Abstract:
Taking the average water content of atmosphere derived from MODIS 3 bands as a parameter, the splitwindow algorithm was employed on Landsat 8 to retrieve the land surface temperature (LST) of the central urban area of Kunming City. Then, the landscape compositions were identified by objectedbased hierarchical classification. After selecting controllablefeatures of landscape composition, stepwise regression fitting was applied to analyze the relationship between LST and these features. The research showed the central parts of Kunming tend to have either high LST or low LST, and the major reason for numerous urban heat islands of Kunming is due to the uneven distribution of vegetation coverage and the high percentage of impervious surface and roads. Regression models indicated that the coverage percentage of vegetation, the shape index of vegetation patches, the size of pool, the length of river and the green rate in impervious surface have negative correlations with LST, while the size and perimeter of impervious surface have positive correlations with LST. It was concluded that there are some weak correlations between LST and some features of landscape compositions. Therefore, further research will be meaningful for the control of city thermal environment and the city planning.