Factors driving human–elephant conflict: statistical assessment of vulnerability and implications for wildlife conflict management in Sri Lanka

Sri Lanka’s population is overwhelmingly rural, with more than 80% living in the countryside (World Bank 2024). The rural poverty headcount—the percentage of rural people living in poverty measured by national poverty lines—has declined from 24.7 to 4.8 from 2002 to 2016, although pockets of poverty remain (CBSL 2019, p. 136). In particular in the research area (see section “Conceptual framework”), paddy (rice) farming and other agricultural activities are the main sources of income. This renders affected village communities vulnerable to negative outcomes of human–elephant encounters, especially crop foraging and post-harvest loss (such as damage to grain storages), as well as bodily harm.

Vulnerability is a central concept in social–ecological research, in particular pertaining questions of hazards, risks, and environmental change (e.g. climate change) (Scoville-Simonds and O’Brien 2018). Vulnerability describes the extent to which a person, household, community, or other societal group will be potentially touched or harmed by a particular hazard or threat. Vulnerability is closely linked to exposition, which means that there are factors that might either shield or expose someone to a given hazard. With regards to wildlife, the ranging grounds of the wild animals define the exposition. Sensitivity refers to the degree, once exposed, people react to a given hazard or threat. A third important dimension to vulnerability is the capacity to adapt, which denotes the ability to handle hazards (Adger 2006).

In the social sciences literature, ‘natural hazards’ and vulnerability to such hazards are seen as products of social and political factors (Blaikie et al. 1994), such as socio-economic and geographic factors, gender, and ethnic group affiliation. This conceptualization understands hazards not as a ‘natural’, but as politically charged and socially constructed phenomena (Hewitt 1995); at the same time, it is important to retain the sense that such environmental hazards are not solely constructed (Nygren and Rikoon 2008), but have serious, sometimes even devastating impacts on vulnerable populations in the ‘real’ world. Marginalization of rural people enhances their vulnerability and erodes resilience (Blaikie et al. 1994).

In several instances and different geographic contexts, a vulnerability lens has been applied to HEC (Nyirenda et al. 2018; Ogra 2008; Twitcher 2018). Where quantitative approaches were used, researchers either established complex general vulnerability indices (Cutter et al. 2003), made use of existing indices (Pereira et al. 2021) or modified existing ones (Aksha et al. 2019) to fit their purposes. In this study, instead of relying on a complex index, a conceptual framework was created with specific relevance and contextual information directly related to human–elephant encounters in Sri Lanka.

The foundation to the conceptual framework used in this study, based on an understanding of existing literature on human–elephant conflict in Sri Lanka (Köpke et al. 2021), lies in the theoretical assumption that human–elephant conflict is driven by factors covering the domains of the economic (including agricultural production), the environmental (including land-use and land-cover change), the political (regarding conservation governance and land-use planning policy) and the social (including cultural dimensions). These domains are translated into concrete independent variables.

Vulnerability to HEC is associated with livelihoods and the coping capacity of rural communities, highlighting the importance of socio-economic conditions (Thant et al. 2021; van de Water and Matteson 2018; Wilson et al. 2015). Further, land use and proximity of fields and settlements to elephant ranges is of utmost importance because it determines exposure to elephants (Hoare 1999; Shaffer et al. 2019; van de Water and Matteson 2018; Sitati et al. 2003; Neupane et al. 2017; Rathnayaka et al. 2022; Bharaty et al. 2022; Withanage et al. 2023). Under the ‘policy’ variable, factors may be understood that relate to the governance of conflict through mechanisms like compensation to damage and official recognition of property and crop damage (Sampson et al. 2021; Guru 2021). Finally, under ‘awareness’ this study categorizes the set of attitudes and practices that is linked to the mitigation of human–elephant conflict, to the perception of elephants and the potential for convivial encounters, and to preventive measures like insurance (Fernando et al. 2005; Talukda and Choudhury 2020; Nayak and Swain 2022).

According to the above-mentioned insights, vulnerability to HEC in the study framework, as the dependent variable, was analyzed in relation to the following independent variables: ‘socio-economic conditions’, ‘land use’, ‘policy’, and ‘awareness’. The questionnaire included a five-point Likert statement intended to capture independent variables. Those independent variables were derived by grouping items from the questionnaire, i.e. occupation, education, loss from crop damage, and the existence of protection mechanisms were grouped to create an indicator of socio-economic conditions, etc. The conceptual framework model for the present study is given below (Fig. 1), and detailed methodological descriptions are presented in the following subsections.

The research, as a part of the collaborative project between Rajarata University of Sri Lanka and University of Kassel, Germany, was conducted from 2018 to 2022 in dry zone areas in Sri Lanka. The key objectives of the project were to investigate the multiple factors, dynamics, and policies driving human–elephant conflict in Sri Lanka. The research team undertook the field survey across the areas covering Anuradhapura, Polonnaruwa, Vavuniya, Trincomalee, Ampara, Matale, Kurunegala, Puttalam, and Mannar districts including 89 Grama Niladhari (GN) divisions (Fig. 2)—GN being the local-level, lowest administrative unit in the country. The total area covered 26,921 km²—30% of regions of the country where HEC is most prevalent (Fernando et al. 2015).

The survey was conducted using a questionnaire with both open-ended and closed-ended questions. The survey participants were selected from the GN divisions that are generally situated within the critical areas affected by the HEC problem. Based on studies from Nyhus et al. (2005) and Okech (2010), we define a critical area where the agricultural and residential areas of communities are located adjacent to a protected area, or newly settled communities encroaching on existing or previously existing wildlife habitat. Some of the GNs are located in between the elephant corridors. In each village, we selected participants for the survey after the pre-study about the village through the village officer (GN officer) and the farmer association or the civil organization in the village. In some cases, we used random sampling for selecting the survey study in the village. The participants in the survey, interviews, and focus group discussions (FGD) were people who had different socio-economic backgrounds.

We designed the survey with the objective of achieving multiple objectives within the main project, thereby facilitating the computation of diverse statistical data pertaining to various subjects that are oriented toward a similar objective. The surveys were completed by research assistants during face-to-face interviews. Some surveys were filled out during the group discussions and individual meetings. The total sample is 651 (N = 651). Among them, 400 were male participants and 251 were female participants. The FGDs and the survey data collection were at the community halls or meetings (farmers’ associations or death donation associations), agricultural farmland, common local shops, or at religious places, such as temples or churches or Hindu-Kovil. Ethical clearance for the study was provided by the Ethics Review Committee of Rajarata University Faculty of Applied Sciences (Ref.-No. ERC/07/22). Informed oral consent was obtained from all participants of the study.

The questionnaire was formulated based on five domains, namely vulnerability, socio-economic conditions, environmental change and land use, policy, and awareness (Table 6 in Appendix). These domains are interrelated and rationally combined with the research hypotheses, survey questions, and the findings from the literature review. To maximize the response rate and minimize the errors in the responses, questionnaires were filled in by independent researchers. In addition to the questionnaire survey, field observations were used as a qualitative approach to gain an in-depth understanding of the coping capacity of the rural community and exposed risk. Furthermore, during field observations and focus group discussions survey data and assigned Likert scale values were screened and verified before statistical analyses. The variables used in this study are in different scales of measurement. In the descriptive analysis, the data were presented with their original scales of measurement, but to model the vulnerability to HEC according to the conceptual framework (Fig. 1) variables were transformed to a Likert scale in order to perform statistical analysis (Jebb et al. 2021).

The transcripts of the individual interviews and the FGD were used as qualitative data in the analysis. MAXQDA 2022 was used for coding the answers to open-ended questions. The categorization system of qualitative analysis followed the theoretical framework described above and illustrated in Fig. 1. The survey data were also transferred to an MS Excel datasheet before the quantitative statistical analysis. The data was transferred to code book for statistical analysis. Descriptive statistical results for each category of predictor and criterion variables are presented in the Table 7 in Appendix. Analysis was performed using SPSS statistical package version 22. Demographic data and the socioeconomic background of the sample were presented using descriptive statistical methods. A Likert scale was used to measure the independent and dependent variables, for each Likert scale item, a numerical value was assigned to each response option (see Tables 6 and 7 in Appendix). The values ranged from 1 to 5, with 1 representing ‘Strongly Disagree’ and 5 representing ‘Strongly Agree.’ Then Likert scaled items were summarized to each variable in the conceptual framework taking the average of each variable. Then, correlations of independent variables (socio-economic conditions, environmental change and land use, policy, awareness) to the dependent variable (vulnerability) were estimated using the chi-squared method. The chi-squared method is the most suitable statistical measure for assessing the correlation between categorical variables (Turhan 2020).

In this study we model the vulnerability to HEC by running a multiple regression analysis with correlated variables. Multiple regression analysis allows us to examine the impact of correlated independent variables on a dependent variable vulnerability of HEC. The significance of the multiple regression model was tested using the ANOVA test at the significance level of p = 0.05. Finally, the association between the level of vulnerability and the relationship between ‘occupation’ and ‘distance to nature’ were estimated using contingency tables.

The level of vulnerability for each GN division was estimated using a developed multiple regression model. However, the independent variables were measured using Likert scales on an ordinal scale, which consequently means that the level of vulnerability is also represented on an ordinal scale. This involved analyzing predictor variables, to predict the vulnerability level of each GN division. Following the regression analysis, the resulting vulnerability values were integrated into a geo-database, employing ArcGIS (version 10.4) (ESRI, Redlands, CA, USA) software. To facilitate the spatial representation of vulnerability across the study area, a two-step process was employed.

Firstly, the X and Y coordinates of the center points of the GN divisions were utilized for data integration within the geodatabase in ArcGIS environment. These center points served as spatial reference points for associating vulnerability values with their corresponding GN divisions.

Subsequently, the estimated vulnerability values were interpolated to cover the entire study area using the Kriging interpolation method (Perera et al. 2019). Kriging, a widely utilized geostatistical technique, leverages spatial autocorrelation by creating weighted predictions based on surrounding measured values (Liu and Miao 2018). This process enables the generation of a continuous raster representation of vulnerability to HEC over the study area, facilitating spatial analysis and visualization.

Figures 3, 4, 5, 6, 7, 8, 9 and 10 present the relevant results of the survey with regard to the research questions. Figure 3 displays clear evidence that respondents were primarily occupied as farmers, with 87% percent of respondents declaring ‘farmer’ to be their occupation, followed by ‘self-employed’ and ‘government-employed’ with 5% each and trailed by ‘unemployed’ (student, monk, no occupation) with 3%. Figure 4 indicates that more than half of the respondents (56.5%) had a mid-level education either up to Grade 8 or O-Levels, while 17.5% had an education up to A-Levels, and 22% had only low formal education up to maximum Grade 6. Only 4% of respondents had a higher education.

The impacts of adverse encounters with elephants are detailed in Fig. 5. Here, crop damage (83.6%) turns out to be the most prevalent consequence of encounters with elephants, with human damage (injury or death) a second (13.8%). Other kinds of damages are comparatively rare, such as damage to houses; damage to livestock was not recorded at all. The elephant-human conflict in affected area is felt more acutely due to crop damage, since the majority of the population is engaged in farming.

The type of housing (Fig. 6) is an important signifier of land use. 35.3% of respondents live in hereditary housing—houses that have been constructed on traditionally owned land that were granted or inherited by parents. 22.9% live in houses under the ‘Suvarnabhumi’ category—a form of government land grant (Marawila 2007). 17.7% of respondents said they lived in licensed (leased) houses, and 11.8% lived in new settlement houses under the Mahaweli resettlement scheme (Brun and Lund 2009), houses that were granted to the inhabitants but could not legally be sold or mortgaged. Only a small minority declared to live in informal (2.3%), reservation (1.4%), or temporarily licensed (0.3%) types of housing, while 8.3% of respondents have been filed under ‘other’. Most of the respondents had settled in accordance with the land laws with the legal permission of the government and there is no noticeable evidence that these respondents had encroached on forest areas for settlement or cultivation.

When asked about prevalent protection measures, responses were heterogeneous, reflecting different levels of engagement, knowledge of available strategies, and overall capacities to fend off elephant attacks. Figure 7 displays the different forms of reaction. Shouting (29.5%) and usage of fire (24.1%) make up the most common reactions. The planting of vegetation disliked by elephants, in particular limes (Santiapillai et al. 2010) and thorny shrubs, is a mid-term strategy employed by a number of respondents. Further measures include the digging of moats, the usage of bees, the use of elephant spells (ali mantra) (Köpke et al. 2023) and the use of an ancient cascade tank system (ellanga). Areas where these cascade tank systems (ellanga) are adequately safeguarded appear to be not as affected by HEC. This can be attributed to the fact that elephants do not traverse the natural barrier to reach the villages. However, in areas where the networked tank systems have been destroyed or compromised, the same level of security is not evident.

Figure 8 displays the assumptions or observations of respondents regarding the main cause of an increase in Human–elephant conflict over the last 20 years. Again, the responses are heterogeneous. Multiple answers were possible; among the most-cited causes, with between 12 and 13% each, respondents answered ‘elephants did not have enough food’, ‘drought’, (lack of proper construction of) ‘elephant fences’, and two causes specific to the respective locales, (noise from weapon training in the) ‘army training grounds’ and (the elephants being attracted by the) ‘garbage disposal areas near the village’. A frequent response was deforestation, either due to ‘development’ (6.1%), ‘illegal deforestation’ (4.2%) or ‘chena cultivation’ (5.8%)—a type of shifting cultivation. Another reason assumed by respondents was ‘water management issues’ (7%).

The above-mentioned reasons have affected the increase in Human–elephant conflict, among which the lack of proper maintenance of the elephant fence and the lack of food required by the elephants have led to the escalation of this conflict. Respondents also said that due to the construction of private electric fences, elephants’ usual paths (elephant corridors) were blocked and as a consequence, elephants moved through other lands. Accordingly, when planning elephant fences on private lands, it should be done with an understanding of the elephants’ route (see 3.4 below).

The presence of electric fences—seen as a largely effective mitigation mechanism for human–elephant conflict, despite some shortcomings—was a major concern. Less than a quarter of respondents (22.6%) claimed to possess a private electric fence, as shown in Fig. 9. Of these, the absolute majority (93.9%) had financed the fence through their own financial means, as presented in Fig. 10. Four persons (2.7%) had received assistance from a village association and 5 (3.4%) from the government; NGOs were not involved in the financing of private electrical fences.

In the geodatabase, vulnerability levels for GN divisions are delineated on an ordinal scale, with the center coordinates of villages serving as spatial reference points. Figure 11 provides a visual representation of the spatial distribution of vulnerability levels to human–elephant Conflict within the study area. This map illustrates several distinct patterns and hotspots.

Notably, the highest concentration of vulnerability to HEC is observed in the Kurunegala, Anuradhapura, and Puttalam Districts. These regions are closely situated to protected areas and agricultural lands. The proximity to these protected areas suggests frequent incursions by elephants into adjacent agricultural lands, thereby exacerbating vulnerability levels. This distribution underscores the significant role of resource competition between elephants and humans in exacerbating conflict dynamics.

Further analysis of the vulnerability map reveals that Puttalam, Anuradhapura, Kurunegala, Matale, and Polonnaruwa districts are significantly affected by elephant-human conflict, while Mannar and Vavuniya districts experience comparatively lower levels of vulnerability. These findings align with observations that elephant corridors often intersect human settlements, main roads, or railways, exacerbating conflict incidences.

Field observations additionally highlight the presence of mini-isolated forest patches that were once connected as larger contiguous forests. Some elephant populations inhabit these mini-isolated forest areas. Conversely, Mannar, Vavuniya, and Ampara Districts exhibit lower vulnerability to HEC, likely due to factors such as habitat suitability and reduced human-elephant interaction.

Overall, the spatial distribution of vulnerability to HEC underscores the complex interplay between ecological factors, human activities, and conflict dynamics, necessitating targeted interventions for mitigating conflict and promoting coexistence between humans and elephants.

In this section, we analyze the direct statements during interviews with individuals and FGD during the field study. To verify some of the factors, we collected information from the respective institutes and authorities. Local communities indicate that there are few reasons to exaggerate the elephant attacks on the village. One of the key reasons that were highlighted by villagers in Singahpur in Walikanda DS Division in Polonnaruwa District is the privately owned large-scale fruit gardens such as marketable fruit crops like mangoes. Though these lands are the property of the Mahaweli Authority, the lands are leased to private entrepreneurs. Per interview data from the officials, the Mahaweli Authority conducted the environmental feasibility study and the private companies submitted their own environmental impact assessment before the project was initiated. However, the FGD interview data revealed the increasing severity of the elephant attacks on the adjacent villages after this large-scale private fruit gardens were set up. The electric fences of these gardens are well-constructed. The attack into the garden is rare according to the information from the guards in a few gardens. Some villagers claim that some of the gardens obstruct the elephant corridors or are located in the natural elephant habitats. As a result of these new geo– and natural-structural changes, elephants enter the villages and damage farmers’ agricultural lands and house properties.

"Rich people from Colombo open up acres of land and cultivate it with fences. These were forests before. When the elephants lose their forests, they come to the surrounding villages and destroy the crops. This harms the poor." – a villager from Sinhapura.

Private electric fences are becoming popular among locals to deal with the elephant attacks on private properties. As can be seen above in Fig. 10, 10.8% of respondents point out that their individual properties (including living premises or agricultural lands) are surrounded by private electric fences. These fences are privately bought from companies. However, per our observation, this type of private fence is owned by only a few villagers in a village. Here, qualitative data serves to amend the statistical data.

In a village in Thirappane GS of ​​Anuradhapura district, and in Kallikulama Village and Omanthei in Vavuniyawa District, some people are building private electric fences to protect their houses and land from wild elephants. In the course of the construction process, elephant corridors are blocked. In reaction to this, wild elephants choose alternative routes when previous routes between forest habitats and lakes (for drinking) are blocked by these private electric fences. Those alternative routes pass through settled areas where private electric fences have not been erected, and here wild elephants destroy crops and damage houses and harm humans. While observing these villages, it became evident to the researchers that the posts of the government-financed elephant fences were damaged by insects and hence lacked proper maintenance. Accordingly, villagers may have built private elephant fences as a reaction to the state of disrepair public fences were in. However, some of the locals criticised this type of private electric fence, mentioning that most villagers are facing heightened vulnerability due to these fences.

“People who have money built electric elephant fences around their land. Some houses are across elephant corridors. When elephants’ usual roads are blocked by houses, elephants cross other roads. Elephants destroy those houses.” – a villager from Thirappane

“See what happened to my land, elephant attacked last month, I do not have any private electric fence, few neighbors they build their own fences. I do not have enough money. However, I see this is not as a solution. These individual fences created a small safety island in the village and elephants could not reach there, so elephants now attacked others heavily. We need a holistic solution for all village This is very selfish way.”

– a villager from Vavuniyawa, DS Kallikulama, Village Omanthei

Meanwhile, the absence of social activities in the evenings and at night has caused social isolation and impacted the social bonds of the people in the village. This has also affected the education of school-going students, who are unable to attend school in the morning and to return home in the evening. The areas heavily affected by HEC are perceived as difficult and isolated, which has led to hesitancy among people to organize events like weddings, funerals, and festivals.