A thesis submitted to the Faculty of Science, University of the Witwatersrand, Johannesburg in fulfillment of the requirements for the degree of Masters in Science. / Satellite imagery and remote sensing techniques provide a way to collect data over large temporal and spatial scales, and build indices that gauge ecological variables. These indices can explain the distribution of animals in their environment. In this study I compare the ability of various indices derived from Landsat 8, to reliably predict locations of different large herbivore species across diverse habitats. The study was undertaken in the Kgaswane Mountain Reserve, North West Province, South Africa. Daily locations of two herds of sable antelope (Hippotragus niger) and one herd of eland (Tragelaphus oryx) were used. One sable antelope herd (vlei herd) occupied a grassland plateau with a wetland and the other sable antelope herd (woodland herd), shared the wooded area at the base of the mountains with the eland herd. I described vegetation communities, burnt areas, geology and soil templates at animal locations during foraging bouts in the dry season; coinciding with the times of the Landsat images. The overall aim of this study was to see whether an index or a combination of indices could better describe animal locations than the normally used NDVI. I calculated a number of indices, and compared their predictive ability to define areas used by the study animals. Specifically, I compared the Normalised Difference Vegetation Index (NDVI) to Soil Adjusted Vegetation Index (SAVI), Visible Atmospherically Resistant Index (VARIgreen), Green Atmospherically Resistant Index (GARI), Normalised Difference Water Index (NDWI), a proxy for soil moisture; and mineral composite indices assessing clay minerals, ferrous minerals and iron oxide. I chose these indices as they describe the basic characteristics of an ecologically functioning unit. The locations of one of the sable antelope herds, located in grassland areas underlined by quartzite, were best described by NDVI, SAVI and VARIgreen. The locations of the other sable antelope herd, occurring in an open wooded area with shallow sandy soils on norite and quartzite, were best described by clay minerals and GARI. Eland locations, found in woodland areas characterised by deep norite soils, were best described by a combination of iron oxide, NDVI and SAVI. Therefore, NDVI proved to be an adequate indicator in open grassland areas, where it could be interchanged with SAVI, and improved by VARIgreen. In closed woodlands NDVI, SAVI and NDWI could all be used to describe browser locations. NDVI was not a suitable index when it came to describe locations of a grazer in a woodland/grassland matrix. However, it is important to keep in mind that my results pertain only to one dry season and two herbivore species, and therefore further studies would be needed to be able to generalise the results further. / MT2016

Fire plays an important role in structuring and maintaining savanna grassland ecosystems. Although regular fires are a characteristic feature of savannas, the effects of fire frequency on these systems are less well known, particularly with respect to how frequency of fire influences large herbivore distribution and behaviour. The expectation is that large herbivores should be attracted to frequently burned sites as a consequence of changes in forage quality and quantity, and/or vegetation structure and composition. The former could be driven by alterations in soil nutrients, such as N and P. Alterations in vegetation also could be important in determining risk of predation. For example, an increase in woody vegetation could decrease predator visibility making large herbivores more vulnerable to predation. The objectives of this study were to investigate the effects of long-term alterations in fire frequency on herbivore distribution and behaviour, as well as the mechanisms (soil nutrients, vegetation structure and composition, and forage quality and quantity) potentially driving the distribution of large herbivores. To address these objectives, I conducted large herbivore surveys on a bi-weekly basis from 2009-2010 in a series of plots in the Experimental Burn Plots (EBPs) burnt at different frequencies (annual, triennial and unburnt) over the last five decades at three study sites in the Kruger National Park, South Africa. Surveys also were conducted on new plots that were established adjacent to the long-term plots. These new plots have a fire return interval of 4 years which is similar to the triennially burned plots of the EBPs. They were established in the landscape adjacent to the EBPs to assess whether the responses of herbivores to fire observed in the EBPs reflected was at landscape level. The distribution of all large herbivore species combined and of grazers (e.g. zebra) or browsers (e.g. kudu) only were not affected by fire frequency. In contrast, the abundance of mixed-feeders, such as impala, was significantly higher in the unburnt (control) and annually burned plots than the triennially burned plots. Although season did not have a significant impact on the distribution of browsers and mixed-feeders, overall more grazers were recorded across all burn treatments in the dry season compared to the wet season. Similar patterns of herbivore distribution were observed between the new plots and the triennially burned EBP plots, suggesting that responses observed to the long-term fire frequency treatments reflects herbivore responses at the landscape level. The long-term fire frequency treatments significantly affected soil nutrients (N, organic C, P, and K were significantly lower with annual burning), vegetation structure (abundance of woody plants were greater in unburned plots), and forage quantity (unburned plots had higher biomass) but not quality. More frequent fires improved visibility by reducing tree height and density and herbaceous biomass, thereby potentially reducing predation risk, when compared to less frequent burning. As a result, herbivores selected sites with more frequent fires. The behaviour of the herbivore species investigated was predominantly influenced by seasonal-induced changes to their environment rather than fire frequency. In the wet season irrespective of the burning treatment visibility was low due to high rainfall that increases plant biomass, whereas in the dry season visibility was improved because there is little to no rainfall. This potential alteration in predation risk likely resulted in herbivores being more vigilant in the wet season than the dry season. Overall, results from this study suggest that the combination of fire frequency and season drive herbivore distribution and behaviour by altering mainly the vegetation structure which can influence predation risk. / Thesis (M.Sc.)-University of KwaZulu-Natal, Pietermaritzburg, 2012.