Abstract:

Coal fire disaster would have a serious negative impact on the natural landscape, ecology and the environment. Land cover change could evaluate the damages to the surrounding environment caused by coal fire disaster. Accurate results from land cover classification is the basis of ecological changes in coalfield fire area. This paper studied the method of land cover classification of Wuda coalfield using Landsat8 data. According to the landform features, surface features and topography features, Wuda coalfield was divided into five sub-regions. Based on the actual situation of different areas, different decision trees were constructed by the NDVI, NDBI, NDWI, altitude, slope, and spectral characteristics. The results showed that the classification method of sub-region decision tress leads to a high overall accuracy up to 87.63% and Kappa Coefficient was 0.86. Compared with the traditional decision tree classification, overall accuracy of our method was increased by 14.75%, and Kappa coefficient was increased by 0.17.

Key words: Wuda coalfield, subregion, decision tree, land cover classification