Arctic Oscillation and Polar Vortex Analysis and Forecasts

April 9, 2018

Special blog on winter 2016/2017 retrospective can be found here -

Special blog on winter 2015/2016 retrospective can be found here -

Dr. Judah Cohen from Atmospheric and Environmental Research (AER) recently embarked on an experimental process of regular research, review, and analysis of the Arctic Oscillation (AO). This analysis is intended to provide researchers and practitioners real-time insights on one of North America’s and Europe’s leading drivers for extreme and persistent temperature patterns.

With the start of fall I will be transitioning to a fall/winter schedule, which is once every week starting sometime in October. Precipitation forecasts will be replaced by snow accumulation forecasts. Also there will be more emphasis on ice and snow boundary conditions (which are both now in their seasonal advance) and their influence on hemispheric weather.

Subscribe to our email list or follow me on Twitter (@judah47) for notification of updates.


  • The Arctic Oscillation (AO) is currently neutral and is predicted to trend slightly negative before trending positive next week and will be mostly negative for much of mid-April.    
  • The current neutral AO is reflective of mixed pressure/geopotential height anomalies across the Arctic and mixed pressure/geopotential height anomalies across the mid-latitudes. The North Atlantic Oscillation (NAO) is currently positive with mixed pressure/geopotential height anomalies across Greenland but mostly positive pressure/geopotential height anomalies across the mid-latitudes of the North Atlantic.  The forecasts are for the NAO to  trend negative for the next two weeks as geopotential height anomalies rise across Greenland and Iceland.
  • The polar vortex (PV) is predicted to linger and finally disappear for the season across Western Siberia  over the next two weeks.  This is reflected in persistent troughing/negative geopotential height anomalies across western Siberia with ridging/positive geopotential height anomalies on either side across Europe and Eastern Siberia.  This will allow result in relatively cold temperatures in the interior of Eurasia with relatively milder temperatures on the edges including much of Europe (including the United Kingdom (UK)) and East Asia over the next two weeks.
  • Currently troughing/negative geopotential height anomalies south of the Aleutians is contributing to downstream ridging/positive geopotential height anomalies across Western North America with additional troughing/negative geopotential height anomalies in eastern North America including the Eastern United States (US). This pattern in general favors normal to above normal temperatures for western North America and normal to below normal temperatures in the eastern North America.
  • Later in the period strengthening ridging/positive geopotential height anomalies across Greenland will help reinforce troughing/negative geopotential height anomalies over the Eastern United States (US).  Therefore little change is expected in the overall  pattern that favors normal to above normal temperatures across western North America with normal to below normal temperatures for much of Southeastern Canada and the Eastern US.


Though the significant PV disruption occurred exactly two months ago it does seem that its influence on Northern Hemisphere (NH )weather will continue over the next two weeks.  Typically warm polar cap heights/negative AO drip down from the stratosphere to the troposphere following a PV disruption or can be described as tentacles extending down from the PV disruption event for up to two months.  In my opinion the influence from the PV disruption will extend for longer than two months thanks to another drip or tentacle predicted in mid- to possibly late-April.  A sun comparable in strength to Labor Day weekend and late August will make sub-freezing daytime temperatures and snowfall increasingly difficult (but based on latest model forecasts still possible) across the Northeastern US but temperatures look to average below normal for the foreseeable future as high latitude blocking is predicted to continue.

I still plan on doing a winter recap or synopsis as soon as I have a chance but before winter gets too far in the rear view mirror, I did want to provide an opinion about what constitutes success in a winter prediction that I feel has important implications heading into next fall and how to approach the next winter forecast. Does one simply rate or score a winter forecast by the end point or should one also include all the variability that takes place from the beginning until the end of the forecast.

Back when I presented verification for winter (Dec-Feb) 2017/18, I wrote that the one cardinal sin in seasonal prediction is probably to declare “problem solved.”  Here I wanted to discuss a related topic – what constitutes a successful winter forecast.  NOAA’s CPC is certainly not claiming “problem solved” but they are feeling good about their forecast and have labelled it “pretty good” (see their blog - ).  I appreciate how difficult it is to issue an accurate winter forecast and by no means want to contradict or detract from their declaration of success.   The discussion that follows is not a criticism and is not about a right and wrong forecast but rather is a discussion of philosophy.  And I do believe that the philosophy of what constitutes a successful forecast will have tangible implications for future winter forecasts.

My AER colleague recently quoted me the British Economist Ronald Coase (Nobel Prize 1991) “If you torture the data enough, nature will always confess.”  Observations are not absolute but contain inherent errors and are displayed with some subjective decisions on how to compute.  NOAA’s CPC is claiming an above average accuracy for their winter forecast.  And based on their own observational plots their forecast looks very good.  My own observational plots showed a degraded forecast relative to their own and I heard from some NOAA scientists that my plots are in error.  I use the NCAP/NCAR reanalysis to both plot “observations” and the forecast and I apply the same smoothing to both.  This decision was probably made over fifteen years ago and was intended to mask artificial bullseyes in the data and not to display the AER forecast in a better light.  In much of the East, temperature anomalies were close to normal and the sign of anomalies are dataset dependent and dependent on some subjective decisions on how to compute.  But regardless, I don’t believe that my arguments are dependent on the subjectivities of the observations.

In Figure i, I include the seasonal forecasts for Dec-Feb, Jan-Mar and individual monthly forecasts from Dec-Mar from the CFSv2.  Much of the CPC official forecasts was derived from the CFS.  The CPC forecast is strongly dependent on the phase of ENSO, which was in La Niña all winter (see Figure ii).  CPC is claiming an overall high scoring forecast and based on their set of observations that is true, something that I am not disputing.  They then attribute the accurate forecast to La Niña.  However composites and averages can cover many blemishes and I question how much to attribute the observed winter anomalies to La Niña.  La Niña was remarkably stable for the entire winter (see Figure ii) and I would argue this is reflected in the monthly forecasts that showed little variability across both North America and Eurasia.  In Figure iii, I show the observed monthly temperature anomalies, which exhibit large month-to-month variability that simply cannot be explained by ENSO that was constant.  One could argue that the constant forcing of ENSO is the signal and the large temperature swings are noise and all we need to focus on is the signal.  Don’t focus on the poor monthly forecasts which are less skillful but more on the seasonal forecast which are more skillful. And this is CPC’s perspective as the blog author concludes in the blog: “This leaves forecasters in a tricky spot in searching for climate signals, which is one reason they often turn to ‘old reliable’ ENSO, and why reality always looks different and noisier than the seasonal predictions.”

Figure i. Surface temperature anomaly forecast for a) December-February 2017/2018, b) January-March 2018, c) December 2017, d) January 2018, e) February 2018 and f) March 2018 from the CFSv2 initialized on November 1, 2017.

I have a different perspective, those large temperature swings should not be simply dismissed as noise but rather as signal related to behavior in the polar vortex and in my winter recap I will explain the wild temperature vacillations to polar vortex related variability.  Therefore the poor monthly forecasts aren’t simply an artifact of lack of time filtering but a lack of the models ability to capture or simulate the main driver of the winter temperature anomalies.  Also the correct winter forecast cannot be separated from the missed forecast elsewhere across the NH including much of Eurasia.  I do consider ENSO as part of the signal across North America but it is not the dominant or at least the only signal.