Arctic Oscillation Analysis and Forecasts April 21, 2016

April 21, 2016

Special recap of this past winter

Below I try to weave a narrative of this past winter.  The predominant thinking in the field is that sequential weather or synoptic events are random and unrelated known as chaos theory.  However I have tried to show how different events throughout the winter are related and one follows the next more akin to a domino effect and events at the beginning of the winter are responsible for events at the end of the winter.  I feel that this is a unique perspective on the winter season that can only be accomplished through consideration of troposphere-stratosphere coupling and involvement of the polar vortex.  Both the hemispheric weather and the polar vortex showed extreme variability this winter and an in-depth understanding of the variability of each is impossible without the other.  Such an explanation of the extreme variability is much more difficult using El Nino Southern Oscillation (ENSO), which showed very little variability throughout the winter.


  • Winter 2015/16 is probably best characterized by record warmth.  Every month since May 2015 has been record warm and the streak of record warm months likely peaked in February 2016.
  • Stratosphere-troposphere coupling was very active this winter and in my opinion dominated the dynamical story of winter 2016.   I believe that the dynamical evolution of the winter cannot be understood without analyzing the polar vortex (PV).
  • The stratospheric PV experienced extreme variability this winter, starting the winter record strong and ending the winter record weak.  But typically it is the first half that is most influential on surface weather and the record strong early PV likely contributed strongly to the warm winter.  Though the record weak PV in March is contributing to a cool spring in the North Atlantic sector.
  • Poleward heat flux started slowly but was unusually active during the winter months.  Spikes in winter surface temperatures coincided with peaks in poleward heat flux.
  • Despite the record strong PV, the Arctic Oscillation (AO) was slightly negative for the winter.  The AO bottomed out in mid-January, which coincided with a record snowstorm for the Mid-Atlantic and was the start of a troposphere-stratosphere-troposphere coupling event that resulted in record cold for the Northeastern US in mid-February.
  • The Niño 3.4 index achieved record high values heading into the winter and the strong El Niño dominated the narrative of this winter.  However the atmospheric response did not match the forecasts and in fact temperature and especially precipitation anomalies more closely resembled those associated with La Niña than El Niño.   This winter should have showcased using ENSO for climate forecasts, instead in my opinion, it highlighted the vulnerability of relying solely on ENSO in producing seasonal forecasts.

Boundary Forcings

For almost two years the climate community had been anticipating a strong El Niño event, which did finally materialize the fall and winter of 2015/16.  The Niño 3.4 index set a new record warm temperature of 29.60 in November 2015 and was record warm for the winter months as well.  El Niño of winter 2015/16 was by all metrics one of the strongest ever observed along with 1982/83 and 1997/98.  The observed impacts of El Niño in the Northern Hemisphere extratropics are strongest across North America.  El Niño is related to below normal temperatures in the Southeastern United States (US) and above normal temperatures in the Northwestern US, Western Canada, Alaska and to a lesser degree the Plains of the Northern US and Southern Canada.  El Niño is also related to above normal precipitation across the Southern US, stretching from Southern California to the Southeastern US with below normal precipitation across the Northern US especially the Pacific Northwest and the Great Lakes. 

Though at AER we use ENSO in producing seasonal forecasts, we have pioneered the use of Arctic boundary forcings in winter seasonal forecasting including Arctic sea ice but especially Eurasian snow cover in October.  We have demonstrated using observational analysis and model perturbation experiments that extensive Eurasian October snow cover is related to/can force a strengthened Siberian high, increased poleward heat flux, a weak PV and culminates in an extended period of a negative AO.  A negative AO is associated with below normal temperatures in the Eastern US and Northern Eurasia including Northern Europe and East Asia.  A negative AO is also related to wet conditions across Southern Europe and the Mediterranean but dry conditions across Northern Europe. Scientists including those at AER, have shown a similar atmospheric response to low Arctic sea ice.  There are different ideas how variability in Arctic sea ice might influence winter hemispheric weather but the trend has been a convergence to a similar set of mechanisms first proposed for Eurasian snow cover.  Also there is growing consensus that it is Barents-Kara sea ice in the late fall and early winter that has the greatest impact.  Therefore low Barents-Kara sea ice in November for example favors a strengthened Siberian high, increased poleward heat flux, a weak PV and finally a negative AO.  An important point in regards to the Siberian high, it strengthens or expands northwest of the climatological center.  For low snow cover and/or high sea ice the opposite occurs. 

October 2015 Eurasian snow cover was the fifth highest observed going back to 1972.  Not a record like El Niño but certainly well above normal and favored a sudden stratospheric warming (SSW) and a weak PV in mid winter followed by a negative AO and cold temperatures across the NH mid-latitudes.  September 2015 Arctic sea ice was the fourth lowest observed and November 2015 Barents-Kara sea ice was third lowest (based on my own calculations) going back to 1979.  Therefore Arctic sea ice similarly favored a weak PV in mid winter followed by a negative AO and cold temperatures across the NH mid-latitudes. 

Also heading into the winter the globe had set record warm months for every month since May 2015.  The biggest contributors to the record warmth were the global oceans.  In November 2015 both the Northern and Southern Hemisphere oceans were record warm in large part due to the record warm sea surface temperatures (SSTs) in the equatorial Pacific.  Finally it is probably important to mention that the quasi-biennial oscillation (QBO) was in its westerly phase.  The QBO is a periodic oscillation of the zonal winds in the equatorial stratosphere and in the westerly phase the zonal winds are stronger.  The westerly phase is thought to inhibit the strongest SSWs where the zonal wind reverses in direction and is referred to as a major midwinter warming.

Early Winter

As mentioned above October 2015 Eurasian snow cover was the fifth highest observed going back to 1972.  Above normal snow cover across Siberia in October favors a strengthened Siberian high in November and into December with the largest positive sea level pressure anomalies northwest of the climatological center (see Figure 1a taken from Cohen et al. 2014).  The Siberian high did strengthen in November 2015 but to the northeast of the climatological center and even into the North Pacific (Figure 1b).  This did not form the tripole pattern, which is optimal for forcing increased vertical transfer of Rossby wave energy (vertical wave activity flux or WAFz) and poleward heat flux but rather more of a dipole pattern.  The WAFz plot in Figure 2 shows that overall the WAFZ was quiet for much of November and early December, with the exception of late November.  Quiet WAFz allowed for the polar vortex to quickly spin up and the plot of the polar cap geopotential height anomalies (PCH) in Figure 3 shows that the stratospheric polar vortex became exceptionally strong (record strong based on the zonal wind at 10 hPa and 60°N) in early to mid-December.  Figure 4a shows the 10 hPa geopotential height and a very strong and circular PV can be seen with the cold air trapped in the center in mid-December.  A strong westerly jet for much of December resulted in a mild westerly flow of air across the NH continents inhibiting Arctic outbreaks to lower latitudes.

Figure 1. a) Regression of November SLP anomalies (hPa) onto October monthly mean, October Eurasian SCE (contouring) and December meridional heat flux anomalies at 100 hPa, averaged between 40-80°N (shading). b) Observed average sea level pressure (hPa; contours) and sea level ressure anomalies (hPa; shading) across the Northern Hemisphere from November 1, 2015 through November 30, 2015.

In my opinion the weather or synoptic events of late December were the most important of the entire winter.  Consistent with the high snow cover, the WAFz became much more active the second half of December (Figure 2).   More active WAFz is related to stronger poleward heat flux and this period coincided with well above normal temperatures across the Eastern US and Europe (Figure 5a).  However it also finally started to perturb the strong PV and caused it to become elongated at the end of December (Figure 4b).  If you follow the isobars around the PV, strong southerly winds extend from as far south as Mexico and all the way to the Arctic region north of Europe.  This likely helped an exceptionally warm mass of tropical origin, warmed by the record strong El Niño, come up through Mexico entering the US through Texas and resulting in an unusual tornado outbreak, pass through the Northeastern US resulting in record warm temperatures during the holiday week and then finally pass into the North Atlantic spinning up a potent extratropical cyclone (named Frank), bringing damaging winds to the United Kingdom and finally entering the Arctic resulting in above freezing temperatures at the North Pole.

Figure 2. Observed daily vertical component of the wave activity flux (WAFz) standardized anomalies, averaged poleward of 40-80°N from October 1 through March 31 (data for first week is missing).

But probably the most critical component of the active WAFz was that even though it perturbed the PV it did little to significantly weaken it.  The stratospheric PV remained strong to record strong during and after the active WAFz in late December.  A plot of vectors of zonally averaged WAFz (Figure 6a) explains why.  Though the WAFz vectors are indeed pointed up they are also pointed towards the equator and away from the North Pole.  Momentum is advected opposite of the arrows so if the arrows are directed towards the equator, momentum is being advected towards the Pole, causing the PV to accelerate. So instead of a weak PV or even a major SSW in early January, the PV remained intact and strong well into February all but insuring a mild winter.  Why did the WAFz go up and towards the equator and not the North Pole? An important question and I don’t know the answer but it may be related to the westerly QBO.

Figure 3. Observed daily polar cap height (i.e, area-averaged geopotential heights poleward of 60°N) standardized anomalies from October 1 through March 31 (data for first week is missing). 

Mid winter

Though the WAFz event in late December failed to significantly weaken the stratospheric PV (and the PCH plot shows a cold stratospheric PV for the duration of January), the incredible advection of heat into the central Arctic had significantly warmed the PCH in the troposphere and forced the AO deep into negative territory.  The warm PCH and the high latitude blocking associated with a negative AO, resulted in the most extensive and sustained period of below normal temperatures across the NH of the winter (Figure 5b).  During the month of January, much of the Central and Eastern US, Central and Eastern Europe and East Asia were cold.  And when the AO rebounds from strong negative values, the US East Coast is particularly vulnerable to a snowstorm and this January was no exception.  The East Coast blizzard of January 22 and 23 was probably the most prolific snow producer between Washington DC and New York City ever observed.  However the warm PCH and negative AO without being coupled to the stratosphere was not sustainable beyond a few weeks and the AO turned positive and temperatures warmed once again across the NH.

Figure 4. a) Observed 10 hPa geopotential heights (dam; contours) and geopotential height anomalies (m; shading) across the Northern Hemisphere for 14 December 2015 b) same as a) except forecasted for 21 December 2015 c) same as a) except forecasted for 21 March 2016.

Still the warm PCH in the troposphere and cold PCH in the stratosphere is a sign of strong baroclinicity (strong gradient or changes with height) and the warm PCH/negative AO of mid January was a robust tropospheric precursor and initiated a troposphere-stratosphere-troposphere (T-S-T) coupling event.  Once again the WAFz and poleward heat transport became very active in late January and early February (Figure 2).  But looking at the WAFz vectors from early February (Figure 6b), they directed more towards the North Pole certainly much more than the previous event in December and the PCH in the stratosphere finally turned warm for the first time all winter the second week of February (Figure 3).  The SSW event was robust enough to result in record warm temperatures at the North Pole in the stratosphere.  There was an almost immediate response in the troposphere as well and the PCH turned warm in troposphere centered on February 12th.   This resulted in a record cold air outbreak into the Northeastern US and Boston recorded its coldest temperatures in 60 years.  I have not done a rigorous study but my sense is that an historic cold air outbreak is highly unusual in a record warm winter and exemplifies the exceptional volatility of the atmosphere.

Figure 5. a) Observed surface temperature anomalies ((°C; shading) for December 15-31, 2015 through February 2016 b) observed surface temperature anomalies ((°C; shading) for January 1-25, 2016 c) observed surface temperature anomalies ((°C; shading) for February 15, 2016 through March 15, 2016.

Late winter

Still the warming of the PCH both in the stratosphere and troposphere was sharp but brief and the PCH turned cold especially in the stratosphere the third week of February (Figure 3).  The last week of February once more featured strong baroclinicity with a warm PCH in the troposphere but cold in the stratosphere and the WAFZ and poleward heat flux remained exceptionally active (Figure 2).  At the surface this meant more record warm temperatures (Figure 5c).  However the WAFz vectors were directed at the North Pole more than any of the previous WAFz events (Figure 6c).  The sustained (and far as I can tell unprecedented) active WAFz finally forced a PV split (Figure 4c; something that I anticipated back in the fall in a Capital Weather Gang column, though admittedly later than I thought), a major mid winter warming and an all time record weak stratospheric PV (based on the zonal wind at 10 hPa and 60°N).

Figure 6. a) Observed vertical component of the wave activity flux (WAFz) zonally and shown as vectors averaged for December 15-23, 2015 b) same as a) except forecasted for February 3-12, 2016 2015 c) same as a) except forecasted for February 21 - March 5, 2016.

The PV split completed the second stage of a T-S-T coupling event associated with Arctic boundary forcings.  The final act is the downward propagation or “dripping” paint of warm PCH anomalies from the stratosphere to the troposphere.  I did question how typical the downward descent would be given the lateness in the season of the SSW.  However the PCH from April (Figure 7) shows a textbook case of the downward propagation of the warm PCH, and the negative AO and the even more elusive negative North Atlantic Oscillation (NAO).  The AO for the latter half of April is predicted to be the most negative since mid-January completing the T-S-T coupling event.  Impacts to the sensible weather from the PV split have included unusual late season snowfalls to the Eastern US and a cool spring to Western Europe.  And as I tweeted out the winter ended the polar vortex opposite of how it began.  Though I do believe that the full impacts to sensible weather are not as strong compared to a similar event occurring in mid-winter and that the impact on our winter weather from the strong PV was greater than that from the weak PV.

Figure 7. Observed and predicted daily polar cap height (i.e, area-averaged geopotential heights poleward of 60°N) standardized anomalies. The forecasts are from the 00Z 18 April 2016 GFS ensemble.

Observed winter circulation

In my opinion the main dynamical forcing that dominated the winter seasonal means was the unprecedented active WAFz or poleward heat transport.  Each event transported warm air from the subtropics and even tropics poleward across Eastern North America and western Eurasia.  And with the global oceans at record warm temperatures fueled by a record El Niño and no subtropical Jet Stream (Figure 12b), warm air flowed northward unimpeded leading to repeated record warm temperatures.

Figure 8. Observed average 500 hPa geopotential heights (dam; contours) and geopotential height anomalies (m; shading) across the Northern Hemisphere from December 1, 2015 through February 28, 2016.

Looking at the winter mean 500 hPa geopotential height (Figure 8) it is easy to see why the WAFz and poleward heat transport was so active and eventually culminated in a PV split and a record weak stratospheric PV (Figure 4c).  The atmospheric circulation stayed for much of the winter in a very distinguishable wave 2 pattern.  Wave-1 and wave-2 are the only waves that generate vertically propagating waves that penetrate into the stratosphere and wave-2 most often results in a PV split.   The strongest geopotential height anomaly in the NH is an area of positive geopotential height anomalies centered near the Ural Mountains and is commonly referred to as Ural Blocking.  This atmospheric feature has been linked to extensive Eurasian snow cover and low sea ice in the Barents-Kara seas.  In my opinion, this feature was critical to the hemispheric atmospheric circulation and supports the idea that the Arctic not only influences the PV but mid-latitude weather.  Moving downstream the next atmospheric feature is an area of negative geopotential height anomalies across East Asia and the North Pacific.  Troughing in East Asia is again linked with extensive Eurasian snow cover and the strengthened Aleutian Low is linked to El Niño.

Figure 9. The latest weekly-mean global SST anomalies (ending 31 January 2016). Data from NOAA OI High-Resolution dataset.

The third dominant atmospheric feature is a positive geopotential height anomaly center over eastern North America and the western North Atlantic.  The final feature is a negative geopotential height anomaly center in the eastern North Atlantic or a deepened Icelandic Low.  Unlike the first two atmospheric features, there are no previously demonstrated boundary forcings discussed in the scientific literature related to the last two features.  However the southwest-northeast dipole in the atmosphere nicely matches a similar dipole in the SSTs (Figure 9) with above normal SSTs off the East Coast of the US and below normal SSTs south of Iceland and west of Europe.  Attributing atmospheric circulation anomalies to mid-latitude SST anomalies is a challenge and becomes a chicken and egg problem.

Figure 10. a) AER Forecasted surface temperature anomalies ((°C; shading) for December 2015 through February 2016 b) observed surface temperature anomalies ((°C; shading) for December 2015 through February 2016 c) NMME Forecasted surface temperature anomalies (°C; shading) for December 2015 through February 2016.

However the cold SSTs in the North Atlantic certainly preceded the winter and are hypothesized to be a consequence of the slowdown of the ocean’s thermohaline circulation due to melting ice and have been referred to in the media as the “cold blob.”  The slowdown of the thermohaline circulation is further predicted to result in cold temperatures for Europe.  This winter was just the opposite and instead was exceptionally warm.  It is plausible that colder than normal SSTs in the northern North Atlantic favored below normal heights while warmer than normal SSTs further south across the North Atlantic, favored above normal heights and together increased the north-south temperature gradient in the North Atlantic and a stronger Jet Stream.  The Jet Stream was strengthened across the mid-latitudes of the North Atlantic this past winter (see Figure 12b) flooding Europe with mild maritime air off of the North Atlantic Ocean.

Figure 11. Zonal mean zonal wind (shading) and correlation (×100) of ENSO index (DJF Niño 3.4) with zonal mean zonal wind (contours) and zonal mean zonal wind (shading) over the North Pacific sector for winters 1979/80-2014/15. First, second and third contours represent 90, 95 and 99% statistical significance respectively.

Together these four features influenced the hemispheric temperature pattern.  The Ural blocking combined with the East Asia trough resulted in a cold winter for East Asia.  The strengthened Aleutian Low cooled central North Pacific SSTs but warmer Alaska and much of the east coast of North America.  Positive geopotential heights/ridging along the east coast of North America warmed the Eastern US and Canada.  A strengthened Icelandic Low and the Ural Blocking resulted in warm temperatures for Europe and Western Asia (see Figure 10b).

Winter forecasts

Before the winter I published the AER winter forecast on the National Science Foundation website, which is still there and here I have included the NH version of the forecast (Figure 10a).   The main predictors in the AER forecast were October Eurasian snow cover, Arctic sea ice and El Niño.  Snow cover was high and sea ice low so therefore the model forecast were consistent with a weak polar vortex and a negative AO.  The model predicted cold temperatures across Northern Asia, East Asia, Western Europe and most of the eastern one third of the US, though the El Niño contributed to the cold temperatures in the Southeastern US as well.  Much of the remainder of the NH was predicted to be mild.  The observed winter temperatures are also shown in Figure 10a.  The temperatures were detrended before computing anomalies to show more of the interannual (and hopefully related to the seasonal atmospheric dynamics) rather than longer-term variability but the warm term warming has been muted in winter and the figure does not differ much from the raw anomalies (I tweeted out the raw temperature anomalies).  The biggest model error were the observed mild temperatures in northwest Asia.  In North America the largest error were the observed warm temperatures throughout the Eastern US.

Figure 12. a) Mean zonal wind (ms-1; contours) and zonal wind anomalies (ms-1; shading) across the Northern Hemisphere for December 1997 through February 1998 b) same as a) except forecasted for December 2015 through February 2016.

Both the extensive Eurasian snow cover and low Arctic sea ice favor a strengthened Siberian high.  But typically the center of the high is over the Arctic Ocean, which results in northerly flow across East Asia and easterly flow across Europe and cold temperatures stretching across much of northern Eurasia.  However this winter the center of the high was shifted southeast over the Asian continent.  This still created northerly flow and therefore cold temperatures across East Asia but western Asia and Europe were more in a mild southerly and westerly flow.

The too cold temperatures predicted in the Eastern US were a combination of the impacts of an El Niño and the expectation of a weak PV and negative AO.  For must of the winter the PV was strong, which kept the polar Jet Stream north of the US and therefore on the mild side.  But also as I show, the subtropical Jet Stream across the Southern US is one the dynamical signatures of El Niño winters but was absent from the Eastern US this past winter.  The strong Jet across the Southern US can result in cooler weather on the northern side but also prevents the very warm subtropical air flowing north of the jet.  With no jet across the Southern US, very warm air flowed unhindered unusually far north time and time again.

Figure 13. a) NMME predicted precipitation rate for North America from December 2015 through February 2016 b) observed average precipitation anomalies (inches; contours) across the United States from December 2015 through February 2016.

The next figure (Figure 11) shows the climatological jet structure in the North Pacific in shading and variability associated with ENSO in contours (this figure is from a manuscript in review).  El Niño strengthens the jet on the equatorward side and weakens the jet on the poleward side and the reverse is true for La Niña.  In Figure 12, I plot the zonal wind at 250 hpa for the winters of 1997/98 and 2015/16.  In terms of the jet, the winter of 1997/98 looked like a canonical El Niño winter with the Jet Stream strengthened on the equatorward side and weakened on the poleward side.  However in winter 2015/16 the reverse is true, the Jet Stream is strengthened on the poleward side and weakened on the equatorward side or in other words is the same atmospheric response as La Niña.  Also similar to La Niña, a strong jet across the Southern US is absent.

Therefore not surprisingly comparing observed temperatures and precipitation to the dynamical forecasts shows that the observations are reversed from the forecasts.  The national multi-model ensemble (NMME) forecast for NH temperatures is included in Figure 10c.  The models predict the canonical El Niño northwest-southeast dipole across the US with the warmest temperatures relative to normal across the Northwestern US and the coolest in the Southeastern US but the observations show the reverse.  Similarly with the precipitation (Figure 13), the models predict the canonical El Niño north-south dipole across the US with the driest conditions relative to normal across the Northern US and the wettest in the Southern US and the observations show the reverse.  The exception being the East Coast but was still wetter to the north than the south and flipped from the forecast.

Figure 14. a) CFS predicted average 700 hPa geopotential heights (dam; contours) and geopotential height anomalies (m; shading) across the Northern Hemisphere from December 1, 2015 through February 28, 2016 b) observed average 700 hPa geopotential heights (dam; contours) and geopotential height anomalies (m; shading) across the Northern Hemisphere from December 1, 2015 through February 28, 2016.

Finally I show the predicted 700 hpa geoptential heights from the CFS model and the observed 700 hpa geoptential heights (Figure 14).  At first glance the model did a good job at predicting the hemispheric pattern.  However the greatest hemispheric anomaly – the positive anomaly center over Northwest Asia is missing in the model forecast.  This resulted in the models incorrectly predicting a warm winter for East Asia.  But this feature is in response to extensive Eurasian snow cover and/or low Barents-Kara sea ice.  That the models completely missed this feature indicates to me that the models are fatally flawed at modeling high latitude surface-atmosphere coupling and troposphere stratosphere coupling.

Concluding remarks

Before the winter I stated that in contrast to other forecasters the strong El Niño lowered my confidence in the winter forecast and did not increase it. That the atmospheric response more closely resembled those associated with La Niña alone I believe has justified that feeling.  But my sentiment truly reflected that if the polar vortex did not weaken sufficiently the NH would be flooded with El Niño fueled warm air.  That turned out to be compounded by the fact that the strong jet along the Southern US, typical of El Niño, never really materialized this winter.  And if next winter La Niña returns that leaves the Eastern US vulnerable to more record warmth if the PV once again stays strong.

This remains an open debate, but the “cold blob” in the North Atlantic has been associated with a strengthened north-south temperature gradient a strengthened Jet Stream a stubbornly positive NAO and a mild Europe.  If nothing changes by next winter I feel it is hard to forecast other than more of the same.

The strong Ural Blocking, the record weak PV and impressive PV split further supports that the Arctic is influencing weather outside of the Arctic.  But in forecasting it is certainly true that “the devil is in the details” and the fact that PV weakening did not climax until mid-March hurt the forecast based on snow cover and sea ice.  But I think the biggest story were the high expectations for accurate forecasts based on a record strong El Niño that never materialized.  One could argue that the true victor this winter is natural variability.  However I hope that it is obvious that I strongly disagree and through the blog I am trying to demonstrate to others but most importantly to myself that there is order in the seemingly chaos, we just need to see the whole picture.