基于GIS和Logistic回归模型的兰州市滑坡灾害敏感性区划研究

Landslide Susceptibility Zoning Study in Lanzhou City based on GIS and Logistic Regression Model

针对兰州市脆弱的地质环境和频繁发生的滑坡灾害,采用Logistic回归模型,以ArcGIS和SPSS软件为工具,选取地层岩性、断层构造、坡度、地貌、植被覆盖度、7～9月平均降水、道路(公路、铁路)作为滑坡灾害影响因子。首先对每个影响因子分级并计算每个因子指标值,然后在Arc-Map中对影响因子图层进行叠加操作,最后在SPSS软件中运用Logistic回归方法,计算出每个影响因子的系数值并建立Logistic回归模型。根据Logistic回归模型在ArcMap中绘制兰州市滑坡灾害敏感性区划图,区划图和实际的滑坡分布情况基本吻合。模型的Kappa系数值和ROC曲线下面积值(AUC值)分别为0.623和0.709,两种方法的检验结果均表明模型模拟效果较好,能应用于兰州市滑坡灾害敏感性区划研究中。

For fragile geological environment and the frequent occurrence of landslides in Lanzhou city,in this paper,based on Logistic regression model and adopted ArcGIS and SPSS softwares,selected elevation,slope,average precipitation from July to September,vegetation cover,roads,railways,lithology,faults as impacted factors of Lanzhou landslide disaster.First of all,impacted factors were classificated and calculated index values of every class at the same time,and then overlapped all impacted factors in ArcMap of ArcGIS software.The end,use of Logistic multiple non-linear statistical methods to calculate the coefficient of each impacted factor in SPSS software,according to the statistical results,Logistic regression model was gained.Drawed the Lanzhou landslide susceptibility zoning map based on Logistic regression model in ArcMap,zoning map of landslides accord with the actual distribution of landslides.Kappa value and AUC value of the model were 0.623 and 0.709 respectively,the tested results of the two methods showed that the model is good and can be applied to the Lanzhou landslide susceptibility zoning studies.