Landslide Susceptibility Mapping And Attribution Of Small And Medium Scale In Northwest Yunnan

Northwest Yunnan is the core hinterland of the World Natural Heritage "Three Parallel Rivers".It is located in the southern extension of the Qinghai-Tibet Plateau,which caused the material to flow eastward in the orogeny 30 million years ago,forming a unique geographic landscape and huge fracture zone in this region.Towering mountains,deep rivers,and frequent earthquakes make it an area with a high incidence of landslide disasters in China,and it is also a key and difficult area for the prevention and control of geological disasters in Yunnan Province.By analyzing the relationship between spatial distribution of landslide in northwestern Yunnan and environmental factors,growth environment and mechanism of landslide occurrence are explored to provide support for the prevention and control of regional landslide disasters.Through literature search,the study summarized the most used basic and inducing factors in landslide sensitivity mapping.Combined with the availability of data,14 environmental variables were initially selected from natural and man-made aspects.In order to overcome the huge amount of calculation and confounding problems caused by too many explanatory variables in the modeling process,we use Geodetector to detect the explanatory power of each factor on the spatial distribution of landslides and the dominant factor with explanatory power greater than 0.01 was retained.According to the above analysis,we use the maximum entropy model,one of the machine learning methods,to build a model.The maximum entropy model is widely used in the field of ecology with its excellent performance in forecasting and forecasting,and is known as the best species distribution model.The advantages of simple and easy-to-understand models and high accuracy are also applicable to landslide sensitive zoning.Northwest Yunnan consists of three prefectures(cities),covering an area of 58,000 square kilometers,with diverse landforms and complex environments,making it difficult to identify landslide factors.In order to improve the prediction accuracy of the model and overcome the influence of the spatial heterogeneity of the geographical environment in northwestern Yunnan on the accuracy of the model,based on expert knowledge,the two factors that have the greatest impact on landslides,lithology and slope,were selected for stratified experiments.Among them,the lithological stratification is clustered by the frequency of landslides per unit area in each type of lithological zone.The 10 lithological zones are divided into 4 layers except for the lithology that never occur.The slope stratification determines the slope of each landslide.The value is divided into 6 layers according to the slope grade.The experiment uses lithology that highly correlated with the spatial pattern of landslide disasters to stratify,and the explanatory power of each factor on the landslide pattern is tested in each layer,and the dominant factor with high explanatory power is retained Perform modeling.The ROC curve shows that the results of layered modeling are higher than those of overall modeling.The main findings of the study are as follows:(1)When modeling the entire region,14 natural and man-made factors selected based on literature retrieval and statistics were included in the experiment to model.AUC value is only 0.793,and the accuracy of the model is not high.In this study,the northwestern Yunnan was modeled according to lithological stratification,and the AUC value in each layer was above 0.85,of which three layers exceeded 0.9,reaching an excellent level.(2)Geographic detector is a statistical tool to explore the degree of influence of environmental factors.Through the explanatory power results of the factors detected by the overall and stratified experiments,it is found that the influence of human factors is deeper than that of natural factors.The impact of the intrusion on the environment continues to deepen.(3)The hierarchical model is used to map the landslide susceptibility in the study area.From the overall spatial pattern of landslide sensitivity in the study area,it can be seen that the high and extremely high risk areas account for 19.63%,showing a duality of clusters and bands distribution status.Among them,Lijiang has the highest risk,accounting for 30.75% of the total area,with 97 very high risk threatening settlements;Nujiang is the second,accounting for 22.51% of the total area,with 54 very high risk threatening settlements;Diqing is the lowest,with very high risk settlements accounting for 17.48%.