Identifying risk zones and landscape features that affect common leopard depredation

Study area

This study was conducted in Gorkha Municipality and Bhimsen Thapa Rural Municipality of Gorkha district (i.e., the epicenter of the devastating 2015 earthquake), located in the mid-hills of Nepal (Fig. 1). Gorkha was selected because Adhikari et al. (2022a) identified this location as a high-risk HWC zone. The district is located in Gandaki Province with latitudes 27°15′N to 28°45′N and longitudes 84°27′E to 84°57′E. The total area is 3614.70 km², while that of Gorkha Municipality (i.e., district headquarters) is 131.56 km² and Bhimsen Thapa Rural Municipality is 101 km² with forest coverage of 33% and 39.1%, respectively (DFRS, 2018). The average altitudinal range of the district is 228 m (e.g., banks of Marsyangdi River) above sea level to 8163 m (e.g., Mt. Manaslu Himalaya). Annual rainfall in the district ranges from 0 mm in December to 529.4 mm in July. Similarly, the maximum temperature (31.9 °C) is observed during June and the minimum (6 °C) in January.

According to Central Bureau of Statistics (2022), the total population of Gorkha Municipality is 53,285 people (15,298 households), and Bhimsen Thapa Rural Municipality is 17,118 people (5,265 households). Major ethnic groups in the study area include Brahmin, Chettri, Newar, Darai, Magar, Damai, Kami, and Kumal, which mainly belong to the Hindu religion. Their primary livelihood strategy is agriculture, especially farming and animal husbandry. Common livestock include buffalo (Bubalus bubalis), domestic goats (Capra aegagrus hircus), and poultry. The relative abundance of plants in the district are Shorea robusta in the southern aspect and Schima- Castanopsis forest in the northern aspect, with other associated species such as Syzygium cumuni, Rhus spp., Terminalia alata, and Semicarpus anacardium in the lower altitudes, whereas Myrica esculanta, Pinus spp. and Madhuca indica in the higher altitudes. Predominant wildlife species include the common leopard, jungle cat (Felis chaus), golden jackal (Canis aureus), yellow-throated marten (Martes flavigula), Rhesus macaque (Macaca mulatta), barking deer (Cervus vagianalis), and Indian crested porcupines (Hystrix indica) and so forth.

Research design

Prior to conducting any fieldwork, the Government of Nepal, Ministry of Forests and Environment, Department of Forests and Soil Conservation, in co-ordination with Divisional Forest Office (DFO), Gorkha issued research permissions for the study (issued #1728-2078/2079). The government has assigned responsibility to verify cases for providing relief to wildlife victims and dealing with issues outside protected areas. From 2019-2021, injuries and fatal cases of common leopard attacks on people and livestock in most affected areas were considered after consultation with the DFO. Meetings of concerned people were organized, including focus group discussions, and recorded locations and topographic, vegetation, environmental, and anthropogenic variables were used as primary data for each attack site. Secondary sources of information included DFO documents, the Manaslu Conservation Area, and other literature.

Data collection

Vegetation related variables

The diet of common leopards consists primarily of wild and domestic herbivores (Devkota, Silwal & Kolejka, 2013; Devkota et al., 2017), underscoring the need for vegetation (Andersen et al., 2000). Therefore, the forest cover of Global Forest Change (GFC) was downloaded from Earth engine partner Appspot and was used as a vegetable variable (Hansen et al., 2013). In addition, the Enhanced Vegetation Index (EVI) time series image was downloaded from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor from the USGS–EVI was used to model possible common leopard attacks. Afterward, ENVI software was used to smooth the data using an adaptive Savitzky-Golay filter in TIMESAT (Jönsson & Eklundh, 2004), which decreased the cloud effect, thus allowing us to obtain EVI mean, maximum, minimum, and standard deviation.

Data analysis

The MaxEnt software was used to generate a predicted livestock depredation risk map using geo-referenced incident points of common leopard attacks and environmental variables as input variables (Table 1) (Elith et al., 2006; Phillips, Anderson & Schapire, 2006; Phillips et al., 2017). The area under the receiver-operator curve (AUC) (Pearce & Ferrier, 2000) was used to validate the model, and true skill statistics (TSS) was used to evaluate it (Allouche, Tsoar & Kadmon, 2006). The AUC value of 1.0 indicates the perfect model, while 0.9−0.99, 0.8−0.89, 0.7−0.79, and 0.51−0.69 indicate excellent, good, fair, and poor model fit (Hanley & McNeil, 1982; Carter et al., 2016). The TSS value ranges between −1 and 1, where the higher value shows the model’s good performance (Allouche, Tsoar & Kadmon, 2006). Multicollinearity was found to be less than 0.7 between the variables, which was acceptable for modeling (Dormann et al., 2013). Seventy percent of the data was utilized to train the model, while the remaining thirty percent was used to validate it. We employed 10 replications, 1,000 maximum iterations, and 1,000 background points during modeling based on Barbet-Massin et al. (2012). Thresholds to optimize the sum of specificity and sensitivity were utilized (Liu, Newell & White, 2016) to compute TSS and construct the binary map from the continuous one.

Distribution of risk area

Of the study area (27.92 km²), 20.67 km² (74%) was identified as a low-risk zone, 5.21 km²(19%) as a medium-risk zone and 2.04 km²(7%) as a high-risk zone. Within the high-risk zone, Gorkha Municipality ward number 3 constituted 1.1 km², whereas Bhimsen Thapa Rural Municipality ward number 1 constituted 0.94 km² (Fig. 1B).

Important environmental variables predicting the common leopard risk zone

The most important variable for predicting common leopard risk zones was the distance to water (25.2%), followed by distance to road (16.2%) and elevation (10.7%). Similarly, the distance to settlement (17.6%) and distance to the road (14.7%) have high permutations, reinforcing their significance in the model (Table 2, Fig. 2).

Response curves (Fig. 3) indicated that the risk of common leopard conflict peaked at the nearest distance from the road and human/livestock path, and vice-versa. The curves also indicated that the HLC risk was highest between 1.5 and 2 kilometers from water sources and between 700 and 800 m in elevation.

Model accuracy

The model’s accuracy was fair, with an average AUC value of 0.726 ± 0.021. The TSS value was 0.61 ± 0.03, which indicates that the model has some discriminatory power because the TSS value ranges between −1 and 1.

Risk zones with respect to common leopard attacks

Our study identified a 2.04 km² area as a high-risk zone for HLC in Gorkha district, with Gorkha Municipality ward number 3 constituting a significant 1.1 km² of this zone. This finding aligns with a parallel study in the neighboring Aadhikhola Rural Municipality (Syangja district), which pinpointed a 2.76 km² very high-risk zone and a 5.56 km² high-risk zone for common leopard attacks (Adhikari et al., 2021). Moreover, Adhikari et al. (2022a) categorized the Gorkha district as a high-risk zone for human-induced common leopard mortality. This classification underscores the pressing need for targeted conservation efforts and management strategies within the region. Identifying high-risk zones provides valuable insights for devising intervention strategies for mitigating HLC while safeguarding villagers and common leopard populations in Gorkha and its neighboring areas.

Potentially, a contributing factor to this observed pattern could be linked to the comparatively low vegetation cover in Gorkha Municipality (33.0%), resulting in lower prey density within the forest in contrast to Bhimsen Thapa Rural Municipality (with 39.1% forest cover) (DFRS, 2018). Recent research has highlighted a concerning decline in several prey species vital to common leopards, such as barking deer and wild boar, particularly in the mid-hill regions (Baral et al., 2021a; Baral et al., 2021b). This decline may force common leopards to seek alternative food sources, as May (1977) suggested, including preying on livestock and even dogs (Athreya et al., 2016). An increase in common leopard attacks near settlements and agricultural lands in the Gandaki province of Nepal was reported by Baral et al. (2021b), emphasizing the growing interaction between common leopards and human-inhabited areas. Additionally, Rostro-García et al. (2016) noted that the prevalence of human settlements, mainly those dependent on animal husbandry, may have increased livestock density, resulting in heightened livestock kills by common leopards in Bhutan. As an outcome, areas inhabited by humans are emerging as crucial zones for wildlife conflict in Nepal (Acharya et al., 2016). Moreover, events such as livestock depredation and severe injuries to humans generate negative attitudes toward marauding wildlife (Treves & Karanth, 2003; Bagchi & Mishra, 2006). Such negative sentiments can lead to the killing or at least favor the killing of wildlife species (Don Carlos et al., 2009), thereby undermining efforts to conserve apex predators in the mid-hills (Baral et al., 2022b.). This interconnected chain of factors underscores the complexity of human-common leopard interactions in these regions and emphasizes the importance of integrated conservation strategies.

Important environmental variables predicting the common leopard risk zone

Key predictors of common leopard risk zones in our study included distance to water, road distance, and elevation. The study by Sharma et al. (2020a) identified crucial factors for HWC in the Kanchenjunga landscape, with distance to road, elevation, livestock density, and mean annual temperature as good predictors. In contrast, Upadhyaya et al. (2020) emphasized the importance of livestock numbers, ethnic groups, and village distance from the park boundary for explaining livestock losses in Bardia National Park, Nepal. Likewise, Sijapati et al. (2021) found that common leopard-induced livestock kills in Bardia National Park were more prevalent at locations distant from the park boundary, subject to seasonality. Key factors of common leopard depredation in South Africa included distance to villages, followed by distance to water, roads, nature reserve, and elevation (Constant, Bell & Hill, 2015).

The risk of common leopard conflict in Gorkha was highest between 1.5–2 Km from water sources. Water attracts domestic and wild animals, thus increasing the number of leopard attacks due to advantageous hunting conditions near these areas. Proximity to water also allows common leopards to conserve energy due to the concentration of animals and dense bushes/vegetation–making these locations favorable for ambushing prey species. A similar observation was reported by Constant, Bell & Hill (2015) in South Africa–most attacks on livestock occurred near water bodies (0–2 Km). Likewise, Miller, Jhala & Jena (2016) found livestock kills were most common between 2.9 ± 1 to 3.9 ± 0.2 km from water sources at the Kanha Tiger Reserve of Central India. Proximity to water is also related to the increased habitat suitability of common leopards (Simcharoen et al., 2008). Increased probability of common leopard presence near water sources was also recorded in previous studies conducted in India (Mondal, Sankar & Qureshi, 2013) and Nepal (Khaiju, 2017). Similarly, Abade (2013) reported distance to water as the strongest predictor of livestock depredation risk (including common leopards) in Tanzania. Consistent with other studies, distance to water will likely increase the risk of common leopard depredation. However, Naha et al. (2020) found that livestock killings increased with distance from water bodies in the Himalayan region and argued that water availability might not be a limiting factor for HWC with carnivores in South Asia. Similarly, Karanth et al. (2013) found that distance to water is not strongly associated with livestock losses in the Western Ghats protected areas.

The risk of common leopard conflict peaks at the nearest distance from the road and human/livestock path. This aligns with the findings of Sharma et al. (2020a), who reported a higher incidence of HWC within 5 kilometers of roads compared to farther distances. Another study by Miller, Jhala & Jena (2016) also discovered increased livestock kills within approximately 1.2 kilometers of roads in the Kanha Tiger Reserve, Central India. Contrary to our observations, common leopard attacks escalated beyond 8 kilometers from roads in South Africa (Constant, Bell & Hill, 2015). The reason for a higher probability of conflict near roads/paths in our study is likely due to rapid infrastructure development in the mid-hills of Nepal, including roads and walking trails (Dhami et al., 2023c), leading to habitat disruption and fragmentation, pushing common leopards closer to human settlements and grazing areas (Dasgupta & Ghosh, 2015; Roy & Sukumar, 2017).

Common leopards prefer living near the forest edge (Lamichhane et al., 2019; Pokheral & Wegge, 2019). Community forests in the Gorkha district are situated near villages (Dhami et al., 2023c). This inclination may increase negative interactions, especially with livestock depredation (Kandel et al., 2023). The study by Lamichhane et al. (2018) documented significant instances of livestock depredation occurring within a 500-meter radius of the boundaries of the forest and the village in CNP. In contrast, a higher frequency of livestock attacks close to forest edges was noted within CNP (Ruda, Kolejka & Silwal, 2020). Similarly, another report shows that the most common leopard attacks occurred within a 1-kilometer radius of forest edges in Baitadi District (Baral et al., 2022a). Furthermore, linear features such as roads and trails serve as linear corridors, influencing both the movement patterns of common leopards and the distribution of ungulate species, notably cattle, and goats, drawn by the availability of abundant forage in landscapes subjected to human exploitation. The increased potential of prey near roads amplifies the risk of common leopard attacks, given that people and animals utilize these well-defined paths frequently (Neupane et al., 2022; Dhami et al., 2023d).

The livestock depredation risk by common leopards was highest between 700 m and 800 m elevation. It can be related to more settlements in the study area and more livestock abundance at that elevation. Topographic factors (i.e., terrain) frequently determine accessibility, which limits human-wildlife interactions (Neilson, Nijman & Nekaris, 2013). As mentioned by Pitman, Swanepoel & Ramsay (2012), the probability of common leopard predation in South Africa increased by 6% at higher elevations where big cats often seek refuge. Maximum HWC was reported at 300–4000 m elevation at the Kanchenjunga landscape in Nepal (Sharma et al., 2020a) due to greater forest fragmentation at that range. The risk of livestock kills was highest at lower elevations (670–780 m) and higher elevations (1540–1760 m) on Blouberg Mountain in South Africa (Constant, Bell & Hill, 2015). Likewise, the greater percentages of common leopard attacks on humans were reported at elevation ranges of 1,000 to 1,500 m and 100 to 500 m in Pauri Garhwal and North Bengal (India) (Naha, Sathyakumar & Rawat, 2018).

Based on our research findings and ensuing discussions, we suggest the expansion of this study to encompass other districts within Nepal where common leopards are prevalent. This extension is crucial for understanding depredation patterns in additional locations that our study did not address. While our research anticipated livestock depredation by considering various bio-physical factors, it did not delve into the identification of causative elements, including but not limited to habitat conditions, prey availability, herding and rearing practices, and the design of livestock sheds/corrals. The study sites are located in the goat farming pocket area where local farmers are involved in goat farming because of high demand and good price of small he-goats that are sacrificed by the devotees in Gorkha Kalika and Manakamana temples. Poorly constructed traditional corrals have been easily broken by the leopards during night time. Increasing emmigration trend of the viillages, fallow lands gradually changing in forested areas that contributed to support leopards’ movements nearby human settlemnts and enough time to find the weaker goat corrals. Analysis of these factors is integral to formulating effective mitigation measures for fostering harmonious co-existence between humans and common leopards across the country. Further exploration and clarification in these areas are essential for informing and refining strategies to manage human-leopard conflict.