Caroline Arnold
Caroline is a research data scientist with the German Climate Computing Centre DKRZ.
Sessions
Lightning Talks are short 5-10 minute sessions presented by community members on a variety of interesting topics.
Earth sytem models (ESMs) can simulate climate projections and weather forecasts. There is growing interest to replace components of the ESM by neural networks, to speed up simulations or to improve the quality of the predictions. We are working with the ESM ICON, which is executed in parallel on high performance computing systems. ICON is written in Fortran, while the neural networks are implemented in Pytorch. To bridge these worlds, we compile Python code into a library using the C Foreign Function Interface (CFFI). This library is then integrated into the ESM. In this talk, we share lessons learned by using CFFI in an ESM and in a parallel setting.