基于信息量模型和数据标准化的滑坡易发性评价

本文以北川曲山-擂鼓片区为研究区,将坡度、坡向、高程、地层、距断层的距离、距水系的距离和距道路的距离作为该区域滑坡易发性评价因子。采用信息量模型计算了各项评价因子的信息量值,并运用4种标准化模型对信息量值进行标准化处理。各评价因子的权重由层次分析法（AHP）确定。在GIS中将权重值和各评价因子的标准化信息量值,进行叠加计算得到区域滑坡总信息量值,并基于自然断点法对其进行重分类,将研究区划分为极高易发区、高易发区、中易发区、低易发区和极低易发区5级易发区。将基于4种标准化模型和信息量模型得到的滑坡易发性评价结果进行了对比分析,结果表明：基于最值标准化信息量模型的滑坡易发性评价结果的ROC曲线下面积AUC值为0.807,高于其余模型的AUC值,说明最值标准化信息量模型的滑坡易发性评价效果最好。极高易发区面积占研究区面积的20.03%,离断层和水系较近,主要分布地层为寒武系、志留系和三迭系。研究结果可为区内滑坡风险评价和灾害防治提供参考。

Assessment of landslide Susceptibility Based on Information Quantity Model and Data Normalization

This paper used an information quantity model and four normalized models to assess earthquake-induced shallow landslide susceptibility in a geographic information system environment. Four normalized models are Min-Max normalization, the zero-mean normalization, the logarithmic logistic normalization and the arc-tangent function normalization. The approach was applied to the Qushan-Leigu area in the Beichuan County, Sichuan Province, where many co-seismic landslides were triggered by the Wenchuan earthquake in May 2008. Seven impact factors, the slope angle, slope aspect, geology, elevation, distance to faults, distance to rivers and distance to roads, were selected as the most important conditioning factors. To assess the shallow landslide susceptibility, a spatial database of conditioning factors and a landslide inventory map were compiled in ArcGIS using data from a topographic map, a geological map and Spot-5 imageries. The information quantity values of the conditioning factors were computed and normalized based on four normalized models. The weighted values of assessment factors were determined using the analytic hierarchy process. Five landslide susceptibility maps were developed. The performance was evaluated using the receiver operating characteristic (ROC) curve. The values of the area under the curve (AUC) for four normalized models and the information quantity model are 0.807, 0.672, 0.592, 0.615 and 0.684, respectively. A comparison of the results of the landslide susceptibility assessment and the landslide inventory indicates that the Min-Max normalization had the highest predictive performance (AUC=0.807). The landslide susceptibility index values were reclassified into five susceptibility classes using the Natural Breaks approach, namely very high, high, medium, low and very low susceptibility. The results show that the very high susceptibility zone obtained using the Min-Max normalization covers about 20% of the study area. Landslides distributed in the very high susceptibility zone are close to faults and rivers. The final landslide susceptibility map has the potential to the regional landslide risk assessment and geohazard mitigation.