Akademik Veri Yönetim Sistemi
Water (Switzerland), cilt.17, sa.20, 2025 (SCI-Expanded)
While weighted multi-model approaches are widely used to improve predictive capability, hydrological models (HMs) and their weighted combinations that perform well under past conditions may not guarantee robustness under future climate scenarios. Furthermore, the extent to which weighting schemes influence the propagation of runoff projection uncertainty remains insufficiently explored. Therefore, this study evaluates the capacity of strategies that weight monthly scale HMs to narrow runoff projection uncertainty. Since standard approaches rely only on historical simulation skill and offer static weighting, this study introduces a refined framework, the Uncertainty Optimizing Multi-Model Ensemble (UO-MME), which dynamically considers the trade-offs between calibration performance and projection uncertainty. In performing the uncertainty decomposition, a total of 140 ensemble runoff projections, generated through a modelling chain comprising five GCMs, two emission scenarios, two downscaling methods, and seven HMs, were analyzed for Beydag and Tahtali watersheds in Türkiye. Results indicate that standard techniques, such as Bayesian model averaging, ordered weighted averaging, and Granger–Ramanathan averaging, led to either marginal reductions or noticeable increases in projection uncertainty, depending on the case and projection period. Conversely, the UO-MME achieved average reductions in projection uncertainty of around 30% across the two watersheds by balancing the influences of climate signals produced by GCMs that are reflected in the projections through HMs while maintaining high simulation accuracy, as indicated by Nash–Sutcliffe efficiency values exceeding 0.75. Although not designed to eliminate inherently irreducible uncertainty, the UO-MME framework helps temper the inflation of noisy GCM signals in runoff responses, providing more balanced hydrological projections for water resources planning.