Detecting irrigation events by integrating multitemporal soil moisture dynamics and water-balance constraints
Abstract. Accurate monitoring of irrigation information is essential for sustainable agricultural water management. However, existing studies have focused on mapping irrigated areas and quantifying irrigation water use, while the timing of irrigation events remains insufficiently characterized. Soil moisture responds directly to irrigation, but rainfall produce similar signals, and response magnitude and duration vary among different irrigation events, making it severely challenging to differentiate irrigation events. To this end, this study developed a novel framework irrigation events detection integrating multitemporal soil moisture dynamics and soil water balance constrains. Multi-depth observations were integrated into unified metrics, and non-irrigation water variations were characterized by the water balance model, while multi-timescale wetting signals and water balance residual jointly identified potential irrigation responses for consolidation into irrigation events through water-source attribution and integration of adjacent responses. This framework achieved robust performance against 149 reference irrigation events at five sites in China, the United States, and Morocco, with mean Precision, Recall, and F1 scores of 0.8321, 0.8111, and 0.8148, respectively, while gaining site-level F1 scores ranging from 0.7097 to 0.9574. Under rainfall–irrigation overlap, 75 % of events were identified. Ablation analyses showed that direct wetting signals and water-source attribution reduced false detections from non-irrigation responses, whereas water balance residual improved detection of weak responses. Irrigation events detectability declined as observation interval increases, with median F1 scores decreasing to 0.62 and 0.53 at 5- and 6-day intervals, respectively. This framework provides a methodological foundation for extracting irrigation timing from satellite soil moisture observations and supporting regional agricultural water use monitoring.