Abstract: Using the temperature data of 128 meteorological stations in Shenzhen, five methods, including inverse distance weighting method (IDW), local polynomial interpolation (LPI), ordinary kriging (OK), Cokriging and thin plate spline interpolation (TPS) of Anusplin, are applied to cross-validate the spatial interpolation of the monthly mean temperature of Shenzhen in 2017. The accuracy of the five interpolation methods in Shenzhen shows that TPS performs the best. The cross validation errors of LPI, IDW, OK and Cokriging are significantly correlated with the altitude. These 4 methods are greatly affected by the altitude and the errors are larger in areas with higher altitude. The correlation coefficient between TPS interpolation errors and altitude is smaller than the other four methods. This result indicates that TPS method is more suitable for the temperature interpolation of complex terrain like Shenzhen. This study has great practical value for meteorological interpolation business.