ABSTRACT The selection of best‐performing CMIP6 Global Climate Models (GCMs) is vital for basin‐scale climate impact assessments (BCIA) and essential to minimize the biases between climate projections and observations. The current study introduces a novel four‐level framework to evaluate and rank 24 GCMs from the NEX‐GDDP‐CMIP6 dataset across 27 Indian River Basins (IRBs) in the Indian Mainland (IM), using a Multi‐Metric Evaluation Approach (MMEA). Downscaled precipitation products from NEX‐GDDP‐CMIP6 for 1981–2010, provided at 0.25° × 0.25° spatial resolution, are considered and evaluated against India Meteorological Department (IMD) gridded data. In Level‐1 , GCMs are assessed at each grid using six categorical and nine statistical performance metrics. Fourteen GCMs ranked among the top five (T 5 ) performers for at least once in any metric. Metrics like correlation coefficient (CC) ≥ 0.6 identified models such as BCC‐CSM2‐MR, IPSL‐CM6A‐LR, MPI‐ESM1‐2‐LR, CMCC‐CM2‐SR5, and MIROC6 as T 5 best performers. In Level‐2 , basin‐wise performance was evaluated using individual and cluster‐wise statistical metrics, revealing that GCM rankings varied significantly with each metric/cluster, emphasizing the need for multi‐criteria assessment. In Level‐3 , GCMs are evaluated using five Multi‐Criteria Decision‐Making (MCDM) methods—CP, CGT, TOPSIS, WA, and PROMETHEE‐2 to refine rankings. The first three methods revealed that the rankings of the T 5 best‐performing GCMs varied slightly across methods, WA and PROMETHEE‐2 yielded consistent results, highlighting the importance of synthesizing outputs. In Level‐4 , a Group Decision‐Making (GDM) approach (net strength method) was employed to finalize consensus rankings per IRB. Notably, all 24 GCMs ranked in the T 5 for at least one basin, reflecting spatial variability in model performance. Additionally, GCM performance in capturing extreme rainfall was evaluated using the 90th and 95th percentile rainfall across each grid over the IM. The analysis identified IPSL‐CM6A‐LR, CMCC‐ESM2, INM‐CM4‐8, ACCESS‐ESM1‐5, and GFDL‐ESM4 as consistently skillful in representing extreme rainfall magnitude, while MRI‐ESM2‐0, INM‐CM4‐8, MPI‐ESM1‐2‐HR, CanESM5, and IPSL‐CM6A‐LR performed well in capturing extreme rainfall frequency. Overall, the results show that GCMs performing well for mean rainfall do not necessarily capture extreme rainfall behaviour. This robust framework facilitates the selection of optimal GCMs for BCIA, thereby supporting resilient water resource planning and adaptive management.
Roulo et al. (Sun,) studied this question.