ENVIRONMENT AND ENERGY

Project High Resolution Climate Modelling (HRCM)

Introduction

The African Continent will be severely hit by climate change: changing weather systems, rising temperatures, dramatic increase of droughts and more extreme events combine with growing population, lack of infrastructure, strong dependence on agriculture, ongoing exploitation and political instability. A necessary building brick for counteraction are reliable projections of the African climate of our century. Climate models are used to do this. While global climate models (general circulation models (GCMs), spatial resolution around 200 km) provide such data at global and continental scales, much more detailed information is required to assess the regional climate and to plan mitigation and adaptation measures. With their considerably higher spatial resolution (refinement by about a factor of 10 compared to GCMs), regional climate models (RCMs) can account for processes which cannot be “seen” by GCMs, like small scale orography, land-sea contrasts and varying land use (e.g. forest – open land), which affect precipitation, temperature and wind. They can also capture extreme events much more realistically.

However, there are differences between models, model setups and emission scenarios used, causing differing results. Therefore, it is necessary to do not just one run, but several runs (an ensemble) which are analysed statistically. This makes regional climate simulations very expensive even on powerful machines such as available at HLRS.

The CORDEX CORE Africa Simulations

For some regions like Europe, a quite large ensemble of high-resolution regional downscaling simulations does already exist. However, for regions like Africa, not many such downscaled results are available. Within the Sixth Climate Assessment Report of the Intergovernmental Panel on Climate Change (IPCC AR6), there will be a special chapter presenting an atlas of regional climate change information for all land regions of the earth.

For this purpose, the CORDEX CORE initiative (https://cordex.org/experiment-guidelines/cordex-core/) was designed. CORDEX is the acronym for the COrdinated Regional Climate Downscaling EXperiment (www.cordex.org), which has been initiated by the World Climate Research Programme (WCRP) in 2009, and the CORDEX CORE regional climate model integrations project provides climate and climate change data for the CORDEX CORE regions, among them CORDEX CORE Africa. Africa has been chosen as a key target region due to its high vulnerability to climate change and lacking quantitative climate change information.

Each modelling group contributing to CORDEX CORE is expected to downscale three different global models to a resolution of 0.22° (about 25 km) for the CORDEX repository. Two emission scenarios covering the range of potential future developments should be used: a) the RCP8.5 emission scenario with an additional radiative forcing by anthropogenic greenhouse gas emissions of 8.5 W/m² by 2100. It is called a worst-case scenario (but actually is quite realistic) with no implementation of emission reduction measures, and b) the RCP2.6 emission scenario which is a lower emission scenario ending up at 2.6 W/m² towards the end of the 21^st century. The three GCMs to be used as forcing for the RCMs are chosen as to represent low, medium and high climate sensitivity, which means a low, medium or high increase of the global mean temperature for a given increase of the greenhouse gas emissions.

Under these prerequisites, the working group "Regional Climate and Water Cycle" of the Institute for Meteorology and Climate Research - Department Troposphere Research (IMK-TRO) at the Karlsruhe Institute of Technology (KIT) (www.imk-tro.kit.edu) used the regional climate model COSMO-CLM (www.clmcommunity.eu) on the CRAY XC40 Hazel Hen system (and later on on Hawk) of HLRS to contribute to the CORDEX CORE Africa project.

Simulation setup

Our new simulations (called AFR-22) use an updated model version with increased horizontal (0.22°) and vertical resolution. In addition, several tuning parameters were optimized to reduce the model bias. Test simulations indicate indeed a reduced model bias in all seasons and different parts of Africa, although a systematic bias persists.

Overall, ten simulations were performed for the AFR-22 ensemble with the new setup:

• An evaluation run for the period 1979-2016.

• Three historical simulations for the period 1950-2005 forced by the NorESM1-M (low climate sensitivity), MPI-ESM-LR (medium climate sensitivity) and HadGEM2-ES (high climate sensitivity) GCMs (our ensemble).

• Three RCP2.6 and three RCP8.5 scenario simulations for the period 2006-2100 using the three GCMs mentioned above.

The simulations took about half a year on the Hazel Hen and produced more than 750 Terabyte of data.

Results

End of the 21^st century climate change signal

Fig. 1 shows the AFR-22 ensemble mean climate change signals for temperature and precipitation over Africa for the high emission scenario RCP8.5 towards the end of the 21^st century compared to the reference period 1971-2000. The results indicate strong changes in the western/equatorial part of Africa from the Guinea coast and Sahel region over Western Central Africa to Namibia and South Africa. In these regions, annual temperature increases by more than 5°C. Even more serious, there is a strong reduction of the annual precipitation amounts for these regions. Dryer conditions are expected for most of Africa, except for the Horn of Africa in the Equatorial East. The ensemble spread is low to moderate in the areas of the strongest changes, indicating robust results.

Temporal evolution and dependence on scenario and GCM

Fig. 2 displays the climate change signals of the ensemble members for both emission scenarios in the Guinea / West Africa-North region (WA-N in Fig. 1). It shows the changes over 30-year climatological time periods throughout the 21^st century in 10 year intervals compared to 1971-2000. Fig. 2 shows that:

• the differences between the scenarios are low to moderate in the first half and increase considerably towards the end of the 21^st century.

• temperature and precipitation changes are greater for HadGEM2-ES (higher global climate sensitivity) than for MPI-ESM-LR (medium) and NorESM1-M (low sensitivity).

• with RCP8.5 the temperature changes increase by more than double that of RCP2.6.

• for precipitation, all simulations indicate a reduction for all periods. The drying is much more severe with RCP8.5 than with RCP2.6.