Key points are not available for this paper at this time.
This study was conducted in the ecologically critical and climate-sensitive Poyang Lake Region of China. Global climate change has increased extreme hydrological events worldwide, necessitating advanced hydrological models to manage escalating risks. This study proposed a hybrid model, SWAT-GCN-BiLSTM, integrating the strengths of SWAT (simulating physical hydrological processes), GCN (capturing spatial topological relationships), and BiLSTM (modeling complex temporal dynamics). The hybrid SWAT-GCN-BiLSTM model outperformed the standalone SWAT and BiLSTM models, with significantly higher NSE and R 2 values of around 0.90. The hybrid model particularly excelled in simulating extreme flows, reducing RMSE by over 20 % for extremely high flows (≥ Q10, Q10 represents streamflow magnitude with a 10 % exceedance probability). Based on the ensemble mean of four Global Climate Models, the hybrid model predicted a substantial increase in streamflow during the wet months of April (24.9 %-44.1 %) and May (11.5 %-20.2 %) compared to the baseline. Furthermore, under all considered climate change scenarios, the Q10 of the 7-day flow was projected to increase by 9.5–19.5 %. Conversely, streamflow in the dry months of November and December was projected to decrease by 21.0–34.7 %. This indicates climate change may exacerbate hydrological extremes, necessitating robust adaptive management strategies to address both increased spring flooding risk and heightened drought conditions during late autumn/early winter in the region under a changing climate. • Propose a novel SWAT-GCN-BiLSTM framework to enhance hydrological extreme simulation. • SWAT-GCN-BiLSTM outperforms both SWAT and BiLSTM with NSE and R² of around 0.90. • SWAT-GCN-BiLSTM excels at extreme flow simulation, reducing RMSE by > 20 % for Q10 flow. • Climate change aggravates spring flood and late fall/early winter drought in the Poyang region.
Zheng et al. (Wed,) studied this question.