Winter '12/ '13 - Arctic Oscillation - Is November A Leading Indicator?

UPDATE (14-DEC-14)
Winter '12 / '13 AO:  -1.222

UPDATE (04-DEC-12)
NOV/s AO:   -0.111.
Probability the average AO is negative for meteorological winter '12 / '13:  76%

---
Little doubt NOV/s index will come in negative despite an early 15-day run of above zero values.

A 2x2 contingency table of NOV/s AO index and winter/s average (D-J-F) index suggests there/s a strong association between the two when NOV/s AO is < 0.



When NOV/s AO is < 0...there/s a 76% chance the winter/s average AO will also be negative.  The table's precision is 63%...where precision is the proportion of negative cases predicted correctly [a / (a + c)].

Winter '12 / '13 - Outlooks

Thread updated periodically as new outlooks and forecasts are issued from a variety of sources.

---
Dr. Todd Crawford - WSI Chief Meteorologist...

While this El Nino event now appears to be a dud, most of our more skillful climate indicators suggest...very cold air will be in plentiful supply across western and central Canada this winter, with frequent border crossings into the northern US.

Currently, we do not expect the kind of frequent atmospheric blocking in the North Atlantic (otherwise known as the negative NAO) that would result in more extreme and more widespread cold in the eastern US.  However, trends in some of the long-lead indicators suggest...this assumption may be challenged, and...the risk to the forecast in the eastern US is towards (sic) the colder side.

In the NE US...
DEC - colder than normal, except northern New England
JAN - colder than normal
FEB - warmer than normal

More...

---
Harvey Leonard, Chief Meteorologist, WCVB-TV (Boston...MA):

...expect a somewhat colder than average winter, with at least average snowfall, and possibly somewhat above average snowfall.

More...

Read more »

Winter '12 / '13 - North Atlantic Oscillation - Trends

Chart 1 - meteorological winter/s average NAO and 9-point binomial filter.
Binomial filter removes noise from signal to reveal trends.

1950s through 1970s...predominantly negative
1980 ...period of transition toward positive
1990s through 2000s...predominantly positive...trending toward negative in the 2010s.



Chart 2 - tally of all monthly NAO sign by calender year.

Noisy signal during most decades; however...positive index began to prevail during the late 1990s.

Value of '6' are years where positive and negative months during the calender year were equal.



Chart 3 (above) - tally of the monthly NAO sign by half-decade.

Slight advantage for negative index during the 1960s through mid-1970s...alternating during late 1970s through the 1980s...strong positive index prevailed during the late 1980s through the early 2000s...and negative ever since.

Value of '30' are periods where positive and negative months were equal.

---
Chart 4 - tally of the monthly NAO sign by decade.

Positive decadal index between the 1970s until 2000s after which negative has been the trend.
Value of '60' are periods where positive and negative months were equal.



The 5-year moving average during meteorological winter turned negative in 2009 has been negative ever since.  The last time the 5-year moving average was negative occurred between 1977 - 1981 and 1954 - 1972.

Trending:  negative NAO.

Winter '12 / '13 - North Atlantic Oscillation - November As Oracle

As it turns out...there/s a strong association between the sign of November/s monthly NAO index and the sign of the average NAO index during meteorological winter (D-J-F).  It's strongest when NOV/s NAO index is positive.

A chi-square test on NOV/s monthly values and average value of winter/s NAO suggests the sign of NOV/s NAO has a measure of predictive skill.

The 2x2 contingency counts the number of events when the the sign of winter/s NAO matched the sign of NAO in NOV.

The distribution of events offers these probabilities:
70%  + NOV ==> + Winter
30%  + NOV ==> - Winer

43% - NOV ==> + Winter
57% - NOV ==> - Winter

---
If the signs of NOV/s NAO and winter/s NAO events were independent / unrelated...then the distribution of events would follow a normal 'bell' curve where...

Positive NOV and positive (negative) winter:  15 (12) events
Negative NOV and positive winter (negative):  19 (16) events

The 2x2 contingency analysis reveals the sign of winter/s NAO is not independent (3.1% chance of Type I error) from the sign of NOV/s NAO.

---
As a footnote...OCT/s NAO has no 'predictive' value b/c it is independent from the sign of winter's NAO.

---
CPC/s monthly NAO index here.

Winter '12 / '13 - North Atlantic Oscillation: Analogs

Analog years selected based on the smallest sum of square errors between current year/s monthly NAO index and historical values since 1950.

It's notable how every analog becomes positive to ever-so-slightly negative during the coming winter. 

ENSO (MEI):
1998 - moderate la Nina.
1993 - warm la Nada after a moderate el Nino
2008 - weak la Nina
2003 - warm la Nada
2004 - weak el Nino

Best match:  2003

PDO:
1998 - weak negative
1993 - moderate positive
2008 - moderate negative
2003 - weak positive
2004 - weak positive

Best match:  2008

QBO:
1998 - weakening easterlies, flip to westerlies in JAN
1993 - strengthening easterlies
2008 - mid-cycle westerlies
2003 - weakening easterlies, flip to westerlies in FEB
2004 - weakening westerlies, flip to easterlies in JAN

Best match:  1998; 2003

---
Season-total snowfall at NEWxSFC Contest stations during Winter '03 / '04 totaled 999"...8" above period-of-record normal.

Next up:  period-of-record trend analysis and the correlation between the northern annular oscillation phase (NAO and the Arctic Oscillation) and sunspots.  Spoiler alert:  not lookin'good.

Winter '12 / '13 - North Atlantic Oscillation: Long-range Signals and Long-term Trends

Up until a few years ago...the UKMET office issued a long-lead forecast of the North Atlantic Oscillation's (NAO) phase for the upcoming winter.  The forecast was based on the correlation between May's sea-surface temperature anomalies (SSTA) in the western Atlantic ocean and the phase of the NAO.  Geo-potential height anomalies at 500 mb in the same general region were also found to have predictive value.

They claimed these correlations had predictive value ~67% of the time...a far cry better than the brain-dean CW about how seasonal NAO forecasting is unpossible.

If May's SSTA matches the pattern in the western Atlantic ocean...then a +NAO is expected to be the dominant state during the upcoming winter.  Above normal SST in the offshore waters of the NE coast...cold water SE of Iceland and warm anomalies to the northeast point toward a +NAO.  If the pattern is reversed...then a -NAO is forecast.

---

This year...May's SSTA were a good match to the anomaly pattern shown above off the NE coast (1) but not so much elsewhere (2,3).

Area 1...the offshore waters of the NE...+SSTAs ==> +NAO

Area 2...SW of Iceland...+SSTAs ==> -NAO

Area 3...NE of Iceland...-SST anomalies ==> -NAO

SSTAs offer a decidedly mixed signal.

---

At 500 mb anomalies for MAY-12...

Positive height anomalies over Greenland ==> -NAO
Positive height anomalies over western Europe ==> +NAO
The effect of a negative action center INVOF the Azores is unknown.
---
This winter's NAO signal based on the UKMET/s long-lead conceptual model is unambiguously ambiguous.

Sea-surface temperature anomalies off the NE coast suggest +NAO while SSTAs off the coast of Greenland suggest -NAO.  Height anomalies at 500 mb over Greenland signal -NAO and a + NAO over western Europe.  The effect of the Azores' strong negative signal is unknown.

Next up:  Analogs

---
Earlier posts about the UKMET NAO forecast here.

Winter '12 / '13 - Eurasia Snow Cover - October

Period of record median:  9,599,610 km²
October-12:  11,106,205 km²

October-12 Eurasia snow cover...as reported by Rutgers Snow Lab...is 16% above normal.  This bodes well for a cold and snowy winter in the eastern US...according to the statistically significant correlation between these global two features discovered by AER's Dr. Judah Cohen.

...Eurasian snow cover during this year’s October has increased far more rapidly than...is normally the case (the corresponding index value exceeds two-standard deviations).

Given the large anomaly for the predictor variable, the following winter will be a good test for assessing if the approach is suitable for real-time seasonal forecasts.¹

Eurasia OCT snow cover : season-total snowfall (D-J-F-M) at NE and MA Contest stations in recent winters...
'09 / '10:  11,288,371 km² - 1,225" (observed*) v. 908" (normal*)
'10 / '11:  10,615,933 km² - 1,426" (observed) v. 921" (normal)
'11 / '12:  9,293,740 km² - 393" (observed) v. 934" (normal)

* SBY removed for lack of station snowfall data.



Blue dotted - observed snow cover
Red - 9-point binomial filter
Orange - period of record median snow cover
Yellow dashed - one standard deviation

The 9-point binomial fiter shows the long-term trend by removing noise from the signal.



---
Reading room...
- A new index for more accurate winter predictions (pdf)
GEOPHYSICAL RESEARCH LETTERS, VOL. 38, L21701

- Eurasian Snow Cover Variability and Links with Stratosphere-Troposphere

Coupling and Their Potential Use in Seasonal to Decadal Climate Predictions (pdf)

Climate Test Bed Joint Seminar Series

NCEP, Camp Springs, Maryland, 22 June 2011