Introduction
Conflict between people and wildlife is a global conservation issue (Woodroffe et al., Reference Woodroffe, Thirgood and Rabinowitz2005). Typically, conflict occurs outside protected areas, where the ranges of people and wildlife overlap. For example, 70% of the range of African elephant Loxodonta africana occurs outside protected areas (Blanc et al., Reference Blanc, Barnes, Craig, Dublin, Thouless, Douglas-Hamilton and Hart2007). Conflict with people therefore seems inevitable. Presently, elephant numbers are increasing in some areas, and their ranges may also be expanding (Blanc et al., Reference Blanc, Barnes, Craig, Douglas-Hamilton, Dublin and Hart2005). Consequently, human-elephant conflict is increasingly reported, particularly where people's livelihoods are affected (Hoare & du Toit, Reference Hoare and du Toit1999; Sitati et al., Reference Sitati, Walpole, Smith and Leader-Williams2003).
Numerous studies promote symptomatic treatments to mitigate human-elephant conflict, although many of these efforts meet with limited success (Hoare, Reference Hoare1999; Osborn & Parker, Reference Osborn and Parker2003; Graham & Ochieng, Reference Graham and Ochieng2008), possibly because they deal with the consequences of people and elephants living together. To manage this conflict better an approach is needed that deals with the underlying causes, and not the consequences, of the problem (Hoare, Reference Hoare1999; Barnes, Reference Barnes2002). The present understanding of human-elephant conflict is fragmented, as factors underlying human-elephant conflict appear to be site-specific and unpredictable. In part this may be due to the spatial scale at which analyses are conducted and the unpredictable behaviour of individual elephants (Hoare, Reference Hoare1999; Sitati et al., Reference Sitati, Walpole, Smith and Leader-Williams2003). Only when the underlying causes of human-elephant conflict have been identified can a non-symptomatic management approach be applied.
Here we examine a key factor underlying human-elephant conflict (the overlap in spatial use between elephants and people), while conceding that other factors such as the nutritional quality of crops for elephants also need to be examined (Osborn, Reference Osborn2004). For elephants, the availability and distribution of water and food underlie patterns of spatial use (Grainger et al., Reference Grainger, van Aarde and Whyte2005; de Beer et al., Reference de Beer, Kilian, Versfeld and van Aarde2006; Harris et al., Reference Harris, Russell, van Aarde and Pimm2008). Elephants remain close to permanent water during the dry season but with the onset of the rainy season may move away and rely instead on rain water that collects ephemerally in natural depressions scattered across the landscape (Owen-Smith, Reference Owen-Smith1996; Verlinden & Gavor, Reference Verlinden and Gavor1998).
When the distribution of people and elephants overlap conflict appears to be correlated to spatial factors such as human population density, the transformation of land through agriculture, distance from roads, and proximity to daytime elephant refuges (Hoare & du Toit, Reference Hoare and du Toit1999; Parker & Osborn, Reference Parker and Osborn2001; Sitati et al., Reference Sitati, Walpole, Smith and Leader-Williams2003). Changes in the seasonal distribution of elephants and people could explain patterns of human-elephant conflict. In the dry season, when water is limiting, elephants and people may compete for this resource. In other instances, however, human-elephant conflict is reported as elephants shift their ranges between seasons (Tchamba, Reference Tchamba1996) or occupy wet season ranges (Thouless, Reference Thouless1994) where they encounter people and agriculture.
In savannah ecosystems, problem elephant activity shows a seasonal peak, usually at the end of the wet season (Hoare, Reference Hoare1999). This is a period when water becomes limiting as seasonal supplies dry up, presumably forcing elephants to return to more permanent water bodies. The period also corresponds to the time when savannah elephants destroy ripening crops (Hoare, Reference Hoare1999).
Here we examine the spatial-temporal distribution of elephants and people and the manner in which this may underlie patterns of crop raiding along the Okavango Panhandle in northern Botswana. The regional elephant population across northern Botswana has apparently stabilized (Junker et al., Reference Junker, van Aarde and Ferreira2008), while the human population continues to grow (Central Statistics Office Botswana, 2002). Given our present understanding of elephant spatial use and human-elephant conflict, we will show how season determines elephant spatial use, that spatial use does not differ between bulls and breeding herds, and that crop damage is a consequence of elephant spatial use and its overlap with people and their livelihoods. Finally, based on our findings, we examine the present scheme used by the Botswanan Government to compensate farmers for crop damage, and suggest a more direct performance payment method to reward farmers effectively for the ecological services they provide.
Study area
Our study area forms part of the Ngamiland District of north-west Botswana. For administrative purposes the area is commonly referred to as NG11. It extends over a controlled hunting area of 5,952 km2. The Okavango River, which forms a Panhandle to the Okavango Delta, marks the southern and western boundaries of NG11 (Fig. 1). The international boundary with Namibia, which is fenced, delineates the northern boundary, and NG11 adjoins NG13 to the east. Rain falls predominantly during the summer (November-April) and the area receives a mean annual rainfall of 500 mm (Department of Meteorological Services Botswana, 2004). We defined the period from November to April as the wet season and May to October as the dry season.
Fig. 1 (a) The location of Ngamiland District 11 in north-west Botswana. (b) NG11 borders the eastern side of the Okavango Panhandle. The road that runs parallel to the edge of the Panhandle served as a survey transect (see text for details).
NG11 is inhabited by c. 13,000 people (Central Statistics Office Botswana, 2002). Seronga is the largest village, with a population of 3,000, and small settlements and cattle posts occur between villages. Soil is largely deep Kalahari sands, and mopane Colophospermum mopane woodlands occur throughout. Fertile soils support subsistence agriculture, mostly in areas adjacent to the Okavango Panhandle. Both people and elephants rely on the Okavango River for water during the dry season when there is no other perennial water (McCarthy, Reference McCarthy2006). Farming is small-scale, and fields are 0.2–6.0 ha in size. Cultivation occurs in the rainy season when farmers plant maize, groundnut, millet and watermelon, which are harvested in April-June.
Methods
Elephant spatial use: aerial surveys
We used a Cessna 206 aircraft to conduct two fixed-wing aerial surveys of the entire area of NG11.These surveys took place over a period of 4 days at the end of the dry (September 2003) and wet (April 2004) seasons. We used strip transects (Buckland et al., Reference Buckland, Anderson, Burnham, Laake, Borchers and Thomas2001) and defined 67 north-south transects, at one nautical mile (c. 2 km) intervals. We flew these transects at an altitude of 300 ft (c. 90 m) (measured with a conventional altimeter) and a speed of 100 kt (c. 185 km h−1). We sampled using calibrated strip widths fixed to 400 m on each side of the aircraft (following Norton-Griffiths, Reference Norton-Griffiths1978). These strips covered 40% of the study area. Two observers conducted each transect, and a survey coordinator sat next to the pilot and used a hand-held global positioning system (GPS) to record the positions of elephants. Each observer used a digital camera to record 2-5 images at each elephant sighting from which we later counted the number of elephants in each herd. We examined numbers at a local scale by dividing our study area into four strata defined by distances to the Panhandle (0-10, 10-20, 20-30 and >30 km). These strata covered areas of 1,716, 1,377, 985 and 626 km2, respectively. Using fixed-width transects of variable length facilitated the application of Jolly's Method II (Jolly, Reference Jolly1969) to estimate population sizes and their variances for each of the strata. We estimated elephant density for a stratum as the mean value for the transects within that stratum.
Elephant spatial use: satellite tagging
We placed GPS satellite collars (Africa Wildlife Tracking, Pretoria, South Africa) on 10 elephants. We immobilized individuals by darting from a helicopter; a veterinarian was present to supervise all collaring and sedation procedures, which adhered to the standards and conditions set by the Animal Ethics Committee of the University of Pretoria (permit number AUCC-040611-013). During October 2003 we fitted collars on four adult females in separate breeding herds and two adult bulls. We present hourly data from these individuals, collected over a 3-day period during October 2003. However, as four of these collars failed within a month we attached a further four collars, on three adult females and an adult male, during June 2004. The collars collected positional data at 24-hour intervals. To reduce bias towards any specific time of day, we advanced the data collection time by two hours each week. We used data collected from September 2004 to August 2005 and global information system software to calculate the distances of individuals‘ locations to the Okavango Panhandle.
Elephant spatial use: spoor counts
We used a public road that runs parallel to the Panhandle as a survey transect. This 50 km transect stretched from Nxiniga, via Seronga, eastwards to Eretsha (Fig. 1). We counted fresh spoor from elephants that crossed from the Panhandle to the hinterland, daily over 06.00–10.00 for 7-13 days each month from October 2003 to May 2004. To prevent recounting old spoor, we cleared this spoor after each survey by dragging a tyre tied to our survey vehicle. We distinguished the spoor of bulls from that of breeding herds, which comprise adult females, their offspring and some males. We divided the survey transect into 100 segments, each 500 m long. We used a χ2 test on 2 § 2 contingency tables to test whether the segments where elephants crossed were associated with the presence or absence of human settlements.
Human spatial use
We used a hand-held GPS unit to determine the position of human settlements along the road transect. The area away from the road is mostly uninhabited and our survey along the road therefore included a relatively complete record of spatial settlement. For each segment of our line transect 50 m each side of the road was sampled as 500 § 100 m (5 ha) quadrats, in each of which we recorded the number of settlements, and questioned inhabitants to estimate the number of people living in these settlements.
Crop raiding by elephants
Each month 40 agricultural fields were randomly selected along the road transect to record crop age and type, area damaged by elephants and the status of these elephants (bulls, breeding herds). To quantify the financial implications of elephants damaging crops we related the area damaged to expected crop production and income using market-related values. Using these data we predicted the income after raids and the actual income after compensation from the Botswana Government, using the compensation standard of BWP 250 per ha damaged. Separate to this, we accompanied personnel from the Botswana Department of Wildlife and National Parks to verify reported elephant damage to crops.
Results
Season and elephant spatial use
The dry season elephant population estimate was 3,579 (95% confidence limit, CL, 2,975–4,183), which is a density of 0.71 elephants km−2 (95% CL=5.1–9.2; Fig. 2, Table 1). The wet season population estimate was 1,060 (95% CL 810–1,310), i.e. a density of 0.21 elephants km−2 (95% CL 1.0–3.2). This suggests a 70.4% reduction in elephant number from the dry to the wet season. The distribution of elephants also changed with season (Fig. 2), with greater proportions of elephants occurring closer to water in the dry season (Table 1). Individuals were only recorded >30 km from permanent water during the wet season (Table 1).
Fig. 2 Distribution of elephant bulls and breeding herds across the NG11 section of Ngamiland (Fig. 1), during aerial censuses conducted in (a) October 2003 (dry season) and (b) April 2004 (wet season).
Table 1 Mean estimates ± SE of elephant numbers and densities from fixed-wing aerial surveys conducted in October 2003 (dry season) and April 2004 (wet season), with 95% confidence interval in parentheses, in the NG11 section of Ngamiland, Botswana (Fig. 1).
The distance that satellite-tracked individuals ranged from the Okavango Panhandle varied during the year (Fig. 3). Elephants were closest to the Panhandle in September, but most distant in April. During the dry season, from June to October, most individuals remained within 10 km of the Panhandle. At this time elephants remained closer to the Panhandle than from November to May, during the wet season, when most elephants moved further away (Fig. 3). In April the furthest elephant from the Panhandle was an average of 60.2 ± SE 0.7 km away (n = 26 GPS fixes). Human settlements were situated 197 ± SE 11 m (n = 305) from the Panhandle. Overall, there appeared to be little overlap in spatial use between elephants and people, as elephants tended to use areas further from the Panhandle than people.
Fig. 3 Mean ± SE of monthly distances at which six GPS-satellite collared elephants were recorded from the Okavango Panhandle. The shaded area indicates the wet season. During the dry season, from June to October, most individuals remained within 10 km of the Panhandle. These distances were closer to the Panhandle than from November to May, mostly during the wet season, when most elephants moved away.
Herd type and elephant spatial use
Both bull and breeding herd numbers close to the Panhandle declined during the wet season when more elephants were recorded in more distant strata (Table 2). From October 2003 to May 2004 we recorded 881 incidences of breeding herds and 861 of bull groups crossing the 50 km road transect. Breeding herds and bulls crossed the transect at as many segments with human settlements as expected during both the dry and wet season (breeding herds, wet season: χ2=2.3, df=1, P=0.13; dry season: χ2=0.0, df=1, P=0.99; bulls, wet season: χ2=0.1, df=1, P=0.72; dry season: χ2=0.0, df=1, P=0.99; Table 3). However, both breeding herds and bulls apparently avoided areas with human settlements, assuming that crossings are independent (Table 4). They did so by crossing the transect at a higher frequency than expected in segments with no human settlements (breeding herds, wet season: χ2 = 40.7, df = 1, P <0.001; dry season: χ2 = 82.9, df = 1, P <0.001; bulls, wet season: χ2 = 57.5, df = 1, P <0.001; dry season: χ2 = 26.8, df = 1, P <0.001
Table 2 Number (with percentage in parentheses) of elephant bull groups and breeding herds observed within distance strata during aerial surveys in October 2003 (dry season) and April 2004 (wet season) in the NG11 section of Ngamiland (Fig. 1).
Table 3 Number of segments along a line transect in which we recorded crossings (with expected number of crossings in parentheses) by elephant breeding herds and bulls in October 2003 (dry season) and April 2004 (wet season), in relation to the presence or absence of human settlements.
Table 4 The number of times that elephant breeding herds and bulls crossed 500 m segments of a line transect (with expected number of crossings in parentheses) in October 2003 (dry season) and April 2004 (wet season) as a function of the presence or absence of human settlements.
Human spatial use
We recorded 5,544 people, 2,686 homes and 111 agricultural fields along the 50 km transect (Fig. 1). Of the 100 survey quadrats, 31 were not inhabited or used for agriculture, and consisted primarily of mopane woodland.
Crop raiding in relation to season and herd type
Crop raiding occurred as crops ripened towards the end of the rainy season, from March until May. This was reflected in both our randomly sampled fields and in fields where farmers reported crop damage (Table 5). The increase in crop raiding during April coincided with the time of our aerial census, when we recorded few elephants within 10 km of the Panhandle. This suggests the increase in crop raiding may be independent of elephant numbers. Raids took place at night, and more bull groups than breeding herds were involved (Table 5).
Table 5 The incidence and type of elephant groups involved in crop raiding along the Okavango Panhandle from October 2003 to May 2004. Reported refers to incidents reported by farmers to the Botswana Department of Wildlife and National Parks, and Random refers to incidences assessed while inspecting 40 randomly selected fields each month.
Financial implications of elephant activity
Farmers applied to the Botswana Department of Wildlife and National Parks for financial compensation for crop damage by elephants (Table 6). Due to the small size of their fields compensation was low, up to a maximum of BWP 250 (c. USD 38) per month. The actual income farmers made following compensation was on average 11% lower than their projected income if elephant raids had not taken place. Considering we recorded 111 fields along our 50 km transect, and that the entire length of the road along the eastern border of the Panhandle through NG11 is 120 km long, we expect 266 fields to be at risk from elephants in NG11. Given an average field size of 1.5 ha (Mosojane, Reference Mosojane2004), and a compensation rate of BWP 250 per ha, this suggests that the Botswana Government would be obliged to compensate the farming community a total of c. USD 10,000 per year if crops were totally destroyed by elephants in NG11.
Table 6 Financial implications of crop damage by elephants raiding fields along the Okavango Panhandle from January to May 2004.
Social implication of elephant activity
We examined hourly variation in spatial use by elephants during the dry season, when elephants were present at their highest densities along the Panhandle. Elephant movements over a 3-day period during this season confirmed they were only present on the Panhandle after dark, moving to areas further away during daytime (Fig. 4). This pattern suggests that these elephants drank at 1–3 day intervals. Thus, while elephants in general were situated further from the Panhandle than people, the highest probability of encountering elephants was at night as they moved onto the floodplain.
Fig. 4 The distances six GPS-satellite collared elephants were recorded from the Okavango Panhandle over a 3-day period in October 2003, during the dry season. Shaded areas indicate hours of darkness. Elephants visited the Panhandle during hours of darkness only and at intervals of 1–3 days, remaining further away from the Panhandle during the day.
Discussion
This study shows that one aspect of human-elephant conflict, crop raiding, was greatest at the transition from the wet to the dry season, during periods of darkness, and dominated by bulls. Elephant spatial use clearly varied with season: they used areas close to the Panhandle during the dry season, moving away at the onset of the wet season. Thus when the distribution of water is limited, dry season refugia played an important role in dictating elephant spatial use. Conversely, during the wet season when the distribution of water is less limiting, elephants tended to range away from perennial water.
Unlike elephants, the spatial distribution of people in NG11 does not vary seasonally, as most people live along the edge of the Panhandle and close to permanent water. Thus, while people and elephants are separated spatially during the wet season, they occur more closely together through the dry season. Even so, elephants tended to remain further away from the Panhandle than people, at least during the day. Critically, elephants moved onto the Panhandle at night, when they also raided fields. Our results thus show the importance of both time and space in determining the likelihood of conflict (Hoare & du Toit, Reference Hoare and du Toit1999).
Unexpectedly, our study did not show a distinct separation in the distances of bulls and breeding herds from water (Stokke & du Toit, Reference Stokke and du Toit2002). Nor did our results suggest breeding herds or bull groups react differently to human settlement patterns, in comparison to Hoare (Reference Hoare1999) who recorded bull elephants significantly closer to human settlements than cows. Despite these differences with other studies, the patterns of elephant spatial use that we recorded are, however, consistent with our prediction that crop raiding would be seasonal.
Given our findings, we suggest several management options to reduce crop raiding by elephants in this area. The most obvious approach would be to modify elephant spatial use. This may be achieved using physical barriers, including various fencing schemes, which are sometimes used to restrain elephant movement (Osborn & Parker, Reference Osborn and Parker2003). Rudimentary fencing using cut branches around fields in our study area were ineffective in excluding elephants. In other areas, chemical deterrents such as chilli peppers may discourage elephants (Sitati & Walpole, Reference Sitati and Walpole2006).
Alternatively, the spatial distribution of elephants may be manipulated by providing mineral licks as forage, as soils and water high in sodium are known to attract elephants (Holdo et al., Reference Holdo, Dudley and McDowell2002). Manipulating the spatial distribution of water during the dry season by creating waterholes away from areas where people live is another way to reduce the spatial overlap of elephants and people. We do not, however, advocate the establishment of permanent water supplies throughout NG11, as this could have deleterious consequences for vegetation (van Aarde et al., Reference van Aarde, Jackson and Ferreira2006). We also do not support the culling of elephants to reduce numbers, as this does not address the underlying causes of human-elephant conflict (van Aarde & Jackson, Reference van Aarde and Jackson2007).
From a human perspective, settlement patterns need to be considered relative to elephant spatial use. Land-use planning is known to influence human-elephant conflicts (Fernando et al., Reference Fernando, Wikramanayake, Weerakoon, Jayasinghe, Gunawardene and Janaka2005). Given the recent increase in human numbers in NG11 (Central Statistics Office Botswana, 2002) land-use and zonation must be carefully planned to ensure that future patterns of human settlement avoid areas that are well used by elephants.
While these ideas require substantial development, farmers in Botswana currently receive government compensation when elephants damage their crops. Despite the goals of relieving social and economic hardships, this type of compensation may be detrimental to conservation efforts (Bulte & Rondeau, Reference Bulte and Rondeau2005). Such schemes are often cumbersome and expensive to administer (Hoare, Reference Hoare1995), and the cost of government bureaucracy to process claims almost certainly exceeds the total value of claims by farmers in NG11.
Critics argue that as farmers know they will be compensated for damage, there is no incentive for them to adopt new or revise current practices to reduce crop raiding: the so-called moral hazard of compensation schemes (Nyhus et al., Reference Nyhus, Osofsky, Madden, Fischer, Woodroffe, Thirgood and Rabinowitz2005). In addition, the costs of compensation schemes run well beyond damage payments and include search and information costs (needed to verify the damage costs) as well as decision making costs (associated with disputes that arise from claims; Schwerdtner & Gruber, Reference Schwerdtner and Gruber2007).
An alternative approach is a payment in advance, or performance payment, to farmers (Nyhus et al., Reference Nyhus, Osofsky, Madden, Fischer, Woodroffe, Thirgood and Rabinowitz2005; Schwerdtner & Gruber, Reference Schwerdtner and Gruber2007). Effectively, this approach rewards farmers for living with elephants rather than compensating individuals for losses incurred through elephant activity. This direct payment approach is more cost-efficient than indirect incentives such as compensatory payments for crop damage (Ferraro & Kiss, Reference Ferraro and Kiss2002). Performance payments are beneficial to farmers, who are effectively rewarded for the ecological services they provide (Ferraro, Reference Ferraro2001; Bulte & Rondeau, Reference Bulte and Rondeau2005), such as sharing space with elephants. From a financial perspective, performance payments do not detract from farmers attempting to maximise crop yield, as any harvested crops will count over and above the performance payment they will receive. The approach relieves farmers from some of the social costs they endure such as sleepless nights or the dangers associated with trying to ward off elephants, knowing they will receive some reward for living with elephants. This may therefore culture a more positive attitude towards elephants, as is presently the case in Namibia's Caprivi Region (G. Owen-Smith, pers. comm.).
Nationally, the fiscal impact of human-elephant conflict to farmers in NG11 is small. At USD 10,000 annually this is less than the license fee to hunt an elephant in Botswana (USD 15,000; Sharp, Reference Sharp2007). The income generated from hunting one elephant would more than cover payments to every farmer should a performance payment scheme be introduced. Similarly, revenue generated from problem elephant control in Zimbabwe went some way to replacing dysfunctional crop damage compensation schemes (Hoare, Reference Hoare1995).
In conclusion, we have outlined, based on our examination of the spatial-temporal distribution of elephants and people, a new approach that could be developed to help alleviate the causes of elephant crop raiding in north-west Botswana, and recommended that a more direct performance payment approach may jointly benefit government, local farmers and elephant conservation. Based on our recommendations the Botswana Department of Wildlife and National Parks (District Office, Maun, Botswana, pers. comm.) has agreed to present the performance payment approach for consideration to communities affected by human-elephant conflict in northern Botswana. This follows the integration of areas of northern Botswana into the Kavango-Zambezi Transfrontier Conservation Area, for which a Botswana government unit was recently established to coordinate activities, and that will include our call for a performance payment approach.
Acknowledgements
The Botswanan Department of Wildlife and National Parks provided logistical support and assistance during the aerial surveys. The project was funded by the Peace Parks Foundation and the US Fish & Wildlife Service, the latter through a grant to Conservation International's Southern African Wilderness Programme.
Biographical sketches
Tim P. Jackson has 17 years of research experience in mammalogy. His present research with the Conservation Ecology Research Unit (CERU) concerns the conservation and management of elephants throughout southern Africa. Sibangani Mosojane was the District Wildlife Coordinator with the Department of Wildlife and National Parks. Until 2007 he worked on mitigation measures for human-elephant conflict across Ngamiland District, where he also administered the management of wildlife resources. Sam M. Ferreira's research centres on mammal and bird conservation biology with emphasis on temporal dynamics and the factors influencing these. He coordinates aspects of the Elephant Programme at CERU. Rudi J. van Aarde is Director of CERU. His research examines the restoration of populations and communities as a contribution to conservation, with a focus on elephant conservation throughout southern Africa.
