Based on temperature and precipitation data from 68 national meteorological stations in Zhejiang Province from 1988 to 2017, mean summer temperature and precipitation were interpolated using thin-plate smoothing splines （ANUSPLIN), inverse distance weighting (IDW) and ordinary kriging interpolation methods. In order to explore the optimal interpolation method of meteorological elements under complex terrain and climate in Zhejiang Province, the cross-validation method was applied to evaluate the accuracy of three methods, and the spatial error was analyzed. The results are as follows: (1) The interpolation accuracy of temperature and precipitation was generally close by using three methods, and the spatial distribution was relatively consistent. However, ANUSPLIN method was significantly better than IDW and ordinary kriging methods in detail for areas with large heterogeneity of element. (2) The interpolation accuracy of air temperature and precipitation based on ANUSPLIN was higher than that based on IDW and ordinary kriging. For example, the mean absolute error (MAE)and root mean square error (RMSE) of air temperature interpolation were both less than 0.5 ℃. The RMSE (0.381 ℃) of temperature interpolation based on ANUSPLIN was smallest, it was second based on ordinary kriging（0.459 ℃）, and it was biggest （0.463 ℃） based on IDW. The RMSE (37.8 mm) of precipitation interpolation based on ANUSPLIN was smallest, it was second (42.2 mm) based on the ordinary kriging, and it was biggest (49.1 mm) based on IDW. (3) The average interpolation error of temperature in plain area was lower than that in mountainous area. The average interpolation error of precipitation in the coastal area had largest error, showing an obvious low estimation. In conclusion, ANUSPLIN was more suitable for the spatial treatment of meteorological elements in Zhejiang  Province under complex terrain and climate background.