Research On Mountain Flood Disaster Risk Dynamic Assessment Method Based On Provincial And County Scale

Mountain flood is the main disaster caused by natural disasters in China,and its prevention is the difficulty and weak link in flood control and disaster reduction.Mountain flood disaster risk assessment is the premise of mountain flood disaster prevention strategy.Through comprehensive analysis of natural environment conditions,social and economic conditions,human activities and other aspects,the mountain flood disaster risk in the study area is divided into different levels,which is of great significance for formulating disaster prevention and mitigation measures in different risk areas,serving local economic and social development,and safeguarding people’s property safety.In this paper,Henan Province and Linzhou City(countylevel city)mountain flood disaster prevention and control areas at two spatial scales were selected as the research area,the consequences of reverse diffusion method was used to construct the risk assessment index system,establish mountain torrent disaster risk assessment model based on random forest algorithm,studies of mountain flood risk zoning evolution under different time and space background,discusses risk grade and the importance index factor on the space correlation,On this basis,the corresponding disaster prevention measures are put forward.The main research contents and achievements are as follows:(1)The risk index systems of different spatial scales are constructed by using consequence backward diffusion method.Mountain flood disaster risk varies at different spatial scales due to climate factors,landform,mountain flood disaster prevention measures and social and economic development.For analysis of mountain torrent disaster risk formation mechanism,relatively comprehensive combing the mountain torrent disaster under different spatial scales of all indicators of risk factors,this paper USES the consequences of the reverse diffusion method from two aspects of natural disaster system and social disasters,converse thinking,reverse diffusion,according to the particular form of mountain torrent disaster losses for mountain torrent disaster occurs,the main factors lead to From the main factors to the specific risk factors,the mountain flood disaster risk index system at two spatial scales of province and county was constructed by comprehensively considering the disaster forms and dominant factors at different spatial scales.(2)Establish mountain flood disaster risk assessment model based on random forest algorithm.In order to avoid the difficulty in accurately determining the weight of index factors in traditional assessment methods such as analytic hierarchy process(AHP)and entropy weight method,this paper introduced random forest algorithm into mountain flood disaster risk assessment and established mountain flood disaster risk assessment models at different spatial scales.The mountain flood risk areas in the study area were divided into five risk levels: very low,low,medium,high and very high,and the accuracy of the model was verified by historical mountain flood disaster points at different spatial scales.The results show that the random forest model has high accuracy in mountain flood disaster risk assessment.The overall risk level of Henan Province was slightly higher from 2000 to 2009,while most of northeast China changed from medium risk level to low risk level with different periods.From2010 to 2019,the overall risk level of Linzhou City was slightly higher,and the risk level of western region changed from high risk level to low risk level in different periods,while the risk level of central Yaocun Town,Hengshui Town and some southern regions increased significantly.(3)The spatial autocorrelation of risk grade and importance index factors under different spatial and temporal backgrounds was analyzed.Moran index and high-low clustering were used to describe the spatial global autocorrelation of mountain flood risk levels under different spatio-temporal backgrounds.The spatial correlation of mountain flood risk levels was analyzed,and the discrete,random or clustering patterns of mountain flood risk were evaluated.Clustering and outliers were used to analyze the local autocorrelation of risk grade and each index factor and measure the degree of local correlation with its surroundings.The spatial distribution of mountain flood disaster risk levels and the dynamic evolution of the spatial interaction of risk indicators under different spatio-temporal backgrounds were obtained through the spatial global autocorrelation analysis of risk levels and spatial local autocorrelation analysis of risk levels and index factors.(4)Put forward suggestions for disaster prevention and mitigation at different spatial scales.Based on the results of risk regionalization and spatial correlation analysis at different spatial scales,and with reference to risk index factors and extreme rainfall conditions,the corresponding disaster prevention and mitigation suggestions were put forward for two research regions at different spatial scales:province and county.