Research On Mountain Flood Disaster Early Warning Method Based On Big Data

Mountain torrents are one of the most serious disasters among flood and drought disasters,which seriously threaten the safety of people's lives and properties.Mountain torrent disasters have the characteristics of strong suddenness.In order to effectively prevent mountain torrent disasters,carrying out research on mountain torrent forecast and early warning has become the focus of mountain torrent disaster prevention work.The study of mountain torrent disaster early warning method is an important part of mountain torrent forecast and early warning.At present,the early warning method with static critical rainfall as the early warning indicator is mainly used in China,but due to the influence of the underlying surface and the temporal and spatial distribution of rainfall,its early warning accuracy is poor.In order to improve the accuracy of early warning,this study relies on the subject of "National Key Research and Development Program(No.2019YFC1510605)",and based on the self-built database,selects typical river basins and disaster prevention objects in Fujian Province,respectively,for the flow early warning based on distributed hydrological model.Researches were carried out on the dynamic critical rainfall early warning method,and the early warning effects of the static critical rainfall early warning method were compared and analyzed,in order to provide better technical support for the forecast and early warning of mountain torrent disasters.The main conclusions are as follows:(1)Collect and sort out the national typical mountain torrent disaster data and build a mountain torrent database.In order to improve data accuracy and utilization,this project uses natural language processing methods to collect multi-level and multisource data in different forms and formats into the database,including basic information on disaster basins,rainfall data for 15 days before and after the disaster,and various survey data,etc.Provide data support for subsequent studies on mountain torrent disaster early warning methods.(2)Using 30 floods in each of the two watersheds of the Minjiang River in Fujian Province(Meixi and Siqian)to calibrate and verify the parameters of the CNFF model for mountain floods in China.Among them,the first 20 fields are used for parameter calibration,and the last 10 fields are used for parameter verification.The results show that the regular mean value of the peak flow error rate of the simulated floods in the two watersheds is less than 6.4%,and the verification period is less than 10.1%;the peak current error mean rate is less than 1.9h regularly,and the verification period is less than 1.5h;the Nash efficiency coefficient rate is regularly and verified The average period is more than 80%,which proves that the CNFF model has good adaptability in the hilly area of northern Fujian and can be used for early warning research of mountain torrent disasters.(3)Construct CNFF distributed hydrological models of Xixi River Basin in Guangze County and Meixi River Basin in Minqing County,Fujian Province,and conduct flow early warning analysis.Two typical flood processes on July 9 and July 25,2020 were selected for the two watersheds,respectively,and the constructed model was used to simulate river floods and analyze the effect of distributed hydrological model flow early warning.The results show that compared with the static critical rainfall early warning effect,the flow early warning effect based on the CNFF distributed hydrological model is better.At the same time,it can deal with the early warning problems of upstream rainfall and downstream disasters.(4)Determination of early warning indicators of dynamic critical rainfall and analysis of early warning effects.Three typical disaster prevention objects in the Renshou River Basin,Shunchang County,Fujian Province were selected to construct the CNFF distributed hydrological model of the Renshou River Basin.Finally,the early warning indicators of each disaster prevention object under the condition of soil moisture content of 20%,50% and 80% were finally determined by comprehensively considering the characteristics of the small watershed where the disaster prevention object is located,and the characteristics of production and confluence.Then combined with the early warning indicators under different soil water contents,the dynamic critical rainfall early warning indicators under the current soil water contents were interpolated,and the 2019-2021 flood season was selected to compare and analyze the effects of static critical rainfall early warning and dynamic critical rainfall early warning.The results show that the dynamic critical rainfall early warning can fully combine the current state of soil water content in the basin for early warning,the results are more accurate,and the early warning rate is reduced by 72%.