Patterns and determinants of elephant attacks on humans in Nepal

Abstract Attacks on humans by Asian elephant (Elephas maximus) is an extreme form of human–elephant conflict. It is a serious issue in southern lowland Nepal where elephant‐related human fatalities are higher than other wildlife. Detailed understanding of elephant attacks on humans in Nepal is still lacking, hindering to devising appropriate strategies for human–elephant conflict mitigation. This study documented spatiotemporal pattern of elephant attacks on humans, factors associated with the attacks, and human/elephant behavior contributing to deaths of victims when attacked. We compiled all the documented incidences of elephant attacks on humans in Nepal for last 20 years across Terai and Chure region of Nepal. We also visited and interviewed 412 victim families (274 fatalities and 138 injuries) on elephant attacks. Majority of the victims were males (87.86%) and had low level of education. One fourth of the elephant attacks occurred while chasing the elephants. Solitary bulls or group of subadult males were involved in most of the attack. We found higher number of attacks outside the protected area. People who were drunk and chasing elephants using firecrackers were more vulnerable to the fatalities. In contrast, chasing elephants using fire was negatively associated with the fatalities. Elephant attacks were concentrated in proximity of forests primarily affecting the socioeconomically marginalized communities. Integrated settlement, safe housing for marginalized community, and community grain house in the settlement should be promoted to reduce the confrontation between elephants and humans in entire landscape for their long‐term survival.


| INTRODUC TI ON
Asian elephant (Elephas maximus, referred to as "elephant" hereafter; Figure 1) is a globally endangered megaherbivore (Williams et al., 2020). It is an umbrella species in tropical and subtropical forests of Asia and has a strong cultural role in various Asian societies (Jadhav & Barua, 2012;Menon et al., 1996;Sukumar, 2003;Vasudev et al., 2020). Once widely distributed in Asia, elephants are now confined to ca. 5% of their historical range in highly fragmented landscapes (Sukumar, 2006). In addition, the rapid development of linear infrastructures including railways, highways, electric transmission lines and irrigation canals cause further obstruction to elephant movement. Elephants require large areas for their survival with long-distance seasonal movements (Goswami, 2017;Leimgruber et al., 2003). However, increasing habitat fragmentation brings them in frequent confrontation with humans. As a result, human-elephant conflict (HEC) is escalating and has become a prominent cause of elephant population decline (Sukumar, 2006). Attack on humans by elephants is the extreme form of HEC. Other effects upon local people from HEC include loss of crops, damage to property, and safety threats (Dickman, 2010;Gross et al., 2021); and a large number of elephants are also killed in retaliation.
Nepal is a typical example of an elephant range country with a small but growing population of >200 elephants in highly fragmented landscape (Ram & Acharya, 2020). Increasing encroachment and forest conversion in the southern lowlands of the Terai and Chure (Himalayan foothills) region have destroyed the traditional migratory routes of the elephants (Ram, 2014). Whereas a few solitary bulls living in protected areas are habituated to visiting agricultural areas for a higher quality diet causing a huge amount of fiscal losses (Koirala et al., 2016). Elephants cause the highest number of human deaths among the wildlife species in Nepal. Due to this, HEC is a serious issue throughout the lowland Nepal (Acharya et al., 2016).
Few studies on human-elephant conflict have been carried out in Nepal primarily focusing on crop and property damage (Graham et al., 2016;Gross et al., 2021;Neupane et al., 2013;Pant et al., 2016).
However, detailed studies of elephant attacks on humans are still lacking. This study attempts to document spatiotemporal pattern of elephant attacks on humans in Nepal, characteristics of the victims, and attacking elephants, determine factors associated with the attacks, and identify human and elephant behavior contributing to deaths of victims when attacked. We tested hypothesis: (a) Human activities are responsible for elephant attacks on humans; (b) encounters leading to attacks by elephants are higher in the proximity to forests, (c) elephant attacks were higher inside protected areas, and (d) solitary bull elephants are responsible for attacks on humans.
The study results have long-term implications for the conservation and management of elephants in the human-dominated landscape of Nepal and beyond.

| Study area
The study was conducted across the Terai and Chure region of Nepal covering ca. 46,000 km 2 of elephant range in 24 districts ( Figure 2).
The Terai and Chure region is densely populated with 391.5 persons/km 2 (CBS, 2012). About 51% of total population of Nepal reside in the region with agriculture and livestock husbandry as the primary occupation. About 42% of the study area is forested providing habitats and migration corridors for the elephants (DFRS, 2015).
Major cities, industrial areas, and highways fragment the forested areas. The forests in the region were intact till 1950s but afterward it is under continuous human pressure from expansion of agriculture, settlements, and built-up areas.
The maximum temperature varies from 35 to 40°C in summer and 14 to 16°C in winter (Jackson, 1994). The mean annual rainfall ranges between 1,138 and 2,680 mm, with over 80% of the rain occurring during the 3 monsoon months (Lamichhane, Persoon, et al., 2018;

| Data analysis
We entered all the questionnaire survey data in MS Excel and prepared descriptive summaries using pivot  (United Nations, 1982). Education level of the victims was categorized into illiterate (who cannot read and write), literate (who can read/write but have not attended formal school), primary (completed primary school), and secondary or above.
Housing of the victim was categorized into cemented house, corrugated galvanized iron (CGI) sheet roof house, tiled roof house, and thatched house.
We ranked models by the small-sampled corrected Akaike's information criteria (AICc, lower AICc value indicates higher model ranking) using multi-model inference in "MuMIn" package in R (Barton, 2020). The top model for making ecological inference was obtained by averaging the models in the candidate set supporting the data equally well (AICc ≤ 2, (Burnham & Anderson, 2001).  (Table 3).

| Elephant characteristics
Most of the elephant attacks on humans (85.2%, n = 412) were caused by solitary adult bulls or subadult male groups and rest of the attacks were caused by the elephants in herd or females separated from the herd ( Table 4). The bulls involved in the attacks were in musth in more than half of the incidents.

| Temporal and spatial distribution of elephant attacks on humans
Elephant-related human attacks varied significantly across months In the forested areas, elephant attacks on humans were at peak in the afternoon (4-5 p.m.), whereas, in settlement areas, elephant attacks peaked in the evening (7-9 p.m.) (Figure 4). incidents (n = 66) from Jhapa and Bardiya districts ( Figure 5). The majority of elephant attacks (67%) occurred within 500 m from the forest edge ( Figure 6).

| Factors associated with human fatality
People who were drunk and chasing elephants using firecrackers were more vulnerable to fatalities while chasing elephants using fire was negatively associated with fatalities (Table 5).

| Characteristics of the victims of elephant attack
Elephants attacked men more frequently than women which can be associated with the high mobility of males and their involvement in chasing the elephants (Sarker et al., 2015).

| Characteristics of elephants attacking humans
Mixed herd elephants rarely attacked humans (<5% of the incidents) although they are involved in crop raiding during migration through agriculture areas or settlements (Naha et al., 2020). Bangladesh (Sarker et al., 2015). In general, bulls range widely, operate solitarily or in small groups and young bulls that disperse out of natal herds lack the social buffering that herd individuals would have. Thus, young bulls can be more excitable and due to their wide movement, bulls can come into frequent contact with people, some of which can turn severely negative (Fernando et al., 2008). However, they are also harassed by people most of the time while raiding crops or grain stores. These irritating actions of humans make them more aggressive resulting in violent attacks (Sampson et al., 2019).
We identified 37 such bulls causing three quarters of all attacks on humans in the last twenty years, some of them caused a disproportionately higher number of attacks (up to 36). Such individuals can be termed "problem individuals" need to be closely monitored, particularly their movement patterns and ranging behavior (Lamichhane et al., 2017). The knowledge thus gained through monitoring can be helpful in prioritizing appropriate management strategies. Note: Significance codes: <0.001 "***"0.001 "**" 0.01 "*"0.05"."0.1 " " 1.

| Temporal patterns of elephant attacks
Documented records of elephant attacks on humans in Nepal goes back to the 1970s (B.N. Uprety, 2020, Personal Communication.) with sporadic records until the late 1990s (Smith & Mishra, 1992). In our study, we only included data between 2000 and June 2020. The wild elephant population has gradually increased in Nepal from 52-53 individuals during the 1990s (Smith & Mishra, 1992)

| Spatial pattern of elephant attacks
Two-thirds of elephant attacks on humans occurred within 500 m from the forest edge. People living in proximity of forests are vulnerable to elephant attacks because (a) chances of encountering elephants are high at close distance to forest, (b) generally economically marginalized communities live in these areas with lack of proper housing (thatched houses), which is often flimsy. Similar finding of a higher number of attacks by wildlife close to forest or park boundary (<1 km) and an inverse relationship between the distance from the forest edge and wildlife attacks is reported in other studies (Gurung et al., 2008;Pant et al., 2016).
A higher number of elephant attacks outside protected areas (59.5%) in our study is consistent with Acharya et al. (2016). Similar results with higher conflict incidents outside protected areas have been reported from north-east India as well (Choudhury, 2004). entering Nepal from the eastern border during September-October and May-June (Mallick, 2012). While migrating, they often came in confrontation with people as they are forced to travel through settlements and agricultural land, with a large part of their historic migration route encroached by people (Choudhury, 2004

| Factors associated with the human fatality
Our results of two third of elephant attacks resulting in the fatality are consistent with Acharya et al. (2016). Human behavior and responses toward elephants were the major factors to cause elephant attacks on humans. Aggressive human behavior toward elephant with intolerance was the major determinant of human fatality in elephant attack (Nelson et al., 2003). People were killed mostly while chasing wild elephants using firecrackers and other high sound and light objects. Such elephant drives when haphazardly done would increase the fear and trigger defensive offensive behavior in elephants increasing the probability of human attacks. In the Terai, consumption of alcoholic beverages by local communities during the leisure hours of evening and night is high. Driving elephants when they come to crop fields in an inebriated state can increase the vulnerability of attacks by elephants . Negative association of fatalities while chasing elephants using fire torch indicates it as a safe and effective method for pushing elephants outside of the village.

| CON CLUS IONS
Human casualties from elephants have been increasing with its mul- Program through WWF Nepal, NTNC, and ZSL Nepal. We like to thank anonymous reviewers and editors for their constructive comments to improve the manuscript.

CO N FLI C T O F I NTE R E S T
None declared. We only carry out stakeholder consultation and questionnaire survey by taking verbal consent from the participants. We removed