Key points are not available for this paper at this time.
Accurate monitoring and prediction are essential for mitigating the impacts of agricultural droughts resulting from water shortages. In many regions, insufficient soil moisture is the primary factor driving agricultural drought. However, the dominant factors that affect both long-term and short-term soil moisture variations at different depths remain not fully explored. In addition, although soil moisture is widely recognized as the most direct and sensitive indicator for agricultural drought monitoring and early warning, few studies have investigated satellite-based soil moisture climatological records for agricultural drought assessment. In this study, 30 years (1990–2019) climatological records of surface soil moisture (SSM) and root-zone soil moisture (RZSM) from the European Space Agency-Climate Change Initiative (ESA-CCI) were used to investigate the characteristics of agricultural drought events in three US states. The results indicate that during prolonged drought events, RZSM remains relatively stable and is mainly influenced by precipitation and infiltration from SSM, whereas SSM shows pronounced fluctuations. For short-duration drought events, precipitation consistently emerges as the dominant factor controlling both SSM and RZSM, and RZSM exhibits a lagged response to precipitation. Furthermore, a novel knowledge-guided machine learning model was developed for agricultural drought prediction. Compared with a standard machine learning model, the proposed model improves RZSM prediction performance by approximately 8% and more accurately reflects drought intensity across the study region. Overall, these findings provide new insights into soil moisture dynamics under drought conditions and offer a robust framework for improved agricultural drought monitoring and forecasting.
Zuo et al. (Thu,) studied this question.