Abstract: Spatial heterogeneity widely exists in the nature.Traditional spatial association data mining assumes that the area on which the mining algorithm performs is evenly distributed,which leads to the mismatch between the mined knowledge and the reality.We suggest that spatial association mining should consider this spatial heterogeneity when designing mining algorithms.The characteristics of spatial association mining were analyzed.Three key measuring indexes indicating spatial association strength were defined.The method of calculating the indexes was presented.The algorithm for mining spatially heterogeneous association patterns and their corresponding subregions in which the pattern shows strong association was proposed.Practical application proved that the proposed strategy was valuable and effective in mining spatial association patterns under spatially heterogeneous environment.