Multi criteria evaluation of downscaled CMIP6 models in predicting precipitation extremes

Abstract

The selection of general circulation models (GCMs) is primary information required for assessing climate change impacts on the hydrological vulnerability of any region. The uncertainties associated with GCMs at the regional scale are mostly attributable to coarser representation of climatic processes, making model ranking an essential step for improving multi-model ensemble (MME) performance. The present study evaluated 13 downscaled-bias-corrected CMIP6 GCMs for eight extreme precipitation indices over the flood-prone Brahmaputra River basin. Precipitation extremes from 1985 to 2014 were employed to evaluate model performance at a grid resolution of 0.25°, while projected events were assessed for the early future (2031–2060) and far future (2071–2100). Individual rankings for precipitation indices were determined using five multicriteria decision-making (MDCM) techniques: TOPSIS, VIKOR, EDAS, PROMETHEE-II, and Performance Matrix. The Criteria Importance Through Inter-criteria Correlation (CRITIC) technique was used to assign weights to each performance indicator for indices-wise ranking. The comprehensive ranking from the various MCDM techniques was further obtained using group decision-making method. The results show that different models are better at capturing different precipitation characteristics, necessitating indices-based rankings for future estimates. The study additionally indicates that Multi-Model Ensemble, MME8, and MME5 outperformed the other ensembles in reducing simulation uncertainty in downscaled GCMs. Future projections indicate an overall increase in precipitation extremes, with the best model ensembles predicting a wetter early future and a drier far future than all model ensembles.