Linkages Between Arctic Warming and Mid‐Latitude Weather Patterns
The Arctic has been undergoing significant changes in recent years. Surface temperatures in the region are rising twice as fast as the global mean. The extent and thickness of sea ice is rapidly declining. Such changes may have an impact on atmospheric conditions outside the region. Several hypotheses for how Arctic warming may be influencing mid-latitude weather patterns have been proposed recently. For example, Arctic amplified warming could lead to a weakened jet stream resulting in more persistent weather patterns in the mid-latitudes. Or Arctic sea ice loss could lead to an increase of snow on high-latitude land; snow expands on land in autumn, which in turn impacts the jet stream resulting in cold Eurasian and North American winters. These and other potential connections between a warming Arctic and mid-latitude weather are the subject of active research.
The National Research Council convened a workshop on September 12-13, 2013, to review our current understanding and to discuss research needed to better understand proposed linkages. The workshop participants were encouraged to take a global perspective and consider the influence of the Arctic in the context of forcing from other components of the climate system, such as changes in the tropics, ocean circulation, and mid-latitude sea surface temperature.
Many workshop participants noted that research on these linkages is still in its infancy, making it difficult to draw conclusions regarding their existence or their mechanisms. Workshop presentations highlighted the complex mechanisms and feedbacks that link the tropics, poles, and mid-latitudes while also noting that they can be bidirectional and not necessarily direct (e.g., the stratosphere and oceans may act as intermediaries). Understanding how Arctic warming could impact these linkages or other large-scale circulation patterns is critical, but the large natural variability of the atmosphere and Arctic sea ice makes attribution difficult.
Furthermore, many of the proposed linkages are composed of multistep hypotheses, and our understanding of the various steps is complicated by inconsistent evidence from observations and modeling studies. Some participants said that the time series for observations is too short to detect statistically significant trends. It was noted by some participants that trends could be clarified through the quantification of uncertainty in data products. Atmospheric reanalyses and efforts to extend Arctic sea ice records back in time might also be useful for detecting trends. Under-utilized observations could also be harnessed from often overlooked sources, such as the National Ice Center (NIC) operational products.
There is also no standard for objective measurement of important mechanisms such as wave amplitude or atmospheric blocks, which lead to a stagnation of weather patterns. A better understanding of the range of blocking systems and how they are impacted by
various components of the climate system is needed. Participants suggested deploying process-based studies on the “weak links” of the specific hypotheses as a way forward.
Results from modeling studies offer a lot of divergence and are difficult to compare systematically because different boundary and/or initial conditions are often used. To address this challenge, many participants would like the modeling community to organize careful model intercomparisons and conduct model sensitivity studies with identical initial and/or boundary conditions. Moreover, given that sea ice loss is only one driver of Arctic amplification, many participants said attribution studies are needed to understand the proportion of Arctic amplification that is due to processes outside of the Arctic. Although models are useful for providing clues on the mechanisms of potential Arctic linkages, as well as detection and attribution, some participants acknowledged that there are important biases and limitations that must be considered.
Many participants said that combining observations with a hierarchy of models is key to making progress on this issue. Furthermore, attention so far has focused on changes in the meridional temperature gradient, but future studies might also consider other mechanisms of Arctic linkages, such as zonal temperature gradients, static stability, and moisture changes. The climate and weather communities could collaborate to apply the sophisticated methods and diagnostics developed in the meteorological community to better understand linkages. Some participants also thought that the possibility of regional impacts of Arctic warming is an important consideration.
Some workshop participants noted that a large-scale research program dedicated to understanding the mechanisms that link a warming Arctic and the heavily-populated midlatitudes is needed, particularly because it could lead to improved seasonal forecasts. Several participants pointed to the analogous example of an improved understanding of El Niño Southern Oscillation (ENSO) resulting in improved seasonal forecast skill. Participants expressed optimism that significant progress could be obtained because of several recent achievements and events: (a) improvements in atmospheric, oceanic, and sea ice models; (b) longer observational records of sea ice and atmospheric circulation; (c) emergence of a reasonably strong sea ice forcing in recent years; and (d) strong interest in the topic from policymakers, scientists, and the public.
On September 12-13, 2013 the Board on Atmospheric Sciences and Climate and the Polar Research Board held a public workshop that brought together a diverse array of experts to examine linkages between a warming Arctic and mid-latitude weather patterns. The workshop included presentations from leading researchers representing a range of views on this topic.
This workshop was planned and organized by an ad hoc committee of the National Research Council1. The committee was tasked to plan a workshop to address the following questions:
- What do we currently understand about the mechanisms that link declines in Arctic sea ice cover, loss of high-latitude snow cover, changes in Arctic-region
1 This report has been prepared by the workshop rapporteur as a summary of what occurred at the workshop. The planning committee’s role was limited to planning and convening the workshop. The views contained in the report are an interpretation of those presented by individual workshop participants and do not necessarily represent the views of all workshop participants, the planning committee, or the National Research Council.
energy fluxes, atmospheric circulation patterns, and the occurrence of extreme weather events?
2. What may be the possible implications of more severe loss (and eventually, total loss) of summer Arctic sea ice upon weather patterns at lower latitudes?
3. What are the major gaps in our understanding, and what sort of observational and/or modeling efforts are needed to fill those gaps?
4. What are the current opportunities and limitations for using Arctic sea ice predictions to assess the risk of temperature/precipitation anomalies and extreme weather events over northern continents? How might these capabilities improve over time?
The workshop discussions focused on questions 1 and 3. As the research matures, it may become possible to fully address questions 2 and 4 in the future. Committee members planned the workshop structure, identified speakers and attendees, and developed background materials for attendees. They also led the workshop and served as session facilitators. Furthermore, during the workshop planning, they decided that the workshop should expand to include links to Arctic amplification as well, because Arctic amplification is not exclusively driven by sea ice loss.
All workshop presentations are available online at http://dels.nas.edu/global/basc/alpresentations. This summary was written by a rapporteur to present the various ideas and suggestions that arose in the workshop discussion and to synthesize the main discussion items. It does not include any conclusions or recommendations, nor does it cover the full spectrum of issues around this topic. For further information, relevant references that were suggested by workshop participants and planning committee members can be found at http://dels.nas.edu/resources/static-assets/basc/miscellaneous/basc-arctic-linkagesworkshopreferences.pdf.
The structure of this report largely follows the structure of the workshop sessions and is divided into three general topics: the big picture context, observations, and modeling and theoretical work. Some speakers touched on all three topics in their remarks so this summary does not step through each presentation in detail, but rather gathers related points thematically. Brief abstracts of each presentation are in Appendix A. The final section lists research needs and key messages that were identified by individuals in breakout groups.
Rising global average temperatures, and especially intense warming in the northern polar regions, are leading to a rapid loss of the sea ice cap that covers the Arctic Ocean. The seven summers with the lowest sea ice minimums all occurred during the past seven years. Workshop presenter James Screen, University of Exeter, noted that Arctic sea ice loss in summer 2012 broke the previous record low set in 2007, with ice cover half as extensive as it was only 30 years ago. Although summer sea ice extent in 2013 did not set a new record low, it still was consistent with a long-term trend of rapid decline. The observed 2012 record low in summer Arctic sea ice extent is extremely unlikely to have occurred due to internal climate variability according to Rong Zhang, National Oceanic and Atmospheric Administration (NOAA).
In addition to sea ice, spring snow cover has also decreased in the Arctic. As discussed by workshop presenter James Overland, NOAA, many areas of the Arctic experienced almost a 40 percent decrease in snow cover in June 2012 compared to the average from 1971-2000 (Figure 1). David Robinson, Rutgers University, noted that annual snow cover extent over
FIGURE 1 Top: Arctic sea ice extent for the years 2007-2013 compared to the 1981-2010 average. Bottom: June snow cover anomalies in June 2012 in the Northern Hemisphere compared to the long-term 1971-2000 average based on the number of days in the year when a location was snow covered. Shades of brown indicate places that experienced up to 40 percent fewer snow-covered days than average in June 2012. Blue indicates areas that experienced up to 40 percent more snow-covered days than average. SOURCE: Climate.gov, http://www.climate.gov/news-features/featuredimages/record-low-spring-snow-cover-northern-hemisphere-2012.
Northern Hemisphere lands has averaged lower since the late 1980s than earlier in the satellite era which began in the late 1960s.
The loss of Arctic sea ice has occurred at rates much faster than climate models have recently projected, even for the worst-case forcing scenarios. The full disappearance of late-summer Arctic sea ice is possible in the coming decades (Massonnet et al., 2012; Stroeve et al., 2012). Many people may consider this stark indicator of climate change to be a distant concern that does not affect our “real lives.” But such trends do indeed matter greatly to society, because the Arctic may play an integral role in the planetary system of climate and weather well beyond the Arctic region.
Although the workshop topic originally focused on the impact of Arctic sea ice loss, because many workshop participants stressed that sea ice loss is only part of the issue, the discussion was broadened to include the impact of Arctic amplification. This term is commonly defined as amplified warming near the Arctic compared to the rest of the hemisphere or globe in response to a change in global climate forcing (e.g., the concentration of greenhouse gases [GHGs] or solar output). For example, as the sea ice melts because of warmer temperatures in the Arctic, the resulting darker ocean waters are exposed to incoming solar radiation. The darker ocean absorbs significantly more energy than the white, reflective sea ice. The result is an increase in water and air temperatures, which melts even more sea ice. Overland noted that this heat does not stay in the ocean; it eventually is re-emitted to the atmosphere. The result is a smaller difference in temperature between the frozen north and the more temperate mid-latitudes (an important concept for several of the proposed Arctic linkages). The temperature change associated with Arctic amplification occurs not only near the surface, but also through a considerable depth of the troposphere.
Some participants noted that there are several other drivers of Arctic amplification and that the impact of albedo on Arctic amplification may not be as significant as previously thought (Pithan and Mauritsen, 2014; Winton, 2006). Martin Hoerling, NOAA, suggested that the estimated observed 1000-500hPa (hectopascal) Arctic warming could be largely due to natural decadal sea surface temperature (SST) forcing. See also Feldstein’s presentation on the role of the tropics in Arctic Amplification (Appendix A).
Extreme Weather Events
This decade has seen a series of record-breaking extreme weather events taking place around the world (Figure 2). Screen highlighted some severe weather events (e.g., heat waves, drought, flooding) in the mid-latitudes, some of which have been high profile and high impact (e.g., Hurricane Sandy, Russian heat wave). Such events can cause human suffering and have a significant economic impact. In 2011, 14 events in the United States caused losses in excess of US$1 billion each (WMO, 2011).
Proposed Arctic Linkages
The recent extreme weather events and the recent significant decline in Arctic sea ice have spurred an increasing interest in the relationship between a warming Arctic and large-scale climate dynamics, which may have tremendous implications for society. Of particular concern is emerging research that may indicate that Arctic warming can have dramatic impacts upon weather patterns across the heavily populated northern mid-latitudes, and
FIGURE 2 World map showing the record-breaking extreme weather events in the past decade. The numbers refer to the year in the twenty-first century. Blue symbols represent rainfall; red symbols represent heat-waves and droughts; yellow symbols represent hurricanes and cyclones; and the green symbol represents a tornado outbreak. SOURCE: Coumou and Rahmstorf, 2012.
that such impacts could increase as ice cover continues to retreat and the Arctic continues to warm in the coming decades (see Box 1).
The cl mate system reflects a complex combination of many interconnected physical and (often haotic) dynamical processes operating at a variety of timescales in the atmosphere, ocean, and land. A significant portion of the workshop was dedicated to understanding the role of Arctic warming compared to other forcing components of the climate system (e.g., variations in the tropics and stratosphere) in influencing weather in the mid-latitudes. Speakers discussed the various connections and linkages within the climate system and considered whether a warming Arctic now has a stronger influence on atmospheric circulation patterns (e.g., polar vortex, jet stream, modes of variability) and the progression of systems (e.g., blocking).
Large-scale Atmospheric Circulation Patterns: What Drives Mid-Latitude Weather?
Workshop presenter Elizabeth Barnes, Colorado State University, provided an overview of atmospheric dynamics of jet streams, mid-latitude waves, and blocking. She stressed that mid-latitude atmospheric variability is composed of dynamical interactions between circulations of a range of spatial and temporal scales.
Examples of Proposed Arctic Linkages
Below is a list of the proposed linkages that were discussed during the workshop. Further details about some of these linkages can be found later in the report.
Increased Arctic warming weakened temperature gradient weakened, more meandering jet streama more persistent weather patterns in the mid-latitude (Francis and Vavrus, 2012)
Arctic sea ice loss increase in autumn high latitude snow cover more expansive and strengthened Siberian high pressure in autumn and winter increase upward propagation of planetary waves more sudden stratospheric warmings weakened polar vortexb and weakened, more meandering jet stream (Cohen et al., 2012; Ghatak et al., 2012)
Arctic sea ice loss changes in regional heat and other energy fluxes unstable polar vortex cold polar air moves to the mid-latitudes (Overland and Wang, 2010)
Arctic sea ice loss amore meandering jet stream and winter atmospheric circulation patterns similar to a negative phase of the winter Arctic Oscillation frequent episodes of atmospheric “blocking” patterns (Liu et al., 2012).
Arctic sea ice loss southward shift of the jet stream position over Europe in summer increased frequency of cloudy, cool, and wet summers over northwest Europe (Screen, 2013)
Arctic sea ice loss winter atmospheric circulation response resembling the negative phase of the Arctic Oscillation rainfall extremes in the Mediterranean in winter (Grassi et al., 2013)
Arctic sea ice loss negative phase of the tripole wind pattern enhanced winter precipitation and declining winter temperature in East Asia (Wu et al., 2013)
a The “typical” jet stream meanders north and south, however it can weaken and slow, allowing the meanders to become larger.
b The polar vortex is a large-scale region of air that is contained by a strong west-to-east jet stream that circles the polar region. The polar vortex extends from the tropopause (the dividing line between the stratosphere and troposphere) through the stratosphere and into the mesosphere (above 50 km). Cold temperatures are associated with the air inside the vortex. Definition from http://ozonewatch.gsfc.nasa.gov/facts/vortex_NH.html
Barnes focused her presentation on the polar jet stream2 in the Northern Hemisphere, which flows over the middle to northern latitudes (50-60) of North America, Europe, and Asia and their intervening oceans. The daily jet location pattern of the north Atlantic jet (Figure 3) has three distinct peaks, but it varies significantly over decadal time scales.
Jet streams and other north Atlantic and north Pacific circulations and variability behave differently depending on the season. Barnes noted that jet streams and storm tracks are strongest over the oceans during winter and shift further north in the summer. Jet and storm track strength and location are driven by changes in:
- wave propagation and momentum fluxes (e.g. Chen and Held, 2007; Chen et al., 2008; Kidston and Gerber, 2010);
- meridional temperature gradients (e.g. Haarsma et al., 2013a; Harvey et al., 2013);
- static stability (e.g. Frierson et al., 2007; Lu et al., 2010);
- ocean influences on the meridional overturning circulation (e.g. Chen and Held, 2007; Chen et al., 2008; Woollings et al., 2012);
- eddy phase speed increases (from increasing upper-troposphere/lower stratosphere winds [e.g. Chen and Held, 2007; Chen et al., 2008]);
2 Jet streams are fast flowing, narrow air currents in the near tropopause. The major jet streams flow west to east and their paths typically have a wavy shape.
FIGURE 3 Jet latitude index3 PDF (probability distribution function) for NCEP/NCAR (National Centers for Environmental Prediction/National Center for Atmospheric Research) reanalysis, where the black line represents the climatological distribution; blue line, Greenland blocking days;4 green line, European blocking days; and orange line, Iberian wave blocking days. The dotted line represents jet latitude index when no blocking in the three sectors is detected. The black vertical line shows the latitude of the central peak. SOURCE: (Davini et al., 2013)
- rising tropopause5 (e.g. Lorenz and DeWeaver, 2007); and
- changes in wind speed over the region (e.g. Haarsma et al., 2013b; Mizuta, 2012).
Modes of Atmospheric Variability
Modes of variability were discussed at the workshop in the context of linkages (see Box 2 for additional details). The dominant mode of atmospheric variability in the Northern Hemisphere extratropics is the Northern Annular Mode/Arctic Oscillation (NAM/AO), which can influence surface temperature and precipitation, especially the frequency of extreme events. There may be potential for seasonal predictability of the AO (Riddle et al., 2013), but it has been found to be shorter in practice (7-10 days). It is not clear why there seems to be skill in predicting the wintertime AO.
The manifestation of the AO in the Atlantic sector is commonly referred to as the North Atlantic Oscillation (NAO). The NAO is also sometimes referred to as a “Greenland Block.” The NAO is the largest contributing pattern to European interannual variability and plays a significant role in predictions of European winter climate. The predictability of the NAO is limited to seasonal timescales (NRC, 2010).
While not extensively discussed at this workshop, participants identified ENSO as an example of a mode of variability in the tropics that can influence weather in the mid-latitudes (i.e., teleconnection). Participants noted that seasonal forecasts improved as our understanding and predictability of ENSO grew; a greater understanding of Arctic influences on mid-latitude weather might allow seasonal (3 months) climate forecasts to be further improved.
3 “Jet latitude index” is defined as the daily latitude of maximum low‐level (mass‐averaged 925 700 hPa) zonal winds zonally averaged over the Atlantic sector (0 –60 W, 15 –75 N; Woollings et al., 2010).
4 A Greenland Block is a very strong area of high pressure located over Greenland.
5 The boundary in the atmosphere between the troposphere and the stratosphere.
Characteristics of NAM/AO, NAO, and ENSO
NAM/AO is a measure of the surface pressure gradient between the polar and subpolar regions of the Northern Hemisphere (Thompson and Wallace, 2000). When the AO index is positive, surface pressure is low in the Arctic region, which is associated with a strong North Atlantic jet (and a strengthening of the polar vortex), thus keeping cold Arctic air in the polar region. When the AO index is negative, the Arctic tends to have high pressure, weaker zonal winds, resulting in a greater movement of polar air into middle latitudes (Overland et al., 2011). The AO was strongly positive in the early 1990’s compared to the previous 40 years; however, the AO has been low and variable for the past 9 years.
The NAO is characterized by changes in the North Atlantic jet stream and storm track location and intensity as well as zonal and meridional heat and moisture transport (Hurrell, 1995). The positive phase of the NAO is associated with strengthened westerlies due to an increase in the pressure gradient over the North Atlantic. Such strong westerlies allow cold air to drain off the North American continent allowing the air to flow northeast across the Atlantic into Europe. The weather for both North America and Europe is mild compared to the weather during a negative NAO. The negative phase of the NAO is characterized by a decrease in the pressure gradient across the North Atlantic, which results in a weakening of the westerlies allowing cold air to build over Canada and the eastern parts of the United States. NAO phases vary from year to year, but they can also remain in one phase for intervals lasting several years.
ENSO refers to a set of weather patterns associated with variations in sea surface temperatures (SSTs) of the tropical eastern Pacific Ocean and in the sea level pressure in the tropical western Pacific. During an El Niño phase, the tropical easterly winds decrease, warm SSTs cover the central and eastern tropical Pacific, and the heaviest rainfall moves east. El Niño is also characterized by sea-level pressure that is lower in the eastern Pacific and higher in the western Pacific. The opposite pattern occurs during a La Niña phase. In North America, El Niño winters tend to be warmer and drier than average in the northern parts of the United States, whereas northern Mexico and the southern United States typically experience wetter and cooler winters On the other hand, La Niña causes above-average precipitation across the northern United States, with precipitation in the southwestern and southeastern states typically below average. El Niño recurs at irregular intervals ranging from 2 years to a decade.
Barnes noted that the Atlantic jet stream variability is affected by the NAO and the east Atlantic pattern,6 which are important to study in the context of Arctic linkages because they are close to the Arctic (and thus more likely to be influenced).
Rossby waves, or large meanders within the jet stream, typically travel eastward. Rossby wave breaking leads to jet stream variations, where cyclonic breaking dominates poleward of the jet and tends to drive the jet southward and anticylonic breaking dominates equatorward of the jet and tends to drive the jet northward (Benedict et al., 2004; Strong and Magnusdottir, 2008).
High-latitude wave-breaking is correlated with NAO on decadal timescales. Rossby wave breaking also serve as the mechanism for the reversal of the meridional gradient of geopotential height fields that is typical of blocks.
Atlantic blocking patterns are a critical component of NAO variability (Woollings et al., 2008). In general, a block is characterized by an atmospheric phenomenon in which a large, quasi-stationary anticyclone develops in the mid-latitudes and persists for several days or longer, blocking the ambient westerly winds and weather systems (Berrisford et al., 2007; Woollings et al., 2008). Currently, there is no consensus on exactly what type of system should be classified as a block. Blocks lead to a stagnation of weather systems and are associated with extreme weather events in the mid-latitudes (e.g. cold snaps, heat waves). They can persist for days to weeks (e.g. Black et al., 2004; Dole et al., 2011).
6 The (EA) pattern is a prominent mode of low-frequency variability over the North Atlantic, appearing in all seasons except for summer. The pattern is structurally similar to the NAO. The anomaly centers of the EA pattern are displaced southeastward to the approximate nodal lines of the NAO pattern. Definition from: http://www.cpc.ncep.noaa.gov/data/teledoc/ea.shtml
FIGURE 4 Atmospheric variables listed in order of ascending difficulty in detection of change. Low values imply a response that is easier to detect than high values. SOURCE: Adapted from Screen et al., 2013.
Barnes said that the e are three main blocking centers in the Northern Hemisphere: the Atlantic Ocean, Europe, and eastern Asia, and most occur in winter season because of greater wave-breaking activity in winter.
Barnes stressed that jet stream shifts, Rossby wave propagation and wave breaking, and blocki g events are all tightly coupled together. Mid-latitude weather variability is also dominated by interactions of the large-scale flow with synoptic-scale eddies. The latter help balance the global energy budget by “stirring” the atmosphere and moving cold air toward the equator and warm air toward the poles, which in turn, reduces the equator-to-pole tempe ature contrast.
Challenges in Detecting and Attributing Changes in Mid-latitude Weather
Variability of the climate system may be due to non-linear dynamical processes intrinsic to the atmosphere (internal variability; as discussed above), or to variations in natural or anthropogenic external forcing (external variability; e.g. GHG emissions, volcanoes). Barnes noted that the internal variability of the atmosphere is high, which complicates attribution of Arctic influences on the mid-latitudes. Screen et al. (2013) suggest that present-day trends in mid-latitude weather driven by Arctic change are most likely masked by internal variability and that the detection of extreme events is even more difficult (Figure 4). Some participants said that the patterns associated with a warming Arctic do not match well with any known natural variability modes (e.g., NAO, AO), which supports that idea that th recent trends are largely anthropogenic.
Barnes noted that mid-latitude jet position has a large amount of internal variability, on varying timescales (i.e., daily, yearly, decadal). Furthermore, blocking frequency is highly variable. She suggested that, given the large internal variability of the jet stream and
blocking, any potential effect of Arctic warming on mid-latitude weather is unlikely to be detectable with current observations. Furthermore, Overland noted that the time series is too short (i.e., significant loss of Arctic sea ice has only been occurring for about the past 6-7 years) to robustly differentiate Arctic forcing of mid-latitude extremes from random events. Overland noted that mid-latitude attribution remains difficult and controversial because interactions between Arctic forcing and chaotic mid-latitude flow are complex. One would not expect events to happen the same way every year even with similar Arctic forcing.
Ultimately, participants noted that the issue of Arctic warming impacting regions outside the Arctic is much larger than just sea ice retreat and includes processes encapsulated under the larger umbrella of Arctic amplification. In addition, participants stressed that it is critical to assess both natural and anthropogenic mechanisms leading to sea ice loss. It is unknown to what extent rapid sea ice loss, particularly after 2005, is a part of multi-decadal oscillation or a result of Arctic warming.
Examples of Connections in the Global Climate System
Several workshop speakers presented examples of the connections and feedbacks linking the tropics, mid-latitudes, and the poles, which highlight the interconnectedness and complexity of the global climate system.
Polar Stratosphere Influence on the Mid-Latitude Troposphere
In his remarks, workshop speaker Paul Kushner, University of Toronto, discussed the coupling between the polar stratosphere and the extratropical troposphere. Stratospheric
circulation variability occurs on longer time scales than tropospheric variability and can be characterized by the strength of the polar vortex, which peaks during winter in the Northern hemisphere and late spring in the Southern Hemisphere (Baldwin and Dunkerton, 2001).
Kushner said that it is well known that extratropical stratospheric variability is controlled by atmospheric composition (e.g., ozone and GHGs) as well as planetary waves (i.e. Rossby waves) propagating upward from the troposphere (Shindell et al., 1999; Sigmond et al., 2008). However, there is also “stratospheric influence”, that is, variability of the stratosphere impacting tropospheric circulation on intraseasonal timescales.
Kushner discussed the pathway for this coupling, which is known as the “NAM/wave mean flow interaction pathway.” Essentially, a strengthened polar vortex leads to positive AO/NAM with a 2- to 4-week lag due to wave mean flow interactions. This interaction has been well studied (i.e., Baldwin and Dunkerton, 2001; Polvani and Waugh, 2004). Kushner also discussed a second pathway for stratosphere-troposhere coupling: the “wave reflection pathway.” The polar vortex can create a reflective surface for upward propagating planetary waves (e.g., Rossby waves). The reflective signal in the troposphere shows up as regional circulation anomalies such as the NAO. This connection has been less well studied, but it continues to be an active area of research (e.g., Perlwitz and Harnik, 2004; Shaw and Perlwitz, 2013).
Kushner said that it is important to examine whether the stratosphere serves as a bridge that connects Arctic sea ice loss to changes in the mid-latitude temperature, storm track, and circulation. As workshop presenter Steve Feldstein, Pennsylvania State University, noted, given the large area (i.e., large spatial scales and long timescales) influenced by Arctic sea ice anomalies, it is likely that the stratosphere plays a role.
Snow Cover Influence on the Mid-Latitude Winter Weather
Judah Cohen, Atmospheric and Environmental Research, said that Siberian snow cover has been shown to influence mid-latitude winter weather. Comparison of correlations between October Siberian snow cover and September Arctic sea ice extent and the winter AO shows that the Siberian snow cover is much more highly correlated with the winter AO than with sea ice.
Kushner said that the stratosphere may act as an intermediary in the linkage between October Eurasian snow cover and wintertime NAM/AO anomalies (e.g., Cohen and Entekhabi, 1999). This linkage likely is driven by a snow-forced planetary wave that induces stratosphere-troposphere coupling (Cohen et al., 2007; Gong et al., 2002).
Mid-Latitude Influence on the Arctic
Workshop speaker John Gyakum, McGill University, presented a case study from a synoptic meteorology perspective to illustrate how mid-latitude processes can influence Arctic weather. He discussed the apparent impacts of extratropical cyclogenesis (i.e., a low-pressure area genesis) in causing an extreme Arctic wind event along the Beaufort Sea coast in September 1999. This case study is associated with at least three extreme phenomena: (1) extreme subtropical water vapor transport; (2) extreme explosive cyclogenesis in the northeastern Pacific; and (3) extreme surface winds in the vicinity of Tuktoyaktuk, a small town on the Beaufort Sea.
This case study, Gyakum said, seems to suggest that if one component of future climate change is an eastward displacement of storm tracks into the northeastern Pacific, then more episodic events such as this one can be expected. Additionally, the decreasing coverage of Arctic sea ice may result in a longer season of Arctic cyclogenesis, with associated extreme winds and coastal erosion.
Tropical Influence on the Arctic
In his remarks, Feldstein presented evidence of the tropics playing a role in Arctic amplification. He noted that there is some disagreement whether Arctic amplification is driven by sea-ice albedo feedback or by poleward transport of heat from the lower latitudes. There has been much research on the former, but relatively little on the latter. The Tropically-Excited Arctic warMing (TEAM) is a proposed mechanism for tropical convection, Arctic amplification, and a reduction of Arctic sea ice. The trend toward stronger and more localized convection over the tropical Indo-Pacific Warm Pool results in the propagation of Rossby waves toward the poles. These Rossby wave trains warm the Arctic by transporting heat and moisture poleward, inducing sinking motion/adiabatic warming7 over the Arctic and increasing the downward flux of infra-red radiation (Lee et al., 2011a; Lee et al., 2011b; Yoo et al., 2011; Yoo et al., 2012a; Yoo et al., 2012b). Dargan Frierson, University of Washington, also discussed the transport of heat to the Arctic but noted that some modeling studies suggest that heat transport to the Arctic is reduced as the Arctic becomes warmer (Appendix A).
Arctic Influences on the Climate System
Arctic Influence on the Tropics
In his remarks, Frierson noted that model simulations show that Arctic regions affect climate elsewhere. For example, Arctic changes can affect tropical rainfall. Work by Chiang and
7 A process whereby the temperature changes because of an expansion or compression.
Bitz (2005) demonstrates a strong sensitivity of tropical rain bands to Arctic sea ice increases. Rain bands shift southward as a result of the subsequent cooling in the northern high- and mid-latitudes because of the increased sea ice. Climate changes in the extratropics also influence the storm tracks over the Southern Ocean.
Furthermore, Frierson noted, there is evidence of “interhemispheric teleconnections”. A recent study shows a poleward shift of the Southern Hemisphere jet stream in response to Northern Hemisphere extratropical cooling alone (Figure 5; Ceppi et al., 2013).
Arctic Influence on the Mid-Latitudes: A Synoptic Meteorology Perspective
Gyakum presented a second case study, which illustrates how Arctic processes may have led to the formation of a mid-latitude storm. He discussed the apparent impacts of Arctic air mass formation on explosive extratropical cyclogenesis in January-February 1979. Northwestern Canada is a primary location for the formation of air masses that impact North America and is also a region of significant cold-season warming during the past several decades. Deep-tropospheric cold air regions facilitate triggers for cyclogenesis in the form of synoptic-scale troughs. This case study highlights Arctic air mass formation mechanisms and possible long-term changes in such mechanisms.
Arctic Influence on the Hydrological Cycle in the Mid-Latitudes and Tropics
Ian Eisenman, University of California, San Diego, said that change in surface albedo (due to Arctic sea ice loss) has been suggested to affect the hydrological cycle in the midlatitudes and tropics. Through modeling studies he found that ice albedo impacts the global hydrological cycle in cold climates, but not in warm climates, with the transition occurring at a point relatively near to the present-day climate. He said these results suggest that ice albedo effects are important for understanding the hydrological cycle in climates colder than today, including many paleoclimate environments.
FIGURE 5 Surface zonal wind change (contours) and control (shading) from cooling at 50 N. SOURCE: Ceppi et al., 2013
Arctic Influence on Ocean Circulation
Arctic change indirectly influences the mid-latitudes through changes in ocean stratification and circulation. Many workshop participants asserted that understanding how longer-term ice loss will impact ocean circulation and eventually mid-latitude weather is critical. In her remarks, workshop speaker Marika Holland, NCAR, presented modeling studies that indicate that freshwater inflow to the Arctic Ocean due to sea ice melt plays an important role in controlling the deep North Atlantic ocean convection and ocean meridional circulation, which in turn affects global climate (Kug et al., 2010). Modeling studies project that significant loss of Arctic sea ice will reduce solid (ice) transport to the North Atlantic but will increase liquid transport to the North Atlantic (Jahn and Holland, 2013). This increase in freshwater liquid transport to the North Atlantic could have implications on North Atlantic circulation (decrease in the Atlantic meridional overturning circulation [AMOC]) resulting in reduced ocean heat transport and reduced warming locally. Regarding the impact of longer-term ice-loss changes on mid-latitude weather, these ocean changes must be considered, she said.
Holland also discussed work by Laura Landrum, NCAR, and colleagues that showed the oceanic response to Arctic sea ice loss. They ran a summer ice free Arctic simulation with summer ice loss forced through albedo modifications and no greenhouse-gas changes. The results show a decrease in AMOC in the first 40 years of the ice free simulation, which is (largely) attributable to ice loss.
Oceanic Feedbacks Influence on Arctic Amplification
Clara Deser, NCAR, discussed the role of oceanic feedbacks in the atmospheric response to GHG-induced Arctic sea ice loss. Her model results show that the primary effect of having an interactive ocean is to deepen the atmospheric temperature response to the mid-troposphere (about 400 hPa) compared to the atmosphere-only run in which the thermal response is confined to the planetary boundary layer (below 850 hPa). This result highlights the contribution of high-latitude ocean feedbacks to Arctic amplification. According to Deser, the near-surface atmospheric circulation response did not appear to be sensitive to ocean coupling.
From this big picture context, where Arctic sea ice, atmosphere, and ocean processes are tightly coupled and are all relevant controls on mid-latitude weather, many participants stressed the importance of understanding the various processes on a global scale and how they interact with one another. This in turn is necessary for understanding not only the impact of Arctic sea ice loss on mid-latitude weather, but also more generally the interactions between Arctic and mid-/low-latitude processes (i.e., the pushing and pulling forces on the mid-latitudes). Furthermore, participants noted that it is evident from the workshop that this issue is much larger than Arctic sea ice loss (the title of the workshop). Rather, it would be more appropriate to place focus on the processes encapsulated under Arctic amplification.
Some of the workshop speakers urged participants to think systematically about how Arctic changes might directly influence the mid-latitudes and about how changes in mid-latitudes might be affected by the Arctic through feedbacks. They discussed components of the climate system as we currently understand it while acknowledging complicating factors such as interacting systems and feedbacks that make attribution difficult. For example, some
participants pointed out that any change that projects onto modes of variability (e.g., AO/NAM) may simultaneously affect the Arctic and mid-latitudes.
One goal of the workshop was to discuss observational evidence of possible Arctic linkages. This section highlights observational evidence of some mechanisms that were raised at t e workshop as possibly linking Arctic amplification and mid-latitude weather:
- Arctic warming faster than the Northern Hemisphere
- Decrease in the temperature gradient between the Arctic and the mid-latitudes
- Slowing of upper-level zonal winds
- Upper-level flow becoming more meridional
- Increase in the amplitude of large-scale waves
- Increase in blocking events
- Large-scale waves progress more slowly eastward
- Increase in extreme events
- Weakening of the polar vortex
Arctic Warming Faster Than the Northern Hemisphere
As discussed in the previous section, the Arctic is warming faster than the tropics and midlatitudes. This has resulted in a greater flux of energy from the newly exposed open ocean (relative to ice-covered waters). The strong Arctic warming is not confined to a shallow surface layer, but extends through a ufficiently deep layer of the atmosphere to influence geopotential thicknesses. In his remarks, Overland presented evidence of anomalous temperature patterns in the Northern Hemisphere (Figure 6). He acknowledged that we cannot prove that these changes are due to changes in the Arctic sea ice loss, but something unusual is occurring.
FIGURE 6 Anomolous temperatures in the Northern Hemisphere. Figure is from December 2009, which had a record negative AO. Note the below-average temperatures throughout the United States and the above-average temperatures around the Arctic. SOURCE: NCEP/NCAR.
In her remarks, Jennifer Francis, Rutgers University, said that current Arctic amplification is strongest in autumn and winter and is a relatively recent phenomenon, which begins to appear in the mid-1990s. She said that this short time series makes it difficult to find a statistical trend in the observations to support the existence of Arctic linkages. Others in the room disagreed that Arctic amplification recently emerged.
Decrease in the Poleward Temperature Gradient
It has been suggested that the increase in the flux of energy from the newly exposed open ocean has led to the weakening of the poleward temperature gradient (i.e., a decline in the tempe ature contrast between the Arctic and mid-latitudes). Francis presented observational evidence of this in Figure 7.
Overland discussed how upper-level atmospheric circulation in north-central Asia in winter responded to the recent large-scale reduction in the north-south temperature gradients, reducing jet-stream zonal velocities. Overland believes that the reduction in jet-stream zonal velocities led to the penetration of storms into northern Asia from the west, increasing the strength and persistence of the Siberian High, and thus increasing the intra-seasonal probability of multiple cold air events over eastern Asia.
FIGURE 7 A negative trend in poleward eopotential thickness gradient (1000-500 hPa) from 1979 to 2012. Left: Change in eopotential thickness in 1000-500 hPa during 1979–2012 in autumn (October/November/December [OND]). Statistically significant changes in geopotential thickness (the white X’s) are apparent over much of the Arctic regions. Right: Trends in poleward geopotential thickness gradient (1000-500 hPa) between 80-70 N and 40-30 N in units of meters per decade corresponding to longitude. The time period is from 1979 to 2012. A value of -10 m/decade is about -2 percent per decade. The red bars are significant at the 90 percent confidence level. Places where the geopotential thickness anomalies are greatest (i.e., where the temperature gradient has weakened the most; left figure8) correspond with a negative trend in poleward geopotential thickness gradient (red bars in the right figure). This signal is strongest in autumn and more variable in summer. Data are from the NCEP/NCAR Reanalysis. SOURCE: Jennifer Francis, Rutgers University.
8 The top figure shows geopotential thicknesses, not temperatures, although a greater temperature will result in a larger 1000-500 hPa distance.
Upper-Level Zonal Winds Decreasing Where Temperature Gradient Weakens
Some participants said hat reduced poleward temperature gradients should result in weakened westerlies. Overland believes that the similar penetrations of cold air into the eastern United States in Decembers 2009, 2010, and 2012 relate to a shift in the long-wave upper-level atmospheric wind pattern (Figure 8). This shift coincides with warmer temperatures and greater geopotential thickness over northeastern Canada, major sea ice loss during October in Baffin Bay, a positive Greenland Blocking Index (greater 500 hPa geopotential heights), and record negative values of the AO index. Such a combination of events amplified and shifted the climatological atmospheric wind pattern westward and allowed deeper southward penetration of cold air into the United States. Northward airflow over Davis Strait acted as a positive feedback to maintain the higher air temperature anomalies. He acknowledged that this is not statistically significant given that this pattern occurred 3 years out of 4, but suggests the wind shift may be the start of a trend.
Francis presented evidence of zonal winds at 500 hPa (roughly 5,500m above sea level) weakening in autumn (ND) and winter (January/February/March [JFM]; Figure 9). Figure 10 shows evidence of the upper-level zonal winds decreasing in areas where the temperature gradient has also weakened. This is more apparent in autumn, but also occurs in summer. John Walsh also presented evidence on zonal wind speeds at 500hPa (Figure 11). They weakened from 1979 to present in autumn (OND), but if observations are extended back to 1948, then zonal wind speed increases from the 1950s to the late 1970s. He noted that there is no known corresponding trend of increasing sea ice cover during the summer/autumn period in these decades. Francis noted that Walsh considered 30-80oN, whereas she took a more regional fo us (40-60oN).
Upper-Level Flow Becoming More Meridional
Another proposed mechanism is upper level winds becoming more meridonal flow due to the decrease in wind speeds. Francis presented evidence of this occurring in autumn (OND; Figure 12).
FIGURE 8 Left: Composite 1000-500 hPa geopotential thickness anomaly field for Decembers 2009, 2010, and 2012. Right: 500 hPa zonal wind anomaly for Decembers 2009, 2010, 2012. Weakened westerlies, especially in the northwest Atlantic, help support the idea that the extraordinary warming over Baffin Bay led to a more meandering jet stream. Data downloaded from the NCEP atlas. Data from NCEP/NCAR reanalysis. SOURCE: James Overland, NOAA.
FIGURE 9 Seasonal zonal wind speed at 500 hPa from 1980 to 2010. Green line: spring (April/May/June [AMJ]); red line: summer (July/August/September [JAS]); blue line: winter (JFM); brown line: autumn (OND). SOURCE: Jennifer Francis, Rutgers University
FIGURE 10 Left: 1000-500 hPa geopotential thickness anomalies in autumn (OND) from 2000 to 2012. Right: Anomalies in zonal winds at 500 hPa from 2000 to 2012. Francis noted that areas where there is a decrease in zonal winds (500 hPa) correspond to areas in t e Northern Hemisphere where there are weakening temperature gradients. SOURCE: Jennifer Francis, data from NCEP/NCAR.
FIGURE 11 Autumn (OND) mean zonal wind at 500 hPa (30-80ºN, 0-360ºW), 1948-2012. Data from NCEP. SOURCE: John Walsh, University of Alaska, Fairbanks.
FIGURE 12 Trends in meridional component of the 500 hPa wind in autumn (OND; 1979-2011). Top: 1000-500 hPA change in geopotential thickness in OND. Bottom: Trend in waviness in OND from 1979 to 2011 in the Northern Hemisphere. Blue is less wavy and red is more wavy. F ancis noted that the areas with a larger change in geopotential thickness (top) correspond well with the pattern of areas with a more meridional flow (bottom). She also noted that there is a less cohesive pattern in summer (JAS), but there is still a correspondence between a weakening temperature gradient and a more meandering upper-level flow.
Some participants said that if there is a statistically significant slowing of waves and weakening of westerlies, then we still should acknowledge that this trend is not necessarily caused by one single mechanism. In addition, it is critical that we can point to other competing mechanisms, not just random “noise” (natural variability).
Increase in Rossby Wave Amplitude
It has also been suggested that the amplitude of Rossby waves is increasing (i.e., an increase in the north-south extent of the waves), which coupled with the decrease in zonal winds, results in a jet stream that is more meandering. Francis said there is a positive trend in the size, or amplitude of the waves, in autumn (OND; Figure 13). She also presented evidence that there is a positive trend in the frequency of ridges according to longitude.
Both Screen and Barnes presented evidence from Screen and Simmonds (2013) and Barnes (2013) that suggest that amplitude trends are metric dependent. Francis also noted this variable is difficult to measure objectively. Screen and Simmonds (2013) measured wave amplitude differently from Francis and Vavrus (2012) and found significant increases in meridional wave amplitude over Europe during spring (AMJ), but not in any other months or seasons. They also found significant decreases in zonal amplitude (a measure of the intensity of atmospheric ridges and troughs at 45 N) in the entire Northern Hemisphere and also individually over Europe and Asia (Figure 14). Screen also noted that they found statistically insignificant positive trends in all seasons in the North America and North Atlantic regions, which is in contrast to the comparatively larger (and significant) increases in summer (JAS) and autumn (OND) found by Francis and Vavrus (2012).
FIGURE 13 Large-scale wave trends in autumn (OND) in the Northern Hemisphere. Top: Latitude of peak of ridges. Middle: Latitude of base of troughs. Bottom: The difference between the peak of ridges and the base of troughs (i.e., a measure of wave amplitude). Amplitude of large-scale waves appears to be increasing since the early 1990s. SOURCE: Jennifer Francis, Rutgers University.
FIGURE 14 Linear trends from 1979 to 2011 at 500 hPa geopotential height in seasonally averaged meridional amplitude. Statistically significant trends are identified by the black dots. The trends in meridional amplitude are positive in summer (JAS) and autumn (OND) and negative in winter (JFM) and spring (AMJ), but none of these is significant. Only three of the trends for individual wavelengths are statistically significant. Left: Meridional amplitude. Right: Zonal amplitude. SOURCE: Screen and Simmonds, 2013.
Increase in Blocks
It has been proposed that blocking trends will be more likely to increase due to the increase in Rossby wave amplitude. Barnes noted that the multiple studies on observed blocking trends have had mixed results, and that it is likely that results are dependent on the methods used by the authors. Francis presented evidence of an increase in blocking events in autumn (OND) in the Northern Hemisphere from 1980 to 2010 (Figure 15). Barnes noted that some studies show that high-latitude blocking over the North Atlantic has decreased in the past 40 years (Barnes, 2013; Croci-Maspoli et al., 2007; Davini et al., 2012; Figure 16) whereas other studies show that low-latitude blocking has increased (Croci-Maspoli et al., 2007; Davini et al., 2012). Barnes (2013) did not find an increase in blocking trends, however, she did point out that detection can be dependent on the algorithm used, so additional studies are needed.
Barnes suggested that reported trend (from Francis and Vavrus, 2012) in wave amplitude do not appear robust because no significant trends are found when (1) daily wave extents are analyzed instead of seasonal maxima and minima and (2) a larger range of isopleths are analyzed. A poleward shift of the isopleths with Arctic amplification may appear as a change in wave extent when a narrower range of isopleths is used instead (Barnes, 2013).
Large-Scale Waves Progress More Slowly Eastward
It has been proposed that both the weakening of west-to-east upper-level winds and the more meandering trajectory of the jet stream have resulted in large-scale waves in the jet stream to progress more slowly eastward. Francis presented evidence of this slowing (Figure 17), but she noted more work is needed to assess changes in the speed of large-scale wave progression.
FIGURE 15 Blocking events have increased while zonal winds have decreased. Blocking events (days/year; blue line) plotted with zonal wind speed at 500 hPa (dashed line) in autumn (OND) in the Northern Hemisphere from 1980 to 2010. SOURCE: Jennifer Francis, Rutgers University.
FIGURE 16 NCEP blocking trends from 1951 to 2010. Blue areas experienced a decrease in blocking events; red areas experienced an increase in blocking events. SOURCE: Davini et al., 2012.
FIGURE 17 The speed of ridges (longitude/day) from 1980 to 2010, which appears to be decreasing. SOURCE: Jennifer Francis.
In agreement with Francis, Barnes (2013) found that Rossby wave speeds at 500 hPa decreased in OND. However, she found no decrease when 250 hPa meridional wind is used from 1980 to 201 (Figure 18). Francis noted that this level is near the tropopause, often above the jet stream, and can be affected by dynamics of the stratosphere.
Barnes stressed that relationships between Rossby wave propagation and zonal wind speeds are complex. She believes that the hypotheses put forth (e.g., of Francis and others) could be correct, but there are not enough data to accurately detect trends.
More Persistent Weather Patterns, Extremes More Likely
Some workshop participants noted that weather conditions associated with the slower large-scale waves become more persistent because of an increase in blocks, increasing the probability of the types of extreme weather associated with long-lived weather conditions.
Francis noted that this variable is difficult to measure. She presented work from Vavrus on upper-air circulation anomalies during extreme weather in Chicago (Figure 19). She also discussed a regression example from Tang et al (2013), which shows that winter sea ice anomalies have a strong association with extreme cold events over the United States.
Weakening of the Polar Vortex
Workshop participants briefly discussed the hypothesis that Arctic sea ice loss leads to changes in regional heat and other energy fluxes, which result in a weak (or unstable) polar vortex. “Normal” strong polar vortex winds, which circle the Arctic from west to east, isolate cold polar air from the mid-latitudes in winter (Figure 20). Thus, it is typically mild across the eastern United States, Europe, and East Asia during winters when the polar vortex is strong. A weak or unstable polar vortex has a more north and south meandering pattern (rather than west to east). This allows cold air from the Arctic to spill into the mid-latitudes and warm air from the subtropics to move into the Arctic. It is typically colder across the eastern United States, Europe, and East Asia during winters when the polar vortex is weak.
FIGURE 18 Seasonal trends of (left) Z500 phase speeds and (right) v250 phase speeds. All trends are for the North America and North Atlantic regions, and averages are taken between 30 N and 70 N. Closed circles denote trends that are statistically different from 0 at 95% confidence. Blue: autumn (OND); red: summer (JAS); black: winter (JFM); green: spring (AMJ). SOURCE: Barnes, 2013
FIGURE 19 Upper-air circulation anomalies in Chicago during 10 coldest days (upper left), 10 hottest days (upper right), 10 wettest days (lower left), and 10 driest summers (lower right). Note the location of the trough (purple) during the coldest days and the location of the ridge (red) during the hottest days. SOURCE: Jennifer Francis.
FIGURE 20 The climatological 850 hPa geopotential height field for December from 1968 to 1996. The map is indicative of normal early winter atmospheric conditions. Note the low geopotential heights of constant pressure surfaces over the Arctic (purples). SOURCE: http://www.arctic.noaa.gov
Overland and Wang (2 10) find evidence that loss of sea ice and the consequent changes in regional heat and other energy fluxes can weaken the polar vortex. Continued loss of snow and sea ice adds additional heat to the atmosphere, increasing the chance of a breakdown of the polar vortex through the thermal wind mechanism. In the Pacific Arctic, reanalysis shows 2005-2011 autumn temperature anomalies reaching the middle troposphere, which supports the fact that sea ice loss impacts the larger atmospheric climate (Figure 21).
Some of the observational evidence presented supports hypotheses for linking Arctic amplification and mid-latitude weather patterns. Other evidence does not. Observations can help decipher whether there is a trend, but some participants noted that there may be several mechanisms behind one trend. Furthermo e, finding observational trends in some of the mechanisms (wave amplitude, blocks, and wind speed) is largely dependent on how they are measured and defined.
Some participants noted that the short time series of observations of the recent significant loss of sea ice is also a limiting factor (less than 10 years) in finding trends to support the existence of these linkages and to robustly differentiate Arctic forcing of mid-latitude extremes from random events. Overland noted that the external forcing of Arctic sea ice loss and snow loss occurs on local scales and on short timescales, so, in his opinion, taking zonal averages and seasonal averages will smooth out some of the effects that we would be looking for in trying to identify trends.
Francis acknowledged more work is needed to assess the following:
FIGURE 21 2005-2011 autumn temperature anomalies reach the middle troposphere. SOURCE: NCEP/NOAA.
- Changing propagation of large-scale waves (speed and cause)
- Changing persistence of weather patterns
- Changing frequency of extremes
- Interactions among Arctic amplification and other large-scale influences (ENSO, Pacific Decadal Oscillation [PDO], etc.)
Another goal of the workshop was to explore theoretical and modeling work to test the mechanisms of proposed linkages. Many participants noted that the model spread9 can offer clues to mechanisms and that models can assess a large number of possible pathways to influence weather because they can simulate complex large-scale dynamics reasonably well. Some participants said that, given the natural variability of the atmosphere and the numerous external factors that influence it, modeling studies are useful for identifying the contributions of each of those external factors. The following variables were discussed in the co text of models that were forced with Arctic sea ice loss or Arctic warming:
- temperature gradients,
- upper-level zonal winds,
- large-scale wave amplitudes,
- blocking, and
- weather patterns and storm tracks.
9 The degree to which models agree with one another. It is often quantified from the across-model standard deviation.
Gudrun Magnusdottir, University of California, Irvine, focused her remarks on a recent modeling study (Peings and Magnusdottir, 2014). The researchers used NCAR’s CAM5 (an atmospheric global climate model) to understand the effect of recent extensive sea-ice loss years (2007-2012; denoted as 2010C) on the atmospheric circulation in winter and then compared those results to response to projected sea-ice change at the end of the century (2080-99; denoted as 2090C). The control experiment (CTL) is a 50-year simulation with sea ice concentration representative of the 1979-2000 period. One result from the study shows that the high-latitude surface warming is greater in 2090C than in 2010C and is significant up to the 500-hPa level. This leads to increased geopotential thickness of the lower troposphere (Figure 22), which was also detected in observations by Francis and Vavrus (2012). The thermal expansion of the Arctic troposphere reduces the meridional temperature and geopotential thickness gradients from the pole to mid-latitudes from 1000 to 500 hPa.
However, Barnes said that model simulations of change in zonal mean temperature in 25 CMIP510 models under RCP8.511 find that the upper-level tropospheric (250 hPa) temperature gradient is projected to increase in all seasons. Near-surface temperature gradient is projected to decrease in the cool months with Arctic amplification. The largest uncertainty (in degrees) is found during the months of December, January, and February on the polar surface.
Upper-Level Zonal Winds
Some modeling studies discussed show a decrease in upper-level zonal winds. Francis presented a CCSM4 run in a 4xCO2 scenario. The run shows that zonal winds are projected to decrease in all four seasons in the future (Figure 23).
Similarly, Magnusdottir noted that Peings and Magnusdottir (2014) found a weakening of the zonal westerly flow in the 2010C run that is due to the decrease in the meridional gradient of atmospheric geopotential thickness between the pole and mid-latitudes.
Large-Scale Wave Amplitudes
Screen and Magnusdottir noted that Peings and Magnusdottir (2014) found weak but significant increases in winter (DJF) mean wave amplitude under 2090 sea ice forcing. However, there was no significant wave amplitude change under 2010 sea ice forcing (Figure 24).
10 In 2008, the World Climate Research Programme (WCRP) Working Group on Coupled Modelling (WGCM), with input from the International Geosphere-Biosphere Programme (IGBP) AIMES project, agreed to promote a new set of coordinated climate model experiments. These experiments comprise the fifth phase of the Coupled Model Intercomparison Project (CMIP5). CMIP5 will provide a multimodel context for (1) assessing the mechanisms responsible for model differences in poorly understood feedbacks associated with the carbon cycle and with clouds, (2) examining climate “predictability” and exploring the ability of models to predict climate on decadal time scales, and, more generally, (3) determining why similarly forced models produce a range of responses. (http://cmip-pcmdi.llnl.gov/cmip5/)
11 Representative Concentration Pathways (RCP) are GHG concentration trajectories used for climate modeling and research. The numbers associated with RCPs (RCP8.5, RCP6, RCP4.5, and RCP2.6) refer to radiative forcings (global energy imbalances), measured in watts per square meter, by the year 2100. The four RCPs include one mitigation scenario leading to a very low forcing level (RCP2.6), two medium stabilization scenarios (RCP4.5 and RCP6) and one very high baseline emission scenarios (RCP8.5; van Vuuren et al., 2011).
FIGURE 22 Increased geopotential thickness of the lower troposphere due to sea ice loss. (a) 2010C zonal mean DJF temperature response (K). (b) 2010C-CTL response of the DJF atmospheric geopotential thickness between 1000 and 500 hPa; (c) 2090C zonal mean DJF temperature response. Contour interval 0.5 K; light (dark) shading indicates the 90 percent (95 percent) significance level. (d) 2090C-CTL response of the DJF atmospheric geopotential thickness between 1000 and 500 hPa. SOURCE: Peings and Magnusdottir, 2014.
FIGURE 23 Percentage change in 500hPa zonal winds in 4xCO2 run of CCSM4. Black (summer), green (spring), yellow (autumn), blue (winter). SOURCE: Jennifer Francis.
FIGURE 24 A weak but statistically significant increase in wave amplitude for 2090C. Top: Range of the 5400 m isoline of height o the 500 hPa surface for one day of February in the control. Bottom: Distribution of the 5400 m isopleth daily wave amplitude change in winter (DJF). White (control), light grey (2010C), dark grey (2090C). Red diamonds show the mean of the distribution, and asterisks indicate the significance level of the change of the mean in 2010C and 2090C, compared to the control value. SOURCE: Peings and Magnusdottir, 2014.
FIGURE 25 RCP8.5 minus historical mul imodel mean. Left: Seasonal cycle of the Northern Hemisphere (NH) blocking frequency as a function of longitude. Units are days per month. Right: Climatology of NH annual-mean blocking frequency. Units are number of blocked days per year. SOURCE: Dunn-Sigouin and Son (2013).
Barnes noted that some modeling studies find a decrease in high-latitude blocking events over the oceans, but an increase over Asia. The decreases are largest over the oceans in autumn and early winter. A couple of modeling studies find that decreases in blocking events are linked to changes in the jet stream (i.e. Barnes et al., 2012; de Vries et al., 2013). She noted that similar responses are found using other blocking identifying schemes (de Vries et al., 2013; Masato et al., 2013). Dunn-Sigouin and Son (2013) found RCP 8.5 integrations compared to historical12 integrations, showing significant decreases in blocking frequency over both the North Pacific and North Atlantic regions, with slight increasing blocki g frequency over western Russia (Figure 25).
Weather Patterns and Storm Tracks
Workshop participants discussed several modeling studies on changing weather patterns and storm tracks. Francis presented output from a CCSM4 RCP8.5 scenario that projected an inc eased ridging over continents associated with higher surface air temperatures. She noted that the ridging over the North Atlantic cools Western Europe. She also presented modeling evidence that projects that light and easterly winds will become more common at the expense of west winds for the 2090s. This will result in higher surface air temperatures in the mid-latitudes, which makes extremes more likely.
Tim Woollings, University of Oxford, discussed a set of experiments with one atmospheric model (HadGAM2) to investigate sources of spread in CMIP model projections of the Atlantic storm track response to climate change. This work focused on the late 21st century and showed that uncertainty in the magnitude of sea ice retreat is a large source of uncertainty in predicting the response of the storm track.
12 Historical runs are 20th century climate integrations with all observed climate forcings.
FIGURE 26 In the Northern Hemisphere (NH), the storm track is weakest during summer (JJA). Projected changes from 1980 to 1999 to 2081 to 2100 in zonal mean pressure-level variance statistics by CMIP5 multimodel ensemble based on RCP8.5 scenario, as a function of latitude and pressure. (a) DJF, (b) JJA, (c) MAM, (d) SON. Black contours indicate model climatology. Red and blue contours indicate projected changes. Shadings indicate regions over which 80 percent (light) or 100 percent (dark) of the models agree on the sign of the change. SOURCE: Chang et al., 2012.
Barnes presented a model study that projects that winter Northern Hemisphere upper-level storm tracks will weaken on equatorward flank by 2100 under RCP8.5; the storm track is weakest during summer (JJA) in the Northern Hemisphere (Figure 26; Chang et al., 2012).
Magnusdottir said that current Arctic sea-ice conditions (2010C) favor more intense cold extremes over mid-latit des (mostly confined to the Asian sector; Peings and Magnusdottir 2014). With stronger sea-ice forcing (2090C), the intensity of cold extremes decreases everywhere north of 45 N because of the extension of the Arctic warm anomaly over northern continents. In 2090C (as in 2010C) cold extremes are more intense south of 45 N. Despite far stronger forcing in 2090C compared to 2010C, the intensity of cold extremes does not change significantly. Magnusdottir noted that this implies that there is a nonlinear relationship between sea-ice retreat and mid-latitude temperature. Liu et al. (2012) also found that a reduction in Arctic sea ice leads to more frequent cold weather outbreaks in the northern mid-latitudes.
Several additional modeling studies were briefly presented which highlight possible regional climate impacts when forced with a reduction in Arctic sea ice (Box 3).
Highlights of Other Modeling Studies Showing Regional Effects
Cold Winters in Northern Europe
Yang and Christensen (2012) found that cold early winter weather in northern Europe is associated with 20-40 percent reduction in Barents-Kara sea ice. Petoukhov and Semenov (2010) found that sea ice loss triples the probability of cold winter over large areas of the mid-latitudes including Europe.
Wet European Summers
There is modeling evidence that sea ice loss may lead to a southward shift of the summer jet stream over Europe, resulting in increased summer precipitation in the United Kingdom (Screen, 2013). Screen noted that the simulated increase is small and closely resembles the spatial pattern of precipitation anomalies in recent summers.
East Asian Monsoon
Wu et al. (2013) found that composites of high ice minus low ice lead to dry air over western Europe and far-east Asia and increased precipitation over west Asia.
Winter Mediterranean Rain
Grassi et al. (2013) utilized simulations that project that sea ice reduction will lead to stronger and more frequent extreme cold events over continental Europe and extreme precipitation events over the entire Mediterranean Basin. In particular, simulations suggest an increased risk of winter flooding in southern Italy, Greece, and the Iberian Peninsula.
Changes in Energy Transport Due to Arctic Amplification
Frierson discussed how poleward energy transport into the Arctic might change in the future (because of global warming). He noted that global circulation models indicate that more Arctic amplification is associated with less heat transport into the Arctic (Hwang et al., 2011). He presented modeling studies that show that energy actually transports toward the mid-latitudes (rather than toward the poles as we might expect) under Arctic amplification. He believes that this anti-correlation is due to the heat transport responding to temperature gradients. Arctic amplification causes weaker temperature gradients which leads to less dry static energy transport into the Arctic. More gradient results in more transport (i.e., transport is diffusive). This “diffusive framework” implies that Arctic warming should spread to lower latitudes (i.e., warmer Arctic leads to warmer mid-latitudes). Models with more Arctic warming have anomalous dry static energy transport southward, back toward the mid-latitudes. This occurs relatively independently of the storm track amplitude and location change within the model.
Arctic Sea Ice Predictability
Some workshop participants noted that Arctic sea ice predictions could eventually be used to assess the risk of temperature and precipitation anomalies and extreme weather events over northern continents. In her remarks, Holland discussed the potential to predict Arctic sea ice on seasonal timescales. She noted that seasonal sea ice prediction is an initial value problem (i.e., the predictability characteristics are largely dependent on the initial conditions). Potential predictability of sea ice decreases during spring and then is regained
during summer and the following winter. Perfect model studies13 consistently show that ocean heat content (temperature) leads to predictability in winter sea ice, in all cases. Summer sea ice predictability is associated with memory that resides in the ice itself—ice thickness. Therefore, there is reduced summer sea ice predictability in thin ice conditions. Model studies suggest limited predictability in sea ice area on 1- to 2-year timescales.
Holland said that realizing even a fraction of the potential predictability of sea ice requires an “adequate” observational network and “adequate” (and compatible) models. However, it is currently unclear what “adequate” means. We do not even know the spatial coverage and variables that are required in an “adequate” observational network. What model biases impact predictability, and what model complexity is required? However, Holland noted that some studies have shown some predictive skill (e.g. Sigmond et al., 2013; Wang et al., 2013).
Holland argued that the inherent limits of sea ice predictions coupled with the need for system improvements to realize predictability should be acknowledged as an important limitation with respect to the possible influence of sea ice loss on extreme weather.
Model Limitations: CMIP Biases
As highlighted in this workshop, CMIP5 model projections can be a useful tool for studying linkages between Arctic amplification and mid-latitude circulation. However, as Barnes noted in her remarks, they contain biases that should be considered. For example, models tend to place the Northern Hemisphere jet too far equatorward (no model is poleward of the observed North Atlantic jet in the annual-mean). Barnes also said that models tend to overestimate the seasonal cycle in the north Atlantic and storm track intensity is too weak (Zappa et al., 2014). It is also well-known that models underestimate observed high-latitude blocking frequencies. These underestimates are largest in the cool seasons. In general, increased horizontal and vertical resolution improves model blocking distributions (Anstey et al., 2013; Matsueda et al., 2009). Blocking biases in global circulation models can be reduced by correcting for mean circulation biases, but this does not remove all errors, especially at high latitudes, said Barnes.
The modeling studies discussed represent a fraction of the work that has been done to test possible mechanisms of the proposed linkages between Arctic amplification and mid-latitude weather. Because the modeling studies are largely divergent, many participants believe research in this area should continue. In particular, many participants said that the modeling community should organize careful model intercomparisons and model sensitivity studies with similar boundary and/or initial conditions to allow for a more systematic review and comparison of the results. Peings and Magnusdottir (2014) provide a detailed comparison of some of the recent modeling work.
Although the workshop presentations highlighted several model limitations and biases, many participants asserted that models are the best tools available to complement observational studies, and to test robustness of the proposed mechanisms.
13 A perfect model study is an idealized exercise where an ensemble of forecasts from a single model are compared to one another. The ensemble is typically initialized from some point in time from the same model from an earlier run. The perfect model exercise is not initialized with observations or any attempt to assimilate observations into the model, and hence the initial state is a state of the model, not nature. The evaluation of the forecast therefore avoids any inaccuracy in knowledge of the observed initial state, and it avoids model errors in reproducing nature.
It was clear from the discussion that research into the linkages is still emerging and will continue in the future. Screen proposed using an “ACID test” as a framework for assessing confidence in proposed linkages: Attribution, Corroboration, Informed by mechanistic understanding, Detection. The first step is to ask whether the proposed link is attributable to sea ice loss (or another forcing). It is important to note that correlation does not necessarily imply causation. Screen stressed that it is difficult (if not impossible) to identify cause and effect in the real world. Carefully designed simulations are key for determining attribution (e.g., decreased mid-tropospheric north-south temperature gradient cannot be attributed to sea ice loss alone). It is also important to differentiate between sea ice loss and Arctic amplification (which has multiple causes). Also, observations do not tell us much about cause and effect, but models can.
The next step is to determine if the proposed link is corroborated by multiple lines of evidence. Is there agreement among multiple studies? Is there agreement between models? Different models with the same forcing can actually give opposite responses. All of the different factors that may affect the model outcome must be considered. Is there agreement among metrics?
Third, is the proposed linkage informed by mechanistic understanding? Confidence in the existence of a particular linkage is clearly strengthened by an in-depth understanding of the mechanisms. Screen acknowledged that much work is left to be done in this area. Do we understand the mechanisms driving the linkages?
The final step in Screen’s ACID test is to ask whether the link is detectable in the real world? Is the signal large enough to be detected over intrinsic/natural variability? Is the signal strong enough to detect over other forcings? Some types of responses are easier to detect than others (see Figure 4).
Screen believes that few, if any, of the proposed linkages pass the ACID test and that caution is needed in linking sea ice loss to recent events and trends. More research is required to make more confident statements. Other participants noted that the ACID test approach is sound, but, given the limitations of available information, there are inherent limitations to the analyses that can be conducted.
Workshop participants divided into three breakout groups to discuss future needs and opportunities in (a) observations, (b) models, and (c) a big picture context. A summary of the discussions in each of these three breakout groups follows.
One theme that emerged was the need to more effectively utilize existing observational capabilities. For example, there was a sense that research aircraft surveys targeted to key seasons and processes are underutilized. Participants also noted that a vast amount of data is never analyzed. Department of Defense products such as Visible and Infrared Scanner (VIRS) and National Ice Center (NIC) operational products could be useful and are not saved unless there is a perceived need for them.
Furthermore, making observations from data available more quickly would allow researchers enough time to analyze them.
Individual participants identified other specific observational and data needs, including
- quantifying uncertainties in data products to clarify trends (e.g., sea ice concentration),
- reprocessing historical snow products,
- utilizing a more robust analysis of hourly temperature data to accurately represent the diurnal cycle in the Arctic,
- deploying field programs that target extreme storm events (especially moisture measurements to clarify role of latent heating),
- obtaining measurements that target:
o clouds and surface fluxes (energy/radiation, moisture, momentum), especially over an increasingly open Arctic, from Arctic buoys (especially under high-wind conditions) and autonomous remote surface observations;
o atmosphere in the Arctic to better understand vertical temperature structure, stratification (e.g. using dropsondes from unmanned aircraft) and changes in cyclone/storm tracks; and
o the Arctic Ocean to improve seasonal predictions, and
- deploying process driven studies, for example, to understand the processes that drive sea ice.
Another theme that emerged from the workshop discussion was the importance of model sensitivity studies and model inter-comparison studies to provide clues on the mechanisms of potential Arctic linkages, as well as for detection and attribution. These studies require models that are consistently developed with similar boundary conditions (e.g., sea ice extent and thickness).
Individual workshop participants identified other needs related to modeling:
- Conducting idealized modeling studies to:
o understand how Arctic amplification and the atmospheric circulation response are coupled (e.g. baroclinic vs. barotropic, seasonality);
o understand future wave propagation with respect to Arctic change and the possible NAO-type response to sea ice loss; and
o capture seasonality and longitudinal variations (by trying perturbed physics runs, and not just fixed boundary conditions).
- Conducting attribution studies to understand the proportion of Arctic amplification that is due to sea ice vs. the proportion that is due to processes outside of the Arctic.
- Identifying the resolution that is required to correctly model fluxes (including horizontal and vertical atmosphere and boundary conditions).
- Applying the sophisticated methods and diagnostics that have been used in the meteorological community (e.g., see John Gyakum’s presentation) to this issue.
- Developing higher resolution models to study processes in the planetary boundary layer (below 850 hPa or roughly 1,500 m above sea level).
- Developing convection-permitting global climate models that resolve to better simulate vertical wind shear.
Big Picture Context
The third breakout group discussed the key points that emerged from the workshop discussions. Individuals in the group identified the following key messages:
- Participants still disagree about the influence of recent Arctic warming on mid-latitude weather patterns.
- Research on the Arctic influence on mid-latitudes is at an early stage compared to influences such as tropical, mid-latitude SST, or the stratosphere. The recent proposed influences are best viewed as hypotheses.
- It is important to retain a global perspective by considering the influence of the Arctic in the context of forcing from other components of the climate system, such as the tropics and the stratosphere. Changes in ocean circulation may be particularly important in Arctic change and its influence on the mid-latitudes. A key limitation is lack of knowledge of the Arctic Ocean energy budget.
- Attention has so far focused on changes in the meridional temperature gradient. Some consideration could be given to other mechanisms, such as zonal temperature gradients, static stability and moisture changes. The possibility of distinct regional responses could also be considered (e.g., east and west Atlantic).
- Application of more sophisticated methods of analysis to determine whether mid-latitude wave activity is changing would be useful. A long history of dynamical analysis of wave activity can be exploited in this field. Additionally, the link between Arctic warming and weakened westerlies is not clear, because of transient eddy feedbacks, which often act to shift the westerlies instead.
- More effort should be directed to understanding the synoptic aspects of possible linkages, considering the variety of synoptic weather systems active in the region, and to bridging the gap between case study analysis and filtered variance approaches. This topic could be further developed through collaboration between weather and climate scientists.
- Blocking is important for weather impacts and was a recurring feature of the workshop discussion, yet it is often interpreted differently by different people or analyses. Improving our understanding of the range of blocking systems and how they may be influenced by forcing is critical.
- It is possible that recent Arctic changes have pushed the atmosphere into a new state with different variability. The strong Arctic forcing has emerged only in the past few years, and development of new methods and approaches may be required to test or account for it. For example, it might be useful to adopt a probabilistic approach to determine whether there are changes in the probability of extreme events.
- Understanding of natural variability should be improved to determine whether changes are taking place. Reanalyses of the atmosphere are generally trustworthy for synoptic and large-scale features, so relatively long records are available. Work to extend sea ice records back in time is also useful. Ocean observations are seriously limited.
- Combining observations with a hierarchy of models is key to making progress on this issue. The detection and attribution framework provides one formal approach.
- Arctic influences could provide additional sources of skill for seasonal forecasting through model improvements or statistical methods.
- A recurring issue is how to respond to the public when asked questions such as “I heard that sea ice loss caused the wet summer, why don’t you include that in your forecasts?”
- A large, policy-relevant research program on these issues would bring the subject to maturity in the same way that has occurred for similar subjects (e.g., stratospheric, ocean influences). The program could focus on specific topics, for example, identifying something we are trying to explain, or by being provocative and picking on the weak links of chains of arguments.
- Although many challenges exist, several new opportunities will make significant progress in this area possible in the near future. These include (a) improvements in atmospheric models, some of which can now simulate blocking well, (b) similar improvements in ocean and sea ice models, (c) longer records back into the past of some sea ice quantities and atmospheric circulation, (d) emergence of reasonably strong sea ice forcing in recent years, and (e) strong interest in the topic from both public and governments.
Overall, many participants thought that, given the far-reaching societal impacts of changes in our climate system and the speed at which the Arctic sea ice cover is disappearing, it is imperative to advance understanding of this issue and to assess what it may mean for the frequency, severity, and persistence of severe weather events in the coming years. The workshop discussions made it clear that progress on this front requires more observational work and modeling experiments to place Arctic linkages within a larger context of the other factors that affect mid-latitude weather.
This page intentionally left blank.