Monday, April 23, 2012

Updating Accu's 25 Day Forecast

For those who buy into the accuracy of long range forecasts, being a day off with a coastal storm three weeks out is not a bad call.  The infamous "April 23rd rain" that AccuWeather's 25 day forecast had projected for our region in its initial call was not a bad forecast in terms of generality.  In terms of getting it right on the right day, which the forecast's mission was supposed to be...it's not horseshoes or hand grenades...

Its revisions as time marches on?  Even worse.

Last week's seven days of forecast tracking compared to reality show an average error of 11.1 degrees per day for the 12th through 18th days of the forecast range.

The forecast missed the intensity of the warm-up (did get the right idea it would be warm) but missed the brief mid-week cooldown that occurred with those showers.  If anything, these long range forecasts might be decent to spot potential trends but not good at determining what the weather will do on a specific day.   I'm not convinced based on the data that having specific high/low forecasts, rainfall are specific days is a good idea as it can merely give the public fuel for the fire in terms of saying that only free swinging first basemen that chase every slider known to man get paid more to do produce less than meteorologists as a collective...or that meteorologists "never get it right."

While the forecast does show the tendency for a cooler/warmer/wetter/drier pattern (although it has occasional misfires)..simple charts and graphics can make that point just as easily as spitting out a high or a low.  I don't think exact temperature predictions are useful in these type of forecasts...and shouldn't be used until we can get the errors down a little bit more.