I forwarded Robert Cohen's request for more information about
weather forecast reliability to our university meterologist, and
this was his reply. He puts out a daily forecast that includes
a detailed explanation of what's going on in the weather patterns
that resulted in the forecast; I've found these to be quite
interesting, and will forward today's when it comes (usually
around noon Central time).
Hope this helps -
Sue Willis
Suzanne Willis, Professor, Physics Department
Northern Illinois University, DeKalb, IL 60115 USA http://niuhep.physics.niu.edu/~willis/ swillis@niu.edu
phone: 815-753-0667 fax: 815-753-8565
Date: Tue, 21 Mar 2000 16:23:17 -0600 (CST)
From: Gilbert Sebenste <sebenste@weather.admin.niu.edu>
To: Suzanne Willis <willis@beauty.physics.niu.edu>
Subject: Re: Any recent data on forecast accuracy?
Hi Suzanne,
I have to make this relatively quick. But to answer your question, in the
last 4 years, there has a been a significant improvement in forecasting
accuracy by computer models. With individual forecasts and forecasters, it
obviously varies. But even here, those scores are climbing.
The best progress over the last few years has been in warning people of
incoming severe thunderstorms and tornadoes. The new Doppler radars have
caused us to go to 25% tornado detection all the way up to over 75% in the
last 5 years. The Probability of Detection (POD) of these events has
increased enormously. However, this is not the only standard they use for
this.
We also use a "false alarm ratio" (FAR). That actually increased quite a
bit. While we are able to see more tornadoes than ever before develop
before our eyes, some almost hit ground...but never do. That's a false
alarm. And because the radar beam can't "see" what's going on at ground
level (only down to about 1-2 thousand feet close to the radar), you have
to guess. Even so, if you have a very strong circulation at 2,000 feet,
you'd be nuts not to put out a warning.
That has what has improved. But even so, accuracy in forecasting depends
on how much actual raw weather balloon and surface observational data we
get, and budget cuts are threatening these data sources. If these cuts
happen, I guarantee you forecast accuracy will decrease. Knowing what has
happened and is happening is paramount for figuring out what *will*
happen.
As for raw numbers, I don't have them, but they do vary by forecaster and
his experience. But the overall trend, even in the last 5 years (including
myself) is up. A "hit" is when a forecaster, in a 24 hour period, gets the
high/low right at or within 3 degrees; precipitation within the category
(which varies), clouds (clear, partly cloudy, mostly cloudy, cloudy),
and wind direction/speed no more than 30 degrees off, speed varies,
allowing for more room for error as the winds increase. Usually +/-5 MPH
*on average* throughout the period.