Spatiotemporal patterns and drivers of landscape change in a semi-arid, southern African savanna: a satellite imaging and intensity analysis

Abstract

Background: The savannas of southern Africa are a highly variable and globally important biome that supports rapidly expanding human populations. They also have one of the greatest concentrations of wildlife on the continent. Savannas occupy a fifth of the earth's land surface, yet despite their ecological and economic significance, little is understood about the complex couplings and feedbacks that drive long-term spatiotemporal patterns of change. Here, we aimed to describe land-cover change in Chobe District, Botswana, and address questions about the complex drivers of patterns of land cover change, with a particular focus on climate, fire, herbivory, and anthropogenic disturbance. Methods: We used Landsat Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+), and Operational Land Imager (OLI) satellite imagery from 1990 to 2013 to explore land-cover change across the 21 000 km2 Chobe District. Net and gross land cover change were assessed for the Chobe National Park, six protected forest reserves, areas surrounding towns and villages, and the riparian corridor of the Chobe River. We also used intensity analysis methods developed by Aldwaik and Pontius to compare observed changes to a hypothetical uniform change to examine how the size and intensity of land cover losses and gains vary between time periods and across different land-cover classes. Findings: Patterns of land-cover change in the Chobe District were not strictly linear in nature, a characteristic noted for other dryland systems, with gross land-cover change and intensity analysis showing extensive, active bidirectional exchange between woodland, shrubland, and grassland. Changes in land cover over the study period were characterised by a loss of woodlands, accompanied by an expansion of shrubland. Woodland losses were greatest along the Chobe River, which provides a large proportion of the drinking water, and other essential resources, for human and wildlife populations. Interpretation: The multi-decadal monitoring of land-cover changes in southern Africa using remote-sensing provides essential information for societies to manage ecosystem resources locally, so that they might increase resilience to future climate change. Our identification of spatial and temporal patterns of land-cover changes supports the need to develop an early warning framework to identify existing and emergent threats to forest resources and services, and reduce potential adverse effects on the livelihoods and health of local communities. Funding: Forest Conservation Botswana (FCB); National Science Foundation.