The Met Office has been promised £1.2 billion from the government for an upgraded supercomputer that will create a digital twin of the Earth’s atmosphere and drive faster weather and climate change predictions.
The government funding announcement comes 10 months after the Met Office, which is home to the world’s twenty-third most powerful supercomputer, said it had finished a procurement exercise that had seen it explore the idea of switching to a cloud-based system, as part of new HPC infrastructure.
Ultimately, it said in May 2019, it had settled on exploring an “integrated solution including physical hosting and all necessary ancillary infrastructure (encompassing network and mechanical and electrical services), supercomputer provision, data storage and ancillary tools/services.”
The Met Office new supercomputer, to be operational from 2022, is expected to be a six-fold improvement on the currently used Cray XC40 system. (That existing two petabyte system is capable of 14,000 trillion operations per second. It is able to amass 215 billion weather observations daily).
The region’s leading universities, the GW4 Alliance, together with the Met Office, HPE, and partners, added today that they have been awarded £4.1 million by the Engineering and Physical Sciences Research Council (EPSRC) to create Isambard 2, the largest Arm-based supercomputer in Europe.
This will be hosted by the Met Office in Exeter. It will double the size of GW4 Isambard, to 21,504 high performance cores and 336 nodes. It will incorporate technologies from HPE and new partner Fujitsu, including next-generation Arm CPUs in one of the world’s first A64fx machines from Fujitsu.
Supercomputing technology is driving improved forecasting as the Met Office now does weather predictions every hour, as opposed to its previous three hour predictions. The recent severe storms, Ciara and Dennis, that battered the UK were all forecast five days in advance thanks to improved technology.
Early warnings help emergency services and local authorities put life-saving measures and procedures in place before conditions become too hazardous for workers.
Professor Penny Endersby, Met Office chief executive said: “This investment will ultimately provide earlier more accurate warning of severe weather, the information needed to build a more resilient world in a changing climate and help support the transition to a low carbon economy across the UK.”
The cited £1.2 billion is the expected investment from government over the next ten years. The actual contract value of the supercomputer operating capability is set at £864 million.
Met Office Supercomputer
Due to a more accurate and detailed simulation of atmospheric factors around the Earth the new system will be instrumental in helping to predict not just storm formation, but will be able to provide researchers with a longer forecast of the effects of climate change and help mitigate some of the damage.
When the Cray system is tasked with creating a digital twin of the atmosphere it divides the earth into a gridded square frame, each square representing 10kms, with more scrutinised grid sections, such as the UK, detailed at 1,500m across. However, the proposed supercomputing system is expected to be able to reduce the 10km grids down to 100ms, giving far more detailed simulations.
The chairman of the Science Review Group Professor Ted Shepherd stated in a release: “The improved processing power will deliver a step-change in weather forecasting and climate modelling capability for the UK, such as the further development of the Earth Systems Model, which involves collaboration with the many UKRI-NERC funded research centres.”
Further Supercomputing Funding Pledged
The government has also pledge to invest £30 million into supercomputing services that are operated by universities across the UK and will support seven High Performance Computing (HPC) programs at places like Queen’s University Belfast, the University of Edinburgh, and Durham University.
Using these systems researchers are discovering new insights into chemicals and materials which can result in improved drug therapies or construction capabilities. For instance the Materials and Molecular Modelling Hub at Queen Mary University of London uses these systems to understand surface degradation on a range of materials and is helping to develop next generation materials that will be used in solar energy technologies.
An AI and data science program led by the University of Oxford will receive £5.5 million to run a Molecular Dynamics program which will help advance our understanding of the structure and function of large biological molecules.
Theses service will provide a computing resource that can be used by the UKRI Artificial Intelligence Centres for Doctoral training.