Met Office's new supercomputer to deliver £2bn of socio-economic benefits

The Met Office hopes to support £2bn worth of benefits to the UK economy through improved supercomputing capabilities over the next five years.

The UK’s national weather forecasting service last year received £97m from the government to fund a new Cray XC40 supercomputer. The investment grant is intended to enable £2bn worth of socio-economic benefits for the UK economy between 2016 and 2020.

Since 2011, the Department of Business, Innovation and Skills (BIS) has managed the Met Office as a "trading fund", which is required to operate on a commercial basis and meet agreed targets set by the department. In August 2015, the Exeter-based organisation lost the contract to provide forecasting services that it had held with the BBC since the 1920s.

Met Office and government officials calculated the projected £2bn return to the UK economy based on quantifying the impact that increased computational power would offer to six key weather-dependent sectors.

Almost half of the planned benefits (£933m) will come from an improved ability to predict the scale and timing of climate change. A further £526m will come through providing better capabilities to the renewable energy sector, which includes wind, solar and hydroelectric power, as well as nuclear power generation.

The Met Office estimates that potential benefits to the aviation industry could be up to £295m. For example, better forecasting for adverse weather conditions could enable airport controllers to plan for disruption and allow flight path optimisation to reduce fuel consumption.

Other industries expected to take advantage of the high-performance computing (HPC) facility include those involved in flood prevention - up to £242m worth of benefits - food suppliers (£104m) and organisers of winter transport logistics (£75m).

The industries are expected to benefit from improved seasonal and longer-term climatic prediction. Other advances include the increased resolution of models and the new Cray’s greater use of ensemble modelling, which samples the uncertainty of future conditions to calculate the reliability of forecasts. The forecasters anticipate they will be able to issue 24-hour weather warnings in situations where currently it is only possible to give 12-hour warnings.

According to the Met Office, the new supercomputer will be one of the fastest in the world. From March 2017 it will be able to perform more than 23,000 trillion calculations per second, or 23 petaflops.

The Cray will weigh 140 tonnes, almost three times as much as the existing hardware and the equivalent of 11 double-decker buses. With 480,000 CPUs, or cores, the Met Office estimates that it will have more power than 100,000 PlayStation 4 consoles.

Two petabytes (two million gigabytes) of memory embedded on chips – or 120,000 iPhones’ worth – will be able to run calculations 18 times faster than the current HPC set-up.

Seventeen petabytes of local memory, attached to the machine, will store data as it is calculated in and out. 

The mass storage, into which the outputs of a model are fed, is not attached to the machine but will be housed in a separate facility. This will be able to handle up to 800 petabytes of data. Liquid coolants will be used to maintain an optimum working temperature.

The Cray replaced the existing HPC capability, provided by an IBM Power 775 supercomputer, in August 2015, five weeks ahead of schedule.

The second phase will be completed in spring 2016 and is expected to provide six times the current processing power. From this point onwards the increased capacity will enable capabilities to be improved.

For the final phase, a new IT hall is being built in Exeter Science Park, at a cost of £22m. This will receive the new hardware in early 2016 and be fully operational a year later. A further £10m has been spent on mass storage for the outputs of climate research runs.

The Met Office paid £60m for the supercomputer itself and will fund operating costs from usual trading expenditure. Cray won the contract through open competition and began implementation at the end of 2014.