Met Office readies £27m weather forecast

The Meteorological Office will go live next month with a £27.5m supercomputer capable of carrying out six times more calculations than its current system, improving forecasting and the study of global warming.

Since January, the Met Office has been running the new SX-6 system from NEC alongside its current Cray T3E computers.

In April it will shutdown the Crays and rely totally on the SX-6.

The new system will be capable of running six times the number of computations than the current machines, explained Steve Foreman, group head of development programmes at the Met Office.

"That will allow us to run many more climate change simulations," he said.

The Met Office takes measurements of temperature, wind, humidity and pressure from across the globe in order to model the atmosphere. By taking more measurements, the accuracy of forecasts can be improved.

But as well as taking more readings, the new supercomputer will have the power to use more complex models to improve long-range forecasting.

Over the summer, the Met Office will undergo extensive software validation to ensure the new models deliver improved forecasts.

"We want to make sure that we chose the best combination for improving our forecasts in the time available," said Foreman.

"There's no value in a forecast of tomorrow's weather if it takes until the next day to produce it."

While weather forecasts help people know whether to carry an umbrella on their way into work, they are also being used increasingly as part of business planning, said Foreman.

"Supermarkets now plan their purchasing based on weather information. As we improve the detail in the forecasts they will be able to make this planning far more specific," he added.

Starting in 2005, the Met Office will add further nodes of NEC's SX-6X next-generation supercomputers. This will take the number of nodes up to 45 and take up the space of two football pitches. It will also double the computational power.

The integration work is being contracted to Scalable Networks.