Stratospheric influence on North Atlantic marine cold air outbreaks following sudden stratospheric warming events

following sudden stratospheric warming events Hilla Afargan-Gerstman1, Iuliia Polkova2, Lukas Papritz1, Paolo Ruggieri3,4, Martin P. King5, Panos J. Athanasiadis4, Johanna Baehr2, and Daniela I.V. Domeisen1 1Institute for Atmospheric and Climate Science, ETH Zürich, Zürich, Switzerland 2Institute of Oceanography, Universität Hamburg, CEN, Hamburg, Germany 3Department of Physics and Astronomy, University of Bologna, Italy 4Euro-Mediterranean Center on Climate Change (CMCC), Bologna, Italy 5NORCE Climate, and Bjerknes Centre for Climate Research, Bergen, Norway Correspondence: Hilla Afargan-Gerstman (hilla.gerstman@env.ethz.ch)

thresholds for the MCAO index (e.g., a threshold of 3 K for moderate MCAO events in Fletcher et al. (2016)), however, the 85 results are not sensitive to small changes of this threshold (on the order of 1-2 K).

Characterization of the large-scale flow
We define a new index based on the 500-hPa geopotential height anomaly from climatology (Z ). The index, named the Zonal Dipole Index (ZDI), is equal to half of the difference in Z between the spatial average over two main areas that modulate the frequency of MCAOs in the Barents and Norwegian Seas (enclosed by the green boxes in Fig. 2d): southeast of Greenland 90 (Z G , 30°W-50°W, 60°N-70°N) and northern Europe (Z E , 30°E-50°E, 60°N-70°N), as follows (2)

SSW events
To assess the impact of the stratosphere on MCAO occurrence, we examine the changes in the MCAO frequency in response to 26 observed major SSW events between 1979-2019. Major SSWs occur when the westerlies associated with the winter 95 stratospheric polar vortex reverse to easterlies. A common definition for the central date of major SSWs is based on a change from westerly to easterly of the daily mean zonal-mean zonal winds at 10 hPa and 60°N between November and March (Charlton and Polvani, 2007). A list of major SSW events in the ERA-Interim reanalysis for the period 1979-2019 is shown in Table 1. The central dates of SSW events between 1979-2014 are based on Butler et al. (2017). Two additional SSW events, on 12 February 2018 and 2 January 2019, are detected according to a wind reversal at 10 hPa and 60°N.

100
The MCAO response to SSW events is here defined as the change in MCAO frequency or index over a period of 30 days after the onset of the SSW (i.e., the SSW central date). The MCAO frequency (P M ) is defined as the percentage of days with MCAOs that exceed a 4 K threshold (M ≥ 4 K) within a period of 30 days. The climatological MCAO frequency (P M≥4K,Clim ) is computed between the December to March and standardized by the number of days in DJFM. Following SSW events, the MCAO frequency (P M≥4K,SSW ) is computed as the percentage of days with M ≥ 4 K within a 30-day period after the SSW 105 central date. Thus, to obtain the MCAO frequency anomaly, the MCAO frequency after SSW events is then compared with the climatological MCAO frequency. Sea and the Barents Sea (black boxes from west to east in Fig. 1a). MCAOs are also most frequent over these regions (Fig. 1b), with a likelihood of more than 40% for an occurrence of moderate-to-strong MCAOs in the Labrador Sea during DJFM, and nearly 35% in the Norwegian Sea and the Barents Sea.

MCAOs in climatology and in response to SSW events
We examine the changes in the tropospheric large-scale flow conditions in response to major SSW events in ERA-Interim. 115 Major SSW events tend to be followed by anomalously cold temperatures over the northeastern North Atlantic and Eurasia (Fig. 1c,d). These anomalies are accompanied by a north-south dipole pattern of 500-hPa geopotential height over the North Atlantic ( Fig. 1e), consisting of a positive anomaly over Greenland, and a negative anomaly southeast of Greenland and over central Europe. This pattern is often associated with a negative phase of the NAO (e.g., Limpasuvan et al., 2004;Butler et al., 2017;Domeisen, 2019). Following SSW events, MCAO frequency exhibits significant regional variability, with the largest 120 increase of MCAO frequency over the western Barents and Norwegian Seas and a decrease along the sea ice edge over the Greenland Sea as well as the Labrador Sea (Fig. 1f).
To assess whether the anomalously cold conditions, which often occur over the western North Atlantic after SSW events, have an impact on the MCAOs in these regions, we analyze changes in the MCAO frequency and strength over a period of 30 days after the onset of a SSW. We focus on the Barents Sea (70°N to 78.5°N, 30°E to 50°E, easternmost box in Fig. 1a), the 125 Norwegian Sea (60°N to 80°N, 15°W to 5°E, central box in Fig. 1a), and the Labrador Sea (55°N to 67.5°N, 40°W to 62.5°W, westernmost box in Fig. 1a). In the next subsections, the link between the large-scale atmospheric circulation and MCAOs in these sub-regions of the North Atlantic is investigated, both in climatology and in connection with stratospheric forcing.

The large-scale atmospheric circulation during MCAOs
In this section we first characterize and establish the large-scale atmospheric circulation patterns associated with anomalously 130 high MCAO occurrences in the North Atlantic for the winter months, without considering the occurrence of SSW events. For this purpose, we examine composites of the geopotential height and meridional wind anomalies in periods of moderate-tostrong MCAOs in DJFM climatology. These periods are identified using the criterion of M ≥ 4 K. We focus on the three regions of interest shown in Fig. 1a: the Barents Sea, the Norwegian Sea and the Labrador Sea.
Periods of moderate-to-strong MCAO intensity in the Barents Sea (left column in Fig. 2) are found to be associated with 135 a zonal dipole pattern of geopotential height anomaly in the mid-troposphere, consisting of a high-pressure anomaly over southern Greenland ("Greenland Blocking") and a low-pressure anomaly over Northern Europe, Scandinavia and the Barents Sea ("Scandinavian Trough") ( Fig. 2d). Cyclone frequency during these periods indicates an increase in storminess primarily east of the Barents Sea ( Fig. 2g)   Periods of moderate-to-strong MCAOs in the Norwegian Sea (middle column in Fig. 2) are similar to those in the Barents Sea but with circulation anomalies shifted slightly to the west. Specifically, they are characterized by a zonal dipole of geopotential height anomalies, with positive anomalies centered over the region south of Greenland and negative anomalies over Scandinavia ( Fig. 2e). Anomalous cyclone frequency is found to be reduced over the western Norwegian Sea, but increased along the Norwegian coast and across the Barents Sea (Fig. 2h). Consistent with that, the wind anomaly indicates a northerly flow over 150 the Norwegian Sea (Fig. 2k). The maximum of the meridional wind anomalies is found further westward as compared to the position of the maximum meridional winds for periods of strong Barents Sea MCAOs (Fig. 2j).
During periods of moderate-to-strong MCAOs over the Labrador Sea and southern Greenland (right column in Fig. 2), finally, the geopotential height pattern is found to be associated with negative geopotential height anomalies centred over the Labrador Sea and positive anomalies over western Europe and Scandinavia (Fig. 2f). Enhanced cyclone frequency south-east 155 of Greenland suggests a connection to increased transient storm activity over this region (Fig. 2i), associated with anomalous northerlies west of Greenland and more southerly winds south-east of Greenland and in the Norwegian Sea (Fig. 2l). Thus, this pattern is, to some extent, opposed to that found for the Barents and the Norwegian Seas but it is consistent with a southward advection of cold air masses into the Labrador Sea, as well as the subsequent eastward advection of the cold air masses south of Greenland (not shown).

160
Thus, more intense MCAOs in all three considered regions of the North Atlantic are clearly linked to specific large-scale circulation patterns with pronounced mid-tropospheric geopotential height anomalies over Greenland and Scandinavia, as well as related shifts of the North Atlantic storm track. These flow anomalies, in turn, cause anomalous advection of cold air masses over the open ocean west of positive cyclone frequency anomalies. In the next subsection, we investigate how the occurrence of SSW events may affect MCAOs by modulation of the prevailing tropospheric conditions. For that purpose, we will focus on 165 the mid-tropospheric geopotential height anomalies. Establishing the link between these stratospheric events and MCAOs can provide further insight on the different pathways of MCAO occurrence, their intensity and frequency.

Stratospheric influence on MCAO occurrence following SSW events
In winter, extreme states of the stratospheric polar vortex can have a significant impact on the tropospheric circulation in the North Atlantic, particularly affecting the state of the NAO (e.g., Baldwin and Dunkerton, 2001;Karpechko et al., 2017; 170 Charlton-Perez et al., 2018;Domeisen, 2019). Here, we explore the influence of the stratosphere on the large-scale tropospheric circulation in the North Atlantic, and on the intensity of MCAOs. For this purpose, we first establish the link between the largescale circulation and MCAOs in climatology, and compare to periods that follow SSW events.
To evaluate the link between the dominant large-scale anomaly pattern and MCAOs, we first consider the extent to which the ZDI index, defined in Eq. 2, is linked to MCAO intensity. By definition, the ZDI is designed to capture the centers of action 175 of the geopotential height anomalies. The dependence between these indices is shown by a linear regression, as follows where M is the MCAO index and b M is the linear regression coefficient. In the Barents and the Norwegian Seas, a positive linear relation is found between the ZDI index (representing the anomalous dipole pattern) and the strength of MCAOs in 180 DJFM (black line in Fig. 3a,b). The opposite relation is found for MCAOs in the Labrador Sea region, which exhibit a negative linear relation with the ZDI index (Fig. 3c). This relation is consistent with the geopotential height pattern shown in Fig. 2f, and corresponds to a negative ZDI index.
After SSW events (triangles in Fig. 3a), a higher correlation between MCAOs in the Barents Sea and the ZDI index is found as compared to climatology (R 2 =0.42 after SSW events versus R 2 =0.24 in climatology). For each SSW, this period is defined 185 as the first 4 weeks after the onset of the SSW. Moreover, in periods that do not include SSW events, the correlation between the MCAO index and the ZDI index is weaker than in the climatology (R 2 =0.21, shown in grey). These periods exclude the first 4 weeks after SSW events (see section 3.3). In the Norwegian Sea the correlation between the ZDI and the MCAO index increases slightly after SSW events (R 2 =0.31) as compared to climatology (R 2 =0.26) and periods without SSW events (R 2 =0.24) (Fig. 3b). In the Labrador Sea, the negative correlation between the ZDI and the MCAO index is weakened after 190 SSW events (R 2 =0.21) relative to climatology (R 2 =0.27). For the Barents Sea region, the correlations between the ZDI and MCAO indices for SSW (orange) and no SSW (grey) periods are significantly different from each other at the 95% confidence level using the Fisher z-test (p=0.03). However, for the Norwegian and the Labrador Sea regions the confidence level is below 95% (p=0.24 and p=0.25, respectively). The autocorrelation of the weekly indices has been accounted for by estimating the number of independent samples, which is found to be one week in DJFM. Using a larger effective sample size after SSW events 195 leads to a qualitatively similar conclusion.
To further understand which of the components of the dipole index has a dominant effect on the anomalous MCAO index, we separately analyse regression of the MCAO index on the geopotential height anomalies over Greenland and northern Europe as represented by Z G and Z E in Eq. 2, respectively. As positive anomalies of the dipole index are centered over Greenland and negative anomalies over Scandinavia, we define the indices GB ("Greenland blocking") and ST ("Scandinavian trough") which 200 correspond to Z G and Z E , respectively.
The results for moderate-to-strong MCAOs following the SSW events indicate a positive correlation between the Barents Sea MCAO index and the GB index (Fig. 4a), however with a larger spread compared to the ZDI index. A negative correlation is found with the ST index, which accounts for approximately 44% of the variance after these SSWs (Fig. 4b) as compared to the variance of 21% in climatology.

205
In the Norwegian Sea, a comparably high correlation is found after SSW events between the ST index and Norwegian Sea MCAOs (R 2 =0.30) (Fig. 4d), whereas in the Labrador Sea region, a considerable correlation is found with the GB index (R 2 =0.40) (Fig. 4e). The occurrence of SSW events somewhat weakens this correlation in the Labrador Sea compared to the periods without SSW events, although the correlation is relatively high in both cases (R 2 =0.40 and R 2 =0.42, respectively).   corresponding R 2 coefficient are computed for the climatology (black) and periods following SSW events (orange). For completeness, R 2 for weekly averages that do not include periods after SSW events is shown in grey. All anomalies are computed with respect to the daily climatology. All correlations are statistically significant (p < 0.05).   Out of the above correlations, only in the Barents Sea region the correlation between the ST index and MCAOs is found to be 210 significantly different at the 95% confidence level between SSW (orange) and no SSW (grey) periods using the Fisher z-test ( Fig. 4b).
Analyzing the linear relationship between MCAOs and the ZDI index over SSW and non-SSW periods demonstrates a link between the large-scale geopotential height anomaly pattern with the strength of MCAOs in these regions. After SSW events, there is an increase of 18% in the explained variance for the Barents Sea, and of 5% for the Norwegian Sea. The presence of a 215 low pressure anomaly over northern Eurasia (as represented by the negative ST index) dominates the relationship in the Nordic Seas, whereas a high pressure anomaly over Greenland (a positive GB index) has a stronger relationship with the Labrador Sea MCAOs. generally shifts towards positive values. However, there is an overlap of the ZDI during periods with and without SSW events.
The Z GB index tends toward more positive values for the periods after SSW events, whereas the Z ST index is rather more negative than neutral after SSW events.
Thus, a modulation of the large-scale flow patterns during DJFM and their projection on the GB and ST indices has an effect on the intensity of the MCAOs in the Arctic. In particular, increasing and decreasing pressure anomalies in the centers of the 225 large-scale zonal dipole pattern, leads to an enhancement of MCAOs in Barents and the Norwegian Seas. As stratospheric precursors such as SSWs often modulate surface weather in the European-North Atlantic regions (Charlton-Perez et al., 2018;Domeisen et al., 2020;Beerli and Grams, 2019), their impact on the large-scale circulation pattern (Fig. 1e) contributes to the increased likelihood of MCAOs in these regions in periods following SSW events (Fig. 1f).

230
In the previous section we have shown how the stratospheric influence following SSW events can modulate the dominant largescale circulation in the European-North Atlantic region, affecting the strength of MCAOs. In winter, the Euro-Atlantic sector may be dominated by cyclonic and blocked large-scale flow features (e.g., Beerli and Grams, 2019;Papritz and Grams, 2018;Domeisen et al., 2020). In a cyclonic flow pattern, a negative geopotential height anomaly (relative to DJFM climatology) dominates at 500 hPa. These negative anomalies are associated with enhanced cyclonic activity, and correspond to more than 235 one dominant weather regime, such as Atlantic or Scandinavian Troughs (Beerli and Grams, 2019 As discussed in subsection 3.3, the ZDI index is found to be positively correlated to the MCAO index in the Barents and the Norwegian Seas (Fig. 3a,b). In fact, a positive ZDI occurs nearly 50% of the time in DJFM, indicating the likelihood of a dipole pattern occurrence, while an opposite dipole pattern is likely to occur for a combination of different circulation patterns.
An analysis of the relation to the GB and ST geopotential height anomalies reveals that most of the variance found for the ZDI index ( Fig. 3a) can be attributed to the ST index (Fig. 4a), while the relation to the GB index exhibits a much larger variability 245 (Fig. 4b).
To assess the contribution of the zonal dipole pattern to the MCAO index in the North Atlantic, periods of GB and ST geopotential height anomalies are analyzed separately (Fig. 6). These periods are defined as days for which the 500-hPa geopotential height anomaly averaged over the GB and ST boxes is positive or negative, respectively. Results show that for both positive GB (Fig. 6a) and negative ST (Fig. 6b) the circulation over the Barents Sea is anomalously cyclonic (Fig. 6c,d), 250 and associated with an increase in the MCAO index over the Barents and the Norwegian Seas (Fig. 6e,f). Interestingly, only the GB pattern is accompanied by a reduced frequency of MCAOs in the Labrador Sea. These differences in MCAO are clearly related to the differences in storminess; Periods of negative ST are associated with increased storminess over Scandinavia and the southern Barents Sea, whereas periods of GB exhibit a strong reduction in cyclone frequency over the Nordic Seas centered over the Irminger Sea.

255
Furthermore, we examine the dependency of the MCAO index on the GB and ST indices in these sub-regions of the North Atlantic during DJFM. In the Barents Sea, stronger MCAOs (represented by light blue marker color) are found to be associated with negative ST index, for either a positive or a negative GB index (Fig. 7a). The most intense MCAOs are found for a negative ST index and a positive GB index, emphasizing the importance of this particular combination for the occurrence of MCAOs in the Barents Sea (this pattern is consistent with a positive ZDI index). In the Norwegian Sea, the MCAO dependency on the GB 260 and the ST indices is found to be similar to that of the Barents Sea, with stronger MCAOs associated with a negative ST index (Fig. 7b). In contrast, in the Labrador Sea, stronger MCAOs are primarily associated with a negative GB index, demonstrating a weaker sensitivity to the sign of the ST index (Fig. 7c).

Conclusions
This study focuses on the influence of the stratosphere on the occurrence of marine cold air outbreaks in the North Atlantic and 265 their connection to the large-scale circulation patterns over the North Atlantic and Europe. Particularly, we investigate how the frequency and the magnitude of such MCAOs in the Barents Sea, the Norwegian Sea, and the Labrador Sea are affected by the large-scale conditions after the onset of extreme events in the stratosphere, known as SSW events.
By analyzing the regional atmospheric conditions in DJFM between 1979 to 2019 we find that a positive 500-hPa geopotential height anomaly over Greenland and a negative geopotential height anomaly over Scandinavia, accompanied by increased 270 storminess and northerly surface winds over the Barents Sea and to the east of the Barents Sea, are strong indicators for enhanced MCAO intensity in these regions. In contrast, the opposite geopotential height anomaly pattern (i.e., lower geopotential height anomaly over Greenland and higher geopotential height anomaly over Scandinavia) and increased storminess in the   After SSW events, significant changes in the tropospheric surface flow tend to occur. These changes involve a negative phase of the NAO and extremely cold temperatures over the northeastern North Atlantic and Northern Europe (Fig. 1c,d). To assess whether these extreme changes have an impact on the occurrence of MCAOs in these regions, we analyze the characteristics of the large-scale atmospheric conditions after 26 SSW events between 1979 to 2019, as compared to climatology. We find that 280 changes in the large-scale atmospheric circulation pattern, represented by a positive zonal dipole index, accounts for 42% of the MCAO variance in the Barents Sea and 31% of the variance in the Norwegian Sea after SSW events. For comparison, the dependency on the zonal dipole index explains only 21% and 24% of the variance in winters without SSW events, respectively.
Thus, the correlation between the zonal dipole index and MCAOs following SSW events is found to be significantly higher than the correlation between the zonal dipole index and MCAOs in periods without SSWs (section 3.3).

285
Furthermore, we find that an intensification of the positive geopotential height anomaly over the southern part of Greenland (as represented by a positive Greenland Blocking pattern) is associated with weaker MCAOs in the Labrador Sea, and accounts for ∼40% of the MCAO variance in this region, both in periods following SSW events and in the DJFM climatology in general.
While the relevance of stratospheric forcing for MCAOs in the Labrador Sea is not found to be as statistically significant as in the Barents and the Norwegian Seas, the importance of MCAOs for dense water formation in this region implies that SSWs 290 might nevertheless have an impact on the North Atlantic overturning circulation, as suggested by Reichler et al. (2012).
Through linear regression analysis we demonstrate a statistical relationship between MCAOs and atmospheric indices that capture the characteristics of the large-scale flow (Figs. 3, 4). Such a connection can be further used for mitigation of societal and economic impacts by providing an estimate of the increase/decrease in MCAO intensity due to a change in the environ-mental conditions. Furthermore, understanding the connection between MCAOs in the North Atlantic and the stratospheric 295 forcing shows potential for improved predictive skill of cold air outbreaks on subseasonal to seasonal time scales. Seas. Thus, this connection can potentially be exploited for improving subseasonal MCAO predictions.
The preferred patterns for MCAOs may also indicate the pathway of cold air masses during MCAO formation. MCAO air masses over the Barents Sea tend to originate in the high or Siberian Arctic, with dominant pathways of cold air masses from Siberia across Novaja Zemlja and the northern sea ice edge into the Barents Sea (Papritz and Spengler, 2017). We show that the northern pathway is largely captured by a positive ZDI (Fig. 2d) and is consistent with a low pressure anomaly over north-305 eastern Europe, bringing cold air masses southward across the Norwegian and the Barents Seas (Fig. 2j,k). A negative ZDI, on the other hand, can be linked to a dominant blocking pattern over the Barents and Kara Seas, possibly related to the eastern pathway for MCAOs in the Barents Sea, consistent with Papritz (2017). A similar link between a positive ZDI index and MCAOs is found over the Norwegian Sea, suggesting the relevance of a northern pathway for the development of Norwegian Sea MCAOs (Fig. 2h,k). In contrast, a pathway for MCAOs in the Labrador Sea is linked to a dominant cyclonic regime over 310 Greenland, bringing a flow of cold air southward into the Labrador Sea (Fig. 2i,l).
We conclude that understanding the connection between the stratosphere and the occurrence of MCAOs in the North Atlantic reveals key ingredients for MCAO formation, which can potentially lead to improved prediction skill on subseasonal time scales due to the long-lasting circulation anomalies associated with stratosphere-troposphere coupling in winter. SSW events are found to have an effect on the large-scale circulation pattern in the troposphere, as evident from the ZDI distribution 315 shift towards positive values after SSW events (Fig. 5a). There is, however, a large variability among SSW events, as also discussed in previous studies (e.g., Karpechko et al., 2017;Afargan-Gerstman and Domeisen, 2020;Domeisen et al., 2020), which imposes some limitations on the predictability that in principle can be obtained in terms of MCAO forecasting.
In addition to the influence of the large-scale tropospheric flow, conditions in the boundary layer also play a role in the formation of MCAOs in the Arctic. Northerly winds during strong MCAOs in the Barents, Norwegian and Labrador Seas 320 are found to be stronger than climatology (Fig. 2j-l), consistent with previous studies (Kolstad, 2017;Fletcher et al., 2016).
Another factor that affects the occurrence of MCAOs is sea ice cover in the Barents Sea (e.g., Ruggieri et al., 2016). In a warming climate, the diminishing sea ice cover over the Barents Sea can potentially modulate MCAO occurrence in this region, by exposing more of the ocean surface to interaction with the atmosphere. Such changes are also likely to be affected by the availability of cold air masses in the Arctic (e.g., Papritz et al., 2019). Further work is thus required for understanding a 325 compound effect of Arctic processes on MCAO formation.