Introduction

Human-wildlife conflict (HWC) is a growing global concern, particularly as human populations expand and natural habitats shrink (Ma et al. 2024). This conflict arises when wild animals enter human settlements or agricultural lands, leading to negative interactions that can damage crops, livestock, and even human life. Rapid urbanization, deforestation, and agricultural expansion have fragmented ecosystems, forcing wildlife to adapt to human-dominated landscapes (Thapa et al. 2024). Across the globe, this has escalated tensions between local communities and wildlife, particularly in regions that border protected areas or biodiversity hotspots (Zhang et al. 2024). The consequences are often more severe, affecting not only the livelihoods and safety of rural communities of mountainous landscapes but also threatening species conservation.

In recent decades, conflicts or negative interactions between humans and wildlife
in the Indian Himalayan Region (IHR) have intensified (Sylvia & Sharma, 2023),
conspicuously involving common leopards (Panthera pardus), Asiatic black bears
(Ursus thibetanus), wild pigs (Sus scrofa), and rhesus macaques (Macaca mulatta)
(Mishra 1996; Chauhan, 2009; Meena et al. 2021; Hussain et al. 2018; Hussain
et al. 2022). These species are reported to cause extensive property loss, such as
livestock kill, human injury/death, property damage, and crop damage (Hussain et al. 2018, 2022). The rise in human-wildlife conflict in recent years has been attributed to anthropogenic pressures, which have led to habitat degradation and fragmentation. Rangelands and grassy slopes have been extensively grazed by livestock (Mishra 1996; Ogra 2008), which leads to competition with wildlife for resources and reduces the tolerance levels among humans towards native wildlife in the landscape (Ogra & Badola 2008; Bargali & Ahmad 2018; Hussain et al. 2018).

Most negative interaction incidents with wildlife in the Himalayas are reported from areas adjacent to forests (Ogra 2008; Hussain et al. 2018), involving crop and property damage, livestock depredation, as well as human deaths and injuries (Negi 1996; Naha et al. 2018; Hussain et al. 2022). Besides economic loss, the aftermath of these negative interactions includes psychological stress, fear of travelling through roads and trails adjacent to forested areas, the pressure of guarding crops, and reduced human activities (Ogra 2008). These factors contribute to negative attitudes among locals towards wildlife and local community conservation initiatives (Ogra 2008; Hussain et al., 2022). Incidents of HWC, such as the death of humans or livestock and damage to crops or property, evoke severe backlash from people and pose setbacks for wildlife conservation (Ogra & Badola 2008; Agarwal et al., 2011). Understanding these interactions is crucial for addressing complex conservation problems (Pradhan 2018), and a thorough understanding of these conflicts is essential for minimizing them (Ogra & Badola 2008; Farooque et al., 2004). One contributing factor, which increases negative interaction with wildlife in the IHR, is human out-migration due to a lack of opportunities and good education for children, leading to abandoned villages and agricultural land (Pathak et al., 2017). This also reduces traditional techniques for minimizing conflict, as fewer people guard, chase, and protect livestock and crops (Hussain et al., 2022).

Studies on HWC have primarily focused on adjoining protected areas (Bargali & Ahmad 2018). However, research on such interactions is limited within the IHR, where abundant forest cover and resources mean that the natural habitats of wild animals often extend beyond government-designated protected areas. Additionally, human settlements in the IHR are patchy and dispersed. The population density of the IHR is also significantly lower than that of peninsular India. The Kailash Sacred Landscape (KSL) is one of the important Himalayan landscapes, and wildlife and humans share the same space (Hussain et al. 2018, 2022). There is scarcely any baseline data on HWC from the Indian part of KSL. Large-ranging wild mammals involved in HWC in the Himalayas are found in almost all the districts and villages in the region (Hussain et al. 2018). Considering the rising cases of negative interactions with wildlife, including crop loss, livestock loss, and human death or injury, this study aimed to assess the patterns of human-wildlife negative interaction in the Indian part of KSL. Furthermore, the Indian KSL holds both cultural and ecological sensitivity, raising the stakes where traditional livelihoods and biodiversity conservation meet.

The primary objectives of this study are (1) to investigate the patterns and factors contributing to HWC in the Indian part of the KSL, a unique landscape where community and wildlife share a common space, and (2) to identify human-wildlife
conflict hotspots focusing four major wild animals.

This research offers two key contributions to the field. First, it investigates HWC in the Indian part of the KSL, which is a distinctive area where wildlife coexists with humans outside of protected zones. Thus, this study fills a gap in the existing literature on HWC, which is primarily concentrated on regions adjacent to protected area. Second, the study identifies specific villages as hotspots for negative interactions. This identification facilitates the development of targeted conservation strategies at a regional scale and allows for more effective resource allocation to mitigate human-wildlife conflicts within these communities.

Materials and Methods

Study area
KSL is an important transboundary landscape comprising portions of the southwestern Tibetan Autonomous Region of Poeople’s Republic of China, adjacent portions of northern India and northwestern Nepal (Zomer & Oli, 2011). The landscape is mountainous and remote, with steep topography, high spatial heterogeneity, and difficult access. The total area of KSL extends over 31,000 km² area, and in the Indian part, it covers the Pithoragarh district in Uttarakhand state (Figure1), with a geographical area of 7,120 km² lying between29° 20’-30° 55’N latitude and 79° 50′- 81° 0’E longitude. The district has seven forest ranges, viz. Askot, Berinag, Dharchula, Pithoragarh, Didihat, Munshiyari, and Gangolihat. The forest types range from moist subtropical broadleaf to temperate oak forests, sub-alpine conifers, high-altitude birch forests, alpine meadows, and grasslands. The forests in this landscape fall under protected areas (PA) and non-protected areas. Askot Wildlife Sanctuary is the lone legally designated PA. However,the non-PA category of forests includes reserve forests under the control of the state forest department and community forests or Van Panchayat, protected by the local community body called panchayat. Among all the seven ranges, density of human settlement is very high in the Pithoragarh range, which encompases the district headquarters. Nearly 80% of the villages are connected with roads in the study area. About 63% of people in the study area depend on agriculture and livestock farming (Zomer & Oli 2011; Hussain et al. 2018). Nearly 53% of the agricultural land shares its boundary with forest and 90% of the farming land is rain-fed. Numerous rivers, tributaries, and springs are spread across the landscape. The study area shows wide altitude variation ranging from < 350 m asl to peaks ranging > 7000 m asl (Zomer & Oli 2011; Hussain et al. 2018).

Compiling conflict data and field visits
To understand the nature of human-wildlife conflict, we compiled records of human injuries and fatalities, livestock depredation, compensation, and crop losses collected and records maintained by the Pithoragarh forest department from 1984 to 2017. The data collected from the forest department included information on the location of conflict incidence, households affected, species causing the problem, date and time, and compensation amount paid (Bargali & Ahmad 2018). Using records of repeated conflict incidents from 1984 to 2017 and data on compensation paid to victims, we determined the most conflicted regions encompassing seven forest ranges in KSL-India. The area was subdivided into 2 km² grids for systematic sampling, with each grid visited for a household survey (Figure1). A total of 584 individuals from the survey region were interviewed to gather information on human-wildlife interactions and to validate data obtained from the forest department (Hussain. et al 2018; Naha et al. 2018). During the interviews, all the respondents were asked about the details of conflict incidents and their companions, as well as more information on age, time, month of the attack, and activity during the attack (demographic details of the respondents in Supplementary Table S1). All the conflict locations collected from the field and the forest department records were marked and mapped using GPS locations in ArcGIS 10.3.

Predicting HWC Hotspot
Using presence data with possible influencing factors, we predicted the conflict hotspots of the common leopard (n = 151), Asiatic black bear (n = 76), wild pig (n = 631), and rhesus macaque (n = 278). The spatial distributions of these locations are provided in the Supplementary Figure S1. We identified two main categories related to conflict distribution, namely abiotic factors (elevation, slope, aspect, land use land cover- LULC, and

Figure 1: Map of the Indian part of KSL, showing the questionnaire survey area, determined based on villages that experienced repeated conflict
incidents from 1984 to 2017 and where the Forest Department compensated affected communities. The survey area covered seven forest ranges of the region. This area was divided into 2 km² grids, with at least one household surveyed per grid.

Table 1. Hypotheses for selecting various independent variables to model the human-wildlife conflict hotspots using MaxEnt in Kailash Sacred Landscape, India

Table 2. The per cent contribution of different variables and AUC values for predicting conflict distribution of rhesus macaque, wild pig, common leopard, and Asiatic black bear in Kailash Sacred Landscape, India.

distance factors) and bioclimatic and biotic factors (livestock density, anthropogenic factors) (Phillips et al. 2006; Sarkar et al. 2018). We selected elevation, slope, aspect, Normalized Difference Vegetation Index (NDVI), and bioclimatic factors for landscape-level analysis. Details of all 14 predictor variables are provided in Table 1. In ArcMAP 10.3, elevation, aspect, and slope were generated using a digital elevation model (DEM) layer, with all variables clipped to our study areas. Land use and land cover were classified using LANDSAT 8 imagery. All predictor variables were re-scale to 1 km² spatial resolution,and we performed pairwise calculations of Pearson’s correlation coefficient (Supplementary Table S2) among the influencing factors, confirming that the correlation coefficient remained within the range of ±0.75 for any combination of variable pairs (Tehrany et al., 2019). Based on Pearson’s correlation coefficient, we excluded human density and LULC from the list of influencing factors, retaining 12 variables for further analysis (Table 2, Supplementary Figures S2-S13).

To address spatial sampling biases and spatial autocorrelation challenges for the location data of four prominent conflicting animals, viz., common leopard, Asiatic black bear, wild pig, and rhesus macaque, we employed a bootstrapping approach using the spThin package in R (Aiello-Lammens et al., 2019) for each species separately. To achieve a 1 km² spatial resolution for the modeling exercise, we conducted 50 iterations of bootstrapping using a minimum thinning distance of 1 km. This process resulted in obtaining 138, 72, 628, and 259 unique locations of common leopard, Asiatic black bear, wild pig, and rhesus macaque, respectively. This approach effectively minimized the removal of records while significantly mitigating sampling bias (Aiello-Lammens et al., 2015).

We used the maximum entropy model (MaxEnt) (Phillips et al. 2006; Naha et al. 2018; Sarkar et al. 2018) to predict the distribution of conflict in the study area under the present scenario. We input all biotic and abiotic variables and presence locations into MaxEnt version 3.4.1. We ran the model through twenty-five bootstrap resampling runs with 75% training data and 25% for model validation. Jackknife and sensitivity analyses were carried out to determine the contribution of variables (Naha et al. 2018). Area under the receiver operator characteristic curve (AUC) was used to assess model accuracy (Phillips et al. 2006; Sarkar et al. 2018). The results were interpreted as conflict hotspots. We used logistic probability, and the gradient map was reclassified in ArcMap 10.3 to show conflict distribution for common leopards, Asiatic black bears, wild pigs, and rhesus macaques, each separately categorized as low (0.0–0.14), moderate (0.14–0.42), and high (0.42–1.0). Hotspot areas with < 0.8 values were extracted from the final shape files of all four conflict hotspot maps and merged all the clip shape flies of the final hotspot.

Results

Human-wildlife conflict in KSL India
A total of 560 crop damage incidences were recorded between 2002 and 2017 from seven forest ranges of KSL India (Figure 2), and 574 livestock depredation and 123 incidences of human injuries and deaths were recorded between 1984 and 2016 (Figure 3). Most human injuries and deaths were associated with the common leopard (75.29%), followed by the Asiatic black bear (24.51%) (Figure 3A). Likewise, the maximum incidences of livestock depredation were also by common leopards (83.62%) followed by Asiatic black bears (16.2%) (Figure 3B). Maximum incidences of human injury/death (n=12) were recorded in 2011, and no incidences were recorded in 2003, 2006 and 2007 (Figure 3A). Maximum livestock depredation (n = 38) was recorded in 2012, and minimum (n = 3) in 2010 (Figure 3B). Maize was the most commonly raid crop (n = 202), followed by wheat (n = 190) and paddy (n =168) (Figure 2). Wild pigs (n = 301) were responsible for most of the incidences of crop raiding, followed by rhesus macaques (n = 259) (Figure 2). Maximum incidences of crop loss were reported in 2017 (n=76), and the lowest number was in 2002 (n = 5) (Figure 2). Common leopards mainly preyed on goats (50.98%), followed by cattle (43.30%), horses (3.54%), and mules (2.16%). For the Asiatic black bear, a similar trend was noted; the primary victims were goats (65.07%), followed by cattle (15.87%), mules (14.28%), and horses (4.76%).

Seasonal and temporal variation of leopard and bear attacks in KSL India
Common leopard caused more incidences of livestock depredation and human attacks than Asiatic black bear (Figure 4A). Moreover, the livestock depredation was highest in the winter season (47.23%), followed by the monsoon (29.27%) and the summer season (23.5%) (Figure 4B). During the monsoon season, common leopards accounted for 75.55% of the depredation incidents, while Asiatic black bears accounted for 24%. In the winter season, the proportion of incidents involving common leopards rose to 83.43%, compared to 16.57% for black bears (Figure 4B). The record of the forest department and informal interactions with the victim’s family members revealed that 36.59% of victims were attacked in villages near the forest, 30.89% of victims were attacked inside the forest, 22.76% of victims were attacked in agricultural fields, and 9.76% of victim’s were attacked near nallah or streams (Figure 5A). About 56% of attacks occurred in the evening between 16:00 h and 20:00 h, 29.27% of the attacks occurred between 21:00h and 24:00h, 10.57% of attacks occurred between 06:00h and 15:00h, and 4.07% of attacks happened between 01:00h and 05:00h (Figure 5B). Among the age groups, the analysis revealed that 43% of victims of the common leopard and black bear attack in the Kailash landscape were between 31-45 years in age, followed by 33.33% between 16-30years age group, 12.20% between the 0-15 years age group and 11.38% above 45 years of age (Figure 5C). Compensation paid by the forest department to the community members in terms of human injury/death and livestock depredation by common leopard and Asiatic black bear was ₹17,51,300 ($20,866) between 2004 – 2016. The data collected from the forest department shows that the compensation amount is gradually increasing at the study site (Figure 5D).

Predictor variables and human-wildlife conflict risk zones in KSL India
We selected 12 independent and ecologically significant variables to predict conflict distribution and hotspots in KSL (Table 2). The AUC values of best performing MaxEnt models for wild pig, rhesus macaque, black bear and common leopard were 0.92, 0.93, 0.97 and 0.95, respectively (Table 2). All the ROC curves predicting conflict hotspots for wild pig, rhesus macaque, black bear, and common leopard are provided in Supplementary Figure S15. The analysis revealed hotspots of wild pig conflict near forested areas or fringes of the forest, particularly in the lower part of the study site, along the forest edges and towards agriculture fields near forest patches and near the water body. In the analyses of conflict hotspots for wild pigs and rhesus macaques, Euclidean distance to agriculture was identified as the primary factor. This variable accounted for 51.2% of the prediction for rhesus macaques and 39.6% for wild pigs. Additionally, distance to settlement emerged as the second most significant factor, contributing 9.3% to the prediction for rhesus macaques and 18.5% for wild pigs (Table 2). However, distance to the forest contributed the lowest, followed by annual temperature and precipitation.

The response curves of Euclidian distance to agriculture demonstrated a high risk of conflict with wild pigs and rhesus macaques closer to agricultural fields, with the conflict probability falling drastically after 100 – 500 meters. MaxEnt fitted the response curve unexplainably at short distances, possibly implying overfit. The response curve for annual average precipitation increased the risk of conflict in areas with higher annual rainfall. The analysis also predicted high risk in areas with very high annual average precipitation, fitting a U-shaped response to the variable. Most models predicted areas with NDVI of 0.4–0.8 having the highest risk of conflict. In the models, the conflict of crop raiding risk rose in areas with relatively low human settlement and high agricultural regions. The models fitted response showing crop raiding by macaques and pigs being more likely in areas between 600 and 3000 m asl. The model response predicted the highest risk for conflict in close vicinity of water sources, decreasing with distance and increasing between 0 to 100 m distance.

Livestock density was overall the most crucial variable in human-bear conflict models, contributing 60.9%, followed by the contribution from elevation at 8.5%, and Euclidean distance to settlement at 7.6%. The least essential variables were the Euclidean distance to the road, slope, annual mean precipitation and temperature (Table 2). In the case of common leopard, the most important variables impacting the distribution of human-leopard conflicts were livestock density contributing 53 %, annual mean precipitation 10.3%, and elevation, contributing 7.%. The models predicted a high risk of conflict with common leopards and black bears in areas with a high density of cattle and human settlements closer to forested areas.