Patial Prediction Of Landslide Susceptibility Based On Information Content And Random Forest Model

Huangzhong County is located in the Hehuang Valley in the transition zone between the Qinghai Tibet Plateau and the Loess Plateau.The climate is changeable,the geological environment is complex,and the landslide geological disasters are relatively developed.Especially in recent years,with the increase of human engineering activities,it is easy to form dangerous slopes and landslide geological disasters,which threaten people's lives and property.The spatial prediction of landslide susceptibility is to predict the spatial distribution and occurrence probability of landslide disaster,which has certain practical significance and reference value for disaster prevention and reduction.Based on the data of historical landslide disaster points and the analysis of the landslide disaster and its environmental characteristics in Huangzhong County,this paper selects 10influencing factors of landslide disaster in five aspects,including the landform(elevation,slope,aspect,curvature),geological elements(rock and soil type,distance from fault),meteorology and hydrology(rainfall,distance from river),vegetation impact(NDVI)and human engineering activities(distance from the road),and each influencing factor is classified and analyzed.Then the information model and random forest model are used to carry out the spatial prediction and evaluation of landslide susceptibility in the study area.The main conclusions are as follows:(1)By analyzing the relationship between landslide disaster points and influencing factors in Huangzhong County,it is concluded that the concentrated area of landslide disaster is within the elevation of 2532-2691 m,the slope of 15-20 m,the slope direction of southeast and south,the curvature of 0-0.5,the type of rock and soil is single-layer Aeolian Loess,the distance from the fault is 4000-5000 m,the rainfall is 500 mm,the NDVI is 0.2-0.4,the distance from the river is 4000-5000 m,and the distance from the road is 0-500 m.(2)Information model and random forest model are used to predict landslide disaster,and the prediction results of landslide susceptibility of different models are obtained.The accuracy of the two models is tested by ROC curve,and the AUC value of information model and random forest model is 0.81 and 0.94 respectively,which shows that the prediction ability of random forest is higher than that of information model.(3)According to the prediction results of random forest model,the landslide susceptibility is divided into five grades:the highest risk area,the high risk area,the medium risk area,the low risk area and the lowest risk area.The results show that:The high risk areas are mainly distributed in the low mountains and hills on both sides of the valley and near the town,accounting for 33.61%of the total area,and the landslide disaster point density is 15/100km~2;The medium risk areas are mainly distributed at the junction of high mountain area and low mountain area,accounting for 17.01%of the total area,and the disaster point density is 1/100 km~2;The low risk areas are mainly distributed in the middle and high mountains above 3000 m above sea level and towns and valleys below 2500 m above sea level,accounting for 49.39%of the total area.The density of geological hazard points is 1/100 km~2,and there is no landslide hazard point in the lowest risk area.(4)Based on the Gini impure,the importance of the landslide susceptibility factors involved in this paper is ranked,in which slope,elevation,vegetation,road,rock and soil type and rainfall are the main factors of landslide occurrence in the study area.On the whole,the terrain factors represented by slope and elevation play a decisive role in landslide susceptibility.Vegetation is also more important to the spatial prediction of landslide susceptibility because of the soil and water conservation capacity of vegetation.On this basis,it is the human activities represented by roads.That is to say,under the natural conditions of landslide formation,the destruction of topography and geology by human activities may aggravate the occurrence of landslide.