Find the EGU on

Tag your tweets with #EGU18


High resolution weather and climate models on large supercomputers (co-organized)
Convener: Peter Düben  | Co-Conveners: Valentine Anantharaj , Reindert Haarsma , Rieke Heinze , Xavier Lapillonne , Malcolm Roberts , Pier-Luigi Vidale 
 / Fri, 13 Apr, 13:30–17:00
 / Attendance Fri, 13 Apr, 17:30–19:00

The quality of predictions of weather and climate depends on both resolution and complexity of the models that are used. However, resolution and complexity are limited by the computational performance that is available on today's supercomputers. While weather and climate models run on some of the fastest supercomputers of the world, models typically fail to run close to peak performance such that there is still room for a significant speed-up if efficiency is improved. The increase in parallelisation in high performance computing and the availability of various computing platforms is imposing significant challenges for the community to find the optimal hardware/model configuration and to achieve the best performance. On the other hand, the evaluation of high resolution simulations is often tedious due to large data volumes, limited statistic that is affordable and changed model behaviour that needs to be studied (e.g. if convection or eddies are resolved explicitly or if non-hydrostatic equations need to be used).
These challenges can only be addressed appropriately in a close collaboration between Computing and Earth System Scientists. This session aims to bring together scientists who run and evaluate atmosphere and ocean models with high resolution and complexity as well as scientists who enable these models to run as efficiently as possible on existing and future high performance computing architectures (regarding both model development and model optimisation). The session will also be an opportunity for scientists from the EU projects PRIMAVERA, ESCAPE and ESiWACE as well as HighResMIP from CMIP6 to meet and interact.

There will be an invited talk by Bjorn Stevens.