A mixed-methods assessment of human-elephant conflict in the Western Okavango Panhandle, Botswana

1. Human-wildlife conflict, where interactions have negative impacts on both people and animals, is complex with underlying drivers and broad ecological and social impacts. From individual incidents and perceptions, to contemporary patterns and long-term trends, a range of information about human-wildlife conflict can help understand and manage challenges. However, many studies focus on a single data type or spatiotemporal scale. 2. In the Western Okavango Panhandle in Botswana, people in rural farming communities share and compete for resources with a growing African savanna elephant population. Few previous studies have focused on human-wildlife interactions in


| INTRODUC TI ON
Understanding, mitigating and managing human-wildlife interactions is a complex global challenge (IPBES, 2019). Increased pressure from human activities, coupled with shifts in wildlife distribution and behaviour linked to climatic and land use changes, has increasingly brought people and wildlife into closer proximity and competition for limited resources (Abrahms, 2021;Durant et al., 2022;Salerno et al., 2021). This can result in localized and widespread biodiversity loss, diminished human health and wellbeing, and decreased support for conservation initiatives (Kansky et al., 2014;Ngorima et al., 2020;Talukdar & Choudhury, 2020). Large terrestrial herbivores are especially prone to negative interactions with humans due to their substantial dietary and space-use requirements (Pascual-Rico et al., 2021;Ripple et al., 2015). This presents a particular challenge for many rural communities living in mixed-use landscapes in sub-Saharan Africa that include African savanna elephants Loxodonta africana.
African savanna elephants are a species of conservation concern that vary widely in their population size and co-occurrence with people (Gobush et al., 2020). Human-elephant interactions can have negative consequences for humans, and disproportionately impact the residents of socioeconomically vulnerable and under-resourced rural communities (Salerno et al., 2021). Foraging of cultivated crops by elephants, typically called "crop raiding", is a common type of HEC. Cultivated crops represent highly nutritious but risky food resources for elephants (Ahlering et al., 2011;Vogel et al., 2019) and can lead to property damage and life-threatening encounters between elephants and people. Here, we use the term "crop raiding" not to indicate malicious intent by elephants but rather to succinctly capture the negative impacts that crop foraging has for farmers. Not all encounters are antagonistic Buchholtz, Redmore et al., 2019), and indeed the concept of "convivial conservation" posits that peaceful cohabitation between people and wildlife in shared landscapes is possible (Buscher & Fletcher, 2020;Keil, 2016). However, human-elephant conflict (HEC) can undermine coexistence as a result of the direct and indirect adverse impacts to both humans (e.g. loss of income and earning potential, injuries and fatalities, food insecurity, increased exposure to disease and genderbased violence, Barua et al., 2013;Khumalo & Yung, 2015;Manoa et al., 2020;Mayberry et al., 2017;Nyumba et al., 2020;Salerno et al., 2020) and elephants (e.g. increased poaching and retaliatory killing, physiological stress that disrupts cognition and behaviour and lowers female reproductive success, disturbance to social structure, and loss of genes from large, reproductively-fit adults, Ahlering et al., 2011;Ball et al., 2022;Campbell-Staton et al., 2021;Chiyo et al., 2011;Compaore et al., 2020;Gobush et al., 2008;McComb et al., 2001;Shannon et al., 2013).
These considerable impacts of HEC mean that it is a social and ecological challenge that requires data at multiple scales to inform effective management and mitigation. Without knowledge and data related to human-wildlife interactions, it can be difficult to fully understand the system's patterns and drivers and how they may change over time (Gross et al., 2022;Tiller et al., 2021). HEC is often analysed using Western quantitative approaches such as estimating elephant population size, recording the number of conflict incidents, and measuring damaged area and economic losses (Deodatus & Lipiya, 1991;Prakash et al., 2020;Hoare, 1999;Pozo et al., 2018;Sukumar, 1991).
Methods for gathering additional data from local community perspectives, including questionnaires and interviews with people experiencing conflict, have also been incorporated in more recent studies (ex: Browne-Nuñez & Jonker, 2008;Ngorima et al., 2020;Sampson et al., 2019). Rural and Indigenous communities frequently have extensive observations of and experiences with local wildlife and ecological processes, and their knowledge of HEC dynamics over time may best be interpreted through interpersonal, qualitative methods (Gilchrist et al., 2005;Kohler et al., 2019). While their worldviews and lived experiences are valuable on their own, documenting and integrating Indigenous and local ecological knowledge within HEC research (a "multiple evidence base approach", Tengö et al., 2014), can provide a more robust understanding of HEC dynamics and support more effective mitigation (Meinecke et al., 2018).
Data that cover the relevant spatial and temporal scales to understand both current conflict conditions and long-term drivers is often lacking. Local knowledge can be highly effective in extending the temporal scope of quantitative studies by drawing on people's memories (Parker et al., 2007). Since HEC patterns and experiences can vary widely across spatial scales (Sitati et al., 2003;Wilson et al., 2015), dialogue with local communities can also provide a nuanced understanding of important fine-scale socioecological interactions Buchholtz, Redmore et al., 2019;. Additionally, interviews can initiate dialogue that may facilitate the development and implementation of participatory research, which can foster more locally relevant and impactful conservation and social outcomes (Fisher et al., 2021;Skroblin et al., 2021). 5. When possible, this mixed-methods approach may facilitate understanding for complex human-wildlife interactions and support the diverse communities and stakeholders involved with conflict-related challenges.

K E Y W O R D S
coexistence, crop raiding, human-elephant conflict, human-wildlife conflict, human-wildlife interactions, indigenous knowledge, local ecological knowledge Despite the advantages of this holistic approach, particularly in understudied areas, local ecological knowledge is often overlooked, and research that links empirical HEC measurements with people's experiences of HEC is relatively limited.
In the Okavango Delta of northwest Botswana, subsistence farming communities share and compete for resources with a growing population of African savanna elephants Buchholtz, Redmore et al., 2019;Chase et al., 2016;Songhurst et al., 2016;Thouless et al., 2016). The Okavango Delta is a key area for elephant conservation, but has also been identified as a HEC hotspot (Botswana Department of Environmental Affairs, 2008). While numerous studies of human-elephant interactions have been conducted in the Eastern Panhandle of the Delta, far less is known about HEC dynamics in the geographically-distinct Western Panhandle. There are limited empirical studies and longterm monitoring data quantifying HEC or its impacts on communities in the region that could inform evidence-based conservation and mitigation strategies.
We aimed to quantify and characterize HEC in the Western Panhandle using a mixed-methods approach. Our main objectives were to characterize current and long-term trends in HEC, specifically patterns in conflict timing, type, severity and spatial variation.
We achieved these objectives through combining data across spatial and temporal scales including incident reports, agricultural field surveys, and farmer interviews.

| Overview
Human-wildlife conflict (HWC) is a multifaceted challenge, and a variety of data exist to help us better understand the drivers, experiences and patterns of conflict. In this study, we focused on data from three sources that span different spatial and temporal scales: government reports, farmer interviews and field assessments. We used incident reports from the Botswana Department of Wildlife and National Parks (DWNP), which span the geographic extent of the Western Panhandle from 2008 to 2016, to quantify regional interannual HEC trends. We conducted semi-structured interviews with farmers about local HEC and used this knowledge to characterize HEC patterns; these surveys were geographically limited but encompassed farmers' memories and experiences from the 1990s to 2016. Finally, we used field assessments of elephant crop raiding incidents across eight villages during the 2016 growing season to analyse spatial patterns and conflict type and severity.
We expected these different data types to provide different information with varying levels of detail across spatial and temporal scales. Although the primary purpose of our study was to use these different data types to identify well-supported patterns of HEC, rather than use them to 'test' each other, we can use a conceptual diagram to understand their differences and better recognize how they can complement each other (Figure 1). People in the Panhandle predominantly live along the A35 road that F I G U R E 1 Conceptual diagram illustrating the potential for variation in geographic (spatial) extent, temporal extent, and level of detail provided by different data sources. These reflect the general patterns within our specific study area, but could be adapted and interpreted for other systems to highlight the way that different sources can complement each other, or where one source may be more relevant than another depending on the spatiotemporal extent and level of detail needed to meet objectives.

| Study area
roughly follows the banks of the Okavango River wetland ( Figure 2).
They range in size from 137 people (Kajaja) to 5234 people (Etsha 6; Statistics Botswana, 2012; Table 2). The Western Panhandle has one of the highest poverty levels in the country (World Bank Group, 2015), making people especially vulnerable to the economic impacts of HEC such as crop loss. Panhandle residents rely on subsistence agriculture and livestock for their nutrition and livelihoods (see Box 1).

| Human-elephant conflict incidents: Longitudinal data
We collated longitudinal data on HEC from records kept by the regional DWNP offices in Shakawe and Gumare. Types of incidents included direct and indirect crop damage, livestock injury or death, property damage (e.g. fences, water tanks (jojos), buildings), and serious human injuries or death. Farmers are responsible for reporting incidents to officers at the DWNP, who then assess the damage as time and resources allow and complete Field Reports. Field Reports contain varying levels of information and detail; summaries of this information are used to record incidents from Field Reports in cumulative Compensation and Occurrence Logs. We copied incident records from hand-written Field Report binders and Compensation and Occurrence Logs from 2008 (the earliest records available) to 2016 and transcribed them into a digital format. DWNP aims to provide compensation for incidents based on severity of incident and standardized rates.
We analysed this longitudinal data using linear regression to characterize annual trends in HEC incident numbers and location.
Since records varied in their level of precision for incident location (e.g., field description, ward name, village name), we attributed all incidents to the village-level for analyses and assessments of HEC variation among villages. These details were used to determine the relative location of HEC in the Panhandle (demarcated as north or south of the latitude of 18°69′56.01″S) and allowed us to investigate broad spatial HEC trends. Total cultivated area in 2016 was roughly equivalent in the north and the south regions of the Panhandle (north = 2,292,702 m 2 , south = 2,319,815 m 2 ).

| Human-elephant conflict incidents: 2016 data
Detailed data on reported elephant crop raiding incidents were collected during the agricultural growing season from February to June 2016. This data supplemented, but did not replace, DWNP

F I G U R E 2
The Western Panhandle of the Okavango Delta, Botswana, including regional towns and the eight 2016 focal villages for field surveys. Farmer interviews occurred in the village of Nxamasere.
assessments. Community enumerators from the eight study villages in the Panhandle were hired and trained through the non-governmental organization Ecoexist. Ecoexist is based in the Eastern Okavango Panhandle and is an NGO focused on "reducing conflict and fostering coexistence between elephants and people" (www.ecoex istpr oject. org). Farmers who experienced elephant crop raiding would report the incident to the community enumerator, local community council official, or police or DWNP officer. The enumerator would then visit the incident with co-author E.K. Buchholtz and field assistant S. Kerumotsemang to assess the damage. These assessments would take place as soon as possible after the incident was reported, often the same day, in order to get the most accurate measure of the damage and the conditions. Every reported incident was assessed, including if elephants raided the same fields on multiple occasions.
We collected data based on standardized methods described by Hoare (1999) and Songhurst (2017), and included: characteristics of the field; farmer; types of crops grown; types of protection or mitigation used; elephant-related damage; and any elephant-related demographics, in addition to a short narrative description of the incident provided by the farmer (see full data sheet in Supporting Information Table S1). We identified the sex of the raiding elephants when possible, using visual observations of elephants or spoor (elephant track) measurements.
We used a linear mixed model to examine what factors could explain variation in the amount of crop damage per reported incident.
We included field characteristics (field size; whether grains, cash crops, or both were grown; location in north or south of the region), mitigation and raid history (whether the field was guarded the night of the incident; if the field had been raided in 2015), and elephant group characteristics (male or female group; number of elephants). BOX 1 Subsistence agriculture is common in the Western Panhandle. Major crops include staple grains like sorghum, maize and millet, and cash crops such as watermelon and sweet reed. The growing season for crops primarily falls from November to May, although timing is rainfall-dependent. Fields may have a guard hut, and these are used either as a seasonal residence throughout the growing season or to sleep in intermittently while protecting the field from elephants. (a) Fields lie outside of the village and vary in size. Proximity to the Okavango Delta can put fields directly in the path of elephants travelling to the river. (b) Fences around fields to exclude free-ranging livestock, such as pole and wire fencing, are typically ineffective at excluding elephants. Electric fencing is rare. (c) When elephants enter fields, crops such as sorghum are damaged through consumption and trampling. (d) Crop loss can be exacerbated when elephants knock over fences, which allows livestock to enter and further consume and trample crops. Photo credits: E. K. Buchholtz 2016.
We used village as the random effect and let intercepts vary. We used the lme4 package to calculate the model and estimates (Bates et al., 2015) and the pbkrtest package to estimate likelihood-based p-values and degrees of freedom (Halekoh & Højsgaard, 2014). We also calculated summary statistics for field report data, including: location; number of previous raids; average field size; crop types damaged; total damage area; and elephant demographics. For all statistical analyses in this study, we used R v3.6.2 (R Core Team, 2021).

| Farmer interviews
We conducted semistructured interviews with farmers in the village of Nxamasere during June and July 2016. This village was selected due to its central location, ongoing rapport and trust with both the farmers and the dikgosi (chiefs/heads of villages), and reliable communication with the Village Development Council. The goal of the interview was to understand people's experiences farming, how elephants and crop raiding impact their farming, and their perception of the local elephant population in terms of size, demographics, and behaviour (full script in Supporting Information Table S2). The trend of increasing HEC incidents over time observed in DWNP reports was supported by farmers' interview responses, with most farmers responding that HEC levels had increased over the last 8-10 growing seasons (Figure 3b). Farmers primarily attributed the increase to a growth in the elephant population and changes in elephant behaviour (47.6%). One respondent said that "as time went on, elephants are more," and others said that "there are a lot more elephants now because of reproducing" and "elephants are reproducing at a higher rate." One respondent specifically stated that more elephants were coming from Namibia into the Panhandle since the end of the civil war. Some farmers believed that elephants had become less fearful over time, with one saying that "it used to be easier…you could just make noise and the elephants would leave the fields, but now they stay longer" and another saying that "now, even if I try to chase them out the elephants will just stay around the field like cattle". Some farmers attributed variation in HEC trends to rainfall levels. Generally, they indicated that years with lower or later rainfall and longer droughts resulted in a greater number and intensity of conflict incidents, with one farmer stated that "It [HEC] only depends on the rainfall." However, there was not a consensus on the magnitude or direction of how rainfall and drought impacted elephant presence and movement.

Increased 14
Due to increased numbers of elephants 5 Due to change in elephant behavior 5

Varies annually 6
Related to rainfall levels 2 Related to number of elephants 4 Related to protection of fields 1    Several farmers described changes in the demographics of the elephant groups that raided their crops (n = 13), reporting that the size of raiding herds was larger. One farmer said "even now there are family herds [raiding], when there used to be only bulls." Other farmers added that "families" and "groups" were coming more frequently, "even coming with their babies now", and that raiding herds "come sometimes in high numbers… sometimes more than five individuals".
Another farmer said that "…with families they [elephants] will even threaten your life…now elephants are in large numbers, and you can chase out a whole group but a new group can then come in." This farmer went on to say "The more elephants in a group, the more damage caused."

| Human-elephant conflict: Spatial variation
Between 2008 Table 2). Most raids were concentrated in the three northernmost villages of Nxamasere, Samochima, and Xhaoga, with nearly a quarter of raids occurring in Xhaoga. More fields incurred repeat raids in the north than the south, and a higher proportion of all raids in the north were repeat raids.
The long-term annual frequency and the proportion of fields  Table S4). In the 2016 field surveys, villages in the south experienced the maximum, second highest, and lowest proportion of fields raided on a per-village basis (100%, 90.00% and 53.33%). The percentage of fields raided in each village in the south had a greater variability than villages in the north (σ(south) = 0.20, σ(north) = 0.03).

| Farmer interviews: Incident reporting
Farmers who experienced crop raiding in 2016 reported the incident to DWNP and/or Ecoexist (27 reported to DWNP; 24 reported to Ecoexist; 20 reported to both). Of those that reported only to Ecoexist, we also informed DWNP. However, only one-quarter of farmers said that DWNP had assessed the reported damages by the time of our survey (25.93%, n = 7). Several farmers said that they were discouraged from reporting field damage because of inaction by the DWNP. One respondent said that "always the wildlife department takes a long time for them to come, and then again a long time to compensate". Another said " [I am] trying to report all the problems, but the DWNP responds late, so [I am] just giving up", and "because even with reports from DWNP the government doesn't do anything".

| DISCUSS ION
Human-wildlife conflict is a multidimensional issue, and in regions where data is limited, it can be difficult to form a comprehensive understanding of the conflict and its impacts. We used three different data types to characterize the complexity of HEC in the Western Okavango Panhandle of Botswana, which we found differs in timing, location, type and severity of conflict incidents. This work provides an example of HWC research that uses a mixed-methods approach to develop a retrospective characterization of long-term trends in an area with limited historical wildlife conflict and population data.
Drawing from a variety of data sources allowed us to build an understanding of past and current trends in HEC and elephant demographics that no single data source could provide.
HEC in the Panhandle, as reflected in reported incidents and farmers' experiences, significantly increased between 2008 and 2016. We found that reported HEC incidents nearly tripled in less than 10 years and farmers perceived conflict and crop raiding as escalating issues. These trends are important to know, because there is limited information for the sizes and interactions of human and elephant populations that could otherwise be used to infer conflict trends. Governmental records that date back to 2008 and span the spatial extent of the western Panhandle are therefore valuable for capturing regional trends. Farmers' experiences likewise indicate more frequent incidents, reflecting the increasing trend over time.
The increasing pattern of annual HEC highlights the growing challenge that conservation, management, and mitigation efforts must address.
Not all conflict incidents are equal, and it is important to consider the type, severity and frequency of conflict. Characterizing the variation in conflict incidents allows us to recognize the different  Lighter grey shading indicates villages in the northern region and darker grey shading indicates villages in the southern region.
outcomes for people and elephants associated with the conflict and to effectively mitigate and manage those impacts (Meinecke et al., 2018). Multiple farmers also reported that the challenge in trying to limit the damage caused by elephants has increased as elephants seem more desensitized to deterrents than they used to be. We did not find any statistical correlation between less damage done in fields that were actively guarded by farmers on the night of the incident.
There was not a strong correlation with the type of crop grown, either; between farmers' experiences and the 2016 field assessments, this suggests some of the primary ways that farmers can attempt to reduce elephant crop raiding damage may have limited effect.
A key finding from both the farmer interviews and field assessments was the relationship between crop damage and the type and size of elephant group involved. Smaller groups, often bachelor herds or single bulls, presented a chronic challenge, occurring more frequently but resulting in less damage on average. Crop raiding by larger, female family groups was less common, but more acute, with a higher likelihood of injuring or killing livestock in addition to damaging crops. Farmers in Nxamasere reported that they had experienced crop raiding every year, at least once or twice a season. In 2016, we found that 80% of the fields were raided at least once, with 52 fields incurring repeat raids during the season. Both chronic and acute impacts can reduce field and farmer resilience throughout the growing season and can lead to food insecurity and loss of income. Previous studies showing that the financial impacts of property damage and trampling of non-target crops can be more costly than the crops consumed (Inogwabini et al., 2013;Naughton-Treves, 1998). This is further exacerbated by our finding that most raids occurred in fields that had been previously raided within the last year.
Conflict and its impacts vary seasonally, as well as in severity and frequency. All three data sources highlighted the strong seasonality of HEC. Farmers said that crop raiding began when crops were at an intermediate stage or ripe. Crops may be more attractive to elephants than wild forage at this stage because of their high nutritional value, despite wild forage also maturing during the rainy season (Chiyo et al., 2005;Gross et al., 2018). If social or climatic factors alter crop planting, this might also shift crop ripening timelines and therefore HEC occurrence. To best anticipate upcoming HEC patterns, mitigation efforts should therefore take into account how farmers adapt and vary crop timing.
We found the number and severity of HEC incidents were not uniform across the Panhandle but varied spatially. Data from the DWNP and field surveys showed that while HEC was increasing, it did not impact villages equally within or between years. Incidents increased exponentially in the northern region between 2008 and 2016, while incidents had no statistically significant trend in the south, although the number of incidents was more variable year-to-year there.
Even within the northern and southern regions, there was variation in crop raiding frequency and intensity. Due to this variation, regionally-aggregated HEC levels could obscure local patterns necessary to manage different communities' needs. Spatial context for HEC has proved important in studying and predicting local patterns, for example in the configuration and connectivity of the landscape  and the distribution of fields and underlying anthropogenic and environmental variables (e.g. distance to water, presence of roads) on conflict incidents (Songhurst & Coulson, 2014 Gilchrist et al., 2005;Meijaard et al., 2011).
Lastly, the field assessment data from a single season provided more in-depth data for HEC incidents as they occurred. This allowed us to record actual incident frequency, which is not recorded by the DWNP, as well as specific variables about the field, damage, elephants, and farmers involved. Repeated efforts to collect this data annually would provide stronger statistical support for understanding the drivers of HEC (Songhurst et al., 2016). driving HEC is key to finding long term solutions to reduce HEC.
As human populations, land use, and other factors that influence where people and wildlife interact change over time, it is crucial to understand and mitigate negative impacts and conflict. By combining existing data sources with interviews and field assessments, we aimed to capture regional patterns in conflict dynamics and context.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare that there are no conflicts of interest.