Where to for large-scale regional climate modelling in developing nations? Are the answers in the clouds?
Domaine:
climate
Through the Coordinated Regional Downscaling Experiment (CORDEX), the African domain enjoys a large model ensemble with over 14500 data files available on ESGF. These downscaled data were generated by 10 modelling groups and have resulted in over 270 research papers by the climate, hydrology, agriculture, health and other communities.
However, none of the CORDEX-Africa downscalings were performed by an African institution. Additionally, as CORDEX has begun downscaling CMIP6 GCMs, no modelling group has committed to the downscaling the African domain yet.
Developing countries scientists face barriers to conducting large climate downscaling experiments including lack of infrastructure (compute nodes, storage, air conditioning), erratic power supply, lack of support (for hardware and software) and of sufficient bandwidth. What potential solutions exist to overcome these barriers and capacitate developing nation scientists to conduct computationally expensive downscaling activities, including CORDEX CMIP6 downscaling?
Here we show how cloud computing offers a potential solution, not only to run regional climate models but also to analyze output from these simulations, all on remote infrastructure.
We present performance results from a test CORDEX-Africa downscaling run on Amazon Web Server (AWS) infrastructure forced by ERA5, some test results created on the cloud through JupyterHub, and project the resource cost of downscaling 10 CMIP6 GCMS over Africa. We discuss the benefits and challenges of this process and conclude by describing how this process, with sustained investment and support, has the potential to capacitate developing nation scientists to contribute to advances in regional climate modelling together with their developed nation colleagues.