“Never make predictions, especially about the future.”    Yogi Berra

The calls or e-mails come in from time to time.  People are looking for a forecast beyond the time frame of our 7 day forecast.  It may be a wedding perhaps a family picnic.  Sometimes it prefaced by a comment about a forecast they’ve already heard:

“I saw on weather.com that it’s only supposed to be 55 two weeks from today what do you think……”

I’ve gotta give you a ‘buyer beware’ statement: If you have to make a decision based on a forecast that far out you are taking your chances. Now I know some of you are probably thinking people in glass houses shouldn’t be throwing stones (yes, our forecasts aren’t 100% either) but hear me out.  The computer models we use to help make our forecasts  are pretty good a day or two out in time but the chances for error increases the farther out in time you go. Right now I think the edge of the computer model reliability for a specific daily forecast is 6 or 7 days. Our temperature error out that far is around 4 or 5 degrees in any given month. That’s why we don’t go any farther out in time in our daily extended forecast.  We might be able to give a general trend 10 or 14 days out  but typically we will give you a range of days when ‘colder’ or ‘warmer’ weather will be coming.

Anyhow, I decided to take a look at the error in the temperatures forecast for some of these really extended forecasts available on the web.  During this past December I looked at Accu-Weather and The Weather Channel’s Syracuse forecasts on the web and here is what I found for a 10 day forecast:

Weather Channel 10 day Forecast              8.21 degrees

Accu-Weather 10 Day Forecast                 7.69 degrees

This is more than 3 times the error of a typical 1 or 2 day forecast. Sure, during the month there were days when the 10 day forecast was close, within a degree or two but 1/3 of the time the forecast was off by 10 degrees or more and there were a handful of days where the error was 20 degrees or more!  If you have an important event that hinges on the forecast are you going to trust a forecast that could be off by 20 degrees? Lets look at it a different way.  What if you looked at the Storm Team calendar and just picked the normal high temperature for each of the days in December to make a your 10 day forecast?  It’s called a climatological forecast.  Guess what: you would have done a better job over the course of December than The Weather Channel and Accu-Weather!

Climatology 10 day Forecast                                  7.28 Degrees

Were the December Accu-Weather and Weather Channel numbers a fluke? A while back (May 2006) and I found similar numbers:

Weather Channel 10 day Forecast (May 2006)       7.45 degrees

Accu-Weather 10 Day Forecast (May 2006)          7.48 degrees

Only Accu-Weather makes specific forecasts 2 weeks out in time and their results were mixed.  Last month their error was 5.77 degrees for a 14 day forecast (better than their 10 day forecast) but when I looked at the same type of forecast in May 2006 the error was almost 10 degrees!

Obviously this isn’t a long term, peer reviewed study.  I only looked at two months and I would want to look at more to make a true scientific conclusion. However, even these rough numbers make me wonder if these very specific long, long range forecasts are doing the profession of meteorology as a whole any good. I think sometime in the future the 10 and 14 day forecasts are going to get more reliable but we’re not there yet.

The Lake effect Challenge is up to $22 for the Leukemia & Lymphoma Society and there are now only 5 days until pitchers and catchers report.