On Saturday, December 31, 2016 at 9:38:30 PM UTC, Tudor Hughes wrote:
On Saturday, 31 December 2016 11:56:09 UTC, Norman Lynagh wrote:
A good number of pundits, including some of the saner variety, have been
getting quite excited by recent operational runs of the GFS showing a very
cold spell setting in during the second week of January. There's been quite a
bit of hope/wish-casting. The 00z operational run today predicted strong E'lies
over the southern half of the British Isles from 11th-16th Jan. In very sharp
contrast, the 06z operational run today predicts strong SW'lies over the
country during the same period. The pundits have gone very silent.

A casual glance at the ensembles recently shows that beyond about 5 days there
is so much divergence that the situation is effectively unforecastable to any
useful level of reliability beyond 5 days. The fact that the model is routinely
run out to 16 days ahead doesn't mean that its predictions have any value. The
operational runs of any model beyond about 5 days are getting into fantasy
land, especially as they are quite often warm or cold outliers when compared to
the ensembles. Unfortunately, it's not easy to build uncertainty into the apps
that are so prevalent these days.

Medium-range NWP output is a tool for meteorologists, not end-user material for
Joe Public.

I'll get off my soap-box now :-)

--
Norman Lynagh
Tideswell, Derbyshire
303m a.s.l.
http://peakdistrictweather.org
@TideswellWeathr



On Saturday, 31 December 2016 11:56:09 UTC, Norman Lynagh wrote:
A good number of pundits, including some of the saner variety, have been
getting quite excited by recent operational runs of the GFS showing a very
cold spell setting in during the second week of January. There's been quite a
bit of hope/wish-casting. The 00z operational run today predicted strong E'lies
over the southern half of the British Isles from 11th-16th Jan. In very sharp
contrast, the 06z operational run today predicts strong SW'lies over the
country during the same period. The pundits have gone very silent.

A casual glance at the ensembles recently shows that beyond about 5 days there
is so much divergence that the situation is effectively unforecastable to any
useful level of reliability beyond 5 days. The fact that the model is routinely
run out to 16 days ahead doesn't mean that its predictions have any value. The
operational runs of any model beyond about 5 days are getting into fantasy
land, especially as they are quite often warm or cold outliers when compared to
the ensembles. Unfortunately, it's not easy to build uncertainty into the apps
that are so prevalent these days.

Medium-range NWP output is a tool for meteorologists, not end-user material for
Joe Public.

I'll get off my soap-box now :-)

--
Norman Lynagh
Tideswell, Derbyshire
303m a.s.l.
http://peakdistrictweather.org
@TideswellWeathr

Stay on your soapbox and shout it even louder, especially the penultimate sentence. This is rather a contrast with the introduction of "long-range" forecasts in the 60's which, as far as I can remember, the Met Office did not want to publish but simply use them for internal research. They were persuaded to publish them by various political/social forces even though they knew, and so did the users, that they were not much good.
I have read some theoretical investigation that says that the position, intensity etc of individual synoptic features are intrinsically unforecastable after a period of about 20 days even if a tendency to a particular pattern may persist. This is due to random effects, i.e. chaos, which cascades up to synoptic scale. This had been predicted by Lorentz some time earlier anyway.

Tudor Hughes, Warlingham, Surrey.

I'd agree with just about everything said above. In winter, for many, it's just wishful thinking at the slightest hint of something cold.

Also, I'd hate accurate forecasts a week or more ahead. It's the unexpected that makes life.

Anyway, Happy New Year everyone.

Graham
Penzance (wet)