区域群发性滑坡危险性评价的累计降雨与短时雨强特征研究OA
Research on the cumulative rainfall and short-term rain intensity characteristics of regional cluster-type landslide hazard assessment:A case study of Zixing,Hunan Province
2024年 7月台风"格美"在湖南资兴市引发极端强降雨,触发大量滑坡灾害.准确评价降雨群发滑坡危险性,对区域防灾减灾与城市规划具有重要意义.文章以该次群发滑坡事件为研究对象,探讨累计降雨与短时雨强特征在滑坡危险性评价中的作用.基于高质量滑坡清单、降雨数据以及地形地貌等环境因子,构建了考虑累计降雨与短时雨强特征的群发滑坡危险性评价指标体系,采用随机森林模型开展滑坡空间危险性评价,并结合特征重要性与沙普利加和解释(shapley additive explanations,SHAP)方法分析各因子的贡献特征.结果表明:随机森林模型在训练集与测试集中具有较好的区分能力,曲线下面积(area under curve,AUC)分别为 0.923和 0.854;研究区滑坡危险性具有明显空间分异特征,中低危险区主要分布于研究区外围,较高及极高危险区主要集中于研究区中部,面积占研究区总面积的 15.2%,极高危险区滑坡密度达到 99.53个/km2;不同乡镇滑坡危险性分布存在差异,其中八面山瑶族乡西部为滑坡高危险区域集中分布区.特征重要性及 SHAP分析结果表明,最大小时雨强与累计降雨量是影响滑坡发生的主导因子,而地形地貌条件则提供滑坡发生的基底孕灾环境.累计降雨量与短时雨强特征能够较好表征极端降雨事件对群发滑坡的触发作用,随机森林模型能够有效识别滑坡高危险区域.研究结果可为强降雨条件下群发滑坡危险性识别与地质灾害防控提供参考.
In July 2024,Typhoon Gaemi triggered extreme rainfall in Zixing City,Hunan Province,China,causing numerous landslides.Accurate assessment of rainfall-induced clustered landslide hazards is important for regional disaster prevention,mitigation,and urban planning.Using this clustered landslide event as a case study,this study investigates the role of cumulative rainfall and short-term rainfall intensity in landslide hazard assessment.Based on a high-quality landslide inventory,rainfall data,and topographic and geological environmental factors,an assessment index system was established by incorporating cumulative rainfall and maximum hourly rainfall intensity.A random forest model was used for landslide hazard assessment,and feature-importance analysis and shapley additive explanations(SHAP)were used to quantify and interpret factor contributions.The random forest model showed good discriminative ability,with AUC values of 0.923 and 0.854 for the training and testing datasets,respectively.Landslide hazard in the study area exhibited significant spatial heterogeneity.Low and moderate hazard zones were mainly distributed around the periphery of the study area,whereas high and very high hazard zones were concentrated in the central part and accounted for 15.2%of the total area.Landslide density in the very high hazard zone reached 99.53 landslides/km2.Hazard distribution differed markedly among towns,with the western part of Bamianshan Yaozu Township identified as the most concentrated high-hazard area.Feature-importance and SHAP analyses showed that maximum hourly rainfall intensity and cumulative rainfall were the dominant controls on landslide occurrence,while topographic and geomorphological conditions provided the fundamental environmental setting.Cumulative rainfall and short-term rainfall intensity effectively characterize the triggering effects of extreme rainfall events on clustered landslides,and the random forest model can identify high-hazard landslide areas.The results provide a useful reference for landslide hazard identification and geohazard prevention under intense rainfall conditions.
李涛;刘毅;赵斌滨;郭威;孔小昂;刘汉星;王秀龙;黄远东;许冲;陈兆宁;梁家荣;金嘉乐;高诗琪
应急管理部国家自然灾害防治研究院,北京 100085||复合链生自然灾害动力学应急管理部重点实验室,北京 100085国网电力工程研究院有限公司,北京 100069国网电力工程研究院有限公司,北京 100069国网天津市电力公司电力科学研究院,天津 300010国网电力工程研究院有限公司,北京 100069应急管理部国家自然灾害防治研究院,北京 100085||复合链生自然灾害动力学应急管理部重点实验室,北京 100085国家电网有限公司,北京 100031应急管理部国家自然灾害防治研究院,北京 100085||复合链生自然灾害动力学应急管理部重点实验室,北京 100085应急管理部国家自然灾害防治研究院,北京 100085||复合链生自然灾害动力学应急管理部重点实验室,北京 100085应急管理部国家自然灾害防治研究院,北京 100085||复合链生自然灾害动力学应急管理部重点实验室,北京 100085应急管理部国家自然灾害防治研究院,北京 100085||复合链生自然灾害动力学应急管理部重点实验室,北京 100085应急管理部国家自然灾害防治研究院,北京 100085||复合链生自然灾害动力学应急管理部重点实验室,北京 100085应急管理部国家自然灾害防治研究院,北京 100085||复合链生自然灾害动力学应急管理部重点实验室,北京 100085
天文与地球科学
降雨滑坡危险性评价累计降雨量短时雨强随机森林资兴市
rainfall-induced landslideslandslide hazard assessmentcumulative rainfallshort-term rainfall intensityrandom forestZixing City
《中国地质灾害与防治学报》 2026 (4)
41-53,13
国家电网公司总部科技项目(5500-202455159A-1-1-ZN)
评论