[Comparison of spatial interpolation methods for daily meteorological elements]

A comparative study was made to evaluate the methods of inverse distance weighting (IDW), co-kriging (CK), and thin plate spline (TPS) in interpolating the average meteorological elements (including maximum air temperature, minimum air temperature, sunshine hours, and precipitation) of the 15th day per month from the 1951-2005 comprehensive observation data of 559 meteorological stations in China. The results showed that the RMSEs for the maximum and minimum air temperature in a year interpolated by TPS were the smallest (1.02 degrees C and 1.12 degrees C, respectively), and the R2 between the observed and predicted values were the highest (0.9916 and 0.9913, respectively), compared with those interpolated by IDW and CK. In four seasons, the smallest RMSEs for the maximum and minimum air temperature interpolated by TPS were observed in autumn (0.83 degrees C) and summer (0.86 degrees C), respectively, and the R2 between the observed and predicted values interpolated by TPS were higher in autumn than in other seasons. The RMSEs for the sunshine hours and precipitation in a year interpolated by TPS were the smallest (0.59 h and 1.01 mm, respectively), and the R2 between the observed and predicted values were the highest (0.9118 and 0.8135, respectively), compared with those interpolated by IDW and CK. In four seasons, the RMSE for the sunshine hours in winter interpolated by TPS was the smallest (0.49 h), and the R2 between the observed and predicted sunshine hours was the smallest (0.9293). The RMSE for the precipitation in winter interpolated by TPS was the smallest (0.33 mm), while the RMSE for the precipitation in summer interpolated by IDW was the smallest (2.01 mm). The R2 between the observed and predicted precipitation in winter interpolated by CK was the highest (0.8781). It was suggested that TPS could be the optimal spatial interpolation method in interpolating and rasterizing the daily meteorological elements in China.