| Human activities often destroy the stability of slope structure.Under the effect of rainfall and other inducing factors,these slopes are prone to landslide disasters.In Jiangxi Province,the hills are widely distributed,the hydrometeorological and environmental geological conditions are complex,and the geological disasters(especially landslides)occur frequently.Jiangxi is one of the provinces with frequent geological disasters,especially,landslides in the country.In this paper,Taking Jiangxi Province as an example,based on the analysis of the relationship between landslide points and disaster causing factors in the region by using geo-information technology,the data-driven machine learning method is used to evaluate the landslide susceptibility in the study area.The following are the main contents and achievements:(1)According to the geographical and geological environment conditions of the study area,the main types of landslide disasters,the temporal and spatial distribution characteristics of landslide points and the formation conditions of landslide are analyzed.The study area is dominated by soil landslides,which are mainly distributed in the surrounding mountainous and hilly areas of roads,river banks and slope sections of Valley engineering buildings.Due to the free face of slope formed under the condition of human engineering excavation,it is easy to breed landslide and other geological disasters under the influence of rainfall and other factors.(2)Through field landslide investigation and analysis of the main disaster factors in the research area.Fourteen geo-environmental factors were selected,including lithology,fault,slope,elevation,land cover,relief degree of land surface(RDLS),aspect,road,river,normalized vegetation index(NDVI),annual rainfall,mean accumulated rainfall from March to July,from March to June and from May to July.According to the similarity of geographical environment and the formation conditions and characteristics of landslide points.All these factors were divided respectively into subsets with same interval to establish the susceptibility assessments system of landslide disaster.(3)The paper selects he logic regression model(LRM)based on information quantity(IV)to evaluate and divide the landslide in the study area.The correlation analysis of the factors of landslide vulnerability evaluation is carried out to improve the feasibility of the evaluation model.The known landslide points are divided into training set and verification set according to the ratio of 7:3,and the prediction model of vulnerability is constructed and evaluated.In this paper,the integrity of two linear elements is used for modeling: Model 1 uses incomplete linear feature elements(such as railways,expressways,national roads and rivers);Model 2 uses complete linear feature elements(such as railway,expressways,national roads and rivers,as well as tributaries,streams,country roads and village roads).the model prediction has improved obviously and the reliability has also been improved with the increase of River tributaries,streams and county-level and township highways and village roads.AUC value has been increased from 0.857 to 0.863.The prediction model is applied to the study area,and the probability of landslide is estimated.According to the equal interval method,the probability value is divided into five grades(stable area,low prone area,medium prone area,high prone area and extremely high prone area),so as to divide the potential area of landslide occurrence.(4)In this paper,30 meter evaluation unit is used to study in the provincial area,and well results are obtained,which reflects the description of surface morphology and landslide points more precisely.The prediction method of landslide susceptibility based on IV + LRM analysis can meet the needs of this study in terms of accuracy and efficiency.With detailed landslide data,the method used in this study can be extended to other areas for regional and national landslide susceptibility assessment. |
PDF Full Text Request