On Jun 7, 3:17*pm, "Will Hand" wrote:
"Dawlish" wrote in message

...
On Jun 6, 7:38 am, Dawlish wrote:





On Jun 5, 11:40 pm, "Dave R." wrote:

"N.E Zephyr" wrote in message

...

LOL in classic fashion your so called forecast has gone titicus
verticus
in a matter of hours.

Yeah but so has Will LOL why shouldn't he get slated too?

--
Dave R. [west London]

Always judge a forecast at outcome. The dry and settled part for
Tuesday still looks good, but I've been commenting on the way that
forecast high has been moving on the models, and a settle to the West
of us is quite possible. Still a few days to go. Will doesn't deserve
any slating whatsoever. Nearly every agency, including the MetO were
talking about a warm-up, with good summer weather last week, until the
models changed yesterday lunchtime. Forecasting at 10 days is
incredibly difficult to do an ever so easy to criticise - but leave
criticism until the outcome date and I'll join in the inquest, as I'll
always return to any forecast I've made.

For my forecast for Tuesday, there is a fly in the ointment. Some
models are, presently, showing a front producing light rain moving SE
across England, but pressure is well over 1020mb and it could be shown
to die as models change over the last few days. I'm not unconfident
about this forecast; temps are being shown to be in the mid-20s,
pressure is forecast to be high and many areas away from NE Scotland
are likely to see some lovely, sunny, settled weather. That could
change and the front shown today could remain to spoil Tuesday's
weather, but the models have been building the high pressure again,
for the last 2 days since Dave and my stalker rashly wrote it off on
the basis of a couple of model runs and any further building of that
high could completely kill that frontal ingress, or prevent it
crossing England and Wales completely.

Never judge a forecast until outcome - but then do have the grace to
come back and judge it and judge it fairly against the original
perameters. If the forecast doesn't do what you want it to, come back
anyway and give credit, if credit is due. There's a measure to judge
yourselves by.

Paul

=====

Well said Paul!

I caught Rob McElwee on News24 at lunchtime and he predicts 26C in London
tomorrow.
He also talked of proper summer weather into next week.

What really grates sometimes is that critics often don't understand
atmospheric dynamics, I carefully explained the process whereby the sunshine
and warmth would come from. i.e. disruption of the upper trough and a build
of pressure across the neck over England and Wales. OK I forgot to mention
that the reason I went for 28-30C "somewhere" was due to a combination of
drying out of the atmosphere due to anticyclonic subsidence and strong
insolation (sunshine) at this time of year.

You cannot just go by model runs per se. You need to understand what is
happening to fully make use of them and that takes years of training and
experience. What the best forecasters do is kind of keep time-lagged
ensembles in their heads, i.e. remember previous runs and spot trends and
forecast accordingly but also understanding the processes that the models
are telling them.

FWIW I think Tuesday's cold front will be just a band of broken cloud in the
SW.

Will
--- Hide quoted text -

- Show quoted text -

My own thinking is that forecasting at 10 days is usually not possible
with any real degree of accuracy, even using all the excellent
forecasting techniques you refer to Will. I've been frustrated for
years by the lack of accuracy at that distance. Out to a week, I agree
wholeheartedly, that a combination of models, plus the ensembles, plus
a good mental schema of how the 3D situation is developing and
training and experience in meteorological processes, is how trained
forecasters should (and do) operate. Using these techniques,
measurable success out towards a week (well, at least at 5/6 days,
using the NASA model comparison output) can be achieved. Further than
that, everyone struggles.

Early last week, I saw the same as you; an extension of the Azores
high NE, warming uppers and building pressure over a 4/5 day period to
the middle of next week, but we see things like that daily. The thing
that convinced me of the possibility of the "start of summer" (no idea
how long it will last!) and to issue a forecast, was the coincidence
of 5 consecutive gfs runs showing very similar output at T240+.
Initial conditions were changing; indeed we were in some extremely
difficult forecasting conditions when I issued the forecast and
conditions were changing every 6 hours, in an Easterly flow. Why would
the gfs keep the similar 10-day output when the initial conditions
were changing? That "time delayed coincidence" is what I look for in
my forecasts. Only the gfs runs charts out past T240 and is therefore
the only one I am able to use, so I use the gfs. It's no better and is
perhaps worse, than other models at T240, but there are no stats.

At 10 days+, the ensebles are spaghetti - almost all the time. Any one
run coming to outcome is probably as unlikely as any other, so I use
the operational run, as that is what the model feels is most likely to
achieve outcome. That was the basis of me issuing that forecast and
I've been doing that for 2 years now, publishing every forecast on the
Internet and coming back to judge every one of 50 odd forecasts. To 50
forecasts, I'd achieved 77% accuracy, but I've stopped monitoring the
forecasts really, as even I have difficulties with exactly how to set
the perameters and judge the outcome properly.

The technique doesn't suit everyone, as you've probably seen (I'd love
the critics to try something they believe will work and monitor the
forecasts that they make to show they could do better - that's a vain
hope!), but maybe there's an element within the technique that could
add to how numerical model based forecasts are compiled. From email
contact, I know some professional forecasters are interested and have
been influenced, but I don't think it is the be-all and end-all of
forecasting at 10 days, which remains awfully difficult on a daily
basis - for anyone, or any agency.

You don't try to forecast all the time, Will, but you do issue
forecasts on an occasional basis. Why don't you do this on a more
regular basis? I suspect you feel the same as me, that much of the
time, there is no model/ensemble agreement and to issue a forecast
would end up as little more than an educated guess.

Paul