为提高滑坡易发性评价的准确性和精度,结合随机森林(FR)模型和信息量(IV)模型2种方法,提出一种新的加权信息量(WIV)模型。以西安市临潼区为研究区,基于野外调查和因素相关性分析,选取坡度、坡向、高程、曲率、地形起伏、岩土体类型、降雨量、断层、水系和道路10个影响因子,以全区72个滑坡灾害点为样本数据,分别采用FR模型、IV模型和WIV模型对研究区开展滑坡易发性评价。结果表明:WIV模型训练样本的成功率和预测率比FR模型分别高4.9%和1.9%,比IV模型分别高17.80%和4.70%;滑坡灾害点高易发区和极高易发区多分布在断层、河流及道路集中区域,且其滑坡灾害点比例WIV模型比IV模型、FR模型分别提高6.37%和4.45%;WIV模型易发性评价分区结果与研究区实际情况相近。研究成果可为防灾减灾工程提供科学依据。

A weighted information value (WIV) model integrating random forest (FR) model and information value (IV) model is proposed in an attempt to improve the precision and accuracy of landslide susceptibility assessment. Lintong District of Xi'an City is taken as the research area. According to field survey and factor correlation analysis, ten factors including slope gradient, slope aspect, elevation, curvature, topographic relief, rock and soil type, rainfall, fault, water system, and road are selected as influence factors for the assessment. With 72 landslide hazard points as sample data, the landslide susceptibility is assessed using RF model, IV model and WIV model, respectively. Results demonstrate that the success rate and prediction rate of WIV model training samples are higher than those of FR model by 4.90% and 1.90%, respectively, and higher than those of IV model by 7.80% and 4.70%, respectively. Highly and extremely highly susceptible areas are mainly distributed in faults, rivers and roads. The rate of landslide hazard points in highly and extremely highly susceptible areas by WIV model is 6.37% and 4.44% higher than that by IV model and FR model, respectively. In conclusion, the results of WIV model are consistent with the actual situation of the research area.