Scholarlink

<p>This study systematically analyzes the spatiotemporal trends and driving mechanisms of maximum temperature (MaxT), minimum temperature (MinT), and precipitation in the Yangtze River Basin from 1984 to 2023 using ERA5-Land reanalysis data and multi-source climate indices. To overcome the limitation of basin-wide extremes, the basin is divided into three sub-regions (upper, middle, lower reaches) and spatial means are calculated. The Mann–Kendall test reveals significant warming across all sub-regions: MaxT increases by 0.382–0.384 ℃/decade (p &lt; 0.01) and MinT by 0.324–0.331 ℃/decade (p &lt; 0.05), while precipitation shows a weak but significant decreasing trend (− 0.087 to − 0.282&#xa0;mm/decade, p &lt; 0.05). A controlled machine learning experiment shows that climate indices (PDO, ENSO, IOD, AO) alone explain 79–89% of the daily temperature variance (R<sup>2</sup>_reduced = 0.79–0.89), and adding lagged meteorological variables further improves R<sup>2</sup> to 0.91–0.98. For precipitation, the reduced model performs poorly (R<sup>2</sup> = 0.06–0.43), highlighting its stochastic nature. Lagged correlation analysis (after detrending) indicates that the Pacific Decadal Oscillation (PDO) exerts the strongest influence on temperature with a 4–5&#xa0;month lag (r ≈ 0.93–0.94, p &lt; 0.001). Seasonal analysis shows PDO positively correlates with temperature in spring and autumn but negatively in winter. The hybrid statistical-machine learning framework provides a robust tool for regional climate prediction, with practical value for water resource management and agricultural planning in the Yangtze River Basin.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Trend analysis and prediction of maximum temperature, minimum temperature, and precipitation in the Yangtze river basin based on multi-source climate indices and machine learning

  • 期刊论文

  • 发表日期：

    2026年7月18日

查看全文
  • Zebin Li,
  • Zhijie Ta,
  • Qian Ren

摘要

This study systematically analyzes the spatiotemporal trends and driving mechanisms of maximum temperature (MaxT), minimum temperature (MinT), and precipitation in the Yangtze River Basin from 1984 to 2023 using ERA5-Land reanalysis data and multi-source climate indices. To overcome the limitation of basin-wide extremes, the basin is divided into three sub-regions (upper, middle, lower reaches) and spatial means are calculated. The Mann–Kendall test reveals significant warming across all sub-regions: MaxT increases by 0.382–0.384 ℃/decade (p < 0.01) and MinT by 0.324–0.331 ℃/decade (p < 0.05), while precipitation shows a weak but significant decreasing trend (− 0.087 to − 0.282 mm/decade, p < 0.05). A controlled machine learning experiment shows that climate indices (PDO, ENSO, IOD, AO) alone explain 79–89% of the daily temperature variance (R2_reduced = 0.79–0.89), and adding lagged meteorological variables further improves R2 to 0.91–0.98. For precipitation, the reduced model performs poorly (R2 = 0.06–0.43), highlighting its stochastic nature. Lagged correlation analysis (after detrending) indicates that the Pacific Decadal Oscillation (PDO) exerts the strongest influence on temperature with a 4–5 month lag (r ≈ 0.93–0.94, p < 0.001). Seasonal analysis shows PDO positively correlates with temperature in spring and autumn but negatively in winter. The hybrid statistical-machine learning framework provides a robust tool for regional climate prediction, with practical value for water resource management and agricultural planning in the Yangtze River Basin.