|Name:||Weather Prediction & High Performance Computing|
|Time:||Thursday, June 26, 2014
09:30 am - 10:00 am
CCL - Congress Center Leipzig
|Breaks:||07:30 am - 10:30 am Welcome Coffee|
|Speaker:||Oliver Fuhrer, MeteoSwiss|
|Abstract:||The societal impacts and economic benefits of weather information for the next hours, days, weeks, seasons and even years have been rapidly increasing over the past decade. Numerical weather prediction (NWP) uses mathematical models to simulate the atmosphere and oceans and to predict the weather forecast based on current weather conditions. These model simulations typically have severe constraints in terms of time-to-solution in order to be useful for dissemination, and thus require large high performance computing (HPC) systems both for operational and research applications. There are three main drivers increasing the computational burden, increase in mesh resolution, increase in number of ensemble members for uncertainty quantification, and increase in model complexity. With the example of the COSMO model used by several operational weather services and research institutions worldwide, we will illustrate the typical model applications, the underlying algorithmic motifs and discuss the learnings from a recent effort to adapt the code base of COSMO to hybrid HPC architectures.