Variation in abundance and habitat use of the critically endangered Microcebus gerpi across its fragmented range

A link between the abundance of species and their degree of ecological specialization has previously been suggested within the primate order. Many species of lemurs were only recently described and even basic ecological data are not yet available for them. We investigated the habitat use, abundance, and habitat characteristics of the critically endangered Microcebus gerpi and evaluated potential impacts of vegetation structure and human disturbances on variations in its abundance. We determined abundance by systematic nocturnal surveys along 13 transects that were also used for characterizing the vegetation structure in seven study sites that were widely distributed within its range. Although M. gerpi occurred in all studied lowland rainforest and littoral forest fragments in central eastern Madagascar and therefore has a higher ecological plasticity and wider distribution than previously thought, its actual Area of Occupancy is very small (339.78 km2) due to an extreme degree of habitat fragmentation throughout its range. M. gerpi occurred with a mean encounter rate of 3.04 individuals/km but abundance varied substantially between sites (0.75–4.5 individuals/km). Statistical modeling revealed that the cover of small‐ to medium‐sized trees had a positive impact on the abundance of M. gerpi, whereas a composite disturbance score (CDS), formed on the basis of information on the prominence of fires, cattle, charcoal production and wood extraction inside and around the forest, had a negative impact on abundance. These results suggest that M. gerpi is slightly less threatened than expected because of its larger geographic range, but also that it responds negatively to human disturbances. These findings raise strong conservation concerns and question the long‐term viability of the remaining small and isolated populations of this arboreal solitary forager.

It is well established that there is a link between habitat specialization and geographical distribution in many taxonomic groups, that is, species with a narrow fundamental niche have a smaller geographical distribution than species with a broader fundamental niche (Brown, 1984;Harcourt et al., 2002;Slatyer et al., 2013).Likewise, a positive relationship has been stated between local abundance and geographical range size from early broad taxonomic comparisons (known as "distribution-abundance relationship," Bock & Ricklefs, 1983;Brown, 1984;Gaston, 1996), although no support could be later on found for this hypothesis in primates in general or in lemurs in particular (Harcourt, 2006;Harcourt et al., 2005).However, these studies did find evidence for a relationship between specialization and abundance (i.e., rarity) within primates (Harcourt et al., 2005).
Extant Malagasy lemurs comprise more than 100 species which fall into five families and 15 genera with almost all species being categorized as threatened with extinction by the IUCN (www.redlist.org, Schwitzer et al., 2014).All of them are endemic to Madagascar and inhabit various forest habitat types on the island.Two nocturnal genera, sportive lemurs (Lepilemur spp.) and mouse lemurs (Microcebus spp.), show a particularly high species diversity with currently 26 and 24 recognized species, respectively (Hotaling et al., 2016;Lei et al., 2017;Schüßler, Blanco, et al., 2020).Most of these species have been described only within the last 25 years and have relatively small, allopatric distributions, but even basic ecological or behavioral data are still missing for most of them.
Population abundances of mouse lemurs were previously shown to vary substantially between sites and species, and were found to be higher in the western dry forests than in the eastern rainforests (Setash et al., 2017).A recent meta-analysis conducted across all available data sets previously published for 16 different species revealed that mouse lemur population density seems to be negatively associated with plant productivity, the Leaf Area Index, and altitude, but positively associated with an index of human disturbance, all derived from remote sensing data (Hending, 2021).
Past and ongoing habitat loss and fragmentation are severe anthropogenic threats to the remaining Malagasy forest habitats and consequently to the lemurs (Schwitzer et al., 2014).It has been estimated that the island has lost about 44% of its forest cover between 1953 and 2014 (Vieilledent et al., 2018), and ongoing deforestation has been documented ever since (Schüßler, Mantilla-Contreras, et al., 2020).However, some species seem to be more resilient than others against anthropogenic habitat disturbances (Knoop et al., 2018).For example, the gray mouse lemur (M.murinus) which has the largest geographical distribution among all mouse lemur species (Mittermeier et al., 2010), can be found in disturbed forest, secondary forest, and sometimes even in very small forest fragments (Andriatsitohaina et al., 2019;Ganzhorn, 1995;Ganzhorn & Schmid, 1998), while their partially sympatric congener, the Critically Endangered Microcebus berthae, is not found in disturbed forests at all and is actually on the brink of extinction (Kappeler et al., 2022;Schwab & Ganzhorn, 2004).
The Critically Endangered Gerp's mouse lemur (Microcebus gerpi) was recently described (Radespiel et al., 2012) and inhabits the lowland forests between the Ivondro and the Mongoro rivers in eastern Madagascar (van Elst et al., 2023), but information on its habitat requirements, abundance variation within and between different forest types, and its tolerance toward anthropogenic habitat disturbance is still completely missing.The aim of this study was therefore to (1) characterize the habitat characteristics, that is, the used vegetation type, height above ground and vegetation structure for M. gerpi, (2) evaluate potential impacts of vegetation structure or disturbance intensity on its abundance, and to (3) determine the degree of habitat fragmentation within its distribution, the IUCNrelevant Extent of Occurrence (EOO), the Area of Occupancy (AOO), and its current threat status.

| Study sites
This study was conducted between August 2018 and September 2019 in eastern Madagascar in nine forest fragments that were located between the rivers Ivondro (north of Anjahamana) and Mangoro (south of Antanambao; Figure 1, Supporting Information: Table S1).The two westernmost study sites were located at >580 m above sea level (a.s.l.) (Vohimana and Ambohimarina), and field work as well as subsequent genetic analyses (results not shown) revealed that they did not contain M. gerpi but Microcebus lehilahytsara.They were therefore not included in the subsequent ecological work.All other sites were located at up to 570 m a.s.l.(Supporting Information: Table S1) and contained M. gerpi which was confirmed genetically in a parallel study (van Elst et al., 2023).These lowland study sites ranged in size between 0.05 and 13.1 km 2 (Supporting Information: Table S1).

| Field methods
The abundance of Gerp's mouse lemurs was determined by systematic nocturnal surveys employing distance sampling along transect paths following standard procedures (Buckland et al., 2001;Rakotondravony & Radespiel, 2009, Supporting Information: Table S1).One to three transect paths were surveyed in each study site with transect length varying between 0.85 and 1.5 km running through forests that were impacted by human activities in different ways (Table 1, Supporting Information: Table S1).A total of four nocturnal surveys was performed per transect, twice in each direction.Upon each encounter, the species, number of individuals, the position on the transect, the perpendicular distance to the animal, and its height above ground was noted.) Two transects in selectively logged primary forest.
Continuous logging for wood transfer to outside markets.Intense charcoal production partly from fast growing tree plantations but also from native forest (to be sold on regional markets).
Only Cheirogaleus sp.along both transects.

Note:
Additional details are provided in Supporting Information: Table S1.
score (CDS) for each of the seven study sites of M. gerpi which could thus range from 12 to 36 (Supporting Information: Table S2).All scores resulted from a composite judgment that was based on the integration of (a) outcomes from an open exchange on the customs of forest use with many villagers during a village meeting that always preceded the work in the forest, (2) our own observations inside and next to the forest during field work, and (3) results from 5 to 10 selected interviews with local field assistants, other people encountered during field work, and with key informants in the nearest village (e.g., chef de village, chef de Fokontany).We acknowledge that this approach has a limited spatio-ecological resolution and cannot clarify causal effects, but has the advantage of incorporating signals of disturbances that reach beyond the immediate location of the transects and may affect animals ranging within the fragment.This approach does not exclude subjective judgments from single informants, but the integration of all information obtained with the three approaches aimed at reducing strong biases and misclassifications as much as possible.This scoring approach was inspired by and adapted from the threat assessment guidelines presented in Margoluis and Salafsky (2001).
Ten vegetation parameters (defined in Table 2) were investigated at 17−30 sampling points (depending on the length of the transect) in 50 m intervals along each transect line.At each sampling point, canopy opening (CaO) and vegetation cover for five strata from ground to the upper canopy level (L1-L5, Table 2) were estimated for a circle with a diameter of 10 m in percent.Moreover, the density of four plant categories (big-sized (DBH > 10 cm), middle-sized (DBH: 5-10 cm), small-sized trees (DBH: 2-4.9 cm) (D1-3), and liana tangles in trees, Table 2) was estimated using the Point-Centered-Quarter method (PCQ, Mühlenberg, 1993) at each sampling point.The area around the point center was divided into four quarters following the four compass directions (N, S, W, E).In each quarter, the distances (d in meter) from the center to the nearest plant of the four density categories above were noted and used to estimate their respective density with the following formula: Density = S/d 2 (Mühlenberg, 1993), with S representing the surface unit of interest (here: 1 ha = 10,000 m 2 ) and d representing the mean of the distances (in m) from the PCQ-center point to the nearest vegetation element in the four quarters.We subsequently calculated the mean for each vegetation parameter over all sampling points per transect to be used for subsequent statistical analyses.

| Spatial analyses
We calculated the degree of forest fragmentation, the Extent of Occurrence (EOO), and the Area of Occupancy (AOO) for M. gerpi based on the inferred altitudinal limit of M. gerpi in conjunction with a digital elevation model (JAXA, 2015) and forest cover data from 2017 (Vieilledent et al., 2018) in between the two large rivers that form the northern and the southern border of the distribution range, according to the IUCN (2012) guidelines.S3).High loadings (>0.7,<−0.7) of variables onto the factors were used for subsequent factor interpretation (Supporting Information: Table S4).The PCA was performed with STATISTICA

| Habitat use and the effects of anthropogenic disturbance
M. gerpi was found in littoral forests and lowland rainforest fragments of various sizes (0.05 -13.1 km 2 , Supporting Information: Table S1).
All forest sites displayed signs of anthropogenic habitat disturbance, such as fires, cattle grazing, charcoal production, or selective logging.
The degree of these disturbances was not equal across study sites (Table 1).The scores of individual pressures varied from 3 to 9  S2).
An average of 3.04 ± SD 1.16 Gerp's mouse lemurs were encountered per kilometer of transect (Table 3).This mean encounter rate was smallest on one transect in Anjahamana (0.75 ind./km) and highest on one transect in Andobo (4.5 ind./ km, Supporting Information: Table S5).Encountered individuals (n = 198) used various forest strata and were observed between 0.4 and 10 m above ground with a mean value of 4.0 ± SD 2.0 m (Figure 2).However, in more than half of the encounters the mouse lemurs were observed between 3 and 6 m above ground (i.e., 55.6%, n = 110).
Forest structure varied substantially between forest sites with minimum values ranging from 1.3% (density of liana tangles) to 69.8% (cover of shrub layer, L4) of the mean, while maximum values were between 27.1% (cover of medium-sized trees, L2) and 178.8% (density of liana tangles) larger than the mean (Table 3).All tree strata (L1-L3) had an average cover between 43.3% and 58.5% with average tree strata covers never exceeding values of 75.2% per transect.Correspondingly, the canopy had gaps along all forest transects for which the average values ranged between 6.0% and 31.2%.The density of large (D1: 279−1926 trees/ha), medium (D2: 524-3344 trees/ha) and small trees (D3: 728-6459 trees/ha) also varied considerably between sites and transects (Table 3, Supporting Information: Table S5).The highest variation between sites, however, was detected for the density of liana tangles for which the standard deviation (SD = 386 tangles/ha) was even larger than the mean (mean = 351 tangles/ha).
T A B L E 2 Definitions for all used vegetation parameters (for more explanations see main text).Univariate generalized additive models (GAMs) revealed that the encounter rates for M. gerpi were significantly impacted only by 2 of the 13 ecological parameters.Encounter rates increased in response to an increasing cover of medium-sized trees (L2) and were negatively impacted by an increasing composite disturbance score (CDS) in the study sites (Table 3, Figure 3).Variations in L2 explained 55.3% (R 2 , Estimate = 0.0728, SE = 0.01829, t = 3.983, p = .00215),whereas the CDS explained 25.6% (R 2 , Estimate = −0.1716,SE = 0.07571, t = −2.267,p = .045) of the variations in the encounter rates.
The multivariate GAM based on the four principal components revealed only one significant influence on the mean encounter rates of M. gerpi and that was by principal component 2 which was mostly impacted by the cover of small trees (L3: +0.859 factor loading) and of the lowest forest stratum (L4: +0.852 factor loading, Table 4, Figure 4, Supporting Information: Table S4).This model explained 56.2% of the variation in the data set (R 2 ).L2 also loaded positively onto this factor (+0.605), but less so than L3 and L4 and slightly less than the chosen threshold 0.7.Given this result, the mean encounter rates for M. gerpi were higher in forests with a higher cover of the lower forest strata.gerpi.Lowland forests at elevations between 550 and 600 m a.s.l.

|
were still mostly connected to rainforests at higher altitudes, belonging to the protected areas of the Corridor Ankeniheny-Zahamena.These already protected forests made up 54.04% (183.6 km 2 ) of the total AOO.At lower altitudes, however, only isolated forest fragments of small sizes remained (Figure 1).The largest of these was Sahafina forest with 13.1 km 2 .Twenty further fragments ranged between 1.0 and 10.0 km 2 in size, while most fragments (n = 6324) were even smaller (Figure 5).( Tinsman et al., 2022).Conversely, the M. gerpi sites showed relatively high densities of small-and medium-sized trees, which may not yet be extensively used for wood extraction.These smaller and medium-sized trees may provide good protection from predators, substrates for locomotion, and food resources for small nocturnal, solitarily foraging mouse lemurs (Mittermeier et al., 2010;Radespiel, 2006).All forest strata of up to 10 m were used (Figure 2) and even higher strata might have been used occasionally with animals remaining undetected because of visibility problems.
This broad usage of forest strata is in agreement with generalist feeding habits, known from other mouse lemur species that descend close to and onto the ground to hunt arthropods, but also feed on various plant items and hunt in all parts of the vegetation including the uppermost canopy (Atsalis, 1999;Dammhahn & Kappeler, 2008;Lahann, 2008;Radespiel et al., 2006;Thorén et al., 2011).The preferential usage of the lower to middle forest strata (3-6 m above ground) shows further similarities with other mouse lemurs (Lahann, 2008;Radespiel et al., 2006), and may also reflect some vertical niche partitioning between different lemur species.Lahann
The abundance of M. gerpi was positively impacted by the forest cover of medium-sized trees (L2, univariate analysis) and, after PCA   and multivariate analysis, by the cover of small trees (L3) and the shrub layer (L4).These results are not contradictory, as all three layers had a positive loading onto the relevant PC2, and L3 and L4 already showed comparable estimates to L2 in the univariate analysis but may have suffered from a lack of power in that analytical step.
These results cannot be explained by a visibility artifact that would predict more encounters under better visibility conditions.By contrast, more mouse lemurs were encountered when these forest strata had a higher coverage and were therefore denser.These findings suggest that these lower to medium-sized forest layers may indeed be important for mouse lemurs, but whether their immediate benefits lie in increased protection from predators (Goodman et al., 1993), access to food (Radespiel et al., 2006) or other resources like for example sleeping sites (Radespiel et al., 2003), cannot be clarified without more in-depth knowledge of the behavioral ecology of this species.

| Conservation implications
The abundance of M. gerpi was negatively impacted by human pressures (CDS) suggesting that the resilience of this species to human activities is limited.(van Elst et al., 2023).Eventually, forest fragmentation in the range of M. gerpi will likely lead to local extinction events, since stabilizing meta-population dynamics and connectivity to a larger network of potential source populations may no longer exist (Steffens et al., 2020).Conservation policies should therefore give a high priority to reforestation initiatives within the range of M. gerpi, not only to relieve pressure of wood extraction from the remaining forest remnants and to increase the livelihoods of the human population, but also to reconnect isolated forest patches and thereby facilitate wildlife dispersal and thereby contribute to the long-term viability of highly fragmented and threatened animal populations.
In view of these severe spatial constraints and ecological threats, the small AOO, and the previously shown low genetic diversity and high genetic differentiation between sites (van Elst et al., 2023), and in accordance with the IUCN criteria (IUCN, 2012), M. gerpi should as of now be considered Endangered (2ab[ii,iii,iv]).Taken together, although M. gerpi may seem to be less threatened today than when it was described in 2012, utmost conservation attention should be given to this small lemur species and the other species inhabiting these very small forest remnants before it is too late.
The research protocols for the capture of mouse lemurs and for the nocturnal survey work along transects adhered to the American Society of Primatologists Principles for the Ethical Treatment of Nonhuman Primates and followed the American Society of Primatologists Code for Best Practices for Field Primatology.Furthermore, they adhered to the legal requirements of Madagascar and were fully approved by the national authority (Ministère de l'Environnement et Developpement Durable) who issued our research permits (N°124/18/MEEF/SG/DGF/ DSAP/SCB.Re and N°242/19/MEDD/SG/DGEF/DGRNE).

Four
types of anthropogenic disturbances were identified to some extent in all study sites: (a) recent fires in the immediate surroundings and inside the forest, (b) the presence of cattle ("Zebu") in the forest, (c) evidence for charcoal production from native forest trees, and (d) wood extraction, either by commercial logging, or by local extraction for selling on local markets or for private use.Because of time constraints in each site, each disturbance type was only scored qualitatively per site based on graded estimates of its intensity (weak [1]-average [2]-high [3]), spatial scale (local [1]zonal [2]-entire fragment [3]), and temporal scale (occasional [1]temporary [2]-regular [3]).The total score for each disturbance type could thus range from 3 to 9. The scores for each of these four potential disturbances were added up to a composite disturbance F I G U R E 1 Map depicting the study region and study sites across the distribution of Microcebus gerpi with populations sampled during this study and all known occurrences (van Elst et al., 2023).Forest cover in 2017 according to Vieilledent et al. (2018) and protected areas from UNEP-WCMC (2023).
forest with low levels of degradation.Recent fires came very close to the lower parts of the transect, while the upper parts were still under low pressure.Temporary, zonal level of wood extraction. in primary forest with low levels of degradation (eastern part of fragment).Transect 3 in degraded secondary forest (western part of fragment).Degradation resulted from previous fires and selective logging for local use.Low levels of all types of threats.Avahi sp.: Along all transects; Cheirogaleus sp.: Observed only along one transect.Andobo (2 transects)

2. 5 |
Analysis of the ecological impacts on abundanceBecause of the generally low number of encounters along the transects and the high variation in vegetation structure and visibility conditions between study sites, no transect-based population density estimate could be calculated.Instead, a standardized mean encounter rate (number of individuals encountered per km transect length) was calculated for each transect.Thirteen ecological variables (Cover of forest layers L1-L5, density of large, medium, and small trees (D1-D3), the estimate of canopy opening, density of liana tangles, CDS, fragment size, and mean altitude) were used for the evaluation of parameters that may have impacted the abundance of the Gerp's mouse lemur in its distribution range.Analyses started with a univariate analysis of the impact of each ecological variable (=fixed factor) on the encounter rate of M. gerpi (=response variable) on the 13 study transects.If a variable was not normally distributed, it was log-transformed (in case of cover of herb layer L5, fragment size, and altitude), and Q-Q plots were visually inspected before and after transformation to ensure improvement.Multivariate modeling was subsequently performed, but because of correlations between various ecological parameters, they were first subjected to a principal component analysis (PCA) after variable standardization.A total of 12 principal components were returned but only 4 of them had Eigenvalues >1 and were thus selected for subsequent modeling.These four factors individually explained between 8.49% and 42.70% of the variance in the data set and together explained 86.20% of cumulative variance (Supporting Information: Table 12(Statsoft, Inc).The coordinates of all cases (=study transects) for the four first principal components were then used together as fixed factors.All univariate and multivariate analyses included fitting a generalized additive model (GAM) to explain variations in the encounter rate of M. gerpi (link function = identity, family = Gaussian) by maximum likelihood estimation.All modeling was performed with the package mgcv(Wood, 2017) and the function gam () in R (v 4.2.1) and RStudio (v.1.4.1717)(Posit team, 2022; R Core Team, 2019).

(
Fires: 3−8, Zebu: 3−5, Charcoal: 3−7, Wood extraction: 3−9).The overall CDS was smallest in Sahafina (CDS = 12) and highest in Anjahamana (CDS = 22, Supporting Information: Table Degree of habitat fragmentation, Extent of Occurrence, and Area of Occupancy Given the observed altitudinal limit of ≤570 m a.s.l. and the distribution between the large rivers Ivondro and Mangoro (van Elst et al., 2023), the Extent of Occurrence (EOO = all areas below 600 m a.s.l., including a narrow altitudinal buffer) for M. gerpi was estimated as 11,212 km 2 , of which only 3.03% (339.78 km 2 ) were still forested in 2017, representing the Area of Occupancy (AOO) for M.

|
Habitat characteristics of M. gerpiGerp's mouse lemur was found in fragments of two main habitat types, littoral forest and lowland rainforest, but not in mid-altitude or montane rainforests.Because of various types of human activities, habitats were partially degraded and, as a consequence, the vegetation structure varied between sites and even between transects within a given site, although causal relationships cannot be evaluated in detail within this study.However, human activities such as wood extraction may explain, for example, the rather limited tree cover for large trees with DBH > 10 cm (D1, mean: 43.3%) and the regular occurrence of canopy gaps (mean = 18.0%), which indicates a higher disturbance of the canopy than in the primary and secondary forest parts of the Lokobe National Park (Nosy Bé), another site of lowland rainforest on Madagascar F I G U R E 2 Use of vegetation height classes by Microcebus gerpi in the study sites (n = 198 sightings).

(
2008) argued that an average height above ground of 3.0 ± 1.8 m (SD) allows Microcebus sp. in the littoral forests of Mandena to coexist with two Cheirogaleus species that preferentially used the intermediate (Cheirogaleus medius: 4.7 ± 1.7 m) and higher strata (C.major: 7.6 ± 2.6 m) of the forest.Whether M. gerpi competes with other lemurs in the study region cannot be clarified.However, they co-occurred with Avahi sp. and with Cheirogaleus sp. in sites.While competition with Avahi sp. is rather unlikely because of their much larger body size (about 1 kg,Mittermeier et al., 2010) and different feeding habits, some vertical niche partitioning could be expected with Cheirogaleus sp.(Lahann, 2008).
Relationship between mean encounter rates for Microcebus gerpi and (a) the forest cover of medium-sized trees (L2), and (b) the composite disturbance score (CDS) in all study sites.Black line: local regression line with Loess smoothing (span: 0.9), in gray: 95% confidence interval.

F
I G U R E 4 Relationship between principal component 2 and the mean encounter rates for Microcebus gerpi estimated as individuals/ km.Black line: local regression line by Loess smoothing (span: 0.9), in gray: 95% confidence interval.
Results of the GAM with all four principal components (=PCs) as fixed factors for explaining variations in mean encounter rates for Microcebus gerpi.
T A B L E 4 Note: Bold values indicate statistically significance.Abbreviations: ER, encounter rates; SE, standard error.*p < 0.05.
To avoid the risk of losing entire forest fragments (Vieilledent et al., 2018)ge information(van Elst et al., 2023), a conservative estimate of the Extent of Occurrence (EOO) for M. gerpi was approximated with 11,200 km 2 which is in the range of the EOO for the Vulnerable Microcebus arnholdi (13,577 km 2 ) or the Vulnerable Microcebus ravelobensis (9486 km 2 ) (www.redlist.org).However, only a very small fraction of the range (approx.3%) of M. gerpi was still forested in 2017(Vieilledent et al., 2018), and a mere 339.78 km 2 was established as M. gerpi's Area of Occupancy (AOO).In addition, the AOO is highly fragmented with thousands of minor patches and only few (n = 39) with sizes above 1 km 2 .The largest forest fragment in the entire range of M. gerpi is its holotype locality (Sahafina) which has a size of only 13.1 km 2 .All these small forest remnants are exposed to edge effects, which could have negative effects on mouse lemur populations as shown for M. murinus(Burke & Lehman, 2014), but moreover suffer from a loss of connectivity which was shown in a parallel molecular study