Estimating occupancy of Chinese pangolin (Manis pentadactyla) in a protected and non‐protected area of Nepal

Abstract Chinese pangolin is the world's most heavily trafficked small mammal for luxury food and traditional medicine. Although their populations are declining worldwide, it is difficult to monitor their population status because of its rarity and nocturnal behavior. We used site occupancy (presence/absence) sampling of pangolin sign (i.e., active burrows) in a protected (Gaurishankar Conservation Area) and non‐protected area (Ramechhap District) of central Nepal with multiple environmental covariates to understand factors that may influence occupancy of Chinese pangolin. The average Chinese pangolin occupancy and detection probabilities were Ψ^ ± SE = 0.77 ± 0.08; p^ ± SE = 0.27 ± 0.05, respectively. The detection probabilities of Chinese pangolin were higher in PA ( p^ ± SE = 0.33 ± 0.03) than compared to non‐PA ( p^ ± SE = 0.25 ± 0.04). The most important covariates for Chinese pangolin detectability were red soil (97%), food source (97.6%), distance to road (97.9%), and protected area (97%) and with respect to occupancy was elevation (97.9%). We recommended use of remote cameras and potentially GPS collar surveys to further investigate habitat use and site occupancy at regular intervals to provide more reliable conservation assessments.

Adhikari, Sharma, & Thapa, 2015;Katuwal, Parajuli, & Sharma, 2016). Chinese pangolin is listed in the IUCN Red List as Critically Endangered  and on Appendix I by the

Convention on International Trade in Endangered Species of Wild
Fauna and Flora.
In Nepal, conservation action plan for Chinese pangolins suggest implementing management strategies to identify drivers of species occurrence and population dynamics (DNPWC & DoF, 2018).
Further, the government of Nepal desires to understand the effectiveness of their protected areas in maintaining viable wildlife populations (DNPWC & DoF, 2018). A challenge to addressing these needs is that the occurrences of pangolin species in Nepal are not well documented due to their low abundance and nocturnal behavior (Bruce et al., 2017;Khwaja et al., 2019). Therefore, documenting Chinese pangolin occurrence through identification of sign, such as burrows (Katuwal, Sharma, & Parajuli, 2017;Thapa, Khatiwada, Nepal, & Paudel, 2014), could be beneficial for long-term monitoring of pangolin species.
Because Chinese pangolin behavior limits the frequency of direct sightings, information on detection probabilities is a prerequisite for understanding their occurrence and habitat use. We employed an occupancy-based modeling approach (MacKenzie, Nichols, Royle, Bailey, & Haines, 2006;MacKenzie, Nichols, Seamans, & Guitierrez, 2009;Miller et al., 2011;Royle & Dorazio, 2008) to estimate detection probability and occupancy of Chinese pangolins through repeated surveys of their active burrows. Due to government of Nepal's role on species conservation inside protected area (PA), we expected that site occupancy and detection probabilities estimates of Chinese pangolin would be greater in PAs than in non-protected areas (non-PA), in response to reduced human disturbance to habitat and greater food availability.

| Study area
We conducted the study in a PA: Gaurishankar Conservation   Katuwal et al., 2017;Wu, Liu, Ma, Xu, & Chen, 2003). We measured the distance of DS, DW, DL, and DR from the center of each plot by using a measuring tape, but if the distance was greater than 500 m from the center of each plot we measured the distance using a handheld global positioning system (GPS). We recorded the GPS location of the center of each plot and nearest DS, DW, DL, and DR, and the respective distances were estimated by overlaying these points in GoogleEarth. We recorded the slope at the center of each plot using a clinometer. We visually identified the habitat type, soil type, and pesticides as presence and absence and noticed by direct observation. The presence or absence of pesticides uses was identified after consultation with the respective land owners. At each plot, we also established five 10 m × 10 m subplots (four subplots at the corner and one at the center of each plot). From the center of each subplot, we measured the canopy cover and ground cover using 16 mega pixel an Android mobile app (canopy cover using Gap Light Analysis Mobile Application (GLAMA) and ground cover using Canopeo). We used the fisheye lens (present in GLAMA app) of radius 5.6 m to assess the tree Canopy Cover (CaCo) Index (Tichý, 2016), and simultaneously, we also measured the ground cover using the Canopeo app (Patrignani & Ochsner, 2015) from the height of two meters based on downward-facing photographs. Both the apps are yet considered to be the powerful tools which captured the photographs and analyzed through the mobile phone application (Patrignani & Ochsner, 2015;Tichý, 2016). We averaged the percentage of canopy cover and ground cover from five different subplots and used the averaged data for the analysis. We noted the presence or absence of Chinese pangolin food sources (i.e., ant nests, termite mounds) in each plot. Distance to water source, distance to nearest human settlement, distance to nearest road, distance to nearest livestock/sign, canopy cover, ground cover, elevation, and slope were standardized before the analysis. We ran correlation analysis for variable selection and did not include variables with |r| > 0.7 in the same model (Dormann et al., 2012; Figure S1). We performed Moran' I to test spatial autocorrelation of study plots.

| Chinese pangolin occupancy and detection probabilities
In each study area, Chinese pangolin occupancy (Ψ) was estimated We used Akaike's information criterion adjusted for small sample sizes (AICc) to rank all the candidate models and calculate their Akaike weights (Burnham & Anderson, 2002). Models with ΔAICc of 0-2 of the best performing model provide most support (best models), while a ΔAICc of 4-7 has reduced support, and models with values >10 were considered not important (Burnham & Anderson, 2002

| RE SULTS
In the PA and non-PA, the total number of Chinese pangolin burrows recorded was 138 and 105, respectively. The results of Moran's I indicated that sites within the PA and non-PA were not spatially autocorrelation ( Figure S2). The simplest model with constant occupancy and constant detection after pooling the data from PA and non-PA was Ψ ± SE, 0.58 ± 0.05 (Model 1.6: Table 1, Figure 2). The potential differences among the candidate models, where Ψ and p were allowed to vary with different environmental covariates, found (1) Pr H 1 = 01011 = Ψ 1 − P 1 P 2 1 − P 3 P 4 P 5 [for protected area] (2) Pr H 1 = 01011 = Ψ 1 − P 1 P 2 1 − P 3 P 4 P 5 [for non-protected area] Note: The covariates used in the study were habitat types (forest or farmland), soil type (red or brown), tree canopy, ground cover, distance to nearest human settlement (DS), distance to nearest road/foot trail (DR), distance to nearest livestock/sign (DL), food source, elevation, and slope after pooling the data from a protected (PA) and non-protected (non-PA) areas in Nepal. Ψ is the probability a site is occupied by Chinese pangolin, and p is the probability of detecting Chinese pangolin in the jth survey where Ψ (.)p(.) assumes that pangolin presence and detection probability are constant across sites, Ψ is the estimated over all occupancy probability, K is the number of parameters in the model, ΔAICc is the difference in AIC values between each model with the lowest AIC model, and W i is the AIC model weight.

| D ISCUSS I ON
Globally, the population of Chinese pangolin is declining rapidly, and therefore, monitoring the occurrence and habitat associations of this species are crucial. To our knowledge, we provide the first occupancy modeling based on habitat use of Chinese pangolin.
The Chinese pangolin burrow detectability was greater in the PA than in the non-PA. We suggest this is a consequence of reduced human disturbance through management intervention for wildlife conservation in GCA (PA). Healthy forests support the occurrence of Chinese pangolin (Katuwal et al., 2017;Sharma et al., 2020), which more common in PAs than non-PAs in Nepal, and likely supports higher occurrence of prey species. Chinese pangolin's occurrence was greater in PAs reduced disturbances, typically >1,000 m from the human settlements, livestock grazing, and road access (Katuwal et al., 2017;Wu et al., 2003). However, 51% of plots with burrows were nearer to human settlements (<1,000 m).
Many settlements practicing agriculture are sparsely distributed within forested areas. Both forest and agricultural lands support the occurrence of Chinese pangolin (Katuwal et al., 2017;Sharma et al., 2020) and though livestock may not directly disturb Chinese pangolins, livestock guard dogs, and people do pose threats. We suggest the observed response of greater Chinese pangolin occupancy nearer to human settlements is in part a consequence of the dispersion of human settlements within forested areas. The GCA is mainly targeted for the conservation of threatened species, such as Himalayan black bear Ursus thibetanus, but Chinese pangolin may have cobenefitted from such effort. GCA regularly implements participation and awareness programs for local people (NTNC, 2015) which have been demonstrated to benefit species conservation (Bajracharya, Gurung, & Basnet, 2007). Generally, wildlife reserve establishment with efficient management practices can lead to the conservation of species threatened with extinction (Hoffmann et al., 2010). For example, average occupancy of dhole was found high within the reserve of Western Ghats of India (Srivastha, Karanth, Kumar, & Oli, 2019).
Chinese pangolin burrow use was not always detected when present, as the detection probabilities (p) in each study area were F I G U R E 3 Detection probabilities and proportion of sites occupied by Chinese pangolin (Manis pentadactyla) in a protected and non-protected area of central Nepal less than 1.0. The naïve occupancy estimate, which assumes p = 1.0, was found to underestimate the occupancy value of Ψ (.) p(.) by 12.00%-14.00% across the two study areas and demonstrates the need to incorporate detection probability to produce more reliable occupancy estimates. Detection probability can be affected by many factors including local density and weather for amphibians (Bailey, Simons, & Pollock, 2004;Pellet & Schmidt, 2005) or road proximity for sun bears (Linkie et al., 2007).
Chinese pangolin burrows were expected to differ within the study areas because of variation in environmental covariates at each plot. The Chinese pangolin's burrow detection is determined by forest, red soil, and food source in PA while food source along with the farmland in non-PA. Elevation was the important contributing factors for estimating occupancy in non-PA compared to PA because Chinese pangolin were assessed from low to high elevations (about 600-1400 m). This finding corroborates an earlier study by DNPWC and DoF (2018) that recorded burrows at 500-2000 m elevation. Generally, Chinese pangolins prefer slopes <50° (Wu et al., 2003), and in our study, most burrows occurred on 15-22° slopes in the PA. This occurrence is probably due to less disturbances and more abundant fallen logs on these slopes which are important for ants and termites. Though fallen log collection in PAs is prohibited, collection that does occur is generally on less steep slopes. Higher livestock grazing in the non-PA can reduce the moisture content of understory vegetation leading to reduced habitat suitability for detritivores (Bromham, Cardillo, Bennett, & Elgar, 1999), which in turn reduces the prey base of Chinese pangolin. Use of pesticides was also reported from non-PA which might reduce the prey base of Chinese pangolin occurrence. We suggest occurrence of Chinese pangolin burrows in red soil habitats was a consequence of increased availability of food compared to brown soil (SS, Pers. Obs.). Further, the PA had less human disturbance than in non-PA which could TA B L E 2 Detection probabilities of Chinese pangolin (Manis pentadactyla) burrow by habitat types, soil type, cover, distance to nearest human settlement (DS), distance to nearest road (DR), distance to nearest livestock/sign (DL), food source, and slope after pooling the data from a protected (PA) and non-protected (non-PA) areas in Nepal Note: The covariates used in the study were habitat types (forest or farmland), soil type (red or brown), tree canopy, ground cover, distance to nearest human settlement (DS), distance to nearest road/foot trail (DR), distance to nearest livestock/sign (DL), food source elevation, and slope after pooling the data from a protected (PA) and non-protected (non-PA) areas in Nepal.

TA B L E 3
Estimate, standard error, confidence interval, and ∑ W i of covariates in both the PA and non-PA F I G U R E 4 (a) Detection probabilities of Chinese pangolin (Manis pentadactyla) by habitat type in a protected and non-protected area; (b) detection probabilities of Chinese pangolin with soil types in Protected and non-protected area; (c) detection probabilities of Chinese pangolin with cover in Protected and non-protected area; and (d) detection probabilities of Chinese pangolin with DW, food source, and slope in protected and non-protected area result in greater food availability in the PA. However, Chinese pangolin have been detected in farmlands in non-PAs, including near human disturbance (Katuwal et al., 2017).
From our combined study areas (183 plots), we predicted that a total of 250 plots would need to be surveyed to obtain Ψ estimates with 0.05 s . This could be achieved by increasing the number of plots within each study area and reducing the number of search occasions (e.g., 3) per plot. Overall, Ψ (.) models received greater support in the model selection procedure, as indicated by higher AIC weightings.
As Chinese pangolins are nocturnal and elusive (DNPWC & DoF, 2018), occupancy estimates of pangolins based on direct observation are difficult (Willcox et al., 2019). We found that using an indirect sign survey was suitable for estimating the occupancy of Chinese pangolin to address the problem of inadequate direct sightings. To validating the indirect sign and survey especially for the study of mammals, occupancy models are considered a powerful approach (Yarnell et al., 2014). However, our survey did not meet the assumption of abundance models of occupancy (Conroy, Runge, Barjer, & Schofield, 2008;Nichols, Hines, MacKenzie, Seamans, & Guiterrez, 2007;Royle & Nichols, 2003).
We recommend further applications of occupancy modeling using alternative data types (e.g., remote cameras) to improve reliability of estimates of Chinese pangolin, particularly detectability. In addition, refined data and associated models would facilitate the identification of spatially explicit priority areas to improve conservation of the Critically Endangered Chinese pangolin in both PAs and non-PAs.

ACK N OWLED G M ENTS
We are grateful to the Rufford Small Grant Foundation (Grant No. 26508-2) and Fresno Chaffee Zoo for funding this research. We thank the Department of National Parks and Wildlife Conservation, Department of Forest, Divisional Forest Office of Ramechhap, Gaurishankar Conservation Area, Rajendra Karki, and Chandeshwor Pattel for assisting in our project. We thank Eric Wikramanayake, Kumar Sapkota, Ugan Manandar, and Damber Bista for their support and guidance.

CO N FLI C T O F I NTE R E S T
None.

AUTH O R S CO NTR I B UTI O N S
SS, HPS, and HBK designed the study. SS and CC carried out the field survey. SS, HPS, and HBK did data analysis. SS, HPS, CC, HBK, and JLB wrote and finalized the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
All the relevant data used in this study will be archived in Dryad after the acceptance of the manuscript. https://doi.org/10.5061/dryad.