Ross Clark forecasts that in spite of its new £150 million headquarters the Met Office will still get the weather wrong
Guests invited to the official opening of the Met Office's spanking new £150 million headquarters outside Exeter should take with them an umbrella. Or perhaps a sunhat. Or a thick coat. Or maybe just bung your entire wardrobe in the back of the car just in case. One thing is for sure: you won't get a lot of guidance from the weather forecast.
Much has been written about the vast and accelerating quantities of taxpayers' money being poured into health and education with little obvious benefit to the public. Less has been made of the expansionism in smaller government departments and agencies. The Met Office is a case in point. Besides its new headquarters and a new corporate logo consisting of green wavy lines on a blue background, the Met Office is about to take delivery of one of the world's largest supercomputers. According to the organisation's blurb, the computer will cover an area of 864 square metres and have the capacity to store one petabyte of data, which, it explains, is equivalent to one million gigabytes. Even if you don't know what a gigabyte is, when you consider that Nasa put a man on the Moon using computer power inferior to that of the crummiest machine you can now pick up at PCWorld, you will get the idea that the Met Office's machine is a jolly big computer indeed.
Notionally, the Met Office's extravagance is of little concern to taxpayers. In 1996 it was made a 'trading fund' of the Ministry of Defence, meaning that it was expected to earn its keep by selling its services commercially. Detailed weather forecasts were moved on to premium-rate phone lines, and long-range forecasts touted to umbrella manufacturers and ice-cream salesmen. Yet seven years on, a mere £21 million of the Met Office's £154 million annual income is raised by selling its services commercially. Many of the services it does sell are to other public-sector operations: the NHS, for example, pays it to forecast the average winter temperature several months ahead, so that NHS managers can predict the number of cases of flu. Given that an NHS winter flu crisis seems to have become a fixture, it is hard to see that the money would not be better spent directly on hospital beds.
The Met Office's claims to have improved the accuracy of its forecasting greatly over the past 20 years – a three-day forecast, it claims, is now more accurate than was a one-day forecast in 1980 – is certainly not borne out by anecdotal evidence. A fortnight ago, the British Association of Leisure Parks, Piers and Attractions complained when its members received only half the 80,000 visitors they would normally expect on a sunny spring Bank Holiday Monday. The missing visitors, it appears, had taken one look at the gloomy weather forecast for the day and forsaken the great outdoors for the plastic-pot-plant exotica of Lakeside shopping centre.
The Met Office does have its uses: thanks to the miserable Bank Holiday forecast, my children and I had the beach virtually to ourselves – and there was hardly a cloud all day. But I know how the British Association of Leisure Parks, Piers and Attractions feels. When you follow weather forecasts hour by hour, day by day, you quickly come to the conclusion that our state weather-forecasting service is employing the same methods as the horoscope writers. In a way they are never wrong. The trouble is that over the course of five days they give so many different forecasts that one of them cannot fail to be right.
To prove what I have long suspected, I decided to log all the forecasts for Cambridge leading up to last Sunday, 9 June. On Thursday evening, the prospects for Sunday in Cambridge were looking promising. A localised forecast on the Met Office website featured a light, fluffy cloud with the sun poking out from behind it. On Friday morning, the light fluffy cloud and the sun were still there, but a single black raindrop had appeared underneath. The forecast on BBC radio suggested that heavy rain would be clearing away from the south-east to give a bright day. Simultaneously on the telly, however, there was a big black raincloud lingering over Cambridge for most of the day.
By Friday evening the forecast for Sunday in Cambridge had changed to a big black cloud, two drops of rain and a bolt of lightning. On Saturday morning it was back to a light cloud with the sun poking out from behind it. By Saturday afternoon, it had changed again, to a dark cloud but with no raindrops. In the space of two days, in other words, the Met Office had issued every conceivable weather forecast save for a snowstorm and gales. As it turned out, on the Sunday it rained from five in the morning until half past eight, followed by a few sunny spells, light showers and a very strong wind. The Met Office will no doubt claim this as a victory, but the truth is that after following the forecasts I was left no wiser as to what the weather was going to do last Sunday than I would have been had I noted whether the cows were lying down.
One sceptical of the Met Office's claims for accuracy is Dr John Thornes, reader in applied meteorology at the University of Birmingham. Ever since the 1880s, he says, the Met Office has been claiming 85 per cent accuracy. The trouble is it has never published its methodology for arriving at such a conclusion. Four years ago Thornes decided to test the accuracy of Met Office forecasts independently and published his findings in Weather, the journal of the Royal Meteorological Society. For 65 days, he and his assistant sat and listened to the 17:55 weather forecast on Radio Four, and each day marked it according to eight criteria: whether the temperature was predicted to within two degrees, whether the wind was strong, medium or light, or whether it rained, and so on.
Sure enough, he concluded, when you tot up the points day to day, the Met Office forecast could be said to be 85.5 per cent accurate. But this is far less impressive than it sounds. You can achieve almost as high a score, he discovered, by the simple device of using today's weather report for tomorrow's forecast. By 'predicting' that the weather tomorrow will be the same as it has been today, you can score an accuracy rating of 76.6 per cent; the point being that the weather usually remains fairly similar for several days in a row.
The real challenge lies in predicting those occasions when the weather does undergo a sudden change. And on this, says Thornes, the Met Office is much less successful. When it comes to forecasting a change in the weather, he calculates, the Met Office's accuracy rating falls to 38 per cent. On the subject of whether it is going to rain or not – surely the most obvious reason why people take notice of weather forecasts – its record is worse still: a mere 30 per cent.
To be fair to the Met Office, it is doubtful whether anybody else could do any better. The weather in a temperate zone is by its nature unpredictable. There are too many factors influencing events in the lower atmosphere, no matter how many gigabytes of information you can collect. 'The other night I heard one of the forecasters admit that he didn't know what the weather was going to be like,' says Thornes. 'It was the first time I've heard them say that, and it was rather refreshing to hear them admit it rather than give a forecast which turns out to be wrong.'
Maybe so, but if the weather is simply too unpredictable to make a forecast, then why the need for the supercomputer and the posh new HQ? The only sure forecast is that large piles of taxpayers' money will be sweeping into the West Country, and there won't be much to show for it coming out the other side.