Hybrid classification of Landsat data and GIS for land use/cover change analysis of the Bindura district, Zimbabwe

Abstract

Rapid land use/cover changes have taken place in many parts of Zimbabwe, including the Bindura district, mainly due to unequal land distribution during the colonial period, demographic pressure, agricultural expansion, government policies and environmental factors such as drought. This study monitored and analysed land use/cover changes in the Bindura district using Landsat data for 1973, 1989 and 2000 and discussed socioeconomic backgrounds for changes. A hybrid supervised/unsupervised classification approach coupled with geographical information systems (GIS) analyses was employed to generate land use/cover maps with six classes; agriculture, woodland, mixed rangeland, settlement, bare land and water. A post‐classification comparison change detection technique revealed different trends in land use/cover changes over the two periods (1973–1989 and 1989–2000). During the colonial and pre‐economic structural adjustment program (ESAP) independent period (1973–1989), the rates of agriculture and bare land changes were greater in the communal areas (CAs) than in the large‐scale commercial farms (LSCF), with ‘woodland to agriculture’ and ‘mixed rangeland to agriculture’ conversions dominating. In contrast, land use/cover changes in the LSCF increased more than the CAs during the post‐ESAP independent period (1989–2000), as ‘woodland to agriculture’, ‘mixed rangeland to agriculture’ and ‘woodland to mixed rangeland’ conversions dominated. The major land use/cover changes in the study area as derived from Landsat data are mainly due to the dynamic nature of agriculture practice such as extension of crop fields on one hand and the abandonment of croplands on the other hand, with additional factors such as wildfires and droughts. Observed trends in land use/cover changes indicate that deforestation and the encroachment of cultivation in woodland areas is a continuous trend in all the land tenure systems. Our results, therefore, suggest that an efficient and sustainable land use plan is required to reduce the rapid land use/cover changes, particularly the loss of woodland areas.

Acknowledgements

We thank Mr Dominic Kwesha of the Forest Commission, Zimbabwe and the Tropical Rain Forest Information Center (TRFIC) through the Michigan State University and the Center for Global Change for making available some of the Landsat datasets used in this study. The valuable comments and suggestions from two anonymous reviewers helped to improve the manuscript.