Evaluation of the Performance of HighResMIP CMIP6 in Simulating Extreme Precipitation in Madagascar

Abstract

This study assesses the performance of 17 High-Resolution Model Intercomparison Project (HighResMIP) from the Coupled Model Inter-comparison Project Phase 6 (CMIP6) and their ensemble mean in simulating extreme precipitation in Madagascar. For this purpose, nine extreme precipitation indices were used, namely consecutive dry days (CDD), consecutive wet days (CWD), heavy precipitation days (R10mm), very heavy precipitation days (R20mm), simple daily intensity (SDII), maximum 1-day precipitation (RX1day), maximum 5-day precipitation (RX5day), very wet days (R95P) and extremely wet days (R99p). Furthermore, two gridded data sets, the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) and ERA5, were employed as the reference data. The performance of the models was assessed using Normalised Root Mean Squared Error (NRMSE), percentage of bias (PBIAS) and Taylor Skill Score (TSS). In addition, this study used a comprehensive model ranking (MR) to provide an inclusive assessment of the models. The results suggest that most HighResMIP models fairly reproduce the precipitation climatology in Madagascar. We also found that except for the simulation of R99P, most of the models have satisfactorily good results in simulating extreme precipitation indices over the study area. A further comparison between HighResMIP models and their CMIP6 counterparts showed that most of the high-resolution models have better performance; yet improving model parametrization is still important. In line with previous research, by crossing the results from CHIRPS and ERA5, we found that the multi-model mean outperformed individual models. Nevertheless, individual models such as HadGEM3-GC31-HH, HadGEM3-GC31-HM, ECMWF-IFS-HR and EC-Earth3P-HR have relatively good results. On the contrary, the worst performance is attributed to HIRAM-SIT-LR and INM-CM5-h. The outputs and results from this research deliver a comprehensive assessment of the performance of the new HighResMIP in simulating precipitation extremes at a local study, which are essential for policymakers and climate modellers.