Abstract: In view of the deficiency in traditional method in feature extraction of complicated surface karst collapse, this paper firstly analyzed the properties of spatial distribution, spatial structure and spatial trend of sample data by ESDA, and then determined the optimum scheme by estimating、examing and optimizing parameters of different model schemes, taking Xu zhou surface karst collapse as example, feature extraction of complicated surface karst collapse model was founded and examined under the GIS platform.The results show that the accurate extraction of the boundary information of the surface karst collapse was made and on the same time the modelling errors was reduced and interpolation accuracy was improved for the method fully considered spatial dependencies of sample data.