Habitat requirements of the Himalayan red panda (Ailurus fulgens) and threat analysis in Jigme Dorji National Park, Bhutan

Abstract Understanding the influence of anthropogenic disturbances on species’ habitat use and distribution is critical to conservation managers in planning effective conservation strategies and mitigating the impact of development. Few studies have focused on the Himalayan red panda (Ailurus fulgens) in Bhutan. This study aimed to assess the habitat requirements and threats to this endangered species in the Khamaed subdistrict of the Jigme Dorji National Park, Bhutan. We employed a transect walk and plot‐sampling survey design across two seasons, that is, winter and spring. In total, we surveyed 84 × 50 m radius circular plots along 51 km of existing trails within a 25.4 km2 study area. At 500 m intervals, we established plots at random distances and direction from the trail. We recorded direct sightings (n = 2) and indirect signs (n = 14), such as droppings and footprints as evidence of red panda presence within an altitudinal range of 2,414–3,618 m. We also noted 21 tree and 12 understory species within plots with red panda evidence; the dominant tree species was the Himalayan hemlock (Tsuga dumosa) and the Asian barberry (Berberis asiatica) as an understory species. Red panda presence showed a significant positive association with distance to water sources and fir forests. Plant disturbance and infrastructure, such as power transmission lines, were identified as prominent anthropogenic threats in the study area. Based on our findings, we recommend the development and implementation of local forest management plans, livestock intensification programs, and strict application of environmental impact assessment regulations to promote the conservation of the red panda in the region.


| INTRODUC TI ON
Information and knowledge on species' distribution are vital to understand their presence and infer habitat and ecological requirements (Elith & Leathwick, 2009;Noon, Bailey, Sisk, & McKelvey, 2012).
Local small-scale distribution surveys can shed light on species habitat preferences, existing threats, and responses to management interventions (Lahoz-Monfort, Guillera-Arroita, Milner-Gulland, Young, & Nicholson, 2010). Such studies can help conservation managers plan and revisit species conservation policies and guide informed management decisions (Rayan & Linkie, 2015).
The Red List of International Union for Conservation of Nature (IUCN) lists the red panda ( Figure 1) as an endangered species (Glatston, Wei, Than, & Sherpa, 2015). Recently, Hu et al. (2020) recognized the red panda of Nepal, Bhutan, northern India, northern Myanmar, Tibet and western Yunnan Province of China as the Himalayan red panda (Ailurus fulgens), and its relative in Yunnan and Sichuan provinces of China as the Chinese red panda (Ailurus styani).
This new taxonomic identification has underpinned the need of more studies to secure the survival of these two red panda species in the wild.
Globally, red pandas are reported from 49 protected areas (PAs) in China; 11 PAs in India; 10 PAs in Nepal; and 3 PAs in Myanmar (Thapa, Wu, et al., 2018). In Bhutan, red pandas are reported in eight PAs and six biological corridors (Dorji, Rajaratnam, & Vernes, 2012) and red panda areas in Bhutan account for about 43.5% of the predicted red panda habitat across range countries (Thapa, Wu, et al., 2018).
However, habitat requirements for the red panda vary across different landscapes. In Phurmsingla National Park, habitats close to water sources were, for example, not a significant predictor of red panda presence (Dendup, Cheng, Lham, & Tenzin, 2016).
Threats to red panda survival are greatest in Bhutan, India, and Nepal than other range countries . Habitat loss, fragmentation, and degradation are some of the major threats to the red panda populations in these countries (Bista et al., 2017;Glatston et al., 2015;Pradhan et al., 2001;Wei et al., 1999;Yonzon & Hunter, 1991). The harvesting of forest resources and infrastructure development are some of the main drivers of red panda habitat destruction and fragmentation (Panthi, Khanal, Acharya, Aryal, & Srivathsa, 2017;Sharma & Belant, 2010;Sharma, Swenson, & Belant, 2014;Williams, 2003). Hickman, Roberts, and Larson (1993) reported habitat loss as a main cause for population decline, especially affecting endangered species that are sensitive to changes in their environment.
Red pandas are known to avoid areas close to human settlements and areas disturbed by livestock (Acharya et al., 2018;Dendup et al., 2016;Sharma et al., 2014;Wei et al., 2000). As habitat specialist, red pandas prefer less disturbed habitats; however, their responses to habitat disturbances may vary across locations (Acharya et al., 2018). Field observations made in some parts of Nepal have revealed extensive spatial overlap between the red panda and the livestock . During winter months when ground vegetation is scarce, livestock are seen feeding on bamboo (Dendup et al., 2016;Sharma et al., 2014). As a generalist species, livestock adapt to feed on any available resources. Habitat-generalist species in a community are reported to overexploit the environment and occupy habitats unexploited by habitat specialists at a larger spatial scale (Morris, 1996).
One of the major goals of conservation biology is to document environmental and anthropogenic factors that influence species' distribution (Robinson, 2006). Understanding species' habitat requirements and disturbances contributing to habitat destruction is very important to conserve endangered species (Hunt, Bayne, & Hache, 2017) like the red panda. Despite Jigme Dorji National Park (JDNP) harboring key red panda habitats in Bhutan (Dorji, Vernes, & Rajaratnam, 2011), very few studies have focused on documenting the conservation threats to this species in this region. This study hence aimed to fill this gap and assess the conservation status of the red panda in a critical but poorly studied area of the JDNP in Bhutan.
In addition, this study sought to identify the threat factors affecting red panda habitat use.  Figure 2b) is one of these subdistricts located in Gasa. The topography, climate, and vegetation structure within the potential red panda habitat (2000-4,500 m) is rather homogeneous across the entire JDNP (Thinley et al., 2015). This is why, we selected Khamaed subdistrict as representative of the entire national park.

| Field survey design
We carried out the 1st phase of surveys between November 2017 and March 2018 (winter and early spring), and 2nd phase in May 2019 (spring). In the 1st phase, we surveyed 43 plots along eight trails (29 km), and in the 2nd phase, we surveyed 41 plots along six trails (21.6 km). During both surveys, we established plots and collected evidence of red panda presence, and recorded vegetation and anthropogenic disturbances.
Following Dendup et al. (2016), circular plots of 50 m radius were used to study red panda presence signs (tracks and dungs). At 500 m intervals along each trail, we established a plot in a random direction and at a random distance of 0-1,000 m (Dendup et al., 2016). Within the plot, each of the team members moved in different cardinal directions and explored for red panda signs moving toward the center of the plot and repeated this till the entire plot was scanned. If red panda signs were found in the location, we shifted the plot center to the location of the animal sign to record vegetation and anthropogenic disturbances. However, plot center was not shifted if no signs were recorded (Dendup et al., 2016). We avoided any overlap between plots.
Vegetation-related data were recorded using tree quadrats (10 × 10 m) which were superimposed on the center of each 50 m radius plot. Understory quadrats (4 × 4 m) were superimposed at F I G U R E 2 Location of the study site: (a) protected area network of Bhutan, with dark green areas representing protected areas and light green areas biological corridors, (b) Jigme Dorji National Park with its 14 subdistricts. Khamaed highlighted in yellow is one of the 14 subdistricts where the study was carried out (c) Study area with sampling plots within Khamaed subdistrict. Red and green features show the sampling plots with red panda presence and absence signs, respectively the center of each tree quadrats, and the ground cover quadrats (1 × 1 m) were superimposed on the center of each understorey quadrat (Dendup et al., 2016). Vegetation data were recorded following Schemnitz (1980) and Dendup et al. (2016 , Table 1).
We also recorded data on disturbance variables (Table 1), including the shortest distance from the plot center to the nearest water source defined as streams or ponds (Bista et al., 2019;Dendup et al., 2016). We also recorded the number of fallen logs and stumps > 30 cm DBH (Dendup et al., 2016;Dorji et al., 2012;Sharma et al., 2014;Zhou et al., 2013).
To assess anthropogenic disturbances in red panda habitat, we recorded the presence-absence signs of plant disturbance (e.g., harvesting, lopping, girdling), livestock (sighting, droppings, hoof prints), infrastructure (power transmission lines, telecom tower, houses, roads), and poaching signs (snares). To assess natural disturbances, we recorded the presence-absence of landslides, dead bamboo, naturally fallen logs/trees (windthrow, snow damage), and carnivore signs (sighting, scats, scrapes marks, pug marks). The same team of trained foresters conducted the surveys in both phases to maintain consistency in data collection.

| Data analysis
To calculate tree species diversity in different forest types (mixed conifer forest (MCF), cool broadleaf forest (CBL) and fir forest), following Margalef (1968), we calculated the Shannon-Wiener diversity index (H') using the following formula: where H' = the Shannon diversity index; ni = number of individuals of the species; and N = number of individuals of species.
To evaluate tree species' dominance, following Phillips (1959), we calculated the importance value index (IVI) using the following formula: We carried out all statistical analysis in R v. 3.5.1 (R Development Core Team, 2018). Prior to performing any statistical modeling, we examined collinearity between the variables (Table 1, except Geographical Location) based on the variance inflation factor (VIF) (2) IVI = relative density + relative frequency + relative basal area  (Naimi, 2015). The variables with VIF > 10 were regarded as highly correlated and omitted from further analysis (Montgomery, Peck, & Vining, 2012;Zuur, Leno, Walker, Saveliev, & Smith, 2009). Since all the habitat covariates had VIF < 2, we retained all the variables (Table 2).
We performed logistic regression using a binomial distribution to model red panda habitat use as a function of habitat and disturbance covariates (Tollington et al., 2015). To investigate the best-fit model, we performed multi-model inference using a dredge function in MuMIN package (Barton, 2016). We then examined the fit of candidate models by selecting the lowest Akaike's information criterion corrected (AICc) for small sample sizes, and final model sets were restricted to ∆AICc < 1 for habitat use variables and ∆AICc < 2 for disturbance variables before model averaging (Bloker et al., 2009;Burnham & Anderson, 2002;Harrison et al., 2018).

| Habitat use
Overall, evidence of red pandas was recorded in 16 of the 84 surveyed plots. In the 1st phase, seven of the 43 plots and, in the 2nd phase, nine of the 41 plots surveyed showed red panda signs.
Red panda presence within KSD was detected through 14 indirect signs in the Gayza, Dompangchung, Zomina, and Chutegompa areas, and two direct sightings in the Jabisa and Phuntshogang areas.
Red panda evidence was observed between 2,414 and 3,618 m with an average elevation of 2,750.8 m (SD = 368.45), and the majority of the evidence was observed < 3,000 m (n = 13).
Interestingly, four plots with red panda signs did not have any bamboo cover. Among plots with signs of red panda, seven plots had 10%-20%, two plots had 30%-60%, and three plots had 90%-100% bamboo cover. Contrary to expectation, bamboo cover was not included in the best-fit model (Table 3).
Other habitat covariates such as MCF, CBL, canopy cover, and the number of understory species did not significantly influence red panda habitat use (Table 4).
Red panda signs and sightings were recorded between 10 and 1,000 m distance to the nearest water sources with an average distance of 241.9 m (SD = 334.65). The majority of red panda records (n = 9) were found within 100 m from water sources. followed by lodh tree Symplocos sp. (n = 10) and Aconogonum molle (n = 6). Six ground cover species were recorded in plots with red panda signs, but most of the plots did not have any ground cover (n = 9).

| Threats
We recorded seven types of disturbances including four natural (dead bamboo, landslides, presence of carnivores, and naturally fallen logs) and three anthropogenic disturbances (plant disturbances, livestock, and infrastructures). Neither dead bamboo nor landslides were retained in the best-fit model examining the relationship between red panda habitat use and disturbance variables (Table 5). Only plant disturbance and infrastructure significantly influenced red panda habitat use (Table 6). Although red pandas avoided areas with plant disturbance, there was a significant positive association between red panda habitat use and infrastructure presence.  (Dorji et al., 2012).

| D ISCUSS I ON
While in another separate study in the PNP, red pandas were recorded in CBL and MCF between 2,860 and 3,597 m (Dendup et al., 2016). In the present study, red panda habitat use was significantly associated with fir forests. These findings may be related to fir trees providing better nesting sites with tree cavities and being evergreen, hence providing better cover and safety (Pradhan et al., 2001;Wei et al., 1999;Williams, 2003). However, Dendup, Lham, Wangchuk, and Tshering (2018)  Bolded are the coefficient estimates where the confidence intervals do not cross zero and the variables that are hence significant.
forest, indicating that red pandas use both higher and lower elevations during the warmer month and lower elevations in the winter.
Bamboo cover did not emerge as a predictor of red panda presence in this study, which contradicts findings from previous studies (Chakraborty et al., 2015;Dendup et al., 2018;Dorji et al., 2011;Johnson, Schaller, & Jinchu, 1988;Pradhan et al., 2001;Sharma et al., 2014). This could be attributed to the availability of other food sources, especially berries. During the 2nd phase of the study, most of the study site had a climber Hedera nepalensis in fruiting stage.
During our study, we indeed observed traces of seeds of Hedera nepalensis in red panda feces.
Fallen logs, tree stumps, and distance to the nearest water source have previously been reported to positively influence red panda habitat use (Dendup et al., 2018;Dorji et al., 2012;Zhou et al., 2013).
However, in the present study, except for the distance to the water source, these other covariates were nonsignificant. Proximity to water sources is evidently a critical predictor of red panda presence, further confirming findings from previous studies (Bista et al., 2017;Dendup et al., 2018;Dorji et al., 2012;Pradhan et al., 2001). Pradhan et al. (2001) reported that due to low water content in the bamboo leaves, red pandas frequently need to drink water. Bista et al. (2017) further reported that the preference for proximity to water sources may also be related to energy conservation and avoidance of predator, hence avoiding longer travel distances to access water.
Red panda presence was low in plots with plant disturbance.
Red pandas generally avoid areas with plant disturbance (Acharya et al., 2018;Dendup et al., 2016). In the study site, plant disturbance was attributed to timber collection by the local people and construc-   (Wangchuk, 2002). There are about 788 cattle in KSD (Dzongkhag Administration Gasa, 2018). The majority of these domestic animals are unproductive breeds and are free ranging, and they share the red panda habitat for grazing. The competition from cattle indeed has a significant negative impact on red panda habitat use (Dendup et al., 2016;Dorji et al., 2011;Panthi, Wang, Sun, & Thapa, 2019;Sharma et al., 2014;Zhang et al., 2017).
Oddly, our study revealed that red panda presence was signifi-   (Freitas, Gonclaves, Kindel, & Teixeira, 2017). There is an urgent need to recruit qualified researchers and conservation practitioners, as well as species-level experts, especially when it comes to endangered species, to generate higher quality EIAs (Freitas et al., 2017). Globally, infrastructure development has been recognized as one of the major factors responsible for habitat loss and fragmentation, declines in wildlife populations and the introduction of invasive species (Trombulak & Frissell, 2000). Buskirk (2009)  (2,322 L in first lactation adjusted to 305 days) than local breeds (529 liters) and are not suitable for rugged terrains (Phangchung et al., n.d.). These cattle are usually grazed in an enclosed pasture or even if they are freely grazed, they graze along the roadsides and do not venture into high forests thereby avoiding competition with the red pandas.

ACK N OWLED G M ENTS
The first phase of this study was funded by the Nagao Natural Environment Foundation, Japan, and the second phase was funded by the School of Anthropology and Conservation, University of Kent, UK. We would like to thank Karma Jangchuk, Sangay Tenzin, Kado Dukpa, Bal Krishna Koirala, Kencho, and Kado for helping with data collection and field logistics. Special thanks to Simon Tollington for his guidance during the data analysis. Mr. Sonam Dorji is well acknowledged for allowing us to use his picture of red panda. We would also like to thank park management, Jigme Dorji National Park, for allowing us to carry out our study in the park.
We are also extremely grateful to the three anonymous reviewers for their invaluable comments and suggestions in improving this manuscript.

CO N FLI C T O F I NTE R E S T
We declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.

DATA AVA I L A B I L I T Y S TAT E M E N T
Since red panda is an endangered species, we are not providing the data under public domain for some conservation reasons.