Risk of real-time irrigation decision-making system for farmland in arid irrigation districts: Methodology and case study

Making precise irrigation decisions several days in advance is of great significance for improving water resource utilization efficiency in arid regions. While crop models such as AquaCrop aid in predicting soil moisture and water requirements, uncertainties arising from soil heterogeneity, management practices, and crop traits can compromise the accuracy of irrigation decisions. Therefore, this study develops a real-time irrigation decision-making (RTID) system and a risk analysis framework based on AquaCrop to evaluate the impacts of uncertainty on a virtual maize field in the Hanyan Irrigation District, an arid region of Northwest China. Results show that uncertainties in weather forecasts, including reference crop evapotranspiration (ET_o) and precipitation, minimally affect net irrigation requirement (NIR) in drought areas. In contrast, sowing date, soil parameters, and crop coefficient (K_c) introduce significant variability. When maximized, these factors cause NIR fluctuations of −15 % to + 13 %, −5 % to + 12 %, and −10 % to + 10 %, respectively. Under the combined influence of these uncertainty factors, the fluctuations in NIR exhibit a saturation effect, meaning that as uncertainty factors continue to accumulate, the magnitude of NIR fluctuations no longer increases obviously. Statistical analysis indicates that when all factors act together, 90 % of NIR predictions remain within ±15 %, while yield losses exceed 1.5 % in ≤ 25 % of cases and 4 % in ≤ 5 % of cases, respectively. Moderately increasing the per-application NIR slightly reduces the risk of yield loss, but the benefits diminish beyond a 10 % increment. These findings provide scientific and practical insights for optimizing precision irrigation in arid regions, highlighting key sources of uncertainty and their impacts on water use efficiency and yield stability.