Scholarlink

<p>Simulating rainfall interception is essential for understanding the global water cycle. Yet most Earth system models rely on leaf area index-based representations, with limited three-dimensional canopy structure, introducing structural uncertainty. Here, we introduce Feature-Constrained Deep Symbolic Regression, an artificial intelligence approach for explicit canopy interception parameterization. Global analyses show that three-dimensional canopy morphology and wind dynamics dominate interception, while precipitation intensity reduces interception efficiency. Morphology-based parameterizations improve site-level interception loss rate KGESS by 0.27–0.39 over traditional approaches. Budyko-based analyses show improved evapotranspiration at nearly 70% of global flux sites and stronger runoff dynamics across six major river basins. CONUS CoLM simulations show improved runoff over 71.5% of the domain relative to GRDC and streamflow at 55.7% of GRFR stations. Global applications reveal nonlinear interception responses to rainfall intensity and vegetation structure. Together, Feature-Constrained Deep Symbolic Regression enables morphology-based parameterization, reducing rainfall partitioning uncertainty in Earth system models.</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

Three-dimensional canopy morphology and wind dynamics govern global rainfall interception

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月29日

查看全文
  • Qingliang Li,
  • Xiaochun Jin,
  • Zhongwang Wei,
  • Cheng Zhang,
  • Wei Shangguan,
  • Jinlong Zhu,
  • Zijian Zhang,
  • Xiaoning Li,
  • Yuguang Yan,
  • Jing Wang,
  • Xiaoming Shi,
  • Yunfeng Lv,
  • Yongjiu Dai

摘要

Simulating rainfall interception is essential for understanding the global water cycle. Yet most Earth system models rely on leaf area index-based representations, with limited three-dimensional canopy structure, introducing structural uncertainty. Here, we introduce Feature-Constrained Deep Symbolic Regression, an artificial intelligence approach for explicit canopy interception parameterization. Global analyses show that three-dimensional canopy morphology and wind dynamics dominate interception, while precipitation intensity reduces interception efficiency. Morphology-based parameterizations improve site-level interception loss rate KGESS by 0.27–0.39 over traditional approaches. Budyko-based analyses show improved evapotranspiration at nearly 70% of global flux sites and stronger runoff dynamics across six major river basins. CONUS CoLM simulations show improved runoff over 71.5% of the domain relative to GRDC and streamflow at 55.7% of GRFR stations. Global applications reveal nonlinear interception responses to rainfall intensity and vegetation structure. Together, Feature-Constrained Deep Symbolic Regression enables morphology-based parameterization, reducing rainfall partitioning uncertainty in Earth system models.