Atmospheric forcing of Fram Strait sea ice export: a closer look
- First Online:
- Cite this article as:
- Tsukernik, M., Deser, C., Alexander, M. et al. Clim Dyn (2010) 35: 1349. doi:10.1007/s00382-009-0647-z
Fram Strait is the primary region of sea ice export from the Arctic and therefore plays an important role in regulating the amount of sea ice and freshwater within the Arctic. We investigate the variability of Fram Strait sea ice motion and the role of atmospheric circulation forcing using daily data during the period 1979–2006. The most prominent atmospheric driver of anomalous sea ice motion across Fram Strait is an east–west dipole pattern of Sea Level Pressure (SLP) anomalies with centers of action located over the Barents Sea and Greenland. This pattern, also observed in synoptic studies, is associated with anomalous meridional winds across Fram Strait and is thus physically consistent with forcing changes in sea ice motion. The association between the SLP dipole pattern and Fram Strait ice motion is maximized at 0-lag, persists year-round, and is strongest on time scales of 10–60 days. The SLP dipole pattern is the second empirical orthogonal function (EOF) of daily SLP anomalies in both winter and summer. When the analysis is repeated with monthly data, only the Barents center of the SLP dipole remains significantly correlated with Fram Strait sea ice motion. However, after removing the leading EOF of monthly SLP variability (e.g., the North Atlantic Oscillation), the full east–west dipole pattern is recovered. No significant SLP forcing of Fram Strait ice motion is found in summer using monthly data, even when the leading EOF is removed. Our results highlight the importance of high frequency atmospheric variability in forcing Fram Strait sea ice motion.
KeywordsFram Strait Sea ice motion Sea ice export North Atlantic Oscillation East–west dipole High frequency atmospheric variability
Fram Strait, located between Greenland and Svalbard, is the primary gateway for the export of sea ice out of the Arctic (e.g., Kwok et al. 2004). Fram Strait sea ice export is highly variable from day to day and from year to year (Vinje 2001; Brummer et al. 2001, 2003; Kwok 2009). Such high variability affects other components of the Arctic climate system: for example, anomalous Fram Strait export has been linked to the “Great Salinity Anomaly” in the North Atlantic (Dickson et al. 1988) and to the recent decline of summer sea ice extent (Rigor and Wallace 2004).
The relationship between the large-scale patterns of atmospheric variability especially the North Atlantic Oscillation (NAO; Hurrell 1995) and the related Arctic Oscillation (AO; Thompson and Wallace 1998) with sea ice export through Fram Strait has been investigated in numerous studies, for example: Kwok and Rothrock 1999; Hilmer and Jung 2000; Jung and Hilmer 2001; Vinje 2001; Rigor et al. 2002; Kwok et al. 2004. During the last two decades of the twentieth century (e.g., 1978–1997) the correlation between the NAO and sea ice export through Fram Strait was highly positive (e.g., Hilmer and Jung 2000, Kwok et al. 2004); however, the correlation during other time periods (e.g., 1958–1977) was near zero or even slightly negative (Hilmer and Jung 2000; Vinje 2001; Jung and Hilmer 2001).
Given the ambiguity in the relationship between the NAO/AO and Fram Strait sea ice export, Wu et al. (2006) and Wu and Johnson (2007) investigated whether other patterns of atmospheric variability are related to the ice export in winter. Wu et al. (2006; see also Koenigk et al. 2006) identified an east–west dipole pattern with centers of action over the Kara/Laptev Seas and the Canadian Archipelago to be an important forcing for sea ice export through Fram Strait, while Wu and Johnson (2007) argued that another pattern with a center of action over the Barents Sea plays even a bigger role. Maslanik et al. (2007) indicated that the strength and position of the centers of action of atmospheric circulation variability associated with sea ice motion within the Arctic basin are affected by cyclone frequency and strength, and that both factors vary considerably from year to year.
To examine the link between sea ice export and atmospheric circulation patterns in more detail, Brummer et al. (2003) analyzed how a single cyclone passing through Fram Strait influences sea ice motion. They found that ice velocity increased by a factor of three during the passage of the cyclone, and that the ratio of ice drift to wind speed also increased. Brummer et al. (2001) analyzed 16 years of cyclone statistics from ERA-40 and corresponding sea ice drift observations. They found that sea ice motion is quite sensitive to the particular cyclone trajectory and concluded that, on average, cyclones increase sea ice export through Fram Strait. Rogers et al. (2005) investigated the role of winter cyclones in Fram Strait sea ice export and found a correspondence between increased cyclogenesis along the northeast coast of Greenland and low sea ice export. High sea ice export years, on the other hand, corresponded to the persistent cyclones in the Norwegian and Barents Seas. Using a case study approach, Tsukernik (2007) illustrated how a particular cyclone trajectory influences sea ice motion: a cyclone passing through Fram Strait can completely reverse the direction of sea ice export, while a cyclone passing east of Fram Strait dramatically increases the sea ice export. The relationship between Fram Strait sea ice flux and the SLP gradient across the Strait has been noted in several studies, including Vinje (2001), Widell et al. (2003), and Kwok (2009). In particular, Widell et al. (2003) found that the SLP gradient explained approximately 60% of the variability in Fram Strait sea ice motion in both daily and monthly averaged data.
Although the topic of atmospheric influence on Fram Strait sea ice export has received a lot of attention, there is still a gap between the monthly averaged studies that relate the sea ice export to large-scale atmospheric patterns and the synoptic-scale studies that investigate the role of high frequency atmospheric disturbances in sea ice export. To bridge this gap, we use daily data to investigate the relationship between the atmospheric circulation and sea ice export over a range of time scales. Due to the scarcity of sea ice thickness measurements, we focus on the areal flux of sea ice through Fram Strait based on satellite estimates of sea ice motion. We investigate the spatial structure and temporal evolution of the SLP patterns associated with variations in sea ice motion through Fram Strait, including its seasonal and frequency dependence. This paper is organized as follows. In Sect. 2 we describe the datasets and methods used in this study. In Sect. 3 we present main results, and in Sect. 4 we summarize our results and discuss them along with findings from previous research.
2 Data and methods
We use linear correlation and regression analysis to define anomalous SLP conditions associated with changes in sea ice export. The statistical significance of the correlation and regression values is assessed using a 2-sided student t test, taking into account the autocorrelation of both series (Press et al. 1986). In order to investigate the relationship between the atmospheric circulation and sea ice export on different timescales we perform cross-spectrum analysis (Bloomfield 1976) and estimate the 99% significance level following Julian (1975) which takes auto-correlation into account. Based on the cross-spectrum results, we define a band pass filter with half-power points at 10 and 60 days (Duchon 1979).
To investigate the seasonal dependence of the sea ice–atmosphere relationship, we divide record into two seasons: winter (15 October–14 April) and summer (15 April–14 October). We subsequently apply all of the techniques described above to the two seasons separately. As the cross-spectrum can only be calculated for a continuous time period, we calculate the spectrum for each winter and summer during 1979–2006 separately and then average individual power spectra together.
3.1 Daily data
Figure 1 depicts the correlation coefficient map between SLP north of 40°N and the Fram Strait sea ice motion index based on 10,227 days of data during 1979–2006. Due to the large sample size, correlation coefficients exceeding ~0.05 in absolute value are statistically significant at the 99% level (outlined by white contour in Fig. 1). There are two main centers of action associated with the anomalous sea ice motion: one over Barents Sea and another one over northern Greenland and Canadian Archipelago. As the sign of the correlation coefficients suggest, southward Fram Strait sea ice motion is maximized with a Barents Sea Low and a Greenland High. Such an east–west dipole pattern is associated with geostrophic northerly winds in Fram Strait and therefore is physically consistent with increased sea ice transport. As previous studies have indicated, sea ice in the Arctic Ocean moves nearly parallel to the geostrophic wind (Thorndike and Colony 1982; Kimura and Wakatsuchi 2000).
We employ the two centers of action revealed by the correlation coefficient map (Fig. 1) to construct a SLP gradient index (SLPGI). For simplicity we define the centers as rectangular boxes, both outlined in Fig. 1. The Barents center of action stretches from 72.5° to 77.5°N and from 17.5° to 50.0°E, and the Greenland center of action occupies the area from 75.0° to 80.0°N and from 60.0° to 42.5°W. We define the SLPGI as the difference between the two. Our results are not sensitive to the exact definition of the Barents and Greenland centers of action—bigger and smaller boxes defining the SLPGI provide similar results (not shown). It is interesting to note that the two centers of action are not significantly correlated with one another (correlation coefficient is 0.1 based on daily SLP anomalies during 1979–2006). However, when the variability associated with the leading EOF is removed from the data, the two centers of action become significantly anti-correlated (correlation coefficient is -0.4, significant at the 99% level). We interpret this as competing influences on the two centers of action, with EOF1 contributing to an in-phase relationship and EOF2 to an out-of-phase connection (e.g., dipole).
Because the Greenland center of action encompasses the region near Iceland (see Fig. 9, middle row, −2 days to +2 days lags) and because the regression values near the Azores (40°N, 30°W) are of opposite polarity, a statistical relationship exists between the NAO-index and the sea ice motion through Fram Strait. To examine this association in more detail, we develop an NAO-like index based on the difference between the Icelandic (55°−65°N and 40°−10°W) and Azores centers of actions (35°–45°N and 40°–20°W). Note that our sign convention is opposite to the traditional definition of the NAO (Hurrell 1995). The lag regression between the NAO-like index and the sea ice motion index based on unfiltered and 10–60 day filtered data for winter only are depicted by green curves in Figs. 5 and 10, respectively. As evident from these figures, the NAO-like relationship with sea ice motion is much weaker than that of the SLPGI, although significant at the 99% level.
3.2 Monthly data
The seasonal structure of the monthly regression map is also noticeably different from that of the daily regression map. With monthly data, the Barents center of action is active in winter only, while no significant relationship between SLP and sea ice motion exists in summer. The latter result is consistent with previous studies that found that summer sea ice export is not correlated with the monthly averaged atmospheric wind forcing (e.g., Kwok et al. 2004; Wu et al. 2006; Wu and Johnson 2007).
Similar results are found for winter (Fig. 13, middle panel), with both centers of the east–west SLP dipole significant at the 99% level. However, the Barents center of action in winter is noticeably weaker when the leading EOF is removed than when it is included (compare Figs. 11, 13, middle panels). This can be partially attributed to the fact that the Barents region is included in the polar center of action of the leading EOF in winter (Fig. 12, middle panel) and thus contributes to the SLP gradient across Fram Strait. In summer (Fig. 13, bottom panel) there is no significant relationship between monthly EOF-residual SLP and Fram Strait sea ice motion. The leading EOF in summer is shifted northward compared to that in winter (Fig. 12, bottom panel). The lack of relationship between monthly SLP anomalies and Fram Strait sea ice motion in summer (Figs. 11, 13, bottom panels) suggests that high-frequency (e.g., sub-monthly) atmospheric variability plays a dominant role in forcing sea ice motion in summer (Fig. 4, bottom panel).
With the help of daily data for SLP and sea ice motion, we found that an east–west dipole pattern with Barents and Greenland centers of action is the most prominent atmospheric driver of sea ice through Fram Strait. The dipole pattern persists year-round, being slightly stronger in winter than in summer. The strongest relationship between the SLP dipole pattern and Fram Strait sea ice motion is simultaneous, with an e-folding time of ~5 days. Spectral analysis shows maximum coherence values in the 10–60 day band. Such a time scale suggests that both high- and low-frequency atmospheric patterns are essential in driving sea ice out of the Arctic.
Atmospheric circulation variability on the 10–60 day time scale has been described in previous studies in association with blocking events in high latitudes (e.g., Michelangeli and Vautard 1998) and westward propagating planetary-scale perturbations (Doblas-Reyes et al. 2001) also known as Branstator–Kushnir oscillations (Branstator 1987; Kushnir 1987). The possible link between these atmospheric phenomena and the Barents–Greenland SLP dipole pattern identified in this study merits further investigation. However, we find no evidence of westward propagation of the SLP dipole (not shown). While the east–west dipole projects onto the zonal wave 1 structure of atmospheric variability identified in Cavalieri (2002) at high latitudes (70°–80°N), it is primarily an Atlantic sector pattern farther south (45°–70°N),
Repeating our analysis using monthly data revealed a modified spatial pattern of SLP anomalies associated with Fram Strait sea ice motion. While the Barents center of action remains prominent, the Greenland center of action and therefore the dipole structure of the pattern disappeared. However, removing the leading EOF from the monthly averaged SLP data (e.g., the NAO) resulted in the return of the east-west dipole pattern. Based on these results, we argue that in monthly data the NAO—the leading intrinsic pattern of atmosphere variability—partially masks the relationship between the SLP dipole pattern and the Fram Strait sea ice motion response. That is, the NAO is not the most dynamically relevant pattern for explaining the variations in sea ice motion through Fram Strait. Rather, the east–west SLP dipole pattern is the important driver of the anomalous sea ice motion both in daily and monthly averaged data. These results help explain why previous studies based on monthly data (e.g., Hilmer and Jung 2000; Vinje 2001; Kwok et al. 2004) found no consistent relationship between the NAO and Fram Strait sea ice motion.
This study investigated the role of atmospheric forcing in driving Fram Strait sea ice motion. It will be interesting to extend this study to examine the relationship between atmospheric forcing and Fram Strait sea ice volume flux by incorporating observational estimates of sea ice thickness. Ice volume changes in the Arctic sea ice are crucial for determining the future behavior of sea ice extent and important for linking the thermodynamic and dynamic components of sea ice change (Holland et al. 2008). As Rigor and Wallace (2004) have argued, the loss of sea ice extent in recent years was preconditioned by the loss of older and thicker sea ice through Fram Strait in the 1990s. We plan to investigate the processes that triggered the sea ice loss of the 1990s in greater detail,
We thank the reviewers for their insightful comments and suggestions that substantially improved the manuscript. We also thank Christophe Cassou for helpful suggestions, and Adam Philips and Dennis Shea for technical assistance in preparation of the figures. This work was supported by a grant from the National Science Foundation Arctic System Science Program. The National Center of Atmospheric Research is sponsored by the National Science Foundation.