An ensemble of CORDEX-Africa climate projections simulated by RCA4

Within the CORDEX-Africa initiative a large ensemble of regional climate simulations over Africa has been produced at the Rossby Centre by dynamical downscaling of a subset of GCMs from the CMIP5 project.

Eight GCMs, based on availability of boundary conditions in the CMIP5 data base, have been chosen for downscaling (see Fig. 1). The downscaling of the GCM simulations has been performed with the latest version of the Rossby Centre Regional Climate Model - RCA4 for the whole African domain at 0.44º resolution (see Fig. 2). First we have downscaled the ERA-Interim Reanalysis (1980-2010) and then all 8 GCMs sampling both RCP4.5 and 8.5 climate change scenarios and running in transient mode for the period 1951-2100 (historical 1951-2005 and scenario 2006-2100). We have also downscaled one RCP2.6 scenario for 2006-2100 (EC-EARTH).

CORDEX Africa GCM location map
Figure 1: A locational map showing the subset of CMIP5 GCMs downscaled by RCA4 at the Rossby Centre (Institute (Country): GCM name).

Figure 2 shows GPCC6 annual mean precipitation and differences from GPCC6 in other observational data sets and the RCA4 simulations, as well as ERAINT and the RCA4(GCMs) ensemble average. A large wet bias evident in ERAINT is strongly reduced in the RCA4(ERAINT) downscaling, although a dry bias appears over the Congo where climatological precipitation maximum is located. We should note that almost no stations report observations in central Africa (Nikulin et al. 2012) that makes evaluation of the simulated precipitation difficult this region. From individual simulations one can see that, in terms of precipitation, RCA4 is not so sensitive to the driving GCMs and simulates similar precipitation climatology when driven by different GCMs. At the same time the driving GCMs produce very disperse precipitation climatology (not shown) with a large spread across the GCMs which is substantially reduced by RCA4. Similar evaluation but for near-surface temperature is presented in Fig. 3. RCA4(EARINT) pretty well simulates near-surface temperature with local biases of order 1-2ºC over most of Africa that is comparable magnitude to difference across the observational data sets. Downscaling of the individual GCMs, on average, shows a large-scale cold bias that is also reflected in the ensemble mean.

Figure 1
Figure 2: GPCC6 annual mean precipitation for 1980-2005 and differences compared to GPCC6 in the other gridded observations, the individual RCA4 simulations, and their ensemble average.
Figure 2
Figure 3: CRU-TS31 annual mean temperature for 1980-2005 and differences compared to CRU-TS31 in the other gridded observations, the individual RCA4 simulations, and their ensemble average.

An example of future climate projections simulated by the RCA4 ensemble over eastern Africa is illustrated in Fig. 4 (temperature) and Fig. 5 (precipitation). The temperature projections across the simulations are consistent and clearly differentiated by the RCP scenarios giving about 3.5ºC (RCP85) and about 2.0ºC (RCP45) increase in the ensemble mean temperature by the end of the century. One simulation with the RCP26 scenario projects about 1.0ºC warming.

Figure 3
Figure 4: Future climate projections in temperature averaged over East Africa for the October-November period (29º-41ºE, 11ºS-4ºN, only land). Temperature anomalies relative to the 1971-2000 period are smoothed by the 31-yr moving average.

Future projections in precipitation (Fig. 5) have larger uncertainties compared to the temperature ones. Large variability at multi-decadal time scales is clearly evident in the individual simulations, for example in RCA4(HadGEM2-ES). Although RCA4 simulates similar precipitation climatology downscaling different GCMs for the recent past (Fig. 2) the future climate projections in precipitation are defined by large-scale boundary forcing from GCMs.

Figure 4
Figure 5: Future climate projections in precipitation averaged over East Africa for the October-November period (29º-41ºE, 11ºS-4ºN, only land). Precipitation anomalies relative to the 1971-2000 period are smoothed by the 31-yr moving average.

Even if the ensembles mean shows an increase in precipitation over eastern Africa for both RCP45 and 85 scenarios some simulations do not show a gradual increase in precipitation towards the end of the century.

Analysis on the RCA4 ensemble over Africa is ongoing now that will lead to a set of articles describing the RCA4-Africa-CORDEX matrix in detail. We are planning to publish full RCA4-Africa-CORDEX matrix via the ESGF system ( by the end of 2012.


Jones C, F. Giorgi and G. Asrar, 2011: The Coordinated Regional Downscaling Experiment: CORDEX, An international downscaling link to CMIP5: CLIVAR Exchanges, No. 56, Vol 16, No.2 pages 34-40. Available from

Nikulin G. and co-authors, 2012: Precipitation Climatology in An Ensemble of CORDEX-Africa Regional Climate Simulations. J. Climate, 25, 6057–6078. doi:

List of acronyms

CORDEX - Coordinated Regional Climate Downscaling Experiment
CMIP5 - Coupled Model Intercomparison Project Phase 5 (
GCM – Global Climate Model
GPCC6 - Global Precipitation Climatology Centre gridded precipitation (
CRU-TS31- gridded temperature and precipitation from Climate Research Unit, University of East Anglia (
UDEL301 – gridded temperature and precipitation from Center for Climatic Research, Department of Geography, University of Delaware (
GHCN-CAMS – Climate Prediction Centre gridded temperature
ERAINT – ERA-Interim Reanalysis (
ESGF – Earth System Grid Federation (