Habitat differentiation among three Nigeria–Cameroon chimpanzee (Pan troglodytes ellioti) populations

Abstract Ecological niche models (ENMs) are often used to predict species distribution patterns from datasets that describe abiotic and biotic factors at coarse spatial scales. Ground‐truthing ENMs provide important information about how these factors relate to species‐specific requirements at a scale that is biologically relevant for the species. Chimpanzees are territorial and have a predominantly frugivorous diet. The spatial and temporal variation in fruit availability for different chimpanzee populations is thus crucial, but rarely depicted in ENMs. The genetic and geographic distinction within Nigeria–Cameroon chimpanzee (Pan troglodytes ellioti) populations represents a unique opportunity to understand fine scale species‐relevant ecological variation in relation to ENMs. In Cameroon, P. t. ellioti is composed of two genetically distinct populations that occupy different niches: rainforests in western Cameroon and forest–woodland–savanna mosaic (ecotone) in central Cameroon. We investigated habitat variation at three representative sites using chimpanzee‐relevant environmental variables, including fruit availability, to assess how these variables distinguish these niches from one another. Contrary to the assumption of most ENM studies that intact forest is essential for the survival of chimpanzees, we hypothesized that the ecotone and human‐modified habitats in Cameroon have sufficient resources to sustain large chimpanzee populations. Rainfall, and the diversity, density, and size of trees were higher at the rainforest. The ecotone had a higher density of terrestrial herbs and lianas. Fruit availability was higher at Ganga (ecotone) than at Bekob and Njuma. Seasonal variation in fruit availability was highest at Ganga, and periods of fruit scarcity were longer than at the rainforest sites. Introduced and secondary forest species linked with anthropogenic modification were common at Bekob, which reduced seasonality in fruit availability. Our findings highlight the value of incorporating fine scale species‐relevant ecological data to create more realistic models, which have implications for local conservation planning efforts.


| INTRODUC TI ON
Ecological niche models (ENMs) are widely used to characterize habitat suitability for a species in a given location, and information from these models may be used to predict the species distribution patterns, densities, and trends (Junker et al., 2012;Sesink Clee et al., 2015). Several recent ENM studies estimate suitable habitats of apes using known ape distributions (Junker et al., 2012;Sesink Clee et al., 2015;Strindberg et al., 2018) and project ape population decline due to anthropogenic pressures and infectious disease risk, most notably resulting from Ebola; and climate change Strindberg et al., 2018;Walsh et al., 2003). These studies rely upon global environmental datasets that describe several abiotic and biotic factors, such as tree cover, surface moisture, precipitation, and seasonality, generally sampled at 1-km 2 resolution (Dimiceli et al., 2011;Farr et al., 2007). However, it is often unclear how these variables directly relate to the resources available to species in their habitats, and most importantly, how the resources used by populations of apes correspond with these remotely sensed abiotic and biotic variables. Understanding these relationships is an important starting point in order to translate the relationships that exist between ENMs based on habitat suitability, and how these models apply at a scale that is ecologically relevant to ape communities and the resources that they rely on for survival.
Habitats occupied by different chimpanzee populations vary (Stumpf, 2011) and could be important in understanding socioecological and genetic diversity in the species. Therefore, understanding the link between ENMs and chimpanzee-specific habitat requirements is important. Four geographically distinct subspecies of chimpanzees occur in Africa, from Senegal in west Africa to Tanzania in the east (Figure 1a,b) (Caldecott & Miles, 2005).

There is a western lineage that includes Pan troglodytes verus and
Pan troglodytes ellioti and a central-eastern lineage that includes P. t. troglodytes and P. t. schweinfurthii (Prado-Martinez et al., 2013).
The processes that have generated the distribution and diversity of these chimpanzee subspecies are largely unexplored (Mitchell & Gonder, 2013). Cameroon is of particular interest in this regard, as it constitutes an area of active chimpanzee diversification (Gonder et al., 2011;Gonder, Locatelli, Ghobrial, & Sheppard, 2009;Mitchell, Locatelli, Ghobrial, et al., 2015). The two main branches of the chimpanzee phylogenetic tree split at the Sanaga River in central Cameroon (Gonder, Disotell, & Oates, 2006;Gonder et al., 1997;Prado-Martinez et al., 2013), and the river also marks the separation of P. t. troglodytes and P. t. ellioti (Gonder et al., 2011;Mitchell, Locatelli, Ghobrial, et al., 2015). There is a further population subdivision found within P. t. ellioti. There is one genetic population, or gene pool, associated with the mountainous rainforest habitats in western Cameroon, and a second genetic population found in the forest-woodland-savanna mosaic (ecotone) in central Cameroon (Mitchell, Locatelli, Sesink Clee, Thomassen, & Gonder, 2015). There is evidence that variation across these habitats plays an important role not only in sex-specific community structuring (Mitchell, Locatelli, Abwe, Ghobrial, & Gonder, 2018), but also in the partitioning of genetic diversity within P. t. ellioti .
Habitat differences are also reflected in chimpanzee nesting patterns with relatively larger parties associated with rainforest compared to drier and savanna habitats (Basabose & Yamagiwa, 2002;Brownlow, Plumptre, Reynolds, & Ward, 2001;Hunt & McGrew, 2002). However, in drier habitats where chimpanzees are also sympatric with predators, nesting parties are larger as smaller foraging parties congregate at nesting sites for safety (Ogawa et al., 2007).
A recent study that modeled habitat suitability for chimpanzees in Cameroon revealed that the two genetically distinctive P. t. ellioti subpopulations reported in Mitchell, Locatelli, Ghobrial, et al. (2015) occupy two significantly different niches that were significantly different from one another and from the niche occupied by P. t. troglodytes in southern Cameroon . Suitable habitats for P. t. troglodytes were fairly homogenous, especially in annual rainfall, forest cover and relief. In contrast, suitable habitats for P. t. ellioti were characterized by greater variation in precipitation and temperature seasonality, forest cover and relief . Differences in these environmental conditions were especially pronounced between the western mountainous rainforest and ecotone habitats , and the differences between them broadly corresponded with the distribution of the two genetically distinctive populations of P. t. ellioti.
However, while niche variation captured through ENMs is salient and informative, ecological details including forest structure, species richness, and fruit phenology that are important to frugivores cannot be depicted through such models. The chimpanzee range across Africa is marked by environmental and ecological variation, and differences in chimpanzees socioecology are tied to this variation, including feeding and nesting behaviors (Stumpf, 2011). Evidence from Cameroon shows that environmental and ecological differences between habitats may be important in the evolution of chimpanzee subspecies. Current evidence about niche differences among the subspecies comes only from habitat suitability models from remote sensing GIS data and is therefore only a starting point for examining the environmental and ecological variation that may contribute to the evolution of chimpanzee populations. In order to place habitat suitability models into a spatial and temporal scale that is ecologically relevant to chimpanzees, it is necessary to ground-truth them with data regarding local environmental factors, as well as the distribution and availability of resources that chimpanzees rely on for survival.

| MATERIAL S AND ME THODS
We collected fine scale environmental and ecological data including variables reported previously to be important in determining regional differences in chimpanzee socioecology (e.g., Stumpf, 2011) at three locations in Cameroon: Njuma, Bekob, and Ganga ( Figure 2).
These three sites represent each of the two gene pools in P. t. ellioti: in a mostly mature rainforest at Ebo (Njuma), at a rainforest location at Ebo that was heavily modified by small stakeholder agriculture until it was abandoned in the 1960s (Bekob), and finally, in a forest-woodland-savanna ecotone at Mbam & Djerem National Park (MDNP: Ganga).

| Ebo forest
The Ebo Forest is in the Littoral Region, Cameroon, and extends for more than 1,500 km 2 , of which approximately 1,200 km 2 is proposed as a national park. With a conservative estimate of at least F I G U R E 2 Study sites. The gradient describes the distribution of the two P. t. ellioti gene pools: P. t. ellioti-rainforest (white) and P. t. ellioti-ecotone (gray) west and east of the Mbam River, respectively, in relation to the study sites gene pool (Mitchell, Locatelli, Ghobrial, et al., 2015). The forest is characterized by closed-and open-canopy semideciduous and evergreen lowland and submontane rainforest of the Atlantic forest dominated by Fabaceae (Letouzey, 1985). The Ebo Forest also harbors a rich assemblage of other diurnal primates including drills

| Mbam & Djerem National Park
Mbam & Djerem National Park (MDNP) is located >200 km north- main conservation threats at MDNP include illegal bushfires, cattle grazing, poaching, and fishing (Maisels et al., 2000). Data were collected at Ganga in the northeast of the park, situated along the Djerem River ( Figure 2).

| Data collection
Based on the previous studies of habitat suitability and niche differentiation among chimpanzee populations in Cameroon , as well as studies of chimpanzee socioecology from other areas of Africa (e.g., Stumpf, 2011), we predicted that (a) the ecotone would have less rainfall volume and seasonality compared to the rainforest, (b) there would be greater variation in plant species diversity within habitats at the ecotone site, (c) plant species diversity would be higher in the rainforest than ecotone, (d) the availability of fleshy fruits would be higher in the rainforest than the ecotone, (e) there would be greater seasonality in fleshy fruit availability at the ecotone than the rainforest, and (f) the incidence of introduced and secondary forest species would be higher at Bekob due to anthropogenic modification. Thus, we designed our data collection to allow us to examine these variables at a fine scale at each of the three study sites.

| Climate
Rainfall data were collected daily at ~7.00H from January 2010 to

| Potential chimpanzee food resources
We assessed the basal area (BA) of woody plants at each site from trees and lianas ≥10 cm DBH along transects. We further determined the BA of tree species that were potentially important in chimpanzee diets (based on macroscopic fecal analysis). Finally, we determined the stem density of tree species whose fruits were recurrent in chimpanzee diets (based on macroscopic fecal analysis) from trees/lianas ≥10 cm DBH along transects at each site (Potts et al., 2009;Worman & Chapman, 2006). Given the seasonal differences in the fruiting phenology of different species, we compared the frequency of plant species with synchronous and asynchronous fruiting patterns during the wet and/or dry seasons (Potts et al., 2009) across the sites.
Given that chimpanzees feed preferentially on THV species from the Marantaceae and Zingiberaceae families (Tutin et al., 1991), we noted the presence/absence of species from these families in each quadrat.

| Fruit availability
We assessed fruit availability monthly by counting fallen fruits (including partly eaten and rotting fruits) within a 1 m band along transects (Furuichi, Hashimoto, & Tashiro, 2001)  March to November, Ganga: April to October) seasons.

| Data analysis
We assembled rainfall and ecological variables and completed a principal component analysis (PCA) to infer the variables that were most important in distinguishing each of these three sites from one another. The PCA was completed using R3.4.3 (R Core Team, 2017) to infer the environmental and ecological variables that contribute to the differentiation among the habitats available to chimpanzees at Njuma, Bekob, and Ganga. Variables included in the analysis were annual rainfall volume and seasonality, and ecological data from transects including tree stem density, liana stem density, number of tree species, mean tree size (diameter), basal area for all tree species, basal area for tree species that were recurrent in chimpanzee diet at each site, dry and wet season fruit availability, and frequency of THV in the Marantaceae and Zingiberaceae families in quadrats.
We also carried out pairwise comparisons for each variable to further distinguish the sites from one another. We calculated measures of species diversity including Jaccard Classic and Shannon Diversity indices in EstimateS, version 9.1.0 (Colwell, 2016). We used the Jaccard Classic index to assess variation in species composition among transects/habitats (beta diversity) across each site, and Shannon Diversity index for species diversity (alpha diversity) among the sites (Magurran, 2013). We generated species accumulation curves to depict species richness in relation to sampling effort across the three sites (Gotelli & Colwell, 2001). We used nonparametric Kruskal-Wallis one-way analysis of variance (ANOVAs) to test for overall habitat differences among the sites including rainfall, plant species and habitat diversity, and fruit availability. We adjusted significant values for multiple comparisons by using the Bonferroni correction. Mann-Whitney U tests were used to test for intrasite seasonality in fruit availability.

| Main factors distinguishing the three sites
In total, PC1 and PC2 accounted for 68.8% intersite variation ( Figure 3). There was a primary separation between the rainforest and the ecotone along PC1 that accounted for 48.8% of the varia- Distinction between the sites in terms of rainfall was characterized by higher annual rainfall at Njuma (mature rainforest) than at Ganga (ecotone). Bekob (human-modified rainforest) received an intermediate amount of rainfall on average through the study period.
The number of tree species along transects at Njuma and Bekob was higher than at Ganga, while the frequency of THV stems (Marantaceae and Zingiberaceae) was higher in quadrats at Ganga and distinguished the ecotone from the rainforest sites.
The separation among the sites along PC2 was linked mainly to tree stem density (42.0%) and liana density (39.0%). The density of tree stems in transects across Bekob and Njuma (rainforest) was higher than at Ganga (ecotone), while the density of lianas was higher for transects at the ecotone than the rainforest sites. To further ascertain inter-site differences, we carried out pairwise analysis of environmental and ecological variables among and within the sites (Table 1).

| Intersite variation in key factors distinguishing the three sites
Overall, the difference in mean monthly rainfall between the rainforest and ecotone was statistically significant: (Kruskal-Wallis:  Figure S1).
The number of tree families and species was higher at Bekob and Njuma than Ganga (Supporting Information Tables S1-S3).
In terms of measures of species diversity among and within the sites, tree species diversity (alpha diversity) was higher for transects in the rainforest than the ecotone (Figures 4 and 5), while variation in plant species composition among habitats/transects within each site (beta diversity) was higher for the ecotone than the rainforest ( Figure 6). The basal area for trees across the sites was significantly higher at Njuma (mature rainforest) than at Bekob (human-modified rainforest) and Ganga (ecotone).
However, there was no difference among the sites in the basal area nor the stem density of fruiting tree species that were commonly consumed by chimpanzees at each site.
The frequency of THV of the Marantaceae and Zingiberaceae families in quadrats was higher at the ecotone than the rainforest sites. Marantaceae species occurred in 11 quadrats at Bekob (5.5%), The overall density of fruitfall was higher at Ganga than at Njuma (Z = 3.553, p < 0.001) and at Bekob (Z = −2.653, p = 0.024).
There was no difference in fruitfall between the two rainforest sites  Table S4).  F I G U R E 5 Tree species richness (accumulation curves) in relation to sampling effort across 10 botanical transects per site: Bekob (5,482 trees), Njuma (5,017 trees), and Ganga (4,908 trees). The species accumulation curves did not asymptote, suggesting the need for a larger sample size in two rainforest locations and one ecotone habitat at a fine geographic scale using chimpanzee-relevant variables. We compared environmental and ecological variables across Ebo Forest (rainforest) and MDNP (ecotone). We also examined differences between two rainforest sites in Ebo Forest that contrasted in levels of anthropogenic modification (high: Bekob, and low: Njuma) to understand how agriculture might also affect chimpanzee natural resource density and their potential utilization of human-modified landscapes.

| D ISCUSS I ON
The impact of both factors on suitable chimpanzee habitat might be lost by considering ecotones and agricultural lands solely as unsuitable habitats, which is often an underlying assumption of ape ENMs (Junker et al., 2012).
Habitat diversity is a function of environmental conditions including rainfall (Chapman, Olson, & Trumm, 2004;Hohmann et al., 2012), relief (Nkurunungi, Ganas, Robbins, & Stanford, 2004;Proctor, Edwards, Payton, & Nagy, 2007), soil moisture (Marshall et al., 2009), and anthropogenic influence (Arnhem, Dupain, Vercauteren Drubbel, Devos, & Vercauteren, 2007;Chapman, Balcomb, Gillespie, Skorupa, & Struhsaker, 2000). Habitat heterogeneity can be advantageous to frugivorous primates when the different categories are rich in plant species that can reduce the incidence of seasonality in fruit availability. Thus, we next investigated habitat diversity within each of the sites. As predicted, there was greater variation in tree species diversity among habitats at the ecotone compared to the rainforest. Within the rainforest sites, there was greater tree species diversity among habitats at Njuma (mature rainforest) than Bekob F I G U R E 6 Variation in tree species composition (beta diversity) among transects/habitats in Ganga, Bekob, and Njuma using the Jaccard Classic index FIGURE 7 Variation in fruit availability for Bekob and Njuma (rainforest) and Ganga (ecotone) using monthly fruitfall for fruit species that were most represented in chimpanzee diets (based on macroscopic fecal sample analysis) from January 2016 to March 2017 In general, tree species diversity was higher at Bekob and Njuma than at Ganga, consistent with the prediction of higher plant species diversity in rainforest than ecotone habitats. Climatic conditions including the length of the wet season were less variable at the rainforest than at the ecotone, and the wider altitudinal range at Bekob and Njuma supported lowland and submontane plant species, respectively. Environmental conditions at MDNP are similar to GGNP, Nigeria, and additionally, tree species richness was similar at both sites (Fowler, 2006). Sites with high species diversity are generally linked with greater fleshy fruit diversity in chimpanzee diets (Head et al., 2011;Newton-Fisher, 1999;Potts, Watts, & Wrangham, 2011;Tutin et al., 1991;Tweheyo & Lye, 2005;Watts et al., 2012) than sites with lower species diversity (Chancellor, Rundus, & Nyandwi, 2012;Hunt & McGrew, 2002;Stanford & Nkurunungi, 2003). The density of lianas was more important at the ecotone compared to the rainforest sites. Fruits from many lianas including Landolphia spp. and Saba spp. are important food sources for chimpanzees (Moscovice et al., 2007;Piel et al., 2017). The frequency and diversity of Marantaceae and Zingiberaceae species were higher at the ecotone and distinguished the site from the rainforest. These are examples of terrestrial herbs, which are also important in chimpanzee diets especially during periods of fruit scarcity (Boesch, Hohmann, & Marchant, 2002;Tutin et al., 1997;Yamakoshi, 2004).
Other structural differences between rainforest and ecotone, and within rainforest habitats were related to tree size and stem density. Tree size and basal area were larger at Njuma than either Bekob or Ganga. However, tree stem density was higher at Bekob than at Njuma and Ganga. These differences could be attributed to the degree of human modification. Most of the lower altitude vegetation at Bekob was relatively young with smaller tree sizes at various stages of ecological succession due to recent anthropogenic modification. Lower stem density and basal area at Ganga could be attributed to climatic conditions and anthropogenic influence including annual bushfires.
Fruit availability was another key component that distinguished the sites from one another. Counter to our prediction, there were more fleshy fruits from trees and lianas available in the ecotone than the rainforest. Higher fruit availability at the ecotone could be linked to swamps along the main rivers and irradiance. The flood zone of the Djerem River and its tributaries seasonally irrigate swamps that store moisture that could alter the effects of the longer dry seasons at Ganga (Maisels et al., 2000). These swamps could also be very fertile due to alluvial deposits from annual floods, but this was not tested. In addition, more open habitats characteristic of the ecotone may benefit from higher irradiance, providing for greater fruit ripening in upper-and lower-canopy species. Fruit production by trees and lianas at GGNP, an ecotone habitat, was higher than at Salonga, a rainforest habitat in DR Congo (Hohmann et al., 2012).
However, there was more marked seasonality in fruit availability at Ganga, consistent with the prediction of greater seasonality in fruit availability at the ecotone compared to the rainforest. The wet season at Ganga was associated with higher fruit availability with many tree and liana species fruiting synchronously. Conversely, fruit phenology in the dry season at Ganga was low and limited to a few species with asynchronous fruiting patterns. There was less marked seasonality in fruit availability at Bekob and Njuma, where many species produced fruits synchronously during the dry and wet seasons, including several species that fruited asynchronously. At Bekob, E. guineensis, M. cecropioides and other human-introduced and secondary forest plant species produced fruits asynchronously in the wet and dry seasons.
Chimpanzees are fruit specialists and their socioecology is largely influenced by the spatial and temporal distribution of fleshy fruits (Anderson, Nordheim, Boesch, & Moermond, 2002;Mitani, Watts, & Lwanga, 2002). Given the greater seasonality in fleshy fruit availability, the chimpanzee communities at the ecotone may be subjected to greater seasonal shifts in dietary components compared to the rainforest. The consumption of fallback food resources including THV, vertebrates, and invertebrates may be more frequent and consistent at the ecotone than the rainforest sites, especially during the dry season. The consumption of vertebrates and invertebrates by P. t. ellioti at Ngel Nyaki, Nigeria, increased during the dry season. The dry season was marked by lower variety in fleshy fruit availability (Dutton & Chapman, 2015), suggesting that this could be a fallback food strategy for chimpanzees at Ngel Nyaki. Seasonality in fruit availability is a major determinant of chimpanzee grouping patterns as it is linked with other determinants including female cycling (Anderson et al., 2002;Mitani et al., 2002;Wallis, 1995). Low availability and/or patchy distribution of fruits increase ranging and grouping costs (Chapman, Wrangham, & Chapman, 1995;Wrangham et al., 1996), and chimpanzees at the ecotone may be subjected to wider ranging and less group cohesion during the dry season. Low fruit availability was associated with lower rates of gregariousness in P. t. ellioti at GGNP (Hohmann et al., 2012 (Kamgang et al., 2018;Morgan et al., 2011). Despite their proximity, there are structural differences between Bekob and Njuma linked to anthropogenic modification, but chimpanzee densities are high at both sites. Behavioral diversity among and within these populations is linked to habitat variation (Abwe, 2018). Fleshy fruits are the most important dietary component in chimpanzee populations across these sites, but the diversity and seasonality of fruit consumption vary. The rainforest chimpanzee populations consistently consume more fleshy fruit species throughout the year, but the incidence of nonfruit plant parts in their diet is higher in the wet season. The consumption of fleshy fruits is higher for the ecotone chimpanzees during the wet season, while the dry season is associated with a higher reliance on nonfruit plant parts including THV. The incidence of introduced and secondary forest fruit species including E. guineensis and M. cecropioides in chimpanzee diets is higher for the population at Bekob (human-modified rainforest), especially during periods of fleshy fruit scarcity. The consumption of meat including mammals, ants, and termites is higher at the ecotone and is more marked during the dry season (Abwe, 2018). Closed-canopy vegetation and steep relief were linked to nesting site location for the rainforest chimpanzees, while fruit availability was related to chimpanzee nesting site selection at the ecotone. Nest group sizes for the ecotone were smaller than for the rainforest chimpanzees. However, larger nest groups were associated with the wet season (higher fruit availability) at the ecotone.
We speculate that adaptations to local ecological conditions including seasonality in fruit availability may be important in promoting genetic diversity within the subspecies in rainforest and ecotone habitats, as it has already been shown that sex-specific patterns of community and population structure are markedly different across the rainforest and ecotone (Mitchell et al., 2018). These important chimpanzee populations and their habitats are increasingly threatened by hunting and the bushmeat trade, habitat destruction linked to subsistence and agro-industrial plantations, pet trade, as well as climate change (Morgan et al., 2011;Sesink Clee et al., 2015). These observations suggest the need for a more realistic landscape planning approach to conservation planning for the remnant populations of the species.

ACK N OWLED G M ENTS
We thank the Ministry of Scientific Research and Innovation, and Nkumbe.

CO N FLI C T O F I NTE R E S T
The authors declare they have no conflict of interest.

AUTH O R S' CO NTR I B UTI O N S
EEA, BF, AM, RCF, BJM, and MKG designed the study and developed the methodology. EEA, FB, BT, RD, MEK, ET, and RA collected the data. EEA, DMV, MWM, and MKG performed the analysis. EEA and MKG led the writing of the manuscript. All authors contributed critically to the drafts and gave final approval for publication.

DATA ACCE SS I B I LIT Y
Data will be deposited at the Global Biodiversity Information Facility.