Evaluation of dry and wet spell events over West Africa using CORDEX-CORE regional climate models

This study investigates the capability of regional climate models (RCMs) in simulating four extreme precipitation indices on an annual and monthly scale over West Africa during the period 1997–2014. Three global climate models (GCMs; HadGEM2-ES, NorESM1 and MPI-ESM) were dynamically downscaled using three high resolution (0.22∘\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.22^{\circ }$$\end{document}) regional climate models (RCMs; RegCM4, REMO2015 and CCLM5-0-15). These simulations were from the Coordinated Output for Regional Evaluations within the Coordinated Regional Climate Downscaling Experiment framework (CORDEX-CORE) publicly available through the Earth System Grid Federation (ESGF) web portals. The capabilities of the RCMs in the representation of maximum consecutive wet day (CWD), maximum consecutive dry days (CDD), number of dry days (NDD), and number of wet days (NWD) were compared with observation/satellites datasets obtained from the Global Precipitation Climatology Project (GPCP), Tropical Rainfall Measuring Mission (TRMM) and Tropical Applications of Meteorology using SATellite data and ground-based observations (TAMSAT). The reference datasets showed similar spatial pattern and magnitude of analyzed precipitation extremes but models exhibit different pronounced discrepancies relative to them. All RCMs consistently captured the spatial patterns of the indices but with some pronounced biases along the Guinean coast and northern parts of Niger. There exists little or no biases in the representation of annual cycle along the Guinea and Sahel for all the indices based on each of the RCMs ensemble, with the exception of RegCM4 which has a more pronounced bias in CWD. Statistical evaluation of the performance of the models over the entire West Africa with respect to the 4 indices revealed that REMO2015 models and its ensemble have overall lowest root mean square error followed by the choice of MPI-ESM GCM downscaled with either of the RCMs. REMO-HAD was found to have the best performance in the representation of consecutive dry days and number of wet days with RMSE values of 25.74 and 18.91 respectively. REMO-MPI has superior performance in the estimation of consecutive wet days and number of dry days with RMSE values of 5.38 and 20.51 respectively. Generally, REMO RCMs ensemble was found to be the best ensemble in all indices except consecutive dry days where REG4 ensembles had better performance. Operational use of these 3 RCMs are recommended with compensation for over-and underestimations.