The following is the established format for referencing this article:
Zvidzai, M., K. K. Mawere, R. J. N'andu, H. Ndaimani, C. Zanamwe, and F. M. Zengeya. 2023. Application of maximum entropy (MaxEnt) to understand the spatial dimension of human–wildlife conflict (HWC) risk in areas adjacent to Gonarezhou National Park of Zimbabwe. Ecology and Society 28(3):18.ABSTRACT
The application of empirical and spatially explicit information to understand the spatial distribution of human–wildlife conflict (HWC) risk zones is increasingly becoming imperative to guide conservation planning and device mechanisms to enhance and sustain the coexistence between wildlife and humans. Spatial information on HWC is scarce in the literature, and previous studies have tended to concentrate more on the human dimensions of HWC. Although normally applied in wildlife studies, species distribution modeling (SDM) is becoming an indispensable tool to predict and visualize potential risk zones for HWC. In this study, we used maximum entropy (MaxEnt), a presence-only SDM to predict the potential distribution of HWC risk zones and to determine ecological variables that significantly explain the spatial distribution of HWC occurrences around the Gonarezhou National Park (GNP) in southeastern Zimbabwe. Our results show that HWC risk zones are not randomly distributed but tend to be concentrated along areas adjacent to protected areas that support potential overlaps and contacts between wildlife and human landscapes. A distinctive HWC high-risk zone is observed north of GNP, around areas such as Chitsa, Mpinga, and Masekesa—communities that should be prioritized for proactive mitigation interventions. In view of limited conservation resources typical of less developed countries, wildlife managers are pressed to explicitly determine zones with the highest HWC risks for effective and targeted interventions. Findings from this study thus provide a crucial baseline for identifying potentially high-risk HWC zones and the main predictors, knowledge that can be streamlined for proactive resource allocation to mitigate the HWC challenge.
INTRODUCTION
Human–wildlife conflict (HWC) is a prevalent challenge, especially in African countries, with the potential to wipe away all the global biodiversity conservation and sustainable development gains achieved so far (Long et al. 2020, Egri et al. 2022). For instance, a study by Nayeri et al. (2022) highlights that more than a quarter of global wildlife mortalities are attributed to HWC, making it one of the major threats to biodiversity. As land under agricultural activities expands to meet the needs of a growing human population, the frequency, intensity, and diversity of HWC tend to escalate, especially in communities located adjacent to protected areas, regions that are often defined as human–livestock–wildlife interface zones (Long et al. 2020). Studies have shown that the intensity and complexity of HWC are positively correlated with an expanding human footprint (Chansa et al. 2012, McNutt et al. 2017, Khattak et al. 2021). The expansion of human footprint in the form of agriculture, tourism, and human settlements and infrastructures complicates any attempts to foster coexistence between wildlife and human population as it aggravates negative repercussions on both biodiversity and human livelihoods (Redpath et al. 2015, König et al. 2021). HWC is often triggered primarily when people and wildlife compete for limited resources in the form of space, food, and water. Besides, Human–wildlife conflict may also be prompted when divergent views, values, and behaviors about wildlife emerge from a wide range of key stakeholders pushing conflicting social, economic, and ecological agendas (Redpath et al. 2015, König et al. 2021).
To this end, several definitions for HWC exist in literature. For example, Makindi et al. (2014) view HWC as a phenomenon that occurs when the needs or behavior of wildlife have a negative bearing on human livelihoods and well-being or when humans engage in activities that impact negatively on wildlife and biodiversity conservation initiatives (Makindi et al. 2014). Frank (2016) argues that HWC discourse may be framed under three fundamental dimensions, that is when wildlife impact negatively on human well-being; when humans impact negatively on wildlife; and when conflicts arise among humans over different attitudes and perceptions on wildlife (Frank 2016). According to Madden (2004), HWC tend to manifest when wildlife physiological requirements are deemed to provoke direct or indirect negative consequences on the aims and aspirations of humans. Accordingly, HWC can manifest in various forms ranging from livestock depredation by carnivores, disease transmission between wildlife and livestock, crop raiding, and human injury or death from wildlife such as African elephant (Loxodonta Africana), African lions (Panthera leo), crocodiles (Crocodylus niloticus), and African buffaloes (Syncerus caffer). On the other hand, people may also perpetrate HWC through deliberate poisoning and retaliatory killing of the perceived problem animals, avian mortalities induced by collisions with human infrastructure, poaching, and habitat destruction and fragmentation, (Distefano 2005, Mateo‐Tomás et al. 2012, Frank 2016, Long et al. 2020, Göttert and Starik 2022). These human-induced disturbances and mortalities have resulted in widespread species extinctions that are well documented but rarely feature under the HWC discourse (Göttert and Starik 2022).
A review of literature shows that studies that model spatially explicit distribution of HWC occurrences are currently few, but on an increasing trajectory (Mohammadi et al. 2021, Egri et al. 2022, Khosravi et al. 2023). This could probably be due to lack of HWC data, which are often sensitive and illegal in nature, thus difficult to access freely (Mateo‐Tomás et al. 2012, Egri et al. 2022). For example, HWC occurrences such as poaching and poisoning are largely disguised, thus making the availability of such data in the public domain a huge challenge. Besides, lack of user-friendly geospatial applications to record and disseminate verifiable HWC occurrence data tends to exacerbate the challenges associated with HWC data scarcity.
For instance, a study by Nayeri et al. (2022) used coarse data from subjective media sources to model the spatial distribution of human-driven hotspots of brown bear (Ursus arctos) mortality events in Iran, and a study by Patricia Mateo-Tomas et al. (2012) in Spain also assessed HWC targeting several species based on data from several sources such as press releases, private institutions, and non-governmental organizations, most of which needed to be verified (Mateo‐Tomás et al. 2012, Nayeri et al. 2022). Other studies have focused on the potential spatial distribution of human–carnivore conflicts hotspots as a tool to prioritize the conservation of lions, whose populations are facing serious threats from human–wildlife conflicts (Broekhuis et al. 2017, Mohammadi et al. 2021). However, indications are that, in the majority of cases, the study of HWC is descriptive in nature without any specific georeferencing thus, rendering its application for understanding the spatial dimension of HWC problematic (Makindi et al. 2014, Mukeka et al. 2018, Long et al. 2020, Khattak et al. 2021). For example, several studies have focused more on the social dimensions of HWC with much emphasis on the social and economic impacts of HWC as well as prevention and mitigation measures (Marchini and Crawshaw 2015, Mhuriro-Mashapa et al. 2018, Long et al. 2020). Other studies have also focused on the characterization of wildlife species that are commonly involved in HWC and the target crops and livestock species (Distefano 2005, Matseketsa et al. 2019, Long et al. 2020, Khattak et al. 2021, Pisa and Katsande 2021).
Application of geospatial and remote sensing technologies, coupled with species distribution modeling (SDM) tools such as maximum entropy (MaxEnt), is increasingly becoming a pressing requirement for explicitly providing a spatial dimension to the HWC discourse (Phillips et al. 2006, Cushman et al. 2018, Sharma et al. 2020, van Bommel et al. 2020). Gaining knowledge of the specific areas prone to HWC risks as well as clear insights of the ecological and social covariates correlated with HWC events is becoming indispensable for developing effective and sustainable mitigatory strategies (Messmer 2009, Songhurst 2017, Mukeka et al. 2018). Although it is traditionally used for species distribution, MaxEnt is increasingly becoming a potent tool across a range of social and ecological applications, especially where data availability is constrained (Phillips et al. 2006, Phillips and Dudík 2008, Sharma et al. 2020). Data on HWC are often scarce and incomplete, thus making MaxEnt a prime candidate modeling tool for statistical treatment of such data (Mateo‐Tomás et al. 2012).
This study expands our knowledge on HWC occurrences as it provides one of the few attempts that integrate the application of SDM tools and ecological covariates to enhance our understanding of the spatial dimension of HWC risk occurrences in a data-scarce region. Specifically, the study aims to determine the ecological variables that significantly predict the occurrence of HWC hotspot zones. The study further uses the SDM techniques to generate a spatially explicit HWC risk map as a tool to inform effective and proactive conservation and mitigation planning. This is imperative for guiding strategic prioritization and deployment of often scarce conservation resources in monitoring and mitigating HWC to maximize benefits for both wildlife and human well-being.
MATERIALS AND METHODS
Study Area
Gonarezhou National Park (GNP) is the second largest national park located in the southeastern region of Zimbabwe and occupies an area of approximately 5053 km² (Gandiwa et al. 2013). Gonarezhou National Park is known to be home to a wide diversity of species, including large herbivores and carnivores, such as African elephant, African buffalo, and African lion, which are also largely the primary perpetrators of HWC incidents (Gandiwa 2012, König et al. 2021, Pisa and Katsande 2021).
The current study was conducted in 10 communities that are located adjacent to GNP namely, Masekesa, Mpinga, Chitsa, Chibwedziva, Tswanani, Batanai, Gonakudzingwa, Chibhavalengwe, Malipati, and Chipimbi. The communities that constitute the study area stretch from 31° 05' 23" to 32° 14' 46" East and from 20° 49' 35" to 22° 23' 46" South and cover about 3956 km² of land bordering GNP. The area is a semi-arid region experiencing high temperatures, averaging around 27°C in winter and 31°C in summer (Musakwa et al. 2020). The average annual precipitation received in the area is approximately 470 mm and is received usually between November and March (Mugandani et al. 2012). The elevation in the study area ranges from about 200 m to 620 m and has the lowest elevation record in Zimbabwe, which is located at the confluence of the Save and Runde rivers in the northern side of GNP (Fig. 1).
Extensive livestock rearing, crop production, and wildlife production are the dominant activities in the region (Gandiwa et al. 2013). However, illegal activities such as poaching, retaliatory poisoning of predators, and deforestation are also rampant (Gandiwa et al. 2012, Mutanga et al. 2017, Makumbe et al. 2022). The study area has a long history of HWC incidents that is well documented (Mombeshora and Le Bel 2009, Gandiwa et al. 2012, Tavuyanago 2016), and that prompted the adoption of Community Areas Management Programme for Indigenous Resources (CAMPFIRE) to enhance tolerance and coexistence between people and wildlife with minimum negative consequences (Gandiwa et al. 2013). Besides, lack of spatially explicit data on HWC occurrences in the study provides a good opportunity to test the applicability of MaxEnt modeling tools to understand the spatial dimensions of HWC.
Human–Wildlife Conflict Incidence Data
We obtained the occurrence data of HWC from the Chiredzi Rural District Council (CRDC) database. The data were recorded in two formats, first as precise GPS locations, and then as descriptive data or the name of the place where the HWC incident was observed. The HWC records showed the type of conflict, such as livestock depredation, crop raiding, wildlife killing, human deaths or injuries, wildlife species involved, crops and livestock targeted, and date of occurrence. We then cleaned the data using Microsoft Excel 2013 and Quantum GIS 3.6. The records with redundant data, missing attributes, incorrect GPS locations, and unclear descriptions were removed. After cleaning, only 150 records of HWC incidents from the 10 communities surrounding GNP collected between 2019 and 2021 were included in the model. The sample size (n = 150) was deemed sufficient as studies have shown that MaxEnt can perform well even with small sample sizes (Phillips and Dudík 2008, Elith et al. 2011).
Environmental Variables Used to Predict Human–Wildlife Conflict Risk Zones
In this study, four environmental variables were considered potentially relevant for predicting HWC risk zones. These include distance of the HWC location from rivers, land use land cover, distance of the HWC location from settlements, and distance of the HWC location from the park boundary (Leta et al. 2021). All raster layers for the predictor variables were resampled to a similar spatial resolution of 30 m. The projection of the output layers was changed from the Geographic Coordinate System to the Universal Transverse Mercator (UTM) Coordinate System. All output layers were then clipped using the boundary of the study area (Fig. 2). To test for multicollinearity between the predictor variables, we used the variance inflation factor (VIF) (Gholami et al. 2020). In this study, the predictor variables with VIF greater than five were considered to show high multicollinearity and were eliminated from the model to improve model performance. However, all the predictor variables were included in the model as they had a VIF less than five.
Distance of the Human–Wildlife Conflict Event from Park Boundary
We calculated the distance of HWC location from the GNP boundary using the Euclidean Distance function in ArcMap Version 10.3 (Fig. 2). Park boundaries were created to separate human population from wildlife as the two are considered incompatible (Mombeshora and Le Bel 2009). As human population grows, the demand for land for cultivation and livestock grazing increases, and this pushes human population to occupy land close to protected areas (Frank and Glikman 2019). This tends to increase the frequency and intensity of HWC as competition for space, water, and grazing land escalates.
Distance of the Human–Wildlife Conflict Event from Settlements
In this study, we used the settlements to indicate the effects of human footprint. We digitized the settlements in the communities surrounding GNP using Google Earth Pro. We calculated the distance of HWC location from the digitized settlements using the Euclidean Distance function in ArcMap 10.3 (Fig. 2). Livestock depredation and crop raids from wildlife normally occur in settlements located closer to protected areas. The further the settlements are from protected areas, the less they are exposed to livestock depredation and crop raids from wildlife. Besides, other HWC such as poaching and wildlife poisoning are also common in protected areas near human settlements. Thus, the distance of HWC events from settlements is an important variable as it tends to influence the nature and extent of HWC (Sharma et al. 2020).
Distance of the Human–Wildlife Conflict Event from Water Sources
We used distance to rivers as a proxy to water availability and to indicate the effects of water sources on the spatial distribution of HWC. We delineated the river network from the digital elevation model (DEM) using the hydro-processing tools in the Integrated Land and Water Information Systems (ILWIS) Version 3.3. The DEM was downloaded from the Open Topography website (https://opentopography.org/). We calculated the distance of HWC location from the delineated rivers using the Euclidian Distance function in ArcMap Version 10.3 (Fig. 2). Water is limited in the study area and is considered one of the major drivers of human–livestock–wildlife interactions and contacts. For example, the co-occurrence of people, livestock, and wildlife at few water sources, may be a pathway for the transmission of diseases, depredation of livestock, injuries and death to people from wildlife attacks, and deliberate poisoning and poaching of wildlife (Zvidzai et al. 2013, Sharma et al. 2020).
Land Use Land Cover (LULC)
We downloaded Landsat 8 Operational Land Imager (OLI) satellite imagery from the United States Geological Survey (USGS) website. Imagery with a resolution of 30 m and cloud cover of less than 30% was selected. Three scenes of the imagery covering the study region were used. We converted the image values from digital numbers (DN) to the Top of Atmosphere Reflectance using the Map Algebra in ArcMap Version 10.3. We then used the following band combinations: natural color composite (red, green and blue) and enhanced color composite (shortwave infrared1, near infrared and green) to enhance feature extraction during the image classification process (Vivekananda et al. 2021). We performed supervised image classification using the maximum likelihood classification algorithm in ArcMap Version 10.3. The imagery was classified into five LULC classes, including water, grassland, woodland, cropped fields, and other (bare land and built-up area) (Fig. 2). A grassland is mainly dominated by extensive grass and grass-like plants with limited woody cover whereas a woodland habitat is mainly composed of trees of many species and is often a transitional zone to forests (Tassicker et al. 2006, Gibson 2009). We used Google Earth Pro to extract reference points that were used as verification data. We then performed image accuracy assessment using the Confusion Matrix and Kappa Statistic in ArcMap 10.3 and Microsoft Excel 2013 (Foody 2020, Verma et al. 2020). Land use land cover is an important variable for explaining HWC events, especially vegetation cover as a source of food to herbivores such as livestock, which subsequently attract carnivores to areas of high vegetation cover to enhance their hunting. In some cases, livestock are forced to track high quality forage in protected areas, and this exposes them to depredation by wild predators. People are also forced to track natural resources such as fruits, thatching grass, and medicinal plants in protected areas, and this equally exposes them to wildlife attacks.
Human–Wildlife Conflict Risk Model Development
To model HWC risk in the 10 communities close to GNP, we used the Maximum Entropy (MaxEnt) approach, a presence-only SDM technique (Phillips et al. 2006, 2017). Previous studies have shown that MaxEnt is robust in modeling ecological interactions (Phillips and Dudík 2008, Baldwin 2009, Elith et al. 2011).
The following variables: distance from water sources, distance from settlements, distance from the park boundary, and Land Use Land Cover were used in the model as predictor variables, whereas HWC incidents were used as a dependent variable. The predictor variables were converted from GeoTiff to the MaxEnt-readable ASCII file format. We then split the data into training and testing data sets, with 70% of the data used for model execution and 30% used for model testing. A study by Tobeña et al. (2016) supported the importance of setting iterations and data for random model testing. A model default setting of 10,000 background points that were randomly selected from the study area were used for the model execution, and it was replicated for 30 times to enhance the precision of the model. For other parameters, the default settings that have been tested and validated were also applied to build the model (Phillips and Dudík 2008). To give insights of the relative importance of each predictor variable to the model, we used the Jacknife significance tests (Padalia et al. 2014). Response curves were used to assess the correlations between the probability of HWC occurrence and the four predictor variables. The output risk map was generated by use of the default format of Cloglog, as it provides an estimate of the relative suitability of one pixel compared with the other. The probability of presence ranges between 0 being the lowest and 1 representing the highest probability of HWC occurrence (Phillips 2005, Sharma et al. 2020). We used the area under the ROC curve (AUC) to assess the performance of the HWC model. The AUC values range between 0 and 1, with values <0.5 suggesting no discrimination, 0.5–0.69 reflecting poor performance, 0.7–0.79 reasonable performance, 0.8–0.89 excellent, and values >0.9 suggesting exceptional performance of the model (Sharma et al. 2020).
RESULTS
The predictive power of the model, based on the Area Under Curve (AUC) shows good predictive performance with an AUC of 0.87 (Fig.3). This shows that the four predictor variables used in the current study significantly explain about 88% of the spatial variability in HWC, which suggests that the model is effective and suitable for explaining the probability of HWC incidences in areas adjacent to GNP.
Among the four predictor variables used in the HWC risk model, we observed that when used in isolation, distance from the park boundary was the most significant variable followed by distance from settlements. Distance from water sources (rivers) was the least contributing variable to the HWC risk model when used in isolation. In addition, the results showed that omitting individual predictor variables has a considerably small effect on model performance and that the highest gain was achieved with all the predictor variables included in the model (Fig. 4).
Results from this study indicate that HWC risk decreases with increasing distance from the park boundary. Findings also show that the highest probability of HWC occurrence is close to human settlements and becoming weaker as distance from settlements increases (Fig. 5). The effect of distance from water sources on HWC risk is weaker when compared with that of distance from park boundary and settlements. We observe that five LULC types (water, cropped fields, other, grassland, and woodland) have varying effects on the spatial distribution of HWC risk. For instance, the lowest risk of HWC is associated with areas covered by woodland, whereas the highest risk is associated with other LULC types.
The HWC risk zone map indicates that the highest risks of HWC are clustered in the northern side around Masekesa, Chitsa, and Mpinga areas. More than 80% of this zone has a very high probability of HWC risk. The general pattern, however, is that HWC risk tends to decrease with increasing distance from the park boundary, except for some isolated pockets of high-risk zones further away from the park boundary (Fig. 6).
DISCUSSION
Understanding the spatial distribution of HWC risk zones and their association with ecological variables is fundamentally important as it provides insights to park managers and communities to develop high impact intervention strategies targeting high-risk areas. Such information may also guide timely and effective deployment of limited resources such as patrol and reaction teams to HWC hotspot areas to reduce adverse impacts of conflicts and support mechanisms for co-existence between people and wildlife (Frank 2016, Frank and Glikman 2019, Long et al. 2020, Khattak et al. 2021).
A key finding from the current study demonstrates that HWC incidences are not distributed randomly across the landscape but tend to cluster close to the park boundary and human settlements. Specifically, the results suggest a significant negative relationship between the likelihood of HWC occurrences and distance from the park boundary as well as from the settlements. A study by Distefano (2005) in the Caprivi region of Namibia is consistent with findings from this study as it also demonstrated that cases of HWC were highest in the villages bordering national parks. Another study by Matseketsa et al. (2019) showed similar findings for areas bordering Save Valley Conservancy in Zimbabwe, and a similar study in Zambia also showed clustering of HWC in Wildlife Management Areas (WMAs) adjacent to protected areas (Chansa et al. 2012). These findings confirm that the proximity of human landscapes to protected areas heightens encounters with wildlife as wildlife species perform their usual activities such as foraging to fulfil their physiological needs. Under these circumstances, management actions such as the creation of buffer zones between protected areas and human settlements is recommended as an appropriate response strategy (McNutt et al. 2017, Songhurst 2017, Teixeira et al. 2021). However, unlike previous studies, the major advantage of the current study is its application of spatially explicit modeling techniques that pinpoint potentially high-risk zones of HWC, thus making findings from the current study of higher utility to park managers, communities, local government departments, and other HWC stakeholders. For instance, where resources are constrained, our results may inform strategic deployment of such limited resources to high HWC risk zones to intensify mitigation actions.
Evidence from previous studies has shown that humans have coexisted with wildlife for several decades and issues of tolerable HWC are well documented (Distefano 2005, McNutt et al. 2017). However, in recent years, HWC incidents have escalated, and this can be attributed to several factors (Messmer 2009). For example, the expansion of human population has increased demand for land for cultivation, livestock grazing, and access to other resources such as water (Mombeshora and Le Bel 2009, Osipova et al. 2018). This has pushed human populations to establish their settlements closer to the protected area boundaries, and in some extreme cases, human settlements have been established inside protected areas boundaries (Mombeshora and Le Bel 2009, Givá and Raitio 2017, Muboko and Bradshaw 2018). Increased human settlements near protected areas often result in competition for resources among humans, livestock, and wildlife, resulting in elevated HWC occurrences. A study by Makumbe et al. (2022) showed that the frequency and intensity of HWC occurrences tends to be highest in newly resettled communities that have encroached much closer to protected areas and transformed wildlife habitats to human landscapes (Le Bel et al. 2011, Matseketsa et al. 2019, Makumbe et al. 2022).
As distance between protected areas and human settlements decreases, the propensity for species such as elephants to stray into agricultural fields during the night increases. Previous studies have shown that HWC, especially involving elephant raids on fields and lion, spotted hyena (Crocuta crocuta), and black-backed jackal (Canis mesomelas) attacks on livestock, tends to increase during the night (Distefano 2005, Le Bel et al. 2011). Our observations in the field have shown that after forays outside the park, elephants and buffaloes especially tend to stampede back to their refugia in protected areas, to avoid danger from legal trophy hunters, poachers, and problem animal control authorities. Such a strategy is only feasible when such crop fields are located close to protected areas. Thus, the further human settlements are from protected areas, the less attractive they become to wildlife as chances to quickly move back to their safe refugia before daylight are limited, thus exposing them to serious threats (Bhatia et al. 2020). Noticeably, livestock are also forced to forage inside protected areas in their proximity due to the scarcity of resources such as water and vegetation, thus exposing themselves to high predation risk by carnivores.
Studies have shown that populations of some wildlife species have increased because of successful conservation efforts from government agencies and their partners (Distefano 2005, Osipova et al. 2018, Makumbe et al. 2022). Increased wildlife populations are creating new challenges as protected areas are no longer adequate to provide enough resources to the expanding wildlife populations (Messmer 2009, Tavuyanago 2016). Due to inadequate resources, wildlife populations are forced to foray out of protected areas into adjacent human-dominated landscapes. Consequently, it can be argued that the expansion of both human and wildlife populations is increasing pressure on the limited resources (Matseketsa et al. 2019), and this has resulted in increased co-occurrence and interactions, a scenario that has escalated HWC occurrences.
Regarding LULC, results suggest that the lowest risk of HWC is associated with areas covered by woodland, and the highest risk is associated with other LULC types, which include bare and built-up areas (Fig. 5). Woodland areas are normally characterized by limited human activities, and livestock normally avoid areas with high vegetation coverage (McNutt et al. 2017). As a result, human–wildlife encounters and livestock predation are less common, which may explain why such land cover areas have the least risk of HWC. On the other hand, previous literature has shown that other LULC types such as built-up and bare areas, landscapes associated with human settlements, tend to have the highest HWC risks (Mukeka et al. 2018). Illegal activities such as poaching are usually common in areas close to settlements to avoid detection by park patrol teams, which are normally deployed to focus on areas inside the park boundary (Long et al. 2020).
We expected distance to rivers, a proxy for water availability, to be one of the most important explanatory variables for a HWC model. This is because water is naturally scarce in the Gonarezhou ecosystem, and this forces people, livestock, and wildlife to co-occur and compete for few available water sources, resulting in conflicts (Zvidzai et al. 2013). Competition for the few available water sources is suspected to be a major determinant for HWC hotspots. For example, co-occurrence at few water points may result in livestock depredation and human injuries and deaths due to attacks from crocodile, elephants, and buffaloes. Studies have also shown that wildlife poaching is common around water sources where wildlife densities are high, thus making it more convenient for poaching activities (Songhurst 2017, Pisa and Katsande 2021, Teixeira et al. 2021). In this regard, the little contribution by the distance from river variable to the model, when used in isolation (Fig. 4), runs contrary to previous studies (Chansa et al. 2012, Matawa et al. 2012), and may call for future investigations. However, based on the response curves, results suggest that the probability of HWC occurrence is highest close to water sources and decreases as distance from water sources increases (Fig. 5). This supports the view that in environments where water is constrained, the location of water sources can indeed be hotspots of HWC incidents such as depredation, human injuries and deaths, and poaching, and thus may require prioritization in terms of mitigatory interventions (Bhatia et al. 2020, Khattak et al. 2021).
Results suggest communities in the northern part of GNP such as Chitsa, Mpinga, and Masekesa to be at highest risk of HWC. Previous studies have shown that this is one of the most contentious areas around GNP, with the history of the Chitsa people’s contestations to the park boundary demarcations well documented (Mombeshora and Le Bel 2009, Muboko and Bradshaw 2018). After the accelerated land reform program, which started in 1999, the Chitsa community, with political support, established their settlements close to the park boundary and with some few villages established inside the park. Their proximity to the park means the frequency and severity of crop raids and depredation of their livestock is much higher compared with settlements that are further away from the park. It also means their conflicts with park rangers are elevated as they tend to illegally graze their livestock, collect natural resources, and hunt in adjacent protected areas. Thus, they have entrenched negative perceptions about wildlife, which they view as pests responsible for destroying their main source of livelihoods in the form of crops and livestock without any recourse from responsible authorities (Mombeshora and Le Bel 2009, Gandiwa 2012, Mutanga et al. 2017, Matseketsa et al. 2019). Yet they face stiff penalties for any retaliatory killing of such problem animals. In this regard, such communities need to be prioritized in the development of any proactive intervention strategies to mitigate HWC. Reactive intervention measures that are implemented after the conflict has occurred may not be helpful as losses and burdens will have already been incurred (van Bommel et al. 2020).
CONCLUSION
In the context of the increasing encroachment of human footprint in areas adjacent to protected areas and the increasing wildlife populations in some protected areas, it is undeniable that HWC may continue to be a permanent feature at the human–livestock–wildlife interface (Madden 2004). Modeling the potential spatial distribution of high-risk zones of HWC and the significant predictors is emerging as a useful tool to design effective and timely interventions. To this end, findings from this study are useful as they can be used to inform strategic deployment of limited resources to zones with the highest risk of HWC to implement innovative and integrative management strategies to mitigate HWC. However, to achieve maximum outcomes, we recommend that spatial models developed in this study should provide guidelines that should be integrated with indigenous knowledge systems on the management systems of HWC (van Bommel et al. 2020). Based on the results from the study, park managers should look beyond protected areas if they are to effectively deal with the threats posed by HWC, and affected communities should be capacitated to develop mechanisms to co-exist with wildlife. In the meanwhile, we draw attention for future studies to incorporate numerous but relevant predictor variables to enhance the usefulness of spatial predictions and to provide a wider spectrum of potential drivers of HWC.
RESPONSES TO THIS ARTICLE
Responses to this article are invited. If accepted for publication, your response will be hyperlinked to the article. To submit a response, follow this link. To read responses already accepted, follow this link.
ACKNOWLEDGMENTS
We sincerely thank the Chiredzi Rural District Council for their willingness to provide data on HWCs. We also thank the reviewers and editors for providing valuable comments to help improve our manuscript.
DATA AVAILABILITY
Data used in this study are spatially geo-referenced and highly sensitive and thus may not be made public as this may violate the ethical principles that were applied during the data collection protocols.
LITERATURE CITED
Baldwin, R. A. 2009. Use of maximum entropy modeling in wildlife research. Entropy 11(4):854-866. https://doi.org/10.3390/e11040854
Bhatia, S., S. M. Redpath, K. Suryawanshi, and C. Mishra. 2020. Beyond conflict: exploring the spectrum of human–wildlife interactions and their underlying mechanisms. Oryx 54(5):621-628. https://doi.org/10.1017/S003060531800159X
Broekhuis, F., S. A. Cushman, and N. B. Elliot. 2017. Identification of human–carnivore conflict hotspots to prioritize mitigation efforts. Ecology and Evolution 7(24):10630-10639. https://doi.org/10.1002/ece3.3565
Chansa, C., S. Ramadhani, C. Harry, M. Jacob, and N. Vincent. 2012. Patterns of human wildlife conflicts in Zambia, causes, consequences and management responses. Journal of Ecology and the Natural Environment 4(12):303-313. https://doi.org/10.5897/JENE12.029
Cushman, S. A., N. B. Elliot, D. Bauer, K. Kesch, L. Bahaa-El-Din, H. Bothwell, M. Flyman, G. Mtare, D. W. Macdonald, and A. J. Loveridge. 2018. Prioritizing core areas, corridors and conflict hotspots for lion conservation in southern Africa. PloS One 13(7): e0196213. https://doi.org/10.1371/journal.pone.0196213
Distefano, E. 2005. Human–wildlife conflict worldwide: collection of case studies, analysis of management strategies and good practices. Food and Agricultural Organization of the United Nations (FAO), Sustainable Agriculture and Rural Development Initiative (SARDI), Rome, Italy. https://tnrf.org/files/E-INFO-Human-Wildlife_Conflict_worldwide_case_studies_by_Elisa_Distefano_no_date.pdf
Egri, G., X. Han, Z. Ma, P. Surapaneni, and S. Chakraborty. 2022. Detecting Hotspots of human–wildlife conflicts in India using news articles and aerial images. Pages 375-385 in COMPASS22, Proceedings of the 5th ACM SIGCAS/SIGCHI Conference on Computing and Sustainable Societies (COMPASS). https://doi.org/10.1145/3530190.3534818
Elith, J., S. J. Phillips, T. Hastie, M. Dudík, Y. E. Chee, and C. J. Yates. 2011. A statistical explanation of MaxEnt for ecologists. Diversity and Distributions 17(1):43-57. https://doi.org/10.1111/j.1472-4642.2010.00725.x
Foody, G. M. 2020. Explaining the unsuitability of the kappa coefficient in the assessment and comparison of the accuracy of thematic maps obtained by image classification. Remote Sensing of Environment 239: 111630. https://doi.org/10.1016/j.rse.2019.111630
Frank, B. 2016. Human–wildlife conflicts and the need to include tolerance and coexistence: an introductory comment. Society and Natural Resources 29(6):738-743. https://doi.org/10.1080/08941920.2015.1103388
Frank, B., and J. A. Glikman. 2019. Human–wildlife conflicts and the need to include coexistence. Human–wildlife interactions: Turning conflict into coexistence:1-19. https://doi.org/10.1017/9781108235730.004
Gandiwa, E. 2012. Local knowledge and perceptions of animal population abundances by communities adjacent to the northern Gonarezhou National Park, Zimbabwe. Tropical Conservation Science 5(3):255-269. https://doi.org/10.1177/194008291200500303
Gandiwa, E., P. Gandiwa, and N. Muboko. 2012. Living with wildlife and associated conflicts in a contested area within the Northern Gonarezhou National Park. Zimbabwe. Journal of Sustainable Development in Africa 14(6):252-260. https://www.academia.edu/en/52208619/Living_with_Wildlife_and_Associated_Conflicts_in_a_Contested_Area_Within_the_Northern_Gonarezhou_National_Park_Zimbabwe
Gandiwa, E., I. M. Heitkönig, A. M. Lokhorst, H. H. Prins, and C. Leeuwis. 2013. CAMPFIRE and human–wildlife conflicts in local communities bordering northern Gonarezhou National Park, Zimbabwe. Ecology and Society 18(4): 7. https://doi.org/10.5751/ES-05817-180407
Gholami, H., A. Mohammadifar, D. T. Bui, and A. L. Collins. 2020. Mapping wind erosion hazard with regression-based machine learning algorithms. Scientific Reports 10: 20494. https://doi.org/10.1038/s41598-020-77567-0
Gibson, D. J. 2009. Grasses and grassland ecology. Oxford University Press, Oxford, UK.
Givá, N., and K. Raitio. 2017. “Parks with people” in Mozambique: community dynamic responses to human–elephant conflict at Limpopo National Park. Journal of Southern African Studies 43(6):1199-1214. https://doi.org/10.1080/03057070.2017.1374810
Göttert, T., and N. Starik. 2022. Human–wildlife conflicts across landscapes—general applicability vs. case specificity. Diversity 14(5): 380. https://doi.org/10.3390/d14050380
Khattak, R. H., L. Teng, T. Mehmood, S. Ahmad, F. Bari, E. U. Rehman, and Z. Liu. 2021. Understanding the dynamics of human–wildlife conflicts in north-western Pakistan: implications for sustainable conservation. Sustainability 13(19): 10793. https://doi.org/10.3390/su131910793
Khosravi, R., H. Wan, M. R. Sadeghi, and S. Cushman. 2023. Identifying human–brown bear conflict hotspots for prioritizing critical habitat and corridor conservation in southwestern Iran. Animal Conservation 26(1):31-45. https://doi.org/10.1111/acv.12800
König, H. J., S. Ceaușu, M. Reed, H. Kendall, K. Hemminger, H. Reinke, E. F. Ostermann‐Miyashita, E. Wenz, L. Eufemia, and T. Hermanns. 2021. Integrated framework for stakeholder participation: methods and tools for identifying and addressing human–wildlife conflicts. Conservation Science and Practice 3(3): e399. https://doi.org/10.1111/csp2.399
Le Bel, S., A. Murwira, B. Mukamuri, R. Czudek, R. Taylor, and M. La Grange. 2011. Human wildlife conflicts in southern Africa: riding the whirl wind in Mozambique and in Zimbabwe. Pages 283-322 in J. L. Pujol. The importance of biological interactions in the study of biodiversity, InTech, London, UK.
Leta, M. K., T. A. Demissie, and J. Tränckner. 2021. Modeling and prediction of land use land cover change dynamics based on land change modeler (Lcm) in Nashe Watershed, Upper Blue Nile Basin, Ethiopia. Sustainability 13(7): 3740. https://doi.org/10.3390/su13073740
Long, H., D. Mojo, C. Fu, G. Wang, E. Kanga, A. M. Oduor, and L. Zhang. 2020. Patterns of human–wildlife conflict and management implications in Kenya: a national perspective. Human Dimensions of Wildlife 25(2):121-135. https://doi.org/10.1080/10871209.2019.1695984
Madden, F. 2004. Creating coexistence between humans and wildlife: global perspectives on local efforts to address human–wildlife conflict. Human Dimensions of Wildlife 9(4):247-257. https://doi.org/10.1080/10871200490505675
Makindi, S. M., M. N. Mutinda, N. K. Olekaikai, W. L. Olelebo, and A. A. Aboud. 2014. Human–wildlife conflicts: causes and mitigation measures in Tsavo Conservation Area, Kenya. International Journal of Science and Research 3(6): 1025. https://www.researchgate.net/publication/273031890_Human-Wildlife_Conflicts_Causes_and_Mitigation_Measures_in_Tsavo_Conservation_Area_Kenya
Makumbe, P., S. Mapurazi, S. Jaravani, and I. Matsilele. 2022. Human–wildlife conflict in Save Valley Conservancy: residents’ attitude toward wildlife conservation. Scientifica 2022: 2107711. https://doi.org/10.1155/2022/2107711
Marchini, S., and P. G. Crawshaw, Jr. 2015. Human–wildlife conflicts in Brazil: a fast-growing issue. Human Dimensions of Wildlife 20(4):323-328. https://doi.org/10.1080/10871209.2015.1004145
Matawa, F., A. Murwira, and K. S. Schmidt. 2012. Explaining elephant (Loxodonta africana) and buffalo (Syncerus caffer) spatial distribution in the Zambezi Valley using maximum entropy modelling. Ecological Modelling 242(10):189-197. https://doi.org/10.1016/j.ecolmodel.2012.05.010
Mateo‐Tomás, P., P. P. Olea, I. S. Sánchez‐Barbudo, and R. Mateo. 2012. Alleviating human–wildlife conflicts: identifying the causes and mapping the risk of illegal poisoning of wild fauna. Journal of Applied Ecology 49(2):376-385. https://doi.org/10.1111/j.1365-2664.2012.02119.x
Matseketsa, G., N. Muboko, E. Gandiwa, D. M. Kombora, and G. Chibememe. 2019. An assessment of human–wildlife conflicts in local communities bordering the western part of Save Valley Conservancy, Zimbabwe. Global Ecology and Conservation 20: e00737. https://doi.org/10.1016/j.gecco.2019.e00737
McNutt, J. W., A. B. Stein, L. B. McNutt, and N. R. Jordan. 2017. Living on the edge: characteristics of human–wildlife conflict in a traditional livestock community in Botswana. Wildlife Research 44(6-7):546-557. https://doi.org/10.1071/WR16160
Messmer, T. A. 2009. Human–wildlife conflicts: emerging challenges and opportunities. Human–Wildlife Conflicts 3:10-17. https://www.researchgate.net/publication/268414579_Human-wildlife_conflicts_Emerging_challenges_and_opportunities
Mhuriro-Mashapa, P., E. Mwakiwa, and C. Mashapa. 2018. Socio-economic impact of human–wildlife conflicts on agriculture based livelihood in the periphery of save valley conservancy, southern Zimbabwe. The Journal of Plant and Animal Sciences 28(3):12-16. https://www.researchgate.net/publication/323858030_Socio-economic_impact_of_human-wildlife_conflicts_on_agriculture_based_livelihood_in_the_periphery_of_save_valley_conservancy_southern_Zimbabwe
Mohammadi, A., K. Almasieh, H. Y. Wan, D. Nayeri, A. Alambeigi, J. I. Ransom, and S. A. Cushman. 2021. Integrating spatial analysis and questionnaire survey to better understand human–onager conflict in Southern Iran. Scientific Reports 11: 12423. https://doi.org/10.1038/s41598-021-91921-w
Mombeshora, S., and S. Le Bel. 2009. Parks–people conflicts: the case of Gonarezhou National Park and the Chitsa community in south-east Zimbabwe. Biodiversity and Conservation 18:2601-2623. https://doi.org/10.1007/s10531-009-9676-5
Muboko, N., and G. J. Bradshaw. 2018. Towards resolving local community and protected area management conflicts: lessons from the Chitsa community and Gonarezhou National Park, Zimbabwe. International Journal of Development Conflict 8:62-79. http://www.ijdc.org.in/uploads/1/7/5/7/17570463/muboko_2018.pdf
Mugandani, R., M. Wuta, A. Makarau, and B. Chipindu. 2012. Re-classification of agro-ecological regions of Zimbabwe in conformity with climate variability and change. African Crop Science Journal 20 (S2):361-369. https://www.ajol.info/index.php/acsj/article/view/81761
Mukeka, J. M., J. O. Ogutu, E. Kanga, and E. Roskaft. 2018. Characteristics of human–wildlife conflicts in Kenya: examples of Tsavo and Maasai Mara Regions. Environment and Natural Resources Research 8(3):148. https://doi.org/10.5539/enrr.v8n3p148
Musakwa, W., T. Gumbo, G. Paradza, E. Mpofu, N. A. Nyathi, and N. B. Selamolela. 2020. Partnerships and stakeholder participation in the management of national parks: experiences of the Gonarezhou National Park in Zimbabwe. Land 9(11): 399. https://doi.org/10.3390/land9110399
Mutanga, C. N., N. Muboko, and E. Gandiwa. 2017. Protected area staff and local community viewpoints: a qualitative assessment of conservation relationships in Zimbabwe. PLoS One 12(9): e0177153. https://doi.org/10.1371/journal.pone.0177153
Nayeri, D., A. Mohammadi, L. Hysen, D. Hipólito, D. Huber, and H. Y. Wan. 2022. Identifying human-caused mortality hotspots to inform human–wildlife conflict mitigation. Global Ecology and Conservation 38: e02241. https://doi.org/10.1016/j.gecco.2022.e02241
Osipova, L., M. M. Okello, S. J. Njumbi, S. Ngene, D. Western, M. W. Hayward, and N. Balkenhol. 2018. Fencing solves human–wildlife conflict locally but shifts problems elsewhere: a case study using functional connectivity modelling of the African elephant. Journal of Applied Ecology 55(6):2673-2684. https://doi.org/10.1111/1365-2664.13246
Padalia, H., V. Srivastava, and S. Kushwaha. 2014. Modeling potential invasion range of alien invasive species, Hyptis suaveolens (L.) Poit. in India: comparison of MaxEnt and GARP. Ecological Informatics 22:36-43. https://doi.org/10.1016/j.ecoinf.2014.04.002
Phillips, S. J. 2005. A brief tutorial on Maxent. AT&T Research 190:231-259. https://biodiversityinformatics.amnh.org/open_source/maxent/Maxent_tutorial2017.pdf
Phillips, S. J., R. P. Anderson, M. Dudík, R. E. Schapire, and M. E. Blair. 2017. Opening the black box: an open‐source release of Maxent. Ecography 40(7):887-893. https://doi.org/10.1111/ecog.03049
Phillips, S. J., R. P. Anderson, and R. E. Schapire. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling 190(3-4):231-259. https://doi.org/10.1016/j.ecolmodel.2005.03.026
Phillips, S. J., and M. Dudík. 2008. Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31(2):161-175. https://doi.org/10.1111/j.0906-7590.2008.5203.x
Pisa, L. S., and S. Katsande. 2021. Human wildlife conflict in relation to human security in the Gonarezhou National Park, Zimbabwe. International Journal of Earth Sciences Knowledge and Applications 3(2):98-106. https://dergipark.org.tr/tr/download/article-file/1636733
Redpath, S. M., S. Bhatia, and J. Young. 2015. Tilting at wildlife: reconsidering human–wildlife conflict. Oryx 49(2):222-225. https://doi.org/10.1017/S0030605314000799
Sharma, P., N. Chettri, K. Uddin, K. Wangchuk, R. Joshi, T. Tandin, A. Pandey, K. S. Gaira, K. Basnet, and S. Wangdi. 2020. Mapping human–wildlife conflict hotspots in a transboundary landscape, Eastern Himalaya. Global Ecology and Conservation 24: e01284. https://doi.org/10.1016/j.gecco.2020.e01284
Songhurst, A. 2017. Measuring human–wildlife conflicts: comparing insights from different monitoring approaches. Wildlife Society Bulletin 41(2):351-361. https://doi.org/10.1002/wsb.773
Tassicker, A., A. Kutt, E. Vanderduys, and S. Mangru. 2006. The effects of vegetation structure on the birds in a tropical savanna woodland in north-eastern Australia. The Rangeland Journal 28:139-152. https://doi.org/10.1071/RJ05029
Tavuyanago, B. 2016. Living on the fringes of a protected area: Gonarezhou National Park (GNP) and the indigenous communities of South East Zimbabwe 1934–2008. Dissertation, University of Pretoria, Pretoria, South Africa.
Teixeira, L., K. C. Tisovec‐Dufner, G. d. L. Marin, S. Marchini, I. Dorresteijn, and R. Pardini. 2021. Linking human and ecological components to understand human–wildlife conflicts across landscapes and species. Conservation Biology 35(1):285-296. https://doi.org/10.1111/cobi.13537
Tobeña, M., R. Prieto, M. Machete, and M. A. Silva. 2016. Modeling the potential distribution and richness of cetaceans in the Azores from fisheries observer program data. Frontiers in Marine Science 3: 202. https://doi.org/10.3389/fmars.2016.00202
van Bommel, J. K., M. Badry, A. T. Ford, T. Golumbia, and A. C. Burton. 2020. Predicting human–carnivore conflict at the urban–wildland interface. Global Ecology and Conservation 24: e01322. https://doi.org/10.1016/j.gecco.2020.e01322
Verma, P., A. Raghubanshi, P. K. Srivastava, and A. Raghubanshi. 2020. Appraisal of kappa-based metrics and disagreement indices of accuracy assessment for parametric and nonparametric techniques used in LULC classification and change detection. Modeling Earth Systems and Environment 6:1045-1059. https://doi.org/10.1007/s40808-020-00740-x
Vivekananda, G., R. Swathi, and A. Sujith. 2021. Multi-temporal image analysis for LULC classification and change detection. European Journal of Remote Sensing 54(S2):189-199. https://doi.org/10.1080/22797254.2020.1771215
Zvidzai, M., A. Murwira, A. Caron, and D. Garine-Wichatitsky. 2013. Waterhole use patterns at the wildlife/livestock interface in a semi-arid savanna of Southern Africa. International Journal of Development and Sustainability 2:455-471. https://isdsnet.com/ijds-v2n2-1.pdf