GFS: Reliable Computer Model? Overview

3167 reads link

By Quincy Vagell on February 17, 2013, 1:50pm Last modified: February 18, 2013, 6:39pm

Use your ← → (arrow) keys to browse more stories.

Next Article
GFS forecast from Saturday morning for heavy snow Sunday morning, which didn't verify in Connecticut.

The Global Forecast System, or GFS, is a commonly used computer forecast model.

There are many different computer models for weather forecasting, but the GFS is basically the standard for American use. Other models, such as the European (ECMWF) are based in other countries. Both, however, do give forecasts for the northeastern United States.

After a series of major weather events and model “mishaps” over the past few months, the reliability of the GFS is in serious question. Whether it was issues with Hurricane Sandy, the Blizzard of 2013 or even this weekend’s snow “storm,” the GFS has shown time after time that it is not the most reliable of computer models.

Hurricane Sandy:
As early as 10 days out, there were hints from the ECMWF that a large storm system could be looming along the East Coast. Within 7 days, the ECMWF locked in on the general idea that a hurricane would transform into a monster storm and make landfall along the mid-Atlantic. The GFS kept predicting that the storm would pass out to sea. It wasn’t until about 4 days before landfall that the GFS gradually shifted the track back into New England, but it was one of the last models to get the landfall correct, in New Jersey.

Blizzard of 2013:
In similar fashion to Hurricane Sandy, the ECMWF was one of the first models to lock in on a historic snow storm for portions of the Northeast. In fact, more than two full days before the storm, the model was forecasting 2 to 3 foot snowfall amounts in Connecticut. The GFS was much more conservative and even on Friday morning, the day that snow stated to fall, was showing closer to 1 foot amounts across Connecticut. This even led some forecasters in the state to decrease snowfall forecasts that morning. The end result was that much of the state received over 2 feet of snow and virtually all of New Haven County ended up with close to 3 feet of snow.

This weekend:
Although a much smaller scale event, the GFS on at least a few computer forecast “runs” was predicting heavy snow across Connecticut. The morning (12z) run on Saturday forecasted 6 to 12 inches of snow across the eastern half of the state. The result was that the National Weather Service (NWS) even issued Winter Weather Advisories. Most areas wound up with less than 2” of snow, except right near the Rhode Island border. Interestingly enough, the ECMWF accurately predicted less than 3” of snow for most of the state and that model was consistent with that.

When it comes to the models, sometimes it’s not just the accuracy of the forecast, but the run to run consistency. The ECMWF creates forecasts twice per day, but the GFS releases forecasts four times. Perhaps that is one of the problems. The GFS not only produces more forecasts, but it has a track record of flip-flopping with each new forecast released every 6 hours. While the ECMWF is usually fairly consistent with each run every 12 hours, the GFS often produces drastically different forecasts twice as often. The end result is a lot of confusion in the weather community, at least for those forecasters who rely heavily upon the GFS for forecasts.

Why bother with the GFS?
There are many reasons…
-Amount of forecasts made publically available
-Frequency of forecast updates (every six hours)
-Based out of the United States
Although there is some benefit of using the GFS, when compared to the ECMWF with the big picture, the model is simply much less reliable.

It’s hard to break a habit:
If we go back a few years, the ECMWF was less commonly used and known about. In fact, going back to my college days, I mentioned the model in 2006 and most students and even faculty members at the time had virtually no knowledge of the model and most had probably never used it.
Television meteorologists, overall, have a reputation of using the GFS more than every other computer model. This isn’t to say that every meteorologist on TV is guilty of this, but it’s more common than not. There are many reasons why this happens. It’s not just habit, but it’s so easy to get updated data from the GFS every 6 hours. This works nicely with on-air broadcasts, as they line up pretty closely with the GFS forecast schedule.

Final thoughts:
The GFS has many uses and it has, on occasion, outperformed the ECMWF. As is with any computer model, it should be used as “forecast guidance.” In combination with other models, data and trends, the GFS can be useful. With that said, when it comes down to a ECMWF vs. GFS forecast, I will usually side with the ECMWF.

For a more technical discussion on this topic, click here.

Use your ← → (arrow) keys to browse more stories.

Next Article


comments powered by Disqus

Quincy Vagell

Town: Marlborough, CT  

Reporting for WXedge since January 2012.

Articles: 1304

Quincy Vagell's Bio

Become a WXedge become a contributor

Let Your Voice Be Heard

Have a question? A comment? A complaint? Connecticut's Weather Experts are here for you. Our team is here to service your every need. Go ahead, let them have it.