军事知识服务平台
基于对象化水文响应单元的流域日径流模拟方法研究
Research on a Watershed Daily Runoff Simulation Method Based on Object-Based Hydrological Response Units
摘要 传统分布式水文模型通常采用规则网格来对流域加以剖分,破坏了地物边界完整性且缺乏物理连接.为解决这一问题,以真实地物实体边界为对象化水文响应单元(OHRU),并借助知识图谱技术构建单元之间的水文级联规则,以增强流域水文过程的表达能力.选取洣水上游流域作为研究区域,对其展开日径流模拟方法的验证工作.结果表明,在引入对象化水文单元之后,其纳什效率系数(NSE)由0.74提升至0.85,均方根误差(RMSE)由19.08 m3/s降至14.25 m3/s;与知识图谱相耦合之后的OHRU-SWAT模型拟合效果最好,NSE为0.93,决定系数(R²)为0.95,RMSE缩减至9.31 m3/s.以真实地理实体为基础单元,构建知识图谱级联规则,能更准确地捕捉产流过程,显著提升地表径流模拟结果的可靠性和可解释性.
Abstract:
Traditional distributed hydrological models typically partition a basin using regular grid cells,which disrupt the integ‑rity of geographic feature boundaries and fail to preserve physical connectivity among landscape elements.To address this limita‑tion,Object-oriented Hydrological Response Units(OHRUs)were delineated based on real-world geographic entity boundaries,and hydrological cascading rules among units were established using knowledge graph technology to enhance the representation of basin hydrological processes.The upper Mi River Basin was selected as the study area for validating the proposed daily runoff simula‑tion framework.The results indicate that,after introducing object-oriented hydrological units,the Nash–Sutcliffe Efficiency(NSE)increased from 0.74 to 0.85,while the Root Mean Square Error(RMSE)decreased from 19.08 m³/s to 14.25 m³/s.Furthermore,the OHRU-SWAT model coupled with a knowledge graph achieved the best performance,yielding an NSE of 0.93,an R² of 0.95,and a further reduction in RMSE to 9.31 m³/s.By using real geographic entities as the fundamental computational units and con‑structing knowledge graph-based cascading rules,the proposed approach can more accurately capture runoff generation processes and significantly improve the reliability and interpretability of surface runoff simulations.
刊 名 水文 ISTICPKU
年,卷(期): 2026, 46(5)
分类号: P344 TV121.2
机标分类号: P333 S271 P208
在线出版日期: 2026年11月9日
基金项目: E4Z202021F:中国科学院空天信息创新研究院自主部署项目202605AF350036:云南省科技人才与平台计划项目42471394:国家自然科学基金,