Reducing Hydrological Uncertainty in Large Mountainous Basins: The Role of Isotope, Snow Cover, and Glacier Dynamics in Capturing Streamflow Seasonality

Hydrological modeling in large mountainous catchments faces challenges owing to the complex interplay of snowmelt, glacier dynamics, and groundwater contributions, which introduce significant uncertainty in streamflow predictions. This study introduces a Bayesian multi-objective parameter estimation framework to reduce predictive streamflow uncertainty in large mountainous catchments by integrating streamflow likelihood with three auxiliary likelihoods, analyzed individually: snow cover area (SCA), glacier mass balance (GMB), and isotopic composition (I). The well-established generalized likelihood uncertainty estimation (GLUE) method is employed to investigate trade-offs among these likelihoods, providing a detailed assessment of their distinct and combined contributions to hydrological model performance across various flow regimes. The semi-distributed tracer-aided Tsinghua representative elementary watershed-tracer (THREW-T) hydrological model applied in this work captures both rapid surface dynamics and slow-response subsurface processes, offering a comprehensive representation of streamflow variability. Results indicate that isotopic likelihood plays a critical role in reducing low-flow uncertainty by effectively constraining subsurface flow and groundwater–surface water interactions, particularly during winter and early spring when these processes dominate. Conversely, while SCA and GMB likelihoods demonstrate some effectiveness in capturing rapid processes such as snowmelt and glacier melt, their influence is most pronounced during the melting season, with limited effect on reducing overall streamflow uncertainty. This seasonality is reflected in sharpness (SH) values, which measure how much uncertainty is reduced, with isotopic likelihood achieving the highest peak of 0.34 in late winter, whereas SCA and GMB reach maximum SH values of 0.19 and 0.16, respectively, during the melting season. Pareto plots further reveal the synergies and trade-offs associated with each likelihood, underscoring the importance of adopting a multi-objective calibration approach that accounts for seasonal variations in hydrological processes. In addition, the results highlight the critical role of seasonality in shaping the effectiveness of auxiliary likelihoods, emphasizing their potential to improve predictive accuracy and reduce uncertainty in hydrological models.