基于多源遥感的闽江流域东溪水库蓄水量变化研究
Study on Reservoir Storage Change of Dongxi Reservoir in the Minjiang River Basin Based on Multi-Source Remote Sensing
何方棱 1冉林巧 1邓海军2
作者信息
- 1. 福建师范大学湿润亚热带山地生态国家重点实验室培育基地，福建福州，35007;福建师范大学地理科学学院，福建福州，35007
- 2. 福建师范大学湿润亚热带山地生态国家重点实验室培育基地，福建福州，35007;福建师范大学地理科学学院，福建福州，35007;福建省陆地灾害监测评估工程技术研究中心, 福建福州 350007
- 折叠
摘要
为解决河流型水库遥感数据缺失，动态监测难度大的问题，以东溪水库为研究对象，基于多源卫星数据（Jason-2/3、SWOT、Sentinel 系列）与机器学习方法，开展水面面积提取、水位构建、蓄水量反演及驱动机制分析。首先采用随机森林模型提取水面面积，再通过线性拟合构建水位-面积关系，结合水位-面积-库容模型反演蓄水量，最后耦合降水与调度数据解析水位变化驱动机制。结果表明：随机森林模型水面提取总体精度达93.315%（Kappa=0.818），显著优于传统指数法；多源卫星协同融合解决了观测间断问题，线性拟合在核心监测区间（255-265m）的RMSE仅1.9 m，且水位呈现"汛期降水主导（r=0.89）、非汛期人为调控（250-255 m平稳区间）"的二元驱动特征；蓄水量反演模型精度高，蓄水量年际波动0-120Mm3，呈汛期峰值（7-9月）、非汛期谷值（1-3月）规律。研究构建的多源数据融合-模型拟合 -要素反演的技术流程，为闽江流域河流型水库蓄水量反演提供参考方案。Abstract
To address the issues of missing remote sensing data and the difficulty in dynamic monitoring of river-type reservoirs, this study targeted the Dongxi Reservoir. Based on multi-source satellite data (Jason-2/3, SWOT, Sentinel series) and machine learning methods, it conducted water surface area extraction, water level construction, storage capacity inversion, and driving mechanism analysis. First, the Random Forest model was used to extract the water surface area. Then, linear fitting was applied to construct the water level-area relationship, which was combined with the water level-area-capacity model to invert the storage capacity. Finally, precipitation and operation data were coupled to analyze the driving mechanism of water level changes.The results show that: the overall accuracy of water surface extraction by the Random Forest model reached 93.315% (Kappa=0.818), significantly better than traditional index methods; the synergistic fusion of multi-source satellites solved the problem of observational discontinuity; the RMSE of linear fitting in the core monitoring range (255-265m) was only 1.9 m; and the water level showed a dual-driver characteristic of "precipitation-dominated in the flood season (r=0.89), human regulation-dominated in the non-flood season (with a stable range of 250-255 m)". The storage capacity inversion model achieved high accuracy, with an annual fluctuation of 0-120 Mm3, showing a pattern of peak storage during the flood season (July-September) and trough during the non-flood season (January-March).The technical flow of multi-source data fusion - model fitting - factor inversion constructed by the research provides a reference scheme for reservoir storage inversion of river-type reservoirs in the Minjiang River Basin. 关键词
闽江流域/东溪水库/水位/蓄水量/多源遥感Key words
Minjiang River Basin/Dongxi Reservoir/water level/water storage/Multi-source Remote Sensing引用本文复制引用
何方棱,冉林巧,邓海军.基于多源遥感的闽江流域东溪水库蓄水量变化研究[EB/OL].(2025-11-26)[2025-11-28].http://www.paper.edu.cn/releasepaper/content/202511-25.学科分类
水利枢纽、水工建筑物/遥感技术/计算技术、计算机技术
评论