|"Alastair" wrote in message
...
|On Nov 27, 7:50 pm, "Col" wrote:
|
| This could be seen as covering all bases, but I think it's just
| being realistic regarding what is currently possible in seasonal
| forecasting.
| A seasonal forecast can't be made along the lines that the
| weather *will* do this, that or the other.
| Rather, it has to be evaluated and presented in terms of
| probabilities. And if there is no strong signal of any definate
| weather pattern then so be it.
| --
| Col
|
| Bolton, Lancashire
| 160m asl
|
|Yes, but when they say there is a 50% chance it will be mild and 50%
|that it will be cool then any fool could of come up with that!
|

They have not said that there is a 50% chance that it will be mild and a 50%
chance that it will be cool. Have a closer look at that forecast.

The prediction is split into three sections (or "terciles"), each of which
would be equally probable if the outcome were totally random. So a "don't
know" forecast would be 33% chance "cold", 33% chance "average", 33% chance
"mild". If you look at the rainfall prediction, you will see that they have
gone for something fairly similar to this.

What the Met Office have gone for with temprerature is the "mild" category
to be as likely as "average" and "cold" *put together*. This is actually a
strong forecast for a mild winter - they are giving "evens" on what should
be "2 to 1 against". Or you could say that they reckon a "mild" winter is
half as likely again (3/6 chance rather than a 2/6 chance) as one would
expect at random.

In fact if you look, you will see their call is for only 20% "cool" - or in
other words they reckon that a mild winter is two and a half times more
likely than a cold winter with a 30% chance the winter will be neither mild
nor cool. Don't know where you got the "50% chance mild and 50% chance
cool" from - that is most definitely NOT what the forecast says!

That's the trouble with probabilistic forecasts. Hardly anyone knows how
statistics actually work...