High precision grids of meteorological data are essential input parameters for most kinds of large-scale global models.
Improvements on data accuracy can make models running more effectively and exactly. At present, IDW, Kriging and
Splines are often used as common interpolation methods, but for meteorological data their interpolation accuracy is not
high enough and the interpolated raster images are sometimes too rough. This paper attempts to use ANUSPLIN, spatial
interpolation software based on the theory of thin plate smoothing spline interpolation, to interpolate average temperature
and precipitation in different time scales as daily, monthly, annual, with source data from 71 meteorological stations in
Anhui Province. Before interpolation, experiments on different ANUSPLIN models were implemented with a
combination of three variants (Longitude, Latitude and Elevation) to ensure the best one correspond each source data in
different scales, the results showed that CO2 (elevation as a covariate and the order of spline is 2) model fits daily and
monthly temperature data, CO3 model is effective for monthly and annual precipitation data. A comparison between the
interpolated surfaces using ordinary kriging method and ANUSPLIN showed the latter one performs more accuracy and
smoothness in all the time scales of temperature and precipitation: the mean error of daily mean temperature
interpolation can be reduced by 0.103 centi-degree, monthly one by 0.091 centi-degree, annual one by 0.078 centidegree,
monthly precipitation interpolation mean error can be reduced by 4.649mm, annual one by 22.194mm. The high
precision of interpolated data can meet the need of many climatic and ecological models.