三峡库区地质灾害趋势预测模型构建与预警系统开发  

Construction of Geological Hazard Trend Prediction Model and Early Warning System in Three Gorges Reservoir Area

作　　者：吴润泽 朱敏毅 付小林 梅红波[2] 李振华[3] 张亮 王红群 涂秋月 WU Run-Ze;ZHU Min-Yi;FU Xiao-Lin;MEI Hong-Bo;LI Zhen-Hua;ZHANG Liang;WANG Hong-Qun;TU Qiu-Yue(Wuhan Center,China Geological Survey(Geosciences Innovation Center of Central South China),Wuhan 430205,Hubei,China;School of Earth Resources,China University of Geosciences,Wuhan 430074,Hubei,China;School of Computer,China University of Geosciences,Wuhan 430074,Hubei,China)

机构地区：[1]中国地质调查局武汉地质调查中心(中南地质科技创新中心),湖北武汉430205 [2]中国地质大学资源学院,湖北武汉430074 [3]中国地质大学计算机学院,湖北武汉430074

出　　处：《华南地质》2023年第3期455-469,共15页South China Geology

基　　金：水利部三峡库区后续地质灾害防治信息系统建设(No.0001212012AC50001)。

摘　　要：地质灾害预先发现、规模预测、受灾损失预测是地质灾害防治信息化建设的核心任务之一。尽管三峡库区地质灾害监测数据丰富,但此前并未建立全面、智能的地质灾害趋势预测体系。本文通过广泛收集三峡库区地质环境数据,选取了地形、地质、植被、气象水文、人类活动等方面的13个因子,通过训练多种机器学习模型,对库区地质灾害进行易发性评价,并根据降雨实况和预报数据的粒度构建短中期和长期趋势预测模型。在此基础上,开发了基于WebGIS的三峡库区地质灾害趋势预测预警系统。系统功能丰富,包括数据管理、地质灾害易发性评价、短中期和长期多尺度趋势预测等功能模块。系统首次实现了与传感网的直接对接,使传感数据能实时参与计算,保证了预测结果的及时性。通过历史灾、险情回代验证,建立的短中期和长期趋势预测模型均具有较好的预测效果,能为三峡库区地质灾害防灾减灾提供决策支持。Early detection,disaster scale prediction,and loss assessment are core tasks of geological hazard prevention and control information systems.Although abundant geological disaster monitoring data exists for the Three Gorges Reservoir area,a comprehensive and intelligent geological hazard trend forecasting system has yet to be established.This study extensively compiled Three Gorges Reservoir geological environment da-ta to select 13 factors related to topography,geology,vegetation,meteorological hydrology,and anthropogen-ic activities,and assessed geological hazard susceptibility with various machine learning models trained in the reservoir area.Then short-term,medium-term,and long-term trend forecasting models were constructed based on the granularity of actual and forecasted precipitation data.Besides that,a Three Gorges Reservoir geological hazard trend forecasting and early warning system was developed on a WebGIS platform,which contains diverse functional modules including data management,susceptibility assessments,and multi-scale short-term to long-term trend forecasts.The system pioneers direct integration with sensor networks,enabling real-time calculation using streaming sensor data for timely forecast updates.Historical disaster case backtest-ing indicates established short-term and long-term trend forecasting models demonstrate favorable forecast-ing capabilities,providing decision support for Three Gorges Reservoir geological hazard prevention and mit-igation.

关 键 词：三峡库区 地质灾害 机器学习 趋势预测 预警系统 

分 类 号：P694[天文地球—地质学]