"Dave.C" wrote in message
. uk...
| When I joined the group a few years ago I tended to blame the forecasters
| for poor forecasts. Then I learned that although it was stated as if fact
it
| was realy the best guess on the information available. Then it was clear
| that anything past 5 days was very unreliable. But, now, despite my best
| efforts, I am wondering just how good computer modelling really is. I mean
| what did those balloons really tell that model that made it change so
| quickly. There lies the mystery and the perpetual "soap" of it all I
guess.
|

Computer modelling is never going to give perfect results. This (as was
discovered by Lorenz several decades ago, now) is because the weather system
is "chaotic" - slight variations in initial conditions are rapidly amplified
as the system evolves until even quite similar initial conditions can give
very different outcomes at a suitable future time. This is why we now see
"ensemble" forecasts - where the models are run with slight perturbations to
the initial fields and variability of the resulting outcomes gives an
indication of how reliable (or otherwise) the forecast is likely to be. And
you think you have problems - in the Royal Met. Soc.meeting on Wednesday
there were two case studies: one on Hurricane Wilma, which one forecast run
predicted to steam up the east coast of the USA to Nova Scotia and the next
predicted it to stall in the SE Caribbean - an error of several thousand
miles between successive forecasts; the second the infamous Epsilon which
refused to weaken in spite of six days of continuous computer guidance that
it would.

The amount of data input to compute models is nowhere near enough to predict
the atmosphere as far forward as we could wish. This not helped by a very
thin cover of observations over large areas of the globe, most notably the
oceans - and there is one of these to the west of us where most of our
weather action comes from. Sattelite observations can fill some gaps, but
these need to be calibrated against "ground truth", so to an extent we have
a "catch 22" situation.

This can be overcome to some extent by a process where the output from
preceding forecasts is used to "fill in" the gaps in the observations
available. But this will to some extent perpetuate errors in the previous
forecast.

No matter how powerful our computers or how great the wit of our
programmers, this problem can never be fully overcome. The errors can be
reduced as far as we are prepared to make the effort to get more
information, computing power, or cleverer programs. But eventually the "Law
of Diminishing Returns" sets in and we have to decide how much more effort
is worthwhile for the reducing increase in skill it will provide.

Personally, I am glad. How boring life would be if we could forecast the
weather exactly a fortnight in advance! Three cheers for the "Butterfly
Effect", deterministic chaos, Sod's Law or whatever you want to call it.
The distant GFS runs are a bit of fun, but remember the wise saying:
"Blessed are they who expect little, for they shall not be disappointed".

And who cares for the cold weather, anyway? I am a shorts-and-T-shirt man
who likes to smell the grass and enjoy the flowers. Bring on the Spring, I
say. If I want months and months of Winter, I could emigrate to Canada.
--
- Yokel -
oo oo
OOO OOO
OO 0 OO
) ( I ) (
) ( /\ ) (

"Yokel" now posts via a spam-trap account.
Replace my alias with stevejudd to reply.