An underlying predictability in winter weather patterns in the North Atlantic Basin


There now exists a number of computer models that show some success in predicting the meteorological conditions in the ensuing winter in the northern hemisphere. These examine the Arctic Oscillation (AO) and are successful up to two months in advance of the beginning of meteorological winter. Other models, which are based upon the North Atlantic Oscillation (NAO), also show some predictive success depending on the period over which they are tested. Here I show that the pattern of winter circulation across the North Atlantic Basin, Arctic Basin and Western Europe can be elucidated from the patterns present as early as the preceding summer and early autumn. These are visible in 500 mb charts that are publicly available. This empirical model has particular success in years where there is neither an El Niño or La Niña event. However, in some recent El Niño or La Niña years, the pattern remains rather predictable making subsequent long term forecasting possible.

Given the impact of severe weather on human activity, there is an obvious interest in producing accurate forecasts weeks or months in advance. However, the success rate is extremely variable (1). In recent years a wealth of observational data has allowed the identification of correlations between autumnal meteorological patterns and weather in the subsequent season. For example, Vladimir Kryjov identified a connection between the surface pressure over the Taymyr Peninsula, in north eastern Russia in October, with the strength of the so-called Arctic Oscillation (AO) in the subsequent winter (2). Moreover, in the last few years a number of superior computer models have been produced that are based on seasonal patterns in the Arctic, related North Atlantic Oscillations (NAO) or more distantly the tropical Madden Julian Oscillation (3-5). For example, the GloSea5 computer model is able to predict the winter season weather in Western Europe with a good degree of accuracy, based upon the NAO in the preceding season (4). These are successfully predictive when made from late October or early November. Meanwhile, Kang et al. show that the, predecessor, GloSea4 model could predict with some accuracy the state of the AO as early as November (3). The accuracy of these forecast models has been assessed by Shi et al. (6) and found to vary depending on the decadal period over which they are tested.

The Arctic Oscillation is a variation in the pressure gradient from the Arctic Basin and the surrounding mid-latitudes, south to 20°N (7,8). When pressure is higher in the Arctic and lower further towards the equator, the Oscillation is said to be negative. Conversely, when pressure is lower over the Arctic, pressure is higher over the Horse Latitudes and the index is said to be positive. Negative AO is associated with weak westerly winds across the mid-latitudes and a stronger prevalence of cold easterly or northerly outbreaks across Western Europe during the winter months. These are typically associated with the formation of warm-cored areas of high pressure, known colloquially as blocking anticyclones. Meanwhile, a positive AO is associated with warmer and stronger westerlies across Europe and limited, if any, blocking of the westerly flow across Europe. Kryjov found that there was a correlation between higher pressure over the region in October and a more negative Arctic Oscillation in the subsequent winter (2).

The North Atlantic Oscillation is an extension of the Arctic Oscillation. During positive phases there is a steep gradient of pressure between Iceland and the Azores, which leads to strong westerly winds. Negative phases are associated with a weakening of these winds, a frequent split in the polar-front jet stream and blocking over Scandinavia or the Norwegian Sea. There is a further correlation between the extent and arrival date of snow cover in Siberia and the severity of the winter in Western Europe (9) and this mirrors an underlying change in the Arctic Oscillation with a more negative AO giving rise to earlier snow cover. Cohen presumes that the snow cover influences the pattern of airflow from Siberia to the west with more snow cover strengthening cooling and thus an expansion of the underlying cold Siberian high pressure area (9). However, what if this assumption is wrong and the extent of snow cover is instead a reflection of the overlying circulation pattern, which in turn is diagnostic of a pressure pattern that has already been established? Instead of driving the observed change, which is generally the assumption, the early winter snow cover is an indicator of the pattern of circulation in the overlying atmosphere during the autumn. This does not negate its effect, but implies that the snow cover is a reflection of the pre-existing pressure pattern.

In essence this is the hypothesis presented here: by adopting a purely empirical approach, it is observed that the pressure pattern in the summer and early autumn relates to the pressure pattern in the subsequent winter. While this pattern is not an exclusive influence over the pressure in the ensuing season, the mean pressure pattern during the summer and early autumn months does appear to possess a considerable influence on the pressure pattern over the Arctic Basin and Western Europe several months later, during the winter. Subsequently, this affects the severity of the winter weather in Western Europe.

Deriving the model
Meteorologists examine the atmosphere in slices. Whereas at the surface pressure varies with the passage of various features, such as frontal storms, meteorologists look at pressure levels and the heights at which these occur. Warmer air raises the pressure level, while colder air has the corresponding pressure at a lower altitude. The mean surface level pressure is 1,000 millibars (1,000 mb). Half this value (500 mb) is found at altitudes of around 5,000 meters, while the polar front jet stream, at the top of the troposphere is found at 200-300 mb, depending on the season. For the purposes of this paper we will look at the 500 mb level. The reasoning for this choice is explained below.

Driven by curiosity, I developed a series of predictive models based purely on observation: i.e. an empirical approach. Initially, these looked at surface level pressure patterns over the North Atlantic through to Western Russia. However, it soon became clear that the observed patterns in the autumn did not provide a very good match to the pattern in the subsequent season. In fact, there were hints that the presence of autumnal blocking over Western and Northern Central Russia might link to conditions in Western Europe later in the year (10,11).

Motivated by this weak correlation, I then examined the 500 mb charts that were commercially available during the late 2000s (12,13). The 500 mb pattern across the North Atlantic gave a much clearer correlation and this agrees with the recent computer models of Scaiff et al. (4). However, it was not until the 500 mb (5,000 metres) level was examined for the whole northern hemisphere that a much stronger correlation was noted and was, therefore, further examined. At the 500 mb level various atmospheric waves, known as Rossby waves, are clearly apparent. It is well established that the pattern of Rossby waves relates to the conditions at the surface as they reflect the path of the polar front jet stream above and, therefore, the passage of frontal systems below. When the Arctic Oscillation is positive (negative) these waves have a low (high) amplitude, thus making them a useful discriminator between different phases of the AO (7,8). The broad pattern of these waves reflects the mobility of the circumpolar circulation and thus forms the basis of this proposition. In essence this is an empirical analysis of the Arctic Oscillation, North Atlantic Oscillations and related oscillations in the North Pacific.

The observed 500 mb pressure pattern over the Arctic Basin and Northern Atlantic Basin in the winter correlates strongly with the observed pressure pattern over and surrounding the Arctic Basin at 500 mb in the preceding two seasons. This leads to a strong predictability of the overall manner in which weather will unfold during that winter. The pressure pattern is modified by the phase and strength of the El Niño Southern Oscillation (ENSO), but is not completely abrogated by it. While other computer models extend predictions for December-February to the beginning of November (3,4), I suggest here that examination of the Arctic Oscillation – and the overriding state of the circumpolar circulation – makes it possible to successfully predict the state of the wintertime AO as far back as the preceding summer (Figure 1).



Figure 1 | 500 mb charts from July 16th 2009 (left) and July 16th 2010 (right). Warm-cored high pressures areas (H) are prevalent across the Arctic Basin. While, this is a snap-shot from the middle of the meteorological summer, it is a consistent pattern. Note, that the geographical region identified by Kryjov and centred on the Taymyr Peninsula (2) is under one such blocking high in both years (arrows). Charts from Colorado State University.


Observations and Supporting arguments
The pressure over the Arctic Basin varies from year to year and is coupled strongly to pressure over the North Atlantic in all seasons. During the summers of 2009 and 2010 surface level pressure and pressure at 500 mb was predominantly high over the Arctic Basin, with a generally weak circulation within the North Polar Basin (Figure 1). Warm-cored blocking anticyclones were evident across much of the basin, particularly in July. In both years the pressure pattern was largely maintained throughout the autumn and into December with pressure remaining generally higher than normal across the basin indicating a negative AO. Each of these seasons was followed by a colder than normal winter, but with some notable differences that will be discussed later.

Conversely, during the summer months of 2013 the 500 mb level was depressed with a stronger, colder circulation and smaller (lower amplitude) Rossby waves surrounding the circumpolar region (Figure 2). This is indicative of a positive AO. This pattern continued to strengthen during the autumn leading to Britain’s wettest and warmest winter on record (at the time of writing). During the late summer and autumn of 2014 the circulation was mixed with a predominantly cold circulation, but with high amplitude Rossby waves evident circulating around the North Pole. Interestingly, while the AO is negative during the summer months, it is evident that the circulation at 500 mb above the basin is as strong as it was in the summer of 2013. Therefore, while there were significant cold circulations across Asia and the northern Pacific, this was offset by the relatively strong circulation around the pole. Thus overall, the pressure pattern during the autumn of 2014 was transitional between the two patterns seen in the autumn of 2009 and 2013.



Figure 2 | The growth of the cold circulation over the Arctic during the late summer and autumn of 2013. While many of the preceding summers and early autumns had evidence of warm-cored blocking anticyclones, the summer and early autumn showed a stronger colder circulation (blue and purple in these charts from This circulation grew in strength with only minimal incursions of warmer air leading to my prediction that the winter would be mild and wet. Charts are for 17th August left); 6th September (middle) and 17th October (right). Charts courtesy of 


At this point it seemed apparent that there was a straightforward correlation between the AO and overall circulation pattern surrounding the Arctic Basin during the summer season and the weather in the subsequent winter across Western Europe and the North Atlantic Basin. This idea was then further tested by retrospective analysis of the pressure patterns in other non El Niño or La Niña years. Data available from Colorado State University (14) allowed the same hypothesis to be tested in various years from the year 2000. Further data was available, extending back to 1899 (12,13,14). However, this was restricted to analysis of the North Atlantic Basin and thus of more limited use. Commercially available numerical data was also examined from National Oceanic and Atmospheric Administration (NOAA) (15).

There was a concern that the observed correlation was either a statistical fluke or one that was related to a particular but non-obvious pattern in the AO that was specific to the last seven years. To test this hypothesis, the summer-winter AO patterns were examined in the years 1995, 1998 and 2004: the same correlation held.

Within this broad pattern you can discern, in most cases, a prevalence for warm cored blocking anticyclones over the Taymyr Peninsula in agreement with the observations of Kryjov (2). However, more generally pressure remains higher than during episodes of positive AO across the central Arctic Basin, as well. Conversely, during July and August, when the AO is positive, there is a preference for the axis of the cold-cored circulation to lie over northern Canada and Greenland. In this situation warm-cored blocking over northern Siberia in the summer and autumn is far more likely to favour the development of a strong surface high pressure area than an early snow cover. This is by virtue of the blocks ability to divert the jet stream north on the western flank and south with cold, moisture – laden winds along its eastern flank. From here positive feedback can further intensify subsidence at the surface and an increase in the intensity of the otherwise cold-cored Siberian high.

Predictability of the winter weather for Western Europe depends on the phase of the Arctic Oscillation. However, this does not readily extend to North America, where cold winters depend principally on the extent and orientation of the polar vortex. This in turn, depends on a number of other factors, such as the phases of the El Niño Southern Oscillation or the (North Atlantic Oscillation). Thus, although the polar vortex may extend across much of the continental landmass during periods of negative AO as was seen during the winter of 2014-2015, it may also be present when the Arctic Basin is experiencing a positive AO as in the winter of 2013-2014. During the latter winter the principle driver of this cold incursion was a strongly modified jet stream with a large amplitude crest (Rossby wave) over the Pacific and corresponding trough over North America at the 500 mb level. During this phase, Western Europe remained predominantly very mild and wet, with a strong circumpolar circulation driving repeated storms across the region. The location of the jet stream was south of its averaged winter time position but a positive AO ensured that the polar front jet was active across Europe, bringing repeated rain storms and record-breaking flooding to the UK. During the subsequent winter of 2014-2015 Western Europe experienced temperatures close to the seasonal average with a good mixture of cold and mild conditions. North America again suffered significant snowfall events. Thus, there was no direct correlation between the weather over North America and that in Western Europe. This lack of correlation is exemplified by further analysis going back to 1950s (9). Looking at data for New York on the Atlantic seaboard, the majority of the coldest winters from 1950 to present day are associated with a positive AO, rather than a negative one. This is readily understood because a positive AO will mean lower amplitude and longer wavelength Rossby waves. With a predominantly westerly wind direction, cold air is directed from the continental interior towards the eastern seaboard. However, a negative AO, with large amplitude waves, will tend to display a trough over the continental interior – an effect driven both by significant land cooling and by the Rocky Mountains. This, in turn directs warm air up, along the east coast of the continent, thus favouring a warmer winter.

The impact of the phase of the ENSO is more problematic, in part because strong El Niños are relatively rare and thus comparing the effect of the ENSO phase with that of the AO suffers from limited data. However, some key themes emerge. Analysis was done of the winters of 1982-1993, 1997-1998, 2009-2010 and 2010-2011, with a comparison of the winter pattern with that of the preceding two seasons. Strong El Niño events began in 1982, 1997 and 2015; while 2009 saw a moderate central Pacific El Niño; and 2010 a strong La Niña.

Typically, El Nino episodes are thought to add additional energy to the polar circulation. They also establish poleward-propagating Rossby waves that strengthen westerlies and bring mild conditions to the North Atlantic Basin, particularly during the first half of any given winter. However, in the winter of 2009, the El Niño failed to override the effect of the already negative Arctic Oscillation and blocking prevailed across Europe. This was despite it driving more typical mild El Niño conditions in the Pacific North-West.

During the ensuing summer of 2010, blocking persisted over the Arctic with a very weak circum-polar circulation during the summer and early autumn (Figure 1). Therefore, I made a prediction of a colder than normal winter, with significant blocking. By late November substantial blocking across northern Europe and the North Atlantic led to the coldest December on record.

Interestingly, there was an additional effect of the La Niña. Although La Niñas typically bring mild winters to Western Europe and the North Atlantic Basin there is an additional effect on pressure patterns with the formation of blocking anticyclones over central and Western Europe in late November and early December. These bring cold and dry conditions. This pattern tends to give way to predominantly westerly conditions from mid-winter onwards. In 2010 the negative AO appears to have conspired with this La Niña preference for early season blocking, producing instead enhanced blocking, but at somewhat higher latitude than normal. This led to the UK’s coldest December since records began. Subsequently, typical late winter La Niña conditions prevailed with mild conditions. Contrast this with the winters of 2007-2008 and 1998-1999 where a positive AO and La Niña conditions conspired to produce mild winters across Western Europe.

The strong El Niño winters of 1982-1983, 1997-1998 all produce predominantly mild conditions. Here, the effect of the El Niño appears to supersede any effect of the AO. However, in the summer of 2015 there was a negative AO. Despite the strong influence of the El Niño, there was still a propensity for blocking over the Arctic and Scandinavia in late December 2015 and early January 2016. Thus, although the El Niño had the strongest meteorological impact, there was still evidence of the impact of the negative AO from the preceding seasons.

I present a simple and testable hypothesis: that the winter weather in Western and Central Europe is determined in large part by the pressure pattern over or around the Arctic in the preceding two seasons. The pattern is modified by strong El Niño and strong La Niña events but is otherwise discernible in every winter tested. While this pattern has been identified in the late autumn as being predictive, these empirical observations take the period from which prediction can be made back two seasons with good reproducibility in the last two decades. Predicting general weather patterns several months in advance has two benefits. Immediate benefits include better preparedness for extreme weather events which are associated with particular weather patterns. However, the ability to predict patterns in advance implies that there are much stronger correlations in weather patterns between seasons. This will likely improve our understanding of feedback mechanisms in the climatic system of our planet as a whole.

Conflicts of interest
The author declares no conflicts of interest.

About the Author
David Stevenson is a former post-doctoral academic with a background in molecular genetics. Subsequent publications include authoring a number of astronomy books for Springer, including “The Exoweather Report”, which is currently in press. Other published titles for Springer covered supernovae, the habitability of planets orbiting red dwarfs and finally star clusters. A number of other popular astronomy articles are available in Popular Astronomy, Astronomy and Sky & Telescope magazines. I currently work at a successful Academy in Nottinghamshire, while writing two further books on planetary geology and astrobiology for Springer.


  1. Corbyn P. Forecast model for July 2014. [Internet] Available from: docs/BI1407JUL30dSlat9cWowContrastsProd27JunTypCor.pdf
  2. Kryjov VN. October circulation precursors of the wintertime Arctic Oscillation. Int. J. Climatol. 2013; 35(2): 161-171.
  3. Kang D, Lee MI, Im J, Kim D, Kim HM, Kang HS, Schubert SD, Arribas A, and MacLachlan C. Prediction of the Arctic Oscillation in boreal winter by dynamical seasonal forecasting systems. Geophys. Res. Lett. 2014; 41:3577–3585,
  4. Scaife AA, Arribas A, Blockley E, Brookshaw A, Clark RT, Dunstone N, Eade R, Fereday D, Folland CK, Gordon M, Hermanson L, Knight JR, Lea DJ, MacLachlan C, Maidens A, Martin M, Peterson AK, Smith D, Vellinga M, Wallace E, Waters J and Williams A. Skilful long range prediction of European and North American winters. Geophys. Res. Lett. 2014; 41:2514–2519.
  5. Saha S, Moorthi S, Wu X, Wang J, Nadiga S, Tripp P, Behringer D, Hou YT, Chuang H, Iredell M, Ek M, Meng J, Yang R, Mendez MP, Dool H VD, Zhang Q, Wang W, Chen M and Beckers E. The NCEP Climate Forecast System Version 2. J. Climate. 2014; 27(6):2185–2208.
  6. Shi W, Schaller N, MacLeod D, Palmer TN and Weisheimer A. Impact of hindcast length on estimates of seasonal climate predictability. Geophys. Res. Lett. 2016; 42(5):1554-1559.
  7. Lorenz EN. Seasonal and irregular variations of the northern hemisphere sea-level pressure profile. J. Meteor. 1951; 8:52–59.
  8. Thompson DWJ, Wallace JM. The Arctic Oscillation signature in the wintertime geopotential height and temperature field. Geophys. Res. Lett. 1998; 25(9):1297-1300.
  9. Cohen J. Eurasian Snow Cover Variability and Links with Stratosphere-Troposphere Coupling and Their Potential Use in Seasonal to Decadal Climate Predictions US National Oceanic and Atmospheric Administration Climate Test Bed Joint Seminar Series NCEP. 2011. Available from:
  10. Tang Y, Lin H, Derome J, Tippett M. A Predictability Measure Applied to Seasonal Predictions of the Arctic Oscillation. J. Climate. 2007; 20(18):4733-4750.
  11. Ambaum MHP, Hoskins BJ, and Stephenson DB. Arctic Oscillation or North Atlantic Oscillation? J. Climate. 2001; 14:3495-3507.
  12. NCEP Reanalysis of NOAA archive charts. [Internet]. [1871 until present]. Available from:
  13. [Internet]. Available from:;sess=
  14. NWS DIFAX weather map archive. [Internet]. Colorado State University. [2000-present]. Available from:
  15. NOAA Arctic Oscillation Index Data. [Internet]. National Oceanic and Atmospheric Administration [1950-present]. Available from:

Share your thoughts

Leave a Reply

You must be logged in to post a comment.