Abstract: The demand for spatial data sets of precipitation in digital form has risen dramatically in recent years. This paper describes a new surface modeling method, high accuracy and high speed method(HASM), which has been successfully used for digital elevation model(DEM) construction and ecosystem changes. We explored the relationship between precipitation and geographical/topographical variables and local topographical factors, developed a polynomial regression model for seasonal precipitation in each region, and then used symmetric successive over-relaxation method and preconditioned conjugate gradient method to solve the matrix equations produced by HASM, which effectively improved the model. We used ArcGIS to create buffers with a specific distance around each subarea, and the actual computational domain expanded to the buffer area. Polynomial regression and residuals interpolation using HASM were applied to develop a gridded precipitation database for China in seasonal scales with a resolution of 1 km in longitude and latitude. We used precipitation data measured at 711 stations during the period of 1951-2010, with 80% of them used for surface development and 20% reserved for validation tests. Accuracy tests revealed that HASM is superior to the classical methods such as Kriging, Inverse Distance Weighting (IDW) and Spline. Precipitation surfaces generated by HASM showed good performance in precipitation research. Therefore, HASM can be considered as an alternative and accurate method for precipitation interpolation in China.

Key words: China, HASM, interpolation, polynomial regression, seasonal average precipitation