In the 43 second, roughly, of this TV news video
clip:


http://news.yahoo.com/p/v?u=/ap_av/2...8&f=5374635 3

(which will undoubtedly go away fairly soon, old
news isn't "news" any more) is a map showing a
cluster of predicted storm tracks decorated with
variously colored triangles and squares.

The software that draws it is the ATCFA software I
maintained and enhanced for the US Navy from
1991-1994, or a very close work-alike (mine was used
on the west coast, but "borrowed" by the east coast
forecast offices as well).

What is being depicted is the cluster of predictions
from _many_ forecast models, one per track shown,
with the colored triangles or squares both
identifying the associated model which predicted that
particular storm track, and also showing the place
the storm center is expected to be at a particular
standard _time_, GMT, usually on 6 hour intervals
from midnight.

While no single forecast model has a splendid record
for consistent accuracy, a hand faired track through
the mean predictions of the models that do agree
(the meteorologist gets to pick from dozens of
models, chooses ones with good past results in
similar situations) tends, these days, to be very
predictive indeed.

Forecasting of tropical cyclones has a goal, to
predict where the storm will be, and a meta-goal, to
build up a sufficient reputation for accuracy that
people will trust the warnings and save their lives
by moving out of harm's way sufficiently early,
based on being in the area of a storm's predicted
landfall.

The meta-goal is really the more important one. Such
storms, singly, have taken in excess of 100,000
lives when inhabitants in the landfall area
disregarded warnings to flee. The only way to avoid
this in the future is for the forecasts to be better
trusted.

Tropical cyclones are _huge_, so Bonnie and Charlie
are affecting one another's motions even though
centered hundreds of miles apart. This makes the
prediction task more difficult, just because there
are many fewer historical examples of tracks of
coupled storms than of single storms, and most of
the forecast models are historical ones (here's what
one did last time) rather than physical ones (here's
where the math says it must go), just because the
math is so stinking hard to get right and slow to
compute. I'd love to see a post-mortem rating of
the predictions in this case.

xanthian.

Do you think, if the naming authority had known in
advance that these two storms would be making near
simultaneous landfall, they'd have had the sense of
humor to name the "C" storm "Clyde"?

Stay safe, Markian, I worry for you. Charlie is a
beast. Even Pammi is on my "stay safe" list, and
she isn't getting Charlie's worst.

"ATCFA" = "Automated Tropical Cyclone Forecasters'
Assistant", FYI.

By the way, I'm sure the sci.geo.meterology folks
are already aware of this, but the rest of the
audience might find it amusing/confusing:

http://www.nrlmry.navy.mil/atcf_web/atcfwb/index1.html

The "graphic" link for each storm shows the current
consensus forecast. The track shown for Charlie has
more of the ATCF output, IIRC the closed dotted
"(vaguely) light bulb shaped area" being the
forecast confidence limit at the outer wind circle
at some percent level I've foregotten, looks like
for 24 hours out, and the circles around the storm
positions showing the wind intensity limits as
forecast, at standard intensity levels.

This all is from nrlmry == Navy Research Lab,
Monterey, California, where I did the above
mentioned software work:

http://www.nrlmry.navy.mil/atcf_web/atcfwb/index.html




--
Posted via Mailgate.ORG Server - http://www.Mailgate.ORG

Kent Paul Dolan wrote:
The meta-goal is really the more important one. Such
storms, singly, have taken in excess of 100,000
lives when inhabitants in the landfall area
disregarded warnings to flee. The only way to avoid
this in the future is for the forecasts to be better
trusted.

This is _mostly_ true, but don't completely rule
out natural selection as a method for increasing
the percentage of the population that heeds
warnings such as, "Hey, dumb****, there's a huge
****ing storm that can pick up you, your trailer,
and your little dog, and set you right down in
Oz. And we mean Australia, not where the Wizard
lives."

The "graphic" link for each storm shows the current
consensus forecast. The track shown for Charlie has
more of the ATCF output, IIRC the closed dotted
"(vaguely) light bulb shaped area" being the
forecast confidence limit at the outer wind circle
at some percent level I've foregotten, looks like
for 24 hours out, and the circles around the storm
positions showing the wind intensity limits as
forecast, at standard intensity levels.

So with some level of confidence, Charlie will
hit somewhere between Panama City and Miami.
It makes evacuation somewhat more difficult
when the entire state has to pick up and drive
to Texas.

Still, you couldn't _pay_ me enough to hang
out in Ft. Myers right now, and I've been through
half-a-dozen Florida Hurricanes.

Kent Paul Dolan wrote:
In the 43 second, roughly, of this TV news video
clip:


http://news.yahoo.com/p/v?u=/ap_av/2...8&f=5374635 3

(which will undoubtedly go away fairly soon, old
news isn't "news" any more) is a map showing a
cluster of predicted storm tracks decorated with
variously colored triangles and squares.

The software that draws it is the ATCFA software I
maintained and enhanced for the US Navy from
1991-1994, or a very close work-alike (mine was used
on the west coast, but "borrowed" by the east coast
forecast offices as well).

What is being depicted is the cluster of predictions
from _many_ forecast models, one per track shown,
with the colored triangles or squares both
identifying the associated model which predicted that
particular storm track, and also showing the place
the storm center is expected to be at a particular
standard _time_, GMT, usually on 6 hour intervals
from midnight.

While no single forecast model has a splendid record
for consistent accuracy, a hand faired track through
the mean predictions of the models that do agree
(the meteorologist gets to pick from dozens of
models, chooses ones with good past results in
similar situations) tends, these days, to be very
predictive indeed.

Forecasting of tropical cyclones has a goal, to
predict where the storm will be, and a meta-goal, to
build up a sufficient reputation for accuracy that
people will trust the warnings and save their lives
by moving out of harm's way sufficiently early,
based on being in the area of a storm's predicted
landfall.

The meta-goal is really the more important one. Such
storms, singly, have taken in excess of 100,000
lives when inhabitants in the landfall area
disregarded warnings to flee. The only way to avoid
this in the future is for the forecasts to be better
trusted.

Tropical cyclones are _huge_, so Bonnie and Charlie
are affecting one another's motions even though
centered hundreds of miles apart. This makes the
prediction task more difficult, just because there
are many fewer historical examples of tracks of
coupled storms than of single storms, and most of
the forecast models are historical ones (here's what
one did last time) rather than physical ones (here's
where the math says it must go), just because the
math is so stinking hard to get right and slow to
compute. I'd love to see a post-mortem rating of
the predictions in this case.

xanthian.

Do you think, if the naming authority had known in
advance that these two storms would be making near
simultaneous landfall, they'd have had the sense of
humor to name the "C" storm "Clyde"?

Stay safe, Markian, I worry for you. Charlie is a
beast. Even Pammi is on my "stay safe" list, and
she isn't getting Charlie's worst.

"ATCFA" = "Automated Tropical Cyclone Forecasters'
Assistant", FYI.

By the way, I'm sure the sci.geo.meterology folks
are already aware of this, but the rest of the
audience might find it amusing/confusing:

http://www.nrlmry.navy.mil/atcf_web/atcfwb/index1.html

The "graphic" link for each storm shows the current
consensus forecast. The track shown for Charlie has
more of the ATCF output, IIRC the closed dotted
"(vaguely) light bulb shaped area" being the
forecast confidence limit at the outer wind circle
at some percent level I've foregotten, looks like
for 24 hours out, and the circles around the storm
positions showing the wind intensity limits as
forecast, at standard intensity levels.

This all is from nrlmry == Navy Research Lab,
Monterey, California, where I did the above
mentioned software work:

http://www.nrlmry.navy.mil/atcf_web/atcfwb/index.html




.... the name's beelzibub and i have net access, FEAR ME!!!

BEELZIBUB
PS; TRUE THIS IS -- YODA

"Kent Paul Dolan" wrote:

The meta-goal is really the more important one. Such
storms, singly, have taken in excess of 100,000
lives when inhabitants in the landfall area
disregarded warnings to flee. The only way to avoid
this in the future is for the forecasts to be better
trusted.

Sigh, and even in the US the message still isn't getting
through. What part of "don't try to ride out a violent
storm in a mobile home" is so difficult to understand
that dying is a simpler option than seeking substantial
shelter?

"Too many bodies to get a count" doesn't sound really
encouraging.

xanthian.


--
Posted via Mailgate.ORG Server - http://www.Mailgate.ORG

On Sat, 14 Aug 2004 14:10:16 +0000, Kent Paul Dolan wrote:

Sigh, and even in the US the message still isn't getting
through. What part of "don't try to ride out a violent
storm in a mobile home" is so difficult to understand
that dying is a simpler option than seeking substantial
shelter?

The fact that people keep doing this should give you a clue that risk
estimation is not a simple thing.

A person will rationally look not just at the worst-case result for
themselves, but the weighted risk, and the costs of avoiding the risk.

Most people in most places can ride out a hurricane, if they take
reasonable precautions. I have done so a couple of times.

How many people would have died without a mass evacuation? A hundred? Two
hundred? How many people die in traffic accidents in that area every
couple of days? Twenty? Thirty? So, riding out the storm is five or ten
times as dangerous as daily transport. Not exactly a risk to cause one to
cringe in fear.

Most people along low-lying coastlines are keenly aware of what areas
flood, and where the tidal surges reach. There is a problem with
newcomers, but that is an educational problem, solved with information.

Having people on-scene after a disaster provides ready labor to quickly
repair or patch things that go bad, and prevent wastage. For instance,
residents of a house can perform minor repairs that might prevent a roof
from tearing off, but they can't do that if they have evacuated.

But, in the extreme, hurricanes and volcanoes and earthquakes can kill a
LOT of people. So, the authorities order very expensive and disruptive
evacuations to avoid the most extreme cases. Individuals make rational
decisions, but sometimes still lose their lives. They take a gamble, and
they lose. Think of the famous hurricane parties on Padre Island, Texas,
and Grand Isle, Louisiana, for example.
--
Incompetence is a double-edged banana.

-- John Perry Barlow, 'Crime And Puzzlement'