On 10 Sep, 10:28, Dawlish wrote:

No Richard. The technique does not allow that and never has. If it
did, it would point to significant and consistent model accuracy at 10
days and we'd have consistently accurate forecasts from the MetO at 10
days and we don't.

Ahh right. Interesting, but doesn't this mean you're just picking and
choosing when you make a forecast? A bit like betting on the favourite
of a horse race when the odds are less than evens? Yes, I'm playing
devil's advocate here !!

Richard

On Sep 10, 11:02*am, Richard Dixon wrote:
On 10 Sep, 10:28, Dawlish wrote:

No Richard. The technique does not allow that and never has. If it
did, it would point to significant and consistent model accuracy at 10
days and we'd have consistently accurate forecasts from the MetO at 10
days and we don't.

Ahh right. Interesting, but doesn't this mean you're just picking and
choosing when you make a forecast? A bit like betting on the favourite
of a horse race when the odds are less than evens? Yes, I'm playing
devil's advocate here !!

Richard

No. I'm determining when a forecast is 75% likely to achieve outcome
at 10 days. Most times, it most certainly isn't; as I'm sure you and
any other forecasters would agree. Will did say that forecasting
anticyclones at this time of year is not easy and the chances of this
one developing (at 10 days out, back on Aug 31st) were no better than
50%. I didn't think that, or I wouldn't have forecast. This was one of
those instances where it was possible to have forecast this
accurately, from 10 days out.

I don't mind you playing devil's advocate. Anyone who issues any kind
of forecast should be prepared to defend both the accuracy and the
methodology. I note you aren't doubting the accuracy! *))

If you follow the earlier posts on this thread, then back on Aug
31st., the MetO were not clearly forecasting the development of such a
dominant high pressure having developed by today. So how could I have
done it? I don't have access to any tools not available to the MetO.

In article bfcb244c-b717-4ea1-bd53-
, says...
If you follow the earlier posts on this thread, then back on Aug
31st., the MetO were not clearly forecasting the development of such a
dominant high pressure having developed by today. So how could I have
done it? I don't have access to any tools not available to the MetO.


What did the MetO predict?

--
Alan LeHun

On Sep 10, 2:08*pm, Alan LeHun wrote:
In article bfcb244c-b717-4ea1-bd53-
, says...

If you follow the earlier posts on this thread, then back on Aug
31st., the MetO were not clearly forecasting the development of such a
dominant high pressure having developed by today. So how could I have
done it? I don't have access to any tools not available to the MetO.

What did the MetO predict?

--
Alan LeHun

Post back on Aug 31st Alan.

On 10 Sep, 12:50, Dawlish wrote:

If you follow the earlier posts on this thread, then back on Aug
31st., the MetO were not clearly forecasting the development of such a
dominant high pressure having developed by today. So how could I have
done it? I don't have access to any tools not available to the MetO.

I don't know how you did it. I just hope unlike the Piers Corbyns and
Jon Wittys (posted on here many moons ago) of this world you'd be open
enough to publish the methodology, or further than that, work with the
MO to improve things unlike Piers was prepared to do. Getting the work
peer-reviewed in a scientific journal is a good starting point.

In terms of a 10-day forecast, it could have been that the Met Office
had access to other forecast tools that you don't have - the ECMWF
ensembles, the Arpege forecasts from France, Deutsche Wetterdienst
etc. that could easily have displayed a bias against an anticyclone
forming. In that sense, having the extra information could be
distracting to the forecast. I don't know, I don't work there !

Richard

On Sep 10, 2:39*pm, Richard Dixon wrote:
On 10 Sep, 12:50, Dawlish wrote:

If you follow the earlier posts on this thread, then back on Aug
31st., the MetO were not clearly forecasting the development of such a
dominant high pressure having developed by today. So how could I have
done it? I don't have access to any tools not available to the MetO.

I don't know how you did it. I just hope unlike the Piers Corbyns and
Jon Wittys (posted on here many moons ago) of this world you'd be open
enough to publish the methodology, or further than that, work with the
MO to improve things unlike Piers was prepared to do. Getting the work
peer-reviewed in a scientific journal is a good starting point.

In terms of a 10-day forecast, it could have been that the Met Office
had access to other forecast tools that you don't have - the ECMWF
ensembles, the Arpege forecasts from France, Deutsche Wetterdienst
etc. that could easily have displayed a bias against an anticyclone
forming. In that sense, having the extra information could be
distracting to the forecast. I don't know, I don't work there !

Richard

Unlike dear Piers, I've always been completely open about my
methodology and it is simple.

If I see 5 consecutive runs of the gfs showing a very similar set up
(consistency) and if the ECM agrees with what the gfs is showing,
there is a 75% likelihood of an event occurring. Without either of
those two criteria being reached, I would not forecast. In doing what
I have done, I know that the knack is knowing when NOT to forecast.

Try it; simple methodology, pretty accurate outcomes. Then it just
takes a nerdy kind of nature to watch almost every single gfs and ECM
run for nearly 4 years to test it over time and a very thick skin to
deal with the criticism when a forecast goes very wrong from the same
people that wouldn't dream of giving praise when a forecast turns out
correct. I can assure you that Internet vultures are a far more
discerning panel of judges than any peer-reviewed journal!! Their
silence is approbation enough. *))

On 10 Sep, 16:53, Dawlish wrote:

Unlike dear Piers, I've always been completely open about my
methodology and it is simple.

Well fair play.

If I see 5 consecutive runs of the gfs showing a very similar set up
(consistency) and if the ECM agrees with what the gfs is showing,
there is a 75% likelihood of an event occurring. Without either of
those two criteria being reached, I would not forecast. In doing what
I have done, I know that the knack is knowing when NOT to forecast.

Is this specific weather types - i.e. a cold easterly or a high
pressure or does this include very short-wave features? Have you had
any luck prediction, for example, a day with a severe gale 10 days
ahead? It would be interesting to see what sort of weather type the
75% of the successes are, and what weather types the 25% of failures
are and whether the two populations are specifically different. Now
I'm assuming that Hurricane Bill in the mid-Atlantic was classed as a
failure? The GFS and the EC had a tight-cored low for several runs and
that didn't come off in the end.

Try it; simple methodology, pretty accurate outcomes. Then it just
takes a nerdy kind of nature to watch almost every single gfs and ECM
run for nearly 4 years to test it over time and a very thick skin to
deal with the criticism when a forecast goes very wrong from the same
people that wouldn't dream of giving praise when a forecast turns out
correct.

That should only really happen if you crow too much when it is right
(see P. Corbyn, 1990-2009) and people want to put you in your place!
If I had the chance to change your method I'd be looking for ECMWF
first and then GFS second (third...fourth?) !

I can assure you that Internet vultures are a far more
discerning panel of judges than any peer-reviewed journal!! Their
silence is approbation enough.

Judges should be fair unless you've upset them !!

Richard

On Sep 10, 5:07*pm, Richard Dixon wrote:
On 10 Sep, 16:53, Dawlish wrote:

Unlike dear Piers, I've always been completely open about my
methodology and it is simple.

Well fair play.

If I see 5 consecutive runs of the gfs showing a very similar set up
(consistency) and if the ECM agrees with what the gfs is showing,
there is a 75% likelihood of an event occurring. Without either of
those two criteria being reached, I would not forecast. In doing what
I have done, I know that the knack is knowing when NOT to forecast.

Is this specific weather types - i.e. a cold easterly or a high
pressure or does this include very short-wave features? Have you had
any luck prediction, for example, a day with a severe gale 10 days
ahead? It would be interesting to see what sort of weather type the
75% of the successes are, and what weather types the 25% of failures
are and whether the two populations are specifically different. Now
I'm assuming that Hurricane Bill in the mid-Atlantic was classed as a
failure? The GFS and the EC had a tight-cored low for several runs and
that didn't come off in the end.

Try it; simple methodology, pretty accurate outcomes. Then it just
takes a nerdy kind of nature to watch almost every single gfs and ECM
run for nearly 4 years to test it over time and a very thick skin to
deal with the criticism when a forecast goes very wrong from the same
people that wouldn't dream of giving praise when a forecast turns out
correct.

That should only really happen if you crow too much when it is right
(see P. Corbyn, 1990-2009) and people want to put you in your place!
If I had the chance to change your method I'd be looking for ECMWF
first and then GFS second (third...fourth?) !

I can assure you that Internet vultures are a far more
discerning panel of judges than any peer-reviewed journal!! Their
silence is approbation enough.

Judges should be fair unless you've upset them !!

Richard

Heh! LOL!

I didn't forecast anything about ex-Hurricane Bill. There was nothing
like enough agreement at 10 days - remember Bill was still tropical 10
days before and its path was by no means predictable, though the
models did a fair job at that range, with a hurricane whose movements
are often notoriously difficult to predict, even at 24-72 hours.

The last time I issued a 10 day forecast using this methodology was on
the 6th June for the 16th, saying that high pressure would be in
charge. It was wrong. Over the summer, there have been no occasions
that I have seen where agreement and consistency have occurred, Hence;
no forecasts - mind you, like I said, I've been mainly looking for
changes of pattern and what changes there have been came under my 10-
day radar and were not showing at T240.

On Sep 10, 7:50*am, Dawlish wrote:

The fine weather should last for a while too, as I've implied with my
similar forecast for 3 days time, but a change may be a-coming for 10
days time - the ECM is leading that possibility but the gfs has not
shown enough consistency yet; some runs having dallied with a
continuance of the anticyclonic weather whilst others have shown a
return to Atlantic weather over the last 48 hours. Another couple of
gfs runs showing windy and wet weather at T240 and I'd be convinced
this blocking pattern won't last to the end of the third week of
September.

This illustrates perfectly the difficulties of forecasting at 10 days
and the need for both gfs consistency and ECM agreement for me to be
able to forecast with any degree of accuracy at that distance.

After writing the above, yesterday, The 00z gfs and the runs last
night, have shown a continuance of the higher pressure out to T240, at
least for England. The 00z ECM shows an Atlantic low stalling. What do
you forecast?



Well the Meto is hedging its bets, but this forecast was issued at
lunch yesterday and, as usual in the morning, is a day out of date.

"UK Outlook for Tuesday 15 Sep 2009 to Thursday 24 Sep 2009:
Much of the country will remain dry through the middle of next week
and over next weekend, but with variable amounts of cloud. Some sunny
spells are likely, more especially in the southwest of England and
Southwest Wales at first. The far Northwest of Scotland could see a
little little rain at times, with also a small risk of showers
extending into far south and southeast of England. Temperatures in all
parts should be near normal, and feeling pleasant in sunshine.
**Southern and Central parts should start mainly dry and fine in the
longer outlook, but there is a low risk of it turning more unsettled
with time. The Northwest is expected to turn more unsettled with
occasional showers or rain.**"

Updated: 1201 on Thu 10 Sep 2009

That's also an interesting comment in the 6-15 day forecast because
that phrase gives an insight into the forecaster's thinking - more
likely to continue settled over England, but a chance of it not doing
so. It also reflects my thinking that an actual forecast would have
less than a 75% chance of being correct.

Therein lies the current problems of forecasting at 10 days. It really
is very difficult.

Still super weather for most. Cool nights and mornings and a
scintillating September day on the south coast yesterday. Hopefully
another one to come. 20C and sunshine is just marvellous weather and
the September light, especially on a morning is quite superb.

On Sep 11, 8:34*am, Dawlish wrote:

Well the Meto is hedging its bets, but this forecast was issued at
lunch yesterday and, as usual in the morning, is a day out of date.

"UK Outlook for Tuesday 15 Sep 2009 to Thursday 24 Sep 2009:
Much of the country will remain dry through the middle of next week
and over next weekend, but with variable amounts of cloud. Some sunny
spells are likely, more especially in the southwest of England and
Southwest Wales at first. The far Northwest of Scotland could see a
little little rain at times, with also a small risk of showers
extending into far south and southeast of England. Temperatures in all
parts should be near normal, and feeling pleasant in sunshine.
**Southern and Central parts should start mainly dry and fine in the
longer outlook, but there is a low risk of it turning more unsettled
with time. The Northwest is expected to turn more unsettled with
occasional showers or rain.**"

Updated: 1201 on Thu 10 Sep 2009

That's also an interesting comment in the 6-15 day forecast because
that phrase gives an insight into the forecaster's thinking - more
likely to continue settled over England, but a chance of it not doing
so. It also reflects my thinking that an actual forecast would have
less than a 75% chance of being correct.

Therein lies the current problems of forecasting at 10 days. It really
is very difficult.

Really very interesting....

The high pressure will still be in charge tomorrow Sunday, with a
front skirting around the northern limb of the high, then slipping
down the North Sea to introduce cloud into some Eastern areas - but
then look at how the forecast has changed next week. A strengthening
north-easterly flow in the south is likely to bring cloud (and speak
it in a whisper) a little rain to the SE (!) as southern areas come
under more of an influence from a large area of low pressure over the
Med. It is likely to be a change which I don't think was possible to
forecast with accuracy, even 6 days ago, never mind 10.

On Sept 11 (see above), the MetO had "Southern and Central parts
should start mainly dry and fine in the longer outlook" i.e. high
pressure would still be in charge in those areas on 17th to (say) 19
Sept (at 6-8 days from the forecast date). That's now unlikely and the
"low risk" of it turning unsettled with time looks like it will happen
well before that 6-15-day forecast was implying, ruining any outcome
accuracy for that forecast. Even at 6 days, the forecast is now much
more likely to be proved incorrect, than correct.

That's why I didn't forecast the high pressure dominance continuing
further than Sunday, but I admit I couldn't see this change next week
at 10 days. All I could see was that there was not enough model
agreement for anticyclonic, settled and fine weather, continuing to be
75% likely. The actual situation in mid-week next week was one that I
feel was not possible to forecast with any current capability, except
by guesswork and the change came well below my 10-day radar.

Out at T240, the gfs has unsettled and the ECM has higher pressure
with an anticyclone again building. You pays your money and you takes
your chance on that one.