Savage downpours. Lightning strikes. Wretched scenes of flooding. The lowest Wimbledon attendances in a decade. Hailstones (not quite) the size of tennis balls, but big enough for children to throw at each other. Monotonous rain soaking the meadows, pounding the pavements, depressing the mood of the populace.

Where are the delicate clouds of English summer, the moderate breezes filled with barbeque smoke, the amber shafts of evening sun? Why aren’t the sounds of cricket and rap music permeating the horizon with remarkable clarity? Where is the elasticity of spirit that causes young women to wear strappy dresses?

June was the wettest since 1914, but it is difficult to say exactly why. It was suggested last week that cooler sea temperatures in the Pacific Ocean prompted by the La Niña weather system (sister to El Niño), were the likely cause of the rainstorms. But many experts say La Niña’s knock-on effect in the northern hemisphere is tenuous. The fact is, you can’t really isolate singular causes, pinpoint-style, in a vastly layered network phenomenon like weather.

The apparent rainfall anomaly (up to 300% in Yorkshire) could just be part of the normal flux and flow. “There’s a huge amount of random variability in the atmosphere and the question of cause and effect doesn’t make sense,” says Rowan Sutton of Reading University’s Walker Institute. “It could be something to do with anomalies in ocean temperatures – the north Atlantic is unusually warm at the moment – but again I wouldn’t like to say so. But we are sure climate change will increase the frequency of heavy rainfalls.”

The Meteorological Office (www.metoffice.gov.uk) says the same thing: “There is an expectation of heavier extreme rainfall events in most places as climate warms and the atmosphere becomes moister.”

That expectation relates directly to global warming on a long scale, but we may already be seeing the effects. As for the current storms, the Met Office says this morning might be fine but heavy showers will move north later. As for the rest of the week, “the cool and unsettled weather looks like continuing for the foreseeable future”.

But what does foreseeable actually mean? People have been grumpily asking a question that goes to the heart of all types of prognostication. The resulting fury when they get it wrong doesn’t mean you should shoot the messenger. I think weather forecasters are heroes, so difficult is the task they face. People assume that because technology has improved we have got the weather taped. But it is still elusive and multi-faceted, a constantly transforming, hydra-headed beast that forecasters must slay every day.

I suppose I’m biased. I’ve been exhaustively researching all of this for my new novel called Turbulence, about the team tasked with forecasting the weather for D-Day.

Why is it so hard to get it right? One answer is that weather itself is always mixing things up. That’s its job. It takes energy in the form of sunlight or wind and distributes it around the globe. Increasing and decreasing turbulence – eddies, vorticity, “whirling-ness” – is the motor for this action. There are also various types of conduit and corridor for weather throughout the atmosphere that in turn act as temporary barriers between evolving weather systems. Related to this is the fiendishly intricate interconnection of weather in different regions: the atmosphere is a complex holistic system, with inbuilt randomness.

Then there is another level of uncertainty because of incomplete measurement. It’s all rather like a game of snakes and ladders, or a tug of war between intelligibility and chaos. Scientists accept all this and many spend their lives trying to reduce the uncertainty in one way or another. The problem with forecasting is not the predictions themselves so much as the rest of humanity’s psychological inability to accept the uncertainty. We keep expecting everything to be stable, simple, coherent . . . but the weather won’t play ball.

Not that we should let those who communicate weather to us off the hook completely. In April the Met Office said there was a high probability that summer would be warmer than normal. This was in the context of a statement about global mean temperatures; 2007 as a whole, it said, “is likely to be the warmest ever year on record going back to 1850, beating the current record set in 1998”.

All this was mistaken by the media to mean “a hot summer in Britain”. But no one mentioned a washout. “The forecast for rainfall is less certain and currently there are no indications of an increased risk of a particularly dry or particularly wet summer,” the April posting said. No warnings of heavy rain were given until June 12.

“In fact, we did mean warm globally,” says Ewen McCallum, chief forecaster at the Met Office, “and perhaps it ought to have been clearer.” Try telling that to the spectators under umbrellas on Centre Court. But McCallum defends his position coolly. “It’s plain sad that people still denigrate forecasters. We just have to keep trying to educate them in what we do.”

Although it’s coolish now, the forecast of warmer average weather has actually been right. Despite the large amounts of rain across many parts of the country, the average UK temperature was above the long-term average last month. June had a mean temperature of 13.7C, while the long-term average for June is 12.6C.

Websites themselves, however, might be part of the problem. Modern methods of communication favour brevity and simplicity, and McCallum wonders if they are best suited for conveying complexity and uncertainty. “We supply Wimbledon with a text forecast. I’d rather it was otherwise, that the organisers sat down with a meteorologist to discuss the relative risk of playing on one day or the next. But that’s up to Wimbledon.” The All England Club actually has a cakewalk compared with the D-Day commanders. [General Dwight D] Eisenhower needed three to five days’ lead time to launch the largest ever mobilisation of military force. What are a few tennis players compared with 2.5m personnel, 500 ships and 3,000 landing craft? All to be moved in secret under specific cloud, wind and moonlight conditions. The odds against the conditions ever being met at all were 13-1. And the unexpected was what they got in spades, with the worst summer storms in 20 years in the Channel and exceptionally deep depressions.

British, American and Norwegian weather forecasters conferred by scrambled telephone for two fraught months. Led by a Scot, James Stagg, they spoke several times a day, arguing like scalded cats. The American forecasters believed they could deliver five-day-ahead forecasts simply by analysing historical weather data statistically; the Brits believed any forecast over two days ahead was speculative; the Norwegian on the team would commit only to evidence-based large-scale patterns. In the event, after advising a one-day delay, they said conditions would be tolerable for June 6. They were, but only by the thinnest of margins.

Complication is all very well and so are slight tolerances, but we do need specifics from our weather forecasters. Can we really get them? “Abso-lutely,” says McCallum. “We do a good deterministic forecast [for] up to five days. But beyond that it’s a case of probabilities. As you move from five to 15 days, the uncertainty increases. You start looking for the broad signal, the likelihood. The words to be used are not yes or no, but maybe, perhaps, probably . . .”

What forecasters actually do is collate data from a variety of sources (observatories, weather ships, balloons, satellites, ocean buoys) and feed them into dynamic computer models that run through mathematical scenario after mathematical scenario. The basis of modern computa-tional forecasting was first proposed by forgotten genius Lewis Fry Richardson in 1922, before electronic computers were available. The rates at which the air, water and heat of the atmosphere evolve are represented as numerical quantities in equations. The calculations start from an estimate obtained from measurements all over the world. With this data, computers “chase the future weather” over a grid of flow boxes wrapped round the globe.

The results are approximate because of the lack of detail (you can’t have weather stations everywhere) but, despite the inexactness of these approximations, advances in technology have brought about a great improvement in accuracy. “A three-day forecast now is as accurate on average as one-day forecasts were 20 years ago,” says McCallum.

To make it more difficult still, when you talk about accuracy, you have to be sure which accuracy. “With temperature and rainfall you are considering different things, however related, says Lord Hunt of Chesterton, one of Britain’s leading climate and weather scientists and an authority on fluid dynamics. “Accuracy of temperature forecasts for a month or even two months ahead is pretty good. But no one in the world can yet tell you precipitation in a month’s time in a specific place. That is a fundamental technical problem. What we do know is that climate change is increasing the intensity of these rainstorms.”

Even though all sensible scientists now agree that climate change is under way, none will blame any single current weather event on it. “It’s unlikely that at present we could relate any unusual pattern to global warming,” says Professor Brian Hoskins, who is renowned for his work on mathematical weather models that link the weather in other parts of the world with the erratic British climate. “We may be able in the next few years to say whether or not the likelihood of such a pattern occurring is increasing due to it. We have the expectation that the same patterns will produce more heavy rainfall.”

Despite the vagaries, three-quar-ters of the scientists I spoke to said climate change will produce more heavy rainfall. The Met Office agrees. Believe it or not, British weather modelling, in general, is among the best available anywhere. But computers can only deal with what you feed into them. Large errors in forecasts will still arise because of incomplete measurements and, as in any computation of a complex system, errors keep growing over time. That is one reason there has been so much argument about climate change.

“Richardson in 1922 was optimistic, thinking that one-day weather forecasts might emulate the perfect accuracy of the astronomical predictions in the nautical almanac,” says Hunt. “But in 1963, when Edward Lorenz [an American mathematician and pioneer of chaos theory] came to use an early electronic computer he found that its predictions were quite chaotic, even though he was calculating the behaviour of a much simpler set of equations than those needed to represent the atmosphere.”

When tiny changes were introduced into starting values, Lorenz found that large variations in the results grew rapidly. The implications of this profound idea were resisted for 30 years but have now become an accepted part of meteorological practice. Bundles of models called “ensembles” run the weather calculation again and again, with slightly different initial data. If all the results begin to converge in one direction, you have McCallum’s broad signal.

“Computer models are not perfect simulators of atmospheric motion,” McCallum sighs. “But using ensembles to give probabilistic outcomes we can turn that weakness into a strength. And we are getting more accurate. Weather forecasting has improved enormously in the past 20 years. However, we still have to get across the message of relative uncertainty with increasing time.”

The chief forecaster is right. Probably. Maybe. Possibly . . .

Giles Foden is author of The Last King of Scotland.