Aided by a battery of new supercomputers and Nasa's most advanced weather satellite, they have battled valiantly to predict every twist and turn of Britain's wettest summer in half a century. From the deluge that devastated the Cornish village of Boscastle to the stifling humidity that created a plague of hoverflies in East Anglia, the one constant of the fickle British climate has been the attempt by Met Office forecasters to give reasonable warning of the vagaries of the elements.

Aided by a battery of new supercomputers and Nasa's most advanced weather satellite, they have battled valiantly to predict every twist and turn of Britain's wettest summer in half a century. From the deluge that devastated the Cornish village of Boscastle to the stifling humidity that created a plague of hoverflies in East Anglia, the one constant of the fickle British climate has been the attempt by Met Office forecasters to give reasonable warning of the vagaries of the elements.

But when it comes to providing a reliable idea of what the weather will be like in five days' time, the Met Office has admitted that its computer-generated forecast is about as dependable as an August bank holiday heatwave. Amid criticism from academic researchers and rival forecasters, the state weather agency confirmed yesterday that its popular online five-day weather service, used by thousands to plan their weekend activities, cannot be relied on to give an accurate forecast.

The admission comes after academics called for all Met Office predictions to be made the subject of independent scrutiny and for them to avoid over-simplification of forecasts. One survey found that over a three-week period, two-thirds of the Met Office five-day forecasts failed to remain consistent, in one case making a prediction for one 24-hour period of heavy thunderstorms that then changed to clear sunshine and then to rain in successive days.

Jim Galvin, senior national forecaster for the Met Office, said that the problems came not with predicting weather for the next day but at a distance of 120 hours. He said: "At anything up to three days our accuracy is pretty good. But after that there is a statistical divergence that becomes magnified. It becomes particularly difficult at day five, when a precise prediction is virtually impossible."

Those expecting the predictions to be the work of an avuncular weatherman should also reconsider. Rather than being the result of a careful analysis by a trained meteorologist, the five-day forecasts are automatically produced by a computer model. The result is a basic graphic for a specific location, such as London or Glasgow, giving day and night temperatures and a pictogram to indicate the prevailing conditions. Other meteorologists complain that this implies certainty when in fact it is statistical guesswork that may change every 24 hours.

Andrew Bond, senior forecaster for Met Check, an independent forecasting company, said: "The Met Office relies on outmoded methods. If you look at the American system, they can provide data that looks 180 hours or 7.5 days ahead. The Met Office is working below that."

Dr John Thornes, a reader in applied meteorology at the University of Birmingham, where researchers have cast doubt on the Met Office's accuracy, said: "There ought to be independent verification of the forecasts."

Even the Met Office's harshest critics concede that extreme weather events, such as the 180mm of rain which fell in five hours to cause the catastrophic flooding of Boscastle on 17 August, cannot be predicted to within a matter of minutes, let alone several days. But the qualms over the ability of the Met Office, which this year celebrates its 150th anniversary, to predict the medium-term weather come despite a significant upgrading in recent months of its equipment.

The installation of two new supercomputers has increased the amount of data that can be processed eightfold. Forecasters said the equipment would more than halve the number of wrong predictions, eliminating blunders such as the failure to foretell the Great Storm of 1987.

Experts point out that the investment, which is expected to help improve accuracy of medium-range forecasting such as the five-day predictions, would also benefit the Met Office's balance sheet. One marketing analyst said: "In commercial terms, an accurate five-day forecast is extremely valuable. If you are a supermarket and can reliably predict the number of lettuces or boxes of lager you'll need because of a hot weekend, then that is very useful information. The Met Office has a lot of commercial clients who want to know the information they're paying for is more than a best guess."

In its defence, the Met Office points out that accuracy has improved threefold in the past 20 years, extending the range for accurate forecasting from 24 hours in 1980 to more than three days in 2004. Mr Galvin said: "Being criticised comes with the territory if you are a weather forecaster. But at least we got the bank holiday right - we said there would be showers spreading into the east with sunshine. And that's broadly what's happened."