The interpolation of air temperature is one of the most important methods in generating high spatial resolution data for un
known points. However, many traditional interpolation methods without geographical information considered were mostly mathematic
ways. The paper, based on digital elevation model (DEM) and the monthlymean air temperature collected from 82 stationswhich loca
ted in and around the eastern regions of Gansu, p roposed a multip le linear regression spatial interpolation (MLR) method via analyzing
the relationship s ofmean air temperature with latitude, longitude, altitude, slope and aspect, and meanwhile carried out the compari
son ofMLR with conventionalmethods such as Inverse Distance Squared method ( IDS) , Sp line method ( SPL INE) and Ordinary Krig
ing method (OK) , which showed that the p recision ofMLR method taking into account terrain factorswas better than that of other con
ventionalmethods. Finally, using theMLR method, a data set ofmonthlymean air temperature with 84 m ×84 m spatial resolution in
the eastern regions of Gansu was obtained, which exhibited that the spatial distribution ofmonthlymean air temperature in the eastern
regions of Gansu, p resented the zonal rule which decreased from the southeast to the northwest, and the extent and trend of fluctuation
with mean air temperature in season internally was different, that is, the lowest undulation extentwas in summer, then in winter, and
the undulation range in sp ring and fallwas the biggest, and the trend of fluctuation p resented the latitudinal zonal from the west to east
in summer, the north to south in winner, the southwest to northeast in sp ring and fall.
