水利水电技术（中英文）2026，Vol.57Issue(3)：91-106,16.DOI:10.13928/j.cnki.wrahe.2026.03.007
集成学习算法驱动的黄河流域干旱监测与评估
Drought monitoring and assessment in Yellow River Basin driven by ensemble learning algorithms
摘要
Abstract
[Objective]To address the intensifying drought under global climate warming and the limitations of existing drought monitoring models that consider overly simplistic factors,a Comprehensive Drought Monitoring Model based on Ensemble Learning Algorithms(CDMMMLEA)for the Yellow River Basin is proposed.[Methods]This model integrates multi-source remote sensing data,comprehensively considers the impacts of crop canopy temperature,crop morphology,vegetation greenness dynamics,soil moisture fluctuations,and crop canopy water status on drought monitoring,and accurately characterizes the spatiotemporal evolution of drought in the Yellow River Basin over the past 20 years.[Results]The result showed that:(1)in drought monitoring of the Yellow River Basin,CDMMMLEA outperformed other models,particularly in the upper reaches,with an average correlation coefficient of 0.46 and a root mean square error as low as 0.81.(2)Compared with the three-month-scale Standardized Precipitation Evapotranspiration Index(SPEI03),demonstrated significant advantages in monitoring drought events in 2002 and 2010.It accurately identified the boundaries of moderate drought areas in the upper reaches and improved the continuity of drought extent.In the middle reaches,it effectively captured abrupt changes in drought intensity and complex fluctuation characteristics.In the lower reaches,it provided smoother transitions in drought severity and significantly enhanced spatial continuity.(3)revealed the seasonal variation characteristics of drought in the Yellow River Basin from 2001 to 2023.In spring,35.2％of the upper reaches showed humidification,while 28.7％of the middle reaches experienced intensified drought.In summer,the Loess Plateau in the middle reaches showed a pattern of"slow in the east and rapid in the west".In autumn,drought frequency increased by 15％in the middle reaches and locally expanded by 15％in the lower reaches.During the growing season,drought intensified across the entire river basin,with the middle reaches most severely affected.Temporally,after 2011,drought intensity across the river basin decreased by 25％,but in the lower reaches,drought duration extended from 2.55 per event to 2.32 months per event.The model effectively captured the spatiotemporal variability of drought across the river basin.[Conclusion]The findings provide scientific method ological support for accurate regional drought monitoring and serve as a basis for optimizing decision-making on drought prevention and mitigation strategies.关键词
集成学习算法/综合干旱监测模型/多源遥感数据/时空变化特征/黄河流域/影响因素Key words
ensemble learning algorithm/comprehensive drought monitoring model/multi-source remote sensing data/spatiotemporal variation characteristics/Yellow River Basin/influencing factors分类
信息技术与安全科学引用本文复制引用
王春晨,马梓策,孙鹏,陈冬花,王玉亮..集成学习算法驱动的黄河流域干旱监测与评估[J].水利水电技术（中英文）,2026,57(3):91-106,16.基金项目
国家自然科学基金项目(42271037) （42271037）
安徽省高等学校科学研究项目(自然科学类)(2024AH051428) （自然科学类）
滁州学院科研启动资金项目(2024qd13) （2024qd13）
滁州学院大学生创新创业训练计划资助项目(2025CXXL030) （2025CXXL030）
