The Kriging method is used to conduct the spatial interpolation of surface air temperature dataset over China for each month from January 1951 to December 2004,and comparisons were made between interpolated and observed series.The results show that high correlations were detected between interpolated and original series,which proves that the interpolation of large scale historic climatic data is acceptable to some extent;and in interpolation,departure values in stead of origional values were recommended and 10 as the radius of the spherical model,which will improve the resultant series greatly,especially in station-sparse areas.While in some regions with particular local climate changes,it deserves more careful downscaling analysis and denser stations distribution.