Cross-scale synergistic optimization for irrigation allocation and fertilizer management under uncertainty

In the face of global water scarcity and food security challenges, improving agricultural water use efficiency under limited resources is essential for sustainable development. This study integrates multi-objective optimization, water-fertilizer regulation models, and large-system decomposition–coordination theory to develop L-FAI, a multi-scale irrigation optimization framework based on iterative “optimization-validation” cycles. The model reveals a three-level parameter transfer mechanism, linking field-scale water-fertilizer response, administrative-level economic transmission, and irrigation-district equity feedback. It dynamically couples the Stewart function with uncertainty to optimize daily water allocation and balance cross-scale resources. Using the Chahayang Irrigation District in China as a case, the L-FAI model achieves three breakthroughs: (1) At the system level, irrigation use is reduced by 12.8 % (2.28 ×107 m3), economic benefits rise by 11 % (CNY 822 million), and allocation equity improves (Gini coefficient = 0.25); (2) At the efficiency level, irrigation schedules tailored to crop-specific water demand patterns across growth stages—e.g., shallow frequent irrigation for rice, rainfed for maize, and water-saving for soybean—boost yield by 3.5–4.5 % and water-fertilizer use efficiency by 10–15 %; (3) At the climate resilience level, irrigation efficiency improves by 10 % under hydrological extremes while yield fluctuations remain within 3 %. The study uncovers how water-fertilizer synergy and cross-scale feedback are coupled, bridging the gap between system-level integrity and local adaptability. It provides a theoretical and technical basis for resilient and efficient agricultural water management.