Television and internet weather forecasts love to look out as many days in the future as possible to please their viewers. 7-day, 8-day, 10-day, and even 15-day forecasts can be found around television stations and websites. The sad thing is, a lot of people buy into those forecasts and some even use them to make outdoor plans.
The National Center for Environmental Prediction (NCEP) runs the majority of computer models used to create forecasts in the United States. Buried on their web site is this statement:
There is some concern about NCEP model variability from run to run, particularly at extended ranges beyond 7-8 days. The American Meteorological Society Statement adopted on January 14, 2001, titled “Statement on Seasonal to Interannual Climate Prediction” (Bull. Amer. Met. Soc., 82, 701), states, in part, the limit of predictability for day-to-day weather changes for deterministic forecasts is “in the range of 1-2 weeks.” NCEP agrees with this statement and notes that for current state-of-the-art models, the limit of “useful skill” is about 7-8 days. Beyond that range, deterministic numerical weather predictions are characterized by little or no skill and suffer from large run to run variability. However, ensemble or probability forecasts may offer improved predictive skill.
In other words, NOAA, the National Weather Service, etc. all agree that forecasting skill is limited to 7-8 days out. Now, consider that even a 24-hour forecast has some degree of unpredictably and you can imagine the decrease in accuracy as you extend out further into the future from the present. Even at 5-7 days, there’s a reasonable chance of being incorrect despite the “useful skill” in the forecast. Note that the NWS graphical forecasts stop at 5 days (the text does extend to 7 days) and the NHC issues forecasts tracks a maximum of 5 days out.
So, I thought it might be interesting to test these extended forecasts from time to time to see how accurate they are.
Our first guest: Syracuse, NY
The time table: Thur-Fri-Sat, August 27-29
The GFS (Global Forecast System) model, which is the long-range model in the U.S., shows some cooler temperatures moving into upstate New York by the end of next week:
This has resulted in the following forecasts:
Thur: 65/47 Fri: 67/48 Sat: 67/51
Thur: 71/57 Fri: 69/55 (no Sat forecast available yet)
So, let’s follow along and see if this materializes. My guess is that it will be closer to normal temperatures (77/57) than the forecast cool temperatures. I’ve seen such cool spells pop up in long-range forecasts the last couple of weeks only to warm up as the dates approach.