The reason for optimism is the service's new Cray T3E supercomputer, capable of carrying out 80 billion calculations per second. Five times faster that its existing Cray model with which it is running in tandem before assuming sole command by the end of the year, the T3E will process information from thousands of worldwide automatic and human operated weather stations, satellites, aircraft and ships.

In the UK alone, 8,500 weather observations are produced each day from 3,000 sites. Even so, the fickle climate will still sometimes defeat the most advanced numerical forecasting techniques because the atmosphere is too large and complex to depict with total accuracy despite continuing advances in computer modelling.

Numerical forecasting drawing upon the principles of physics was first devised in the 1920s but proved of little use until the advent of computers because, by the time lengthy calculations were completed, the weather had already happened. Similarly, before Morse telegraphy developed in the 1840s no forecasts, however accurate, could be transmitted quickly enough to be of any use except in the immediate vicinity. Successful numerical models of the atmosphere require a vast number of calculations to be made in the very short time that only a computer can accomplish.

A spokesman at the Meteorological Office (or Met Office, for short) at Bracknell, 30 miles (48 kilometres) west of London, explained that where predictions failed - currently about one in seven for next-day forecasts in the UK - it was most often because an error in timing outwitted the human-machine mix making the calculations.

The Met Office, which employs some 2,200 people in 80 locations, issues nearly 3,000 tailored forecasts and briefings daily for major customers, such the Ministry of Defence Ministry and the Civil Aviation Authority, as well as commercial enterprises such as the offshore industry and shipping. Internationally, it participates in the voluntary cooperation programme of the World Meteorological Organisation for providing equipment and training to developing countries.

The service's first large commercial licensing and development deal for another national service is the present contract with Thailand to set up a supercomputing capability to begin forecasting by November 1998.

Bracknell's new supercomputer will also help research into the El Nino weather pattern, named after the Christ child because its impact is strongest around Christmas. It occurs when the trade winds that usually blow from South America to Australia subside, causing the warm waters of the Pacific to drift eastwards provoking worldwide weather chaos from drought to hurricanes, flooding and the destruction of crops, most seriously in the developing countries of Latin America, Africa, India and South-East Asia.

An alternative approach to forecasting has been developed by Weather Action, a company founded by Piers Corbyn, an astro-physicist who lectures in computing and mathematics at the South Bank University in London. A staff of only 11 full-timers in this competitor to the public sector Met Office, produce forecasts on premium telephone and fax lines for the public and more tailored predictions for insurance companies, power generators and retailers in the UK. Swedish firms in the petro-chemical and energy sectors have been among the first large overseas customers.

Mr Corbyn's system is described as enhancing the traditional methods that involve forecasting the present atmosphere forward in time by additionally identifying key changes in weather types that he sees as being caused by solar activity. Information on solar activity is collected from satellites and scientific sources, and forecasting - particularly long term - is largely based upon the effect it is believed to have on the distribution of high and low pressure areas that control weather types worldwide. The precise method of calculation is a trade secret, although it is intended to make it public eventually.

Mr Corbyn claims that his company's long-term forecasts far surpass in accuracy those of forward projection models by supercomputers, which he views as having a reliable time horizon of only three or four days. An independent study by the University of Sunderland, north-east England, took one aspect - Weather Action's gale forecasting data - and concluded that the possibility of the success rate being due to chance was one in a thousand.

The Met Office maintains that it keeps an open mind but that at present it does not feed solar activity into its calculations as it has not found any correlation that would affect forecasts. Competitive prediction claims are hard to evaluate because the basic rules are so difficult to define. One national newspaper lets it readers decide by printing Weather Action's forecasts alongside those of the Met Office.

Mr Corbyn, who is sceptical of the Met Office's claim of open-mindedness, makes no claim for infallibility. In his October 1997 long-range forecasting bulletin he explains that although Hurricane Erika's crossing of the Atlantic in September 1997 was correctly predicted by his service months ahead, the arrival of tropical storm remnants in Britain in the form of heavy rain had been lighter than expected.