Land cover change detection based on MODIS 250m vegetation index time series data

Abstract

Land cover is a critical variable that links many aspects of the physical and human environments. In the present study, the post-classification comparison method is employed for detecting land use/land cover changes (LUCC) of Dongjiang River basin during the period of 2001 - 2008 mainly based on MODIS 250 m Enhanced Vegetation Index (EVI) time series data, and in addition MODIS land surface temperature data and digital elevation data. The Hybrid ISODATA-DT classifier for land use/land cover (LULC) classification is adopted for classifying LULC for each year, which uses the ISODATA algorithm to carry out unsupervised classification and decision trees (DT) to refine the classification obtained with ISODATA. Based on the classification result for each year, LUCC of the Dongjiang River Basin for the period of 2001 to 2008 are assessed. It is found that, the areas of forest and shrubland exhibit about 3-4 year cycles and the total areas changed for forest and shrubland are greatest among all the six land cover types in Dongjiang basin; the area of grassland is most changeable; the area of cropland and waters decreased, whereas the area of urban/built-up area significantly increased.