Back the Favourite - What Are the Odds for a Good Long Range Forecast?

27/05/2015 16:53 BST | Updated 27/05/2016 10:59 BST

Long-range forecasts have been a political hot bed for meteorologists for some time. People want to know what the weather will be like months in advance to plan their summer holiday or to find out if there will be clear skies on their wedding day. Unfortunately for those brides to be and staycation planners, this is not as simple as it may sound.

So many people are under the misapprehension that somehow, modern meteorologists can give a detailed forecast for a specific location months in advance. This is, sadly, but a myth - not helped by those meteocharlatans who peddle their long-range weather forecasts triggering media headlines of heat waves, floods and big freezes. It goes without saying that these forecasts don't stand up to verification.

A few years ago a journalist phoned me to get an insight into typical summer weather for the UK, which places were warmest, wettest and so on. All fine and good. Then came the clincher, please could I give a forecast for a variety of summer events? This was many, many months before said events were taking place.

Not possible, was my response, but I can give you an idea of what to typically expect and any historical extremes, such as wet Wimbledons or flooded festivals. This wasn't what the journalist had in mind, the paper wanted a simple weather forecast and temperature for each venue on a specific date.

Again I said it wasn't possible so far ahead, so instead I was asked if I knew who could provide that information. I can only presume either they thought I was refusing to look at my crystal ball or that I mustn't be a very good forecaster. Believe me, if I could give you an accurate forecast that far ahead then I would be a very rich meteorologist. We can make educated guesses based on climatology but it's not a 'proper forecast'. Here's why.

Our atmosphere is chaotic. Sometimes it's a struggle to pin down the details for the next day's forecast, let alone next month. Looking further in advance offers up many possible outcomes. Long range forecasts express what might happen over a larger area during a longer time frame compared to normal climatology using probabilities.

So how do we do that? First up we need a forecast. To do this the current state of the atmosphere is observed and numerical weather prediction computer models then calculate how this state will evolve in time.

But there's a problem. Our computer models don't truly represent all the intricacies, interactions and processes of our atmosphere - the computing power required for that is just too much. And even more of an issue is this: we can never observe every detail of the atmosphere's initial state so the computer doesn't get accurate starting conditions. And as the atmosphere is a chaotic system, very small errors in its initial state can lead to large errors in the forecast. The further ahead you try to look, the greater these errors become.

The second step is to run another forecast. And another, and another, in fact to run a number of forecasts, each with slightly different starting conditions - a trick known as ensemble forecasting. Each forecast is just as likely as the other but the different start point can lead to a different outcome. The ensemble reveals which outcomes are more likely, for example, if half your forecasts indicate above average temperatures, with a quarter opting for average and the remainder for below average, then it's 50% likely we're in for a hot spell. Pretty promising, but there's still a 50% change it won't happen!

Think of it like a horse race. Different rainfall and temperature scenarios over a three-month period - such as dry, near average or wet for rainfall - are the different horses and each option is given odds. Just like racing the favourite doesn't always win. Only people are not quite as understanding if the favourite forecast isn't the one that actually comes to pass.


And that's partly down to how we present the weather to the British public. Ever heard probabilities or percentages during a forecast? Instead of giving the likelihood of rain and expressing uncertainties over timing and amounts, we've fallen into the trap of delivering the most likely outcome as a definite.

No surprise then that forecasters are often accused of getting the weather wrong, most memorably in the year of the "barbecue summer". Back in April 2009, the Met Office explained there was an 80 per cent chance of average or above average temperatures that summer. The media latched onto their barbecue summer description. Sadly an outsider won the race that year and July was one of the wettest on record.

It's also all about interpreting the stats. If we were to talk about a warmer than average summer to come, that doesn't mean a guaranteed heat wave. In fact, nor would it guarantee sunshine as we can still get warmer than average temperatures when it's cloudy and wet.

And just to make it more complicated, even if temperatures are below average some months we could still end up with an above average result over the whole three month period. Not to mention regional variations in temperature across the UK.

So, who benefits from long-range forecasts? They are useful to people who understand probabilities and who plan on longer timescales such as energy suppliers and retailers. It may not be useful to help you decide where and when to go on your summer staycation or to pick a wedding date, so you'll have to be a little patient if you really want to know what the weather will be doing on a particular date nearer the time.