The role of precipitation deficit and potential evapotranspiration excess in triggering flash droughts over China

Flash droughts are characterized by a rapid decline in soil moisture, transitioning from normal to drought conditions within weeks. Driven by water-limited and energy-limited conditions, their swift onset and intensification can severely impact agriculture, ecosystems, and socioeconomic systems. This study analyzed the spatiotemporal patterns of negative precipitation (P) and positive potential evapotranspiration (PET) anomalies by introducing four indices including, time length, average, maximum, and total level during the onset-development of flash droughts across China from 1979 to 2019. Using random forest (RF), Gradient boosting decision tree (GBDT), and extreme gradient boosting (XGBoost), we constructed the relationship between the rapid rate of intensification (RI) of soil moisture and anomaly features of P and PET. The models’ performance in simulating RI and capturing flash droughts was evaluated, and the contributions of P deficit and PET excess were further quantified. The results indicated that absolute negative P anomalies showed longer time lengths and larger total values than positive PET anomalies, while the opposite trend was for average and maximum values. The proportion of total positive PET anomalies gradually increased over the past 40 years across China. For the models’ performance, RF presented the best performance and showed stronger detection capabilities in most China sub-regions. P deficit was the primary driver of flash drought, especially in northeastern, northern, and southern China, contributing more than 60 %, while PET excess played an important role in the Tibetan Plateau and Xinjiang regions. This study is valuable to advance the understanding of flash drought mechanisms by highlighting the dual roles of P deficit and PET excess.