Assessing the extent of and changes in the wildlife sector in Limpopo province, South Africa

Abstract

The wildlife sector has grown rapidly over the past few decades and is considered a valuable asset for South African ecotourism, economy and conservation. However, there has been an increasing concern around its conservation efficacy, particularly with the industry becoming more intensive regarding its animal production. The growth of the wildlife sector, especially intensive breeding practices, has proliferated the use of fencing. Fences establish boundaries and protect wildlife, but may also cause mortality, inhibit animal movement, and can ultimately lead to landscape fragmentation which has been shown to have adverse effects on wildlife and the environment. To infer spatial changes in the wildlife sector across a ten year time frame, I used remote sensing procedures to manually map and quantify the changes in fences and camps (fenced areas) of wildlife properties based on satellite images of south-west Limpopo during 2007, 2012 and 2017. Results show an increase in intensive wildlife properties, total length of fences, and total number of camps from 2007 to 2017. The mean area of camps decreased over the ten year time period, accompanied by an overall increase in the number of smaller camps (≤200 ha) and a general decrease in larger camps (≥500 ha). Furthermore, the areas covered by smaller camps (≤200 ha) increased whilst the areas covered by larger camps decreased (≥500 ha) over the entire time period. The biggest changes in the wildlife sector occurred between 2012 and 2017, which suggest that the changes may be occurring progressively more and should therefore be urgently addressed. As fence maps would be very beneficial to wildlife researchers and managers, I pursued an alternative method to ‘automate’ the mapping of fences through image classification. Two image classification methods were used, namely Support Vector Machine (SVM) and Random Forest (RF), to classify the satellite images of the wildlife sector in south-west Limpopo. The fence area obtained from the classified images did not however correspond with the manual fence map, due to the high variability in accuracy values, specifically overall accuracy and kappa index. The SVM and RF methods were statistically identical in accuracy values. Furthermore, it was found that some landscape characteristics, such as percentage elevation and presence of water, correlated with the overall accuracy of certain classified images. Therefore, image classification methods have the potential to map fences of the wildlife sector, and needs to be improved for future use. The extent of increase in intensive wildlife production and the rise of fences are disconcerting trends that may have detrimental consequences to wildlife and their environment. It is vital to increase research efforts to assess the extent and effects of fencing, and inform landowners of fence impacts in South Africa so as to mitigate the ecological effects of fencing. Remote sensing and image classification methods can be used to map the full extent of fences in the wildlife sector. Ultimately, the reduction and regulation of intensive wildlife management practices and fencing may significantly aid in conserving South African wildlife.