Computer-generated weather forecasts could become more accurate thanks to an initiative just launched by the universities of Cambridge and Oxford in collaboration with the Meteorological Office.

Cambridge's Isaac Newton Institute for Mathematical Sciences has brought together 60 experts from mathematical, meteorological and oceanographic university departments and research institutes around the world - including Leeds, Reading and Imperial College, London.

The idea is to examine the mathematical concepts underpinning computer models used for weather forecasting. Ian Roulstone, programme coordinator at the Newton institute, says that this could provide novel insights into how, for instance, large-scale weather systems would alter if initial data such as wind speed and direction, temperature, humidity and cloud changed.

He explains: "Given the current state of computer software and hardware we cannot hope to resolve all the fine-scale structure of real weather conditions.

ÖйúAƬ

"Using some of the latest mathematical ideas, we want instead to identify and quantify the dominant processes that govern the behaviour of large scale weather patterns."

The work could lead to improved forecasts for extreme events such as severe weather storms. It could also help to explain how errors in forecasting specific events such as blizzards, dense fog and heavy rain grow over time.

ÖйúAƬ

Dr Roulstone says that currently the same numerical models are used to forecast the weather a few hours ahead as well as thousands of years ahead in climate change studies.

The researchers aim to identify which features of the atmosphere have to be represented most accurately by numerical models when they are used on these different timescales.