Management of human–carnivore conflict using expert-based science (EBS) has been proven to be ineffective, and the conflicts have continued to have socio-economic implications on the livelihoods of pastoralists residing near protected areas globally (Goodrich et al. 2011; Inskip & Zimmerman 2009). In response, pastoralists developed retaliatory behaviour by killing (both legal and illegal) or poisoning carnivores (Ogada et al. 2003), leading to the observed decline of carnivore populations. Information on distribution and intensity of human–carnivore conflict is essential in managing such conflicts around the world, and obtaining adequate information to come up with effective and efficient strategies requires extensive data on wildlife populations which is often insufficient in technical EBS (Gilchrist, Mallory & Merkel 2005).Techniques involved in EBS are usually expensive and demanding in terms of inputs, thus complicating research (Anadón et al. 2009). However, it has been indicated that information from local knowledge and skills of communities could be of high quality at a relatively low cost (Hartwig 2009a). Despite appeals to integrate this knowledge in scientific research to enhance natural resource management, such efforts are seldom put into practice (Sutherland et al. 2013).

There is a widespread recognition that people residing in an area for long periods develop important and established relationships with and knowledge about those areas (Prado & Murrieta 2015). Various terminologies have been proposed to describe this knowledge, such as local ecological knowledge (LEK) (Anadón et al. 2009), traditional ecological knowledge (TEK) (Phuthego & Chanda 2004) and indigenous knowledge (IK) (Roba & Oba 2009). Although many authors use this terminology interchangeably, Gilchrist et al. (2005) and Neagh et al. (2008) suggested that a distinction that comprehends the time scale from which this knowledge and skills has been acquired is required. The distinctions being that TEK and IK are more of knowledge and practices acquired through adaptive management, by an ethnic group through generations interacting with nature and is believed to be transferrable from one generation to another (Neagh et al. 2008), while LEK, on the contrary, describes the skills and practices gained and developed by an individual’s annotations over his or her periods of interaction with nature, and this knowledge is not transferred from one generation to another (Gilchrist et al. 2005). In many studies, LEK is associated with the perception or skills possessed by individuals given their periods of interaction with nature. In this study, we assessed the competences of LEK possessed by livestock herders to investigate carnivore species that predate on livestock around Makgadikgadi and Nxai national parks. We define LEK in this study as the skills possessed by herders by virtue of experience from duties linking them with livestock depredation issues on a daily basis, and herders as individuals that are employed to look after livestock throughout the year, and this includes self-employed farmers. Herding activities included general husbandry of keeping livestock (taking livestock to pasture and water in the morning and bringing them back to the kraal [boma – a place where they are kept safe during the night]), looking for lost and unhealthy animals and reporting any problem concerning the keeping of the livestock to the owner (in case of employed herders).

Livestock farming in Botswana is very important because majority of the rural communities depend on it as a source of income and protein (Statistics Botswana 2009; Statistics Botswana 2013). As such, livestock production, especially cattle, plays a critical socio-economic role for farmers in Botswana. Depredation of livestock by carnivores therefore poses economic loss to farmers. Although the Botswana Wildlife Conservation and National Park Act of 1992 permits lethal control of wildlife in defence for human property damaged, Botswana is still one of the countries in Southern Africa that pay compensation for human properties damaged by wildlife as a measure to balance the trade-off between wildlife conservation and the impact of conflict they bring to the farming communities. Before compensation is paid, a thorough investigation on determining the wildlife species that has caused damage is mandatory. Because of insufficient resources by the Department of Wildlife and National Parks (DWNP), the initial investigation is carried out by herders. Necessary skills are needed to best find out which carnivore species is causing damage. Three basic indicators that are used to determine carnivore predation are sighting the carnivore at the kill (hereafter referred to as visual), tracks of the predator (hereafter referred to as tracks) and examining the carcasses (bite marks, body parts eaten and location of the carcasses, hereafter referred to as kill patterns).

Because information on livestock raiding carnivores in Botswana is currently insufficient (Hemson 2003; Valeix et al. 2012), this study used LEK possessed by herders to assess herders’ ability to identify the carnivore species causing damage using the three indicators mentioned above and mapped their perceptions on magnitude and spatiotemporal patterns of large carnivore predation on livestock. One of the significant aspects of using LEK is to enhance the contribution of local people in policy decision-making in order to promote their participation in proper management of natural resources around them (Calamia 1999). The objectives of this study were to (1) compare carnivore predation reported to and confirmed by the DWNP (hereafter refer to as PAC) with reports made during interviews (hereafter referred to as LEK) for lion, leopard and wild dog; (2) compare the carnivore predation risk at cattle post (an area where the main activity is livestock farming) as determined by data collected by PAC and LEK and (3) assess herders’ ability to identify carnivore species from visual photos, tracks and kill patterns.

Although LEK is not commonly used in mapping the spatial and temporal patterns of human–carnivore conflict, the results of this study show that confirmed carnivore predation incidents and reports gathered through interviewing herders did not significantly differ for all the three carnivores and distances from protected areas. This shows that carnivore predation information collected through LEK and the standard protocols used by DWNP were complementary. Similar results have been observed with other species such as tortoise distribution in south-eastern Spain (Anadón et al. 2009), fish and fisheries management in the Beaufort Sea, Canada (Hartwig 2009b), southern Mexico (Espinoza-Tenorio et al. 2013), south-eastern Brazil (Teixeira et al. 2013) and wild mammals in northern Gonarezhou National Park, Zimbabwe (Gandiwa 2012). The results further show that both LEK and PAC data were in agreement over the fact that lion predation decreases with increasing distance from protected areas (Figure 3a) and leopard incidents were evenly distributed throughout the study area (Figure 3b). Contrary to the other two carnivores, there was disagreement between LEK and PAC data on the distribution of wild dog predation incidents in the study area, with PAC data indicating that wild dog predation incidents were evenly distributed and LEK indicating that wild dog predation incidents increased with decreasing distance from protected areas. The results are partly comparable to the findings that stronger carnivores such as lion are more problematic in the periphery of protected areas, while outcompeted smaller ones are forced to forage deeper into human-dominated landscapes (Gusset et al. 2009; Mogwera & Rutina 2015; Rutina et al. 2015). It is also apparent that during the dry season, carnivore problems are more pronounced, particularly because the preferred wild prey has migrated to their dry season habitats, thus leaving livestock in high abundance and as alternative prey (Valeix et al. 2012).

In general, LEK and PAC were in agreement in categorising cattle posts for carnivore predation risk areas. In addition, the study shows that the chances that LEK predictions for carnivore predation risk to be similar to confirmed incidents were high for both the three carnivore at all distances from protected areas. This is consistent with other studies that showed LEK information complement scientific information in mapping environmental hazards (Anh, Phong & Mulenga 2014; Raungratanaamporn & Pakdeeburee 2014; Udmale, Ichikawa & Manandhar 2014), fisheries (Silvano & Begossi 2012), natural resource management (Gilchrist et al. 2005; Ruddle & Calamia 1999), environmental planning (Kujala et al. 2013), natural resources research (Anadón et al. 2009; Gandiwa 2012; Silvano & Begossi 2012) and HWC hotspots (Calamia 1999; Hartwig 2009a; Rutina et al. 2015; Strindberg et al. 2007). This significant contribution of LEK to map carnivore predation areas indicates the potential impotence of LEK in HWC management in the region.

We expected that that herders, having grown up with wildlife could easily identify carnivores visually and their tracks. However, we found that herders can identify large carnivore species visually but had difficulties in identifying carnivore tracks and kills patterns. This could be attributed to the current herding practises that are carried in the study area. Gontse et al. (2012) found that because of the paying schedule from the Department of Labour Affairs, farmers are not able to pay herders for all the activities involved in the practice of herding. As such, herders are now taking livestock to water points and kraal that arrive back during the afternoon (Gontse et al. 2012; Rutina et al. 2015). Contrary to the results of this study, information from the DWNP investigation diaries suggests confirmed 82% of lion-related incidents, 67% of wild dog-related incidents and 100% of leopard-related incidents identified by herders who used their ecological knowledge. Furthermore, data from diaries suggest that herders use tracks and/or patterns of kill to identify carnivores causing damage. As we have shown above, herders had difficulties in identifying carnivore tracks and kill patterns (Figure 5). We thus recommend that more education regarding identification of carnivore tracks and kill behaviour is needed for herders in the study area. Nevertheless, the results of the study show that most of the carnivore predation on livestock information was from LEK (e.g. number of cases [Figure 3], risk areas [Figure 4] and visual identification of carnivores [Figure 5]) mentioned by the herders corresponded with information obtained from scientifically recognised methods.

Lack of resources by the DWNP to mitigate and attend to carnivore predation on livestock in the study area have been suggested as one factor contributing to human–carnivore conflict in the study area (Gontse et al. 2012; Hemson 2003; Hemson et al. 2009; Rutina et al. 2015; Valeix et al. 2012). Because of the inadequate resources by DWNP, most of HWC mitigation measures are conducted at a large scale and use a ‘one size fits all’ strategy. Despite different opinions among scientists and policy-makers as to the use of LEK in natural resource management (see Mclain et al. 2013) and given the complexity and nature of HWC, mitigation strategies need to be area and species specific. The difficulties are mostly because of lack of mutual understanding among stakeholders, including local communities, towards wildlife management approaches, and this undermines the effectiveness of integrating LEK in HWC. Coordination among government, communities and scientists has been suggested as the best strategy to manage natural resources (Espinoza-Tenorio et al. 2013; Gilchrist et al. 2005; Mclain et al. 2013; Neagh et al. 2008; Phuthego & Chanda 2004). Based on the results of this study, we argue that decentralising HWC to local communities using existing government structures that exist at local level, such as district conservation committees, village development committees and village farmer committees will improve the problem. This will not only supplement the inadequate resources by the DWNP to effectively mitigate the problem, but also empower local communities in wildlife management and tap on the LEK the community have. The implementation of community-based natural resources management goes along training communities to collect data for monitoring purposes and inform government strategies using a Management-Oriented Monitoring System (MOMS). The use of MOMS in carnivore predation investigations will significantly improve information on management of human–carnivore conflict. This should include the recognition that in order to increase the effectiveness of reducing or mitigating carnivore predation on livestock, communities should play important roles by applying their LEK understanding of the spatiotemporal dynamics of carnivore predation rather than focusing on increasing DWNP capability to mitigate HWC.

Over the past 10 years, southern African governments have been engaged in creating Transfrontier Conservation Areas, consisting of a mosaic of different land uses. Within this mosaic agricultural landscapes play an important role as habitat and corridors for wide-ranging wildlife, such as large carnivores, because the sizes of most protected wildlife areas are not large enough to adequately accommodate their home ranges. Our study site forms part of the Kavango–Zambezi Transfrontier Conservation Area (KAZA TFCA), a conservation area covering part of Angola, Botswana, Namibia, Zambia and Zimbabwe. Both countries, through revision of their National Biodiversity Strategy and Action Plans (NBSAPs) to align them with CBD 2015 goals and 2020 Aichi targets, recognise that HWC hinders the promotion of integration between biodiversity conservation and human well-being and have prioritised HWC as priority areas that need to be addressed in their Integrated Development Plans (IDPs) for their KAZA TFCA components. The results of our study suggest that involving local communities in managing HWC could help the KAZA TFCA mandate to be realised.

The results of the study showed that both PAC and LEK information were complementary in reporting incidents of carnivore predation on livestock and categorising cattle posts for carnivore predation risk for the three carnivores studied. This calls for involvement of local communities in carnivore predation investigations and management. This will not only complement the shortage resources by government to manage conflict, but also reduce the spatial scale at which conflict management is conducted. This can be done by decentralising HWC management to local communities using government structures that exist at local level. However, herders had difficulties in identifying carnivore tracks and kill patterns. This could be due to partly poor pictures used because information from the DWNP investigation diaries suggests that almost all confirmed incidents were identified by herders using tracks and/or patterns of kill.