In order to compensate for the limited number of meteorological stations in Shaanxi Province and the uneven distribution, this paper makes a comparative study on the common spatial interpolation methods to explore the suitable spatial interpolation method of air temperature for Shaanxi Province. According to the monthly average temperature of 110 meteorological stations in Shaanxi Province from 1981 to 2010, the inverse distance weight (IDW), ordinary Kriging (OK) and multiple linear regression (MLR) were used to perform the null interpolation, and the accuracy was evaluated by the actual verification method. The results show that the interpolation accuracy of the MLR, which takes into account the longitude, latitude and elevation, is the highest, the accuracy of the OK ranks the second, and the IDW has the lowest precision. The MLR is most suitable for the temperature interpolation in Shaanxi Province.