Drought signals in CORDEX-Africa RCA4 climate simulations and CMIP5 Projections

The Rossby Centre has produced a large ensemble of regional climate simulations by downscaling a subset of CMIP5 GCMs under the CORDEX-Africa initiative. Can this, as well as an ensemble of the driving GCMs, give us a hint about the possible future of one of Africa’s drought hotspots? Visiting scientist to the Rossby Centre, Bode Gbobaniyi, explains.

The African continent, which is now home to over a billion people, is vulnerable to the adverse impacts of a changing climate. This vulnerability is further compounded by economic constraints and dearth of strategic initiatives for adaptation and mitigation of emerging climatic threats. Many economies whose livelihoods and sustainability depend on rain fed agriculture are periodically experiencing drought or threatened by drought conditions.

Eight atmosphere-ocean general circulation models (AOGCMs) from the fifth phase of the Coupled Model Intercomparison Project (CMIP5) were downscaled over the Africa-CORDEX domain (Jones et al. 2011) at the Rossby Centre (SMHI) by a regional climate model, RCA4. The entire simulated period spans 1951-2100 with the first 54 years (historical: 1951-2005) forced according to the non-SRES 20C3M scenario with greenhouse gases increasing as observed through the 20th century. The remaining period (2005-2100) were forced by the representative concentration pathway scenarios RCP4.5 and RCP8.5 with boundary conditions and sea surface temperatures (SSTs) from the eight AOGCMs.

Various indices have been devised for characterizing drought. The Standardized Precipitation index (SPI) formulated by McKee et. al (1993) is widely used. We explore the sensitivity of projections of future drought in Africa to the SPI and two other indices of moisture availability namely the moisture index (MI) and the difference of precipitation and evaporation (P-E), over the so called drought hotspots of Africa. Here we show as an example, some SPI results over the Sahel.

Spatial Variations

Figure 1 shows the three-month standard precipitation index (SPI-3) over Africa for the July – September (JAS) season for the late 21st century period 2070-2100 with respect to a control “present day” period 1971-2000. Mean precipitation for the control period, inter-model spread, as well as agreement maps signifying the numbers of models showing an increase in SPI-3 is presented for both the RCM and underlying GCM ensembles. In the boreal summer (JAS), the slightly wet signal over the Sahel in the GCM ensemble is amplified in the downscaled simulations over the central region with moderately wet conditions all the way down to the Guinea coast. The inter-model spread is pronouncedly lower than in the underlying GCMs and there is more agreement on the projected increase amongst the regional ensemble members.

CORDEX Africa Precip Index
Figure 1: Ensemble mean precipitation (left) in the control period 1971-2000 and SPI-3 change (second from left) in the RCP8.5 scenario comparing 2071-2100 to 1971-2000 for the three-month season July-September. The spread between the models is shown on the third column as the standard deviation and model agreement is shown on the right. The upper row shows RCA4 simulations and the lower row the corresponding results from the underlying GCMs.

Temporal evolution

The simulated temporal evolution of SPI-3 over the Sahel region of West Africa reveals a wide spread in the GCM ensemble members which is considerably reduced in the RCA4 ensemble (Figure 2). With regard to this, some RCA4 ensemble members change both in sign and amplitude of the signal compared to the driving GCMs. The RCP4.5 and RCP8.5 emission scenarios start to diverge by the middle of the 21st century in the downscaled projections but this is not evident in the underlying GCMs. The time series also show considerable decadal variability. Interestingly, when compared with the driving GCMs, the ordering of the emission scenarios in terms of which gives the strongest climate change signal is reversed in the downscaled ensemble mean with RCP4.5 indicating a higher amplitude of wetness by the end of the century.

Temporal Evolution SPI
Figure 2: Change in SPI-3 over the 1960-2100 period with respect to a reference climatology (1971-2000) for the JAS season in the Sahel. The data has been smoothed using a 31-year moving average. The upper panels show RCA4 results and the lower ones the underlying AOGCMs.

Changes with temperature increase

Global mean temperature through the 21st century as simulated by CMIP5 models (within the envelope of all RCP scenarios) could increase by up to 4 °C by the end of the century (Knutti & Sedláček 2012). The regional change in temperature could even be higher than the global mean in certain parts of the world including Africa. A common trend in the regional models ensemble for all regions is that the amplitude of change increases progressively with increasing temperature from the early decades until the end of the century (Figure 3). The amplitude of change is higher for the RCP8.5 scenario than the RCP4.5 scenario. In West Africa (Figure 3) as well as East Africa (not shown), SPI-3 in the downscaled ensemble indicate a progressively wetter Sahel although this progression is not detectible in the underlying GCM ensemble mean. In Southern Africa, however (not shown), the regional ensemble mean SPI disagree on the sign of change, the later indicating slightly drier than normal conditions while the former indicates a slightly wetter than normal regime. Each respective regional ensemble mean index also disagree with the corresponding underlying GCM ensemble mean in terms of the sign of the projected change. It is difficult to qualitatively conjecture the reasons for such a disagreement between downscaled output and the driving models. Also, the spread of the underlying GCMs is quite wide but noticeably reduced in the downscaled ensemble members. This reduction in spread is strongest over West Africa.

Temp change Sahel
Figure 3: Change in SPI-3 versus temperature for the peak rainy season (July-September) in the Sahel. Panel to the left show the RCA4 simulated change while the right panel show the underlying GCMs (indicated by 1-8 in the legend). Changes are shown over three different time periods and for theRCP4.5 and RCP8.5 emission scenarios.


The amplitude of the drought signals, where present, are higher in the RCP8.5 scenario than the RCP4.5. Also the spread between members of the underlying GCM ensemble is reduced in the regional simulations and this varies from region to region. This was also found by Nikulin et al (2012) to be the case for downscaled historical simulations over Africa with RCA4. For more robust conclusions, we will have to wait for downscaled simulations from other regional models involved in the CORDEX-Africa initiative. However, there is now a basic framework to build on more rigorous assessment of regional simulations and their underlying CMIP5 GCMs towards obtaining a coherent and actionable message of change for the highly vulnerable African continent.


Jones C, Giorgi F, Asrar G (2011) The coordinated regional downscaling experiment: CORDEX. an international downscaling link to CMIP5. CLIVAR exchanges 16:34-40

Knutti R, Sedláček J (2012) Robustness and uncertainties in the new CMIP5 climate model projections. Nature Climate Change

McKee TB, Doesken NJ, Kleist J The relationship of drought frequency and duration to time scales. In. Proc Proceedings of the 8th Conference on Applied Climatology. American Meteorological Society Boston, MA

Nikulin G, Jones C, Giorgi F, Asrar G, Buchner M, Cerezo-Mota R, Christensen OB, Deque M, Fernandez J, Hansler A, van Meijgaard E, Samuelsson P, Sylla MB, Sushama L (2012) Precipitation Climatology in an Ensemble of CORDEX-Africa Regional Climate Simulations. Journal of Climate 25:6057-6078