Weather Banter - View Single Post

Dawlish · October 21st 09, 10:57 PM posted to uk.sci.weather

On Oct 21, 10:07*pm, Dawlish wrote:
On Oct 21, 8:30*pm, "Norman" wrote:

John Hall wrote:
In article ,
*Will Hand writes:

"Col" wrote in message news:4NSdnYGhC8U4zkL
...

"Dawlish" wrote in message

...
On Oct 21, 6:48 pm, "Col" wrote:
"Dawlish" wrote in message

...

It's just not there. The 06z sinks the high!
http://www.wetterzentrale.de/pics/Rtavn2402.png

But back again on the 12z.....
--
Col

Bolton, Lancashire
160m asl

Very true Col! The 06z *was probably an outlier, but the interruption
to the consistency begs a couple more runs and agreement from the ECM
before a forecast.

Yes, I am beginning to understand how you approach these10
day forecasts. It's all about consistency and ultimately 'confidence'.
If you get a run that suddenly goes against what the recent runs
have been showing then that confidence is shaken and it takes a
few more runs showing the original sceanario to get that confidence
back to it's original level.
That run could be an outlier, but then it could depict what the
weather actually does!
-- *Col

Of course professional forecasters don't have that luxury. Customers
want forecasts and they want them now!

But they do have the big advantage of access to the ensembles, whereas I
believe that Dawlish uses only the production runs. Thus he can't tell
whether or not a particular production run is likely to be an outlier
other than by relying on consistency between models and from one run to
the next.

But why not look at the ensembles? The GFES ensembles are widely promulgated on
the net.

Norman- Hide quoted text -

- Show quoted text -

Mainly because, at 10 days, ensembles are a waste of time. At that
distance, the error in every run is so great that you can't trust what
you are seeing with the spread. It tends towards randomness. If
ensemble prediction was the answer, we'd have better accuracy at 10
days, but the outcomes show that ensembles are certainly not the
answer. Neither are they "a part of the mix" because the "mix" doesn't
work either. In fact, nothing works at 10 days on a regular basis; if
it did, someone would be able to demonstrate accuracy on a daily basis
and they can't. That is why I only forecast when my technique suggests
I will have at least 75% success and over 74 forecasts, that is the
success rate that I can demonstrate using that method. That has to
count.

I take entirely what Will says about public expectation and he is
completely right. The public expect forecasts every day at 10 days and
more. If it wasn't on the website, there would be complaints galore!
Fortunately, the public don't keep any accuracy records and the MetO
can say just about what they want in the 6-15 day forecasts, with no
comeback whatsoever - everyone has forgotten what they said at
outcome! Well, almost everyone *)). Who tracks the accuracy of that
6-15 day forecast? The MetO don't; NOAA don't and I would suggest that
it is impossible to do that from anyway from the dearth of information
that it contains.

At 10 days we are presently beyond the limits of forecasting with
accuracy on a daily basis. Simple as that.- Hide quoted text -

- Show quoted text -