Scholarlink

<p>This study develops an “Attention Mechanism + Random Forest” model integrating GRACE/GRACE-FO, GLDAS, and ERA5-Land data to downscale GWSA resolution from 0.25° to 0.1° in Jiangsu Province. The downscaled data effectively captures local groundwater variations, showing strong agreement with original observations (R² = 0.95). Results reveal an overall GWSA increase of 4.824 mm/yr with distinct spatial heterogeneity: southern regions (e.g., Suzhou) show significant increases (up to 25.754 mm/yr) while northern areas (e.g., Lianyungang) experience declines (as low as −7.063 mm/yr). The GWSA centroid has shifted toward Changzhou-Taizhou, with seasonal variations primarily driven by precipitation (weight 0.32) and evapotranspiration. Analysis of heritage sites indicates pronounced fluctuations in lake-wetland areas compared to stable forest regions. This research provides crucial support for refined groundwater monitoring and sustainable management.</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

Groundwater storage downscaling and regional water resource analysis based on an attention-enhanced RF model

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月16日

查看全文
  • Yan Zhang,
  • Ning Zhang,
  • Chaoyu Zhang,
  • Zhenfei Pei,
  • Xinyu Yuan,
  • Yufan Hu

摘要

This study develops an “Attention Mechanism + Random Forest” model integrating GRACE/GRACE-FO, GLDAS, and ERA5-Land data to downscale GWSA resolution from 0.25° to 0.1° in Jiangsu Province. The downscaled data effectively captures local groundwater variations, showing strong agreement with original observations (R² = 0.95). Results reveal an overall GWSA increase of 4.824 mm/yr with distinct spatial heterogeneity: southern regions (e.g., Suzhou) show significant increases (up to 25.754 mm/yr) while northern areas (e.g., Lianyungang) experience declines (as low as −7.063 mm/yr). The GWSA centroid has shifted toward Changzhou-Taizhou, with seasonal variations primarily driven by precipitation (weight 0.32) and evapotranspiration. Analysis of heritage sites indicates pronounced fluctuations in lake-wetland areas compared to stable forest regions. This research provides crucial support for refined groundwater monitoring and sustainable management.