<p>Reducing uncertainty in climate change projections is essential for developing reliable impact assessments and effective adaptation strategies in water resource management. Climate model sub-selection has become a key tool for managing large ensembles; however, trade-offs between accuracy, uncertainty coverage, and spatial consistency remain insufficiently explored. This study introduces a standardised multi-criteria framework to evaluate sub-selection methods across basins with contrasting hydrological regimes. The framework integrates five indicators across three dimensions—performance in projecting observed change signals in a second historical or “<i>future</i>” period reserved for validation of the projections, uncertainty representativeness, and computational simplicity—into two comprehensive indices: average performance and spatial consistency. Results show that no single method optimises all criteria, confirming the existence of trade-offs. Methods based on future change diversity and adaptive consensus approaches perform best, striking a balance between accuracy and uncertainty representation, and consistently outperform mono-model strategies. Crucially, spatial variability proved to be a decisive factor. Some methods performed adequately at aggregated scales but lost consistency across heterogeneous sub-basins, underscoring the value of spatial consistency as an explicit criterion for selection. By incorporating spatial robustness into ensemble evaluation, this study advances a reproducible and low-complexity procedure for identifying balanced sub-selection methods. Beyond the case study, the framework is transferable to other regions, climate variables, and sectors, offering transparent and standardised guidance for model selection. These findings enhance the credibility of climate impact assessments and support more equitable, robust adaptation planning and policy in climate-sensitive domains, such as water management, agriculture, and energy.</p>