A multi-metric approach for assessing CMIP6 GCM suitability for temperature extremes in Semi-Arid Regions

Climate change has intensified temperature variability and extreme heat events, particularly in semi-arid regions where agriculture, water resources, and livelihoods are highly climate-sensitive. Reliable temperature projections are therefore essential for effective adaptation planning, yet the performance of General Circulation Models (GCMs) varies across regions and temperature characteristics. However, existing studies have largely focused on mean temperature, with limited attention to extreme temperature dynamics (intensity, duration, and frequency), and often rely on multiple statistical metrics without systematically selecting an optimal combination, leading to potential biases in model evaluation and selection. To address these limitations, this study develops a comprehensive and relatively robust framework for evaluating CMIP6 GCM performance by integrating both mean and extreme temperature indices defined by the Expert Team on Climate Change Detection and Indices (ETCCDI). The framework employs ten statistical metrics, applies a leave-one-out approach to identify the optimal combination of metrics, and incorporates multi-criteria decision analysis (MCDA) techniques, namely TOPSIS and VIKOR, for relatively robust model ranking. Empirically, there remains a lack of systematic assessment of GCM performance in the semi-arid region of Rajasthan, India, a climate-sensitive area where temperature variability significantly affects agriculture and livelihoods. The proposed framework is applied to this region using 28 CMIP6 GCMs for the historical period (1961–2014), with ERA5 as the reference dataset. Results indicate that GISS-E2-1-G, KACE-1-0-G, EC-Earth3, UKESM1-0-LL, and NorESM2-MM perform most consistently based on the selected optimal metrics (RMSE, MAE, NSE, and R²), although no single model performs best across all temperature indices. The proposed framework provides a transparent and reproducible basis for relatively robust GCM selection for improved regional climate projections and adaptation planning in semi-arid regions.