In order to check the status of the immense cold waves that the GFS predicted last week, I want to use some tools available on AccuWeather.com Professional (30-Day Free Trial!) today to gauge the model's confidence level.

What we talked about last week were the "operational" model run's temperature predictions. The Operational GFS is the version of the model that the government believes is the most accurate. But the hidden truth is that there are up to 10 other versions ("members") of the model that are run each time the Operational produces maps. When you look at those, you get two important things: a "range of possibilities" and a "confidence level" based on how far apart the different predictions are.

Above is an ensemble "plume graph" for the temperatures in New York City. From this, we can tell that the different members of the ensemble are predicting temperatures to be between 1 and 4 degrees F (above zero) Friday morning with the Mean (average) being about 2. Notice, however, that the lines are fairly close together compared to the day before, which means that the model has a good bit of confidence in that range of numbers (a box graph shows that a little more clearly, but removes the extreme outliers). One, and only one member, thinks that it could dip to -2 on Saturday morning, and a couple members believe that the cold snap could repeat itself next week, and possibly dip that low again.

For those of you who don't live in New York City, but are hoping for sub-zero temperatures, here's a map showing where each member is plotting the 0 F (-18 C) temperature line as of Friday morning:

This is called a "spaghetti plot" because it looks as random as spaghetti on a plate after a few days. The colder air is to the north of the lines. This tells me (interestingly) that the mode is more confident about the location of the 0 degree line near New York City than it is over Pittsburgh. All in all though, this is pretty good agreement. But to quantify even further, and produce a more meaningful map, so the last thing we can do is try to translate this into probabilities.

NOTE: The above paragraph was rewritten less than an hour after I launched the blog, when realizing I had pasted in an image from the wrong day above it.

Now it's easy to see that, if you want to be below zero next Friday morning and you live in... Albany, great, enjoy. New York City? Probably not going to happen according to the GFS.

Who else can we ask? Well, the SREF Ensembles, which contain members from the NMM, WRF and RSM models, don't quite stretch out that far yet - they forecast out through Wednesday night. So, we might get a second opinion from them tomorrow. Data from joint project between the U.S. and Canadian government (called the NAEFS Ensembles) is also available. I downloaded this map but am waiting to get confirmation that it is, in fact, Celsius. If so, the average of those ensembles has NYC around 16 F (-9 C) so that would be much warmer than the average that we saw above from the GFS. (On that map, the color indicates the member spread, or confidence, which is pretty good in the NYC area at that time.

Report a Typo