More than 40 Below Average in Peoria?
Image credit for all of the ones in this entry: WSI
So, today at the morning TWC meteorology discussion I led I cynically showed the above map, and with a smile on my face I exclaimed to TWC winter weather expert Tom Niziol, "I dare you to show that on the air!"
Then I posted an entry on my TWC Facebook page, with that graphic and this text: "[Stu Ostro] Will be interested to see in two weeks how this model forecast compares to the actual temperatures where it's predicting them to be more than 40 degrees below average..."
Based on the comments and shares, my cynicism might have been too veiled!
Such a model forecast can get people's attention, with GFS output now publicly and freely available these days out to 384 hours in advance, and such low temperatures would be of much interest and effect.
I followed with a comment which linked to this map, of ECMWF ("European") model ensemble mean departure from average temperatures for approximately the same time (six hours apart, the closest I could get to match 06z GFS operational and 00z ECMWF ensemble runs):
A bit of a difference in the Mississippi Valley!
Let's dig a little deeper ...
While such ensemble mean values might be damped by averaging all of the individual members, not all of them likely to have produced such an extreme deviation, in this case actually virtually all of them show something decidedly different than the solution predicted by that GFS (U.S. NOAA model) operational run.
This map will fry your brain! But it tells the story, and I'll try to walk you through it.
I've drawn, with thick blue, light green, red and purple lines (and with a narrow white line in each to make them more visible), that GFS operational run's values for 516, 546, 576 and 588 decameter 500 millibar heights.
All the narrow squiggly "spaghetti" lines show the corresponding heights for the individual ECMWF ensemble members.
Not only the ensemble means (gold lines) but nearly every single member shows what I've annotated in white type and arrows, compared to the GFS operational run: a stronger ridge over the southeast U.S., a less-sharp trough over the north-central states, a sharper trough with its axis west of Baja, and -- importantly, since it's the difference between more of a Pacific trajectory and a "Polar Express" straight from the North Pole -- a much flatter ridge over western Canada and the Gulf of Alaska.
That's key, as with the polar vortex near Hudson Bay, if a strong arctic surface high could build up and come down from northwest Canada, that'd be a way for the atmosphere to have a shot at producing such frigid air in that part of the U.S.
All of this is less mind-numbingly visualized by these maps which average it all together and plot the 500 mb heights and departures from average.
I'm comparing the GFS operational run to the ECMWF ensemble both because the ECMWF operational run does not go out beyond 240 hours in advance, and because the ECMWF ensemble scheme is more sophisticated than the GFS's with many more members than the GFS ensemble one. Even the GFS ensemble mean, though, was much closer to the ECMWF ensemble mean than to the GFS operational run, with many of its members not showing such an extreme cold solution, and some of them much warmer for that part of the country.
One thing that both the GFS and ECMWF ensembles seem to be more consistently supportive of in their forecasts for a couple weeks from now is a mean trough position and below-average temperatures in the West, and in the meantime there's confidence of another surge of well-above-average temperatures in the eastern two-thirds of the Lower 48 next week ... which BTW will erode the snow cover that has built up, and in turn be a factor working against extreme cold immediately thereafter.
Notwithstanding all that has been documented about the superiority of the ECMWF compared to the GFS and this hilarious animated cartoon about that (warning: not rated G, and NSFW), and the ECMWF's remarkably good long-range forecast for Sandy 204 hours in advance, not every GFS forecast is less accurate than the ECMWF's (case in point Tropical Storm Debby last year, when the GFS predicted it to go to Florida and the ECMWF had it going in the nearly opposite directly to Texas).
And I don't have hard statistics at my fingertips to support this ... but, that having all been stated ...
It's been my (and I'm not the only one) anecdotal impression that in recent winters the GFS has had a bias in the Day 10-16 time frame of frequently overdoing the degree (pun intended) of cold air coming south of the border when it has shown extreme Arctic outbreaks.
We'll see what happens with this case, and maybe on January 17 I'll be posting an entry titled, "The GFS was right: 40 below average in Illinois!"
What's the likelihood of that?
In addition to the questions regarding the models' handling of the upcoming pattern, there's the climatological context about which I posted on FB in another follow-up comment (it would be one of the most extreme arctic outbreaks on record in that part of the country):
The average daily low temperature at that time in Peoria, Ill. in the core of that extreme cold anomaly is 17F. Forty-two degrees below average, as suggested by this model forecast map, would be a low of -25. That would make it the 3rd coldest temp recorded there in 130 years of records and within 2 degrees of the coldest temperature on record in Peoria, -27 in January 1884. The chance of that happening there in two weeks is about the same as _____________________. (fill in the blank)