Peak-cluster depression catchment in southwest of China This study investigates the influence of antecedent rainfall characteristics on streamflow dynamics in karst catchments. Autocorrelation and cross-correlation analyses of rainfall–streamflow series were conducted to assess the memory effect and determine the duration of antecedent rainfall impacts on streamflow. Machine learning models (random forest, RF; support vector machine, SVM; and artificial neural network, ANN) were applied to identify key antecedent rainfall indicators. These indicators were subsequently used to enhance streamflow predictions and explore ecohydrological mechanisms linking rainfall patterns to streamflow variability. The study emphasizes the critical role of antecedent rainfall characteristics, such as maximum consecutive five-day precipitation and consecutive drought days, in influencing streamflow dynamics. By incorporating these indicators into hydrological models, both the prediction accuracy and understanding of ecohydrological processes were significantly enhanced. Climate change emerged as the predominant driver of streamflow variability, contributing 73.8% to the observed changes in streamflow patterns. In contrast, vegetation restoration had a negative impact on streamflow, with a contribution of −3.8%. Specifically, extreme rainfall events were found to be major drivers of interannual streamflow variability, whereas moderate rainfall events primarily affect streamflow persistence. These findings have significant implications for the management of water resources and ecosystem services in karst regions, particularly in the context of continuing climate change. • Karst streamflow exhibits similar rainfall memory lengths across different hydrological years. • Adding memory time-based rainfall characteristics boosts streamflow prediction accuracy. • Extreme and moderate rainfall characteristics shape amplitude and persistence of annual streamflow.
Wang et al. (Sun,) studied this question.