Global climate models performance: A comprehensive review of applied approaches, recognized issues and possible future directions

Future climate projections of global climate models (GCMs) are widely used for climate change impact assessments. However, increasing available GCMs and computational constraints necessitate using a selected subset for such applications. The selection of the GCM subset is a critical step in climate change impact assessment, given the computational constraints and inherent uncertainties associated with climate projections. This review comprehensively evaluates existing GCM evaluation and selection methods, focusing on past performance, future projections, and hybrid approaches. It highlights the challenges in selecting GCMs due to varying resolutions, interpolation techniques, reference datasets, and the sensitivity of evaluation metrics. Traditional performance-based methods emphasize replicating historical climate, while envelope and advanced hybrid approaches integrate future projection consistency and past performance to address uncertainties. The review underscores the growing importance of multi-model ensembles (MMEs) in reducing projection uncertainties and enhancing the reliability of climate simulations. Advanced statistical techniques, including hierarchical clustering and divergence analysis, offer promising tools for addressing structural uncertainties and improving model evaluation. The study identifies key gaps, such as the need for standardized metrics, the integration of extreme event simulations, and the development of region-specific selection frameworks. Future research directions include leveraging machine learning techniques, expanding reference datasets, and incorporating stakeholder-driven criteria to refine GCM selection processes. By addressing these challenges, the study aims to enhance the robustness and applicability of GCM selection methodologies, contributing to more reliable climate projections and informed decision-making in climate adaptation and mitigation strategies.