1. Introduction
An ostensibly large number of record-breaking climate events have affected the Arctic region in recent years, gaining widespread scientific and media coverage (e.g., Perovich et al. 2008; Wormbs 2013; Moore 2016; Kim et al. 2017). The most notable have been historic minima in sea ice extent and multiyear ice during both the summer and winter seasons (e.g., Comiso 2006; Stroeve et al. 2008; Maslanik et al. 2011) and unprecedented warm wintertime temperatures (e.g., Cullather et al. 2016; Kim et al. 2017).
The temperature extremes have been linked to a number of drivers, ranging from perturbations in the polar vortex (Moore 2016) to tropically forced planetary waves [the so-called tropically excited Arctic warming mechanism (TEAM); Lee et al. 2011a,b; Lee 2012; Flournoy et al. 2016] and the constructive interference between stationary waves and transient eddies (Baggett and Lee 2015; Goss et al. 2016; Baggett et al. 2016). A common feature of these mechanisms is that they typically lead to a more meridionally oriented circulation, which favors the intrusion of midlatitude air masses into the Arctic region. A number of recent studies have highlighted that these intrusions result in very discontinuous meridional moisture fluxes into the Arctic region, with a small number of extreme events effectively setting the net seasonal transport value and resulting in significant positive temperature anomalies (Woods et al. 2013; Liu and Barnes 2015; Woods and Caballero 2016). The sea ice loss is closely associated to these moisture intrusions, which lead to a downward infrared (IR) radiation forcing (D.-S. R. Park et al. 2015; H.-S. Park et al. 2015a,b). Furthermore, years with the lowest September sea ice minima are characterized by an enhanced springtime meridional transport of moist air masses into the high latitudes (Kapsch et al. 2013).
At the same time, the Arctic warming and below-average sea ice cover in autumn and winter have been linked to cold winters in the midlatitudes, especially over Eurasia [the so-called warm Arctic–cold Eurasia (WACE) pattern]. A number of authors have ascribed this pattern to a reduction in midlatitude westerlies (e.g., Honda et al. 2009; Tang et al. 2013), suppressed eastward cyclone propagation owing to reduced sea surface temperature gradients (and hence, baroclinicity) over the Barents Sea (Inoue et al. 2012), and a transition to a more blocked midlatitude flow (Mori et al. 2014; Walsh 2014; Luo et al. 2016), which in turn favors a northerly flow of cold air over eastern Siberia and severe temperatures (Kug et al. 2015). However, no conclusive evidence has been reached regarding the importance of specific mechanisms or the role of decreased sea ice cover (Screen and Simmonds 2013; Cohen et al. 2014; Barnes and Screen 2015; Sun et al. 2016; McCusker et al. 2016; Seviour 2017; Garfinkel et al. 2017).
There is, therefore, a broad set of dynamic and thermodynamic interactions between the middle and high latitudes, leading to temperature extremes in both regions. A large part of the vast literature on the topic has focused on the drivers and consequences of sea ice loss and on wintertime temperature anomalies on monthly or longer time scales. Comparatively less attention has been devoted to winter temperature extremes in the high Arctic on synoptic time scales. Here, we specifically aim to address the following knowledge gaps:
-
While numerous warming mechanisms relevant to synoptic time scales have been discussed (e.g., Woods et al. 2013; H.-S. Park et al. 2015b; Woods and Caballero 2016; Baggett et al. 2016; Graversen and Burtu 2016), a systematic characterization of extreme warm spells is largely lacking.
-
Although some attention has been given to cold extremes occurring over the continental subarctic regions (Yu et al. 2017), cold extremes in the high Arctic have mostly been overlooked.
-
The variability in meridional moisture transport into the Arctic on seasonal scales has been linked to high-latitude cyclone activity (Sorteberg and Walsh 2008; Sepp and Jaagus 2011; Kim et al. 2017). However, the link between extreme moisture intrusions associated with positive temperature extremes and cyclone activity still needs to be investigated systematically.
2. Data and methods
a. Datasets
The analysis is primarily based on the European Centre for Medium-Range Weather Forecasts interim reanalysis (Dee et al. 2011) over the period January 1979–December 2016. This product outperforms other reanalyses in the Arctic region (Jakobson et al. 2012; Lindsay et al. 2014). We use data with a 6-hourly temporal resolution and a horizontal resolution of 1° on pressure levels and 0.75° for surface variables.
Because the study of extreme events can be severely limited by the length of the dataset, we verify the robustness of our conclusions using the longer NCEP–NCAR reanalysis (Kalnay et al. 1996), from which we select the 1950–2016 period. The data have a 6-hourly temporal resolution and a horizontal resolution of 2.5°. The daily North Atlantic Oscillation (NAO) and Arctic Oscillation (AO) indices we use are taken from NCEP’s Climate Prediction Center. We standardize the indices to have zero mean and unit standard deviation over the extended winter season and time period analyzed here.
All the analysis is based on an extended winter [November–March (NDJFM)] season. Statistical significance is evaluated using both a Monte Carlo random sampling procedure and a sign test. The latter is applied to the composite maps and identifies at each grid point the fraction of the composite members that have the same sign as the composite. Unless otherwise stated, all composite maps only show anomalies exceeding the 5% one-sided significance bounds, obtained by random sampling. Regions where more than two-thirds of the composite members agree on sign are cross-hatched, except in Fig. S5 of the supplemental material, where the cross-hatching shows anomalies exceeding two standard deviations of the local (single grid point) anomaly distribution. We note that under the assumption of a binomial distribution with 50 members and equal chances of positive and negative outcomes, the two-thirds threshold exceeds the 1% significance level.
b. Temperature extremes
Warm and cold spells are defined over a high Arctic domain, covering the polar cap north of 80°N. We choose this relatively narrow domain to focus specifically on events characterized by a deep penetration of midlatitude air masses into the Arctic basin, as opposed to events that lead to intense warming or cooling over the Siberian shelf seas or the Nordic seas but perhaps weaker anomalies over the pole. A brief analysis of events selected over a cap north of 70°N is presented in the supplemental material (see Figs. S2–S4 in the supplemental material). Two-meter air temperature anomalies are computed as deviations from the daily climatology. Because the Arctic region has experienced a rapid warming trend over the last decades (Cohen et al. 2014), the climatology is computed using a 9-yr running window. For example, the climatological value for 3 January 2000 is the mean of every 3 January between 1996 and 2004. Similarly, the climatological value for 3 January 2001 is the mean of every 3 January between 1997 and 2005 and so on. This procedure ensures a smooth variation of the seasonal cycle and a relatively uniform distribution of extreme events across our analysis period. If a simple daily climatology were computed over the full dataset, almost all warm spells and virtually no cold spells would fall in the last decade (cf. Fig. 1 with Fig. S1 in the supplemental material). This is consistent with observed decreasing trend in cold Arctic extremes (Matthes et al. 2015). Similarly, a simple linear trend removal was deemed ill-suited for our purposes because the Arctic warming in the past decades has been highly nonlinear (e.g., Johannessen et al. 2004). The daily climatology is then smoothed with a 21-day running mean, and the November–March seasons are selected. As a caveat, we note that at the beginning and end of the data series, the window is fixed and covers the first nine seasons (years 1–9 of the datasets) and the last nine seasons (last 9 years of the datasets), respectively. For an increasing temperature trend, this means that we will underestimate the frequency of cold events in the final years and of warm events in the initial years and, conversely, overestimate the frequency of cold events in the initial years and of warm events in the final years.
The temperature anomalies are area-weighted and averaged over the high Arctic domain defined above, thus providing a single temperature anomaly value per day. Days are then ranked according to their respective temperature anomalies. A 5-day running mean is applied to the anomaly time series to ensure we retain events that correspond to persistent deviations from the climatology. Only the warmest and/or coldest days within any one week are considered. For example, if the four warmest episodes in our time series were found to occur on days 201, 205, 47, and 798, ranked by decreasing temperature anomaly, only days 201, 47, and 798 would be retained. This ensures that we do not double-count extremes that might be detected over several consecutive days. The 50 warmest and coldest events are then retained for analysis. This number is chosen as a balance between the competing needs to select events that are unusual enough to warrant the definition of “extreme” while having a sufficiently large sample size to provide meaningful statistics. We further note that the running window climatology, relative to which the temperature anomalies are computed, implies that the events we select will be different from those identified in studies using a climatology defined over the full reanalysis period. All lags discussed in the paper are relative to the day of peak positive or negative temperature anomalies.
Our compositing approach is designed to limit aliasing effects for the peak temperature anomalies but, as a caveat, might provide an inaccurate view of the onset phase of the events. We have therefore produced composites centered on the warm and cold spells onset dates. These are defined as the closest days preceding the selected warm (cold) extremes when the domain-averaged temperature anomaly goes above the 90th (below the 10th) percentile of the November–March distribution. The average interval between the onset and peak of the spells is 4.04 days for the warm episodes and 4.7 days for the cold episodes. Consistently with this, the lag 0 and lag +5 day 2-m temperature onset composites (not shown) match very closely the lag -5 and lag 0 day peak composites, respectively. This suggests that the lag −5 day composites (shown in Figs. 2–4) provide a good representation of the onset phase of the warm/cold spells. Similarly, it points to the fact that the majority of the selected events share a relatively similar evolution.
c. Moisture intrusions
Intrusions of anomalously moist air from lower latitudes (termed moisture intrusions) exert a significant influence on the surface climate over large parts of the Arctic Ocean during winter (Woods et al. 2013; Woods and Caballero 2016). As part of this study, we explore the association between the presence (absence) of such events and the emergence of Arctic warm (cold) extremes. To achieve this, we employ the moisture injection detection algorithm of Woods et al. (2013). Moisture injections are defined as events in which the vertically integrated meridional moisture flux at 70°N maintains values in excess of 200 Tg day−1 (° lon)−1 over a contiguous zonal extent of at least 9° longitude for a minimum of 1.5 days. Our minimum flux threshold corresponds to the 88th percentile value of all vertically integrated northward moisture fluxes at 70°N during the analysis period. Implementation of this algorithm results in a dataset containing 1234 events. Following Woods et al. (2013) and Woods and Caballero (2016), we reject events that fail to penetrate sufficiently deep into the high Arctic and, therefore, cannot be associated with large precipitable water anomalies and surface radiative perturbations near the pole. Thus, for each moisture injection event detected, we compute an ensemble of 5-day forward and backward trajectories, initiated at 900 hPa at every grid point and time step for which the detection criteria were satisfied at 70°N. This allows us to track the propagation of a moist air mass through the Arctic, from its initial region of origin to entry at 70°N and eventual point of exit elsewhere along the Arctic boundary. The 900-hPa level is chosen, as this corresponds to the climatological northward moisture flux maximum during winter (Woods et al. 2013). Injections with less than 40% of their representative forward trajectories reaching 80°N are rejected from the dataset. Our final dataset consists of 844 moisture injection events—an average of approximately one per week.
We note that no detrending was applied to the moisture data prior to the detection of the moisture injections, and there has been a significant positive trend in the frequency of moisture injection events during NDJFM between 1980 and 2016 (Fig. S6). One may therefore wonder to what extent this has influenced the conclusions presented hereafter. We refer the reader to the supplemental material for a detailed discussion of this point (text section S2 and Figs. S6–S8 and S16 in the supplemental material), where it is shown that the study’s main conclusions are largely insensitive to the observed trend in moisture intrusion events.
d. Cyclone tracking
We base our analysis of cyclone tracks (section 5) on a cyclone track dataset obtained using the algorithm of Hanley and Caballero (2012). The method utilizes the ERA-Interim 6-hourly sea level pressure (SLP) field. Cyclones are identified as local minima in SLP. Each identified cyclone is assigned a “cyclone center” as the location of the local SLP minimum, and cyclone centers appearing in subsequent 6-hourly SLP snapshots are joined into cyclone tracks following the criteria specified in Hanley and Caballero (2012). Cyclogenesis and cyclolysis events are identified as the beginning and end points of each cyclone track, respectively. Cyclone tracks that pass over terrain higher than 1500 m above mean sea level (where the extrapolation involved in computing SLP is dubious) are eliminated, as are those with a lifespan of less than 24 h. The cyclone track dataset is part of the Intercomparison of Mid Latitude Storm Diagnostics (IMILAST) cyclone-tracking intercomparison dataset (Neu et al. 2013), which spans the period 1989–2009. This time period contains 30 and 28 of the warm and cold temperature extremes, respectively. Composite figures using the cyclone dataset are based on this shorter time period. This choice enables us to verify the robustness of our results to the choice of cyclone tracking algorithm by repeating the analysis using the combined statistics for the 15 cyclone tracking algorithms used in the IMILAST dataset. Very similar qualitative results are obtained (cf. Figs. 11–13 with Figs. S9, S13, and S15 in the supplemental material). As a caveat, ending the analysis in 2010 excludes recent years, during which unprecedented Arctic warm extremes have occurred.
3. Statistics of temperature extremes
The temperature extremes, selected as described in section 2b, are shown in Figs. 1a and 1b for both ERA-Interim (top 50 events 1979–2016) and NCEP–NCAR datasets (top 38 events 1950–78 and top 50 events 1979–2016, to have approximately the same average number of events per winter as in ERA-Interim), respectively. While there is not always a one-to-one correspondence, the two datasets show an overall good agreement in the timing of extremes during the period of overlap. However, we note that the NCEP–NCAR dataset tends to display larger area-weighted anomalies. This is not an artifact of our deseasonalization process, and, indeed, the same pattern is seen when the anomalies are computed as deviations from a simple daily climatology (Fig. S1). The larger magnitude anomalies—at least for positive excursions—may be related to the fact that the NCEP–NCAR reanalysis exhibits robust positive biases with respect to ERA-Interim in the mean northward moisture transport across 70°N during winter (Woods et al. 2017; see their Fig. S1i in the supplemental material). The majority of this bias is contributed by instantaneous fluxes of moisture greater than 200 Tg day−1 (° lon)−1 within the Atlantic sector, which are also those that exert significant control over the wintertime surface climate. The extreme events are distributed throughout the analysis period; however, both cold and warm extremes tend to cluster within periods of frequent occurrence followed by gaps of two or more seasons. This interannual–decadal modulation, although beyond the scope of the present paper, has previously been noted by Matthes et al. (2015). This suggests that for specific seasons, the temperature anomalies might display an enhanced persistence and that the 7-day separation between warm and cold spells we adopt here might not be sufficient to fully separate successive warm or cold episodes. The two datasets typically capture the same periods of frequent warm or cold episodes. We further note that it is not unusual for a given winter to display both warm and cold extremes.
Timing of the Arctic cold (blue) and warm (red) extremes in the (a) ERA-Interim and (b) NCEP–NCAR reanalyses. defined as described in section 2b. Over the period 1979–2016, the top 50 events are selected for both datasets. An additional 38 events are shown for the NCEP–NCAR reanalysis over the period 1950–79. (c) PDF of month of occurrence of warm (red) and cold (blue) Arctic extremes in ERA-Interim.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Timing of the Arctic cold (blue) and warm (red) extremes in the (a) ERA-Interim and (b) NCEP–NCAR reanalyses. defined as described in section 2b. Over the period 1979–2016, the top 50 events are selected for both datasets. An additional 38 events are shown for the NCEP–NCAR reanalysis over the period 1950–79. (c) PDF of month of occurrence of warm (red) and cold (blue) Arctic extremes in ERA-Interim.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Timing of the Arctic cold (blue) and warm (red) extremes in the (a) ERA-Interim and (b) NCEP–NCAR reanalyses. defined as described in section 2b. Over the period 1979–2016, the top 50 events are selected for both datasets. An additional 38 events are shown for the NCEP–NCAR reanalysis over the period 1950–79. (c) PDF of month of occurrence of warm (red) and cold (blue) Arctic extremes in ERA-Interim.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
The extremes are not evenly distributed within the extended winter season, but rather tend to favor the canonical winter months (December–February; see Fig. 1c). This is consistent with the higher variability seen in winter relative to summer, and we hypothesize that November and March might already be transition months from and to the warm season in this respect.
4. Composite structure of warm and cold spells
Here, we construct a climatology of both warm and cold wintertime extremes, based on the top 50 warmest and coldest events in the analysis period (see section 3). To give an idea of the variability of individual episodes relative to the composites, we present case studies of a warm and a cold extreme in Fig. S5.
The temperature footprint of the warm events is initially characterized by warm anomalies over the Greenland and Barents Seas and the Bering Strait (Fig. 2a). By lag 0, these have joined, resulting in very warm temperatures across the Arctic basin. The peak positive anomalies at lag 0 exceed 15 K (Fig. 2b). A large positive polar anomaly is still evident at lag +5 days (Fig. 2c). At the same time, twin cold anomalies emerge over North America and Eurasia. The Eurasian anomaly peaks at lag 0 but is still significant at lag +5 days. On the contrary, the North American cold region weakens as the warm polar extreme develops, yielding to a warm anomaly over the southeastern portion of the continent starting from lag 0. These temperature anomalies map closely onto the downward longwave radiation (DLWR) anomalies, as seen by a comparison of Figs. 2a–c with Figs. 2g–i.
Composite 2-m temperature anomalies (K) for (a)–(c) warm and (d)–(f) cold extremes at lags of (a),(d) −5, (b),(e) 0, and (c),(f) +5 days relative to peak temperature anomaly. Corresponding composite downward thermal radiation (W m−2) at lags of (g),(j) −5, (h),(k) 0, and (i),(l) +5 days relative to peak temperature anomaly. Only statistically significant anomalies are shown; cross-hatching shows regions of high sign agreement (see section 2a).
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Composite 2-m temperature anomalies (K) for (a)–(c) warm and (d)–(f) cold extremes at lags of (a),(d) −5, (b),(e) 0, and (c),(f) +5 days relative to peak temperature anomaly. Corresponding composite downward thermal radiation (W m−2) at lags of (g),(j) −5, (h),(k) 0, and (i),(l) +5 days relative to peak temperature anomaly. Only statistically significant anomalies are shown; cross-hatching shows regions of high sign agreement (see section 2a).
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Composite 2-m temperature anomalies (K) for (a)–(c) warm and (d)–(f) cold extremes at lags of (a),(d) −5, (b),(e) 0, and (c),(f) +5 days relative to peak temperature anomaly. Corresponding composite downward thermal radiation (W m−2) at lags of (g),(j) −5, (h),(k) 0, and (i),(l) +5 days relative to peak temperature anomaly. Only statistically significant anomalies are shown; cross-hatching shows regions of high sign agreement (see section 2a).
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
The SLP anomalies preceding the warm events are characterized by an NAO-like dipole over the Atlantic region (Fig. 3a). By lag 0, the pattern displays an almost continuous band of positive pressure anomalies in the mid-to-high latitudes and a deep low over Greenland, the North Pole, and northern Canada (Fig. 3b). This configuration creates a natural corridor for an anomalous southerly flow from the North Atlantic into the Arctic basin and an anomalous westerly advection of cold air over central-northern Siberia. At the same time, warm subtropical air is advected over the east coast of North America. This is particularly evident in the absolute SLP field (Fig. 3h). By lag +5 days, the polar low has mostly dispelled. Especially at lags −5 and 0 days, the 500-hPa geopotential height anomalies (Figs. 4a,b) are relatively closely aligned with the SLP anomalies. A similar large-scale pattern was also noted by H.-S. Park et al. (2015b), who were investigating strong DLWR events in the Arctic. The circulation anomalies revealed by our composite analysis therefore appear to be robust in the sense that they are evident for composites based on both warm spells (as in this study) and DLWR (as in H.-S. Park et al. 2015b).
Composite SLP anomalies (hPa) for (a)–(c) warm and (d)–(f) cold extremes at lags of (a),(d) −5, (b),(e) 0, and (c),(f) +5 days relative to peak temperature anomaly. Corresponding absolute SLP composites (hPa) at lags of (g),(j) −5, (h),(k) 0, and (i),(l) +5 days relative to peak temperature anomaly. In (a)–(f), only statistically significant anomalies are shown; cross-hatching shows regions of high sign agreement (see section 2a). [The green line in (b) corresponds to the transects shown in Figs. 5–8.]
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Composite SLP anomalies (hPa) for (a)–(c) warm and (d)–(f) cold extremes at lags of (a),(d) −5, (b),(e) 0, and (c),(f) +5 days relative to peak temperature anomaly. Corresponding absolute SLP composites (hPa) at lags of (g),(j) −5, (h),(k) 0, and (i),(l) +5 days relative to peak temperature anomaly. In (a)–(f), only statistically significant anomalies are shown; cross-hatching shows regions of high sign agreement (see section 2a). [The green line in (b) corresponds to the transects shown in Figs. 5–8.]
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Composite SLP anomalies (hPa) for (a)–(c) warm and (d)–(f) cold extremes at lags of (a),(d) −5, (b),(e) 0, and (c),(f) +5 days relative to peak temperature anomaly. Corresponding absolute SLP composites (hPa) at lags of (g),(j) −5, (h),(k) 0, and (i),(l) +5 days relative to peak temperature anomaly. In (a)–(f), only statistically significant anomalies are shown; cross-hatching shows regions of high sign agreement (see section 2a). [The green line in (b) corresponds to the transects shown in Figs. 5–8.]
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Composite 500-hPa geopotential height anomalies (m) for (a)–(c) warm and (d)–(f) cold extremes at lags of (a),(d) −5, (b),(e) 0, and (c),(f) +5 days relative to peak temperature anomaly. Corresponding composite 300-hPa wind (vectors) and wind speed (m s−1, color shading) anomalies at lags of (g),(j) −5, (h),(k) 0, and (i),(l) +5 days relative to peak temperature anomaly. Only statistically significant anomalies are shown; cross-hatching shows regions of high sign agreement (see section 2a).
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Composite 500-hPa geopotential height anomalies (m) for (a)–(c) warm and (d)–(f) cold extremes at lags of (a),(d) −5, (b),(e) 0, and (c),(f) +5 days relative to peak temperature anomaly. Corresponding composite 300-hPa wind (vectors) and wind speed (m s−1, color shading) anomalies at lags of (g),(j) −5, (h),(k) 0, and (i),(l) +5 days relative to peak temperature anomaly. Only statistically significant anomalies are shown; cross-hatching shows regions of high sign agreement (see section 2a).
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Composite 500-hPa geopotential height anomalies (m) for (a)–(c) warm and (d)–(f) cold extremes at lags of (a),(d) −5, (b),(e) 0, and (c),(f) +5 days relative to peak temperature anomaly. Corresponding composite 300-hPa wind (vectors) and wind speed (m s−1, color shading) anomalies at lags of (g),(j) −5, (h),(k) 0, and (i),(l) +5 days relative to peak temperature anomaly. Only statistically significant anomalies are shown; cross-hatching shows regions of high sign agreement (see section 2a).
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
The upper-level wind pattern closely reflects the geopotential height anomalies. At lag −5 days, the geopotential height dipole over the North Atlantic results in a northward shift of the midlatitude jet (Fig. 4g). By lag 0, the wind anomalies over the eastern North Atlantic are predominantly meridional (Fig. 4h) and are associated with a large-scale cyclonic circulation corresponding to the negative geopotential height anomaly centered over Greenland. At positive lags, the winds over the North Atlantic gradually return to a more zonal configuration (Fig. 4i).
In the case of the cold spells, below-average temperatures over the Arctic are evident throughout lags from −5 to +5 days; peak anomalies at lag 0 exceed −10 K (Figs. 2d–f). Significant anomalies outside the Arctic are largely limited to warm anomalies over Québec and Japan at negative lags and a cold anomaly over the southeastern United States that develops starting from lag 0. These anomalies, although significant, mostly have a relatively low sign agreement (hatching in Figs. 2–4; see section 2a), and, indeed, are often entirely absent from individual cold episodes (Figs. S5d–f). As for the warm spells, the temperature anomalies map closely onto the DLWR anomalies (cf. Figs. 2d–f with Figs. 2j–l).
The SLP pattern associated with the cold spells is initially relatively weak, with lows over the Arctic shelf seas and the Azores and highs over the East Asia and the Canadian Arctic (Fig. 3d). By lag 0, the only feature showing a high sign agreement is a dipole anomaly over the Arctic basin, with the negative pole dominating in both extent and intensity (Fig. 3e). Compared to the SLP field seen for the warm extremes, this configuration leads to a zonal intensification of the flow rather than favoring meridional advection (cf. Figs. 3h,k). This is consistent with the findings of Goss et al. (2016), who noted that during periods of weak stationary wave interference the hemispheric flow is anomalously zonal, and the Arctic is anomalously cold. Similarly, Lee (2012) showed that during El Niño years, when the large-scale circulation is more zonal than the climatological flow, the Arctic region is colder than average. The geopotential height anomalies display a significant shift relative to the SLP ones, and at lag 0, they display strong negative values over the Arctic basin (Fig. 4e). These anomalies are associated with an anomalous circumpolar westerly upper-level flow and a southward shift of the North Atlantic jet (Figs. 4j,k). This zonalized configuration is largely consistent with the thermal wind response expected for a colder Arctic. By lag +5 days, the configuration has shifted to a Pacific dipole, with a high over Alaska and a low centered around 50°N, evident in both the SLP and geopotential height fields (Figs. 3f,l and 4f). The anomalies around the Arctic become more meridionally oriented, especially over eastern Siberia and Canada, and the cold spell begins to tail off.
We note that the SLP anomaly dipole seen at lag 0 over the Arctic for the cold spells is almost exactly the opposite of what is seen for the warm extremes. As previously noted, the latter favors southerly advection from the North Atlantic into the Arctic basin, while the former impedes it. For a more detailed perspective on the vertical structure of these advective processes, we examine vertical cross sections of wind, temperature, and humidity along a transect across the Arctic basin. The transect line, shown in Fig. 3b, has one end in Scandinavia (60°N, 20°E,), passes over Svalbard and the North Pole, and ends in the Bering Strait region (60°N, 160°W). It is roughly aligned with the node in the Arctic SLP dipole anomaly characterizing both warm and cold extremes (Figs. 3b,e), and thus follows the direction of anomalous meridional flow. We present the transects with the Atlantic sector on the left and the Pacific on the right (i.e., viewing the cross section from Scandinavia).
During warm events (Figs. 5 and 6), the composites show a strong wind anomaly oriented from left to right along the transect (i.e., from the Atlantic sector toward the Pacific) with only a moderate vertical shear, consistent with that seen in Figs. 3a–c and 4a–c. The winds advect warm moist maritime air from the Barents Sea over the Arctic Ocean, inducing large temperature and humidity anomalies, with the largest amplitudes near the surface. At lag −5 days, most of the region north of 80°N (shown by the thick black line under each transect) is covered by a near-surface temperature inversion, with its top at around 900 hPa; as the event progresses, the inversion is strongly weakened or removed through most of the region, consistently with the bottom-amplified structure of the temperature anomaly. At lag −5 days, the potential temperature (white lines in Figs. 5a,c,e,g) shows a “cold dome” structure centered close to the pole. Wind vectors cross the isentropes in the left-hand portion of the dome, implying warm potential temperature advection, so that the temperature anomalies observed there in the subsequent days are largely advective in origin. By lag −2 days, the dome has shifted to the right, and the flow is mostly along the isentropes; at this point, advective forcing of the temperature anomaly has largely ceased, and the anomaly decays radiatively in the free troposphere, though it continues to be maintained near the surface. The along-isentropic airflow over the left-hand part of the dome implies lifting of moist low-level air, consistent with positive cloudiness anomalies there (Fig. 6). By lag +5 days, temperature and humidity anomalies have largely relaxed back to climatology, except for substantial anomalies that persist close to the surface.
Composite transects for warm extremes. The transect line is shown in Fig. 3b; Scandinavia is to the left and the Bering Strait to the right. Shading shows (a),(c),(e),(g) absolute temperature and (b),(d),(f),(h) temperature anomaly. Vectors show the component of the absolute wind in (a),(c),(e),(g) and anomalous wind in (b),(d),(f),(h) in the plane of the transect, with horizontal and vertical velocities scaled appropriately for the axes. White contours in (a),(c),(e),(g) show potential temperature (contour interval of 4 K).
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Composite transects for warm extremes. The transect line is shown in Fig. 3b; Scandinavia is to the left and the Bering Strait to the right. Shading shows (a),(c),(e),(g) absolute temperature and (b),(d),(f),(h) temperature anomaly. Vectors show the component of the absolute wind in (a),(c),(e),(g) and anomalous wind in (b),(d),(f),(h) in the plane of the transect, with horizontal and vertical velocities scaled appropriately for the axes. White contours in (a),(c),(e),(g) show potential temperature (contour interval of 4 K).
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Composite transects for warm extremes. The transect line is shown in Fig. 3b; Scandinavia is to the left and the Bering Strait to the right. Shading shows (a),(c),(e),(g) absolute temperature and (b),(d),(f),(h) temperature anomaly. Vectors show the component of the absolute wind in (a),(c),(e),(g) and anomalous wind in (b),(d),(f),(h) in the plane of the transect, with horizontal and vertical velocities scaled appropriately for the axes. White contours in (a),(c),(e),(g) show potential temperature (contour interval of 4 K).
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
As in Fig. 5, but with shading showing (left) specific humidity and (right) specific humidity anomaly. Magenta lines in (a),(c),(e),(g) show cloud fraction, contoured at 30% (thin), 50% (medium), and 70% (thick). Magenta lines in (b),(d),(f),(h) show cloud fraction anomaly, contoured at +20% (solid) and −20% (dashed).
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
As in Fig. 5, but with shading showing (left) specific humidity and (right) specific humidity anomaly. Magenta lines in (a),(c),(e),(g) show cloud fraction, contoured at 30% (thin), 50% (medium), and 70% (thick). Magenta lines in (b),(d),(f),(h) show cloud fraction anomaly, contoured at +20% (solid) and −20% (dashed).
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
As in Fig. 5, but with shading showing (left) specific humidity and (right) specific humidity anomaly. Magenta lines in (a),(c),(e),(g) show cloud fraction, contoured at 30% (thin), 50% (medium), and 70% (thick). Magenta lines in (b),(d),(f),(h) show cloud fraction anomaly, contoured at +20% (solid) and −20% (dashed).
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
During cold events (Figs. 7 and 8), the wind anomalies are oriented right to left along the transect, bringing cold, dry air from the Chukchi and Beaufort Seas toward the pole. At lag −5 days, the cold dome is centered somewhat to the right of the pole but is advected leftward, so that it is precisely over the pole by lag −2 days, where it remains up to lag +5 days. This displacement is associated with a strong negative temperature anomaly with maximum values near the surface, implying an intensification of the low-level temperature inversion. After lag −2 days, the flow is along the isentropes, so the cold anomaly over the polar cap is presumably maintained largely by radiative cooling after this time. Figure 8 shows widespread negative anomalies of humidity and cloudiness throughout the polar cap, which are consistent with the latter process.
As in Fig. 5, but for the cold extremes.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
As in Fig. 5, but for the cold extremes.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
As in Fig. 5, but for the cold extremes.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
As in Fig. 6, but for the cold extremes.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
As in Fig. 6, but for the cold extremes.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
As in Fig. 6, but for the cold extremes.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
5. The role of moisture intrusions and cyclones
a. Moisture intrusions
The above analysis suggests that the influx of warm, moist air from the Atlantic sector is a primary driver of the warm extremes, while below-average meridional advection favors radiative cooling and cold extremes. Here we examine the statistical relationship between warm and cold extremes and moisture intrusions as defined in section 2c.
The association is diagnosed as follows. First, all moisture injection events that existed for at least one time step over lags from −22 to +12 days, relative to the temperature extremes, are identified. Next, the total duration and number frequency of these injections is calculated for 5-day windows centered on each lag in the range from −20 to +10 days. This choice is motivated by the fact that on average, only one moisture injection event is found during any 5-day window, and this time scale also captures the typical advective time scale of the moisture from 70°N to the polar region (Woods and Caballero 2016; see also discussion below). For example, the number of injection events associated with lag −6 days, relative to a warm extreme occurring on day 20 of our dataset, will be the number of injections that existed for at least one time step over days 12–16 of the dataset, that is, lags from −8 to −4 days relative to the warm extreme. Figure 9 shows a lagged composite of these metrics centered on the previously discussed warm and cold extremes. As expected, anomalies in the frequency of moisture injection events at 70°N lead the temperature extremes, with the largest deviations from climatology occurring roughly 5–7 days earlier. This provides an indication of the time scale of the moisture advection from 70°N to the polar cap. The anomalies are generally larger for the warm extremes than for the cold extremes, consistent with the analysis in the previous sections that found that the large-scale pattern linked to the cold mode is generally more similar to the climatology than that of the warm extremes. We also note that the duration of the injection events has a more significant association with warm temperature extremes than the number frequency [cf. the relative anomalies at day −5 for duration (>100%) and number frequency (~75%)]. This is consistent with the fact that the positive temperature anomalies north of 80°N are advective in nature and, therefore, to some degree proportional to the time scale of the advection from lower latitudes. Persistence in the atmospheric circulation appears to be an important factor in the emergence of these extremes. Overall, the life cycle of moisture injections associated with warm and cold extremes takes place over a period of roughly 30 days, although the most significant association takes place over a window of roughly 10–15 days (Fig. 9). The statistics of moisture injections associated with warm and cold spells are summarized in Table 1. Probability distribution functions for the cumulative duration and number of injections during a pentad are determined from 10 000 sets of 50 randomly sampled dates in the NDJFM 1979–2016 range. The 1st and 99th percentile values of these metrics occur at approximately ±2.38σ from the mean, somewhat similar to the values for a Gaussian distribution (±2.56σ). We also note that there is a slight positive skew in the distribution of the duration metric. Any given 5-day window will on average overlap with 0.99 injection events, and the average cumulative duration is 3.16 days (Table 1).
Mean injection (a) event duration and (b) frequency integrated over 5-day windows, for all injection events that existed for at least one time step over lags from −22 to +12 days relative to the 50 warm (red) and cold (blue) temperature extremes. The x axis displays lags relative to the peak temperature anomaly; values are shown as a function of the central day of each window. Lower and upper dashed lines indicate the 1st and 99th percentile values of the two metrics, obtained from random sampling, as well as their means.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Mean injection (a) event duration and (b) frequency integrated over 5-day windows, for all injection events that existed for at least one time step over lags from −22 to +12 days relative to the 50 warm (red) and cold (blue) temperature extremes. The x axis displays lags relative to the peak temperature anomaly; values are shown as a function of the central day of each window. Lower and upper dashed lines indicate the 1st and 99th percentile values of the two metrics, obtained from random sampling, as well as their means.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Mean injection (a) event duration and (b) frequency integrated over 5-day windows, for all injection events that existed for at least one time step over lags from −22 to +12 days relative to the 50 warm (red) and cold (blue) temperature extremes. The x axis displays lags relative to the peak temperature anomaly; values are shown as a function of the central day of each window. Lower and upper dashed lines indicate the 1st and 99th percentile values of the two metrics, obtained from random sampling, as well as their means.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Statistics of moisture injection events and their association with warm and cold extremes. From left to right the columns under each heading indicate: 1) the average cumulative duration and number of injection events existing over 5-day periods for all dates in the 1979–2016 NDJFM range; 2) the average cumulative duration and number for the randomly sampled 50-event sets (see text); 3) the standard deviation of the random sample distribution; and the distribution’s 4) 1st and 5) 99th percentile values, respectively. Values for warm and cold extremes are shown for lag −5 days relative to peak temperature anomaly.
Motivated by the highly significant association between the moisture injection events and temperature extremes, we assess how the spatial structure of the moisture intrusion trajectories varies between warm and cold extremes. For reference, Fig. 10a shows the climatological density of intrusion centroid trajectories, averaged over all 844 injection events, with arrows indicating the preferred direction of intrusion trajectories. During moisture intrusion events, a cyclonic flow is apparent between Greenland and Svalbard, while an anticyclonic circulation appears over Siberia. This is consistent with the large-scale anomalies shown in Figs. 2a–c, 3a–c, and 3g–i. A similar pattern is apparent on the Pacific side of the basin, with a cyclonic circulation centered over eastern Siberia and an anticyclone centered over the North Slope of Alaska. This is consistent with previous work showing that blocking-like patterns are an important factor in episodes of extreme moisture transport into the Arctic (Woods et al. 2013; Liu and Barnes 2015; H.-S. Park et al. 2015b). The connection between cyclonic systems and temperature extremes will be examined in more detail in the next section. Again in agreement with the large-scale analysis of section 4, the Atlantic appears to be the dominant pathway for intrusions of moist air from lower latitudes into the high Arctic. Intrusion centroid trajectory frequency is at a maximum between the Greenland and Norwegian Seas, with an intrusion centroid trajectory being present overhead roughly once every day. Figure 10b shows the mean anomalies of intrusion centroid trajectory density (with respect to the climatology in Fig. 10a) for all moisture injection events occurring from 7 to 3 days before the warm extremes. This 5-day window is chosen for consistency with Fig. 9. Warm extremes are accompanied by a systematic increase in the amplitude of intrusion centroid trajectory density, as evidenced by the widespread positive anomalies. Interestingly, the pattern of positive anomalies is also rotated counterclockwise with respect to the climatological flow, such that the anomalies are mostly oriented meridionally in the region north of 70°N. Significant positive anomalies extend back into the North Atlantic, where the bulk of the moist air masses presumably originates. Warm extremes therefore appear to be characterized by anomalously persistent periods of moisture advection over the North Atlantic and into the Norwegian Sea, while simultaneously the large-scale circulation favors meridional advection through the Fram Strait, directly toward the pole, rather than along the climatological intrusion path across the Barents Sea (vectors in Fig. 10a). By separately counting the number of injection events occurring within the Atlantic (70°W–110°E) and Pacific (110°E–70°W) sectors, we determine that 49 out of the 50 warm events analyzed here are primarily associated with Atlantic intrusions. As expected, cold extremes systematically display negative intrusion density anomalies (Fig. 10c), and these again have lower absolute values than the positive anomalies associated with the warm extremes. The few moisture intrusions that do occur during cold extremes tend to have no preferred direction in their anomalous flow over the high Arctic.
Shading shows (a) climatological density of intrusion centroid trajectories per moisture injection event; (b) anomalies with respect to (a) for all injection events that existed for at least one time step between 7 and 3 days before the warm extremes; and (c) as in (b), but for the cold extremes. Trajectory densities are calculated by dividing the study region into 350 km × 350 km grid boxes and counting the number of 6-hourly centroid trajectory points contained within each grid box. Vectors in (a) indicate the typical direction in which the intrusion centroid trajectories move. They are computed as the unit tangent vector at each time step along each centroid trajectory; the tangent vectors falling within each grid box are then averaged. If the trajectories are isotropic, the vectors average to zero. Vectors in (b) and (c) show the vector differences between the warm and cold extremes, respectively, and the climatology in (a). Shading and vectors in (b) and (c) are shown only for anomalies significant at the 2% level. Dashed circles show the 70° and 80°N parallels.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Shading shows (a) climatological density of intrusion centroid trajectories per moisture injection event; (b) anomalies with respect to (a) for all injection events that existed for at least one time step between 7 and 3 days before the warm extremes; and (c) as in (b), but for the cold extremes. Trajectory densities are calculated by dividing the study region into 350 km × 350 km grid boxes and counting the number of 6-hourly centroid trajectory points contained within each grid box. Vectors in (a) indicate the typical direction in which the intrusion centroid trajectories move. They are computed as the unit tangent vector at each time step along each centroid trajectory; the tangent vectors falling within each grid box are then averaged. If the trajectories are isotropic, the vectors average to zero. Vectors in (b) and (c) show the vector differences between the warm and cold extremes, respectively, and the climatology in (a). Shading and vectors in (b) and (c) are shown only for anomalies significant at the 2% level. Dashed circles show the 70° and 80°N parallels.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Shading shows (a) climatological density of intrusion centroid trajectories per moisture injection event; (b) anomalies with respect to (a) for all injection events that existed for at least one time step between 7 and 3 days before the warm extremes; and (c) as in (b), but for the cold extremes. Trajectory densities are calculated by dividing the study region into 350 km × 350 km grid boxes and counting the number of 6-hourly centroid trajectory points contained within each grid box. Vectors in (a) indicate the typical direction in which the intrusion centroid trajectories move. They are computed as the unit tangent vector at each time step along each centroid trajectory; the tangent vectors falling within each grid box are then averaged. If the trajectories are isotropic, the vectors average to zero. Vectors in (b) and (c) show the vector differences between the warm and cold extremes, respectively, and the climatology in (a). Shading and vectors in (b) and (c) are shown only for anomalies significant at the 2% level. Dashed circles show the 70° and 80°N parallels.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
b. Cyclones
Having identified a highly significant link between surface temperature extremes in the high Arctic and moisture intrusions, we investigate whether these can further be linked to synoptic-scale cyclonic systems identified and tracked using the algorithm described in section 2d. Climatological NDJFM frequencies of cyclone centers, cyclogenesis, and cyclolysis obtained using this tracking algorithm are shown in Figs. 11a–c. In the North Atlantic, cyclone center frequency peaks between Greenland and Iceland, where cyclogenesis and cyclolysis also reach local maxima. There is a secondary cyclone frequency maximum in the Barents Sea south of Svalbard, with a local cyclogenesis maximum to its west. All these features agree well with those obtained using other cyclone-tracking methods (Hoskins and Hodges 2002; Neu et al. 2013).
Climatological frequencies of (a) cyclones, (b) cyclogenesis, and (c) cyclolysis for NDJFM based on the cyclone detection algorithm of Hanley and Caballero (2012). Mean anomalies over the 5 days prior to 30 warm extremes of (d) cyclone frequency, (e) cyclogenesis, and (f) cyclolysis are also shown. (g)–(i) As in (a)–(c), but for 28 cold extremes. The thick dashed contour in (b) shows the climatological NDJFM sea ice margin (15% sea ice concentration) in ERA-Interim. Thin dashed circles show the 70° and 80°N parallels and the 85°N parallel in (a). Frequencies are computed by counting the number of features within a radius of 564 km (equivalent to 5° latitude) of each point on a 3° × 3° spherical grid. The area enclosed by the 85°N parallel is approximately equal to the area of a circle with radius 564 km.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Climatological frequencies of (a) cyclones, (b) cyclogenesis, and (c) cyclolysis for NDJFM based on the cyclone detection algorithm of Hanley and Caballero (2012). Mean anomalies over the 5 days prior to 30 warm extremes of (d) cyclone frequency, (e) cyclogenesis, and (f) cyclolysis are also shown. (g)–(i) As in (a)–(c), but for 28 cold extremes. The thick dashed contour in (b) shows the climatological NDJFM sea ice margin (15% sea ice concentration) in ERA-Interim. Thin dashed circles show the 70° and 80°N parallels and the 85°N parallel in (a). Frequencies are computed by counting the number of features within a radius of 564 km (equivalent to 5° latitude) of each point on a 3° × 3° spherical grid. The area enclosed by the 85°N parallel is approximately equal to the area of a circle with radius 564 km.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Climatological frequencies of (a) cyclones, (b) cyclogenesis, and (c) cyclolysis for NDJFM based on the cyclone detection algorithm of Hanley and Caballero (2012). Mean anomalies over the 5 days prior to 30 warm extremes of (d) cyclone frequency, (e) cyclogenesis, and (f) cyclolysis are also shown. (g)–(i) As in (a)–(c), but for 28 cold extremes. The thick dashed contour in (b) shows the climatological NDJFM sea ice margin (15% sea ice concentration) in ERA-Interim. Thin dashed circles show the 70° and 80°N parallels and the 85°N parallel in (a). Frequencies are computed by counting the number of features within a radius of 564 km (equivalent to 5° latitude) of each point on a 3° × 3° spherical grid. The area enclosed by the 85°N parallel is approximately equal to the area of a circle with radius 564 km.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Figures 11d–i show composite anomalies of cyclone center, cyclogenesis, and cyclolysis frequencies averaged over the 5 days prior to Arctic temperature extremes. There is a shift in cyclone center frequency toward the east coast of Greenland, with a local maximum collocated with the cyclonic circulation noted in Fig. 10a and a deficit of cyclone activity in the Barents and Kara Seas. In addition, there is also a significant positive anomaly in the western Arctic basin, with local values reaching around 200% of the climatology. This situation, with cyclones simultaneously present off the east coast of Greenland and near the pole, is reminiscent of the 2015 warm event case study by Moore (2016). Given the similarity among the anomaly patterns in Figs. 3a,b and 11d, one may wonder to what extent the cyclone tracking algorithm—which identifies local minima in SLP and is not scale-aware—is simply picking up the surface signature of planetary-scale waves. Conversely, one cannot exclude the opposite case—that by selecting specific periods in which synoptic-scale cyclones happen to be clustered in particular regions and then compositing over many such periods (as done here), the synoptic-scale pressure minima end up being smoothed into larger-scale features. This issue cannot be solved without a more formal spectral decomposition of the variability into planetary- and synoptic-scale components, which we do not attempt in the present study. This, however, remains an interesting avenue for future research (see also section 7).
The pattern of cyclone frequency anomalies seen in Fig. 11d—with positive anomalies stretching almost continuously along the east and north coasts of Greenland—gives the impression that Atlantic cyclones are being deflected northward, propagating from the North Atlantic all the way into the high Arctic. However, Fig. 11e indicates anomalously high cyclogenesis in the high Arctic; the positive cyclone center frequency anomalies north of 80°N in Fig. 11d therefore have a large contribution from tracks originating within the polar cap itself. Furthermore, Fig. 11f shows large anomalous cyclolysis at around 80°N in the Fram Strait region. These results suggest that cyclones from the North Atlantic are indeed deflected northward during warm events but reach the end of their life cycle near the Fram Strait, while separate cyclonic anomalies are simultaneously forming to the north of Greenland. To confirm this picture, Fig. 12 shows cyclone center, cyclogenesis, and cyclolysis frequencies computed as in Figs. 11a–c, but selecting only those cyclone tracks that were present north of 80°N for at least one 6-h time step during the 5 days preceding warm extremes. Figure 12a shows a large maximum in cyclone center frequency northwest of Greenland—consistently with the anomaly pattern shown in Fig. 11d—but very small frequencies in the North Atlantic south of 80°N, implying that the selected cyclones spend most of their lifetimes within the Arctic basin. Moreover, Fig. 12b shows that these cyclones belong to tracks originating almost exclusively to the north of 80°N, with only a small fraction originating at lower latitudes along the eastern coast of Greenland and, subsequently, propagating into the region. We therefore conclude that although it is occasionally possible for North Atlantic cyclones to propagate all the way into the Arctic basin, the predominant case is one where warm anomalies are associated with separate North Atlantic and Arctic cyclones, the latter originating in situ.
Mean frequency over 30 warm extremes of (a) cyclones, (b) cyclogenesis, and (c) cyclolysis for the cyclones that existed north of 80°N for at least one time step during the 5 days prior to each warm extreme. Dashed circles show the 70° and 80°N parallels.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Mean frequency over 30 warm extremes of (a) cyclones, (b) cyclogenesis, and (c) cyclolysis for the cyclones that existed north of 80°N for at least one time step during the 5 days prior to each warm extreme. Dashed circles show the 70° and 80°N parallels.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Mean frequency over 30 warm extremes of (a) cyclones, (b) cyclogenesis, and (c) cyclolysis for the cyclones that existed north of 80°N for at least one time step during the 5 days prior to each warm extreme. Dashed circles show the 70° and 80°N parallels.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
For the cold extremes, a reversal of these patterns is generally observed. Cold extremes are favored by anomalously high cyclone frequencies in the eastern Arctic basin and weak negative frequency anomalies in the western basin (Fig. 11g). The anomalies are consistent with a conceptual picture whereby cold extremes are characterized by a large-scale flow that is predominantly zonal across the Nordic seas, leading to a buildup of cyclone counts and cyclolysis over the Barents and Kara Seas, with a relative deficit north of 80°N (Fig. 11i). Generally, the spatial patterns of the anomalies associated with cold extremes (Figs. 11g–i) are much closer to the respective climatological patterns than those of the anomalies preceding the warm extremes. To allow for a more immediate comparison with the results presented in section 4, we have repeated the analyses shown in Figs. 11 and 12 for 5-day periods centered on lags of −5, 0, and +5 days (Figs. S10–S12 and S14, respectively).
c. Relation between cyclones and intrusions during warm events
We have shown above that Arctic warm events are associated with two distinct sets of cyclones: one in the North Atlantic, to the east of Greenland, and another in the high Arctic. This suggests a conceptual picture whereby the atmospheric moisture contained in the moisture intrusion events is relayed into the Arctic via an interaction of several cyclonic systems centered at different latitudes. To better understand this interplay between cyclones and moisture intrusions, Fig. 13 shows cyclone-frequency anomalies averaged over 2-day segments between lags −10 and 0 days, relative to the warm extremes (instantaneous composites, though noisier, present the same qualitative features), together with concurrent intrusion trajectory density anomalies, calculated relative to the climatological density shown in Fig. 13f. During days from −10 to −4 (Figs. 13a–c), significant positive anomalies in the density of intrusion trajectories are present in the northern North Atlantic, collocated with the climatological maximum (Fig. 13f). The lead up to warm extremes appears to be characterized by an increase in the amplitude of the climatological pattern. After day −4, robust positive anomalies in cyclone frequency emerge in the region north of 80°N, while strong negative anomalies are apparent over the Kara Sea (Figs. 13d,e). The warm extremes therefore appear to be preceded by a buildup of intrusion trajectories—and presumably of warm, moist air—in the Norwegian Sea, which is then transported into the high Arctic by the cyclonic anomalies near the pole (which, as mentioned above, are mostly generated north of 80°N).
Composite anomalies of cyclone frequency (shading) and intrusion density (contours) for 30 warm extremes at lags (a) from −10 to −8, (b) from −8 to −6, (c) from −6 to −4, (d) from −4 to −2, and (e) from −2 to 0 days relative to peak warmth. For comparison, (f) climatological intrusion density plotted in the same units as in (a)–(e). Contours in (a)–(e) begin at 0.4 (thickest contour) and increase by 0.15. Dashed circles show the 70° and 80°N parallels.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Composite anomalies of cyclone frequency (shading) and intrusion density (contours) for 30 warm extremes at lags (a) from −10 to −8, (b) from −8 to −6, (c) from −6 to −4, (d) from −4 to −2, and (e) from −2 to 0 days relative to peak warmth. For comparison, (f) climatological intrusion density plotted in the same units as in (a)–(e). Contours in (a)–(e) begin at 0.4 (thickest contour) and increase by 0.15. Dashed circles show the 70° and 80°N parallels.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Composite anomalies of cyclone frequency (shading) and intrusion density (contours) for 30 warm extremes at lags (a) from −10 to −8, (b) from −8 to −6, (c) from −6 to −4, (d) from −4 to −2, and (e) from −2 to 0 days relative to peak warmth. For comparison, (f) climatological intrusion density plotted in the same units as in (a)–(e). Contours in (a)–(e) begin at 0.4 (thickest contour) and increase by 0.15. Dashed circles show the 70° and 80°N parallels.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
6. Relation to large-scale modes of variability and midlatitude features
The large-scale SLP composites discussed in section 4 display some features reminiscent of the canonical NAO and AO dipoles, with the warm spells composite indicating a positive projection on these modes and the cold spells composite suggestive of a weak negative projection (Figs. 2a–f). This is confirmed in Fig. 14a, which shows that warm spells indeed display a significant positive projection on both modes of variability, which peaks around lag −1 days and then switches to a negative projection as positive SLP anomalies start expanding across the North Atlantic region. On the contrary, the cold extremes initially display a weak negative projection, which subsequently shifts toward positive values (Fig. 14b). Unlike for the warm spells, where the NAO and AO indices are tightly coupled, the projections of the cold spells on the two modes differ significantly. This is due to the influence of the Pacific pole of the AO, which corresponds to an anomalous high at small negative lags and up to the peak of the cold spell but is then affected by a growing low pressure center over the eastern Pacific at positive time lags.
Projections of the (a) warm and (b) cold spells onto the NAO (red) and AO (black) indices. Projections of the (c) warm and (d) cold spells on a standard SHI (red) and the SHI_N (black). The dashed horizontal lines mark the one-sided 5% significance levels.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Projections of the (a) warm and (b) cold spells onto the NAO (red) and AO (black) indices. Projections of the (c) warm and (d) cold spells on a standard SHI (red) and the SHI_N (black). The dashed horizontal lines mark the one-sided 5% significance levels.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Projections of the (a) warm and (b) cold spells onto the NAO (red) and AO (black) indices. Projections of the (c) warm and (d) cold spells on a standard SHI (red) and the SHI_N (black). The dashed horizontal lines mark the one-sided 5% significance levels.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
Another notable feature of the SLP patterns associated with the Arctic temperature extremes is the strong footprint on northern Eurasia (Fig. 3). For the warm spells, this takes the form of a persistent anomalous high around Novaya Zemlya and the Barents and Kara Seas, which at lag 0 covers most of Eurasia and stretches across the North Atlantic. For the cold spells, the pattern is more localized, with a negative pressure anomaly stretching across the Siberian shelf seas and the Russian Far East.
The most notable SLP feature during wintertime over northern Eurasia is the Siberian high (SH)—a semipermanent surface high centered over northern Mongolia (Lydolf 1977; Sahsamanoglou et al. 1991). The Siberian high has a strong impact on both local and remote climate: its buildup and variability are associated with the very low temperatures found in eastern Siberia and with cold air surges affecting East Asia (Yihui 1990) and, more recently, has been associated with teleconnection patterns stretching from the Arctic to the tropical Pacific (e.g., Panagiotopoulos et al. 2005; Huang et al. 2016). For example, a stronger SH has been linked with warm air advection from eastern Europe across the Kara and Laptev Seas, leading to above-average temperatures in the region (Panagiotopoulos et al. 2005).
Here, we define a Siberian high index (SHI) as the standardized area-averaged SLP anomaly over the domain 40°–65°N, 80°–120°E, previously used by Panagiotopoulos et al. (2005). Because the SLP anomalies in our composites are centered farther north than the climatological SH center, we further define a modified SHI (SHI_N) over the domain 60°–80°N, 80°–120°E. The projections of the warm and cold spells on these two indices are shown in Figs. 14c and 14d, respectively. The warm spells are systematically associated with a strengthened SH, and both indices display significantly heightened values over the range from −2 to +4 days. The cold spells again present a weaker negative projection, which is significant only for SHI_N.
The intensified SH is consistent with the intense negative temperature anomalies seen over northern Eurasia during the Arctic warm extremes. This so-called WACE pattern has been extensively discussed in the literature, motivated by the rapidly warming Arctic temperatures, the decreasing sea ice cover, and the repeated cold extremes that have affected the northern mid-to-high latitudes in recent winters (e.g., Cohen et al. 2014; see also section 1). However, the focus has mostly been on seasonal time scales (e.g., Mori et al. 2014; Sun et al. 2016) or on seasonal frequency of cold spells (e.g., Tang et al. 2013) rather than on the link between individual warm and/or cold episodes. Here, we verify whether the warm extremes in the high Arctic systematically match cold extremes over Eurasia. We define a northern Eurasian temperature index (NETI) and northern Eurasian cold extremes using the same methodology outlined in section 2b for the Arctic warm and cold spells, but now applied to the domain 37.5°–60°N, 50°–120°E. This is chosen to match the strongest temperature anomalies seen in Figs. 2a–c, while at the same time encompassing densely populated regions across Asia.
Figure 15a shows the match between warm Arctic extremes and cold NETI extremes. While there is no one-to-one correspondence between the two sets of events, 14 of the top 50 Eurasian cold extremes fall within 5 days of an Arctic warm extreme, well above the 99th percentile obtained from random sampling. Similarly, the warm Arctic extremes are systematically associated with significantly below-average temperatures over Eurasia at both negative and positive lags. The mean area-weighted temperature anomaly over the NETI domain for lags from −5 to +5 days relative to the warm extremes is −1.44 K, while the 1st percentile obtained from random sampling is −0.80 K. The scatterplot of positive Arctic temperature anomalies versus the corresponding NETI values and of negative NETI values versus the corresponding Arctic temperature anomalies (Fig. 15b) confirms this picture. There is no systematic correspondence between the two, but in general, warm extremes in the Arctic favor negative NETI values (39 out of 50), while cold NETI extremes favor warm anomalies in the Arctic (38 out of 50).
(a) Timing of the Arctic warm extremes (red bars with circles) and Eurasian cold extremes (blue circles). The top 50 events for both classes are selected over the period 1979–2016. The blue line is the NETI (see section 6). (b) Scatterplot of all negative NETI episodes vs the corresponding Arctic temperature anomalies (blue circles); all the positive Arctic anomalies vs the corresponding NETI values (red circles); all the cold Eurasian extremes vs the corresponding Arctic temperature anomalies (blue-filled circles); and all the warm Arctic extremes vs the corresponding NETI values (red-filled circles). Note that the warm and cold anomalies have been selected enforcing a minimum 7-day separation and excluding warmer and colder days, as described in section 2b. The red horizontal lines mark the mean (solid) and median (dashed) NETI values corresponding to warm Arctic extremes. The blue vertical lines mark the mean (solid) and median (dashed) Arctic temperature anomalies corresponding to cold Eurasian extremes.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
(a) Timing of the Arctic warm extremes (red bars with circles) and Eurasian cold extremes (blue circles). The top 50 events for both classes are selected over the period 1979–2016. The blue line is the NETI (see section 6). (b) Scatterplot of all negative NETI episodes vs the corresponding Arctic temperature anomalies (blue circles); all the positive Arctic anomalies vs the corresponding NETI values (red circles); all the cold Eurasian extremes vs the corresponding Arctic temperature anomalies (blue-filled circles); and all the warm Arctic extremes vs the corresponding NETI values (red-filled circles). Note that the warm and cold anomalies have been selected enforcing a minimum 7-day separation and excluding warmer and colder days, as described in section 2b. The red horizontal lines mark the mean (solid) and median (dashed) NETI values corresponding to warm Arctic extremes. The blue vertical lines mark the mean (solid) and median (dashed) Arctic temperature anomalies corresponding to cold Eurasian extremes.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
(a) Timing of the Arctic warm extremes (red bars with circles) and Eurasian cold extremes (blue circles). The top 50 events for both classes are selected over the period 1979–2016. The blue line is the NETI (see section 6). (b) Scatterplot of all negative NETI episodes vs the corresponding Arctic temperature anomalies (blue circles); all the positive Arctic anomalies vs the corresponding NETI values (red circles); all the cold Eurasian extremes vs the corresponding Arctic temperature anomalies (blue-filled circles); and all the warm Arctic extremes vs the corresponding NETI values (red-filled circles). Note that the warm and cold anomalies have been selected enforcing a minimum 7-day separation and excluding warmer and colder days, as described in section 2b. The red horizontal lines mark the mean (solid) and median (dashed) NETI values corresponding to warm Arctic extremes. The blue vertical lines mark the mean (solid) and median (dashed) Arctic temperature anomalies corresponding to cold Eurasian extremes.
Citation: Journal of Climate 31, 4; 10.1175/JCLI-D-17-0386.1
We further note that warm Arctic extremes display a robust cold footprint over eastern North America (Figs. 2a,b), albeit weaker and less persistent than the Eurasian anomalies. Northerly advection over the region is associated with the meridional pressure dipole spanning the eastern half of the North Atlantic basin (Figs. 3g–i), and the general atmospheric configuration closely resembles that associated with westerly cold air outbreaks over the Irminger Sea. These outbreaks are associated with lee cyclogenesis or the intensification of pre-existing cyclones in the Irminger Sea; the cyclones then typically proceed along a northeastward trajectory into the Nordic seas (Papritz 2017). This is consistent with the positive anomalies in cyclone frequency seen along the east coast of Greenland and around Iceland in Fig. 11d.
7. Conclusions
Our analysis of wintertime (November–March) temperature extremes in the high Arctic highlights a number of systematic large-scale circulation features and synoptic-scale drivers common to the vast majority of episodes. The warm extremes are characterized by an anomalous SLP and geopotential height dipole, with a low over the Arctic and a high over northern Eurasia, conducive to meridional advection from the Atlantic sector into the Arctic basin. This wavenumber 1 configuration is consistent with the dominant role of planetary waves in impacting Arctic temperatures (Graversen and Burtu 2016). A similar large-scale pattern has also been associated with enhanced meridional moisture transport and wintertime sea ice decline over the Barents and Kara Seas (Luo et al. 2017). Indeed, the warm extremes are systematically preceded by a large number of intense moisture transport episodes into the high latitudes, here termed moisture intrusions. At synoptic scales, these intrusions are further favored by cyclones that entrain moist air masses residing in the Norwegian Sea. We note that the cyclones generated in the North Atlantic do not generally penetrate into the Arctic, and the high-latitude cyclones are generated locally. These results lead us to propose a conceptual picture whereby the atmospheric moisture contained in the moisture intrusion events is relayed into the Arctic via an interaction of several cyclonic systems centered at different latitudes. The moisture intrusions lead to a weakening of the near-surface temperature inversion in the Arctic basin, while their uplift drives positive cloudiness anomalies there. An additional consequence of the large-scale anomalies associated with the warm extremes is the advection of cold polar air masses across Siberia and into central Eurasia, leading to cold anomalies there—a situation resembling the so-called warm Arctic–cold Eurasia pattern. Conversely, the Arctic cold extremes appear to arise mainly due to the Arctic being sealed off from intense moisture advection from lower latitudes through an enhanced zonality of the high-latitude atmospheric flow. This then allows for a rapid radiative cooling, which results in unusually low temperatures across the region.
The large-scale circulation anomalies described here are likely driven by specific planetary wave-breaking patterns. The latter have indeed been linked to large meridional moisture transport into the Arctic (Liu and Barnes 2015). More generally, planetary-scale motions have been shown to play a major role in affecting Arctic temperatures (Baggett and Lee 2015; Goss et al. 2016; Graversen and Burtu 2016). At the same time, our analysis highlights the important role played by synoptic-scale motions, which can interact with planetary-scale perturbations and lead to a very large moisture transport into the high latitudes [see also Baggett et al. (2016), who focus on the Pacific sector, and Messori and Czaja (2014, 2015) and Messori et al. (2017), who focus on moist static energy]. A promising pathway for future studies might therefore be to partition the contribution of the different scales to the moisture extremes, following the approach of Graversen and Burtu (2016). Additionally, Goss et al. (2016) noted that Arctic warming episodes are enhanced and prolonged when constructive interference with the climatological stationary wave occurs in concert with warm pool convection, although the latter is not a necessary condition for the warming to occur in the first place. A further avenue for future research would therefore be to investigate whether the association of warm spells to prior warm pool convection depends on the former’s duration. Concerning the synoptic scales, the mechanism generating the Arctic cyclonic anomalies seen during the warm extremes and the potential role of cyclone clustering in driving more persistent than usual warm spells remain unclear and are future additional targets.
The present analysis has focused on the variability and extremes of the detrended wintertime temperature signal. However, the surprisingly rapid warming of the Arctic in the last decades (e.g., Vinnikov et al. 1980; Polyakov et al. 2002; Serreze and Barry 2011) and the large number of extremes affecting it (e.g., Perovich et al. 2008; Wormbs 2013; Moore 2016; Kim et al. 2017) open the question of whether and how the large-scale patterns associated with temperature extremes have changed over time and whether and how they may change in the future. A stimulating hypothesis could be that part of the Arctic amplification signal derives from a higher frequency of warm extremes, induced by a more frequent recurrence of the large-scale conditions that favor them.
Acknowledgments
G. Messori has been funded by a grant from the Department of Meteorology of Stockholm University and by Vetenskapsrådet under Contract 2016-03724_VR. C. Woods and R. Caballero acknowledge the support of Vetenskapsrådet under Contract E0531901. ERA-Interim data are freely available from ECMWF (http://apps.ecmwf.int/datasets). The authors thank an anonymous reviewer, R. G. Graversen, and S. Lee for their constructive feedback and J. M. Monteiro for helpful discussions.
REFERENCES
Baggett, C. , and S. Lee , 2015: Arctic warming induced by tropically forced tapping of available potential energy and the role of the planetary-scale waves. J. Atmos. Sci., 72, 1562–1568, https://doi.org/10.1175/JAS-D-14-0334.1.
Baggett, C. , S. Lee , and S. Feldstein , 2016: An investigation of the presence of atmospheric rivers over the North Pacific during planetary-scale wave life cycles and their role in Arctic warming. J. Atmos. Sci., 73, 4329–4347, https://doi.org/10.1175/JAS-D-16-0033.1.
Barnes, E. A. , and J. A. Screen , 2015: The impact of Arctic warming on the midlatitude jet-stream: Can it? Has it? Will it? Wiley Interdiscip. Rev.: Climate Change, 6, 277–286, https://doi.org/10.1002/wcc.337.
Cohen, J. , and Coauthors, 2014: Recent Arctic amplification and extreme mid-latitude weather. Nat. Geosci., 7, 627–637, https://doi.org/10.1038/ngeo2234.
Comiso, J. C. , 2006: Abrupt decline in the Arctic winter sea ice cover. Geophys. Res. Lett., 33, L18504, https://doi.org/10.1029/2006GL027341.
Cullather, R. I. , Y.-K. Lim , L. N. Boisvert , L. Brucker , J. N. Lee , and S. M. J. Nowicki , 2016: Analysis of the warmest Arctic winter, 2015–2016. Geophys. Res. Lett., 43, 10 808–10 816, https://doi.org/10.1002/2016GL071228.
Dee, D. P. , and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
Flournoy, M. D. , S. B. Feldstein , S. Lee , and E. E. Clothiaux , 2016: Exploring the tropically excited Arctic warming mechanism with station data: Links between tropical convection and Arctic downward infrared radiation. J. Atmos. Sci., 73, 1143–1158, https://doi.org/10.1175/JAS-D-14-0271.1.
Garfinkel, C. I. , S.-W. Son , K. Song , V. Aquila , and L. D. Oman , 2017: Stratospheric variability contributed to and sustained the recent hiatus in Eurasian winter warming. Geophys. Res. Lett., 44, 374–382, https://doi.org/10.1002/2016GL072035.
Goss, M. , S. B. Feldstein , and S. Lee , 2016: Stationary wave interference and its relation to tropical convection and Arctic warming. J. Climate, 29, 1369–1389, https://doi.org/10.1175/JCLI-D-15-0267.1.
Graversen, R. G. , and M. Burtu , 2016: Arctic amplification enhanced by latent energy transport of atmospheric planetary waves. Quart. J. Roy. Meteor. Soc., 142, 2046–2054, https://doi.org/10.1002/qj.2802.
Hanley, J. , and R. Caballero , 2012: Objective identification and tracking of multicentre cyclones in the ERA-Interim reanalysis dataset. Quart. J. Roy. Meteor. Soc., 138, 612–625, https://doi.org/10.1002/qj.948.
Honda, M. , J. Inoue , and S. Yamane , 2009: Influence of low Arctic sea-ice minima on anomalously cold Eurasian winters. Geophys. Res. Lett., 36, L08707, https://doi.org/10.1029/2008GL037079.
Hoskins, B. J. , and K. I. Hodges , 2002: New perspectives on the Northern Hemisphere winter storm tracks. J. Atmos. Sci., 59, 1041–1061, https://doi.org/10.1175/1520-0469(2002)059<1041:NPOTNH>2.0.CO;2.
Huang, W. , B. Wang , J. S. Wright , and R. Chen , 2016: On the non-stationary relationship between the Siberian high and Arctic Oscillation. PLoS One, 11, e0158122, https://doi.org/10.1371/journal.pone.0158122.
Inoue, J. , M. E. Hori , and K. Takaya , 2012: The role of Barents Sea ice in the wintertime cyclone track and emergence of a warm-Arctic cold-Siberian anomaly. J. Climate, 25, 2561–2568, https://doi.org/10.1175/JCLI-D-11-00449.1.
Jakobson, E. , T. Vihma , T. Palo , L. Jakobson , H. Keernik , and J. Jaagus , 2012: Validation of atmospheric reanalyses over the central Arctic Ocean. Geophys. Res. Lett., 39, L10802, https://doi.org/10.1029/2012GL051591.
Johannessen, O. M. , and Coauthors, 2004: Arctic climate change: Observed and modelled temperature and sea-ice variability. Tellus, 56A, 328–341, https://doi.org/10.3402/tellusa.v56i4.14418.
Kalnay, E. , and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471, https://doi.org/10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.
Kapsch, M.-L. , R. G. Graversen , and M. Tjernström , 2013: Springtime atmospheric energy transport and the control of Arctic summer sea-ice extent. Nat. Climate Change, 3, 744–748, https://doi.org/10.1038/nclimate1884.
Kim, B.-M. , and Coauthors, 2017: Major cause of unprecedented Arctic warming in January 2016: Critical role of an Atlantic windstorm. Sci. Rep., 7, 40051, https://doi.org/10.1038/srep40051.
Kug, J.-S. , J.-H. Jeong , Y.-S. Jang , B.-M. Kim , C. K. Folland , S.-K. Min , and S.-W. Son , 2015: Two distinct influences of Arctic warming on cold winters over North America and East Asia. Nat. Geosci., 8, 759–762, https://doi.org/10.1038/NGEO2517.
Lee, S. , 2012: Testing of the tropically excited Arctic warming mechanism (TEAM) with traditional El Niño and La Niña. J. Climate, 25, 4015–4022, https://doi.org/10.1175/JCLI-D-12-00055.1.
Lee, S. , S. B. Feldstein , D. Pollard , and T. S. White , 2011a: Do planetary wave dynamics contribute to equable climates? J. Climate, 24, 2391–2404, https://doi.org/10.1175/2011JCLI3825.1.
Lee, S. , T. Gong , N. Johnson , S. B. Feldstein , and D. Pollard , 2011b: On the possible link between tropical convection and the Northern Hemisphere Arctic surface air temperature change between 1958 and 2001. J. Climate, 24, 4350–4367, https://doi.org/10.1175/2011JCLI4003.1.
Lindsay, R. , M. Wensnahan , A. Schweiger , and J. Zhang , 2014: Evaluation of seven different atmospheric reanalysis products in the Arctic. J. Climate, 27, 2588–2606, https://doi.org/10.1175/JCLI-D-13-00014.1.
Liu, C. , and E. A. Barnes , 2015: Extreme moisture transport into the Arctic linked to Rossby wave breaking. J. Geophys. Res. Atmos., 120, 3774–3788, https://doi.org/10.1002/2014JD022796.
Luo, B. , D. Luo , L. Wu , L. Zhong , and I. Simmonds , 2017: Atmospheric circulation patterns which promote winter Arctic sea ice decline. Environ. Res. Lett., 12, 054017, https://doi.org/10.1088/1748-9326/aa69d0.
Luo, D. , Y. Xiao , Y. Yao , A. Dai , I. Simmonds , and C. L. E. Franzke , 2016: Impact of Ural blocking on winter warm Arctic–cold Eurasian anomalies. Part I: Blocking-induced amplification. J. Climate, 29, 3925–3947, https://doi.org/10.1175/JCLI-D-15-0611.1.
Lydolf, P. E. , 1977: Climates of the Soviet Union. Vol. 7, World Survey of Climatology, Elsevier, 443 pp.
Maslanik, J. , J. Stroeve , C. Fowler , and W. Emery , 2011: Distribution and trends in Arctic sea ice age through spring 2011. Geophys. Res. Lett., 38, L13502, https://doi.org/10.1029/2011GL047735.
Matthes, H. , A. Rinke , and K. Dethloff , 2015: Recent changes in Arctic temperature extremes: Warm and cold spells during winter and summer. Environ. Res. Lett., 10, 114020, https://doi.org/10.1088/1748-9326/10/11/114020; Corrigendum, 11, 029501, https://doi.org/10.1088/1748-9326/11/2/029501.
McCusker, K. E. , J. C. Fyfe , and M. Sigmond , 2016: Twenty-five winters of unexpected Eurasian cooling unlikely due to Arctic sea-ice loss. Nat. Geosci., 9, 838–842, https://doi.org/10.1038/ngeo2820.
Messori, G. , and A. Czaja , 2014: Some considerations on the spectral features of meridional heat transport by transient eddies. Quart. J. Roy. Meteor. Soc., 140, 1377–1386, https://doi.org/10.1002/qj.2224.
Messori, G. , and A. Czaja , 2015: On local and zonal pulses of atmospheric heat transport in reanalysis data. Quart. J. Roy. Meteor. Soc., 141, 2376–2389, https://doi.org/10.1002/qj.2529.
Messori, G. , R. Geen , and A. Czaja , 2017: On the spatial and temporal variability of atmospheric heat transport in a hierarchy of models. J. Atmos. Sci., 74, 2163–2198, https://doi.org/10.1175/JAS-D-16-0360.1.
Moore, G. W. K. , 2016: The December 2015 North Pole warming event and the increasing occurrence of such events. Sci. Rep., 6, 39084, https://doi.org/10.1038/srep39084.
Mori, M. , M. Watanabe , H. Shiogama , J. Inoue , and M. Kimoto , 2014: Robust Arctic sea-ice influence on the frequent Eurasian cold winters in past decades. Nat. Geosci., 7, 869–873, https://doi.org/10.1038/ngeo2277.
Neu, U. , and Coauthors, 2013: IMILAST: A community effort to intercompare extratropical cyclone detection and tracking algorithms. Bull. Amer. Meteor. Soc., 94, 529–547, https://doi.org/10.1175/BAMS-D-11-00154.1.
Panagiotopoulos, F. , M. Shahgedanova , A. Hannachi , and D. B. Stephenson , 2005: Observed trends and teleconnections of the Siberian high: A recently declining center of action. J. Climate, 18, 1411–1422, https://doi.org/10.1175/JCLI3352.1.
Papritz, L. , 2017: Synoptic environments and characteristics of cold air outbreaks in the Irminger Sea. Int. J. Climatol., 37, 193–207, https://doi.org/10.1002/joc.4991.
Park, D.-S. R. , S. Lee , and S. B. Feldstein , 2015: Attribution of the recent winter sea ice decline over the Atlantic sector of the Arctic Ocean. J. Climate, 28, 4027–4033, https://doi.org/10.1175/JCLI-D-15-0042.1.
Park, H.-S. , S. Lee , Y. Kosaka , S.-W. Son , and S.-W. Kim , 2015a: The impact of Arctic winter infrared radiation on early summer sea ice. J. Climate, 28, 6281–6296, https://doi.org/10.1175/JCLI-D-14-00773.1.
Park, H.-S. , S. Lee , S.-W. Son , S. B. Feldstein , and Y. Kosaka , 2015b: The impact of poleward moisture and sensible heat flux on Arctic winter sea ice variability. J. Climate, 28, 5030–5040, https://doi.org/10.1175/JCLI-D-15-0074.1.
Perovich, D. K. , J. A. Richter-Menge , K. F. Jones , and B. Light , 2008: Sunlight, water, and ice: Extreme Arctic sea ice melt during the summer of 2007. Geophys. Res. Lett., 35, L11501, https://doi.org/10.1029/2008GL034007.
Polyakov, I. V. , and Coauthors, 2002: Observationally based assessment of polar amplification of global warming. Geophys. Res. Lett., 29, 1878, https://doi.org/10.1029/2001GL011111.
Sahsamanoglou, H. S. , T. J. Makrogiannis , and P. P. Kallimopoulos , 1991: Some aspects of the basic characteristics of the Siberian anticyclone. Int. J. Climatol., 11, 827–839, https://doi.org/10.1002/joc.3370110803.
Screen, J. A. , and I. Simmonds , 2013: Exploring links between Arctic amplification and mid-latitude weather. Geophys. Res. Lett., 40, 959–964, https://doi.org/10.1002/grl.50174.
Sepp, M. , and J. Jaagus , 2011: Changes in the activity and tracks of Arctic cyclones. Climatic Change, 105, 577–595, https://doi.org/10.1007/s10584-010-9893-7.
Serreze, M. C. , and R. G. Barry , 2011: Processes and impacts of Arctic amplification: A research synthesis. Global Planet. Change, 77, 85–96, https://doi.org/10.1016/j.gloplacha.2011.03.004.
Seviour, W. J. M. , 2017: Weakening and shift of the Arctic stratospheric polar vortex: Internal variability or forced response? Geophys. Res. Lett., 44, 3365–3373, https://doi.org/10.1002/2017GL073071.
Sorteberg, A. , and J. E. Walsh , 2008: Seasonal cyclone variability at 70°N and its impact on moisture transport into the Arctic. Tellus, 60A, 570–586, https://doi.org/10.1111/j.1600-0870.2008.00314.x.
Stroeve, J. , M. Serreze , S. Drobot , S. Gearheard , M. Holland , J. Maslanik , W. Meier , and T. Scambos , 2008: Arctic sea ice extent plummets in 2007. Eos, Trans. Amer. Geophys. Union, 89, 13–14, https://doi.org/10.1029/2008EO020001.
Sun, L. , J. Perlwitz , and M. Hoerling , 2016: What caused the recent “warm Arctic, cold continents” trend pattern in winter temperatures? Geophys. Res. Lett., 43, 5345–5352, https://doi.org/10.1002/2016GL069024.
Tang, Q. , X. Zhang , X. Yang , and J. A. Francis , 2013: Cold winter extremes in northern continents linked to Arctic sea ice loss. Environ. Res. Lett., 8, 014036, https://doi.org/10.1088/1748-9326/8/1/014036.
Vinnikov, K. Ya. , G. V. Gruza , V. F. Zakharov , A. A. Kirillov , N. P. Kovyneva , and E. Ya. Ran’kova , 1980: Recent climatic changes in the Northern Hemisphere. Soviet Meteor. Hydrol., 6, 1–10.
Walsh, J. E. , 2014: Intensified warming of the Arctic: Causes and impacts on middle latitudes. Global Planet. Change, 117, 52–63, https://doi.org/10.1016/j.gloplacha.2014.03.003.
Woods, C. , and R. Caballero , 2016: The role of moist intrusions in winter Arctic warming and sea ice decline. J. Climate, 29, 4473–4485, https://doi.org/10.1175/JCLI-D-15-0773.1.
Woods, C. , R. Caballero , and G. Svensson , 2013: Large-scale circulation associated with moisture intrusions into the Arctic during winter. Geophys. Res. Lett., 40, 4717–4721, https://doi.org/10.1002/grl.50912.
Woods, C. , R. Caballero , and G. Svensson , 2017: Representation of Arctic moist intrusions in CMIP5 models and implications for winter climate biases. J. Climate, 30, 4083–4102, https://doi.org/10.1175/JCLI-D-16-0710.1.
Wormbs, N. , 2013: Eyes on the ice: Satellite remote sensing and the narratives of visualized data. Media and the Politics of Arctic Climate Change, M. Christensen, A. E. Nilsson, and N. Wormbs, Eds., Palgrave Macmillan, 52–69.
Yihui, D. , 1990: Build-up, air mass transformation and propagation of Siberian high and its relations to cold surge in East Asia. Meteor. Atmos. Phys., 44, 281–292, https://doi.org/10.1007/BF01026822.
Yu, L. , C. Sui , D. H. Lenschow , and M. Zhou , 2017: The relationship between wintertime extreme temperature events north of 60°N and large-scale atmospheric circulations. Int. J. Climatol., 37, 597–611, https://doi.org/10.1002/joc.5024.