Use of Habitat Suitability Model as a Tool to Highlight Best Conservation Area for the Red-bellied Monkey (Cercopithecus erythrogaster erythrogaster) in Southern-Benin, West Africa

© 2022 Academia Sinica, Taiwan Open Access Use of Habitat Suitability Model as a Tool to Highlight Best Conservation Area for the Redbellied Monkey (Cercopithecus erythrogaster erythrogaster) in Southern-Benin, West Africa Omobayo Ghislain Zoffoun1,2,* , Côme Agossa Linsoussi2, Georges Nobimè2, and Brice Augustin Sinsin2 1GeoEnvironment and Biodiversity Conservation (GeoEBC-NGO), Cotonou, Benin. *Correspondence: E-mail: [email protected] (Zoffoun). Tel: +229 67706663. 2Laboratory of Applied Ecology (LEA) of the Faculty of Agronomics Sciences (FSA) of the University of Abomey-Calavi (UAC), Abomey-Calavi, Benin. E-mail: [email protected] (Linsoussi); [email protected] (Nobimè); [email protected] (Sinsin) Received 4 February 2022 / Accepted 30 May 2022 / Published 29 September 2022 Communicated by Teng-Chiu Lin Wildlife habitats are increasingly degraded as a result of anthropogenic pressures. The IUCN recently updated the red list category of the red-bellied monkey (Cercopithecus erythrogaster erythrogaster) from Endangered to Critically Endangered due to its population decrease, habitats degradation and various threats to its conservation. It is therefore important to identify areas of great importance for the sustainable conservation of the subspecies. The Species Distribution Model (SDM) is a method increasingly used by conservationists to help find these areas and thus limit areas of intervention. In this study, maximum entropy model was used to identify suitable habitats for the red-bellied monkey in landscape of southernBenin from occurrence data and selected predictor variables according to ecological habitat requirements of the subspecies. The suitable habitat model for the red-bellied monkey has a good predictive power (AUC = 0.97). The variables that contributed most to the final model, as indicated by the permutation importance, were: Distance to Water (47.7%), Land Cover Class (23.1%), Brightness (17.0%), Wetness (4.7%), Human Population Size (2.8%) and Elevation (2.2%). Thus, using Maximum Training sensitivity and Specificity threshold, 3.62% of the landscape was classified as suitable and 96.38% was classified as unsuitable for the red-bellied monkey. The largest area of suitable habitat is found in protected areas (57.46%), mainly in the Lama Forest Reserve central core (49.5%). The landscape is fragmented and 91.49% of suitable habitats are between 0 and 0.01 km2 in size. The mean size of suitable habitats in the landscape is 0.017 ± 0.545 km2. Nevertheless, there is no significant difference between the mean size of suitable habitats in protected areas and those in the unprotected area (P = 0.061, Mann-Whitney U tests). The Average Nearest Neighbor Distance of suitable habitats in the landscape is low (0.139 km) and the Average Nearest Neighbor Ratio (R) is less than 1 (R = 0.408, p < 0.001). Those features indicate a clustered pattern of suitable habitats for the red-bellied monkey in the landscape. This makes it possible to foresee the establishment of connections between the isolated suitable habitats and thus allow for the long-term conservation of the species populations. Key words: Red-bellied monkey, Species Distribution Model, Suitable habitat, Protected areas, Anthropogenic pressures. Citation: Zoffoun OG, Linsoussi CA, Nobimè G, Sinsin BA. 2022. Use of habitat suitability model as a tool to highlight best conservation area for the red-bellied monkey (Cercopithecus erythrogaster erythrogaster) in Southern-Benin, West Africa. Zool Stud 61:47. doi:10.6620/ZS.2022.61-47. Zoological Studies 61:47 (2022) doi:10.6620/ZS.2022.61-47 1 © 2022 Academia Sinica, Taiwan BACKGROUND In Africa, non-human primate populations have been most affected by the loss and alteration of forest habitats resulting from factors such as intensive logging, shifting cultivation, wildland fire, excessive cutting of firewood to commercial and domestic purposes, pasture creation and clearing for crop establishment (Estrada et al. 2017; Badiella-Giménez et al. 2021). Bergl et al. (2008) noted that primates are particularly vulnerable to deforestation because of their dependence on tropical forest habitats and low reproductive rates. Therefore, a better understanding of environmental factors such as forest cover distribution and species distribution modelling would be vital to improve the conservation and appropriate management of forest protected areas in Africa. Benin, a country in West Africa, is home to a diversity of eleven primates species (Nobimè et al. 2010), including the red-bellied monkey (Cercopithecus erythrogaster erythrogaster) subspecies endemic to Dahomey Gap (Nobimè 2012). Cercopithecus erythrogaster erythrogaster is listed as Critically Endangered by the International Union for the Conservation of Nature (Goodwin et al. 2020) and is listed in Appendix II of the Convention on International Trade in Endangered Species of Wild Fauna and Flora (UNEP-WCMC 2014). According to Sinsin et al. (2000), suitable habitats for the red-bellied monkey in Benin are highly fragmented. Pressure from farmers who continue to cut trees for firewood and clear land for agriculture is predicting a continuing decline in the area of occupancy of the red-bellied monkey (Nobimè et al. 2009). Large area of the species range then appears to be unoccupied and those portions where the habitats are suitable and the species are truly present are not equivalent. It is therefore important to determine what proportion of the suitable habitats are found in protected areas and in what part of the protected area they are (Rodrigues et al. 2003). In order to develop species conservation plans in areas threatened by deforestation and habitat change, it is vital to collect data on their habitat preferences, as well as on the availability and location of suitable habitats (Estrada et al. 2017). In research, collecting extensive data on a large scale can be very costly, time-consuming, if not impossible. Available field data are often insufficient to meet all conservation needs (Polasky et al. 2000; Wilson et al. 2005). This difficulty has led biologists to develop methods of modelling the distribution of taxa over the last thirty years. On the one hand, they make it possible to better understand the potential distribution of a species and its habitats and, on the other hand, these models help researchers better understand conservation issues. The Species Distribution Model (SDM) can provide information on the potential distribution of a species when occurrence data are insufficient, and can assist with conservation planning (Guisan and Zimmermann 2000) by highlighting unknown populations (Pearson et al. 2007), suitable sites for reintroduction (Hernandez et al. 2006), and key areas for fieldwork (Papes and Gaubert 2007), improve the assessment of threat status (Solano and Feria 2007). This method will then enable us to identify suitable habitats for red-bellied monkeys, which is necessary in conservation ecology for the protection of species. This study aims to contribute to a better assessment of the conservation status of the red-bellied monkey in Benin. The specific objectives were to (1) identify precursor habitat variables suitable to the redbellied monkeys, (2) identify suitable habitats for the red-bellied monkey in the study area in southern Benin, and (3) analyze the spatial structure of suitable habitats for the red-bellied monkey. MATERIALS AND METHODS Study area This study was carried out in southern Benin, a region characterized by a subequatorial climate with two dry seasons (a small one from August to September, the large one from December to March) and two rainy seasons (a large one from April to July and a small one from September to November) and is found in the Guineo-Congolian phytogeographic area. The study area covers an area of 1997 km2 (Fig. 1) and covers four administrative regions: the Zou, the Atlantic, the Ouémé and a very small part in the Couffo. The average rainfall of the study area recorded at the Bohicon synoptic station by ASECNA was 1131 mm for the period ranging from 1988 to 2018. The average minimum temperature for the same period is 28°C with maximums of 37°C and minimums of 21°C. The relative humidity is 65% on average and ranges from 25% to 97%. Specifically, four forests where the red-bellied monkey has been previously reported as present (Campbell et al. 2008; Nobimè et al. 2009) were prospected and represent the survey sites: the Lama Forest Reserve (16,250 ha), the Lokoli swamp forest (3000 ha), Gnanhouizounmè forest (6 ha) and the Togbota forest (2 ha) (Fig. 1). The landscape including the four forests object of the present study is the preferred area of the red-bellied monkey in Benin (Sinsin et al. 2002) and is, therefore, our study area. In the study area there are three protected areas: Lama Forest Reserve, Agrimey Forest Reserve and Djigbé Forest Reserve. The Lama page 2 of 15Zoological Studies 61:47 (2022) © 2022 Academia Sinica, Taiwan Forest Reserve extends between 6°55' and 7°00' North and between 2°04' and 2°12' East and covers an area of 16,250 ha. But the portion of semi-deciduous dense forest, commonly called the Central Core, in this area covers only 4,785 ha (ONAB 2011). About 10,046 ha of the Lama Classified Forest are covered by teak plantations and 1,418.83 ha by resettlement perimeters for local populations, which are used for housing and agricultural production. The Agrimey Forest Reserve extends between 7°1' and 7°4' North and between 2°2' and 2°12' East. The forest covers an area of 2624.39 ha, including 2495.38 ha of mainly teak plantation for the production of timber, 124.67 ha of natural vegetation and 4.34 ha on which infrastructure is located (ONAB 2005). The Djigbé Forest Reserve extends between 6°81' and 6°91' North and between 2°29' and 2°36' East. The forest covers an area of 3720.80 ha, including 3616.70 ha of mainly teak plantation for the production of timber, 102.06 ha of natural vegetation and 2.043 ha on which infrastructure is located (ONAB 2005). Data collection Occurrence points for the Species Distribution Model (SDM) for the red-bellied monkey The occurrence points used for the modeling of suitable habitats for the red-bellied monkey are those resulting from the counts of the red-bellied monkey in southern Benin taking place in 2019 in four forests: Lama Forest Reserve (16,250 ha), the Lokoli swamp forest (3000 ha), Gnanhouizounmè forest (6 ha) and the Togbota forest (2 ha). In Lama Forest Reserve, data were collected along 07 linear transects with a total length of 57.163 km, existing in the central core of the Lama Forest Reserve (Fig. 1) representing the tracks already traced by ONAB (Benin National Wood Office), the structure responsible for the management of the forest. Each transect crosses a diversity of forest vegetation, such as: typical dense forests, secondary forests, teak plantation areas, fallow land with Chromolaena odorata and former settlements area dominated by Elaeis Guineensis. They are all parallel and separated from each other by a distance about 1 km. The length of the transects varies between 3.98 and 9.04 km. The transects were walked in October 2019 at an average speed of around 1.25 km/h (Peres 1999). Surveys on the different transects occurred at 6 a.m. to 8 a.m. and from 4 p.m. to 6 p.m. (Peres 1999). For each observation of red-bellied monkey groups, the geographic coordinates using a GPS Garming 64 st, the type of habitat, the number of individuals, the height of the animal in the tree, and the species of the tree are noted. In the Togbota and Gnanhouizoumè forests, the line transect method was also used to collect occurrence data. Three (3) transects of 400 m were surveyed in Fig. 1. Location of the study area in Southern Benin (West Africa). The red points represent visual contacts GPS locations of the red-bellied monkey (N = 22) used in this study. Source: Field work, 2019 and Benin IGN Map, 2018. Spatial reference system: WGS 1984 UTM Zone 31. N page 3 of 15Zoological Studies 61:47 (2022) © 2022 Academia Sinica, Taiwan Gnanhouizoumè forest and tree (3) transects of 250 m in Togbota forest. In Lokoli swamp forest, navigable only by canoe, the data were collected along navigable tracks of 12.505 km traversing the forest in the length with our canoes. Twenty-two (22) occurrence points were collected during fieldwork. This small number of occurrence points resulting from the fieldwork is due to the ability of the red-bellied monkey to quickly hide under the undergrowth and thus makes difficult to observe the subspecies in the field. The occurrences data was inspected to remove occurrence points that fell within the area of a single pixel of 900 m2. However, after this inspection, no two occurrence points fall within a single pixel of 900 m2, so number of occurrence points then remained 22. Predictive biophysical variables for SDM Raster layers of predictor variables dealing with landscape structure and land cover, called biophysical variables, were prepared at a spatial resolution of 30 meters. A total of 13 variables (Table 1) were generated: Land Cover Class (LCC), Normalized Difference Vegetation Index (NDVI), photosynthetically active vegetation (Greenness), soil moisture (Wetness), soil and surface luminosity (Brightness), Elevation, Aspect, Slope, Heat Load Index (HLI), Compound Topographic Index (CTI), Distance to Water (DW), Distance to Main Roads (DMR) and Human Population Size (PS). A supervised classification of a Landsat 8 OLI / TIRS image of December 2018 was carried out with the Semi-Automatic Classification Plugin (Version 6.3.1) for QGIS 3.6 to define the Land Cover Classes (LCC) using the maximum likelihood method (Congedo and Munafò 2012). The seven Land Cover Classes identified were Dense Forest, Mixed Forest, Forest Plantations, Farms and Fallows, Water, Settlement and Swamp. Vegetation indices were then calculated to capture differences at the micro scales because primates are reputed to have mental mapping abilities capable of perceiving their environment at the level of individual trees and forest plots (Ban et al. 2014; Fitzgerald et al. 2018). To do this, a Normalized Difference Vegetation Index (NDVI) raster was generated from the Landsat 8 OLI / TIRS image. NDVI is an indicator of biomass (i.e., healthy, photosynthetically active vegetation) within each raster cell and can range from -1 (water or bare soil) to 1 (dense and healthy vegetation). It is calculated using the near-infrared and red bands of a satellite image ((NIR - R) / (NIR + R)) (Campbell and Wynne 2011). In addition, the characteristics of the microhabitats in the study area were captured using a tasseled cap transformation from the original Landsat 8 OLI / TIRS image (ToA value i.e., Top of Atmosphere reflectance) (Ali and Salman 2016). This process transforms the original spectral data into a new coordinate system with six orthogonal axes. The first Table 1. Variables used to model habitat suitability of red-bellied monkey in response to the heterogeneity of the Southern Benin Forest landscape. After performing a correlation analysis, the 11 variables selected for use in the final model are marked with * N° Variables name Abbreviation Description Resolution (m) Units Source 1 Brightness * Brightness Soil and surface brightness 30 −Landsat 8 OLI 2 Greenness Greenness Measure of photosynthetically active Vegetation 30 −Landsat 8 OLI 3 Wetness* Wetness Soil moisture 30 −Landsat 8 OLI 4Normalized Difference Vegetation Index NDVI Index of relative biomass 30 −Landsat 8 OLI 5 Land Cover Class* LCC Categorization of land cover types in landscape: Dense Forest, Mixed Forest, Forest Plantations, Farms and Fallows, Water, Settlement and Swamp 30 −Landsat 8 OLI 6 Elevation* Elevation Height above sea level 30 Meters ASTER DEM v.3 7 Aspect* Aspect Direction a slope face 30 Degrees ASTER DEM v.3 8 Slope* Slope Steepness of a surface 30 Degrees ASTER DEM v.3 9Compound Topographic Index* CTI Relative variation in water availability on the ground 30 −ASTER DEM v.3 10 Heat Load Index* HLI Temperature experienced on the ground 30 −ASTER DEM v.3 11 Distance to Water* DW The minimum distance to the nearest water area 30 Meters Water shapfile from LCC 12 Distance to Main Roads* DMR The minimum distance to the nearest main roads 30 Meters Roads shapfile from IGN Benin 13 Population Size* PS Number of persons per pixel considered 30 Persons per grid WorldPop datasets page 4 of 15Zoological Studies 61:47 (2022) © 2022 Academia Sinica, Taiwan three axes that were used represent the following indexes: (1) the luminosity of the soil and the surface (Brightness), (2) the photosynthetically active vegetation (Greenness), and (3) the soil humidity (Wetness) (Serckx et al. 2016). These are good indicators of how climate and vegetation can play an important role in determining suitable habitat. Wetness is an important climatic factor for the red-bellied monkey presence (Kassa et al. 2007; Nobimè et al. 2009). According to Sinsin et al. (2002), the red-bellied monkey is only present in natural vegetation where the topographic situation and the edaphic conditions make it possible to keep the water on the surface periodically and to maintain a high humidity throughout the year. Therefore, the following topographic variables included in the initial model were chosen for their ability to serve as explanatory variables for this factor. These variables were generated using QGIS v.3.6.3 and derived from a Digital Elevation Model (DEM) with a resolution of ~30 m from Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) version 3 (NASA 2018): Elevation, Slope and Aspect. Aspect is the direction a slope faces (Flat, North, South, West, East, Northeast, Northwest, Southeast, Southwest). The sighting areas of the red-bellied monkey population are characterized by wet habitats: swamp forests, forest galleries and dense semi-deciduous forests that flood during the rainy season (Kassa et al. 2007; Nobimè et al. 2009). For this purpose, the variable Distance to Water (DW) was generated in QGIS v.3.6.3 as the euclidean distance of each raster cell of 30 × 30 m to the nearest area with water. In addition, two topo climatic indexes derived from DEMs were added to the other variables for their ability to determine wetlands characteristic of the red-bellied monkey habitat: Heat Load Index (HLI) and Compound Topographic Index (CTI). The first, Heat Load Index (HLI), provided a relative indication of temperature experienced on the ground, and the second, Compound Topographic Index (CTI), describes relative variation in water availability. HLI relates to evapotranspiration rates and soil temperatures and is a direct measure of incident radiation (McCune and Dylan 2002; Evans et al. 2014). CTI is a metric of potential ground wetness that is considered steady-state, or based on variables that remain relatively constant over time. CTI models water flow accumulation as a function of upstream contributing area and slope (calculated by percent rise). Prolonged exposure to water is a key factor in determining soil type, and CTI has been shown to be strongly correlated with many soil properties, including depth, texture, organic content and moisture (Moore et al. 1993; Evans et al. 2014). All the variables derived from the DEMs were finally resampled at 30 × 30 m using the bilinear method. Since human settlement development can induce disturbance and persecution that negatively affects the habitat of primates (Estrada et al. 2017), we also generated a raster with lowest Distances to the Main Roads (DMR), provided by IGN Benin (National Geographic Institute of Benin), as well as the Population Size (PS) i.e., persons per grid square, downloaded from https://www.worldpop.org/doi/10.5258/SOTON/ WP00023 at a resolution of ~100 m and resample at 30 m. Data analysis Species Distribution Modelling technique First, the correlation among the predictive variables was examined in order to reduce the effect that collinearity could have on the interpretation of Maxent’s results (Kumar et al. 2014). The Pearson (r) correlation was used for this purpose in R 3.5.3 (R Development Core Team 2019). For a set of highly correlated variables (| r | > 0.7, P < 0.05), the variable with the strongest predictive power in the preliminary model using all the 13 initial variables was chosen (Estes et al. 2010) (Table 1). To map the suitable habitats for the red-bellied monkey in the study area and analyze the biophysical variables contributing to their adequacy, we used a software that only works with presence data, namely Maxent 3.4.1 software based on maximum entropy (Phillips et al. 2017). Maxent’s algorithm has been shown to work well with presence data only and often outperforms other SDM methods (Elith et al. 2011; Wilson et al. 2013). Maxent estimates the relative probability of species presence from occurrence data (Fig. 1) and user-selected predictor variables (Phillips et al. 2006) (Table 1, Fig. 2). The result is a perfectly adapted model for classifying the locations in the study area according to the probability of presence (0 to 1, with 1 indicating the highest probability of presence). The predictive performance of the model is evaluated using the Area Under Curve (AUC). The AUC was chosen over other assessment measures because it does not require arbitrary threshold selection (Phillips et al. 2006). For presence data only, the AUC describes the probability that the model assigns a presence site a higher score than that of another site in the study area (Phillips et al. 2009). The performance of the model can be classified into three categories based on the value of AUC: fair (0.7–0.8), good (0.8–0.9) and excellent (0.9–1.0) (Phillips et al. 2006). To use the model in Maxent, we relied on recommended default values for the convergence page 5 of 15Zoological Studies 61:47 (2022) © 2022 Academia Sinica, Taiwan threshold (10-5) and maximum number of iterations (1000) (Phillips and Dudik 2008). The species presence data were randomly divided into 75% as the training dataset and 25% as the validation dataset. A crossvalidation procedure was replicated 10 times to account for uncertainty introduced by training and validation set splits and to obtain an average AUC value for the final model (Kumar et al. 2014). Since formal absence data were not available, a maximum of 10,000 background points was randomly generated to represent the availability and range of environmental conditions within the study area (Wilson et al. 2013). However, to limit the generation of background points only to the area in which occurrence data were collected—i.e., in the different prospected forests—a minimum convex polygon around the occurrence points was created. This ensures that the sampling of the background points is limited to the same region from which the occurrence points were collected and helps account for sampling bias (Phillips et al. 2009). The logistic outputs of habitat suitability were converted into binary outputs of unsuitable and suitable habitats using the threshold of Maximum Training Sensitivity and Specificity (Max TSS) as explained for the model generated employing presence-only data by Liu et al. (2013). According to Liu et al. (2013), Maximum Training Sensitivity and Specificity (Max TSS) is a promising method for threshold selection when only presence data are used and is more accurate than over methods. Besides, Maxent’s results generate response curves showing the relationship between each predictor variable and the predicted probability of the red-bellied monkeys’ presence, and the importance of permutation is reported for each variable. The importance of permutation is the contribution of each variable to the final model and is a measure of how the AUC changes when a variable is removed from the model and is not sensitive to the order in which the variables are inserted into the model (Wilson et al. 2013; Fitzgerald et al. 2018). Analysis of the spatial structure of suitable habitats using fragmentation parameters To analyze the spatial structure of suitable habitats to red-bellied monkeys in our study area, we estimated the following parameters: (1) area of suitable habitats, (2) number of fragments, (3) the mean size of the fragments and (4) the Average Nearest Neighbor Distance and Average Nearest Neighbor Ratio (R) (Moore and Carpenter 1999). An R value less than 1 indicates that the distribution is clustered and a value more than 1 indicates that distribution is dispersed or uniform. R = 1 indicates a completely random distribution pattern. In addition, the spatial distribution of suitable habitat fragments was analysed by pair correlation function g(r) following Pebesma and Bivand (2005) using spatstat package under R 3.5.3 (R Development Core 2019). Finally, the mean size of the suitable habitats fragments in the protected areas was compared to those on unprotected areas using the Mann-Whitney U test under R 3.5.3 in order to assess the contribution of protected areas in the conservation of wildlife habitats, especially red-bellied monkey. RESULTS Habitat suitability model for the red bellied monkey After checking the correlation between the variables, three pairs of variables were found to be highly correlated (| r | > 0.7, P < 0.05): Wetness and NDVI (r = 0.75), Greenness and NDVI (r = 0.98), and Greenness and Wetness (r = 0.73). The variable with the lowest permutation importance for the contribution to the model of the 13 starting variables was removed for each highly correlated pair. Finally, 11 variables were selected for the final model: Wetness, Brightness, Land Cover Class (LCC), Elevation, Aspect, Slope, Heat Load Index (HLI), Compound Topographic Index (CTI), Distance to Water (DW), Distance to Main roads (DMR) and Population Size (PS) (Table 1). The model of red-bellied monkey habitat suitability performed well, as indicated by the high AUC of 0.974, which indicates that the potential distribution of the subspecies fits well with our data and is therefore ecologically useful. The final model was classified according to the probability of presence and Fig. 2. Logic model representing the model inputs used to find suitable habitat for the red-bellied monkey in the study area in southern Benin. Red-bellied monkey occurrencePredictor variables Geo-statistical Analysis Habitat suitability model Threshold Suitable habitats / Unsuitable Overlay page 6 of 15Zoological Studies 61:47 (2022) © 2022 Academia Sinica, Taiwan is presented in figure 3A. Then, the habitat suitability model was converted to the binary outputs of unsuitable and suitable habitats using the threshold of Maximum Training Sensitivity and Specificity (Max TSS = 0.0898) (Fig. 3B). Thus, 3.62% of the landscape was classified as suitable and 96.38% was classified as unsuitable for the red-bellied monkey. The variables that contributed most to the final model, as indicated by the importance of the permutation, were: Distance to Water (47.7%) (this variable contributed the most to the model), Land Cover Class (23.1%), Brightness (17.0%), Wetness (4.7%), Population Size (2.8%) and Elevation (2.2%) (Table 2). These variables contributed to 97.5% of the final model and are therefore the precursors of the Fig. 3. Probability of red-bellied monkey presence (A) and suitable habitats (upper 0.0898 of red-bellied monkey presence probability) identified with the Maximum Training Sensitivity and Specificity (Max TSS) threshold (B). N Table 2. Permutation importance of each biophysical predictor variable used to create the final habitat suitability model Variables Permutation importance (%) DW 47.7 LCC 23.1 Brightness 17.0 Wetness 4.7 PS 2.8 Elevation 2.2 DMR 1.5 Aspect 0.4 Slope 0.4 CTI 0.2 HLI 0 page 7 of 15Zoological Studies 61:47 (2022) © 2022 Academia Sinica, Taiwan presence of the red-bellied monkey in the intervals of the values of these variables inducing a high probability of presence (Fig. 4A to 4F). The HLI variable did not contribute to the model (0%) and the Distance to Main Roads, Aspect, Slope, and Compound Topographic Index variables contributed very little (< 2%), and the response curves of these variables are not presented. The response curves of the Maxent model provides information on the relationship between the probability of presence (probability of suitable habitat) and the input biophysical variables. The spatial distribution of the variables is displayed above each response curve (Fig. 4A to 4F). After analyzing the response curves, it appears that the high probability of presence of the red-bellied monkey is obtained for area with the low Distance to Water (0–2000 m). The Land Cover Class inducing the greatest probability of the red-bellied monkey presence is Dense Forest (0.68) followed by Mixed Forest (0.08) and the one inducing the lowest probability of presence is Farms and Fallow (0.01). As for Brightness, low value (0.28–0.37) induced the high probability of presence. There is a strong positive correlation between the probability of presence and Wetness. Indeed, positive great wetness value (0.02–0.1) induced a high probability of presence. Besides, the probability of red-bellied monkey presence and Population Size evolves in two diametrically opposite directions. When the number of people in each 900 m2 grid is close to 0 the probability of presence is high (between 0.5 and 0.9) then decreases as the number of people tends to move away from 0, and when the number of people per grid reaches 5 the probability of presence cancel. Also, the probability of presence is high for positive low elevation value between 40 and 100 m. Spatial structure of suitable habitats The largest area of suitable habitats was found in protected areas (57.46%) (Fig. 5 and Table 3) and mainly in the Lama Forest Reserve and particularly in its central core with 35.81 km2 so 49.5% of the total area Fig. 4. Plots of the response curves for each variable depending on the probability of presence and map of their spatial distribution. Each plot represents a Maxent model using only the corresponding variable. The plots are given for the six biophysical variables (A to F) with highest permutation importance > 2% (percent shown on plot). The plots show the average response (red line) and the standard deviation (blue interval around the average). For Land Cover Class, two shades (blue and green) represent upper and lower limit defined by standard deviation. N N page 8 of 15Zoological Studies 61:47 (2022) © 2022 Academia Sinica, Taiwan Fig. 4. (continued) N N N N page 9 of 15Zoological Studies 61:47 (2022)