Abstract: Spatial interpolation is an approach to estimate data at an un-sampled points based on known observations at sampled stations and generate a continuous surface model. In order to obtain the estimated value for missing data and precise spatial surface models, it is necessary to develop a high-performance spatial interpolation method. Based on the typical adaptive inverse-distance weighting (AIDW) spatial interpolation method, a new method, called adaptive cluster gradient inverse-distance weighting (ACGIDW) is presented in this paper. Considering the effect of latitude, longitude, elevation and complex topography factors, this method offers a more accurate result, 1) it adjusts the distance-decay parameter in the IDW method to improve the adaptability of the ACGIDW according to the spatial distribution pattern of the stations; 2) it was tested by using two groups of different actual meteorological data. The experimental results demonstrate its superiority and feasibility.