基于信息量法的多模型对比分析石家庄西部山区滑坡易发性研究

Study on Multi-model Comparative Analysis of Landslide Vulnerability in Mountainous Areas of West Shijiazhuang Based on Informative Approach

石家庄市西部山区地质状况不良,气象方面常在雨季形成暴雨,因而地质灾害频发,其中滑坡灾害占比较重,因此针对石家庄市西部山区的滑坡灾害进行易发性分析有重要意义。现有单一模型在滑坡灾害易发性评价方面应用广泛,但仍有局限性。为提高对区域滑坡灾害易发性评价精准度,采用单一模型和复合模型等多模型对比分析研究区滑坡易发性,得到评价精度更好的评价模型。首先通过对研究区孕灾条件筛选出降雨量、坡度、高程、坡向、曲率、地质岩性、距水系距离、距断层距离、地震烈度、距道路距离、植被覆盖率和土地利用类型共12项评价因子。基于信息量法,用层次分析法、熵权法以及组合赋权法进行对比分析,最终得到易发性分区图,通过ROC曲线和灾害点密度验证其可靠性。结果表明,以上三种分析方法在ROC曲线中的AUC值分别为0.732、0.749、0.766,三种模型评价结果都有一定的可靠性,其中组合赋权法模型精度更高,可靠性更好,确定采用组合赋权模型。

The geological condition of the western mountainous area of Shijiazhuang city is poor,and the topography of this often forms heavy rainfall in the rainy season,so the geologic disasters are frequent,among which landslide disasters account for a relatively heavy proportion.Therefore,it is of great significance to analyze the susceptibility of landslide disasters in the western mountainous area of Shijiazhuang city.The existing single model has been widely used in landslide disaster susceptibility evaluation,but it still has some limitations.In order to improve the accuracy of the evaluation of regional landslide disaster susceptibility,multiple models such as single model and compound model were used to compare and analyze the landslide susceptibility in the study area,and an evaluation model with better evaluation accuracy was obtained.Firstly,a total of 12 evaluation factors including rainfall,slope,elevation,slope direction,curvature,geological lithology,distance from water system,distance from fault,seismic intensity,distance from road,vegetation coverage and land use type,were screened out by evaluating the evaluation factors related to forming disaster in the study area.Based on the informative method,the hierarchical analysis method,entropy weighting method and combined weighting method were used for comparative analysis,and the susceptibility zoning map was finally obtained,and its reliability was verified by ROC curve and density of disaster points.The results show that the AUC values of the above three analysis methods in the ROC curve are 0.732,0.749,0.766,respectively,and the evaluation results of the three models have a certain degree of reliability,among which the combined weighting method has higher accuracy and better reliability.Finally,the combined weighting model is determined to be adopted.