| 引用本文: | 俞佳固,刘思果,杨 可,等.长江流域气象干旱演变规律及其成因研究[J].灌溉排水学报,2026,45(3):120-132. |
| YU Jiagu,LIU Siguo,YANG Ke,et al.长江流域气象干旱演变规律及其成因研究[J].灌溉排水学报,2026,45(3):120-132. |
|
| 摘要: |
| 【目的】阐明长江流域气象干旱演变规律及其成因。【方法】基于长江流域站点和遥感数据，估算不同时间尺度的标准化降水指数（SPI）、标准化降水蒸散指数（SPEI）和蒸发需求干旱指数（EDDI），基于游程理论和小波分析，阐明长江流域干旱时空变化规律。利用共线性分析和皮尔逊相关分析筛选关键环流因子，揭示长江流域干旱演变的环流驱动机制。通过多元线性回归和XGBoost算法进行干旱预报。【结果】夏季极端降水与高温并存，冬季干旱与低温明显；下游平原暖干化趋势加剧，上游山区受地形调节作用显著。SPI、SPEI和EDDI在干旱监测中具有差异性。SPEI因考虑温度和蒸散的综合效应，更适用于气候变暖下的干旱识别，但多指数联合可提升极端事件的监测精度。干旱风险呈向西部转移、东部有所缓解的空间格局。1981年后，干旱严重度重心向西部转移，干旱历时缩短但峰值强度向西部高海拔地区迁移。关键环流因子（AMO）通过多尺度协同作用驱动干旱演变。AMO与SPEI的耦合周期随季节和年际尺度变化，反映海洋-大气系统的非线性响应。温度升高（尤其是最低温度）与风速下降是干旱恶化的主要成因，但蒸发需求与降水的非线性关系导致区域干旱特征分化（如干旱峰值西迁、中下游干旱历时缩短）。XGBoost算法在非线性干旱预测中表现优异，通过集成决策树优化残差，其长期预测精度高于多元线性回归模型（R2=0.56）。【结论】长江流域未来干旱风险预计显著增加，2021—2050年长江流域干旱次数、历时与严重度普遍加剧，但干旱峰值在2031—2050年回落至基准期水平，需警惕严重或极端干旱事件带来的不利影响。 |
| 关键词: 长江流域；干旱演变；环流指数；干旱预测；XGBoost |
| DOI：10.13522/j.cnki.ggps.2025233 |
| 分类号: |
| 基金项目: |
|
| Spatiotemporal dynamics, drivers and prediction of drought in the Yangtze River Basin under climate change |
|
YU Jiagu, LIU Siguo, YANG Ke, YAO Ning
|
|
College of Water Resources and Architectural Engineering, Northwest A&F University, Yangling 712100, China
|
| Abstract: |
| 【Objective】The Yangtze River Basin is a major economic and ecological region in China and has experienced increasing drought frequency and intensity in recent decades. This paper investigates the spatiotemporal variations in drought and its underlying drivers in this basin.【Method】Seasonal and interannual variations in meteorological variables and drought indices were analyzed using data measured from weather stations across the basin. Different meteorological indices were compared as proxies for drought monitoring and prediction. Drought duration, intensity and spatiotemporal evolution were analyzed using run-length theory and wavelet analysis; dominant atmospheric circulation factors influencing drought were identified using collinearity diagnostics and Pearson correlation analysis. Drought prediction models were developed using multiple linear regression and the extreme gradient boosting (XGBoost) algorithm, based on optimal drought and circulation indices.【Result】①Summer was characterized by extreme precipitation and high temperature, whereas winter was associated with low temperature. Warming and drying intensified in the downstream plains, while drought in the upstream mountainous regions was modulated by topography. SPI, SPEI and EDDI were complementary in drought monitoring. SPEI, incorporating temperature and evapotranspiration effects, was more accurate under warming conditions. Integrating multiple indices improved early prediction. Spatially, drought risk shifted westward in that, since 198, the centroid of drought severity has shifted westward, accompanied by shorter drought duration and increasing peak intensity in high-altitude western areas. ②Large-scale circulation patterns, particularly the Atlantic Multidecadal Oscillation (AMO), dominantly influenced drought through multi-scale interactions. The AMO-SPEI coupling varied seasonally and interannually, reflecting nonlinear ocean-atmosphere feedbacks. Increases in temperature, especially minimum temperature, and decreases in wind speed were responsible for drought intensification. Nonlinear interactions between evapotranspiration and precipitation explained the spatiotemporal variations in drought. The XGBoost model was superior to multiple linear regression in capturing nonlinear drought dynamics; its coefficient of determination for long-term drought prediction was R2=0.56. Projections indicated that drought frequency, duration and severity in the basin are likely to increase during 2021—2050, although peak drought intensity during 2031—2050 may return to baseline levels. 【Conclusion】Flooding and drought management in the Yangtze River Basin should consider both large-scale climatic drivers and regional heterogeneity. SPEI can be used as a drought proxy under warming conditions, supplemented by other indices to ensure predictive accuracy. Accounting for multi-scale circulation interactions and nonlinear climate processes is essential for drought risk assessment and management. The projected increase in drought underscores the urgency of developing robust mitigation strategies to reduce the impacts of severe and extreme droughts. |
| Key words: Yangtze River Basin; evolution of drought; circulation index; drought prediction; XGBoost |
