Using 40year meteorological data at 660 conventional stations in China, the influence of direct interpolation （DI） and integrated remainder method （IRM） on temperature was analyzed by 3 intepolation methods of IDW, Kriging and Spline. A method of removing deterministic part in temperature was put forward which can improve temperature calculation precision. The error was contrasted from time to space and the interpolation sensitivity to station density was investigated finally. The result shows that: excluding some months， mean absolute error （MAE） of IRM was in 1 ℃ which was 0.6-1.6 ℃ lower than DI, the precision was improved by 38％-85％. The precision of IRM did not rely on station density and interpolation method which has a high accuracy and stability. That means it is not the main accesses of increasing observation stations and modifying interpolation methods on improving precision of temperature calculation.