In message , Dave.C
writes
I have been an advocate of probability forecasting on here for a few years
now. I have however been giving it some more thought and do have some
misgivings. What I have noticed in a forecast such as, say, "90% confidence
dry and mild, 10% cold and heavy snow" is that the interpretation by some of
the outcome is:-
1. If it is dry and mild then it was a good forecast.
2. If it is cold and there is heavy snow then there is the comment along
the lines "brilliant forecast, only a 10% chance and he/she was the only one
to spot it." (Not 90% inaccurate!)

This is happening to some extent, I feel, with the UKMO winter forecast. In
the overall expectation of 66% chance of being colder than average, people
that think it hasn't been that cold are saying it was good because it never
really said it would be and people that think it has been cold are saying it
was good because there was a 66% chance.

Perhaps the best option is to give a confidence in a particular forecast
with no other scenarios or their associated confidence.
(Which of course, the Met Office has in effect)
This is just some thoughts I've been having and am quite open to other
viewpoints as there is clearly more to it (or at least it's intrepretation)
than meets the eye.

Dave

As I have said a couple of times recently a single probability forecast
cannot be judged to be either accurate or inaccurate. Accuracy, or
perhaps better described as reliability, can only be judged by
evaluating a representative sample of probability forecasts. I would
think that a minimum of something like 20 forecasts would be needed.

If an event occurred, on average, on 3 occasions out of every 4 in which
it was forecast to have a 75 percent probability of occurrence then that
probability forecasting would have been spot on.

The Met Office forecast of a 2-1 probability of a colder than average
winter cannot, on its own, be evaluated as being either accurate or
inaccurate.

Norman.
(delete "thisbit" twice to e-mail)
--
Norman Lynagh Weather Consultancy
Chalfont St Giles 85m a.s.l.
England