Spatial interpolation of temperature in the United States using residual kriging

• We developed two statistical models for estimating air temperature in the U.S.
• Our models have better performances than the kriging function of ArcGIS.
• The spatial variability of temperature is sensitive to season.

Temperature is one of the most important factors influencing every aspect of life. In response to the increasing greenhouse effect in recent years, the demand for understanding the spatial variability of temperature in the U.S. has risen dramatically. To meet this need, we developed a statistical model for constructing a gridded temperature dataset over the mainland United States. Based on the data collected from 922 meteorological stations in the U.S., temperatures at over 5000 unknown locations were predicted in January and July, 2010. This study utilized variables of latitude and longitude (model 1), and latitude, longitude and elevation (model 2) as inputs in a residual kriging method to interpolate the average monthly temperature. We also estimated temperatures at the same locations with the kriging function of ArcGIS and compared the performances of our models with that of ArcGIS. We found that, by adding an elevation factor, our model (model 2) had a better predicting performance than that of ArcGIS kriging function in both January and July. However, only estimation in July was not different from the observation. This suggests that our kriging model is capable of capturing the spatial variability of temperature, but it is sensitive to season. The successful interpolation of July temperature indicates that the accuracy of interpolation can be improved by adding appropriate variables. Seasonal models developed in future research can be valuable tools for meteorological and climatological research.