Comparison and integration of heuristic and statistical models of landslide susceptibility mapping

Heuristic and statistical models are both important models of landslide susceptibility mapping. Experts´ Scoring Model (ESM) and Generalized Linear Model (GLM), typical heuristic and statistical models respectively, are adopted to carry out a case study of Shenzhen where landslides are increasingly serious. The outputs of the two models obviously differ in terms of receiver operating characteristic (ROC), susceptibility distribution, information entropy, and prediction efficiency. Two integration methods, one to change the ranks of the factors and the other to change the weights of the factors, are evaluated and the outputs of the integration models appear to be more comprehensive and reasonable than ESM and GLM. This paper will provide a guide to future studies on landslide susceptibility mapping.