基于信息量和逻辑回归耦合模型的新疆S101公路山区路段滑坡危险性评价

    LANDSLIDE RISK ASSESSMENT OF S101 HIGHWAY MOUNTAIN SECTION IN XINJIANG BASED ON INFORMATION QUANTITY AND LOGISTIC REGRESSION COUPLING MODEL

    • 摘要: 新疆S101公路位于北天山西段低中山区,道路沿线地质情况复杂,岩性变化频繁,滑坡地质灾害频发。以S101公路山区路段线路两侧500 mm集水区为研究区域,通过对路段进行详细的现场调查,选取高程、坡度、坡向、与断层距离、与公路距离、与水系距离、地震动峰值加速度、地层岩性、年平均降水量和归一化植被指数(NDVI)10个因子,运用信息量模型、逻辑回归模型、信息量-逻辑回归耦合模型分别对路段进行了危险性评价。研究结果表明:(1)区内有利于滑坡发生的条件分别是:高程1.4~1.8 km;坡度20°~30°;坡向NE、N、NW;与断层距离0.4~0.8 km和6.4~12.8 km;与公路距离100 m内;与水系距离400 m内;地震动峰值加速度为0.30g;工程地质岩组为软弱和较坚硬;降雨量200~260 mm;NDVI范围0.05~0.35。(2)耦合模型中高和极高危险区滑坡点占比(83.62%)和AUC值(0.885)都要高于单独使用的信息量模型(79.51%和0.741)和逻辑回归模型(78.69%和0.875),耦合模型对该路段滑坡灾害危险性评价具有更高的精度。(3)以信息量-逻辑回归耦合模型危险性评价为最终结果,对新疆S101公路山区路段进行危险性分区,其中有107起滑坡发生在中、高、极高危险区,占比达到87.7%。(4)本研究可为后期新疆S101公路的养护、灾害点排查和防治以及品质提升等工作提供参考。

       

      Abstract: The Xinjiang S101 highway is located in the low and middle mountainous area of the western section of the northern Tianshan Mountains. The geological conditions along the road are complex, the lithology changes frequently, and landslide geological disasters occur frequently. Taking the 500 mm catchment area on both sides of the S101 highway mountain section as the research area, a detailed field investigation of the road section was conducted. Ten factors were selected, including elevation, slope, aspect, distance from fault, distance from highway, distance from water system, peak acceleration of ground motion, stratum lithology, annual average precipitation, and normalized difference vegetation index (NDVI). The information quantity model, the logistic regression model, and the information quantity-logistic regression coupling model were used to evaluate the risk of the road section. The results showed that: (1) The conditions conducive to landslide occurrence in the area are:elevation 1.4~1.8 km; slope 20°~30°; slope aspect NE,N,NW; distance from fault 0.4~0.8 km and 6.4~12.8 km; within 100 m from the road; within 400 m from the water system; peak ground motion acceleration 0.30 g; engineering geological rock group weak and hard; rainfall 200~260 mm; NDVI 0.05~0.35. (2) The proportion of landslide points in high and extremely high risk areas (83.62%) and the AUC value (0.885) of the coupling model were higher than those of the information model (79.51% and 0.741) and the logistic regression model (78.69% and 0.875) used alone. (3) Taking the risk assessment results of the information-logistic regression coupling model as the final outcome, the risk zoning of the S101 highway mountain section in Xinjiang was carried out. Among them, 107 landslides occurred in medium, high and extremely high risk areas, accounting for 87.7%. (4) This study can provide a reference for the maintenance, disaster point investigation and prevention, and quality improvement of the S101 highway in Xinjiang in the later period.

       

    /

    返回文章
    返回