Abstract: Most researches on collapse gully,a frequent natural disaster in south China,focus on qualitative analysis rather than quantitative risk assessment and prediction. With southeast Guangxi Province as the study area,we established a susceptibility prediction model for collapse gully using Logistics model with 80% of the acquired collapse gully based upon RS and GIS images as training sample. We selected seven indicators inclusive of elevation,slope gradient,aspect,soil,geology,land use,and vegetation coverage,and calculated the weight of each index. Furthermore,we assessed the model’s fitting degree using ROC curve,and took 20% of the total collapse gully for model verification. Results showed that the AUC(area under curve) value of the model was 0.718,implying good prediction accuracy. Geology is the key factor of erosion,followed by elevation and slope aspect. Extremely susceptible,highly susceptible,and moderately susceptible areas accounted for 10.17%,19.30% and 26.49%,respectively of the total area. The zoning map of collapse gully accorded with the actual distribution of collapse. Results demonstrated that Logistic model is applicable to assess the key influencing factors and susceptibility of regional collapse gully.

Key words: collapse gully, Logistic model, influence factors, susceptibility, southeast Guangxi Province