Landslide Hazard Assessment Based On Logistic Regression And SINMAP Model In Bailong River Basin

The Bailong River Basin is located in southern of Gansu Province. With steep mountains, deep valley, concentrated rainfall and heavy rains, the environment is quite weak and that makes it become one of the four major geological disaster regions that take the highest in China. Studying on the main induced factors and space distribution of landslide disaster, establishing suitable landslide hazard assessment models and making landslide hazard assessment research has important theoretical significance and practical value. On one hand, it provides a research basis for landslide monitoring and early warning, on the other hand, provies decision support for local government to prevent and mitigate the disasters.This paper analyzes16landslide hazard impact factors. Under the support of GIS technology, the Logistic Regression Model and the SINMAP Model are applied to assess landslide hazard of the region and study the applicability of landslide hazard assessment of the two models. Landslide hazard assessment of Logistic regression models are set up based on grid cell unit and slope unit respectively, then assessment results of different model are compared, and the regional multi-scale statistical and deterministic compound assessment model was proposeed. Finally, on the basis of the SINMAP model, the regional slope stability under different rainfall conditions is analyzed. On the basis of those, this research paper draws the following main conclusions.Based on grid units Logistic regression model and impact factors analysis of landslide, this paper states the spatial distribution law of landslide, and determines that the main impact factors of landslide, in this region, include elevation, NDVI, slope, lithology, land use, distance to river and average rainfall within an hour. The secondary factors include24-hour average rainfall and aspect.With Logistic regression model of the grid cell and slope unit respectively, it assesses the landslide hazard. On the whole, both two results are similar. Both can perform the regional hazard zonation of regional changing trends, but the second method can well distinguish the stability between different slope unit and it has a more practical significance. The two-scale Logistic regression model both can be used to assess the landslide hazard in the study area. The slope unit is a reasonable model unit in the region, and the Logistic regress model with slope unit is more suitable in this region for landslide assessment. The evaluation results on the regional stability by SINMAP model show that the area of unstable partition is4339.8km2, accounting for64.49%of the total area. The evaluation results of the unstable region and the existing landslide point distribution are basically the same, indicating SINMAP model has a certain applicability in the region.The evaluation accuracy of these two model is compared according to landslide validation data. The accuracy of the Logistic regression model with grid cell is70.24%and SINMAP model is69.35%. The Logistic regression model can well reflect the regional distribution of landslide disaster risk trend and performance better on average. But SINMAP model is better at reflecting the stability of the partial slope.Half quantitative analysis of landslide hazard under different conditions is included based on SINMAP model. The result shows that rainfall has a distinct impact on landslide. The percentage of landslide in unstability part is more than50%when rainfall in an hour is over80mm.