20th and 21st Century Climate Simulations and Projections in Central Africa by CMIP5 Climate Models

Global and regional climate change assessments rely heavily on the Global Climate Model (GCM) outputs provided by the IPCC's Coupled Model Inter-comparison Project (CMIP5). In this study, we evaluate the ability of 25 GCMs to simulate historical precipitation and near surface temperature fields in Central Africa, apply a quantile-mapping based bias correction to monthly climate fields, and develop three-hourly, daily, and monthly bias-corrected fields for the period 1948-2099. The dataset, at 1.0o latitude/longitude horizontal resolution, is constructed by combining a suite of global observation and reanalysis based monthly and three-hourly data, monthly GCM simulations for the twentieth century, and twenty-first century projections for the IPCC medium mitigation (RCP45) and high emission (RCP85) scenarios. The GCMs simulate historical temperature better than precipitation, but substantial spatial heterogeneity exists among models. Many models show limited skill in simulating the seasonal evolution of present day precipitation, but none of them reveal changes in the seasonality in the future at monthly scale. We present the comparison of historical model performance by individual GCMs as well as several combinations of multimodel ensemble averages. Our results do not reveal any improvement in model performance between high- and low-resolution GCMs during the historical period. But, the multimodel averages of better performing models show greater skills in reproducing the historical climate over randomly selected GCM averages in Central Africa. Our analyses also show that the choice of GCM and emission scenario will dominate the uncertainty in climate change projections. Although our analyses are done for the Central African region, the final dataset is available for global land areas, which will be useful for a variety of climate impact, assessment, and adaptation studies.