What are the physical links between Arctic sea ice loss and Eurasian winter climate?

Remarkable changes have been taking place in the Arctic and, in particular, highly statistically significant decreases in sea ice extent have occurred in all calendar months since 1979 (Simmonds 2015), a year that marks the start of the modern satellite era. These reductions of ice cover have been associated with a number of changes in the Arctic basin, including those of the temperature and the energy and moisture budgets (Screen and Simmonds 2010, Bintanja and Selten 2014).

The presence of the changes have naturally led to the question as to whether the impact of these are confined to the Arctic region or whether remote weather and climate is also influenced in a significant manner. A number of contrasting views on this subject have been expressed and several studies have purported to demonstrate connections between warming and/or ice decline and midlatitude weather and climate extremes (Petoukhov and Semenov 2010, Francis and Vavrus 2012, Petoukhov et al 2013, Tang et al 2013, 2014, Coumou et al 2014). Others have questioned whether these associations are statistically and/or physically robust (Screen and Simmonds 2013, 2014, Barnes 2013, Barnes et al 2014, Screen et al 2014, Hassanzadeh et al 2014), while some investigations suggest that the ostensible associations may have their origin, in part, in remote influences (Petoukhov and Semenov 2010, Screen et al 2012, Peings and Magnusdottir 2014).

In light of this divergence of views on cause-and-effect and of the importance of clarification, the investigation of Sato et al (2014) makes a very valuable and illuminating contribution. Their results, obtained with reanalyses and model experiments, suggest that apparent links between Barents Sea ice coverage and cold Eurasian winters are in fact associated with part of a teleconnection pattern which originates from outside the Arctic, in the North Atlantic Gulf Stream region. They posit that the remote planetary wave atmospheric response to the poleward shift of a sea surface temperature front over the Gulf Stream is associated with warm southerly advection and consequent sea ice decline over the Barents Sea, and a cold anomaly over Eurasia. This remote response would be amplified over the Barents Sea region via interacting with the sea ice anomaly, promoting the warm Arctic and cold Eurasian pattern. However, Sato et al conclude that while the Barents Sea has been regarded a key sector, a complete understanding of the chain of events requires the consideration of midlatitude Gulf Stream forcing.

The broad concept of midlatitude conditions potentially impacting on Arctic sea ice and Eurasian winter extremes is not new, but had not been clearly and fully quantified to date. Scoccimarro et al (2012) explored the sea level pressure teleconnection patterns that were set up by strong tropical cyclones in the tropical Atlantic Ocean. They identified a pattern not dissimilar to that found by Sato et al, this being associated with changes in sea ice coverage, a large high pressure region extending from the Barents Sea to the Black Sea and a broad low pressure region to its southeast, centred at 90°E (their figure 2). Petoukhov and Semenov (2010) reported on the response of the ECHAM5 model to changes in prescribed sea ice concentration in the Barents-Kara Sea. They found their Arctic ice experiments could simulate cold continental temperature anomalies, but that these were 'some 2–3 times weaker' than those observed. They offered some general thoughts on the possible impact of midlatitude and tropical sea surface temperatures (SSTs) but did not comment on where key 'hotspots' for forcing may be located. Peings and Magnusdottir (2014) focussed on the consequences of broad Atlantic SST variations (in the form of the Atlantic Multidecadal Oscillation (AMO)) and suggested that it is plausible that the AMO plays a role in the recent resurgence of severe winter weather over Eurasia. They remarked that because the AMO cycle is associated with Arctic sea, some of the recent winter climate anomalies that have been attributed to sea ice loss may therefore have been related to AMO variability. The authors observed that not taking this factor into account may lead to overestimation of the midlatitude consequences of the Arctic ice loss. This view is consistent with the too-weak response simulated by Petoukhov and Semenov (2010).

One of the substantial advances revealed in the Sato et al study is the clear identification of a localised midlatitude hotspot influencing both the Arctic and Eurasia. They contrasted hemispheric atmospheric circulation and SST patterns in eight Decembers which experienced anomalously warm conditions in the Barents Sea with nine Decembers which had cold conditions there (these relatively large ensembles enhance the statistically significance and robustness of the findings). In the warm-minus-cold composite they documented strong activity-wave fluxes emanating out of the Gulf Stream region, and propagating on to the Barents Sea and thence to central Eurasia. (Strongly consistent with this we have identified an analogous pattern with the very different approach of Rossby wave source identification (Sardeshmukh and Hoskins 1988), emphasising the important role played by this 'trans-Arctic' teleconnection.) Supporting the discovery of this key hotspot they show highly significant positive correlations between winter SST over the Gulf Stream region and lower-troposphere temperature over the Barents and Norwegian Seas, and similarly robust (negative) correlations over central Eurasia. Sato et al back up this compelling observational analysis with numerical model experiments in which diabatic forcing was imposed over the entire Atlantic sector and then only over the Gulf Stream region. The results confirmed that the structure of the response pattern was essentially driven from the hotspot Gulf Stream region. They also found that when the forcing in their linear barotropic model was restricted to the Barents and Norwegian Seas the steady response did not propagate to central Eurasia. This last result is strongly in accord with the interpretive comments of Peings and Magnusdottir (2014) referred to above.

There is still much to learn of the complex interactions between the Arctic and midlatitude climate extremes. There have been a number of recent comprehensive reviews of the thorny issue and its ambiguities (e.g., Cohen et al 2014, Fischer and Knutti 2014, Overland 2014, Wallace et al 2014, Walsh 2014) and a common theme to come from these is the critical need to increase our understanding of how Arctic warming and/or sea ice loss could impact the large-scale circulation patterns. The Sato et al paper represents a significant step along his road.