Hello JCW!
You are right. I have explained the method a number of times, but your posting
prompted me to do it one last time. I'll try to make it short.

I've had many years experience and this has taught me that the NWP method of
forecasting is unreliable after 4 or 5 days. in fact it is always unreliable. The
computers are doing a fine job. It is the method which causes the errors. Any
system which requires a master chart (analysis chart) and then runs forward in a
series of stages is bound to be limited in the end.

Now imagine for a moment that the hourly weather was acted upon by certain factors
(I'm leaving this vague on purpose). In other words, the factors produce the
weather, and there is a direct 1 to 1 agreement of factor to weather. It's clear
that if you know the factor you know the weather. All you have to do is to know the
date and know the time and calculate the factor. ( It's taken me about 50 years to
work this out so so I'll leave it at that.) What you finish up with is an index
number. From the index number you can look in a catalogue of charts - about 40
charts inall - and read off the pressure situation for the area in question. You
can have catalogues for different areas, the index numbers are the same. - the
index numbers change when the hour changes for that day. It sounds complicated the
way I'm describing it, but routine makes it quite easy.

It could happen that a particular sort of weather is not included in the catalogue.
The index number still sorts out the chart in question and this is usually the
neaarest similarity in the batch. If you remember the 4th December was way out.
The reason was the build up of HP from the SW which developed between two LPs, was
not in the catalogue. The catalogue behaved as if the HP had not been there and
showed the two LPs being joined together without an HP between them.

Calculations from the date and the time can be done for any time. I don't need to
know what weather went on before or after, ie. I don't need a master chart to work
on. I can make master charts for myself for any time. And I can let them run on
for the next day, and the next, and the next, without doing any calculations. (just
like a NWP forecast) and the resulting forecast days, jusst like the NWP, fall off
after a few days and are unreliable.

So what have I got?. I can pick out a date and time to forecast without knowing
what type of weather surrounds the date in question. I make one shot only whether I
do the forecast a week before, or five years before (not ten shots at a ten day
forecast like the NWP).

And that's it, JCW. No cyclic patterns, no visualised pressure patterns; no notice
taken of the calendar as far as the pressure situation is concerned, but the
'effects' of the seasons have to be taken into account.

How do I measure my success? Well, when you have been at long range forecasting as
long as I have you can see at a glance what is on the right path and what isn't-
As a yardstick I have a number of historical weather charts and each new method has
to get everyone correct. Then each new method is progressed in a certain situation
over a number of days. I know what I expect, and each new method has to deliver.
If it doesn't deliver, it is discarded, unless a slight change in the method can
bring it back into line again. The forecasting methods which have come through are
then checked to see if their success is related to one season only. Then whatever
method is left over is checked for cirrent weather situations. After that, whatever
method is left over is tested over a long term to find out that no drop-outs
occur.
In the end you can say that the final winners are mostly successful at delivering
what you require.

As I set the limits I don't need publication to tell me what I already know about
my own method. I set my own goals and I'm responsible for my own succes and my own
downfalls.

As I mentioned above, the forecasts can be made for any place and any time. I have
Christmas forecasts running since summer 2003 for Europe, America, Falklands,
Australia and Japan. They run from 24th December till 1st January 2004. I expect
them to give a rough guide to the daily pressure situations. The question is, how
much DETAIL will be shown to be correct in the various areas of the earth?

Well JCW, I think I've answered all you questions as well as I can. Thanks for
the interest.

Cheers, Keith

JCW schrieb:

"Keith Darlington" wrote in message
...
Mike - I would like to come back and reply to some of the points you have
made.

SNIP
Bill Giles uses a sequential method of forecasting - ie. he needs to know
the latest
analysis in order to project into the future. I don't. I go directly to
the date I
require. Furthermore, Bill Giles explains much about the likely atmospheric
weather. I
don't. My interest lies in forecasting the expected pressure situation.

Keith, no doubt you have probably explained your methodology more than once on
this forum if not others - unfortunately I do not recall them. I hope you can
take the time to answer these queries.

Is your methodology based on some form of cyclical pattern or what? When you say
you go directly to a date then is it that you have visualised some sort of
pressure pattern based on a prior result? Or is it that you determine the
pressure pattern is, forgive me, "calendar" driven? I don't mean literally but
some sort of referral to specific time of season / year, etc.?

Given the difficulty I have understanding how you derive your forecasts you'll
excuse me, I hope, if I question how conditions that fail to arise on any
particular day in your forecasts do not interfere or alter the weather you had
expected to occur in following days?

Finally, and I do mean this constructively, how do you yourself measure your own
success and have you published any specific results based on this criteria? Also
to what "scale" to you find your forecasts reliable, i.e. are your forecasts
generally restricted to an area the size of the British Isles or can you be even
more specific?

J.