GIS-based rainfall-triggered landslide warning and forecasting model of Shenzhen

Landslide warning and forecasting is essential to landslide risk management. This paper establishes a GIS-based landslide warning and forecasting model of Shenzhen where rainfall-trigged landslides have caused great property and life loss. Based on the high-resolution observation data of rainfall and landslide occurrences in Shenzhen, Kriging interpolation model is applied to derive the regional rainfall data prior to the occurrence of landslides. The binary logistic model is then trained by the forward stepwise method, in which the precipitation of the first, second and fourth day prior to the occurrences of landslide are selected as significant explanatory factors. Then, the confusion matrix and AUC/ROC are calculated to verify the model. Finally the critical precipitation of landslide warning and forecasting of Shenzhen is determined. This paper may provide concrete support for landslide risk management work in Shenzhen and helpful reference for related studies.