Towards Global Cloud-Resolving Weather and Climate Prediction at Exascale
TimeTuesday, June 26th2:29pm - 2:51pm
DescriptionGlobal high-resolution, that is kilometer-scale, simulations enable reliable and accurate weather and climate predictions. By resolving clouds and storms, uncertainties in the model parametrization are reduced. This is particularly expected to improve future climate models and thus enable a more accurate prediction of, for example, extreme climate events. Applying a kilometer-scale model and aiming at a production-ready simulation throughput of 1 simulated year per compute day will, however, result in a huge amount of data (in the order of TB and PB) and require exascale compute capacity.
By enhancing the productivity and efficiency of the entire simulation procedure, the Centre of Excellence in Simulation of Weather and Climate in Europe (ESiWACE) intends to advance numerical climate and weather simulation on high performance computing platforms. One major attempt is the establishment of high-resolution demonstrators, which showcase the computability on PetaFLOP supercomputers. With these demonstrators, the simulation of atmosphere-only and atmosphere-ocean coupled processes is to be conducted.
In my talk, I will present ESiWACE and address current developments of different models used in the community with regard to performance and scalability. I will further discuss a performance prediction approach which shall facilitate the inter- and, at a later stage, extrapolation of performance data to predict the actual run times of the complex (in terms of large number of parameters) weather and climate simulations. I further point at a recently evolving inter-comparison project that is meant to foster international exchange on global high-resolution models.