Illegal bushmeat hunting is a global threat to wildlife, but its secretive and unregulated nature undermines eﬀorts to mitigate its impacts on wildlife and wildlife-based industries. We investigated the scale of illegal bushmeat hunting in the Okavango Delta, Botswana (~20,000 km2) to assess its potential contribution to wildlife population declines. Approximately 1,800 illegal hunters each harvest an average of 320 kg of bushmeat annually, though some reported harvesting ≥ 1000 kg. While impala were the most commonly hunted species, buﬀalo and greater kudu accounted for most bushmeat. Hunters remove ~620,000 kg of medium-large herbivore biomass (equivalent to 15,500 impala) annually from the delta and humans are the fourth most prominent predator in the delta. Cumulative harvest by humans and other predators likely exceeds the intrinsic population growth rate of several species of ungulates in the delta, and helps explain purported declines in ungulate populations. Competition between humans and other apex predators for limited prey reduces the ecosystem’s carrying capacity for large carnivores. Illegal bushmeat hunting represents an economically ineﬃcient use of the delta’s wildlife and a threat to the region’s tourism industry. Strategies are required that provide clearer avenues for communities to beneﬁt legally from wildlife, while concurrently curbing illegal hunting through eﬀective law enforcement.
Hunting wild animals to consume or sell their meat (“bushmeat hunting”) is one of the most intensive and extensive threats to wildlife in African protected areas, with ecological and economic consequences (Lindsey et al., 2017). Bushmeat hunting drives declines in wildlife that locally extirpate populations and threaten many species with extinction (Brashares et al., 2004; Hilborn et al., 2006; Ripple et al., 2015, 2016; Harrison et al., 2016). Hunting pressure and corresponding wildlife declines are typically most severe close to human settlements or reserve boundaries (Henschel et al., 2011; Lindsey et al., 2011a; Mgawe et al., 2012; Koerner et al., 2016). Where hunting is most intensive, however, negative impacts permeate even large protected areas (Kiﬀner et al., 2013; Midlane, 2013). Other impacts from bushmeat hunting include skewed sex ratios and altered community composition (Holmern et al., 2006; Marealle et al., 2010; Koerner et al., 2016). These outcomes, coupled with non-lethal eﬀects such as increased prey vigilance and altered habitat use, induce secondary population impacts (Setsaasetal., 2007; Marealle et al., 2010). By disproportionately killing rare species (Martin etal., 2013), bushmeat hunting aﬀects vulnerable species most. The impacts of bushmeat hunting extend far beyond species targeted for bushmeat. Indiscriminate hunting methods such as snaring kill non-target species (Becker et al., 2013). By removing prey, bushmeat hunting limits the abundance and distribution of apex predators, and thus their ecological role. (Henschel et al., 2011; Ripple et al., 2014). Excessive hunting drives trophic cascades, which disrupt ecosystem processes and services (Abernethy et al., 2013; Galetti and Dirzo, 2013). Handling bushmeat increases risks of disease spillover to humans and livestock (Karesh and Noble, 2009; Kock et al., 2009; Alexander et al., 2012). While bushmeat contributes to food security (Rentsch and Damon, 2013; Schulte-Herbrügen et al., 2013), the beneﬁts are unlikely to persist in the face of increasing demand and declining wildlife resources (Bennett, 2002). Unfettered hunting can harm human communities by restricting ecosystem services, wildlifebased industries, and sustainable yields of wild meat (Lindsey et al., 2013).
While most research into the ecological impacts of bushmeat hunting has focused on tropical forest ecosystems, bushmeat hunting in savannah ecosystems poses a unique set of socioecological challenges. Savannahs host rich vertebrate communities that are ecologically and economically valuable. Photographic tourism, which depends on intact wildlife communities, can be a social and economic boon (Snyman and Spenceley, 2012). Where the potential for wildlife-based tourism exists, illegal bushmeat hunting is an ineﬃcient use of wildlife resources (Lindsey et al., 2015). Too frequently, however, the ﬁnancial beneﬁts of wildlife-based tourism do not reach impoverished communities near or within protected areas (PAs; Mbaiwa, 2005a; Wong, 2013). Consequently, local households may not receive direct beneﬁts from wildlife other than through hunting (Lindsey et al., 2013). Additionally, these same communities may endure most costs associated with wildlife, such as crop raiding, livestock depredation, and opportunity costs when land is gazetted (Adams et al., 2004; Snyman and Spenceley, 2012).
Although many African countries have laws regulating legal hunting of game species, illegal bushmeat hunting is pervasive and increasingly commercial (Brashares et al., 2004; Fa and Brown, 2009; Lindsey et al., 2013; Rentsch and Damon, 2013). Illegal hunting is a particular challenge to wildlife conservation because of its pervasive, unregulated, and secretive nature (Knapp et al., 2010). In some African countries, perceptions that hunting is a traditional subsistence activity with limited eﬀects on wildlife undermine enforcement of hunting laws (Lindsey et al., 2013). Improved understanding on the nature and impacts of illegal bushmeat hunting is essential to guide interventions addressing the issue.
We investigated illegal bushmeat hunting in the Okavango Delta in northern Botswana (hereafter “the delta”), in order to establish its scale and ecological impacts and to inform conservation interventions. Although Botswana is not traditionally associated with high levels of poaching, we found that illegal hunting occurs on a signiﬁcant scale in the delta, poses a risk to wildlife populations, results in direct competition with large predators, and may ultimately impair the local tourism industry.
2.1. Study area
The delta is a depression in the Kalahari basin that encompasses a mosaic ofpermanently drysavannahandMopane woodland, seasonally ﬂooded plains, and permanent swamp (McNutt, 1996). In 2014, UNESCO designated 20,236 km2 of the Okavango as a world heritage site with the 4610 km2 Moremi Game Reserve (MGR) at its core (Mbaiwa, 2005b). Surrounding the game reserve are 18 Wildlife Management Areas (WMAs) the government leases to private entities or to community-based trusts under the Community-Based Natural Resource Management (CBNRM) program. Most of these WMAs are exclusively for wildlife-based uses and all of them prohibit livestock. The 18 WMAs and MGR span 18,195 km2 (Fig. 1).
We conducted interviews in 12 villages located within mixed-use land concessions adjacent to WMAs (hereafter designated “residential concessions”). Residential concessions include residential communities, small-scale crop and livestock farming, and some ecotourism enterprises. The residential communities consist of formal villages with civic infrastructure and informal cattle posts – single households or small settlements – located outside the village proper where farmers tend remote crop ﬁelds and free-ranging herds of livestock. Veterinary control fences separate residential concessions from WMAs to prevent livestock-wildlife interactions. The few villages inside WMAs participating in Botswana’s CBNRM program were excluded from this study.
All study villages were located within 35 km of WMAs, a distance consistent with previous reports of how far bushmeat hunters travel (Loibooki etal., 2002). We included 39% of villages in the study area as focal villages(Statistics Botswana, 2014) to ensure that selected villages were representative of the study area as a whole. Study villages varied geographically, ethnically, and economically. Six villages were part of ongoing CBNRM projects and six were not. Study village economies varied from mainly pastoral to predominantly crop farming. Communities ranged in size from 35 residents to ~9000 (Statistics Botswana, 2014).
Due to the size of the delta and the challenges inherent in convincing people to acknowledge and discuss illegal activities, we concentrated eﬀorts to identify and interview hunters in six focal villages adjacent to four WMAs in the western delta. We chose this focal area because we consider those WMAs representative of the delta and because of documented illegal hunting there (911 Anti-poaching Operations Centre, unpublished data). The focal WMAs encompassed a variety of habitat types from permanent swamp to dry woodland. They exhibit management scenarios ranging from minimal management with minimal tourism (NG 24) to intensive tourism and private antipoaching operations (NG 26). One concession, NG 24, is leased to a community trust under the CBNRM program, while the other three are leased to private companies. We supplemented western-delta hunter interviews with opportunistic interviews with hunters in the northern and southern portions of the delta.
Botswana is widely credited for strong commitments to wildlife conservation (Lindsey, 2010) and anti-poaching (Henk, 2005; Hoon, 2013). It has sought to engage local communities through CBNRM initiatives (Pienaar et al., 2013). The safari industry predominantly follows ahigh-value low-impact modeloftourism designed to minimize its environmental footprint (Magole and Magole, 2011). With a sparse human population and large PAs, northern Botswana may appear relatively insulated against the threat of bushmeat hunting. However, in 2014, the government curbed hunting on public lands due to concerns that wildlife populations were in decline. Some experts cited illegal bushmeat hunting as a potential contributing factor (Rogan et al., 2015). Recent wildlife surveys found decreasing populations across northern Botswana in some herbivore species (Chase, 2011; DWNP, 2013). However, imprecision restricts these surveys’ power to identify population trends conclusively for most species. Ascertaining the extent of illegal hunting and its contribution to potential wildlife declines is a key pre-requisite to conserving the Okavango ecosystem and eﬀectively managing the wildlife therein.
With permission from the Ministry of Environment, Wildlife and Tourism [research permit EWT 8/36/4 xxvii (25)], we conducted questionnaire surveys and integrated available records on wildlife populations and hunting events to assess the scale and impacts of bushmeat hunting in the delta. Participation in interviews was voluntary and anonymous.
We hired local citizen interviewers based on their familiarity with study villages, language skills, knowledge of the bushmeat industry, and prior interviewing experience. They conducted interviews from August 2014–March 2015 in one of ﬁve local languages, depending on respondent preference. Interviews consisted of semi-structured questionnaires for hunters and heads of households. Questionnaires, adapted from Lindsey et al. (Lindsey et al., 2011b), included questions on households’ social and economic characteristics, consumption patterns, perceptions of wildlife, and, in the case of self-identiﬁed bushmeat hunters, on hunting practices and bushmeat harvests (see Supplementary Materials 8.1–8.2).
We sampled the heads of households at randomly distributed points within each village. This sampling approach excluded households situated outside the village core as deﬁned by manually drawn polygons. To compensate, we also conducted interviews with heads of households based at cattle posts. It was not possible to identify all cattle post locations and thus sample a truly random subset. We therefore used local knowledge to identify clusters of cattle posts in disparate geographic directions from the village centre and conducted 1–4 interviews at each cluster. The non-random cattle post component was critical to capturing the variation within delta communities.
Interviewers networked in study villages to build trust with community members and to identify hunters. We also identiﬁed hunters at households that acknowledged consuming bushmeat during household interviews, and through snowball sampling (Goodman, 1961). We attempted to interview every identiﬁed hunter in each village sampled. Hunters were informed that no information gathered would be used against them in any way and that the surveys were anonymous.
2.3. Supporting data
We examined records of all court cases in Ngamiland related to wildlife crimes between 1 January 2009 and 31 December 2014. We acquired human population data from the 2011 census (Statistics Botswana, 2014) and aerial wildlife census data from a 2014 dry season (August – November) survey (Elephants Without Borders, EWB, unpublished data). We generated population estimates for the large carnivore guild (lions Panthera leo, spotted hyaenas Crocuta crocuta, leopards Panthera pardus, cheetahs Acinonyx jubatus, and African wild dogs Lycaon pictus) from density estimates derived from studies conducted in similar habitats elsewhere in the Okavango (Cozzi et al., 2013; Golabek et al., unpublished data).
We conducted all statistical analyses using R version 3.2.2 (R Core Development Team, 2015) and spatial analyses in ArcGIS 10.2 (ESRI, 2013). For data that were non-normally distributed, we used nonparametric simulations to estimate 95% conﬁdence intervals by resampling with replacement 10,000 times.
2.5. Common hunting practices
Hunters reported up to three species they most target during hunts. We categorized species according to size, with species> 20 kg classiﬁed as large and< 20 kg classiﬁed as small (Coe et al., 1976). We then characterized hunters as targeting primarily small, large, or mixed-size species and compared the distribution of animal size classes relative to hunting methods to an expected even distribution of size classes using a Chi-Square goodness-of-ﬁt test.
2.6. Estimating the number of illegal hunters
Due to the secretive nature of illegal hunting, estimates based on interview surveys are prone to underreporting (Knapp et al., 2010). Combining interview data with arrest records can validate and improve estimates of illegal hunting prevalence (Knapp et al., 2010). We therefore used three complimentary methods to estimate the total number of bushmeat hunters in the delta:
2.6.1. Random sample Hunting intensity varies between small villages and larger population centres (Nielsen, 2006). We therefore classiﬁed villages as small (< 2000 residents) or large (> 2000), based on 2011 census data. We estimated the number of households in each village using the average number of people per household reported in the surveys. Asymptotic Gaussian approximation produced 95% conﬁdence intervals for the proportion of households involved in hunting in small and large villages. We obtained an estimate (and a minimum and maximum) by summing the results for small and large villages. We estimated on average 1.5 hunters per hunting household, based on evidence from surveys (including intra-family sharing and generational transfer of hunting skills and weapons within some households) (Brown and Marks, 2007; see Supplementary Materials 8.3–8.4).
2.6.2. Conviction rates We recorded the total number of Ngamiland residents convicted of wildlife crimes in or around the delta as described in magistrate court records of cases that opened and closed from 2009 to 2014. Using the sample of interviewed hunters, who reported whether they had convictions for wildlife-related crimes, we used non-parametric resampling to estimate with 95% conﬁdence the minimum number of hunters necessary to result in the observed number of convictions. Due to low conviction rates among interviewed hunters, a maximum estimate was unobtainable (i.e., the upper bound is inﬁnite).
2.6.3. Integrated method To address the propensity for random samples to underestimate illegal activity (Knapp et al., 2010), we generated a “revised” estimate byintegratingthreeadditionalsourcesofinformationintotheestimates ofhuntingrateswithinsmallandlargevillages. Weincorporatedresults from interviews conducted at cattle posts, the proportion of households reported by hunters in focus groups, and the number hunters interviewed in study villages. The resulting revised estimate was consistent with other estimates (see Supplementary Materials 8.3–8.4) and therefore was used as a ‘best estimate’ in subsequent analyses.
2.7. Bushmeat harvest
We deﬁned each individual hunter’s oﬀtake as the cumulative weight of all animals he reported killing in the previous twelve months divided by the number of hunters in his typical hunting party. We used non-parametric resampling to estimate with 95% conﬁdence the true mean oﬀtake per hunter. We also estimated rates of commercial hunting, which we deﬁned as the proportion of interviewed hunters who reported exchanging bushmeat for cash.
We modelled biomass consumption from bushmeat hunting and the large carnivore guild in relation to available biomass from medium to large herbivores (excluding elephants) and herbivore species’ intrinsic growth rates. A species’ intrinsic growth rate is a life-history trait that characterizes the maximum potential growth rate of a population given stable age demographics (Tanner, 1975). To capture spatial variation across the study area, we divided the area into four sections delineated by aerial transect routes (Fig. 2) conducted across 1–2 concessions. We estimated the intrinsic growth rate of biomass (φ) in the herbivore community (S) in a region (r) using Eq. 1 where Psr represents the estimated population of herbivore species s in region r from the 2014 EWB aerial survey, Ms. represents body mass of species s (from Coe et al., 1976). We assumed an intrinsic growth rate for herbivore species that breed annually and produce a single young per year of 1.2 per year (Owen-Smith, 2015). Warthog produce more than a single oﬀspring per year and therefore intrinsic growth rate was estimated using the expected relationship between ungulate bodymassandintrinsic growth rates, giving an intrinsic growth rate of 1.27 (Bothma, 2002).
Total estimated annual biomass oﬀtake from poaching (β) in a region (r) is given by Eq. 2 where Hr is the estimated number of hunters in region r, and− ks and− w s represent the average number of individuals from species s that hunters kill and wound respectively. We derived Hr from the revised estimate of hunters apportioned to each region based on the population of adjacent villages. Anti-poaching reports from one focal concession (NG 26) indicate that un-butchered carcasses with gunshot wounds are common (Bushcraft APU, unpublished data). For the purposes of this model, we assumed half of all animals hunters report wounding die. We estimated annual biomass consumption bythe large carnivore guild (βCr) using Eq. 3 where βdc is the average daily adult consumption for carnivore species c (Mills and Biggs, 1993), Prc is the estimated population of carnivore species c in region r, and 0.9 is the estimated adult proportion of carnivore populations. We reduced annual spotted hyena consumption by 50% to account for scavenging (including mortalities resulting from gunshot wounds).
We assumed that hunting levels in the delta “core” (see Fig. 2) are insigniﬁcant, based on testimony from illegal hunters and observations from land managers. We tested the eﬀects of hunting in 1) the western delta (where data are most comprehensive), 2) the three peripheral regions collectively, and 3) the entire delta.
Thus, we summarize the impact of bushmeat hunting is reﬂected in the diﬀerence, with and without hunting, between total biomass consumption and the intrinsic growth of herbivore biomass.
We applied a sensitivity analysis to the herbivore biomass model to determine the relative importance of variables and to test the robustness of underlying assumptions. We varied predator (individual predator population size), hunter (poacher numbers, average oﬀtake per hunter) and prey population (starting size and intrinsic growth rate) characteristics simultaneously and ran 1000 Monte Carlo simulations. For all count data we drew values from a random uniform distribution± 10% of observed values and for rate variables we drew random variables from a random uniform distribution± 0.1. To assess the relative importance of each variable we estimated the beta coeﬃcient from linear models between the resulting change in biomass consumption (growth − oﬀtake) and the spread of each variable’s input values as drawn from the uniform distribution.
2.8. Edge eﬀects from hunting
We designated areas< 15 km from a residential concession as edge habitat and areas> 15 km as interior habitat based on travel distances reported by illegal hunters and WMA managers. We tested the hypothesis that commonly-hunted species will be less abundant near the edge where they are susceptible to hunting (Lindsey et al., 2013) than in the interior (distinguished here from the “core” area referred to previously in Section 2.7) using transects of a 2014 dry-season aerial census (EWB 2014 aerial survey).
Using transect segments ≥ 10 km we calculated encounters/km for each species and used the two-sample Kolmogorov-Smirnov to test for diﬀerences. To control for diﬀerences in habitat between the edge and interior, we compared the habitat composition using a habitat classiﬁcation ofthree habitat classes (swamp,grassland, andwoodland) from 30 m Landsat 5 Thematic Mapper imagery (N.A.S.A. Landsat Program, 1994; K. Collins, unpublished data).
Ninety-one self-identiﬁed bushmeat hunters and 431 heads of households (367 randomly-sampled households and 64 selected households in cattle posts) participated in interviews. Non-response rates were below 3%, indicating a low risk of response bias (Groves, 2006).
3.1. Common hunting practices
Firearms were the most common hunting weapon, with 58% of interviewed hunters reporting their use (Fig. 3). Interviews also revealed that many hunters (24%) run animals down on horseback and kill them with axes or clubs (thus avoiding the sound of gunshots). Thirteen percent of interviewed hunters acknowledged using snares and none reported using gin traps. The choice of hunting weapon depended on the size of prey that hunters targeted (n =139 as some hunters used multiple methods, χ2 =76.7, df = 12, p < 0.001). Hunters using ﬁrearms or hunting from horseback primarily targeted large game, while those using dogs targeted small mammals. Hunters used airguns exclusively for hunting game birds. Interviews revealed that most hunting occurs outside of protected areas between the veterinary fences that encompass the delta WMAs and the roads that encircle the delta 10–30 km from the WMA boundaries. Twenty-one percent of hunters, however, acknowledged hunting inside protected areas. Delta hunters frequently kill, and occasionally target, small carnivores. Twenty-seven hunters reported killing small carnivores, including black-backed jackals Canis mesomelas, civets Civettictis civetta, and African wild cats Felis silvestris, but only two reported targeting them. Among medium and large prey species, hunters reported killing impala Aepyceros melampus most often, followed by kudu Tragelaphus strepsiceros and buﬀalo Syncerus caﬀer (Fig. 4). Our data contrast with oﬃcial reports of hunting incidents, which identiﬁed buﬀalo and giraﬀe Giraﬀa camelopardalis as the most commonly-hunted bushmeat species. One private anti-poaching unit accounted for most records of other bushmeatspecies.Many of the small, most commonly hunted species did not appear in any records of illegal hunting incidents.
3.2. Number of illegal bushmeat hunters
Among randomly-sampled heads-of-households, 13.7% of responents in small villages (n = 248; village range: 3–50%) and 7.6% in large villages (n =119; village range: 0–13%) reported that household members hunt for bushmeat (Table 1). Based on these proportions, the delta hosts approximately 1239 illegal hunters (95% CI: 576–1902; see Supplementary Materials 8.3–8.4). The 95th percentile bootstrapped conviction rate for illegal hunters suggested that at least 1523 hunters operate in the delta. The revised method data produced an estimate of 1787 hunters in the delta, 42% above the random-sample estimate but within the 95% conﬁdence interval.
3.3. Bushmeat harvest
Among hunters who reported their bushmeat harvest for the previous 12 months (n= 90), the average harvest was 320 kg, (95% CI: 230–422 kg, range: 0.5–2625 kg), which corresponds to 172 kg dressed weight (median 88 kg). On average, hunting parties consisted of 3.3 people. Three hunters reported harvesting at least 1800 kg per member of their respective hunting parties. These three individuals accounted for 22% of all meat that interviewed hunters reported harvesting. The top 20% of hunters (n = 18) harvested 68% of the bushmeat. Buﬀalo accounted for 30% of all bushmeat production, followed by kudu (17%) and giraﬀe (12%). Impala, the most commonly hunted ungulate, accounted for 9% of all bushmeat.
3.4. Impacts of bushmeat hunting on large herbivore populations
The estimated harvest of individual species by bushmeat hunters ranged from ~2–20% of populations (Table 2). Kudu suﬀered the highest proportional oﬀtake and was the only species harvested at rates approximately equivalent to its intrinsic growth rate (not considering other sources of mortality). Large carnivores consume an estimated 19.7% of herbivore biomass in the delta annually, just under the intrinsic biomass growth rate of 20% (Table 3). This proportion varied by region, however, ranging from 14.4% in the delta core to 23% on the periphery. Oﬀtake from bushmeat hunting (mean: ~348 kg/hunter/year) was 4.7% of herbivore biomass in the western delta, and 3.7% in the three peripheral regions. Consequently, total consumption exceeded the intrinsic growth rate in both the western delta and the periphery. Across the delta as a whole, humans and carnivores consumed 2.0% more biomass than herbivore populations can produce annually. Hunters removed more biomass from the western delta than all carnivores except for lions and spotted hyaenas,andmorethancheetahsandwilddogs inthedeltaasa whole. Over 1000 Monte Carlo simulations, 98.9% of runs resulted in a decline in herbivore biomass (i.e. total oﬀtake exceeded intrinsic growth) at the ecosystem scale, with 95% of runs producing a decline of 0.5% or greater. Herbivore populations and intrinsic growth rate exhibited the strongest eﬀects on the model outcomes, with beta coeﬃcients of 0.54 and 0.53, respectively (Fig. 5). Predators (β= −0.31) exhibited an impact nearly four times greater than bushmeat hunting (β= −0.08). Average oﬀtake per hunter (β= −0.077) was half again as important as the number of hunters (β= −0.051). When considering only the western delta, results were similar, but predation had an eﬀect only 60% greater than bushmeat hunting, and the number of hunters had an eﬀect broadly equivalent to average oﬀtake.
3.5. Edge eﬀects from hunting
Five species exhibited higher average relative abundance in the interior than in the edge, and ﬁve species had higher average relative abundance in the edge (see Supplementary Materials 8.5). Impala was the only species that was< 75% as abundant in the edge as in the interior, however the diﬀerence was not signiﬁcant (KolmogorovSmirnov Test, p =0.11).
4.1. Common hunting practices
Illegal bushmeat hunting is common and widespread in the delta. The prevalence of ﬁrearms is evidence that bushmeat hunting is lucrative (Damania et al., 2005). The practice of running animals down on horseback, which is a recent development (M.A.P. Ives, personal communication, July 2014), mirrors ﬁndings in Tanzania where poachers employ a similar tactic with motorbikes (Kiﬀner et al., 2014). Although ﬁrearm use and conviction rates suggest that the risk of detection is low, the rise of hunting on horseback without ﬁrearms suggests that detection is a concern for some hunters, particularly those hunting inside protected areas. We hypothesize that secretive methods will become more proliﬁc if law enforcement eﬀorts intensify. We would expect to see the hunters employ the most eﬀective methods most frequently. This was not always the case. Airguns and horses were more eﬀective than hunting with dogs but employed less often. The size of targeted species explains some of this diﬀerence, with airguns used almost exclusively to hunt game birds and horses used only to hunt large game. The skill necessary to hunt from horseback, and access to horses, may also limit its use. In contrast, dogs are eﬀective across a wide range of small and medium-sized species, and may be practical for the most opportunistic hunters. The frequent killing of small carnivores is an unusual phenomenon in bushmeat hunting. The resilience of jackals, and potentially other mesocarnivores, to persecution may result in sustained populations close to villages despite intense hunting pressure (Minnie et al., 2016). In other respects, species selection mirrored Martin et al.’s (2013) ﬁndings, with hunters primarily targeting large common species but actually killing higher proportions of rare species. Kudu, wildebeest, and zebra epitomize this trend, as they are common on the edges of the delta closest to human settlements, but less common at the ecosystem scale. Only a ﬁfth of hunters reported hunting inside protected areas. The abundance of wildlife outside protected areas appears to be the main reason, rather than risk concerns (Rogan et al., 2015). We expect that hunting inside protected areas will increase if wildlife declines outside PAs. The discrepancy between the species reported killed in oﬃcial records and the species that hunters reported killing most frequently suggests a systematic failure to detect illegal hunting of all but the largest species. Detection bias likely accounts for the higher prevalence of giraﬀe and buﬀalo in oﬃcial reports because hunters are generally notable to remove the carcasses of tho sespecies. Localized reports from Bushcraft APU (unpublished data), however, show that eﬀective enforcement allows for detection of the hunting of smaller animals even when hunters remove entire carcasses.
4.2. Numbers of illegal bushmeat hunters
The multi-method approach suggests that there are 1500–2000 operating inthe delta. Both the revised estimate and theconviction rate estimate suggest that the random sample was prone to undercounting as expected (Knapp et al., 2010). Therefore, we consider a reasonable possibility that thetruenumber of hunters exceeds therandom sample’s 95% conﬁdence limit of 1902 hunters. The best estimate of 1787 hunters represents 5% of men aged 15–64 in Ngamiland (Statistics Botswana, 2014). Observed hunting rates were consistent with those reported across Tanzania (Ceppi and Nielsen, 2014), and lower than reported in a localized study of Tanzanian communities close to a nationalpark (Kiﬀneretal.,2015). Inrural communitieswith abundant wildlife, this hunting rate is thus lower than we might expect, particularly in a society in which hunting was traditionally ubiquitous and where ﬁrearms are common. Bushmeat hunting is nevertheless suﬃciently pervasive and intensive to be a major issue for wildlife managementinNgamiland.Ourestimateprovidesasnapshotofcurrent hunting intensity and ongoing monitoring is critical to track the extent of illegal hunting and to assess the eﬀectiveness of interventions.
4.3. Bushmeat harvest
While most hunters reported hunting exclusively for home consumption, several hundred also sell meat. The large quantities of bushmeat (~2000 kg) reported by some hunters suggests the existence of an organised commercial element to the industry, with capacity to harvest, transport and dispose of signiﬁcant volumes. Even less intensive hunters are harvesting suﬃcient bushmeat to meet a considerable portion of household demand. The median dressed weight bushmeat harvest (88 kg) is equivalent to 40% of an average 6-person Botswanan household’s annual meat consumption (FAO, 2011). For non-commercial hunters (~70% of hunters), bushmeat represents a nearly costless substitute for livestock meat, and thus allows for signiﬁcant savings. Demand for bushmeat among consumers who purchase the product is driven by various factors, including price, availability and taste preferences (Rogan et al., 2015).
4.4. Ecological impacts
With the exception of kudu, harvests from bushmeat hunting were well below estimated intrinsic growth rates. However, the delta supports important populations of apex predators that depend on many of these same species for prey. Bushmeat hunters compete directly with carnivores for prey and as a result reduce the carrying capacity of the area for those species (Henschel et al., 2011). We acknowledge that our biomass oﬀtake model relies on coarsely estimated inputs, yielding imprecise estimates of the relationship between biomass oﬀtake and growth. The sensitivity analysis, however, illustrates that even accounting for potential error in inputs, consumption of predators and bushmeat hunters almost certainly exceeds sustainable levels. Illegal bushmeat hunting thus poses a real threat to the delta’s wildlife. These eﬀects are consistent with outcomes observed in a wetlandwoodland ecosystem of Rukwa-Katavi, Tanzania, where comparable levels of oﬀtake, including disproportionate hunting of uncommon species, coincided with steep declines in wildlife populations (Martin and Caro, 2013; Martin et al., 2013). Our ﬁndings help explain observed declines in some species of herbivores, particularly intensively hunted species such as kudu, giraﬀe, and wildebeest (Chase, 2011; DWNP, 2013). The sensitivity analysis also revealed the importance of hunting intensity (i.e., oﬀtake per hunter) relative to the number of hunters. This relationship suggests that delta wildlife populations are particularly vulnerable to the commercialization of bushmeat hunting. Demand for bushmeat in Ngamiland is likely to increase with human population growth, which, in the absence of increased law enforcement, could lead to further commercialization of the bushmeat trade. Numerous additional factors, including climate change and reduction of the delta ﬂoodplains through decreased ﬂood volumes, will further compromise the present-day productivity and herbivore carrying capacity of the Okavango ecosystem (Murray-Hudson et al., 2006). These trends suggest that the relative impacts of bushmeat hunting quickly will become more severe. In addition, prey depletion caused by bushmeat hunting may increase human-carnivore conﬂict (Valeix et al., 2012; Khorozyan et al., 2015), and retaliatory killings of wild carnivores. These factors may explain the delta’s declining lion population over the previous decade (Fuller and Sievert, 2001; Winterbach et al., 2012; Bauer et al., 2015).
4.5. Edge eﬀects from hunting
We did not ﬁnd evidence supporting our hypothesis that bushmeat hunting depresses wildlife populations on the delta’s periphery compared to its un-hunted interior. Our ﬁndings may reﬂect variation in habitat suitability or seasonal wildlife movements (Bartlam-Brooks et al., 2013; Bennitt et al., 2014), but they may also reﬂect consistent
movement of herbivores from a population source in the core to a population sink in the edge. Because the delta core is highly productive and experiences minimal bushmeat hunting, local movement could theoretically compensate for the oﬀtake of species in peripheral WMAs. Trends from other savannah ecosystems suggest that in actuality locally concentrated bushmeat hunting can negatively aﬀect wildlife populations across an entire ecosystem rather than conferring localized impacts (Kiﬀner et al., 2013; Midlane, 2013).
4.6. Economic and social impacts
Excessive bushmeat harvests have potential to confer negative economic impacts on both a local and national level. The Okavango’s globally renowned wildlife is critical to Botswana’s high-value tourism and is the key attraction for foreign tourists (Mladenov et al., 2007). Wildlife based tourism is a driver of economic growth and job creation in Botswana, contributing 70,000 jobs and nearly 10% of GDP (World Travel and Tourism Council, 2015). Illegal hunting for bushmeat threatens the growth of this important industry through its potential to impart population declines of herbivores and reduce the carrying capacity for large predators, on which the tourism industry depends (Di Minin et al., 2013).
4.7. Implications for conservation
We recommend a holistic approach to addressing illegal hunting in the Okavango, including eﬀective law enforcement on the delta’s periphery (Hilborn et al., 2006) and the development of wildlife tourism ventures in the Okavango’s peripheral areas that generate employment for members of local communities. Interventions should focus on increasing incentives to conserve wildlife while simultaneously increasing the risks of hunting illegally(Biggsetal.,2017;Knapp,2012; Rogan et al., 2015). Success may depend on changes to legal frameworks that empower communities to beneﬁt from wildlife signiﬁcantly, in ways other than through illegal hunting (Lindsey et al., 2015). Wildlife managers must seek ways to engage local communities in wildlife tourism industries more eﬀectively through partnerships and employment. Finally, policies are required that promote legal wildlife ranching and the distribution of game meat so that legal harvests can meet demand for the product. Communities in the lands adjacent to the delta need opportunities to participate in such land uses.
Our study underscores the threat illegal bushmeat hunting poses to wildlife in African savannahs and to the prospects of wildlife-based tourism industries. Even in Botswana, a middle-income country with a relatively well-resourced wildlife authority and low human population density, illegal hunting constitutes a serious threat to wildlife populations.The bushmeat trade is a complex problem that links food security, development, culture, and conservation, and which demands multifaceted interventions.