Shifting weather patterns, particularly increasing extreme events, are observed in populous midlatitudes, and connections to climate change are becoming clearer. The specific roles of rapid Arctic warming and sea ice loss in this story are a topic of intense research and controversy. Instead of converging on answers however, the science seems as unsettled as ever. Many new studies find evidence supporting Arctic–midlatitude linkages, while others report no robust signals. This article presents my interpretation of recent literature to illuminate possible causes of this discrepancy, which include a low signal-to-noise ratio, shortcomings in experiment design, ill-suited metrics of circulation changes, and deficiencies in models used to simulate atmospheric responses to rapid Arctic warming.
Possible explanations for the ongoing controversy surrounding connections between the rapidly warming Arctic and changing weather patterns in midlatitudes are explored.
A blizzard of new studies has investigated connections between the rapidly warming Arctic and extreme weather events in Northern Hemisphere midlatitudes. Media and public interest in the topic has also been high, as the loss of Arctic sea ice is one of the most conspicuous symptoms of human-caused climate change (Notz and Stroeve 2017) and severe weather directly affects individuals, agriculture, and commerce. The recent pace of sea ice loss has been nothing short of breathtaking, as about 80% of the volume has disappeared in a human generation. Possible links between the Arctic meltdown and extreme weather, first hypothesized in 2012 (Francis and Vavrus 2012; Liu et al. 2012), has piqued the public’s interest (Hamilton and Lemcke-Stampone 2014) even before thorough testing, which raised objections by some in the scientific community (Wallace et al. 2014; Kintisch 2014). The hypothesis purports that disproportionate Arctic warming may increase the frequency of persistent weather patterns that can cause drought, heat waves, cold spells, and floods. In the specific case of extreme cold events, overall warming is expected to reduce the occurrence of record-breaking cold (e.g., Ayarzagüena and Screen 2016; Screen et al. 2015), but it is often the duration of a cold spell rather than extreme temperatures that is most disruptive.
Despite intense efforts to understand Arctic–midlatitude linkages, the controversy surrounding its significance and even its existence rages on. New studies come to differing conclusions. While natural variability (such as large-scale ocean temperature fluctuations, volcanic eruptions, and short-term weather) in the climate system—both in observations of the real world and in model simulations—contributes substantially to inconsistencies among studies, other explanations are also possible. I extend the discussion by McCusker et al. (2016) on some sources of uncertainty and offer several others that may be fueling the ongoing dispute.
Sea ice loss ≠ Arctic amplification.
Several recent studies employ model simulations to assess the influence of Arctic sea ice loss on Northern Hemisphere weather patterns, particularly on unusual cold spells in winter continents (e.g., Sun et al. 2016; Meleshko et al. 2016; Chen et al. 2016; McCusker et al. 2016). Each of them aims to isolate the atmospheric response to sea ice loss by forcing a model with reduced sea ice conditions and comparing the resulting atmospheric response to that from simulations with more expansive sea ice. Three of the studies find no robust connection between ice loss and winter weather patterns, while one finds only an intermittent relationship. In contrast, a number of other recent investigations based on observations and model simulations do find a robust connection (e.g., Kim et al. 2014; Mori et al. 2014; Kug et al. 2015; Zhang et al. 2016; Kretschmer et al. 2016; Pedersen et al. 2016), which raises the question, how can these two groups of studies come to such disparate conclusions?
It has been known for many years that the decline in sea ice extent accounts for only part of the rapid warming that has been observed in the Arctic atmosphere in recent decades, known as Arctic amplification (AA), and that the sea ice–induced warming is confined mostly to the lower atmosphere. Some estimates of the direct influence of reduced sea ice extent on AA are as low as 20% (Perlwitz et al. 2015). Several other factors are known to contribute, such as thinning sea ice, to which Lang et al. (2017) ascribe about 37% of observed surface warming, and declining spring snow cover, which augments high-latitude warming in late spring and summer (Estilow et al. 2015). Warming aloft is caused by a combination of heat trapped by additional water vapor from both local and remote sources (Porter et al. 2012), more abundant clouds, and latitudinally varying changes in atmospheric temperature profiles (Pithan and Mauritsen 2014; Burt et al. 2016). Any additional heat that finds its way to the Arctic is amplified by a variety of positive feedbacks in the region. Consequently, in modeling experiments that exclude natural or forced energy exchanges with more southerly zones, a large fraction of AA is missed. Other contributing factors may include model inaccuracies in boundary layer stratification and radiative transfer (Bintanja and Krikken 2016) that affect energy exchanges, as well as heat transported by ocean currents (Polyakov et al. 2017).
The contrast in atmospheric conditions with and without the full contingent of AA factors is conspicuous in Fig. 1 (from Meleshko et al. 2016). The winter atmospheric response to sea ice loss alone (left) exhibits a far weaker AA than the full global warming response (right), both in terms of warming (top) and raised geopotential heights (bottom). Consequently, the weakening of poleward gradients—a major driver of jet stream winds—is greatly understated. It should come as no surprise therefore that circulation changes are also muted in model experiments that target the response to sea ice loss alone and that exclude influences, both natural and anthropogenic, from regions beyond the Arctic.
It has been suggested that the warm-Arctic/cold-Asia pattern observed in recent winters is simply due to natural and/or internal variability (Deser et al. 2017; McCusker et al. 2016; Sorokina et al. 2016). The extremely chaotic nature of the midlatitude circulation makes this a real possibility, but it also seems improbable that the Arctic could lose half of its summer sea ice extent and ∼80% of its volume—accompanied by record-setting AA and water vapor content during the last few decades (Francis et al. 2017)—without affecting the large-scale circulation. Distinguishing this recent signal from the noise of natural variability, however, is a challenge (e.g., Alexander et al. 2004; Barnes 2013; Screen and Simmonds 2013), but new analysis methods are making headway (e.g., Kretschmer et al. 2016; Mann et al. 2017; Peings et al. 2017).
What about the summer months? Evidence supporting a linkage between AA and extreme summer weather is also accumulating. Building on studies by Petoukhov et al. (2013) and Coumou et al. (2014), Mann et al. (2017) find that a weakening poleward temperature gradient favors a split jet stream and the formation of a waveguide. This configuration tends to trap a certain wavenumber range of jet stream undulations, causing persistent weather conditions that often lead to summer heat waves, drought, and flooding. This general relationship is supported by model projections analyzed by Vavrus et al. (2017), who find a belt of weakened 500-hPa zonal-mean zonal winds near the latitude of earlier spring–summer snowmelt in North America, creating relative wind maxima to the north and south. The in-between band of weak winds is associated with amplified jet stream waviness measured as “sinuosity” (Fig. 2, left), which favors drought and heat spells in summer. Wintertime warming over the Arctic produces a similar association between weakened zonal winds and increased sinuosity (Fig. 2, right), suggesting a potential for more persistent weather patterns in the North American cold season, as well.
Realistic stratosphere necessary.
Several recent modeling studies have investigated the role of troposphere–stratosphere coupling in linkages between AA and winter midlatitude weather (Liu et al. 2012; Cohen et al. 2014; Kim et al. 2014; Handorf et al. 2015; Wu and Smith 2016; Zhang et al. 2016; Kretschmer et al. 2016; Zou et al. 2017; Nakamura et al. 2016a). These studies identify a mechanism involving sea ice loss in the Barents–Kara Seas and anomalous Eurasian snowfall in late autumn that favors an enhanced upward transfer of wave activity from the troposphere in late fall that disrupts and weakens the stratospheric polar vortex. Weeks to months later, the weakened polar vortex transfers the wave anomaly back to the troposphere, evident as a negative Arctic Oscillation, which effectively perpetuates the influence of autumn AA on late-winter circulation patterns. Other studies comparing simulations with well versus poorly resolved stratospheres find differing responses, implying an important role for troposphere–stratosphere exchanges (Sun et al. 2015). These results suggest that models used to simulate Arctic–midlatitude linkages may not capture the late-winter atmospheric response to autumn AA if they do not include realistic exchanges of troposphere–stratosphere wave energy (Wu and Smith 2016).
Right time, right place.
Arctic amplification is occurring on a backdrop of large natural variations in the climate system and in combination with other human-caused changes. Isolating the influences of AA on weather changes in the real world, therefore, is challenging. As AA intensifies, however, evidence of its impacts is becoming clearer, particularly the regional and seasonal variations, and its dependence on the background state (Overland et al. 2016; Sung et al. 2016; Nakamura et al. 2016b). The abrupt flip of the Pacific decadal oscillation (PDO) index from negative to positive in late 2013, coinciding with substantial ice loss in the Arctic’s Pacific sector (Chukchi–Beaufort Seas), provides an example of this relationship. Anomalously warm SSTs in the northwest Pacific (such as observed during positive phases of the PDO and North Pacific mode) tend to promote ridging and dry conditions in the western United States (Hartmann 2015; Lee et al. 2015). Abnormal ice loss north of Alaska contributed to regional AA, which further intensified the existing ridge and increased its longevity (Lee et al. 2015; Kug et al. 2015; Sung et al. 2016). During negative PDO conditions, ridging along the coast of western North America is inhibited and consequently the jet stream is not collocated with positive height anomalies in the Pacific sector of the Arctic (if they exist), precluding constructive interference. Similar relationships have been demonstrated in the North Atlantic (e.g., Overland et al. 2016) and near the Barents–Kara Seas (Kim et al. 2014; Kug et al. 2015). The magnitude and spatial pattern of the large-scale atmospheric response to sea ice loss has also been shown to depend on the background climate state (Screen and Francis 2016; Osborne et al. 2017; Overland et al. 2016). These emerging spatial/temporal dependencies suggest that the typical analysis approach that averages over large regions, long periods, and/or many ensemble members will obscure the types of responses caused by regional forcing and/or seasonal mechanisms. If these specific responses in combination with varying background states are not considered appropriately, then the results may cloud, not clarify, our understanding of AA effects on midlatitude weather.
Tropical side of tug-of-war too strong?
In addition to AA, another region of greenhouse gas–induced amplified warming is projected to occur in the tropical upper troposphere (e.g., Barnes and Polvani 2015; Cattiaux et al. 2016; Oudar et al. 2017), although it has yet to emerge in reanalyses. Thus, while AA weakens the poleward gradient in the lower troposphere, tropical amplification (TA) strengthens the gradient in upper levels. The result is expected to be a “tug of war” between these two influences as global warming continues unabated. Two recent studies based on observations suggest that the AA side of this battle has been gaining the upper hand (Feldstein and Lee 2014; Cohen 2016), contributing to an enhanced meandering character to the upper-level midlatitude flow (Cattiaux et al. 2016; Di Capua and Coumou 2016), albeit with large longitudinal and seasonal variability. Analyses of future projections by fully coupled models, however, find no consistent changes in waviness, either in location or season (Barnes and Polvani 2015; Cattiaux et al. 2016; Peings et al. 2017), implying a relative increase in TA’s influence. But are models overstating the magnitude of future tropical warming? Two new studies indicate that some global climate models simulate too much tropical warming compared with satellite retrievals and reanalyses (Santer et al. 2017), which is consistent with findings of an overly stable tropical atmosphere (Sohn et al. 2016). Clearly, the realistic simulation of amplified warming in both the Arctic and tropics is crucial for accurately assessing the combined effects on midlatitude weather patterns as greenhouse gas concentrations continue to rise.
The effects of climate change on extreme weather are a topic of intense scientific interest and of vital societal impact. Some of these effects are clear—such as more severe heat waves, more frequent heavy precipitation events, and more persistent droughts—but other less direct influences are still “up in the air.” The role of a rapidly warming and melting Arctic is one of these factors that challenges present modeling capabilities and dynamical understanding. These limitations are now coming into focus as changes in the real world either confirm or oppose expectations based on simulations, offering avenues to resolve disputes in our understanding of Arctic–midlatitude linkages. Future work using targeted simulations based on suitable models, targeted experimental design, and relevant circulation metrics, as further described in Overland et al. (2016), will undoubtedly dissipate some of the cloudiness obscuring the impacts of a rapidly warming and melting Arctic on weather patterns in temperate latitudes.
Thanks to two anonymous reviewers, Dr. John Fyfe, BAMS Editor Dr. Michael Alexander, and Dr. Steve Vavrus for their constructive comments on the manuscript. JAF is supported by NSF Grant 1304097 and NASA Grant NNX14AH896.
Publisher’s Note: This article was revised on 21 September 2018 to correct the vertical axis in Figure 1.