I've never really had a good feel for any correlation between the QBO and SE winter temperatures because there are so many ways to look at the QBO magnitude, sign, and pattern that it can make for very small sample sizes. However, after reading DT's winter outlook, I just decided to analyze similar QBO autumns (back to 1948) to 2017. These are the most similar in regard to sign/magnitude and the magnitude is still slowly rising as of Oct. using this link:
https://www.esrl.noaa.gov/psd/data/correlation/qbo.data
1956-7, 1958-9, 1962-3, 1965-6, 1972-3, 1989-90, 2000-01, 2009-10
I'm throwing out these because they're El Nino winters: 1958-9. 1965-6, 1972-3, and 2009-10.
That leaves these 4, which are all between neutral and weak La Nina: 1956-7, 1962-3, 1989-90, 2000-01
These 4 had the following during winter in the SE US:
Dec: 3 of the 4 very cold; other one (1956) very warm
Jan: varies (2 warm; 2 cold)
Feb: 3 of the 4 very warm, the other one very cold (1963)
Based on the above, alone, for the SE, there is a rather strong signal for a very cold Dec., a hard to predict Jan., and a very warm Feb.
Now looking more closely at these 4 analogs, the very best QBO analog patterns are 1962-3 and 2000-01 because those two shifted from +QBO to -QBO the prior spring, similar to 2017, after a typical 1 year+.+QBO period. In contrast, 1956-7 first shifted into -QBO back in the prior Jan (several months earlier) and 1989-90 first shifted into -QBO way back in May of 1988. So, that is why those not as good 2 analogs' then current -QBO shifted back to +QBO by the following spring whereas 1962-3 and 2000-01 didn't shift back to +QBO until the subsequent fall or winter. Looking at 2017's QBO pattern, it, too, would likely not go back to +QBO until at least the fall of 2018. (This is the problem with using QBO. To get something really similar, you have to eliminate so many potential analogs, thus leaving one with a very small sample....but I digress. And I'm not even considering the potential effect of GW to throw the older analogs off.)
So, we're now left with the very best 2 QBO analogs: 1962-3 and 2000-01. They both had a very cold Dec. This is more evidence that especially if the first part of Dec actually ends up not cold as the models are saying, to look out for the potential of a VERY cold middle and end of Dec. and possibly into early Jan. in the SE. That is the strongest signal from these 2 analogs fwiw with even Jan cool to cold for both of these analogs. Feb. was very warm for one and very cold for the other...so no signal there.
Now, let's look at the -AO in Nov. Like 2017 is about to do, 1962 and 2000, in addition to being the very best QBO analogs with similar ENSO, each went sub -3 in Nov. whereas 1956 and 1989 did not. So, that's another reason to use 1962-3 and 2000-01 over 1956-7 and 1989-90.
The last thing I'll say about these analogs is that 2000-01 was a weak La Nina while 1962-3 was cold neutral. being that I feel we're most likely going to end up with a weak La Nina, I'd say that the very best analog is 2000-01. 2000-01 had a very cold SE Dec. (Aside: 3" of SN KATL, the most in Dec. since way back in 1917!), a cold first half of Jan., a mild 2nd half of Jan. and a mild Feb. DJF averaged ~1.5 colder than normal at KATL thanks to that cold Dec-1st half of Jan.
*****For the fun of it since I just put a decent amount of this time into it, I've just decided to make a winter forecast for the SE US based on the above analysis:
a near normal first half of Dec. (in deference to the current not cold early Dec. model runs and likely temporary warmer change from the recent cool domination), a mainly cold Dec. 16 through Jan 15th, and a mainly mild Jan 16th through Feb 28th. DJF would come out within 2 of normal up or down for most of the SE...so call it near normal DJF (a soooooo much better winter than the last one temperaturewise). I'm still not touching overall wintry precip with a 10 foot pole, at least not right now, though I will go ahead and say that the best shot at a widespread significant SE winter storm would be Dec. 16th through Jan 15th based on my projection of cold then with 12/25-1/15 the most favored as per the longterm climo of few winter storms before Christmas.
https://www.esrl.noaa.gov/psd/data/correlation/qbo.data
1956-7, 1958-9, 1962-3, 1965-6, 1972-3, 1989-90, 2000-01, 2009-10
I'm throwing out these because they're El Nino winters: 1958-9. 1965-6, 1972-3, and 2009-10.
That leaves these 4, which are all between neutral and weak La Nina: 1956-7, 1962-3, 1989-90, 2000-01
These 4 had the following during winter in the SE US:
Dec: 3 of the 4 very cold; other one (1956) very warm
Jan: varies (2 warm; 2 cold)
Feb: 3 of the 4 very warm, the other one very cold (1963)
Based on the above, alone, for the SE, there is a rather strong signal for a very cold Dec., a hard to predict Jan., and a very warm Feb.
Now looking more closely at these 4 analogs, the very best QBO analog patterns are 1962-3 and 2000-01 because those two shifted from +QBO to -QBO the prior spring, similar to 2017, after a typical 1 year+.+QBO period. In contrast, 1956-7 first shifted into -QBO back in the prior Jan (several months earlier) and 1989-90 first shifted into -QBO way back in May of 1988. So, that is why those not as good 2 analogs' then current -QBO shifted back to +QBO by the following spring whereas 1962-3 and 2000-01 didn't shift back to +QBO until the subsequent fall or winter. Looking at 2017's QBO pattern, it, too, would likely not go back to +QBO until at least the fall of 2018. (This is the problem with using QBO. To get something really similar, you have to eliminate so many potential analogs, thus leaving one with a very small sample....but I digress. And I'm not even considering the potential effect of GW to throw the older analogs off.)
So, we're now left with the very best 2 QBO analogs: 1962-3 and 2000-01. They both had a very cold Dec. This is more evidence that especially if the first part of Dec actually ends up not cold as the models are saying, to look out for the potential of a VERY cold middle and end of Dec. and possibly into early Jan. in the SE. That is the strongest signal from these 2 analogs fwiw with even Jan cool to cold for both of these analogs. Feb. was very warm for one and very cold for the other...so no signal there.
Now, let's look at the -AO in Nov. Like 2017 is about to do, 1962 and 2000, in addition to being the very best QBO analogs with similar ENSO, each went sub -3 in Nov. whereas 1956 and 1989 did not. So, that's another reason to use 1962-3 and 2000-01 over 1956-7 and 1989-90.
The last thing I'll say about these analogs is that 2000-01 was a weak La Nina while 1962-3 was cold neutral. being that I feel we're most likely going to end up with a weak La Nina, I'd say that the very best analog is 2000-01. 2000-01 had a very cold SE Dec. (Aside: 3" of SN KATL, the most in Dec. since way back in 1917!), a cold first half of Jan., a mild 2nd half of Jan. and a mild Feb. DJF averaged ~1.5 colder than normal at KATL thanks to that cold Dec-1st half of Jan.
*****For the fun of it since I just put a decent amount of this time into it, I've just decided to make a winter forecast for the SE US based on the above analysis:
a near normal first half of Dec. (in deference to the current not cold early Dec. model runs and likely temporary warmer change from the recent cool domination), a mainly cold Dec. 16 through Jan 15th, and a mainly mild Jan 16th through Feb 28th. DJF would come out within 2 of normal up or down for most of the SE...so call it near normal DJF (a soooooo much better winter than the last one temperaturewise). I'm still not touching overall wintry precip with a 10 foot pole, at least not right now, though I will go ahead and say that the best shot at a widespread significant SE winter storm would be Dec. 16th through Jan 15th based on my projection of cold then with 12/25-1/15 the most favored as per the longterm climo of few winter storms before Christmas.
Last edited:
)
