Research On Prediction And Evaluation Of Geological Hazard Risk Based On Multi-source Data And Deep Learning

With the intensification of human activities,geological disasters have also had an increasingly serious impact on human production and life.Geological disasters in human life gathering areas are often of high risk and are prone to large casualties and property losses.Therefore,the evaluation and prediction of geological hazard risk is particularly important.It has important theoretical and practical significance for disaster prevention and reduction,forecasting and early warning of geological disasters.In the past studies on the prediction of geological hazard risk,the traditional analytic hierarchy processes or conventional neural network methods are mainly used.Based on the geological environment of the Anhui section of the Tanlu fault zone and the results of field investigations,the geological hazard development characteristics and distribution of the Anhui section of the Tan-Lu fault zone are carried out by means of aerial interpretation,field investigation,theoretical research and model simulation.Based on the research mentioned above,the deep learning theory is introduced into the research of geological disaster risk prediction.Based on the comprehensive analysis of the relationship between the geological hazard level and the rainfall data that have a great impact on geological disasters,the deep neural network model for predicting the geological hazard of the Anhui section of the Tanlu fault zone is built,and the risk of geological disasters in the study area is predicted.The main achievements of this research are as follows:(1)Based on multi-source remote sensing data,the remote sensing interpretation of the Anhui section of the Tanlu fault zone is carried out,and the spatial distribution characteristics of the Anhui section of the Tanlu fault zone are studied by using GIS spatial analysis and spatial statistics.The structural interpretation of the Tanlu fault zone indicates that the fault zone is mainly composed of the Wuhe-Hefei Fault,Shimenshan Fault,Chihe-Taihu Fault,Jiashan-Lujiang Fault and 40 secondary faults in the Dabie Mountains.(2)Using remote sensing,GIS spatial analysis and spatial statistical techniques,preliminary exploration of the boundary problem of the Anhui section of the Tanlu fault zone in the Dabie Mountains is conducted.The fault strike line on the west side of the fault-intensive area of the NNE direction,which is interpreted by remote sensing in the Dabie Mountains,is determined as the boundary of the Anhui section of the Tanlu fault zone in the Dabie Mountains.The fault strike line is determined as the boundary of the Anhui section of the Tanlu fault zone in the Dabie Mountains.Through analyzing comprehensive interpretation results and tectonic background,the Anhui section of the Tanlu fault zone is divided into two sections by the Xiaotian-Mozitan Fault,and the average length of the Minjiang-Susong section(or southern section)is about 136.74 km,with an average width about 38.38 km,and the overall trend is about N34.9°E.The average length of the Jiashan-Yujiang section(or north section)is about 261.09 km,the average width is about 25.99 km,and the overall trend is about N23.5°E.The total length of the Anhui section of the Tanlu fault zone is about 397.83 km,the average width is 30.35 km,and the overall trend is about N33.3°E.The fault zone generally exhibits the spatial distribution characteristics of“south short and north long”,“south wide and north narrow”and“the overall trend from the south to the northern fault zone gradually shifts northward”.(3)Within the scope of the Anhui section of the Tanlu fault zone determined by the above method,Twelve factors such as“engineering geological rock group”,“fracture density”and“24-hour maximum rainfall”are identified as risk assessment indicators.The analytic hierarchy process is used to evaluate the geological hazard risk in the Anhui section of the Tanlu fault zone.The results show that the low-risk area in the study area is the largest,with an area of about 4603 km~2,accounting for 42.84%of the study area.The high-risk area is second,with an area of about 2844 km~2,accounting for 26.47%of the study area.The area occupied by the risk zone is about1658 km~2and 1639 km~2,accounting for 15.43%and 15.26%,respectively.The area occupied by the two is basically the same.(4)This paper exploratoryly introduces the method of deep learning into the prediction of geological hazard risk,and uses the results of geological hazard assessment in the study area obtained by the analytic hierarchy process.The“24hours maximum rainfall”and“five-day rainfall”are two factors that have a major impact on geological disasters is used as the input layer.The“geological hazard level”is used as the output layer and Deep Neural Network(or DNN)models is used to predict the risk of geological hazards in the study area.The results show that the accuracy of the DNN model for the prediction of geological hazard in the study area is 74%,which has a good predictive effect.