related to temperature, indicating that this species may be more sensitive to high temperatures than other Cheirogaleus or that the upper thermal limits of its preferred feeding/sleeping trees may be low (Sentinella et al., 2020). Precipitation appears the primary driver of C. major distribution; this species is a lowland rainforest specialist (Blanco et al., 2020d) and is thus sensitive to drought and unable to survive in dry habitats. The occurrence of C. medius appears to be influenced by vegetation presence and productivity (Table III). This indicates that this species may be sensitive to habitat degradation (but see Hending et al., in press). Dwarf lemurs spend large periods of the austral winter in seasonal hibernation, a trait evolved to enable their survival during periods of water crises and food shortages associated with the dry season (Dausmann & Blanco, 2016). Heterothermy might therefore buffer dwarf lemurs from spatial variations in temperature and rainfall and might allow some species to persist in a wider range of seasonal environments and be more widespread. Whereas the influence of climate on the ENMs can be explained by dwarf lemur heterothermy, the influence of NDVI and LAI is primarily explained by habitat requirements, because, such as many other primates, dwarf lemurs require forest habitat for their survival (Schwitzer et al., 2013). Species with higher NDVI and LAI ENM percentage contributions are more sensitive to vegetation density and have niche preferences of dense-vegetation forest (eastern rainforest: C. major), rather than more open-vegetation forest (western/southern dry forests: C. medius).

Significant niche divergence among species pairs, as per our original prediction (Table IV), suggests that each dwarf lemur occupies a distinct ecological niche (Chetan et al., 2014). The taxonomy of the dwarf lemurs has undergone significant upheaval in recent years (Lei et al., 2014, 2015), and while visual discrimination among members of this cryptic primate group may be difficult (Groeneveld et al., 2009), our niche-overlap results support the existing genetic and morphological data that classifies these species as distinct (Frasier et al., 2016; Herrera et al., 2016; McLain et al., 2017). Dwarf lemur sympatry does however occur in some areas (Blanco et al., 2009; Lahann, 2007). Local-scale resource partitioning needs to be further studied so that we understand how dwarf lemur coexistence can occur.

Realized Niches

Unfortunately, many areas of the fundamental niches identified in species ENMs often do not contain the habitat types that the organism would require to survive (realized niche) due to habitat loss and disturbance (Escobar et al., 2015). This is the case for our study, as is demonstrated when the suitable habitat areas are clipped to account for the availability of forest (Vieilledent et al., 2018). Although dwarf lemur populations are likely to exist outside of natural forest (Hending et al., 2018; Webber et al., 2020), the conservative habitat area estimates of the forest-clipped maps (realized niches) are the most appropriate means to gauge their current status. This is because some dwarf lemur species now only occur over small areas, particularly C. shethi, C. lavasoensis and C. thomasi (Fig. 1), although we did not construct ENMs for these taxa due to low occurrence point sample sizes. Coupled with the high rate of deforestation in Madagascar (Vieilledent et al., 2018), this information highlights the necessity for urgent conservation of the remaining dwarf lemur populations and the habitats within their ranges.

Protected Areas and Anthropogenic Disturbance

The results of our study show that 58.9% of the realized niche areas of the analysed species were within Madagascar’s protected area network (Fig. 4A). Such results provide information on which species are well-protected (or underprotected) within their geographic range, and they are therefore important to assign conservation priorities so that key habitat areas are preserved (Sharma et al., 2018; Thorn et al., 2009). Our results also suggest that dwarf lemurs, and their forest habitats, are well-protected within their ranges. As deforestation, habitat fragmentation, and climate change continue to threaten Madagascar’s forests (Hending et al., 2022; Vieilledent et al., 2018), this is encouraging for their survival. However, the percentage of the realized niche area that is protected varies between species, and more than 50% of the realized niche of C. medius is under no protection at all. Due to its wide distribution, this is unlikely to be detrimental for C. medius conservation. However, species with smaller distributions (e.g., C. lavasoensis and C. thomasi) would be much more threatened and likely to depend on protected areas for survival. Species-specific conservation measures therefore need to be implemented to protect their remaining populations, including ecological research and conservation action plans. For example, less than half (47.8%) of C. medius’ realized niche area is protected, and so conservation efforts for this species could focus on identifying additional areas for consideration for protection. Furthermore, improved enforcement within protected areas and community conservation initiatives to counteract local community reliance on cutting down forest are needed. If such actions are successfully implemented, extensions of Madagascar’s protected area network may be a viable strategy.

Although some of the results are encouraging for dwarf lemur conservation, this study also highlights that 63.9% of realized niche areas are within 10 km of an anthropogenic feature. Considering that people venture several kilometres from roads and villages to hunt and harvest from neighbouring forests (Peres & Lake, 2002), this of great conservation concern and it suggests that dwarf lemurs may be at great risk. The effect of anthropogenic disturbance on habitat suitability is not a significant factor for some species (Table VI) and may in fact be a positive influence (Hending, 2021a), which is reflected by observations of dwarf lemurs within anthropogenic habitats (Hending et al., 2018; Webber et al., 2020). However, habitat suitability of C. crossleyi appears to be negatively impacted by human activity. These results suggest that this species may be more-specialized than others that can tolerate and adapt to anthropogenic disturbance (Devictor et al., 2008). It has been hypothesised that disturbance gradients have greater effect on ecologically specialized, narrow-niched species than broader-niched generalists (Botts et al., 2013; Devictor et al., 2008). Our results demonstrate that the dwarf lemurs reflect this to some degree.

Study Limitations

While this investigation has provided an overview of dwarf lemur conservation biogeography, there are some limitations that originate from the occurrence point dataset and the environmental layers. Foremostly, the number of occurrence points of six species (C. andysabini, C. grovesi, C. lavasoensis, C. shethi, C. sibreei, and C. thomasi) were too low to construct reliable ENMs. Increased sampling efforts of the occurrence of more widespread species may alleviate this issue, which would allow the construction of reliable ENMs for these species. This would be highly useful to inform their conservation. However, this would not work for species with small, restricted geographic ranges, such as C. thomasi. Also, the resolution of our environmental layers was limited to 1 km2, which was the finest resolution available (resolution of all input-layers must be identical). Higher-resolution layers often result in improved Maxent model performance (Ross et al., 2015), but until such layers become available, this aspect cannot be improved. Higher-resolution layers would enable studies such as this to find out more information about species with restricted ranges (i.e., more pixels would be available). Finally, the percentage values of dwarf lemur realized niches that occur within protected areas may be influenced by the preference of researchers to work within protected areas. However, we were able to account for this sampling bias by including bias raster layers in our Maxent models.

Implications and Conclusion

The analyses of our ENMs provide evidence to suggest that dwarf lemur distribution is affected by both climatic and vegetation cover-related factors. Our results support the theory of divergent ecological niches among closely related species (Kamilar et al., 2016; Rissler & Apodaca, 2007). The dwarf lemurs, and indeed many other Cheirogaleid lemurs, often are described as generalist, adaptable species in the literature (Kamilar & Muldoon, 2010; Schäffler & Kappeler, 2014). Although our results provide evidence of this for C. medius, our ENMs suggest that others, such as C. crossleyi, are much more specialized and as a consequence are at much greater risk from anthropogenic disturbance. These findings underpin the importance of ecological niches for the informed understanding of the taxonomy and biogeography of cryptic organism groups, such as nocturnal strepsirrhine primates, and highlight their importance for the assessment of threats and assignment of conservation priorities.