GIS and remote sensing applications for modelling the distribution of elephants and their interaction with vegetation

Abstract

Knowledge of elephant (Loxodonta africana) interaction with vegetation is critical for conservation of the mega-herbivore and of other wildlife species found in the ecosystem. Although the impact of elephants on vegetation structure has been investigated before, location and time specific knowledge on changes in the landscape has remained largely inconclusive. This is because most of the early studies largely depended on plot-based observations that are limited in scope both spatially and temporally. This thesis develops and applies GIS and remote sensing methods aimed at understanding the spatial pattern of elephant-vegetation interaction in a predominantly savannah landscape. Specific objectives of the study were to: (1) understand the predictive ability of elephant distribution models developed using presence data collected from GPS collars and compare them to those developed from aerial survey data; (2) develop and test new methods for correcting locational error in aerial survey data for improving models of elephant distribution; (3) test whether elephant presence peaks farther from water points in addition to the known peak near water; (4) investigate whether elephants selectively utilise a heterogeneous landscape; and (5) test whether and how the rate of change in vegetation structure differs across a heterogeneous landscape. Firstly, results of the study show that elephant presence models built from GPS collar data utperformed those built from aerial survey data. Secondly, a new method suggested for correcting error in aerial survey data shifted location by 143 to 177m from the line of flight. In addition, the models of elephant presence built from the corrected dataset had better predictive ability than those built from uncorrected data. Thirdly, elephant presence peaked at places located farther from water sources in addition to the known peak near water. The peaks occurred in areas of high vegetation cover. Fourthly, elephant speed of movement and utilisation of the landscape (i.e., speed, Linear Time Density and the Kernel Density Estimator) differed by vegetation/cover type. Finally, the rate of tree cover change differed by vegetation/cover type. The change was also observed to be correlated with elephant movement and utilisation of the landscape. Results of the thesis thus suggest that GIS and Remote sensing-based methods improve our understanding of elephant-vegetation dynamics in space and time. These findings underscore the utility of GIS and remote sensing in studies that investigate the spatial pattern of elephant interaction with vegetation. Knowledge of those patterns could be applied in the formulation of strategies aimed at conserving the African elephant as well as other wildlife species that co-occur with the megaherbivore.