Abstract: As a typical rapid urbanization city, the thermal environment in Dongguan, Guangdong has changed much in the past 28 years because of the economy development, the population growth, the underlying surface change, and the greenhouse effect enhancement. Taking Dongguan as an example, the paper retrieves the land surface temperature using Mono-window Algorithm proposed by Qin Zhihao from Landsat/TM data, meteorological measurement data and basic geographical data. Based on the Vegetation-Impervious surface-Soil model which was proposed by Ridd in 1995 according to the basic elements of underlying surfaces in a city, it defines the typical underlying surfaces as water, high vegetation region and dense building region,and gets the distribution information of them from the threshold division of normalized difference vegetation index. It takes an overlay analysis between the typical underlying surface information and land surface temperature data. The result shows that this method can effectively get the temperature differences of typical underlying surfaces. From 1988 to 2005, the land surface temperature of dense building region is higher than that of vegetation and water in Dongguan. The distribution of high land surface temperature region is consistent with the distribution of dense building region, and the low is consistent with the vegetation and water. Under the background of global warming caused by enhanced greenhouse effect, the temperature difference of dense building region and water, as the same as the temperature difference of dense building region and high vegetation region, are apparently increased. The former is bigger than the latter.