Revised long term forecast

[Mulder]

Two thirds of the way through summer, the met office has revised its long term forecast (http://news.bbc.co.uk/1/hi/sci/tech/8173533.stm)for the season!

They got the original long term forecast wrong ("barbecue summer"). Now they are saying it will be more like what we can see it already is! That seems fair enough, experience always beats prediction and, taking the season as a whole, there are already enough cool days to make the whole thing cool overall. But considering they got it wrong before, does that mean we should be looking forward to a hot August after all?

[Croydon Bob]

But considering they got it wrong before, does that mean we should be looking forward to a hot August after all?

Unfortunately, like any other purveyors of woo, the Met Office can be correct sometimes. So you can't just assume the opposite of whatever they say. They wouldn't be able to win the JREF $1m by getting above average by chance statistics but they'll be correct about the same number of times that you or I would be using guesswork.

[FarSideOfTheMoon]

They didn't get it wrong. They said there was a 65% probability of a long hot summer. As most people know (Daily Mail headlines excepted), that means there was a 35% probability it wouldn't be any different from our normal summers.

I think the Met office's interactions with the media would need to be the main focus here. I suspect they are constantly producing updated long range forecasts on a daily basis, and those become more accurate as the forecasted period approaches.

[FarSideOfTheMoon]

Unfortunately, like any other purveyors of woo, the Met Office can be correct sometimes. So you can't just assume the opposite of whatever they say. They wouln't be able to win the JREF $1m by getting above average by chance statistics but they'll be correct about the same number of times that you or I would be using guesswork.

I'm quite interested in how their predictions are evaluated. You would expect to see short range and very short range predictions showing some degree of accuracy, possibly a high degree, whilst the longer you make the prediction for, it obviously becomes less accurate. At what point do they match that predicted by chance. And is this point likely to be extended further into the future as methods are improved and technology is implemented?

I don't believe they are completely woo when making long range forecasts, but I would like to see them describe the methodology where they feel vindicated in making long range forecasts.

[Mulder]

They didn't get it wrong. They said there was a 65% probability of a long hot summer. As most people know (Daily Mail headlines excepted), that means there was a 35% probability it wouldn't be any different from our normal summers.

On that basis they can never be wrong!


I suspect they are constantly producing updated long range forecasts on a daily basis, and those become more accurate as the forecasted period approaches.

That's certainly what they do with short term forecasts. I frequently check the forecast and I've noticed how it changes it for the next few hours to fit what is actually happening now! In other words, their forecast for 10am tomorrow might be rain but if, at 9am on the actual day there is sunshine, you'll find their forecast for 10am has changed to sunshine. Almost as though they just looked out of the window!

[Croydon Bob]

Almost as though they just looked out of the window!

Unfortunately they don't always manage even that. The stupid twittering Scottish woman who does BBC breakfast telly will often stand in the Blue Peter Garden in pouring rain saying "it's a bright sunny morning!" Or vice-versa. She seems incapable of deviating from her script even when it is obviously wrong.

[Trinoc]

What is it with the British and the weather? They insist on forecasts for everything from the walk to the shops in 5 minutes to the whole of the next year, and then complain if they are not 100% accurate!

The weather is a chaotic system ... if anyone doesn't believe the weather is more likely to be as forecast than not, then they should ignore the forecast and always carry a complete set of clothing for all weather wherever they go.

[Croydon Bob]

The weather is a chaotic system ...

Of course.

But that's not what the Met Office say is it? They'll give a forecast saying that a particular rain system will be exactly here at 1pm and exactly here at 4pm and exactly here at 7pm and in fact it'll actually head off in the opposite direction and rain on 10 million people who were told that they would get dry weather. If the Met Office was honest there wouldn't be anything to complain about, but they lie.

When did a weather presenter say: "this system could go this way or that, it may rain on you lot up here or you lot down there"? They never do.

There was a period of time when they'd say: "40% chance of rain here". But they've stopped doing that and will now wave an arm vaguely across the South East and say: "light showers here". Wrong for 60% of the people in that area but counted as a hit in the world of woo-woo.

[Tony Williams]

I agree that they are often far too definite in their forecasts (but not always - you do occasionally hear things like: "we're not sure exactly when this front will arrive") and, to add insult to injury, they rarely comment on the accuracy of their previous forecast even when they get it badly wrong (a big bouquet to the wonderful NW BBC weather girl who I once heard say "I'll give myself one out of ten for yesterday's forecast, and that was for getting the day right").

The problem is I think that they are under pressure to come up with crisp and snappy summaries which Joe Average can understand, rather than hedging them around with statistical probabilities. Joe has no concept whatever of probabilities, or he wouldn't buy lottery tickets.

[Trinoc]

I suspect the media insist that the Met Office keep it simple for all of us poor, thick punters. From the current hoohah it is clear that even quoting a percentage which amounts roughly to "we are 2/3 sure" does no good -- everyone complains anyway if it is the 1/3 that actually happens.

Very occasionally you hear a weather forecaster mention that a forecast has a low confidence level due to running parallel predictions with slightly different starting points which turned out very differently, but if they did that regularly I'm sure it would be stamped out rapidly by the "dumb it down" brigade.

[chaggle]

What should we do then if the Met Office only does woo? Not have a weather forecast at all? As a (previously) regular user of the shipping forecast, I must say the the Met Office's forecast is vital. It has probably saved my life and at least got me back to work safely for Monday morning. Of course the shipping forecast is designed for people who have been trained to interpret it.

With the complexity of the weather systems over the British Isles it is hardly surprising that it's wrong from time to time. The weather can be different in adjacent valleys or to windward or leeward of a hill. The problem, as stated in previous posts, is that the weather forecast is presented as a kind of certainty.

Whereas the average poster on this forum would probably be able to interpret a shipping-style forecast for inland, the majority of the populace probably wouldn't and therefore need a spoon-fed, simplistic formula and it doesn't particularly matter if it's a bit wrong.

[Matt]

I don't mind the existence of a dumbed down forcast. Is the unavailability of a forecast containing confidence measures that's a bit rich.

[Tony Williams]

I don't mind the existence of a dumbed down forcast. Is the unavailability of a forecast containing confidence measures that's a bit rich.
You can apply your own. If the wind speeds given are low (isobars far apart, if you're lucky enough to be shown them), then the weather is likely to be very slow to change and therefore easier to predict accurately. If wind speeds are high (isobars tightly packed) then it means a frontal system is going over, so the weather will change quickly. The main problem with fronts is getting the timing right, because this cannot be predicted with high confidence. This is where most of the short-term weather forecasts trip up: they will correctly identify that a certain pattern of weather is going to cross the country, but if their calculation of the transit speed is off by a few hours (as it often is) that can mean that the rain falls during the day rather than at night, which is a minor issue to meteorologists but makes all the difference to the punters planning their outdoor activities.