How accurate is the weather forecast? Explaining a few of its off-target predictions this summer, Met Éireann says people need to think in terms of probabilities, not certainties

THE BALCONY ON the third floor of Met Éireann’s pyramidal headquarters, on Glasnevin Hill in Dublin, was built to give panoramic views over the city. One morning this week, everywhere you look the sky is grey. We are living through one of those horribly wet Irish summers where the weather is the first topic of conversation.

The nerve centre of Met Éireann’s forecasting services contains banks of computers modelling the weather for the next 10 days – and the news is not good. Jean Byrne, Ireland’s best-known meteorologist, is looking at predictions computed by the European Centre for Medium-Range Weather Forecasts (ECMWF), to which Met Éireann and dozens of other national weather services subscribe. The computers have come up with 50 scenarios for what might happen in both the short term, which is to say the next three days, and the medium term, which is to say the next seven. Known as ensemble forecasting, it follows the principle that the more scenarios are modelled, the better the chance of coming up with the right one. None of the 50 is promising.

“There’s just low pressure everywhere,” says Byrne. Dublin could get between 1mm and 12mm over 12 hours on Tuesday, according to the predictions. In fact, no rain falls in that period. “If the atmosphere was completely stable, [the predictions] all should be the same in theory, but that never happens in practice, and this shows how uncertain the forecast is,” she explains.

Met Éireann’s head of forecasting, Gerald Fleming, shows the 10-day model for Athens. All the predictions converge: there’s not a cloud in the sky. Weather forecasters there must have nothing to do for four months of the year.

Closer to home, farmers, fishermen, hoteliers, holidaymakers and concertgoers are among those who depend on accurate predictions. When the forecasters say it is going to be sunny, they expect it to be sunny; when they say it is going to rain, they expect it to rain. There is no small amount of irritation when forecasters get it wrong. Heavy rain was forecast for the Dublin area last Thursday week, for example, which might have altered people’s plans for going to the Stone Roses concert in the Phoenix Park. It was the right forecast but the wrong day. The heavy rain fell on Friday. Last Saturday’s weather in Dublin was due to be largely cloudy with some showers, and a maximum temperature of 17-18 degrees. Instead the weather station in the Phoenix Park recorded no rain and temperatures of almost 20 degrees in more than 12 hours of sunshine. “The weather system moved very slowly and did not develop how we expected it to happen. The general character of the weather was correct, but there was an error in timing,” says Fleming.

Met Éireann’s computers, which are operated by the Irish Centre for High-End Computing, a State-supported technology agency, process many billions of bits of information each day. These include atmospheric pressure at different altitudes, wind speed and direction, moisture content, hours of sunshine and soil temperatures, among a plethora of other variables.

Fleming acknowledges that Met Éireann might do more to stress that forecasts are predictions, not certainties, even with all the computer power behind them. “We can describe the weather mathematically; we know the movement of the air obeys the laws of physics; but the equations are horrendously complicated,” he says. “We cannot know everything there is to know about the weather even now. We can’t establish the base line as to what today’s weather is with exactitude.”

Even with accurate forecasts, he says, there is many a slip between what the forecaster says and what the listener or viewer thinks was said. “There are a lot of steps where the information can get confused or lost or enhanced.”

Other weather services use percentage models to display the weather – saying, for example, that a location has a 70 per cent chance of rain. This is commonly used in the US and some European countries. “We don’t do it numerically in this country. We tend to do it verbally. A lot of people, we discovered, pay attention to the nuance, but ideally we should move to a numerical model,” he says.

The chance of a one or two-day forecast being accurate is between 85 and 90 per cent. That drops to 70 per cent at three days, but even the seven-day forecast has a 50 per cent chance of being accurate. The UK Met Office does a 10-day forecast but – wisely, given its battered reputation – has ditched its seasonal forecasting. Technology has improved accuracy somewhat. A six-day forecast now is as good as a five-day forecast at the turn of the millennium.

This summer’s weather charts have been dominated by low pressure. When pressure is low, air rises. As it rises it cools and any moisture in it condenses as rain. When the pressure increases, the air sinks and clouds evaporate. Fleming compares it to water going down a plughole. There is an element of unpredictability about how the water disappears. High pressure is much easier to predict. Low-pressure systems can move faster or slower than expected, take a different track, or deepen or lighten in intensity.

“We know the general trend, but there are all sorts of smaller-scale things that are happening that we know nothing about,” he says. “If rain is predicted tomorrow, I can’t tell you if there’s going to be a shower at Fairyhouse at a particular time. That level of prediction is too difficult.”

Can you trust the forecaster? Bad weather predictions

* On the evening before the worst storm to hit the UK for almost 300 years, the BBC forecaster Michael Fish breezily declared on the night of October 15th, 1987: “Earlier on today, apparently, a woman rang the BBC and said she heard there was a hurricane on the way. Well, if youre watching, don’t worry, there isn’t.” He was wrong, and the UK Met Office has never been allowed to forget it.

* The UK Met Office unwisely issued a press release in April 2009 about the forthcoming summer, predicting “barbecue weather”. That summer was a washout. The Met Office soon dropped its seasonal forecasts.

* In the big freeze of January 2010, Batt O’Keeffe, as minister for education, directed all schools to close for three days, based on a Met Éireann forecast of more snow and ice. No sooner had he made the order than a thaw set in, and he rescinded his decision.

Met Éireann issued an extreme-weather warning in October 2011 for heavy rain in the Dublin area. The rain materialised on the night of Monday, October 24th – with twice as much as predicted, causing widespread flooding. “When an extreme event happens, just getting the numbers right is very hard,” explained Gerald Fleming of Met Éireann.