Assessment of Extreme Precipitation Indices over India by CMIP6 Models

To simulate the extreme precipitation events through GCMs has become a challenge due to discrepancies in spatio-temporal resolution, physics, and parameterization schemes of the models along with deficiencies in the observed datasets. In this study, the performance of 27 CMIP6 models and their Multi model mean (MMM) in simulating extreme precipitation indices has been compared to the observed precipitation datasets (APHRODITE and IMD) over India during JJAS for 1975-2014. Meanwhile, the MMM shows a close agreement in simulating the indices derived from APHRODITE with PCC >0.6 for all indices with higher skill score (0.54), lower NRMSE than IMD. However, the MMM over- (under)-estimate the number of consecutive wet days (total precipitation) with a median relative error of 64% and 160% (5% and 20%) respectively, as compared to APHRODITE and IMD. Which inferred that similar biases still persist in the newly released CMIP6 GCMs with inter-observation dissimilarity in reproducing the indices. In general, the MMM is unable to replicate the very heavy precipitation (R20mm), with negative median relative errors. However, for all three aforementioned precipitation indices the extent of over- and under-estimation is less while comparing against the APHRODITE than IMD. For consecutive dry days (CDD), the MMM over- (under)-estimate over the North west (northern tip and peninsular as well as lee side of Western Ghat) parts of India, where the biases relative to APHRODITE (IMD) is large (less). The MMM simulates precipitation indices well, instead of using individual model. Whereas, the variation of NRMSE values of individual models are less with the exception of CDD and CWD, where the disagreement between the models with observation is large with larger interquartile model range. Comparing the relative errors between the different homogenous regions of India, all the regions are marginally performing good in simulating the different indices except the NW region, which is appended with larger relative error. It was worth noting that the models having higher spatial resolutions simulate the indices realistically with high (low) PCC (NRMSE), whereas the reversal is not valid for the worst performing models.

 

Key Words: Extreme Precipitation, CMIP6, MMM, IMD, APHRODITE