Evaluation and correction analysis of the regional rainfall simulation by CMIP6 over Sudan

Abstract

This study utilizes satellite-based rainfall CHIRPS to evaluate GCMs-CMIP6 models over Sudan from 1985 to 2014. Overall, the GCMs of BCC-CSM2-MR, CAMS-CSM1-0, CESM2, ECEarth3-Veg, GFDL-ESM4, MIROC-ES2L, and NorESM2-MM are well reproduced in the unimodal pattern of June to September (JJAS), and hence employed to calculate Multi-Model Ensemble (MME). Then, we examine the capability of the GCMs and MME in replicating the precipitation patterns on annual and seasonal scales over Sudan using numerous ranking metrics, including Pearson Correlation Coefficient (CC), Standard Deviation (SD), Taylor Skill Score (TSS), Mean Absolute Error (MAE), absolute bias (BIAS), and, normalized mean root square error (RMSD). The results show that the MME has the lowest bias and slightly overestimates rainfall over most parts of our study domain, whilst, others (ACCESS-CM2, BCC-CSM2-MR, CAMS-CSM1-0, CESM2, CNRM-CM6-1, CNRM-CM6-1-HR, CNRM-ESM2-1, FGOALS-f3-L, FGOALS-g3) consistently overestimate rainfall in referring to CHIRPS data, respectively, but FIO-ESM-2-0 underestimates bias value. Moreover, MIROC-ES2L and NorESM2-MM demonstrate better performance than the other models. Finally, we employed a bias correction (BC) technique, namely Delta BC, to adjust the GCMs model products through the annual and monsoon seasons. The applied bias correction technique revealed remarkable improvement in the GCMs against the observations, with an improvement of 0 - 18% over the original. However, MME and MIROC-ES2L show better performance after correction than other models.