What's inside an SPD?

[JohnW2]

That doesn't look at all probable.

I've actually given up 'apologising' to you every time I use the word "probability"! We live in different worlds - there are many days when I type the word hundreds of times, and almost invariably in relation to very small numbers (i.e. things which are extremely 'improbable')!

...but what does a 23% chance of rain tomorrow mean? Anything?

Well, as I'm sure you understand, in mathematical terms it means a lot - essentially that (assume everything is 'random') if one could re-run history 100 times, it would 'rain tomorrow' with about 23 of those re-runs, but not on the other 77 of them.

In practical, non-mathematical, terms, it is clearly more difficult, and all one can really do is play with words. A 90% probability that it will rain tomorrow could be expressed as it being 'very likely' that it will rain, whereas a 10% might be said to indicate that rain was "unlikely" - leaving you to choose what words to use for probabilities between 10% and 90%. As I think I have illustrated, the real problems arise with very small probabilities, not the least because one runs out of 'extreme adjectives'.

Whilst the 're-running of history' mathematical meaning of probability is very abstract, and sometimes difficult to conceptualise, it is sometimes possible to look at things in a more easily appreciated fashion, if one can, say, think about lots of people, rather than lots of 're-runs of history'. For example, if I told you that you had a 0.0002% probability of contracting a certain very rare disease in the next 12 months, it would probably be very difficult for you to really conceptualise what that meant. However, if I told you that if you selected 1 million people at random, then about 2 of them would contract the disease in a next 12 months, that would probably be more meaningful to you.

Kind Regards, John

 

[EFLImpudence]

Ah. I hadn't considered that it meant the weather forecasters had looked back and found that it rained in my area on 23% of the February 23rds.

I just wondered if it meant that there will be 23% rain cloud cover over the whole area at any one time but they don't know where the gaps will be, and of course, it does not state any time perameters.
So, a five minute shower or constant rain all day could mean the forecast was correct.

 

[JohnW2]

I would assume they have a definition of that statement. It doesn't necessarily make it any more useful.

Given the way in which weather forecasting is done, the derivation of such a statement is almost certainly that, when similar prevailing conditions have been seen in the past, on 23% of those occasions there has been rain.

However, as I think we are all agreed, 'usefulness' is much more complicated. Just as with the lighting strikes, one is interested in the yes/no dichotomy of "will it rain". As I've just written to EFLI, it's relatively straightforward at the extremes, provided one accepts that there are 'no certainties' - i.e. a very high probability means that it is 'very likely' that it will rain, and a very low probability means that it is 'very unlikley' that it will rain - the greater difficulty arises with intermediate probabilities.

Anything probabilistic obviously, by definition, has uncertainties, but probabilities can be used for practical numerical estimation, although there will always be an element of 'gambling'. If one has a restaurant which serves about 100 customers per day, and if statistics show that there is a 23% probability of any given customer ordering a fillet steak, then one can estimate that one has to buy abut 23 steaks (probably plus a little 'overage') each day.

Kind Regards, John

 

[JohnW2]

Ah. I hadn't considered that it meant the weather forecasters had looked back and found that it rained in my area on 23% of the February 23rds.

Not quite ...

... as I've just written, they do 'look back in history' but not just at "February 23rds in your area". They will have looked back to situations in which all the prevailing meteorological conditions (and the location, time of year etc.) were similar to what they believe they will be tomorrow, and ascertained that on 23% of those previous similar occasions it had rained.

Weather forecasting utilises incredibly powerful computing facilities, and has always been at the forefront of usage of the 'state of the art' of what technological advances have to offer.

Kind Regards, John

 

[JohnW2]

I just wondered if it meant that there will be 23% rain cloud cover over the whole area at any one time but they don't know where the gaps will be, and of course, it does not state any time perameters. So, a five minute shower or constant rain all day could be a true result.

When I look (as I often do) at the Met Office current weather forecast for my location, it gives a probability of rain (as well as temperature, wind speed, visibility etc.) for each hour for the next 7 days - and those figures are updated very frequently (I think several times per hour) on the basis of the evolving situation.

Kind Regards, John

 

[JohnW2]

When I look (as I often do) at the Met Office current weather forecast for my location, it gives a probability of rain (as well as temperature, wind speed, visibility etc.) for each hour for the next 7 days - and those figures are updated very frequently (I think several times per hour) on the basis of the evolving situation.

Here you go ... from a couple of minutes ago, for my location ...

Kind Regards, John

 

[EFLImpudence]

That shows a constant 10% chance of rain from 1800 to 0700.

 

[DetlefSchmitz]

My issue is that if it rains in my 'area' it may still not rain where I am. Very often we see rain in one location during the day and none 5 miles away. Personally if it rains 5 miles away I don't really care. This complicates the interpretation.

 

[EFLImpudence]

Or - it could rain non-stop for the fourteen hours.

Surely, the figures are meaningless.

 

[JohnW2]

That shows a constant 10% chance of rain from 1800 to 0700.

It does, and that's quite common. In fact, I think the lowest probability they ever give is "<5%" - and when they say that (or 10%) it's very rare for us to get any rain.

I think most people would regard 10% (or less) as meaning at least "likely", probably "very unlikely".

Kind Regards, John

 

[JohnW2]

My issue is that if it rains in my 'area' it may still not rain where I am. Very often we see rain in one location during the day and none 5 miles away. Personally if it rains 5 miles away I don't really care. This complicates the interpretation.

Sure, but that's just a fact of life. We had a situation quite recently when, for a good few minutes, it was raining quite heavily on one side of my garden but not at all on the other side. With something as 'granular' as weather, no highly-localised predictions can be perfect.

However, much higher resolution, essentially real-time, data does exist. A friend of mine works in Air Traffic Control at Heathrow, and they have data, and predictions on a minute-by-minute basis, of weather conditions at all points in the airport and the surrounding area. That's were the 'state-of-the-art incredibly powerful computing' comes into play.

Kind Regards, John

 

[JohnW2]

Or - it could rain non-stop for the fourteen hours. Surely, the figures are meaningless.

No, not 'meaningless', but subject to inevitable uncertainty.

If I were contemplating planning an outdoor event for tomorrow afternoon, my decision would be highly influenced by the difference between a 10% probability of rain throughout the afternoon and a 90% probability of rain throughout the afternoon. There is, of course, an element of 'gambling', since my decision based on those predictions could prove to have been wrong - but that's what 'uncertainty' is unavoidably all about.

Kind Regards, John

 

[EFLImpudence]

A friend of mine works in Air Traffic Control at Heathrow, and they have data, and predictions on a minute-by-minute basis, of weather conditions at all points in the airport and the surrounding area. That's were the 'state-of-the-art incredibly powerful computing' comes into play.

Surely that's just looking at radar -

and like Formula 1 - that does not always prove correct.

However, that is different than forecasting 10% chance next Wednesday at 1700 - which is just of no value at all.

 

[JohnW2]

... pressed button prematurely!

 

[JohnW2]

Surely that's just looking at radar -

The literally real-time data obviously is - the predictions regarding the probable situation in 5, 10, 15, 30 or whatever minutes' time is far more complex, since radar has no crystal ball.

However, that is different than forecasting 10% chance next Wednesday at 1700 - which is just of no value at all.

There's obviously always going to be an element of uncertainty/gambling, but if you were trying to decide whether or not to hold a 'garden party' next Wednesday evening, would your decision not perhaps be different depending upon whether the prediction of rain at that time was 10% or 90%?

Kind Regards, John