Elephant space-use is not a good predictor of crop-damage

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 Elephant space use is not a good predictor of crop damage. Rocío A. Pozo1,2,3, Jeremy J. Cusack3, Graham McCulloch1,2,3, Amanda Stronza2,4, Anna Songhurst1,2,3 and Tim Coulson1 1 Department 2 Ecoexist of Zoology, University of Oxford, United Kingdom Project, Maun, Botswana 3 Biological and Environmental Sciences, University of Stirling, United Kingdom 4 Applied Biodiversity Science Program, Texas A&M University, United States of America Corresponding author: Rocío A. Pozo Biological and Environmental Sciences, University of Stirling Stirling FK9 4LA, UK +44 (0) 7850509716 rocio.pozo@stir.ac.uk 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 ABSTRACT Elephant crop-damage is a consequence of interactions between people and elephants that impact people’s livelihoods and biodiversity conservation efforts. Conflicts between people and elephants usually occur when there is overlap in elephant and human space-use leading to competition for resources. Therefore, understanding space-use patterns by elephants is key to alleviating negative human-elephant interactions. In the eastern Okavango Panhandle (Botswana), more than 16 000 people share resources with 18 000 elephants. Using data from 20 GPS-collared elephants, we investigated elephant space-use in relation to landscape variables during the day and night throughout the year and during the dry, wet and crop-damage seasons. We compared elephant space-use and crop-damage occurrence during the cropdamage seasons of 2014-2016. We found that elephant space-use was determined primarily by distance to waterholes and areas away from agricultural fields. However, predicting elephant space-use at the large scale was challenging. In particular, during the crop-damage season when the relationship between crop-damage events and elephant distribution was found to be non-linear. This revealed that areas that elephants frequently use might not be good indicators of the likelihood of crop-damage. Based on our findings, we suggest deterring elephants from peoples’ crops at the local scale is the most appropriate strategy for reducing elephant impacts on crops, alongside landscape scale interventions. We encourage future studies to use combinations of spatiotemporal methods, as well as practitioners to focus their efforts at the local scale, protecting elephant corridors, and supporting farmers to collaboratively work to decrease elephant crop-loss. KEYWORDS Human-wildlife conflict; crop-raiding; crop-loss; movement; satellite collar; Botswana. 1. INTRODUCTION Human-wildlife interactions have increased in the last decades as a result of expanding human activities and a concurrent loss of natural habitats for wildlife (Barnosky et al. 2011; Liu et al. 2003). Increased interactions between people and wildlife can result in competition for natural resources. Over time, this competition may develop into conflict between those affected by wildlife and groups wishing to conserve biodiversity (Redpath et al. 2015; Young et al. 2010). ‘Conservation conflicts’ are not only one of the most pressing challenges facing biodiversity conservation (Redpath et al. 2013; Woodroffe et al. 2005), but they are also hugely detrimental to the economic development and wellbeing of local people (Redpath et al. 2015; Thirdgood et al. 2005). Although numerous factors make each conflict unique (Redpath et al. 2013), a common aspect across conflicts is that species frequently move across human dominated landscapes, which highlights the fact that these conflicts are often spatial in nature. Thus, studying the space use of species is a key step in understanding how and why conflicts might occur, providing vital information for conservation strategies aimed at reconciling the interests of wildlife conservation and those of local stakeholders (Douglas-Hamilton et al. 2005; Treves et al. 2004). Crop-damage by elephants is a significant problem in many countries in Africa, affecting the local livelihoods of countless rural human populations 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 (Hoare 2000; Sitati et al. 2003). Elephant crop-damage (ECD) generally takes place during the night (Barnes et al. 2007; Jackson et al. 2008), and occurs particularly in isolated fields rather than in populated areas (Graham et al. 2010; Songhurst and Coulson 2014). Bulls have been more frequently recorded damaging crops (Hoare 1999; Sitati et al. 2003; Smit et al. 2017), because of their risk-taking behaviour to maximise reproductive success (Sukumar 1991), although cow-calf groups have also been observed to feed on crops (Sitati et al. 2003; Smith and Kasiki 1999). It is extremely dangerous for farmers to defend their fields because – unlike other species – elephants can cause severe harm, and even death, to people (Naughton-Treves et al. 1999; Thirgood et al. 2005). In retaliation for ECD, farmers will sometimes kill elephants, closing in this way a cycle of conflict affecting both humans and elephants. Negative perceptions and lethal retaliation as a result of ECD are some of the most important threats to elephant populations worldwide (Hoare 1999; Lamarque et al. 2009). The incidence of ECD has increased in recent decades partly because of the expansion of agricultural land into natural habitats. This increase is reflected in the number of studies describing the distribution and drivers of crop-damage (see Chiyo et al. 2005; Graham et al. 2010; Sitati et al. 2003, 2005; Songhurst and Coulson 2014). However, those studies that have focused on the spatial factors determining crop-damage incidence have not necessarily combined these insights with observed space-use by elephants. On the other hand, elephant movement studies have predominantly focused on ecological questions relating to home ranges, individual trajectories, and the speed of elephants in relation to specific landscape features such as water sources or protected areas (see Birkett et al. 2012; Loarie et al. 2009; Polansky et al. 2015). Few studies have used the whereabouts of elephants to better understand their spatial preferences and whether these relate to ECD locations at different times of the year (Graham et al. 2010; Jackson et al. 2008). Such a spatiotemporal approach may be far more useful to land-use planning efforts, as well as to better understand if the spatial overlap of elephant populations with people is a good indicator of the level of conflict with agricultural activities (Pozo et al. 2017a; Neumann et al. 2012). The eastern Okavango Panhandle (Botswana) is home to the largest unprotected populations of African elephants (Loxodonta africana) in the world (Chase et al. 2016). The size of this population has increased in recent years (DWNP 2013; Pozo et al. 2017a; Songhurst et al. 2016), having benefitted from the continuous supply of water from the Okavango River and a low level of poaching relative to neighbouring countries (Chase et al. 2016). Here, more than 18 000 elephants share resources with an increasing number of people (CSO 2011; Pozo et al. 2017a) throughout the dry (May-October) and wet (November-April) seasons. In this area, farmers plough their fields in fertile soils away from the Okavango River, close to which land is of less good quality for crops. Ploughing takes place at the beginning of the wet season, and the harvest happens between January and April, a period known as the cropping season (Songhurst and Coulson 2014). This is also the time when elephants are more likely to forage in fields, and thus this period is also known as the crop-damage season. To minimise ECD, a number of mitigation strategies have been implemented in recent years (Pozo at al. 2017b), such as land-use management and the identification and protection of elephant 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 corridors (Songhurst et al. 2015). All of these would benefit from a clearer understanding of elephant space-use in the region (Pozo et al. 2017a,b). In this study, we analysed and predicted elephant space-use patterns across the eastern Panhandle from Global Positioning System (GPS) telemetry data, and compared these to the distribution of contemporaneous crop-damage incidents. Our goal was first to characterise elephant distribution in the eastern Panhandle across seasons (dry, wet and crop-damage), as well as during the day and at night. Secondly, we used the distribution of reported crop-damage locations to identify vulnerable agricultural land and predict a risk map of crop-damage in the study area. Lastly, we combined both predictions (distribution of elephants and crop-damage locations) to determine the intensity of space-use by elephants at which crop-damage incidents are more likely to occur in the study area. 2. MATERIALS AND METHODS 2.1 Study area The 8 732 km2 non-protected study area is located in the eastern Okavango Panhandle, in northern Botswana. Deep Kalahari sands cover the majority of the region, with fertile soils away from the Okavango River. Vegetation cover is represented by mopane (Colophospermum mopane) and acacia (Acacia erioloba, Acacia tortillis) woodlands, and mixed marginal floodplain (Acacia nigrescens, Hyphaene petersiana) (Roodt 1998). The Delta has a continental climate with 360-500 mm annual rainfall during the wet season (NovemberApril) (Ramberg et al. 2006). Daily temperatures range from 25-35°C during the day to an average of 8°C during the night (Ramberg et al. 2006). The hottest month of the year is October, at the end of the dry season (MayOctober). The study area is delimited by the Namibian border to the north, the Okavango River to the southwest and the northern buffalo fence on the southeastern edge (Fig. 1). Elephants stay in this area due to the presence of artificial (veterinary fencing) and natural (the Okavango River) barriers. More than 16 000 people live in 13 villages along the Okavango River (CSO 2011; Fig. 1). Local farmers cultivate fields from the river edge up to 15 km inland. Ploughing takes place during the wet season and crops are harvested every year between April-June (Jackson et al. 2008). Most ECD events take place at the end of the wet season between January-April (Songhurst 2017). 2.2 Data collection Elephant relocations were collected by the Ecoexist Project from GPS collars (Iridium Vectronic) fitted to 10 females and 10 males in April 2014 (Appendix, Table A1). Each collar was set to give hourly GPS fixes. Individuals were selected using a spotter plane, and subsequently darted and immobilized from a helicopter. All collaring procedures were supervised by a veterinarian and performed under the research permit EWT 8/36/4 XVII (79) as well as affiliated immobilization permits. To reduce bias towards any specific area within the eastern Panhandle, individuals were selected from independent herds. Females were selected based on their body size (larger individuals were preferred) and on the age of their calf (>3 years) if any. All collared males were older than 20 years (Songhurst 2014). For this study, we considered data collected between April 2014 and April 2016. 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 Using Quantum Geographic Information System (QGIS version 2.12.1Lyon) and Google Earth (v 7.1.5.1557, 2016) images, we digitised layers pertaining to human land-use (i.e. settlements and agricultural fields), water sources (i.e. the Okavango River and waterholes) and elephant corridors. The resulting layers were ground-truthed using GPS locations taken on the ground as well as from aerial surveys and large-scale aerial photographs (1: 50 000). Elephant corridors were determined using ground monitoring, field vulnerability surveys and participatory land-use planning (Songhurst et al. 2015). In addition, we analysed crop-damage data collected in the study area during the same time as elephant movements were recorded. Crop-damage data was collected following Hoare (1999), by selecting local enumerators and training them to identify crop-damage events in each village. Enumerators visited damaged fields throughout the year, taking a GPS location and additional data on the scale of foraging damage at each crop-damaged plot (Songhurst 2017). 2.3 Data analysis 2.3.1 Elephant distribution We estimated a utilization probability distribution (UD; Worton 1995) for each of the 20 collared elephants using a Brownian bridge movement model (BBMM, hereafter; Horne et al. 2007) applied to the corresponding trajectory. The BBMM is a continuous-time stochastic model of movement in which the probability of being in an area is conditioned on the time between consecutive locations of an individual and its estimated mobility. It uses a sequence of time-specific location data, the estimated error associated with the location data, and a grid-cell size for the output UD (Sawyer et al. 2009) to estimate animal space-use. Given elephants move between 2-6 km a day in our study area (Loarie et al. 2009), we use grid-cells of 5 km2 to facilitate data analysis, provide realistic space use patterns and adequate mapping resolution. Assumptions of the BBMM are that location errors correspond to a bivariate normal distribution, and that relocations are not independent. Given that our GPS relocations are recorded every hour and that the assumption of normally distributed errors is appropriate for GPS data (Horne et al. 2007; Sawyer et al. 2009), our dataset fulfilled both BBMM requirements. Due to satellite failure, 8.4% of all fixes were collected with more than 1-hour interval. This was not considered a problem because BBMMs accounts for unequal time intervals between locations (Horne et al. 2007). We implemented BBMMs using functions in the R packages adehabitatHR and BBMM. Specifically, we used the function brownian.bridge to estimate UDs across the grid. The brownian.bridge function estimates the motion variance given an individual trajectory and a specified location error (here set to 100 m). In the following analyses, we consider UDs conditional on the corresponding 99% home range contours in order to avoid estimating utilisation probabilities in areas outside the study area (e.g. the western Panhandle). We first estimated four different UDs per individual elephant, each of these representing one of four temporal periods: all study years combined, dry (May-October), wet (November-April) and crop-damage (January-April) seasons. In our study area, elephants use space differently during the day and night (Loarie et al. 2009; Pozo et al. 2017b), therefore joint UDs for all 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 elephants (hereafter, population UD) during the day (06:00-18:00) and night (18:00-06:00) were created. These describe the probability of finding any of the collared elephants, in a given grid-cell and period of time. Joint UDs were obtained by summing individual UDs and re-scaling to sum to one. We repeated the same procedure across years to obtain a single population UD per season. This approach resulted in a single probability map for all individuals throughout the year, as well as during the day and night of the dry, wet and crop-damage seasons. All UDs were projected onto a 1 112-cell grid (i.e. excluding grid-cells not occupied by elephants), each with a specific probability of finding at least one individual from a particular category (e.g. elephants during the day in the dry season). We performed our statistical analysis in 2 stages. As a first step, we used generalised linear models (GLMs) with binomial errors to model the probability of cell use by each of the 20 collared elephants across all years, as well as during the dry, wet and crop-damage seasons. Our response in this case consisted of a binary variable indicating occurrence (1) or not (0) of each individual in each cell. Secondly, to determine the drivers of space-use intensity, we used beta regressions (Ferrari and Cribari-Neto 2004) to model population UD values (excluding cells with UD=0) against the distance to five landscape features: peoples’ settlements, agricultural fields, elephant corridors, waterholes and the Okavango River. However, we found high levels of multicollinearity among some of the explanatory variables (i.e. variance inflation factor, VIF > 5) and as a result our final models across seasons only considered distance to agricultural fields, elephant corridors and waterholes as predictor variables. We used the betareg function (betareg R package) applied to each population UD throughout the year and per season during the night and day (i.e. 7 models). For all seasons, Moran’s I test revealed significant spatial auto-correlation (Legendre and Legendre 1998) between neighbouring UD probabilities, and as a result we included a distanceweighted autocovariate obtained using the function autocov_dist (spdep R package) as an additional explanatory variable in all regressions (Chen et al. 2015; Dorman et al. 2007). Lastly, we predicted the expected distribution for the entire population of elephants (i.e. not only the 20 collared individuals) in each season. To do this, we used our full models (i.e. those including agricultural fields, corridors and waterholes) across years, and for each season (i.e. dry, wet and crop-damage). In doing this, we assumed all of the explanatory variables influenced elephant space-use. Because of the suspected variation in space-use across individual elephants, we additionally performed an analysis at the individual level. To do this, we used beta regressions to determine the relative influence of the distance to each of the three landscape variables on cell UD for collared females and males. To avoid bias due to spatial autocorrelation, we calculated and included a spatial autocovariate as an explanatory variable in each individual model. In both cases (population and individual analysis), we scaled continuous variables to a mean of zero and standard deviation of one prior to model implementation so that resulting coefficient estimates could be compared. 2.3.2 Crop-damage distribution We used crop-damage incidence data to: a) calculate a traditional fixed kernel 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 distribution (Calenge 2015) and contour isopleths of 50%, 20% and 10% of foraging incidents in the study area; and b) to predict a risk map of ECD occurrence for the eastern Panhandle. To compare the distribution of damage incidents with that of elephants, we mapped crop-damage data onto the same 1 112-cell grid as for the elephant space-use analysis. For each cell, we allocated a 0 when no foraging incidents had been recorded, and a 1 if they had (i.e. binary response). We applied a traditional kernel method via the kernelUD function in the adehabitatHR package in R to produce a distribution of crop-damage events across the study grid (Calenge 2011; Chen et al. 2015), and subset the 50%, 20% and 10% contours to identify areas that had been more affected by ECD. To predict crop-damage incidents we used generalised linear models (GLMs) with quasibinomial errors and a logit link function to regress the binomial crop-damage response against the same landscape features used for the elephant distribution analysis (i.e. distance to agricultural fields, elephant corridors and waterholes). Moran’s I test revealed spatial auto-correlation between ECD locations, and so we added an autocovariate as explanatory variable to the global model. Lastly, we used our full models to predict the likelihood of crop-damage occurring throughout the eastern Panhandle. To better understand the spatial relation between the likelihood of ECD events and elephant space-use, we used Generalised Additive Models (GAMs; Hastie and Tibshirani 1990) with Gaussian error structure and identity link function. For this analysis, we used the mgcv package in R to regress predicted ECD values against the predicted distribution of elephants during the crop-damage season. We used R (ver. 0.13.17) for all analyses. 4. RESULTS 4.1 Elephant distribution Elephant occurrence throughout the year was primarily determined by proximity to waterholes and increased distances to agricultural fields (Table 1). In addition, our most general model (i.e. all seasons at all times) showed that, within areas used by elephants, the intensity of space-use by the species was significantly determined by shorter distances to elephant corridors (Fig. 2; Table 2). Despite providing an overview of the occurrence and use of space by collared elephants in our study area, both general models (all season at all times; Tables 1 & 2) provided little information on seasonal processes. This is the reason why we also investigated each season separately. Without exception, all models (i.e. across seasons and during the day and night times) showed a negative effect of distance to waterholes on elephant occurrence, and a positive effect of distance to agricultural fields (Table 1). Similarly, all models regardless of the season showed a negative effect of distance to corridors on the presence of elephants (Table 1). However, the effect of distance to elephant corridors on the occurrence of the species was only significant during the crop-damage season (Table 1; cropdamage season, day & night). The dry season analysis showed that elephants spent time in areas close to waterholes throughout the day and night (Fig. 2; Table 1). During the driest months of the year (May-October), the intensity of elephant space-use was restricted to areas close to water sources in comparison to other seasons, with highly used sites more likely to be ‘inter-connected’ via pathways (Fig. 2). 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 During the wet season (November-April), our analysis showed that the space-use patterns of collared elephants were more dispersed across the study area than during the dry season (Fig. 2). Our GLMs and beta regression models revealed elephants prioritised areas close to waterholes throughout the wet season, and collared individuals only stayed away from agricultural land during the day (Table 1) and used corridors more intensively at night (Table 2). Lastly, during the crop-damage season (January-April), elephants showed a clear spatial avoidance of the north-eastern section of the study area, as well as of some of the villages along the Okavango River (Fig. 2). The local elephant population showed significant spatial preferences for areas closer to waterholes and corridors, and away from agricultural fields during the day and night (Table 1). However, our analysis suggested an absence of effect of any of the landscape features on intensity of use during the months of the crop-damage season (Table 2). At the individual level our analysis showed a significant variation in space-use patterns across elephants, but a non-significant difference between females and males (Appendix, Fig. A1). In general, across all seasons, females and males stayed closer to waterholes and corridors, but females seemed to use corridors during the dry, wet and crop-damage seasons more intensively than males (Appendix, Fig. A1). Although we found that both females and males were more likely to use areas close to agricultural land at night-time during the dry and wet seasons, respectively (Appendix, Fig. A1), we did not find significant trends for proximity to fields during the crop-damage season. On the contrary, females and males seemed to avoid agricultural land during the day and at night during the busiest harvesting months (Appendix, Fig. A1). 4.2 Crop-damage distribution The analysis of risk of ECD showed that more than 50% of recorded cropdamage events took place in the north-west and southern parts of our study area (Fig. 3). Our risk map predicted that agricultural land close to Tobera and Mohembo-East, as well as fields between Gunotsoga and Beetsha represented areas of high risk (Fig. 3). The GLM analysis showed that fields close to waterholes (-3.8621 ± 0.89, P < 0.001) were more likely to be damaged by elephants, and that distance to corridors did not have a significant effect on the likelihood of elephants damaging crops (-0.0355 ± 0.19, P > 0.05). Our GAMs revealed a non-linear effect of predicted elephant distribution on crop-damage events during the day (Deviance explained = 29.6%; N=1112; P < 0.001) and at night (Deviance explained = 33.2%; N=1112; P < 0.001). The non-linear association between elephant distribution and crop-damage incidence during the night (smooth term = 7.185; F = 66.04; P < 0.001; Fig. 4) was stronger than the effect during the day-time (smooth term = 5.634; F = 67.42; P < 0.001; Fig. 4). This translates as a higher likelihood of ECD in areas with intermediate to low levels of elephant spaceuse intensity during the night. 5. DISCUSSION Our findings are in agreement with previous research showing that the use of space by elephants is influenced by water availability (Jackson et al. 2008; 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 Loarie et al. 2009; Polansky et al. 2015) and the presence of corridors as a risk avoidance strategy in proximity to people’s settlements (DouglasHamilton 2005; Graham et al. 2009; Songhurst et al. 2015). During the cropdamage season, collared individuals in the eastern Panhandle stayed away from people’s fields. To this end, we developed a crop-damage risk map to better understand the relative likelihood of damage incidents and its relation to landscape variables in the study area, and we compared this with the intensity of elephant space-use measured over the same time period. We found elephant space-use to be correlated in a non-linear way with the occurrence of crop-damage events during the day and night throughout the crop-damage season. Our study shows that elephant distribution in the eastern Panhandle is challenging to predict at the population level because of the high variation across individuals. Although observed patterns were primarily related to the availability of water, the presence of elephant corridors, and the absence of people’s fields, our analysis explained comparatively little variance in elephant space-use. Equally, the non-linearity in the relationship between elephant space-use and the occurrence of crop-damage events signifies that the intensity of space-use by elephants is not a good predictor of the likelihood of crop-loss. Although our models explained a small amount of variation, they suggest that throughout the year elephant distribution is primarily determined by proximity to waterholes and areas away from agricultural fields. We expected elephants to be away from waterholes during the dry season when there is little water available in them, and close to them during the wet season when more water resources are available away from the Okavango River. In addition, since most human settlements in our study area are close to the river, we also expected elephants to show a preference for waterholes during the wet and the crop-damage seasons to minimise contact with human activities. Nevertheless, our study agrees with previous work suggesting elephants rely primarily on water sources throughout the year, and that they avoid the risk associated with human inhabited areas by adapting their spatial behaviour (Douglas-Hamilton et al. 2005; Sitati et al. 2003). Elephant space-use during the dry season hinted at a reliance on “routes” connecting highly used areas. This may arise from pre-existing knowledge of reliable water and foraging sources during the driest months (Polansky et al. 2015). Moreover, owing to a drop in temperature during the wet season, elephants may be able to cover greater distances during the day and not only move during the coolest hours as they usually do during the dry season (Wittemyer et al. 2008). These contrasting patterns of elephant spaceuse between the dry and wet seasons are consistent with previous data collected in the eastern Panhandle and elsewhere in Africa (Birkett et al. 2012; Jackson et al. 2008; Loarie et al. 2009). Consequently, future land-use planning should aim to maintain connectivity between key sources of water in the region (Goswami and Vasudev 2017). We found marginally different patterns of elephant space-use during the night than during daylight hours, which could be a consequence of elephants moving at times of lower temperatures in the dry season. However, because this pattern was present throughout the year, we argue this behaviour may also be the result of elephants prioritising times of low human activity (Douglas-Hamilton et al. 2005; Hoare and Du Toit 1999; Songhurst et 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 al. 2015). Such a risk-avoidance strategy, whereby elephants use the cover of darkness, has been demonstrated previously. For instance, Graham et al. (2009) showed elephants in Tanzania and Kenya were more active at night outside protected areas. It is interesting to find free-ranging elephants using the same strategies in an unprotected area such as the eastern Panhandle. Our individual-level analysis showed that, during the crop-damage season, elephants stayed close to waterholes, tended to stay away from people’s fields and used corridors significantly more than during any other time of the year. This contradicts our expectation of finding a strong effect of proximity to fields on elephant space-use. A possible explanation for this behaviour might be that none of the 20 collared elephants feed on crops, and they remain in areas further away from human activities. However, we found this to be unlikely because of the strong correlation our results showed between the likelihood of ECD and the distribution of the local population during the crop-damage season. We, therefore, think that if any of the collared individuals in our study feed on crops, they probably do it opportunistically and for short periods of time, which cannot be represented as trends in our individual space-use analysis. In addition, elephants are probably using corridors more intensively to move away from areas of high risk during the harvesting months, as has been found in previous studies in the same area (Songhurst et al. 2015). An alternative explanation to the field-avoidance behaviour might be that elephants are also using the higher availability of natural resources away from fields exclusively during the crop-damage season, which are less risky to forage on than are crops from fields. This is also supported by our results in the dry and wet seasons when elephants occurred more often close to agricultural land. It is important to emphasise that we were unable to account for additional environmental variables, such as vegetation types and rainfall across seasons. Given that vegetation and crop growth in northern Botswana rely on seasonality (i.e. variation on temperature, rainfall, etc.), our findings should be taken with some caution. Nevertheless, they provide a much-needed starting point to understand elephant space-use patterns and their relation with the levels of crop-damage in the region. In this context, our study suggests that the relationship between the distribution of crop-damage and elephant space-use is non-linear. We found that ECD had a higher probability to occur within a low intensity of elephant space-use, rather than at the highest levels. This likely reflects that ECD occurs opportunistically when elephants transit through highly used sites (e.g. water sources, vegetation patches) across the study area. This finding could have consequences for future mitigation strategies to reduce ECD. Indeed, studies that only consider elephant distribution as a direct indicator of ECD make the underlying assumption that mitigation should be targeted at areas that are most highly used by the species. Such a strategy would overlook the pathways between highly used areas, where elephants might opportunistically damage crops. Our study highlights the importance of considering elephant and ECD distributions when focusing mitigation efforts across space, and validates the protection of elephant corridors in the eastern Panhandle as a strategy to minimise ECD. Despite finding that certain landscape covariates determined elephant space-use better than others, predicting the use of space by the species 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 remained challenging, particularly during the crop-damage season (Pittiglio et al. 2014; Sitati et al. 2003). Our results revealed spatial variation in the distribution of crop-damage based on ECD events, but we did not find strong evidence that elephants were attracted to people’s crops. Moreover, high variation in space-use patterns across individual elephants suggests larger sample sizes would be needed to reliably predict crop-damage. Our findings also support the idea that to effectively manage ECD and inform appropriate land-use planning interventions it is important to focus on deterring elephants from people’s crops at the local scale alongside understanding elephant space-use patterns (Pozo et al. 2017b). 6. CONCLUSION Our study shows that elephant distribution in the eastern Panhandle is primarily determined by the proximity to waterholes and areas away from agricultural land. Elephants tended to stay away from people’s fields and close to elephant corridors during the crop-season; suggesting crop-damage may be an opportunistic behaviour. This was also reflected in a non-linear, negative relationship between the likelihood of ECD and elephant space-use intensity. Thus, we concur with previous studies that high elephant space-use may not be a good indicator of the likelihood of ECD. We highlight that a combination of different spatiotemporal methods, such as those used in this study, represents an integrated and more reliable approach to better understand crop-damage dynamics and tackle ECD issues. This will help prevent negative human-elephant interactions and perceptions developing into conflicts. 7. ACKNOWLEDGEMENTS We are thankful to the Government of Botswana for granting permission for this study and to the Ecoexist Project for the use of their data. This project was funded by the Ecoexist Project and the National Commission for Scientific and Technological Research (CONICYT, Chile). In addition, financial support was granted by Kellogg College and the Department of Zoology (University of Oxford). We are also thankful of the ConFooBio Project (University of Stirling) for its support during the preparation of our manuscript. 8. REFERENCES Barnes, R.F.W., Dubiure, U.F., Danquah, E., Boafo, Y., Nandjui, A., Hema, E.H. and Manford, M. (2007). Crop- raiding elephants and the moon. African Journal of Ecology 45, 112–115. Barnosky, A.D., Matzke, N., Tomiya, S., Wogan G.O.U., Swartz, B., Quental, T.B., Marshall, C., McGuire, J.L., Lindsey, E.L., Maguire, K.C., Mersey, B. and Ferrer, E.A. (2011). Has the Earth’s sixth mass extinction already arrived? Nature 471, 51-57. Birkett, P.J., Vanak, A.T., Muggeo, V.M.R., Ferreira, S.M. and Slotow, R. (2012). Animal perception of seasonal thresholds: changes in elephant movement in relation to rainfall patterns. PLOS ONE 7, e38363. Calenge, C. (2011). Exploratory analysis of the habitat selection by the wildlife in R: the adehabitatHS package. Office national de la chasse et de la faune sauvage Saint Benoist 78610. Auffargis, France. 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 Calenge, C. (2015). Home range estimation in R: the adehabitatHR package. Office national de la chasse et de la faune sauvage Saint Benoist 78610. Auffargis, France. Chase, M., Schlossberg, S., Griffin, C.R., Bouché, P.J.C., Djene, S.W., Elkan, P.W., Ferreira, S., Grossman, F., Kohi, E.M., Landen, L., Omondi, P., Peltier, A., Selier, S.A. J. and Sutcliffe, R. (2016). Continent-wide survey reveals massive decline in African savannah elephants. Peer J doi:10.7717/peerj.2354. Chen, Y., Marino, J., Chen, Y., Tao, Q., Sullivan, C.D., Shi, K. and Macdonald, D.W. (2015). Predicting hotspots of human-elephant conflict to inform mitigation strategies in Xishuangbanna, southwest China. PLOS ONE 11(9): e0162035. doi:10.1371/journal. pone.0162035. Chiyo, P.I., Cochrane, E.P., Naughton, L. and Basita, G.I. (2005). Temporal patterns of crop raiding by elephants: a response to changes in forage quality or crop availability? African Journal of Ecology 43, 48-55. CSO (2011). Population and housing census, Ministry of Finance and Development Planning, Central Statistics Office (CSO), Government of Botswana. Dorman, C.F, McPherson, J.M., Araújo, M.B., Bivand, R., Bolliger, J., Carl, G., Davies, R.G., Hirzel, A., Jetz, W., Kissling, W.D., Kühn, I., Ohlemüller, R., Peres-Neto, P.R., Reineking, B., Schröder, B., Schurr, F.M. and Wilson, R. (2007). Methods to account for spatial autocorrelation in the analysis of species distributional data: a review. Ecography 30. 609-628. doi: 10.1111/j.2007.0906-7590.05171.x. Douglas-Hamilton, I., Krink, T. and Vollrath, F. (2005). Movements and corridors of African elephants in relation to protected areas. Naturwissenschaften 92, 158-163. DWNP (2013). Aerial census of animals in northern Botswana, Dry season 2013. Department of Wildlife and National Parks (DWNP), Government of Botswana Gaborone, Botswana. Ferrari, S.L.P. and Cribari-Neto, F. (2004). Beta regression for modeling rates and proportions. Journal of Applied Statistics 31, 799-815. Google Earth version 7.1.5.1557, May 2016. www.rmit.edu.au/library/learningrepository Goswami, V.R. and Vasudev, D. (2017). Triage of conservation needs: the juxtaposition of conflict mitigation and connectivity considerations in heterogeneous, human-dominated landscapes. Frontiers in Ecology and Evolution 4:144. doi: 10.3389/fevo.2016.00144. Graham, M.D., Douglas-Hamilton, I., Adams, W.M. and Lee, P.C. (2009). The movement of African elephants in a human-dominated land-use mosaic. Animal Conservation 12, 445-455. Graham, M.D., Notter, B., Adams, W.M., Lee, P.C. and Ochieng, T.N. (2010). Patterns of crop-raiding by elephants, Loxodonta africana, in Laikipia, Kenya, and the management of human–elephant conflict. Systematics and Biodiversity 8, 435–445. Hastie, T.J. and Tibshirani, R.J. (1990). Generalized Additive Models. Chapman & Hall, London. Hoare, R.E. (1999). Determinants of human-elephant conflict in a land use mosaic. Journal of Applied Ecology 36, 689-700. 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 Hoare, R.E. (2000). African elephants and human conflict: the outlook for coexistence. Oryx 34, 34-38. Hoare, R.E. and du Toit, J.T. (1999). Coexistence between people and elephants in African savannas. Conservation Biology 13, 633-639. Horne, J.S., Garton, E.O., Krone, S.M. and Lewis, J.S. (2007). Analyzing animal movements using Brownian bridges. Ecology 88, 2354-2363. Jackson, T.P., Mosojane, S., Ferreira, S.M. and van Aarde, R.J. (2008). Solutions for elephant Loxodonta africana crop raiding in northern Botswana: moving away from symptomatic approaches. Oryx 42, 83-91. Lamarque, F., Anderson, J., Fergusson, R., Lagrange, M., Osei- Owusu, Y. and Bakker, L. (2009). Human-wildlife conflict in Africa. Causes, consequences and management strategies. Rome, FAO. Legendre, P. and Legendre, L. (1998). Numerical ecology. Developments in Environmental Modelling 20. Elsevier, Amsterdam, the Netherlands. Liu, J., Daily, G.C., Ehrlich, P.R. and Luck, G.W. (2003). Effects of household dynamics on resource consumption and biodiversity. Nature 421, 530-533. Loarie, S.R., Van Aarde, R.J. and Pimm, S.T. (2009). Fences and artificial water affect elephant movement patterns. Biological Conservation 142, 3086-3098. Naughton, L., Rose, R. and Treves, A. (1999). The social dimensions of human-elephant conflict in Africa: a literature review and case studies from Uganda and Cameroon. A report to the African elephant specialist, humanelephant conflict task force, International Union for the Conservation of Nature (IUCN), Glands, Switzerland. Neumann, W., Ericsson, G., Dettki, H., Bunnefeld, N., Keuler, N.K., Helmers, D.P. and Radeloff, V.C. (2012). Difference in spatiotemporal patterns of wildlife road-crossings and wildlife-vehicle collisions. Biological Conservation 145, 70-78. Pittiglio, C., Skidmore, A.K., van Gils, H.A.M.J., McCall, M.K. and Prins, H.H.T. (2014). Smallholder farms as stepping stone corridors for cropraiding elephant in northern Tanzania: integration of Bayesian expert system and network simulator. AMBIO 43, 149-161. doi 10.1007/s13280013-0437-z Polansky, L., Kilian, W. and Wittemyer, G. (2015). Elucidating the significance of spatial memory on movement decisions by African savannah elephants using state–space models. Proceedings Royal Society 282: 20143042. http://dx.doi.org/10.1098/rspb.2014.3042. Pozo, R.A., Coulson, T., McCulloch, G., Stronza, A. and Songhurst, A. (2017a). Determining baselines for human-elephant conflict: a matter of time. PLOS ONE 12(6): e0178840. https://doi.org/ 10.1371/journal.pone.0178840. Pozo, R.A., McCulloch, G., Stronza, A., Coulson, T. and Songhurst, A. (2017b). Chilli-briquettes modify elephant temporal behaviour but not numbers. Oryx doi:10.1017/S0030605317001235. Quantum GIS Development Team (2016). Quantum GIS Geographic Information System. Open Source Geospatial Foundation Project. http://qgis.osgeo.org R Core Team (2018). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/. 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 Ramberg, L., Hancock, P., Lindholm, M., Meyer, T., Ringrose, S., Sliva, J., Van As, J. and VanderPost, C. (2006). Species diversity of the Okavango Delta, Botswana. Aquatic Sciences 68, 310-337. Redpath, S.M., Gutiérrez, R.J., Wood, K.A., Sidaway, R. and Young, J.C. (2015). An introduction to conservation conflicts. In: Redpath, S.M., Gutiérrez, R.J., Wood, K.A., and Young, J.C. (Eds.), Conflicts in conservation navigation towards solutions. Cambridge University Press, Cambridge, pp. 3-18. Redpath, S.M., Young, J., Evely, A., Adams, W.M., Sutherland, W.J., Whitehouse, A., Amar, A., Lambert, R.A., Linnell, J.D.C., Watt, A. and Gutiérrez, R.J. (2013). Understanding and managing conservation conflicts. Trends in Ecology and Evolution 28, 100-109. Roodt V. (1998). Trees and shrubs of the Okavango Delta: medicinal uses and nutritional value. In: Lahm Shell Field Guide Series, Part I. Sawyer, H., Kauffman, M.J., Nielson, R.M. and Horne, J.S. (2009). Identifying and prioritizing ungulate migration routes for landscape-level conservation. Ecological Applications 19, 2016-2025. Sitati, N.W., Walpole, M.J., Smith, R.J. and Leader-Williams, N. (2003). Predicting spatial aspects of human-elephant conflict. Journal of Applied Ecology 40, 667-677. Sitati, N.W., Walpole, M.J. and Leader-Williams, N. (2005). Factors affecting susceptibility of farms to crop raiding by African elephants: using a predictive model to mitigate conflict. Journal of Applied Ecology 42, 11751182. Smit, J., Pozo, R., Cusack, J., Nowak, K. and Jones, T. (2017). Using camera traps to study the age–sex structure and behaviour of crop-using elephants Loxodonta africana in Udzungwa Mountains National Park, Tanzania. Oryx, doi:10.1017/S0030605317000345. Smith, R.S. and Kasiki, S.M. (1999). A spatial analysis of human–elephant conflict in the Tsavo Ecosystem IUCN African Elephant Specialist Group Report. Songhurst, A. (2014). Ecoexist report on elephant collaring exercise Eastern and Western Panhandle, Ecoexist Project, Botswana. Songhurst, A. (2017). Measuring human-wildlife conflicts: comparing insights from different monitoring approaches. Wildlife Society Bulletin. doi:10.1002/wsb.773. Songhurst, A. and Coulson, T. (2014). Exploring the effects of spatial autocorrelation when identifying key drivers of wildlife crop-raiding. Ecology and Evolution 4, 582-593. Songhurst, A., McCulloch, G. and Coulson, T. (2015). Finding pathways to human–elephant coexistence: a risky business. Oryx. doi:10.1017/S0030605315000344. Songhurst, A., McCulloch, G. & Stronza, A. (2016). Ecoexist project end year 3 report, Ecoexist Trust, Botswana. Sukumar, R. (1991). The management of large mammals in relation to male strategies and conflict with people. Biological Conservation 55, 93-102. Thirdgood, S., Woodroffe, R. and Rabinowitz, A. (2005). The impact of human-wildlife conflict on human lives and livelihoods. In: Woodroffe, R., Thirdgood, S., and Rabinowitz, A. People and wildlife conflict or coexistence? Cambridge University Press, Cambridge, pp. 13-26. 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 Treves, A., Naughton-Treves, L., Harper, E.K., Mladenoff, D.J., Rose, R.A., Sickley, T.A. and Wydeven, A.P. (2004). Predicting Human-Carnivore Conflict: a Spatial Model Derived from 25 Years of Data on Wolf Predation on Livestock. Conservation Biology18, 114-125. Wittemyer, G., Polansky, L., Douglas-Hamilton, I. and Getzb, W.M. (2008). Disentangling the effects of forage, social rank, and risk on movement autocorrelation of elephants using Fourier and wavelet analyses. PNAS 105, 19108-19113. doi/10.1073/pnas.0801744105. Woodroffe, R., Thirgood, S. and Rabinowitz, A. (2005). The impact of humanwildlife conflict on natural systems. In: Woodroffe, R., Thirdgood, S., and Rabinowitz, A. People and wildlife conflict or coexistence? Cambridge University Press. pp.1-12. Worton, B.J. (1995). Using Monte-Carlo simulation to evaluate kernel-based home- range estimators. Journal of Wildlife Management 59, 794-800. Young, J., Marzano, M., White, R.M., McCracken, D.I., Redpath, S.M., Carss, D.N., Quine, C.P. and Watt, A.D. (2010). The emergence of biodiversity conflict from biodiversity impacts: characteristics and management strategies. Biodiversity Conservation 19, 3973-3990. 750 751 752 753 754 755 756 757 758 Season 9. TABLES Table 1. Summary of generalized linear models (GLMs) for elephant occurrence at all seasons, and during the dry, wet and crop-damage seasons. Each model shows elephant binomial response (i.e. presence or not) as a function of the effect of distance to waterholes, elephant corridors and agricultural fields; as well as its respective distance-weighted autocovariate. The period column shows models including data collected 24 hours a day, during the day (6:00-18:00) and night (18:00-6:00). Period all all times dry day dry night wet day wet night cropdamage day cropdamage night Binomial GLMs Explanatory variable Estimate intercept 6.453 distance to waterholes -1.160 distance to corridors 0.132 distance to fields 0.533 autocovariate 9.571 intercept -0.820 distance to waterholes -0.673 distance to corridors -0.237 distance to fields 0.265 autocovariate 2.217 intercept -0.894 distance to waterholes -0.596 distance to corridors -0.124 distance to fields 0.266 autocovariate 1.687 intercept 1.043 distance to waterholes -0.620 distance to corridors -0.180 distance to fields 0.414 autocovariate 2.695 intercept 1.018 distance to waterholes -0.691 distance to corridors -0.242 distance to fields 0.201 autocovariate 1.983 intercept 0.134 distance to waterholes -0.472 distance to corridors -0.663 distance to fields 0.903 autocovariate 1.925 intercept -0.012 distance to waterholes -0.566 distance to corridors -0.822 distance to fields 0.868 autocovariate 1.234 Std. Error 0.548 0.237 0.217 0.186 0.898 0.104 0.189 0.172 0.117 0.224 0.103 0.199 0.161 0.119 0.184 0.121 0.143 0.148 0.110 0.270 0.110 0.144 0.146 0.107 0.225 0.093 0.160 0.181 0.130 0.167 0.080 0.154 0.170 0.121 0.122 z-value 11.782 -4.900 0.608 2.868 10.664 -7.901 -3.562 -1.372 2.264 9.901 -8.713 -2.994 -0.767 -2.235 9.167 8.652 -4.326 -1.216 3.777 9.989 9.263 -4.795 -1.658 1.876 8.811 1.438 -2.948 -3.669 6.931 11.524 -0.144 -3.673 -4.848 7.187 10.104 Pr(>|z|) < 2e-16 *** 0.000 *** 0.543 0.004 ** < 2e-16 *** 0.000 *** 0.000 *** 0.170 0.024 * < 2e-16 *** < 2e-16 *** 0.003 ** 0.443 0.254 * < 2e-16 *** < 2e-16 *** 0.000 *** 0.224 0.000 *** < 2e-16 *** < 2e-16 *** 0.000 *** 0.097 . 0.061 . < 2e-16 *** 0.150 0.003 ** 0.000 *** 0.000 *** < 2e-16 *** 0.886 0.000 *** 0.000 *** 0.000 *** < 2e-16 *** 759 760 761 762 763 764 765 766 Table 2. Summary of beta regression models for the intensity of elephant space-use at all seasons, and during the dry, wet and crop-damage seasons. Each model shows the elephant population UDs as a function of the effect of distance to waterholes, elephant corridors and agricultural fields; as well as its respective distance-weighted autocovariate. The period column shows models including data collected 24 hours a day, during the day (6:00-18:00) and night (18:00-6:00). Beta Regression Models Season Period all all times dry day dry night wet day wet night cropdamage day cropdamage night Explanatory variable intercept distance to waterholes distance to corridors distance to fields autocovariate intercept distance to waterholes distance to corridors distance to fields autocovariate intercept distance to waterholes distance to corridors distance to fields autocovariate intercept distance to waterholes distance to corridors distance to fields autocovariate intercept distance to waterholes distance to corridors distance to fields autocovariate intercept distance to waterholes distance to corridors distance to fields autocovariate intercept distance to waterholes distance to corridors distance to fields autocovariate Estimate -7.098 -0.080 -0.170 0.062 0.422 -6.155 0.015 -0.181 -0.112 0.170 -6.238 -0.045 -0.171 -0.047 0.207 -6.708 -0.094 -0.123 0.040 0.253 -6.767 -0.002 -0.147 0.021 0.292 -6.438 -0.052 -0.126 0.108 0.167 -6.396 0.052 -0.148 0.078 0.133 Std. Error 0.029 0.057 0.050 0.032 0.013 0.082 0.140 0.110 0.071 0.025 0.078 0.142 0.095 0.070 0.023 0.043 0.077 0.072 0.046 0.020 0.039 0.073 0.064 0.042 0.019 0.052 0.084 0.092 0.064 0.028 0.052 0.088 0.096 0.062 0.031 z-value -243.200 -1.416 -3.386 1.924 31.342 -74.959 0.105 -1.641 -1.574 6.687 -79.519 -0.315 -1.812 -0.672 8.889 -155.842 -1.217 -1.706 0.087 12.868 -175.014 -0.021 -2.275 0.494 15.637 -122.672 -0.620 -1.370 1.675 5.938 -123.764 0.592 -1.546 1.249 4.336 Pr(>|z|) < 2e-16 *** 0.157 0.001 *** 0.054 < 2e-16 *** < 2e-16 *** 0.917 0.101 0.116 0.000 *** < 2e-16 *** 0.753 0.070 . 0.501 < 2e-16 *** < 2e-16 *** 0.224 0.088 . 0.386 < 2e-16 *** < 2e-16 *** 0.983 0.023 * 0.622 < 2e-16 *** < 2e-16 *** 0.536 0.171 0.094 . 0.000 *** < 2e-16 *** 0.554 0.122 0.212 0.000 *** 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 10. FIGURE CAPTIONS Figure 1. Study area in the eastern Okavango Delta Panhandle, Botswana (a). White circles represent the thirteen villages (i.e. Mohembo-East, Kauxwi, Tobera, Xakao, Sekondomboro, Ngarange, Mogotho, Mokgacha, Seronga, Gunotsoga, Eretsha, Beetsha and Gudigwa) along the Okavango River (in light grey). White and dark-grey polygons represent the thirteen elephant corridors and agricultural land respectively. Light-grey circles show waterholes. The small southern Africa inset map (b) shows the location of the study area in northern Botswana in white. Figure 2. Utilisation distribution (UD) maps of the elephant population in the eastern Okavango Delta Panhandle. Maps represent population UDs during the (a) whole year, (b) the dry (May-October), (c) wet (November-April) and (d) crop-damage (January-April) seasons, at day (06:00-18:00; in orange) and night (18:00-06:00; in light blue) times, respectively. Legend shows symbols for landscape variables (i.e. villages, fields, elephant corridors, waterholes, and the Okavango River) and UD proportions in all maps, respectively. Figure 3. Crop-damage risk map for the eastern Okavango Delta Panhandle. (a) Circled areas show 50% (continuous black line), 20% (dashed black line) and 10% (dotted black line) likelihood of crop-damage events in the study area based on traditional fixed kernel analysis. Risk-map (in red) shows predicted distribution of damage events by elephant during the crop-damage season (January-April), and it spatial overlap with landscape variables (i.e. the Okavango River, waterholes, corridors, fields and villages). Small maps (b and c) represent the overlap between predicted crop-damage events and predicted elephant distribution during the (b) day (06:00-18:00; in orange) and (c) night (18:00-06:00; in blue). Figure 4. Polynomial regression of predicted crop-damage events as a function of predicted elephant distribution. Orange and blue lines represent the best-fitted models for the (a) day (06:00-18:00) and (b) night (18:00-06:00) elephant distributions, respectively. 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 11. APPENDIX Table A1. Summary of the 20 collared elephants in the Easter Okavango Delta Panhandle. Each column gives specific information about individuals ID, sex, number of relocations received between April 2014 and April 2016, and the mean number of relocations per elephant per hour, respectively. ID Sex No. relocations E01 E02 E03 E04 E05 E06 E07 E08 E09 E10 E11 E12 E13 E14 E15 E16 E17 E18 E19 E20 male male male female female female male male female female male female male male female female female male male female 16839 11814 16193 17133 9848 14239 16951 11783 16831 8668 5150 8484 11444 11135 16606 16210 5924 16724 11426 5878 Mean time between relocations (hours) 1.05 1.5 1.09 1.04 1.8 1.24 1.05 1.51 1.05 2.05 3.43 2.09 1.55 1.59 1.07 1.09 2.99 1.06 1.55 3.01 843 844 845 846 847 848 849 850 851 852 Figure A1. Individual elephant space use intercepts and standard deviations during the dry (a and b), wet (c and d) and crop-damage (e and f) seasons at day (06:00-18:00; in white) and night (18:00-06:00; in grey) times. Each graph represents one of the three landscape variables included in our analysis (i.e. elephant corridors [Corridors], agricultural fields [Fields], and waterholes [Waterholes]). Black circles and grey squares represent female and male intercepts, respectively. Both separate by a dashed blue line. 853 854