Scholarlink

<p>Irrigation water productivity (IWP) in arid agricultural regions is under increasing threat from climate change, while field-level water-saving practices offer potential pathways for adaptive management. However, the complex, nonlinear, and crop-specific responses of IWP to interacting environmental and management factors remain poorly understood. This study presents an explainable machine learning framework integrating remote sensing and environmental data to quantify, interpret, and predict IWP dynamics for wheat, maize, and sunflower in the Jiefangzha Irrigation Area of China’s Hetao Irrigation District - the largest artesian irrigation system in Asia. Among six candidate machine learning models evaluated, a Bayesian-optimized CatBoost model achieved the highest predictive accuracy (R² &gt; 0.95 across three crops). SHAP-based interpretation identified irrigation volume, sunshine hours, and groundwater evaporation as key IWP drivers, with substantial variation in influence patterns among crops. By 2030, the region’s IWP will be reduced by an average of 0.105&#xa0;kg/m<sup>3</sup> compared to the year of 2006. We further simulated future IWP trajectories under coupled climate change scenarios (CMIP6) and water conservation measures, including canal lining and field irrigation upgrades. Results indicated that these interventions can partially offset climate-driven declines in IWP with nearly 0.051&#xa0;kg/m<sup>3</sup>, though their effectiveness is both crop- and location-specific. This study offers a comprehensive, robust, data-driven approach for evaluating IWP under environmental and managerial uncertainties, providing transferable insights for sustainable irrigation in other arid and semi-arid regions.</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

Interpreting and forecasting crop-specific irrigation water productivity in an arid irrigated area using explainable machine learning and scenario simulation

  • 期刊论文

  • 发表日期：

    2026年1月12日

查看全文
  • Yang Lei,
  • Liuyue He,
  • Shouzheng Jiang,
  • Zailin Huo,
  • Isaya Kisekka,
  • Jingyuan Xue

摘要

Irrigation water productivity (IWP) in arid agricultural regions is under increasing threat from climate change, while field-level water-saving practices offer potential pathways for adaptive management. However, the complex, nonlinear, and crop-specific responses of IWP to interacting environmental and management factors remain poorly understood. This study presents an explainable machine learning framework integrating remote sensing and environmental data to quantify, interpret, and predict IWP dynamics for wheat, maize, and sunflower in the Jiefangzha Irrigation Area of China’s Hetao Irrigation District - the largest artesian irrigation system in Asia. Among six candidate machine learning models evaluated, a Bayesian-optimized CatBoost model achieved the highest predictive accuracy (R² > 0.95 across three crops). SHAP-based interpretation identified irrigation volume, sunshine hours, and groundwater evaporation as key IWP drivers, with substantial variation in influence patterns among crops. By 2030, the region’s IWP will be reduced by an average of 0.105 kg/m3 compared to the year of 2006. We further simulated future IWP trajectories under coupled climate change scenarios (CMIP6) and water conservation measures, including canal lining and field irrigation upgrades. Results indicated that these interventions can partially offset climate-driven declines in IWP with nearly 0.051 kg/m3, though their effectiveness is both crop- and location-specific. This study offers a comprehensive, robust, data-driven approach for evaluating IWP under environmental and managerial uncertainties, providing transferable insights for sustainable irrigation in other arid and semi-arid regions.