Spatial data of soil temperature is one of the basic datasets of many large scale researches, such as the International Geosphere-Biosphere Program and DIVERSITAS. Thus, it is really meaningful to derive spatial continuous data of soil temperature from discrete data using the spatial estimation method. Based on the data of mean annual soil and air temperatures in the period from 1971 to 2000 collected from 698 meteorological stations in China and the data of the digital elevation model of China, analysis was conducted of effects of meteorological and topographic factors on mean annual soil temperature. In light of different sources of data, three groups of affecting factors, that is, 1) mean annual air temperature; 2) longitude, latitude and elevation; and 3) mean annual air temperature, longitude, latitude and elevation; were designed as auxiliary variable, and the regression kriging method was adopted to predict the spatial patterns of mean annual soil temperatures across the country. Results showed that mean annual air temperature, longitude, latitude and elevation all displayed significant impacts on spatial variation of mean annual soil temperatures. Validation of the results revealed that prediction with auxiliary variable of group 2 was in the lead in accuracy, and followed by that with Group 3 and then with Group 1. Data accuracy of the auxiliary variables and their relativity with mean annual soil temperature would significantly affect precision of the prediction.