Multitemporal LANDSAT Imagery Analysis to Study the Dynamics of Land Cover over Lake Kivu Region

In this study, maximum likelihood supervised classification and post-classification change detection techniques were applied to cloud-free Landsat mosaic scenes formed for three years, 1987, 2001 and 2010, to map land cover changes in the Lake Kivu region in Central Africa. A supervised classification was carried out on the stacked vector of six reflective bands and two vegetation index images for the three years individually with the aid of ground data. Since ground data was not available for 1987, visual interpretation was used to aid supervised classification. The overall seven-class classification accuracies averaged to 77% for the three years. Post-classification change detection technique was used to evaluate land cover change statistics in the region using cross-tabulation. Changes among different land cover classes were accessed. A very severe land cover change has taken place in the region over the past 25 years as a result of human migration and agriculture practices. The major land cover changes in the area include rapid deforestation, urbanization and cropland expansion.