Regional land cover mapping and change detection in Central Asia using MODIS time-series

Abstract

In Central Asia major alterations in land use and land cover occurred in the past decades due to political collapse of the Soviet Union, human forces, and climate change. In this context accurate land cover information for the region of Central Asia is important. In this study we present a classification approach with implemented C5.0 algorithm addressing regional land cover characteristics of Central Asia. The classification is performed on seasonal features derived from MODIS time-series for the years 2001 and 2009, which allows us to analyse possible land cover and land use changes. Training and validation are based on a reference dataset collected from high resolution remote sensing data. The overall accuracy of both classifications is above 90%. The results show some significant changes between both years in different land cover classes. Human induced alterations of water bodies, variability in sparsely vegetated areas due to seasonal precipitation and forest loss caused by forest fires and logging are exemplarily depicted and discussed in this study.