Energy-efficient Scalable Algorithms for Weather Prediction at Exascale (ESCAPE2)
ESCAPE-2 will develop world-class, extreme-scale computing capabilities for European operational numerical weather and climate prediction systems.
The new generation of Numerical Weather Prediction (NWP) capabilities are required to target global operational forecasts at 1 km horizontal resolution by 2025. Weather- and climate-related disasters have caused trillions in economic losses and nearly two million deaths globally since 1971.
ESCAPE-2 aims to attack three sources of enhanced computational performance of NWP models. Namely (i) developing and testing bespoke numerical methods that optimally trade off accuracy, resilience and performance, (ii) developing generic programming approaches that ensure code portability and performance portability for High Performance Computing (HPC) (iii) testing performance on HPC platforms offering different processor technologies.
ESCAPE-2 builds on the pioneering work of the ESCAPE project. ESCAPE2 consortium comprises one international European organisation, three national meteorological centres, two universities – including Loughborough –, two HPC centres, and one hardware vendor.
The research undertaken by Loughborough forms the basis of the project work package 1: Algorithms and Mathematics. This package aims to develop mathematical methods and implement advanced and disruptive algorithms suitable for extreme-scale parallelism that achieve major improvements in the accuracy, efficiency, fault-tolerance, and scalability of dynamical cores and of physical parametrizations for next-generation weather and climate prediction models.
Loughborough has already contributed to ESCAPE with the development of a class of scalable, energy efficient Krylov solvers bespoke for NWP. The concepts developed have been demonstrated in the form of “Weather & Climate Dwarfs”, as well as integrated into next-generation operational models. Going forward we will contribute to increased efficiency and reliability of Finite Volume Model (FVM) of IFS --- the novel nonhydrostatic dynamical core complementing the existing spectral transform hydrostatic dynamic core of IFS. IFS is a world leading Numerical Weather Prediction System for medium range forecasts used by ECMWF to provide weather forecasts to 34 counties and to monitor the Earth system.
Dr Joanna Szmelter, Reader in Computational Fluid and Solid Dynamics
“ESCAPE2 is the part of our long term collaborative research with the European Centre for Medium Range Weather Forecast (ECMWF) and other European partners leading to the next-generation operational models for weather and climate predictions. The numerical methodologies proposed by Loughborough have already made fundamental contributions to the developments of the highly accurate and efficient Finite Volume module of IFS. The next generation tools are needed to address the formidable challenge facing numerical weather predictions of resolving rather than parameterizing essential small-scale forcings and circulations in the multi-scale global flow in the Earth’s atmosphere.”