Nonlinear Multi-field Early Warning Model And Method Of Mountain Flood Geological Disaster In Zhongshan County

China is affected by many factors such as special natural and geographical environment,severe weather with high intensity precipitation,and unreasonable land use by human beings,that mountain flood geological disasters,floods,mudslides,and landslides frequently occur.Mountain flood geological disaster will cause damage to water conservancy facilities,roads,bridges,houses,and agricultural production,and even cause human casualties,that causing huge losses to people's lives and property.Accelerating the prevention and control of mountain flood geological disaster is a major pressing task related to China's economic and social development.Mountain flood geological disasters occur mostly in the windward slopes of the mountains and the subtropical areas in the center of the storm,they are the key defense areas for mountain torrents in China.Zhongshan County is selected from Guangxi Zhuang Autonomous Region as a subtropical mountain flood geological disaster assessment and early warning study object.Based on the disaster characteristics of subtropical mountain flood geological disaster,the nonlinear characteristics and the laws of pregnancy disasters of subtropical mountain flood geological disasters are studied and explored,as well as establish scientific and effective early warning technologies and methods.The nonlinear multi-field pregnancy disasters mechanism was extracted,and RS and GIS technologies were combined with artificial neural network technology to explore new technologies for mountain flood geological disaster assessment and early warning.Go to the site to investigate and collect the river systems,meteorological conditions,landforms,geological soils,etc.in Zhongshan County of the study area and statistic human economic and social information.Analyze the characteristics of historical mountain floods,susceptibility points and hidden danger points of mountain flood geological disasters,and analyze the causes of mountain flood geological disasters from the aspects of rainfall,topography,stratigraphic lithology,river systems,vegetation coverage,population distribution,and human activities.Using historical data,remote sensing images,and field surveys as data sources,DEM data,multi-spectral data,meteorological and hydrological data are processed on ENVI,Arc GIS,and MATLAB platforms to obtain a full range of elevation,slope,vegetation cover index,soil looseness coefficients,valley ridge types,rainfall,and other data in the study area.These data are input as input factors of the artificial neural network model,and take the risk level of mountain flood geological disasters as output factors.BP artificial neural network,radial basis artificial neural network and generalized regression artificial neural network model for risk assessment and early warning of mountain flood geological disasters in Zhongshan County are constructed.The network models are trained and tested,as well as compared and analyzed.The nonlinear multi-field artificial neural network models that suitable for simulated mountain flood geological disasters are discussed.For the mountain flood geological disaster fitting models,the radial basis artificial neural network is more suitable for risk level fitting,and the accuracy and fitting time have advantages.The generalized regression artificial neural network is the second one;the accuracy rate of the BP artificial neural network is lower,but it also exceeds 0.87,and the fitting time is slow,but it does not exceed 1 second.For the mountain flood geological disaster early warning models,the radial basis artificial neural networks are not suitable for prediction and have poor adaptability to the new data added.The BP artificial neural network and generalized regression artificial neural network model are suitable for the early warning of mountain flood geological disasters.The accuracy of the BP artificial neural network model is about 86.7%;the accuracy of the generalized regression artificial neural network model is more than 90%,and the forecasting time is shorter and the overall effect is better.However,in order to avoid the errors of few cases,the two models can be used together.Specific complex conditions can be combined with other conditions to analyze.In the database of elevation,slope,and valley ridge types obtained from the DEM data,the elevation,slope,and valley ridge types of the prediction point are extracted;the vegetation coverage index and soil loosening coefficient at this point are extracted from the database of vegetation cover index and soil looseness coefficient obtained from real-time updated g multispectral remote sensing data.Through real-time meteorological forecasting,rainfall is used as a variable element of early warning,and rainfall in real time or hours later is input into the artificial neural network model.The six kinds of impact factors are calculated in a mountain flood artificial neural network model after big data training to obtain the risk level of mountain flood geological disaster and realize the early warning of mountain flood geological disasters.