On Mon, 24 Apr 2023 12:18:31 -0700 (PDT), "
funkma...@hotmail.com"
<
funkma...@hotmail.com> wrote:
>On Monday, April 24, 2023 at 1:04:14?PM UTC-4, Jeff Liebermann wrote:
>> On Mon, 24 Apr 2023 03:16:59 -0700 (PDT), Andre Jute
>> <
fiul...@yahoo.com> wrote:
>>
>> >Historian Mark Lewis Wonders About Weird Weather:
>> >
https://townhall.com/columnists/marklewis/2023/04/24/weird-weather-n2622307
>> >
>> >Andre Jute
>> >Computer models predicting events a century hence are not Science, they're witchcraft with a thin veneer of crooked statistics.
>> What would recommend as a suitable replacement for today's witchcraft
>> based long range weather predictions? Perhaps a farmers almanac?
>> <
https://www.almanac.com/weather/longrange>
>> I agree, that the current technology isn't very accurate. As the
>> computer models improve and better satellite are launched, the
>> forecasts should improve. However, weather is a chaotic system and
>> modeling chaos is not (yet) a science.
>
>Sure it is. Chaos theory is widely studied and applied across many different fields of study - weather, economics, evolution, linguistics, gaming......it's based on predictive analysis of emergent systems.
I really don't know how much of today's weather forecasts include
chaos theory in the calculations. I'll ask (time permitting).
Edward Lorenz did his work in chaos theory initially by trying to
improve weather forecasting:
<
https://www.encyclopedia.com/environment/energy-government-and-defense-magazines/chaos-theory-and-meteorological-predictions>
"Based on his results, Lorenz stated that it is impossible to predict
the weather accurately."
The basic problem is the weather forecasting by computer is largely an
exercise in getting the initial conditions for the calculations
correct. Give the computer the wrong initial conditions and the
results will soon drift off the charts. Without sufficient and
accurate data, chaos theory is prone to be umm... chaotic. In my
never humble opinion, the real progress in forecasting comes from the
availability of better sensors, better satellite data, and larger
sensor networks. That enables climate scientists to view a 3D picture
of the atmosphere, which produces a much better map of atmospheric
conditions. Try it... Go to:
<
https://www.windy.com/?51.998,-84.902,3>
At the right, click on "Wind". Slide the "altitude" slider up and
down. On the surface, the pictures look like the views shown by the
TV weather personalities. Go up a few thousand feet, and everything
looks very different. Weather events are generated at higher
altitudes and not at ground level.
>I personally don't believe in chaos.
Everywhere I look, I see nothing but chaos and entropy. Organization
is just a temporary illusion that occurs just before everything falls
apart.
>To me, it's just a way of copping out that we don't know enough
>about the system and or the history of it to be predictive beyond
>a certain level of accuracy. "Chaos" and "random" are shorthand
>for "we don't know all the factors influencing the outcome".
Yep, that's about it. Whenever we can't deliver accurate answers, we
back step and produce probabilities. Weather is certainly a
probabilistic system, where the best that can be done is something
like 50% chance of rain, which isn't very useful, but better than zero
information.
>AFAIAC, computer modeling of weather systems is a classic example.
>Your anecdote of not covering your woodpile as an example of how
>the forecast was wrong actually wasn't really wrong. The forecast
>gave you a probability of 5%. Based on the information the models
>came up with, there was a 5% chance due to certain unknowns in
>the data set that it would rain.
Nope. Wrong definition of probability of rain:
<
https://www.abc15.com/weather/what-does-a-30-chance-of-rain-really-mean-how-meteorologists-determine-rain-chances>
"The official definition of the probability of precipitation by the
National Weather Service is the chance of precipitation (rain, snow,
etc.) occurring at any one spot in the area covered by the forecast.
(...)
P = C x A, or the probability of precipitation equals the
meteorologist’s confidence that it will rain, times the percentage of
the area that is expected to get rainfall."
As of 2016, the grid size for the "area covered" is 9km by 9km for the
ECMWF model. That's noticeably better than the previous 16km grid.
<
https://www.ecmwf.int/en/about/media-centre/news/2016/new-forecast-model-cycle-brings-highest-ever-resolution>
"Ensemble forecasts, which describe the range of possible weather
scenarios and the likelihood of their occurrence, usually use twice
the grid spacing of high-resolution forecasts. They are now at 18 km
up to forecast day 15 and 36 km thereafter."
>The forecast was not only not wrong, it was right. Those unknowns
>fit the the 5% model, and it rained.
My problem wasn't the 5%. It was the sudden appearance, apparently
out of nowhere, of the 5% after I went to bed. I was irritated enough
turn on my desktop and check if anything was forecast. Yep, it was
there at 3am, but wasn't there the previous night at about 10pm.
NWS has 3 and 7 day histories on their web pile. The drizzle was so
little, that it didn't show on the accumulated rainfall graph.
However, the history did show two large peaks of 100% RH (relative
humidity), which is a good indication that it might be raining.
<
https://www.weather.gov/wrh/timeseries?site=BNDC1>
It might be gone by the time you look at the above link, so I saved a
screen dump:
<
http://www.learnbydestroying.com/jeffl/crud/Drizzle.jpg>
The red circle shows when it rained on Apr 19 (at 3am). There's a 2nd
peak, where it probably tried to rain, but I didn't notice. This kind
of post-mortem comparison between forecast and reality should probably
be done more often:
"How Reliable Are Weather Forecasts?"
<
https://scijinks.gov/forecast-reliability/>
"A seven-day forecast can accurately predict the weather about 80
percent of the time and a five-day forecast can accurately predict the
weather approximately 90 percent of the time. However, a 10-day or
longer forecast is only right about half the time."
>If they said there was a 0% chance of rain, and it rained anyway,
>you could have said they were wrong.
True, but is that because they think they are highly confident that it
will rain in a very tiny part of an expected area, or that they are
lacking in confidence that it will rain over a large area? I've seen
it rain quite heavily in a small area but nowhere else (in Hawaii).
>But even then we're dealing with a statistical probability, which
>in reality is _never_ 0% when you're dealing with a 'chaotic' system.
True. Most models fall apart at extremes.
>The NOAA claims 36 hour weather forecasting has increased in accuracy
>from ~ 25% in the 1930s to well over 90% today.
>
https://www.weather.gov/about/models.
>The reason is that they have access to more data to build the
>predictions and better tools to process it.
Right. It's the mountains of data that make today's forecasting
possible. The next improvement might be weather data gathering
drones, to deal with urban microclimates. However, we first need to
cleanup our existing sensor network. Here's the sensor head at a
local TV station office sited over the HVAC system:
<
http://www.learnbydestroying.com/jeffl/crud/KSBW-WX-Station.jpg>
Anthony Watts used to have photos of similarly badly sited weather
stations on his blog, but I can't find them.
>This brings me back to my original point: the more we know
>about the past, the better we will be about predicting the future.
"Past performance is no guarantee of future results"