The underlying surface of more than 700 national meteorological stations in different historical periods is distinguished based on the MSS (MultiSpectral Scanner) images of the 1970s, TM (Thematic Mapper) images around 1990, and ETM (Enhanced Thematic Mapper) images around 2000 and 2005. The stations with apparent changes in their underlying surface and surrounding building environment are picked out and verified by telephone consultation. Based on the discriminant result, annual and seasonal change characteristics of air temperature are analyzed, and it is concluded that the change rate at air temperature of the stations “entering city” is higher than that at the other stations. The result validates the discriminant method based on RS (Remote Sensing) images, and conforms that the observed air temperature at the stations “entering city” increase more than the other stations affected by the urban circumstance.