Abstract: Based on total solar radiation observation data for 1961-2015, we used the Thin Plate Spline (TPS) method to obtain a surface fitting model that is suitable for the distribution of solar radiation in China. Based on the grid data, the error distribution and seasonal variation of radiation grid data are comprehensively analyzed by means of average deviation, average absolute error and root mean square error (RMSE). The results show that: (1) the interpolation model got a good effect in the radiation field for 1961-2015, the square root of the total radiation value of the generalized cross validation (RTGCV) mainly in 1.95～6.24 MJ·m^-2·d^-1, (2) the interpolation precision is relatively high. The averaged errors of interpolated values for total radiation from that of observations are mainly concentrated in the ±0.5 MJ·m^-2·d^-1 for 1961-2015, the averaged deviation of the interpolated lattice values from that of the acture values is 0.0009 MJ·m^-2·d^-1, the RMS error is 0.9710 MJ·m^-2·d^-1. The variation of error statistics with seasons is obvious. The seasonal variation of the RTGCV and RMSE are obvious, the biggest one is in summer, the second in spring and autumn, and the smallest in winter.