Mammals of the Bhagirathi basin, Western Himalaya: understanding distribution along spatial gradients of habitats and disturbances

Abstract Understanding the distribution of wildlife species and their response to diverse anthropogenic pressures is important for conservation planning and management of wildlife space in human-dominated landscapes. Assessments of anthropogenic impacts on mammals of the Indian Himalayan Region have mostly been limited to locations inside protected areas. We studied the occurrence of mammals in an unexplored landscape, the 7,586 km2 Bhagirathi basin, at an altitude of 500–5,200 m. The basin encompasses wilderness areas of various habitat types and protection status that are exposed to a range of anthropogenic pressures. Camera trapping at 209 locations during October 2015–September 2017 confirmed the occurrence of 39 species of mammals, nine of which are categorized as threatened (four Vulnerable, five Endangered) and four as Near Threatened on the IUCN Red List. We recorded five mammal species that were hitherto undocumented in Uttarakhand State: the argali Ovis ammon, Tibetan sand fox Vulpes ferrilata, woolly hare Lepus oiostolus, Eurasian lynx Lynx lynx and woolly flying squirrel Eupetaurus cinereus. In addition, we recorded two Endangered species, the dhole Cuon alpinus and tiger Panthera tigris. Threatened species such as the sambar Rusa unicolor, common leopard Panthera pardus and Asiatic black bear Ursus thibetanus occur in a wide variety of habitats despite anthropogenic disturbance. We recorded the snow leopard Panthera uncia in areas with high livestock density but temporally segregated from human activities. The musk deer Moschus spp. and Himalayan brown bear Ursus arctos isabellinus were recorded in subalpine habitats and appeared to be less affected by human and livestock presence. Our findings highlight the potential of the Bhagirathi basin as a stronghold for conservation of several threatened and rare mammal species.


Introduction
T he mammals of the Indian Himalayan Region are exceptionally diverse, and many are endemic to the region (Schaller, ), but they are threatened by persecution (Mishra, ; Naha et al., ), habitat loss and degradation (Namgail et al., ; Kittur et al., ), and competition with livestock (Bhatnagar, ; Mishra et al., ; Bhattacharya et al., ). The type and intensity of these threats often vary across the seasons (Bhattacharya & Sathyakumar, ).
Anthropogenic disturbance is dynamic, and responses of wildlife are likely to be influenced by human density and location at a given time, and the duration of human activity (Rogala et al., ; Carter et al., ). Understanding these dynamics facilitates conservation planning and illuminates the processes governing wildlife behaviour in humandominated landscapes (Hojnowski, ). The , km  Bhagirathi basin in Uttarakhand, India, is recognized for its ecological, socio-cultural and conservation significance (Rajvanshi et al., ). This landscape encompasses wilderness areas of various habitat types and protection status that are exposed to a range of anthropogenic pressures. The only protected area in the Bhagirathi basin is the , km  Gangotri National Park, which provides protection to species of the Trans-Himalaya and Greater Himalaya.
Anthropogenic activities in the Bhagirathi basin include seasonal grazing (May-October) above , m altitude. There are local livestock herders and pastoral migrant communities such as Gujjar (outside the protected area) and Gaddis (in the Trans-Himalayan part of Gangotri National Park), with large herds of livestock (c. , sheep, goats and mules) grazing the alpine pastures of the National Park (Chandola et al., ) for  months annually (June-September). Anthropogenic activities in the lower and mid-altitude forests (-, m) of the Bhagirathi basin include livestock grazing, extraction of non-timber forest products and collection of fuelwood, activities that have been conducted for centuries (Awasthi et al., ; Rana et al., ). Additionally, tourism, mountaineering, and pilgrimage attract numerous visitors during April-November. As the northern boundary of the Bhagirathi basin also forms the international border with the Tibet region of China, patrol camps and small settlements of the Indo-Tibetan Border Police and other security agencies are present in the area. Recurrent deliberate burning of hill slopes to stimulate regrowth of grasses has altered the vegetation structure and composition of the Bhagirathi basin (Mehta, ; Gurumni, ).
The conservation importance of the Bhagirathi landscape, and that of its mammals in particular, has been described in an environmental impact assessment of hydropower projects (Rajvanshi et al., ) and several short-term surveys (Uniyal & Ramesh, ; Bhardwaj & Uniyal, ). These studies are based on observations of species or evidence encountered during trail and ridge walks. However, the distribution of mammals in this area has not yet been assessed with robust scientific methods such as camera trapping and genetic sampling.
Here, we describe () the occurrence of mammal species in the Bhagirathi basin, () the occurrence of threatened mammals in relation to human activities, and () conservation prospects for threatened mammals under the existing protection measures in the Bhagirathi landscape.

Study area
The , km  Bhagirathi basin in Uttarakhand State, India, is drained by the Bhagirathi river (c.  km) and its tributaries. The study area encompasses altitudes of -, m (Fig. ). The major habitat types of the basin are () subtropical deciduous forest (-, m) characterized by broadleaved and needle-leaved species such as Pinus roxburghii, () temperate forest (,-, m) with montane broadleaved and conifer species such as Quercus semecarpifolia, Quercus floribunda, Abies pindrow, Cedrus deodara, and Pinus wallichiana, () high altitude alpine and subalpine vegetation (,-, m) with Rhododendron spp., Betula utilis and alpine herb and forb species, and () Trans-Himalayan landscape (,-, m) with alpine dessert steppe plants such as Eurotia sp., Caragana sp., Lonicera sp. and Rhamnus sp. Summer (or monsoon, April-September) and winter (November-February) are more pronounced than the short autumn (October) and spring (March-April) seasons. The economy of the region is largely dependent on agriculture. There are  villages in the Bhagirathi basin, mostly below , m.

Methods
After a -month reconnaissance survey (July-September ), we conducted a camera-trap study at  locations, using  Cuddeback C (Cuddeback, De Pere, USA) camera traps, during October -September . We positioned camera traps along an elevational gradient (-, m) representing various habitats. At each site, camera traps were deployed in locations likely to be used by animals, affixed to trees or, in alpine meadows, to a pile of stones, at a height of c. - cm above the ground (Sathyakumar et al., ; Bashir et al., ). To survey evenly across the various habitats, we divided the basin into  grid cells of  km  each ( ×  km), which corresponds to the average home range of the largest mammal in the area, the Himalayan brown bear Ursus arctos isabellinus. We subdivided these cells into  ×  km cells and deployed camera traps in - of these smaller cells within each  km  cell (Fig. , Table ). In the fragmented forests of the lower areas,  camera traps were stolen, which prevented adequate coverage in all grid cells (Fig. , Table ).
We examined all camera-trap photographs of large and medium-sized mammals (except families Muridae and order Chiroptera) and identified species with the help of Prater () and Menon (). We assessed the elevational range (minimum and maximum elevation of occurrence) and habitat types used by each identified species based on camera-trap locations where they were captured. We calculated photo-capture rates as the number of captures per  trap days, following Bashir et al. (; Table ), and camera trapping days as the number of -hour periods from placement of the camera until the memory card was full or the camera was retrieved. Multiple captures of the same species within  hour at a camera site were excluded from trap rate calculation (Sathyakumar et al., ). We used photocapture rates (mean ± SE) to assess the relative abundance of each species and anthropogenic disturbances (people, dogs and livestock).
We examined the effect of habitat and human disturbance with generalized linear mixed models, using the glmmTMB package (Magnusson et al., ) in R .. (R Core Team, ). For the generalized linear mixed models we used cameras ( locations, , trap nights) that were active in both seasons or either summers (April-September ) or and winters (November-February - and ). Some of the smaller grid cells had more than one camera location. We therefore tested for spatial autocorrelation among sampled locations, using the weighted correlation coefficient of Moran (Moran's I) in ArcGIS . (Esri,  Redlands, USA). We used data from repeated sampling at the same sites (summer and winter) and incorporated site as a random effect variable. We used captures of species as the response variable and number of trap days (logtransformed) as offset, to account for variation in the trapping effort between sites. Habitat features (elevation, ruggedness, slope) and anthropogenic pressures (capture rate of humans, dogs and livestock) were used as fixed predictor variables ( Table ). We acquired data on elevation from Shuttle Radar Topography Mission (Jarvis et al., ), at a resolution of  ×  km pixels. Slope and ruggedness were calculated from the elevation layer in ArcGIS. For the brown bear, we examined only summer data as they hibernate in winter, using a Poisson-distributed generalized linear model. We tested for the presence of over dispersion in the dataset and selected the appropriate error distribution (i.e. Poisson, negative binomial). We also evaluated the data for zero-inflation. We used Akaike's information criterion adjusted for sample size (AICc) to rank models, and we considered the best supported models to be those with ΔAICc values ,  units (Arnold, ). To examine any multicollinearity between predictor variables, we performed Pearson correlation tests, correlated variables (Pearson correlation coefficient . .) were not used in the same model. We decided on the suitable habitat for each species based on the elevation range and habitats in which camera traps recorded them ( Table ). For example, we used greater Himalayan and Trans-Himalayan habitats (,-, m) for snow leopard analysis, and temperate, subalpine and alpine habitats (,-, m) for musk deer. Based on the time stamp on the camera trap images, we assessed temporal overlap between each threatened species and occurrence of human disturbance (records of people, livestock and domestic dogs) using the kernel density method (Ridout & Linkie, ) in R.

Results
The total number of camera-trap days was ,, with a mean of  trap days per camera. We recorded  species of mammals belonging to  families in five orders ( Table ). Carnivora was the most diverse order with  species, followed by Artiodactyla (), Rodentia (), Lagomorpha () and Primates (). Of the  species recorded, nine are categorized as threatened (four Vulnerable, five Endangered), four as Near Threatened and  as Least Concern on the IUCN Red List (IUCN, ). We recorded five mammal species (Fig. ) that were hitherto not known to be present in Uttarakhand State: the argali Ovis ammon, Tibetan sand fox Vulpes ferrilata, woolly hare Lepus oiostolus, Eurasian lynx Lynx lynx, and woolly flying squirrel Eupetaurus cinereus. Argali, sand fox, woolly hare and lynx were recorded in the Trans-Himalayan landscape (,-, m) of Nelong valley in Gangotri National Park, which is a typical cold desert characterized by rock fields with sparse vegetation (Fig. ). The woolly hare was captured widely ( locations) and regularly ( captures) throughout the survey, whereas the sand fox was captured on only three occasions, Argali on four occasions and Eurasian lynx on one occasion. The woolly flying squirrel was captured once, during the sampling period in temperate habitat at , m in Harsil valley (Pal et al., a). Apart from these new records, we also captured photographs of the dhole Cuon alpinus and tiger Panthera tigris. A tiger was photographed only once, in February  (at , m altitude) in subalpine broadleaved forest dominated by Quercus semecarpifolia.
Six threatened species were captured regularly throughout the survey: the Himalayan brown bear, Asiatic black bear, snow leopard, common leopard, musk deer and sambar. Records of Himalayan brown bears (n = ,  locations), musk deer (n = ,  locations) and snow leopards (n = ,  locations) were confined to elevations . , m in the Trans-Himalayan areas, alpine and subalpine forests. Asiatic black bears (n = ,  locations) were distributed throughout the study area except for dry Trans-Himalayan scrub, and sambar (n = ,  locations) were captured in all forest types up to , m. The common leopard was the most frequently captured (n = ,  locations) large carnivore, in the subtropical forest and temperate habitats (-, m).
Seasonal comparison of capture rates of people, livestock and dogs showed that during the winter there was a comparatively low presence of people and associated activities in both protected and non-protected areas. During summer, photo-capture rates of people inside the National Park (mean . ± SE .) were lower than outside (. ± .) but captures of livestock (. ± .) and dogs (. ± .) were higher inside the National Park than outside (livestock . ± ., dogs . ± .). Anthropogenic disturbance in high altitude Trans-Himalaya was higher in summer (people . ± ., livestock . ± .) than winter (people . ± ., no livestock). We observed a similar seasonality in alpine and subalpine habitats, with disturbance being higher in summer (people . ± ., livestock . ± .) than in winter (people . ± ., no livestock). In temperate habitats, disturbance was high in summer (people . ± ., livestock . ± .) but only slightly less in winter (people . ± ., livestock recorded in only  of  locations). In subtropical habitats, mean photo-capture rates of people were similar in summer (. ± .) and winter ( ± .).
The Pearson test showed a significant correlation between livestock and dogs (r = .) and between ruggedness and slope (r = .), and therefore these variables were not used together in the models. Spatial autocorrelation ( Table ) was insignificant for all species across sites (all Z-scores were between −. and .). The most supported model (Tables  & ) showed that sambar, common leopard and Asiatic black bear occurred in areas with high levels of human disturbance. Asiatic black bears had lower capture rates in winter ( Supplementary Fig. ), when they hibernate (Sathyakumar et al., ). The sambar and Asiatic black bear avoided steep slopes and rugged areas, respectively ( Supplementary Fig. ). Musk deer and brown bear were found in narrow elevation zones of subalpine habitats. In winter, musk deer capture rates declined with increasing elevation (Supplementary Fig. ), which could be associated with snowfall at high altitudes. Snow leopards were recorded in areas with high human presence and showed a negative response to livestock (Supplementary Fig. ). They occurred at an altitude of ,-, m, with lower capture rates at higher elevations ( Supplementary Fig. ), and were rarely detected in the high elevation plateau habitat of the National Park. Snow leopard capture rates were higher in winter, when there was less disturbance by livestock and people ( Supplementary Fig. ).
Human presence in the Bhagirathi basin was comparatively low in winter. We therefore analysed temporal overlap between each of the six regularly detected threatened species and people only for the summer. Temporal overlap during summer was highest between the Himalayan brown bear and livestock and domestic dogs (Fig. ), followed by the Asiatic black bear and livestock and domestic dogs (Fig. ). There was also considerable overlap between musk deer and domestic dogs (Fig. ). The snow leopard, common leopard and sambar showed minimal temporal overlap with any anthropogenic disturbance (Fig. ).

Discussion
Human encroachment on wildlife habitats has caused the decline of large mammals globally (Ceballos & Ehrlich, ). Some species persist in human-dominated landscapes by changing their behaviour in response to human presence (Frid & Dill, ). The Bhagirathi basin is one such landscape, where large mammals, including some threatened species, occur across a gradient of habitat types and human disturbances.
Our model did not show a significant influence of anthropogenic pressures on the Himalayan brown bear, but the high capture rates of livestock and high temporal overlap with livestock suggest there could be a high probability of livestock depredation by the species, which could lead to retaliatory killings. Such incidents are relatively common in Himachal Pradesh (Rathore, ; Sathyakumar et al., ). Similarly, temporal overlap with domestic dogs in the summer can negatively affect musk deer, which occur in subalpine habitat. Studies in Mongolia (Young et al., ), Lahual Spiti (Pal, ) and other areas (Home et al., ) describe the need for the exclusion of feral dogs from critical wildlife habitats. Musk deer are also vulnerable to poaching, but we could not quantify this and hence did not include poaching in our analysis. Camera-trap photographs of people with guns in subalpine and temperate forests outside the protected area (N = ) and presence of snares that we found during monitoring of camera traps (N = ) in subalpine habitats showed that hunting occurs in this region. Frequent removal of individuals can increase the chance of local extirpation of the remaining subpopulations. Snow leopard habitats, which generally consist of alpine areas, are under pressure from livestock grazing in the summer even inside the National Park.
In comparison with Gangotri National Park and high altitude areas (. , m), lower, non-protected areas of Bhagirathi basin are more fragmented and more densely populated by people. Three species were found in areas of high human activity: the common leopard, sambar and Asiatic black bear. The leopard and Asiatic black bear are hunted and their body parts traded (Sathyakumar & Choudhury, ; Raza et al., ). The populations of all three species are declining as a result of habitat Because of its rugged terrain and inaccessibility, the Bhagirathi basin contains some areas with little or no direct human disturbance, which may act as refugia for some threatened and rare species. We recorded four typical Trans-Himalayan mammals (argali, Tibetan sand fox, woolly hare and Eurasian lynx) and the woolly flying squirrel in the Bhagirathi basin. The Near Threatened argali has declined significantly (Harris & Reading, ). The Tibetan sand fox, although categorized as Least Concern, occurs in low densities (Harris, ). The woolly hare has been assessed as Endangered in India (Molur et al., ). The population of the Eurasian lynx is declining and the species is believed to be close to extinction in India (Breitenmoser et al., ). The population of the woolly flying squirrel is believed to have declined by % during the last decade, largely because of deforestation and grazing pressure (Zahler, ). In addition to these threatened species, we recorded two Endangered large carnivores, the dhole and tiger, which were hitherto not known from the area. These new records and the high mammal diversity in the Bhagirathi basin are a result of a wide range of habitats, including many areas with low anthropogenic pressures. In the Indian Himalayan Region, information on the distribution of the dhole (Bashir et al., ; Johnsingh & Acharya, ; Pal et al., b) and tiger (Gopi et al., ; Bhattacharya & Habib, ) is limited. Presence of the wild dog was recently reported from sub-alpine and temperate habitats of Uttarkashi district in Uttarakhand (Pal et al., b). The presence of these two Endangered carnivores in high-altitude forests emphasizes the need for regular monitoring of these areas over a longer period. Long-term monitoring could elucidate if and how these species persist in these habitats and climatic conditions. Our findings highlight the potential of the Bhagirathi basin as a stronghold for several threatened and rare mammal species. Persistence of these species can be attributed to the presence of remote, rugged and undisturbed habitats, and seasonal absence of people and livestock. Nonetheless, the distribution of threatened species overlaps with human activities both spatially and temporally, and thus these species remain vulnerable to anthropogenic pressures. A better understanding of their distribution, abundance and resource utilization, and of the anthropogenic pressures they are exposed to, is required for conservation planning.
FIG. 2 The activity overlap (grey area) and overlap coefficient (Δ) of anthropogenic disturbance (people, livestock and domestic dogs) with musk deer, Himalayan brown bear, Asiatic black bear, sambar, snow leopard and common leopard in Gangotri National Park and outside during summer in the Bhagirathi basin.