Delineating dynamic hydrological response units to improve simulations of extreme runoff events in changing environments

Abstract

Accurate understanding of watershed hydrology is vital for managing water resources and predicting extreme events. However, existing distributed hydrological models depend on static hydrological response unit (HRU) clusters, which do not adequately reflect temporal variability driven by climate and human activities over time. In this study, we introduce a novel dynamic HRU clustering methodology that considers time-varying environmental factors to improve the accuracy of hydrological simulations. Using a distributed hydrological model (i.e., the Water and Energy Transfer Processes model), we identified the dynamics of HRU clusters from 1951 to 2020 and assessed the effectiveness of the model when dynamic HRU clusters were used in capturing extreme hydrological events across Southeast Asian watersheds. Our results indicate significant spatiotemporal variability in HRU clusters, with 51.8% of the study area experiencing changes during the last few decades. Hydrological simulations using dynamic HRU clusters showed improved performance, with the average Nash-Sutcliffe efficiency coefficient (NSE) for the runoff simulations across 14 hydrological stations improved from 0.49 to 0.72. Additionally, the simulations of extreme events using dynamic HRU clusters demonstrated high accuracy, reducing the differences between the simulated and observed runoff thresholds for drought and flood events. We further analyzed the contributions of climate change and human activities to runoff dynamics, revealing that air temperature and human activities are the primary drivers of runoff changes, with spatial heterogeneity observed across different watersheds. Our study indicates that incorporating dynamic environmental factors into hydrological models enhances the simulation accuracy and supports water resource management and climate adaptation strategies.