基于栅格最大值法的县域综合地质灾害建模

四川省地形高低悬殊, 岩性构造发育, 各类地质灾害频发, 开展地质灾害易发性评价具有重要意义。崩塌、泥石流属于广义上的滑坡, 以四川省丹巴县为例, 从考虑不同滑坡类别的区域性地质灾害易发性出发综合考虑崩塌、滑坡、泥石流的空间概率分布。基于ArcGIS通过高精度数字高程模型共选取高程、坡度等10个地质灾害关键控制因素, 采用信息量模型对综合地质灾害进行了易发性评价。最终通过ArcGIS的单元统计(Cell Statistics)功能实现多个栅格图层最大值法合成综合易发性, 进一步利用受试者工作特征曲线(ROC)验证单种滑坡类别易发性模型的精度。按照自然断点法将研究区划分为极低、低、中、高、极高易发区, 高易发区和极高易发区主要集中分布在章谷镇、太平桥乡以及甲居镇等地。研究结果证明信息量模型能对单类地质灾害进行评价, 栅格最大值法是获取综合易发性的一种有效评价方法。 

County comprehensive geohazard modelling based on the grid maximum method

Sichuan Province is characterized by great differences in topography, lithologic structure and frequent occurrence of various local disasters. Therefore, it is of great significance to carry out evaluations of the vulnerability of geological disasters. Rockfall and debris flows are landslides in a broad sense. Taking Danba County, Sichuan Province, as a case study, the spatial probability distributions of collapse, landslide and debris flow are comprehensively considered from the perspective of the susceptibility of different types of landslides to regional geological disasters. Based on ArcGIS, 10 key control factors of geological hazards, such as elevation and slope, were selected by a high-precision digital elevation model, and the susceptibility of comprehensive geological hazards was evaluated by an information content model. Finally, the Cell Statistics function of ArcGIS was used to realize the synthesis and comprehensive vulnerability of the maximum value method of multiple raster layers, and the ROC curve was further used to verify the accuracy of the vulnerability model of landslide categories in a single area. According to the natural break point method, the very low-, low-, medium-, high- and very high-prone areas were divided, and the high- and very high-prone areas were mainly concentrated in Zhanggu Town, Taipingqiao Township and Jiaju Town.This paper shows that the information model can evaluate a single type of geological hazard and that the grid maximum method is an effective evaluation method to obtain the comprehensive vulnerability.