Chronology Current Month Current Thread Current Date
[Year List] [Month List (current year)] [Date Index] [Thread Index] [Thread Prev] [Thread Next] [Date Prev] [Date Next]

Weather Forecasts (was Laws and Theories)



A couple of weeks ago, Leigh posted the story below about the early state
of weather forecasting. Rather than get in a debate over the relative
scientific merit of weather forecasting, I thought I'd look for some
statistics pointing out the success (or lack thereof) of modern
forecasting.

Unfortunately, I wasn't able to find much. About all I could find was
information about 1-day forecasts of daily high temperatures. A study in
1996 looked at how well two computer forecasts did compared with
"persistence", where "persistence" is the method of making a forecast by
just repeating the previous day's high temperature*. It found that the
average of the two computer forecasts produced a forecast that was off by
4.0 degrees F (rms) compared to 9.2 degrees F (rms) for the persistence
forecast.

Since these computer models use a grid that is not as fine as what is
currently available, I tried to find statistics for a more recent model.
What I found looked at the temperature forecast (by hour) for the Pacific
Northwest but they didn't list the persistence forecast for comparison.
FWIW, it was shown that the temperature forecast for the Pacific Northwest
was off by between 2-3 degrees F over a 48-hr forecast (rms).

For precipitation, I was able to find the following page:
http://sgi62.wwb.noaa.gov:8080/verf/pcpgifs.html
which gave various maps of forecasted and observed precipitation. For
more information, you might want to try
http://www.hpc.ncep.noaa.gov/html/model2.html

There's probably more information out there than this. If anyone finds
anything, let me know.

By the way, considering that the computer models are using observations of
the weather obtained from balloons that do not measure liquid water
content (satellite data is also used to fill in the gaps), I am impressed
with what a few physics equations and a good computer can do.

*I don't debate Leigh's version of the story, but my guess is that
nowadays "experiential" forecasting refers to the forecast made by using
the forecaster's understanding of climatology or "experience" (i.e., what
is the average for that day and what is normal given the conditions).

----------------------------------------------------------
| Robert Cohen Department of Physics |
| East Stroudsburg University |
| bbq@esu.edu East Stroudsburg, PA 18301 |
| http://www.esu.edu/~bbq/ (570) 422-3428 |
----------------------------------------------------------

On Wed, 8 Mar 2000, Leigh Palmer wrote:

[snip]
"Excuse me, sir, but you didn't say anything about the accuracy of
your computer weather forecasting methods. How accurate are your
predictions?"

The presenter made some comments about this being early days in the
development of the methods, but he said that already they had achieved
a rate of success which was 90% that of experiential forecasting. He
then went on to another questioner, but the WHE stuck his hand up
again. Finally it was the only hand still up, and the speaker
recognized him.

The WHE spoke: "Excuse me, sir, but this "experiential forecasting"
to which you referred - just what is experiential forecasting?"

The speaker replied "What is experiential forecasting? Experiential
forecasting is predicting that tomorrow's weather will be like today's."

Leigh