Detection of land cover and thermal environment change in Beijing from TM images

Abstract

This paper focuses on the detection of land cover and thermal environmental change over 20 years in Beijing using TM data collected on 17 June, 1991, and 8 June, 2011, respectively. For observing the urban expansion and thermal environment change in the whole Beijing governmental region, the central urban area and satellite towns were firstly recognized and detected, and then the central urban areas was divided into 5 areas based on the ring roads. After image mosaicking, masking and clipping, sample data of land cover types as water bodies, green land, bare land, construction land, farm land, and forest were collected and trained. Classification was performed using object-based support vector machine algorithm. Accuracy assessment of the classification results was conducted according to ground truth of regions of interest. The conversion matrix and band math were used to detect the land cover and thermal environment change respectively from 1991 to 2011. The results showed that, over the last 20 years, the central urban size was 5 times larger than before, construction land increased, and farm land and water bodies decreased, especially for areas outside the third ring road. Besides, UHI intensity in nearly 60% of the areas within the sixth ring road increased by more than 4 degree, and only 2% of the areas decreased. Both are consistent with the real status of the rapid development of economy and urbanization in Beijing.