Construction and application of a real-time irrigation decision system based on the multiple sources data assimilation and weather forecasts

Improving irrigation strategies is essential for increasing irrigation water productivity (IWP) while maintaining crop yield. We developed a real-time irrigation decision system that couples the soil-water-atmosphere-plant (SWAP) crop simulation model with real-time dynamic calibration by simultaneously assimilating available soil moisture (SM) and leaf area index (LAI) observations using the iterative ensemble smoother (IES). Driven by short-term weather forecasts, the system predicts future crop water stress and generates irrigation decisions through two algorithms: SWAP-IES-Iteration (SII), which is suitable for farms with a stable water supply, and SWAP-IES-Optimization (SIO), which is suitable for farms that require multiple feasible options under constrained water supply. Field experiments with silage maize in Ningxia, China, in 2024 and 2025 showed that assimilating SM and LAI improved the simulation accuracy of soil water and crop states. The data assimilation process reduced the soil moisture root mean square error (RMSE) to 0.019–0.032 cm³/cm³, increased the Nash-Sutcliffe efficiency (NSE) to 0.53, and narrowed percent bias to between −16.54% and 10.66%. Compared with two farmer-managed treatments, the proposed system achieved competitive yields while using much less irrigation water. As a result, IWP increased to 21.16–33.88 kg/m 3 , compared with 13.70–19.66 kg/m 3 under the farmer-managed treatments. Further analysis showed that observation costs can be reduced through a staggered observation schedule, and the framework shows potential for application in other water-scarce regions when public meteorological data and field observations are available.