Understanding the role of local scale climate drivers over Africa using the GA3-based regional climate model

Climate variability has huge impacts on food security, water availability, human health and social and economic infrastructures, particularly in Africa where vulnerability to hazardous weather and the natural vagaries of the climate is already high. With the IPCC AR4 concluding that climate change is now virtually inevitable, clear research efforts are needed to better understand the driving processes of climate variability and change over Africa, and to improve their representation in the current generation of climate models. In this regard, we have developed the GA3-based regional climate model (RCM), a merged forecast and climate dynamical downscaling system derived from the latest configuration of the Met Office Unified Model (MetUM), in order to further examine the local and regional scales components of climate drivers. The present study investigates first the ability of the GA3-based RCM to reproduce the key features of the mean climate and variability over Africa using the CORDEX framework, a novel framework for evaluating and improving regional climate downscaling techniques. The GA3-based RCM is run continuously through 20-yr (1989-2008) integration, over a domain covering the entire African continent, with a horizontal grid-spacing of ~50km and 63 vertical levels, and using large-scale atmospheric conditions from ERA-Interim reanalysis dataset to drive the edges of the model domain. Our analysis focuses on intraseasonal and seasonal timescales, and uses a wide range of gridded observational datasets and metrics to assess the model performance. Secondly, we explore the effect of horizontal resolution on the model ability to simulate the characteristic of rainfall and the regional water cycle through series of repeated 20-yr integrations which are performed at 210km, 135km, 90km, 25km, and 12km horizontal resolution. We also investigate the effect of mineral dust, biomass burning, aerosols, convective, and land-surface-exchange processes trough sensitivity experiments with targeted changes in the physical parameterizations. Results indicate significant agreements between the GA3-based RCM and the observational datasets across Africa, despite the uncertainty in the observations. The RCM also shows improvement in comparison to the PRECIS model, that is the current Met Office dynamical downscaling system.