Research On Dynamic Prediction Model Of Landslide Susceptibility Based On CA-Markov

Due to the unstable geological conditions and complex terrain,geological disasters occur frequently in China,causing serious casualties and economic losses,which have a serious impact on social and economic development.Landslide is one of the most common and serious geological disasters,so regional landslide susceptibility evaluation is a hot research topic.According to geological conditions and other factors,the distribution of regional landslide susceptibility can be evaluated,which can provide reference for the prevention and control of landslide.Then,the dynamic prediction of landslide susceptibility will be helpful to the landslide prevention and control,which is conducive to the arrangement of production and livelihood.Starting from the research of landslide evaluation factor,regional landslide disaster susceptibility evaluation,Logistic Regression model(LR),Cellular Automata-Markov model(CA-Markov),using the remote sensing image processing techniques,GIS techniques,statistical analysis models,and combining with LR model and CA-Markov model,this paper carried out the research on the landslide susceptibility evaluation model based on LR and the dynamic prediction model of landslide susceptibility based on CA-Markov,and designed and carried out experiments.Based on the 40 groups of experiments,a high accuracy dynamic prediction model of landslide susceptibility based on CA-Markov was established.The prediction of landslide susceptibility in the study area in 2017 and 2020 were finally obtained.After theoretical research and model experiments,the thesis drew the following three conclusions:1)After statistical analysis of the classified section of the evaluation factors,the distributions of the landslide in the most concentrated section of each evaluation factor are obtained,namely: DEM(1620-2014m),slope(31.487586 °-38.753952 °),aspect(southeast: 112.5°-157.5°),plan curvature(3 classes: 0.8393°-0.5622°;6 classes:-0.3003°-0.3466°),lithology(Ar3-Pt1: plagioclase amphibolite,mixed gneiss,granulite and fine granite),distance from stream(0-300m),distance from fault(0-400m),land use(bare land)and NDVI(0.236391-0.358369).2)The evaluation model of landslide susceptibility based on LR has a high accuracy rate for the evaluation of landslide susceptibility in the study area(AUC average: 81.37;area under curve abbreviates AUC);the LR model is suitable for the research of landslide susceptibility evaluation.3)Based on CA-Markov's predictive model of landslide susceptibility,the prediction results of landslide susceptibility evaluation in the study area were compared with the real values(year 2016),which has a high consistency(Kappa: 0.7509),indicating that the model can predict the landslide susceptibility in the study area.CA-Markov model is suitable for dynamic prediction of landslide susceptibility.Based on the study of the dynamic prediction model of landslide susceptibility,this paper aims to explore the method of predicting the results of regional landslide susceptibility evaluation,and to provide support for the effective work of landslide disaster prevention and control.