Abstract

Observations about inadequacy of GeneralClimate Models (GCMs) of the sixth phase of Climate ModelIntercomparison Project (CMIP6) in simulating differentclimatological variables in several regions of the worldprompted the examination of the performance of GCMs of thepreceding fifth phase (CMIP5) phase vis-à-vis those of CMIP6in simulating mean and extreme temperatures over aphysiographically diverse mainland India. Selected statistics ofdaily temperature, simulated at 0.25×0.25 grid by 30 GCMs ofCMIP6 and 28 of CMIP5, were compared with those ofreference data of 27 years for performance assessment of thetwo phases. Significance Score, Mean Absolute Error and Indexof Agreement, were applied to explore the spatial distribution ofthe best-performing model of each CMIP phase, and aninnovative rank-based approach was devised and Taylordiagrams were used to examine the spatially aggregatedperformance of the models. GCMs of both CMIP6 and CMIP5were found as being valuable in simulating one or the othertemperature statistic in one or the other part of the country withno indication of one phase being consistently better than theother. The results of this extensive study shows that, the focus ofdevelopment of each CMIP phase being unique, a model ofCMIP5 could also have the potential of outperforming themodels of CMIP6 in simulating climate variables, and that, inthe case of mainland India, a GCM of either of the two phaseswould be appropriate for reliable projection of location-specificfuture temperature under different climate change scenarios.