Resolution dependence in simulating the African hydroclimate with the HadGEM3-RA regional climate model

Abstract

This study documents the effect of horizontal resolution on the ability of the Met Office third-generation Global Atmosphere Regional Climate Model (HadGEM3-RA), a regional atmospheric configuration of the HadGEM3 model, to simulate rainfall variability over Africa. It is based on six 20-year long RCM simulations driven by ERA-Interim reanalysis and performed at 12, 25, 50, 70, 90, and 150 km over the CORDEX-Africa domain. To gain further insight into model errors, we also compared the HadGEM3-RA’s performance to that of the parent General Circulation Model using three different spatial resolutions (70, 100, and 150 km), and to HadRM3P—the current Met Office regional climate model. It is found that the 50 km resolution RCM reproduces reasonably the spatial and temporal features of rainfall variability across regions. These include the seasonal progression of the tropical rainbelt, its extent and location, the annual cycle and interannual variability. Although model biases vary across seasons and locations, a prominent feature is the over-prediction of rainfall totals over Central Africa, and underestimation of rainfall in coastal areas of the Guinea Gulf during boreal spring and autumn. HadGEM3-RA improves with increased horizontal resolution, but some model errors persist. Comparison with the parent global model simulations demonstrates generally a realistic and consistent behaviour over large scales—suggesting that the physical formulation is able to capture the key driving processes, but also confirms the benefit of increasing the model horizontal resolution. Despite the model errors, HadGEM3-RA rainfall shows superiority over that from HadRM3P, ERA-Interim and MERRA datasets—indicating that the associated dynamical features of HadGEM3-RA can complement the physical understanding gained from reanalyses. This article also highlights the challenges for evaluating climate models in data sparse regions where satellite derived rainfall and gridded observational datasets often diverge.