I could be completely wrong but I don't see how you can take a mean of
different evolutions. For example, in the GFS 850 hPa ensembles for Feb 28th
there are 4 members around +3C and six around 5-10C giving a mean about -4C.
Isn't this a bit like saying the mean of a warm sunny day and a cold snowy
day is an average cloudy day. Surely one type of synoptic evolution will
prevail and it is the mean within that particular evolution which is likely
to provide the best outcome. So that would mean February 28th is likely to
have max temperatures of around, say, 4 deg C or 10 deg C depending on the
resulting evolution. Not the mean of 7C.
Please correct me if this is not basically correct.

Dave

"Dave.C" wrote in message
k...
I could be completely wrong but I don't see how you can take a mean of
different evolutions.

Although this page deals with EC output, the basic principle is the same
....

http://www.ecmwf.int/products/foreca...mble_mean.html

It is acknowledged that you shouldn't just look at the mean of the
ensemble; in the EC output, with many more members, then sophisticated
analysis of the 'clustering' is performed, and it is these rather than
the crude mean solution that is used to assess the likelihood of future
developments. I use the mean on the GFS output simply as a 'eyeball
guiding' line (to judge how the outcomes are clustering about it) rather
than trying to use it to explicitly work out the future state of the
atmosphere. It's not a 'fallacy', just another tool.

Martin.


--
FAQ & Glossary for uk.sci.weather at:-
http://homepage.ntlworld.com/booty.weather/uswfaqfr.htm
and
http://booty.org.uk/booty.weather/metindex.htm

In article ,
Dave.C writes:
I could be completely wrong but I don't see how you can take a mean of
different evolutions. For example, in the GFS 850 hPa ensembles for Feb 28th
there are 4 members around +3C and six around 5-10C giving a mean about -4C.
Isn't this a bit like saying the mean of a warm sunny day and a cold snowy
day is an average cloudy day. Surely one type of synoptic evolution will
prevail and it is the mean within that particular evolution which is likely
to provide the best outcome. So that would mean February 28th is likely to
have max temperatures of around, say, 4 deg C or 10 deg C depending on the
resulting evolution. Not the mean of 7C.
Please correct me if this is not basically correct.

I think you are right, but I suppose a simple mean has the advantage of
simplicity. Most of the time, the ensemble members will probably be
spread fairly uniformly about the mean, with the bi-modal situation that
you've identified being the exception rather than the rule. In any case,
so long as you aren't only provided with the mean, with the individual
members being hidden, then I don't see that there's a problem,

Another point is that I believe that the "operational run" is calculated
on a finer grid than the other members, so arguably should be given more
weight when calculating the mean.
--
John Hall
"Honest criticism is hard to take,
particularly from a relative, a friend,
an acquaintance, or a stranger." Franklin P Jones

Another point is that I believe that the "operational run" is calculated on
a finer grid than the other members, so arguably should be given more weight
when calculating the mean.

I don't suppose you know if it is, do you John?

Thanks,
Dave

In article ,
Dave.C writes:

Another point is that I believe that the "operational run" is calculated on
a finer grid than the other members, so arguably should be given more weight
when calculating the mean.

I don't suppose you know if it is, do you John?

Given more weight? I don't think so, but that's only based on eyeballing
the ensemble graphs.
--
John Hall
"Honest criticism is hard to take,
particularly from a relative, a friend,
an acquaintance, or a stranger." Franklin P Jones