人民珠江2025，Vol.46Issue(6)：52-59,8.DOI:10.3969/j.issn.1001-9235.2025.06.006
基于历史排位降雨阈值的粤港澳大湾区滑坡危险性预警
Landslide Hazard Warning in Guangdong-Hong Kong-Macao Greater Bay Area Based on Historical Ranking Rainfall Threshold
摘要
Abstract
This study focused on the Guangdong-Hong Kong-Macao Greater Bay Area and constructed a grid-based landslide hazard assessment model to enhance regional disaster prevention and mitigation capabilities.A semi-supervised learning method was used to optimize the proportional selection of landslide points and non-landslide points to reduce the uncertainty of susceptibility modeling.A historical ranking rainfall threshold-based method was proposed to classify daily rainfall,3-day cumulative rainfall,and 7-day cumulative rainfall data.The spatial susceptibility of landslides and rainfall-induced probability were quantitatively coupled to establish a dynamic landslide hazard warning system.The results indicate that when a 12.5-meter evaluation unit scale is used within the Guangdong-Hong Kong-Macao Greater Bay Area,the optimal ratio of landslide points to non-landslide points is 1:4.Furthermore,the area under curve(AUC)value of the susceptibility model reaches as high as 0.973.In practical application during June 2018,the warning system accurately predicted 25 rainfall-induced landslide events,with 72%occurring in extremely high-risk warning zones and 28%in high-risk warning zones,validating the model's effectiveness.This system achieves fine-scale landslide hazard warnings in the Greater Bay Area,providing scientific support for regional landslide risk management.关键词
半监督机器学习/非滑坡样本/降雨阈值/滑坡危险性/粤港澳大湾区Key words
semi-supervised machine learning/non-landslide sample/rainfall threshold/landslide hazard/Guangdong-Hong Kong-Macao Greater Bay Area分类
水利科学引用本文复制引用
金毅,于海霞..基于历史排位降雨阈值的粤港澳大湾区滑坡危险性预警[J].人民珠江,2025,46(6):52-59,8.基金项目
广东省基础与应用基础研究基金项目(2025A1515011898) （2025A1515011898）
国家重点研发计划(2021YFC3001000) （2021YFC3001000）
