考虑日降雨强度的滑坡双指标降雨预警模型研究  

A Dual-Indicator Rainfall Model for Landslide Early Warning Based on Daily Rainfall Intensity

作　　者：王家柱 铁永波 徐伟 白永健 张鸣之 WANG Jiazhu;TIE Yongbo;XU Wei;BAI Yongjian;ZHANG Mingzhi(Chengdu Center,China Geological Survey,Chengdu,610218;Engineering Technology Innovation Center of Geological Disaster Risk Prevention and Control,Ministry of Natural Resources,Institute of Prospecting Technology,Chinese Academy of Geological Sciences,Chengdu,611734;Chengdu Geological Hazard Field Scientific Observation and Research Station,Ministry of Natural Resources,Chengdu Institute of Geo-environment Monitoring,Chengdu,610000;Engineering Technology Innovation Center of Geological Hazards Intelligent Monitoring and Risk Early Warning,Ministry of Natural Resources,China Institute of Geo-environment Monitoring,Beijing 100081)

机构地区：[1]中国地质调查局成都地质调查中心(西南地质科技创新中心),成都610218 [2]中国地质科学院探矿工艺研究所自然资源部地质灾害风险防控工程技术创新中心,成都611734 [3]成都市地质环境监测站自然资源部成都地质灾害野外科学观测研究站,成都610000 [4]中国地质环境监测院(自然资源部地质灾害技术指导中心)自然资源部地质灾害智能监测与风险预警工程技术创新中心,北京100081

出　　处：《山地学报》2025年第1期157-166,共10页Mountain Research

基　　金：自然资源部地质灾害智能监测与风险预警工程技术创新中心项目(TICGM-2023-09);中国地质调查局成都地质调查中心“刘宝珺院士基金”;四川省自然科学面上基金(2023NSFSC2086)。

摘　　要：基于降雨特征的滑坡预警模型构建是地质灾害防控体系建设的关键环节,其模型精度提升对地质灾害风险管控具有重要实践意义。传统单指标I(Intensity)-D(Duration)模型,多基于平均降雨强度(I)-历时(D)构建滑坡事件与降雨参数的统计关系确定临界阈值,但弱化了短时强降雨事件的触发作用。本研究以凉山州喜德县94个典型滑坡案例及其灾前降雨序列为研究样本,通过最小二乘回归建立降雨强度-历时(I-D)基础模型,在此基础上耦合当日降雨强度(D值为1)参数,创新性构建双指标协同预警模型。研究揭示:(1)喜德县滑坡降雨触发机制可划分为先雨后滑型、暴雨致滑型及连雨致滑型三类典型模式。(2)基于74个训练样本建立的I-D基础模型,其预警响应在暴雨型滑坡中呈现明显滞后性,通过引入当日降雨强度指标并采用双指标极值判别准则,验证集20个案例的红色预警占比提升至40%;(3)模型验证表明,双指标体系对先雨后滑型保持I-D模型的识别精度,而对暴雨致滑型的预警时效性提升2 d。研究成果不仅完善了降雨阈值理论体系,更为滑坡灾害防治提供了新的技术路径。The construction of a landslide warning model based on rainfall characteristics constitutes a critical component of a geological disaster prevention and control system,with model accuracy enhancement holding significant practical implications for geological risk management.Traditional single-indicator I-D models,predominantly established through statistical relationships between average rainfall intensity(I)and duration(D)to determine critical thresholds,often overlook the triggering effects of short-duration rainstorm events.In this study,94 typical landslide occurrences and related pre-failure rainfall sequences in Xide County,Liangshan Prefecture,Sichuan Province,China,were used as research samples.A basic rainfall intensity-duration(I-D)model was established using least squares regression,and on this basis,the daily rainfall intensity(D value is 1)parameter was coupled to innovatively construct a dual-indicator collaborative warning model.(1)The rainfall-triggered landslide mechanisms in Xide County were classified into three typical patterns:antecedent rainfall-induced,rainstorm-induced slide,and prolonged rainfall-triggered slide.(2)The I-D basic model established based on 74 training samples showed a significant lag in early warning response for rainstorm-induced landslides.By introducing the daily rainfall intensity indicator and adopting dual-indicator extreme value discrimination criteria,the proportion of red warnings in the validation set of 20 cases increased to 40%.(3)By model validation tests,it confirmed that the dual-indicator system maintained I-D model recognition accuracy for antecedent rainfall-induced landslides,while the timeliness of warnings for rainstorm-induced landslides was improved by 2 days.The research results not only improve the theoretical system of rainfall thresholds but also provide a new technical path for the prevention and control of landslide disasters.

关 键 词：降雨预警 I-D模型 双指标 滑坡 雨型 

分 类 号：P642[天文地球—工程地质学]