Why is the Arctic climate and ice cover so variable?

by Judith Curry
A discussion of Section 8.3 of Alan Longhurst’s book Doubt and Certainty in Climate Science.

For context on the topic of the variability of Arctic climate, see these previous blog posts:

• New presentations on sea ice
• Early 20th century Arctic warming
• Uncertainty in Arctic temperatures
• Unprecedented(?) Arctic warming
• Unprecedented(?) Arctic warming II
• Historic variations in Arctic sea ice
• Likely causes in the recent changes in Arctic sea ice

For background to this section in Longhurst’s book, it is useful to also read sections 8.1 and 8.2. Some excerpts from Section 8.3:
8.3  Why is the Arctic climate and ice cover so strongly variable?
Observations  suggest that variability in oceanographic conditions in the Arctic is very largely driven by the consequences of the flows through open passages to both Atlantic and Pacific Oceans, which themselves respond to the different and characteristic variability of the circulation patterns of each ocean: each inflow is not only variable in volume of water transported but also in the temperature of the water imported. 
JC note:  here is a map for geographic reference

 
Of the 8.5 Sv of warm, salty Atlantic water that passes north across the Greenland‐Scotland Ridge annually, about 4.0 ±2.5 Sv passes into the Barents Sea either directly to the north of Norway as a barotropic flow, or along the western coast of Spitzbergen as a baroclinic flow. These fluxes of warm water (6-‐8C) carry almost 100 TW of Atlantic heat into the eastern Arctic Ocean annually, while another 10­‐20 TW passes into the western Arctic basin through the Bering Strait in a flow of about 0.8 ±0.2 Sv of Pacific Ocean water.  The high‐salinity water Atlantic water fills the Arctic basin between the low­‐salinity surface water and the Arctic bottom water while the small flow of Pacific shelf water passes to the east along the Alaskan­‐Canadian shelf.
Although the Arctic Ocean comprises only 3.7% of the surface of the global ocean, it receives an input of freshwater from Asia and North America that is equivalent to 11% of the flow of all rivers, whose flow across the northern continents and into the arctic seas is at least as variable as it is in other regions. This fresh water, together with melt‐water from the melting ice­‐pack in summer forms a permanent superficial layer (usually about 200m deep) of low salinity over the entire Arctic Ocean, without which much less seasonal ice would form. The flow of freshwater from the northern continents represents an export to the world ocean that goes almost entirely into the Atlantic, about 5.1 Sv passing as relatively low salinity water through the passages between Greenland and Ellesmere Island into the Labrador Sea, a flow of low salinity water that can subsequently be traced around the subpolar gyre. Balance is also maintained by flow from the Arctic Ocean through the western part of Fram Strait to enter the East Greenland Current. The strength of both of these annual fluxes during summer will have consequences for the salinity of the surface water mass of the Arctic Ocean and hence on the strength of the freezing cycle during the following winter.
JC note:  The freezing temperature of water depends on the salinity of the water — more salt means a lower freezing temperature (the same effect of salting your driveway in winter).   The freezing/melting point of pure water is 0C; the freezing point of seawater with salinity 35 psu (~ parts per thousand) is -1.92C.  Because the the Arctic Ocean is fresher than average seawater, the average salinity is closer to 28 psu.
Because the incoming and the outgoing flows, warm and cold respectively, lie side‐by‐side between Greenland and Scandinavia, an asymmetry is induced in the distribution of ice-­‐cover on the Arctic Ocean; this is generally dense to the west of Fram Strait while, to the east of Spitzbergen, much of the Barents Sea – at similar latitudes – remains ice‐free even in winter due the eastward flow of warm Atlantic water.
The outgoing flow through Fram Strait carries with it large volumes of fresh water as fragmented pack ice, a flow that is strongly episodic at decadal scale and is associated with the series of so­‐called Great Salinity Anomalies observed within the circulation of the subarctic gyre and in the Nordic seas that were discussed in the previous chapter.
The significance of these events continues to be revealed: a new synthesis of circulation in the Arctic basin has been made from almost 3000 oceanographic profiles obtained in the central Arctic Ocean since the 1890s, which were not previously accessible. This makes it clear to what extent the variability in the inflow of ‘warm and salty’ North Atlantic water at times of positive values of the NAO (North Atlantic Oscillation) dominates the temperature of the Atlantic water mass by importing ‘vast quantities of heat’ into the Arctic Ocean to induce core temperatures in the intermediate layer in Nansen Basin that are much warmer than in the Canadian Basin, far downstream. This warm intermediate layer has the potential for significant control of the annual cycle of formation and melting of arctic ice.
On the other hand, during the negative phase of the AO (Arctic Oscillation), water motion in the Arctic Ocean is anticyclonic and the Beaufort gyre is strengthened, so that ice is retained and thickened both in the Canada Basin and along the Siberian coastline, where it may survive summer melting. Similarly, records of fast ice thickness and extent in four Arctic marginal seas (Kara, Laptev, East Siberian, and Chukchi) indicate that long­‐term trends are small and generally statistically insignificant, although correlation degrades eastwards and is absent in the Chukchi Sea. That a simple warming trend throughout the 20th century does not characterise arctic conditions is also confirmed by records of ice­‐cover in the four seas that lie north of Siberia (Kara, Laptev, East Siberian and Chukchi); these show clearly that ice variability in these seas is dominated by a low­‐frequency oscillation of frequency 60‐80 years that “places a strong limitation on our ability to resolve long­‐term trends”. This low frequency signal is strongest in the Kara Sea (where very strong ice minima occurred in 1940 and at 2000 the end of the data series studied) and decays eastward so that in the Chukchi Sea ice cover is dominated by decadal fluctuations. Only in the Kara Sea is ice cover dominated by thermodynamic factors, while ice cover in the other basins is dominated by the effects of wind and currents.
Transport of warm water on this scale may be expected to be directly related to the pattern of low and high pressure cells in the atmosphere. A stubborn, positive state of the NAO characterised the final decades of the 20th century, and was associated with transport of Atlantic water into the Arctic Basin that significantly reduced ice coverage.
Since 2002, this process has accelerated due to very thin spring ice and to the “memory of the system to the positive winter AO state that characterised the mid-­‐1980s and 1990s” as Stroeve et al. put it. As well, these authors note that the character of sea ice has also progressively changed after so long a period of positive NAO values, particularly in the progressive loss of multi­‐year ice. The single, strongly-‐negative NAO index during the winter of 2009/2010 was not sufficient to reverse the process.
The first evidence that a warm pulse had entered the Arctic Ocean in 1990 was the occurrence of anomalies of order 1C in the Atlantic water mass of the Nansen Basin. These were transported in the anticyclonic gyral circulation along the Asian continental slope through the Makarov Basin to reach the Canadian Basin 7 or 8 years later as a warm anomaly of about 0.5C. A second set of warm pulses was detected at Fram Strait in 2004 were a little warmer, but followed the same trajectory as in 1990 so that peak warming in the Eurasian Basin occurred in about 2007.
Warm anomalies such as these, transported within the Atlantic sub‐surface water mass, are not in direct contact with the pack ice that is insulated across a steep pycnocline from the warmer water by a <50m layer of cold surface layer of low salinity. The heat lost by each warm anomaly as it passes eastwards must in part be lost into the bulk of the Atlantic water mass below, but there is good evidence also of significant upward heat flux during transit along the slope: despite microstructure observations that suggest that mixing is very weak across the Arctic halocline, heat budget estimates nevertheless yield significant vertical fluxes. These in turn suggest that decreases in ice thickness of <30 cm may be attributable to this flux, rather than to the supposed consequence of a warming atmosphere over the Arctic Ocean.
The pulses of warm Pacific water that pass north through the Bering Straits are rather variable. Sea surface temperatures at the source of these fluxes in the Bering Sea  closely matches the evolution of the value of the PDO (Pacific Decadal Oscillation). The significance of this observation is that it confirms that the inflow of Pacific summer water (PSW) in the late 1990s through wind forcing of near-­‐surface transport was both unusually warm and unusually strong. The area of this interflow in the southern Canadian basin and the Chukchi Sea corresponds with the area of summer ice reduction during the late 1990s. However, increasing Bering Sea temperatures at the end of the 20th century cannot be formally correlated with relative ice loss in the Arctic Ocean, and an alternative mechanism has been proposed: that the warm pulse of PSW retards winter ice formation and so ensures a more efficient transfer of momentum from wind to the coastal water mass which “in turn causes an imbalance between ice growth and ice melt”.
Variability in summer ice‐cover in the Chukchi Sea, as in the Barents Sea, has been correlated with the values of the AO and the NAO, and hence with the frequency of cyclonic depressions over the Arctic Ocean. During the years 1979-­‐2009 there was an increasing frequency and strength of extreme wind events on the north coast of Alaska during late summer and autumn. Such conditions will not only hasten melting of ice formed the previous winter but, independently of that process, will also increase the apparent area of open water by rafting and compacting small, isolated ice floes.
To summarise the arguments presented so far concerning ice-‐loss in the arctic basin, at least four mechanisms must be recognised: (i) a momentum-‐induced slowing of winter­‐ice formation, (ii) upward heat-‐flux from anomalously warm Atlantic water through the surface low‐salinity layer below the ice, (iii) wind patterns that cause the export of anomalous amounts of drift ice through the Fram Straits and disperse pack-­‐ice in the western basin and (iv) the anomalous flux of warm Bering Sea water into the eastern Arctic of the mid­‐1990s.
These and other observations can be integrated into a model with feedbacks and having two unstable end‐points that is consistent both with classical studies of past climate states, and also with recent analysis of ice dynamics in the Arctic basin by Zhakarov, whose oscillatory model identifies feedback mechanisms in atmosphere and ocean, both positive and negative, that interact in such a manner as to prevent long‐term trends in either ice‐loss or ice‐gain on the Arctic Ocean to proceed to an ultimate state.

The key to this model lies in the distribution of precipitation on Earth, with maxima in the tropics and in high latitudes, so that the Arctic receives an excess of precipitation over evaporation of about one third, which is associated with the permanent presence of the low salinity surface water mass of the Arctic Ocean, separated by a halocline from the saltier Atlantic water below. The presence of this low salinity surface water mass would enable ice cover to fully recover in winter, even in the extreme case in which it was totally absent by the end of summer.
The model also provides a key to understanding the causes of the natural oscillation of the Arctic climate between two states of relative ice cover, depending on the balance of the relative volume of freshwater in the Arctic basin. Zhakarov’s model is conceptually simple: during periods of high precipitation when winter ice forms readily, summer ice cover increases, the atmosphere cools, the arctic front together with its associated rain belt shifts south so that freshwater input to the Arctic Ocean decreases, and winter ice cover is thicker, has a deeper draft, and so survives better in summer.
All this has been available to arctic science since the 1990s, but has been widely neglected perhaps because it suggests that when we are predicting change in arctic conditions we should look to the ocean for the major forcing, rather than to local atmospheric temperature. It emphasises that there is a strong internal relationship between the formation, stability and extent of sea‐ice and the structure of the upper layer of the Arctic ocean: it is the relative area and depth of low-­‐salinity arctic water above the halocline that are paramount to ice formation and its summer survival.
JC reflections
I find this section to be an excellent summary of what is going on in the upper Arctic Ocean and sea ice — and Arctic climate dynamics is one of my areas of expertise.  Longhurst’s argument is consistent with the stadium wave hypothesis [link], and clarifies the role of salinity and freshening from river runoff (from land precipitation) into the Arctic Ocean.  In particular, Longhurst highlights Russian research (which was also integrated into the stadium wave), notably this paper by Zakharov Sea Ice in the Climate System – A Russian View, which is a must-read for anyone interested in this topic.  Further, Longhurst interprets recent sea ice variability in context of the ideas put forth by Zakharov
The complete text of the chapter further clarifies how this view of Arctic sea ice differs from the prevailing view that atmospheric warming and ice-albedo feedback are conspiring to amplify the Arctic warming and melt the sea ice.
Section 8.2 is on Arctic temperatures, in which Longhurst concludes:
There is, therefore, very little support for the implications of the widely-­‐ disseminated and quoted NOAA analysis and graphic used to introduce this section; the 1.5‐3.5C warm anomaly for Arctic regions that it suggests is most probably a result of gridding, and of adjustment and homogenisation of station data.
I agree with Longhurst that the extrapolations/kriging exercises of GISS and Cowtan and Way are misleading; see my previous post [link].  I think the only way to approach the Arctic-wide temperature changes is through reanalyses (data assimilation by numerical weather prediction models) [link]; see this figure from the ECMWF reanalyses [link]:
The narrower, darker bars denote complete global averages, while the lighter, broader bars denote averages taken only over grid boxes which exclude most of the Arctic and Antarctic. Evidently the ranking of average temperatures depends on the data coverage, although the differences are within the bounds of uncertainty associated with the dataset. For the purpose of this illustration we used HadCRUT4 geographical coverage for each month to sample ERA-Interim estimates. The November 2014 HadCRUT4 coverage was used for December 2014 as HadCRUT4 data were not yet available for the latest month.
The narrower, darker bars (including both polar regions) don’t show any evidence of a ‘polar amplification’ of the warming, although the signals from a warming arctic are combined here with a mostly cooling antarctic.  Interesting that when you include the polar regions, that 2005 and 2010 were substantially warmer than ‘warmest year’ 2014.
The processes summarized by Longhurst are occurring in the Arctic Ocean (and are consistent with the stadium wave hypothesis).  The key question is to what extent  atmospheric warming/ice albedo feedback is influencing the recent sea ice decline relative to the natural variability.  In my 2013 Climate Dialogue essay on The Decline of the Arctic Sea Ice, I stated that I figured that the human contribution was 50%, +/- 30%.  After reading Longhurst’s chapter, I am inclining more towards a natural dominance for the recent sea ice decline.  I have argued (consistent with the stadium wave hypothesis) that a recovery of the sea ice is occurring already in the Atlantic sector of the Arctic.  It seems that the warm blob in the northern Pacific is dissipating and Alaska is having a very cold and snowy autumn so far, so perhaps we will soon see a recovery start in the Siberian sector.  The 2015 Arctic sea ice melt season was a little lower than 2013 and 2014, but sea ice volume continues to increase.  The next 10 years of sea ice observations should clarify all this.
Moderation note:  This is a technical thread, keep your comments on topic and civil.  Moderation will be heavier than usual.Filed under: Oceans, Polar regions