多源遥感数据融合调查的复杂山区滑坡易发性评价方法研究

STUDY ON THE EVALUATION METHOD OF LANDSLIDE SUSCEPTIBILITY IN COMPLEX MOUNTAINOUS AREAS BASED ON MULTI-SOURCE REMOTE SENSING DATA FUSION SURVEY

我国山区地质灾害频发,其中滑坡占比超过1/3,但传统的、单一的技术手段对于高山地、高植被覆盖的复杂山区滑坡调查来说显得力不从心,这极大影响了滑坡易发性结果的可靠性。本文选取九寨沟部分区域作为研究区,将光学遥感、合成孔径雷达(InSAR)、机载LiDAR等技术融合应用于滑坡的调查,对研究区典型滑坡进行分类,编录了滑坡样本库,筛选了地形地貌、地质条件、水文条件、人类工程活动4类等作为易发性因子,并引入Maxent模型进行滑坡易发性评价研究,它结合了机器学习和统计模型被认为是一种强大的建模方法。应用受试者工作特征曲线下面积(ROC-AUC)对模拟结果进行检验,结果显示平均ROC-AUC值达到0.855,预测效果优秀,Maxent模型适用于滑坡易发性评价。该研究对山区防灾减灾具有支持意义。

Geological disasters in China's mountainous areas are frequent,of which landslides account for more than one-third,but the traditional,single technical means are not sufficient for the investigation of complex mountain landslides with high vegetation cover,which greatly affects the reliability of landslide susceptibility results.In this paper,some areas of Jiuzhaigou are selected as the research area,and optical remote sensing,synthetic aperture radar(InSAR),airborne LiDAR and other technologies are integrated into the investigation of landslides,the typical landslides in the research area are classified,the landslide sample bank is cataloged,the topographic landform,geological conditions,hydrological conditions,and human engineering activities are screened as the susceptibility factors,and the Maxent model is introduced to evaluate the susceptibility of landslides,which combines machine learning and statistical models to be considered a powerful modeling method.The simulated results were tested by applying the area under the working characteristic curve(ROC-AUC)of the subjects,and the results showed that the ROC-AUC value reached 0.851,and the prediction effect was excellent,and the Maxent model was suitable for landslide susceptibility evaluation.The study is of supportive significance for disaster prevention and mitigation in mountainous areas.