Abstract

Understanding the adaptability of hydrological models under changing climate conditions is crucial for enhancing the accuracy of streamflow simulations in regions with high rainfall variability, such as southern Thailand. Although the GR2M (Génie Rural à 2 paramètres) model, a two-parameter monthly conceptual hydrological model, has shown reliable performance, it underestimates streamflow, particularly during high-flow periods. To address this limitation, this study introduces a flexible parameter-switching mechanism (GR2M-S) that allows model parameters to adapt to different rainfall conditions. The proposed model was compared with the original GR2M (GR2M-O), which applies constant parameters throughout the simulation period. Rainfall conditions were classified using k-means clustering to divide the data into two regimes. The main parameters, X<inf>1</inf> (soil water storage capacity) and X<inf>2</inf> (groundwater exchange coefficient), were then adjusted separately for each regime to improve model flexibility and realism. Sensitivity analysis showed that X<inf>2</inf> had the strongest and consistently positive influence on streamflow, reflecting the role of groundwater exchange and routing processes, while X<inf>1</inf> exhibited a non-linear response, indicating the complexity of soil water storage processes. The results indicate that the GR2M-S model outperformed GR2M-O across all studied catchments, particularly during high-flow periods. The Kling-Gupta Efficiency (KGE) and Overall Index (OI) increased, while the Mean Absolute Error (MAE) and Percent Relative Peak Error (PE) decreased. The combined metric (CM) improved by up to 11.82, demonstrating the effectiveness of the proposed approach. Overall, the GR2M-S model provides a more flexible and reliable representation of rainfall-runoff relationships and shows strong potential for streamflow forecasting in data-scarce and climatically variable regions.