Susceptibility Analysis Of Debris Flow In Yongji County Based On Watershed Units

China is one of the countries which have the most serious debris flow disasters in the world.Yongji County,as one of the vulnerable areas of debris flow in Jilin,has suffered major debris flow disasters frequently.Debris flows destroyed houses and buried roads,which seriously restricted the development of local economy and brought serious economic losses to the people.Therefore,it is very important to analyse the susceptibility of the debris flow in Yongji County,which can provide the basis for the assessment,control and prevention of debris flow disaster.Based on the “Geological disaster survey and zoning(1:50000)project in Yongji County,Jilin”,the influence of different statistical methods and mapping units on analysis of debris flow susceptibility were discussed in this paper,we selected two kinds of statistical methods,frequency ratio method and statistical index method,four mapping units,grid units and watershed units with different thresholds,constructed five debris flow susceptibility analysis model,mapped debris flow susceptibility of Yongji County,the main work is as follows:(1)Through field investigation,the number,distribution and development characteristics of debris flow disasters in Yongji County were recorded,and debris flow database was constructed as a basis for establishing debris flow susceptibility mapping model.A total of 123 debris flow hazards were identified.The debris flow disasters are mainly distributed across the southeast mountain area,and less in the northwest plain area.In this paper,70%(n=86)of the debris flows were selected randomly for training data,which is used to create the DFSM models.The remaining 30%(n=37)were used as testing data,which is used to validate the DFSM model.(2)Grid unit is the most commonly used mapping unit for analysing the susceptibility of debris flow,but grid unit is not related to the geological and geomorphic information of debris flow.As a basic unit of debris flow formation and activity,watershed unit can take into account the whole geological environment background of debris flow formation area,circulation area and accumulation area.Based on the Model Builder of ArcGIS,the steps of extracting watershed units based on DEM is developed into a tool,and the efficiency is improved.(3)In order to get better debris flow susceptibility analysis,the influencing factors we chose are not exactly the same for grid cell unit and watershed unit.For grid cell unit,we choose 8 factors: DEM,slope,aspect,landforms,lithology,land use,distance to fault and precipitation;for watershed unit,we choose 8 factors: watershed area,relative height difference,roundness,ratio of poor vegetation area,watershed slope,lithology,landform and precipitation.(4)Chose the watershed unit of threshold 10000 as the mapping unit,based on frequency ratio method and statistical index method,constructed two debris flow susceptibility analysis models,analysed the susceptibility of debris flow and mapped debris flow susceptibility.By ROC curve and AUC,the success rate of model was tested using the training data,the prediction rate was tested using testing data,the results shown that accuracy rates of 0.896 and 0.893 for frequency ratio method and statistical index method,respectively.The associated prediction accuracy rates were 0.875 and 0.865,respectively,showed that the prediction ability of two models are high,but the analysis results of frequency ratio method are more accurate than statistical index method.The model based on frequency ratio method is more suitable for the debris flow susceptibility of Yongji County(5)Chose frequency ratio method as statistical method,chose four kinds of mapping units,grid cell units,watershed units of thresholds 30000,10000 and 5000,respectively,constructed four debris flow susceptibility analysis models,analysed the susceptibility of debris flow in Yongji County and mapped debris flow susceptibility.By ROC curve and AUC,the results shown that accuracy rates of 0.916,0.877,0.896,0.887 for grid cell units,watershed units of thresholds 30000,10000 and 5000,respectively.The associated prediction accuracy rates were 0.877,0.818,0.875,0.841,respectively,which means the four models are accurate in predicting the debris flow occurrence,and the analysis results of the GCU are more accurate than others,followed by the watershed unit of thresholds 10000.The accuracy of the thresholds 30000 is the lowest and the area of watersheds of thresholds 30000 are very large,which is explained as an overestimation of debris flow susceptibility.Therefore,for Yongji County,it is more appropriate to choose watershed unit with threshold of 10000 and 5000 as mapping units.