Landslide Development Characteristics And Hazard Assessment Of Mianning-xichang Section Of Anning River

The section from Mianning to Xichang of Anning River is located in the southeastern margin of the Tibet Plateau.In addition to the Anning Valley Plain,the rest of the terrain is cut strongly,the canyon is vertical and horizontal,and the terrain is undulating.Because of the complex geological and environmental conditions,landslide disasters occur frequently in this area,which has a great impact on the safety of people's lives and property,infrastructure construction and operation(such as highways,power stations,reservoirs,etc.)and even the ecological environment.Based on the analysis of regional geological and environmental conditions in the section from Mianning to Xichang of Anning River,the development characteristics and controlling factors of landslides in this area are analyzed in detail.The information entropy method is introduced to evaluate the hazard of 415 landslides in the study area.The results of single landslides are applied to the regional landslide hazard assessment,and the shortcomings of the traditional regional landslide hazard assessment methods are pointed out.The method of XGBoost machine learning is introduced for the first time to complete the regional landslide hazard assessment,and the following understandings are obtained:(1)According to the methods of remote sensing interpretation and field investigation and verification,415 landslides have been identified in the study area,and the development density is 1.76/10km~2.The landslide mainly develops along the east branch fault zone of Anning River and Hongmo fault.The distribution of landslides on both sides of Anning River is very uneven,and the development density of landslides on the left bank is 3.11 times higher than that on the right bank.The development degree of landslides is obviously different in all directions of slopes,among which there are more landslides in four directions:southeast,southwest,south and west.in addition,landslides are mainly developed in slopes with a slope of 15°~35°.The development density is the highest in the area with terrain fluctuation degree of 100~150m,and the profile shape of most landslides is convex.(2)Stratigraphic lithology has obvious control effect on landslide development.According to the number of landslides,64.34%of the landslides in the study area are developed in the Upper Triassic,Jurassic and Lower Sinian strata,and the most slippery strata in the area with known density of landslides are the Lower Sinian strata.The development density of landslides in each slope structure type has obvious differentiation,which indicates that the slope structure type has a controlling effect on the distribution of landslides,and landslides are prone to occur in the along slope,the loose structure slope and the inclined outward slope.(3)The development of landslides in the area is strongly affected by tectonic activities.According to the characteristics of spatial distribution,there are two obvious concentrated distribution zones of landslides in the area,in which the concentrated zone A is located on both sides of the fault zone of the east branch of Anning River.The concentrated zone B is located on the northwest side of the Hongmo fault in the Reshuihe River.The landslide development density in the two concentrated zones is about 8.70/10km2,which is about 5 times of the landslide development density in the region.The river system and road engineering activities also play a controlling role in the development of landslides in the study area.(4)Taking three groups of landslides in Luonaige Village,Heisha River right bank landslide and Yizi Amu landslide as examples,the risk qualitative assessment of landslides investigated in the field in the analysis area was carried out.Considering the factors affecting landslide hazard,under the guidance of the principles of systematicness and independence,the hazard assessment index of single landslide is established.The information entropy method is introduced to objectively assign weight to each index,and the hazard value H of 415 landslides is calculated.According to the equidistant grading method,the hazard of single landslide is divided into three grades:low hazard(H<0.333),medium hazard(0.333<H<0.667)and high hazard(H>0.667).The results of landslide hazard assessment based on information entropy method are consistent with those landslides investigated in the field,which indicates that the results of single landslide hazard assessment are more reliable.(5)Based on the evaluation of the hazard of single landslide,the hazard assessment results of single landslide were applied to the regional landslide hazard assessment.The XGBoost machine learning method was introduced for the first time to evaluate the regional landslide hazard.The learning rate(learning_rate),the number of iterative steps(n_estimators),the maximum depth(max_depth)and the minimum sample weight sum(min_child_weight)of the regression tree in XGBoost model are adjusted.The prediction accuracy of the training machine learning model is as high as 87.74%,and the training results are applied to the whole region.The evaluation results have high accuracy.