Using the daily maximum temperature and minimum temperature data from 550 statio ns for the period 1960-2000, the heat wave days index (HWDI) and the warm night index (Tn90) are calculated and their temporal and spatial variation are analyze d. The first two modes are able to represent the spatial distribution characteri stics. The spatial distribution of first eigenvector of the two indexes is gener ally similar, with the same increase and decrease trend. Their interannual and interdecadal variability in temporal characteristic is remarkable. The change s of heat index and warm night index can be divided into 11 and 10 districts by using REOF and CAST Clustering Analysis Method, which can overcome the subjectivity of the boun dary of the load values by REOF and the uncertainty of climate center by CAST an d make the division more objective.