1. Introduction
Arctic cyclones during the summer months occasionally cause large reductions in sea ice cover in just a few days (Zhang et al. 2013; Stern et al. 2020; Wang et al. 2020; Lukovich et al. 2021), making them important drivers of short-term sea ice variability. Wang et al. (2020) found that rapid ice loss events associated with Arctic cyclones have become more frequent in recent years. Given that stronger cyclones tend to be more destructive to sea ice during the melt season (e.g., Zhang et al. 2013; Lukovich et al. 2021), it is plausible that an increase in the intensity or frequency of Arctic cyclones would explain this increased likelihood for rapid ice loss events. However, several climatological studies of Arctic cyclones find no significant increase in the frequency or intensity of Arctic summer cyclones in recent decades (Koyama et al. 2017; Zahn et al. 2018; Vessey et al. 2020). Crawford et al. (2022) showed that an increase in surface enthalpy fluxes from the ocean due to declining Arctic sea ice cover has led to stronger cyclones only during the winter months. Therefore, there is little evidence to support the idea that, during the summer months, stronger or more frequent cyclones are responsible for the recent increase in rapid sea ice loss events. In this study, we will show that cyclones at the end of the melt season are becoming more destructive to sea ice not because they are getting stronger or more frequent, but because the ice is getting thinner and less concentrated and the upper ocean is getting warmer. Both of these changes have made Arctic sea ice cover more vulnerable to atmospheric forcing during the melt season.
Synoptic-scale cyclones introduce transient anomalies in surface wind and energy forcing that impact sea ice on short time scales. Increased cloud cover and warm-air advection associated with Arctic cyclones in spring can trigger the onset of the melt season by temporarily increasing the downward flux of longwave radiation at the surface and raising surface temperatures above the freezing point (Persson 2012; Maksimovich and Vihma 2012; Liu and Schweiger 2017). In June and July, when the sun angle is at its highest and melting snow and ice begins to lower the surface albedo, the importance of shortwave radiation in the surface energy budget over sea ice briefly surpasses that of longwave radiation (Kay et al. 2008; Perovich 2018). As a result, clear-sky conditions associated with anticyclones early in the melt season accelerate seasonal sea ice loss (Wernli and Papritz 2018), while increased cloud cover from early-summer cyclones locally slows down seasonal ice loss (Schreiber and Serreze 2020; Finocchio et al. 2020; Finocchio and Doyle 2021). By August, however, cyclones once again become capable of locally accelerating sea ice loss due to warm air advection over sea ice (Park et al. 2015; Woods and Caballero 2016; Stern et al. 2020; Fearon et al. 2021) and turbulent mixing of relatively warm seawater from just beneath the oceanic mixed layer (Jackson et al. 2010; Steele et al. 2010; Zhang et al. 2013; Stern et al. 2020). Additionally, the lower concentration sea ice at the end of the melt season is more prone to drift in response to strong surface winds (Lei et al. 2020), which further mixes the upper ocean and facilitates sea ice divergence. The reduction in sea ice concentration and the warming of the upper ocean over the course of the melt season thus enables Arctic cyclones to become more destructive to sea ice as summer progresses.
Because the sea ice response to weather forcing depends strongly on the antecedent ice and upper-ocean conditions, it stands to reason that the large changes in the sea ice pack and upper ocean due to climate change in recent decades have significantly altered the short-term impacts of Arctic cyclones on sea ice. Thinner and/or lower-concentration sea ice at the beginning of the melt season allows for more insolation to be absorbed at the surface around the solar maximum, leading to greater ice melt and upper-ocean warming under clear-sky conditions (Perovich et al. 2007). Increased upper-ocean warming early in the melt season results in warmer seawater that can be mixed upward in response to Arctic cyclones later in the melt season. Furthermore, a transition toward thinner sea ice in recent years is likely to enhance sea ice divergence under cyclonic wind forcing because thinner ice drifts at a greater fraction of the surface wind speed (Spreen et al. 2011; Kwok et al. 2013; Itkin et al. 2017). Based on all of these factors, we hypothesize that the recent changes to the Arctic environment are making cyclones more destructive to sea ice.
Few studies have examined how the short-term impacts of Arctic cyclones on sea ice have changed in recent decades. As mentioned above, Wang et al. (2020) found that the frequency of rapid ice loss events during summer has increased in recent years in the central Arctic and in the Beaufort, Chukchi, and East Siberian Seas. They attributed the more frequent rapid ice loss events to lower sea ice concentrations and a warmer lower atmosphere, rather than to changes in the Arctic cyclones. Schreiber and Serreze (2020) computed trends in the four-day reductions in sea ice concentration (SIC) following Arctic cyclones between June and August from 1979 to 2018. They found that areas of sea ice impacted by summer cyclones in recent years still tend to experience lower rates of SIC loss in the ensuing days than sea ice that is outside of a cyclone’s influence. However, they found that the 4-day SIC loss in cyclone-influenced areas is increasing at a faster pace, especially within the Chukchi and East Siberian Seas (their Table 2). A drawback of their approach is that they composite cyclone impacts over June, July, and August, which could average together opposing trends in the earlier and later parts of the melt season.
In this study, we examine how the impact of Arctic cyclones on sea ice has changed in recent years by comparing the 1–7-day changes in sea ice area and thickness following cyclones from two recent decades: 1991–2000 and 2009–18. We utilize multiple reanalysis and observational datasets of the atmosphere, ocean, and sea ice (described in section 2) to accomplish this. Because cyclone impacts on sea ice evolve considerably from June through August (Finocchio et al. 2020), we composite our results by month rather than over the entire summer. Using this approach, we reveal that only August cyclones are becoming significantly more destructive to sea ice on short time scales. In contrast, we find no significant changes in sea ice loss following June cyclones between the two decades, but significantly more sea ice loss in the absence of June cyclones in the more recent decade. These results suggest that recent climate change in the Arctic has made relatively quiescent conditions at the beginning of the melt season and stormy conditions at the end of the melt season more likely to locally accelerate seasonal sea ice loss.
2. Data and methods
a. Atmospheric reanalysis and sea ice concentration datasets
The atmospheric and SIC data for this study come from the ECMWF fifth-generation reanalysis (ERA5; Hersbach et al. 2020). All variables in ERA5 are available on a common 0.25° latitude–longitude grid. Sea ice concentration in ERA5 originates from the real-time European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) Satellite Application Facility on Ocean and Sea Ice (OSI-SAF) product (Hirahara et al. 2016) and is updated daily at 0000 UTC. Sea ice concentration from ERA5 has been shown to be highly correlated (correlation coefficient r > 0.9) with SIC from the NASA Team algorithm during the melt season (Crawford et al. 2022).
b. Arctic cyclone case selection
We utilize a database of cyclone position and intensity that was generated by Michael Sprenger (ETH Zürich) using the Sprenger et al. (2017) cyclone tracking algorithm applied to ERA5 atmospheric fields from 1991 to 2018. Although we initially considered cyclones in this database from June, July, and August, we found that the sea ice response to July cyclones has intermediate characteristics of the responses to June and August cyclones. For the sake of brevity, we only present results from June and August so as to compare cyclone impacts on sea ice in the earlier and later parts of the melt season.
All cyclones in the database are located north of 70°N and maintain a minimum sea level pressure of ≤1000 hPa for at least 48 consecutive hours. From these cyclones, we select cases that are located over grid points with SIC > 0 in order to guarantee that the cyclones analyzed impact sea ice. As a further geographical constraint, we only consider cyclones located from the prime meridian eastward to 120°W. We focus on this region of the Arctic because it is mostly covered by thinner seasonal sea ice, it exhibits large interannual variability in sea ice cover (Finocchio and Doyle 2021), and it has experienced dramatic declines in summer ice cover in recent decades (Deser and Teng 2008; Ogi and Rigor 2013). We further subdivide this region into the Eurasian sector (0°–140°E), which includes the Barents, Kara, and Laptev Seas, and the Amerasian sector (140°E–120°W), which includes the East Siberian, Chukchi, and Beaufort Seas. The analysis domain excludes the region north of Greenland and within the Canadian Arctic Archipelago because cyclones in this region primarily affect thick, multiyear, and land-fast ice that is less responsive to transient atmospheric forcing. The domain also excludes the Greenland Sea because the large-scale export of sea ice through the Fram Strait governs the short-term evolution of ice cover in this region (Spreen et al. 2020). Within the selected analysis domain from 1991 to 2018, about 70% of the candidate cyclone cases in June have their centers located over SIC > 0. This fraction decreases to 51% of cases in August because of the larger open water areas within the Arctic later in the melt season.
To understand how Arctic cyclone impacts on sea ice during the summer months have changed in response to recent climate change, we compare June and August cyclones from two different 10-yr intervals. The first interval is from 1991 to 2000 (“early decade”) and the second is from 2009 to 2018 (“recent decade”). These two decades are chosen not only because they represent the earliest and most recent decades available in the ERA5 cyclone database, but also because they bracket a period of remarkable transformation in the Arctic environment. Sea ice loss in the Arctic began to markedly accelerate starting in the late 1990s and early 2000s (Deser and Teng 2008; Ogi and Rigor 2013; Choi et al. 2019), with average ice thickness across the Arctic basin estimated to be declining at a rate of 6 cm per year from 2000 to 2012 (Lindsay and Schweiger 2015). This acceleration of summer ice loss has been attributed to atmospheric warming, as well as to notable shifts in the mean atmospheric circulation pattern. Since the start of the twenty-first century, the large-scale wind pattern in the Arctic has more frequently exhibited meridional flow, which promotes the reduction of sea ice extent by enhancing poleward heat transport (Zhang et al. 2008) and ice export through the Fram Strait (Overland et al. 2012; Kwok et al. 2013; Choi et al. 2019; Heo et al. 2021). Comparing cyclone cases from the early and recent decades chosen for this study enables us to examine how sea ice impacts and the underlying physical processes have changed since the Arctic began this rapid transition to its present state.
Figure 1 shows the selected cyclone cases in June and August from each decade considered in this study (red dots) as well as the SIC averaged over all of the cyclone days (shading). There are slightly more cyclones in June and in the early decade than in August and in the recent decade. This is because we require cyclones to be located over SIC > 0, and the ice-covered area in the Arctic declines seasonally and from the early to the recent decades. Consistent with previous studies (e.g., Crawford and Serreze 2016; Fearon et al. 2021), cyclones in June are more frequent on the western (Eurasian) side of the Arctic domain, while cyclones in August are more frequent on the eastern (Amerasian) side of the domain.
Cyclone cases (red dots; size corresponds to intensity) and average sea ice concentration on cyclone days (shading) in (top) June and (bottom) August (a),(c) from 1991 to 2000 and (b),(d) from 2009 to 2018. The black contours are the average 15% and 80% ice concentration contours that bound the marginal ice zone. The number of cyclone cases is given in the upper right of each panel. Outlined regions denote the Eurasian and Amerasian sectors defined in the text.
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
Figure 2 shows distributions of the intensity (minimum sea level pressure; Figs. 2a,d), latitude (Figs. 2b,e), and longitude (Figs. 2c,f) at the centers of selected cyclone cases in each decade in June and in August. August cyclones are slightly stronger in terms of minimum sea level pressure (MSLP) and are located slightly more north and east than June cyclones. The eastward shift in cyclone positions from June to August is also apparent in Fig. 1. The focus of this analysis, however, is on how the location and intensity of cyclones in each month has changed from the early to the recent decade. The distributions of cyclone intensity, latitude, and longitude are not normally distributed, so we use the nonparametric Mann–Whitney U test and the Kolmogorov–Smirnov test for evaluating differences in the means and distributions, respectively, of each quantity between the early and recent decade (note that we will use nonparametric significance tests in place of t tests throughout this paper because of the nonnormal distributions of the datasets being compared). We find no significant change in the means or distributions of cyclone intensity from the early to the recent decade in either month (Figs. 2a,d). Because this is only a subset of Arctic cyclones, we cannot conclude from Fig. 2 that the average intensity of all Arctic cyclones has not changed significantly from 1991 to 2018, but only that there is no significant difference in cyclone intensity between the early and recent decades in this sample of cyclone cases.
Distributions of cyclone (left) minimum sea level pressure, (center) latitude, and (right) longitude in (a)–(c) June and (d)–(f) August. Blue lines correspond to cyclone cases from 1991 to 2000 (early decade), and red lines correspond to cases from 2009 to 2018 (recent decade). Vertical lines denote the average of each quantity. The values in the upper right of each panel indicate the confidence level at which averages from each decade differ according to a nonparametric Mann–Whitney U test (MW), the confidence level at which distributions from each decade differ according to a standard Kolmogorov–Smirnov test (KS), and the average confidence level from 1000 KS tests applied to different equal-sized random subsets of cyclone cases from each period [i.e., bootstrap (BS)].
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
The only variable shown in Fig. 2 for which there is a statistically significant difference in the averages (significance level p < 0.05) between the early and recent decades is cyclone latitude in August, which is more poleward in the recent decade. However, this poleward shift in cyclone cases is likely an artifact of the method of excluding cyclones that are not over sea ice, and it simply reflects the poleward retreat of the sea ice edge from the early to the recent decade. Moreover, despite the differences in the means, the overall frequency distributions of August cyclone latitude between the early and recent decades are not significantly different according to a standard Kolmogorov–Smirnov (K-S) test or a bootstrapped K-S test (see Fig. 2 caption).
The key point from Fig. 2 is that the intensity of the selected cyclone cases in both June and August has not changed significantly between the two decades. Based on this, we do not believe that differences in cyclone intensity between the two decades can explain the changing impacts on sea ice that we will show in section 3.
c. Sea ice motion and thickness datasets
For computing sea ice motion and divergence, we utilize daily ice motion vectors from version 4.1 of the Polar Pathfinder sea ice motion vector dataset (Tschudi et al. 2020). We remapped these data from the native 25-km Equal Area Scalable Earth (EASE) grid to the ERA5 grid using bilinear interpolation. We also examine the sea ice thickness evolution using output from the Pan-Arctic Ice-Ocean Modeling and Assimilation System (PIOMAS; Zhang and Rothrock 2003), which assimilates ice and ocean observations into a coupled ice–ocean model forced with the NCEP–NCAR atmospheric reanalysis. Because the atmospheric conditions in NCEP–NCAR can differ from ERA5, cyclones in PIOMAS may not always be in the same location or have the same intensity as in ERA5. Consequently, the PIOMAS sea ice thickness evolution will occasionally be inconsistent with what would occur if the atmospheric forcing had been from ERA5. Nevertheless, assuming that significant cyclones are in similar locations in NCEP–NCAR and ERA5, using PIOMAS enables us to at least estimate the sea ice thickness response to cyclones, which is not possible using ERA5.
d. Upper-ocean datasets
The ocean plays a key role in short-term sea ice loss during the summer, especially at the end of the melt season, when cyclones can melt sea ice by mixing warm water upward from beneath the mixed layer (Steele et al. 2010; Zhang et al. 2013; Stern et al. 2020; Peng et al. 2021). Therefore, we also examine the upper-ocean thermal structure and the changes that have taken place between the two decades using upper-ocean temperature data from the 1/12° Global Ocean Forecast System, version 3.1 (GOFS3.1), reanalysis (https://www.hycom.org/dataserver/gofs-3pt1/reanalysis; Metzger et al. 2017), and from the Copernicus 1/12° Global Ocean Eddy-Resolving Reanalysis (GLORYS; Lellouche et al. 2021).
The GOFS3.1 reanalysis assimilates ocean observations daily into the global Hybrid Coordinate Ocean Model (HyCOM) using the Navy Coupled Ocean Data Assimilation (NCODA) system (Cummings 2005; Cummings and Smedstad 2013). HyCOM in GOFS 3.1 is coupled to version 4.0 of the Community Ice Code (CICE) model (Hunke et al. 2015) and uses hourly atmospheric boundary conditions from the NCEP Climate Forecast System Reanalysis (CFSR) through 2012 and from CFSRv2 after 2012. Output from GOFS3.1 is available from 1994 to 2015. We use daily ocean temperature and sea ice thickness fields at 0000 UTC. Although ocean temperature is available at 41 vertical levels, we only use the 19 levels from the sea surface to a depth of 90 m because wind-induced ocean mixing is not observed to extend below this depth (Timmermans 2015; Smith et al. 2018).
The GLORYS reanalysis utilizes the Nucleus for European Modeling of the Ocean (NEMO; Madec et al. 2008), which consists of interacting global ocean and ice model components at 1/12° horizontal resolution. The NEMO ocean model in GLORYS assimilates a similar set of ocean observations as GOFS3.1, using the reduced-order Kalman filter of Brasseur and Verron (2006). GLORYS uses 3-hourly atmospheric boundary conditions from the ERA-Interim reanalysis (Dee et al. 2011). Although GLORYS is available from 1993 to near the present day, we only use daily output from the period that overlaps with the GOFS3.1 reanalysis (1994–2015). The ocean output from GLORYS has somewhat higher vertical resolution in the upper ocean than GOFS3.1, and we use the 22 levels that extend from the sea surface to a depth of 92 m.
We assess the accuracy of the output from both ocean reanalyses by comparing upper-ocean temperature profiles with observations from ice-tethered profilers (ITPs; Toole et al. 2011; Krishfield et al. 2008) in the Canada Basin. We also compare the sea ice thickness from both reanalyses with the ice thickness from the PIOMAS reanalysis, which has been shown to provide ice thickness estimates that are generally within 1 m of satellite, airborne, and submarine observations (Schweiger et al. 2011).
We note that the ocean temperature output in GOFS3.1 is the in situ temperature, while that in GLORYS is potential temperature. For the upper 100 m of the ocean considered in this study, the difference between potential and in situ temperature is O(0.01)°C, which is negligible relative to the differences in the average temperature profiles between GOFS3.1 and GLORYS and between each product and ITP temperature profiles. Therefore, we use the respective raw “temperature” output from each product in our analysis. We also note that GOFS3.1 and GLORYS both use atmospheric boundary conditions that differ from ERA5. However, because we are primarily interested in comparing the ocean initial conditions rather than the ocean response to cyclone forcing, we do not believe that the use of different atmospheric boundary conditions affects this analysis considerably.
3. Results
a. Total sea ice area response
We start by considering how the total sea ice area over the entire domain considered in this study (60°–90°N, 0°–120°W; Fig. 1) evolves following cyclone days. This provides a sense for whether cyclones have a noticeable impact on the short-term variability of pan-Arctic sea ice area (SIA). We compare SIA changes following cyclone days with those following noncyclone days, which we define simply as days without cyclone cases anywhere in the domain. Figure 3 shows the 1–7-day change in SIA following cyclone days (solid lines) and noncyclone days (dashed lines) in each decade. Large dots in Fig. 3 indicate when the cyclone and noncyclone averages at a particular lag time are significantly different from each other (p < 0.05) according to a nonparametric Mann–Whitney U test. In June (Fig. 3a), cyclone days tend to be followed by less total SIA loss than noncyclone days in both the early and recent decades, which is consistent with Schreiber and Serreze (2020) and Finocchio et al. (2020). In August, noncyclone days are only associated with more SIA loss than cyclone days in the early decade (significant at 1-, 3-, 6-, and 7-day lags; Fig. 3b). In the recent decade, however, cyclone days in August are followed by more SIA loss than noncyclone days (significant at 1–4-day lag).
Average change in sea ice area 1–7 days after cyclone days (solid lines) and noncyclone days (dashed lines) in (a) June and (b) August in the region from 60° to 90°N and from 0° to 120°W. Blue lines correspond to the period from 1991 to 2000 (early decade), and red lines correspond to 2009–18 (recent decade). The shaded regions (noncyclone days) and error bars (cyclone days) represent the 95% confidence intervals around the mean values. Thick dots are drawn at lag days for which the differences between cyclone and noncyclone SIA change is significant (p < 0.05) according to a Mann–Whitney U test.
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
Although the impacts of cyclones on total SIA within this large region of the Arctic are occasionally significant at the 95% confidence level, the difference between the average ice loss following cyclone versus noncyclone days is almost an order of magnitude smaller than the total ice area loss at each lag time. Cyclone impacts on sea ice relative to seasonal ice loss might be more substantial if we only consider the local area around the cyclones, as is done below.
b. Local sea ice response
To only account for the local impacts of cyclones on sea ice, we compute the 1–7-day changes in sea ice area and thickness within a 500-km radius of cyclone positions (we also tested radii of 400 and 600 km and find qualitatively similar results). For this analysis, we adopt a different technique for defining the noncyclone cases: for each cyclone case, we search the cyclone database for other nearby cyclone cases on the same day of the year, but in different years. The ERA5 fields from the nearest year in the database for which there is no cyclone within 500 km of the current cyclone position are used to define the noncyclone case. This technique produces a corresponding noncyclone case at the same location as each cyclone case. For about 12% of cases, the year of the noncyclone case differs from that of the cyclone case by more than 10 years. However, there are no cases in which a cyclone case from the recent decade has a corresponding noncyclone case from the early decade, or vice versa.
Figure 4 shows the percent change in SIA and sea ice thickness (SIT) within 500 km of cyclone and noncyclone cases in the following 1–7 days. Consistent with Fig. 3a, there is more SIA and SIT loss following noncyclones than cyclones in June (Figs. 4a,c) in both the early and recent decades. The greater amount of ice area and thickness loss following June noncyclones is statistically significant (p < 0.05) at multiple lag times beyond two days according to a paired Wilcoxon test (we use this test instead of the Mann-Whitney U test here because each cyclone case has a corresponding noncyclone case). At 7-day lag, the SIA loss has increased from the early to the recent decade by 67% more around noncyclones than around cyclones in June (Fig. 4a). SIT loss has also increased by a larger amount following noncyclones than following cyclones in the recent decade (Fig. 4c).
Percent change in (a),(b) sea ice area and (c),(d) sea ice thickness within 500 km of all cyclone (solid lines) and noncyclone (dashed line) cases shown in Fig. 1 in (left) June and (right) August. SIA changes are calculated from 1–7 days after cyclone/noncyclone dates. As in Fig. 3, blue lines correspond to the early decade (1991–2000), red lines correspond to the recent decade (2009–18), and shading (for noncyclones) and error bars (for cyclones) represent the 95% confidence intervals. Note that SIT data in (c) and (d) are from the PIOMAS reanalysis (Zhang and Rothrock 2003). The number of cyclone/noncyclone cases in each decade is given in the upper-left corner of each panel. Thick dots are drawn at lag days for which the differences between cyclone and noncyclone SIA change are significant (p < 0.05) according to a paired Wilcoxon test.
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
Unlike in June, cyclones in August tend to be followed by more SIA loss than noncyclones (Fig. 4b, significant at 3–7-day lag in the recent decade). At 7-day lag, the percent reduction in SIA around August cyclones has increased from 3.3% in the early decade to 10.4% in the recent decade—nearly twice the increase in local SIA loss around noncyclones between the two decades. However, the SIT changes following cyclones in August are indistinguishable from SIT changes following noncyclones in both decades (Fig. 4d). We initially attributed this to SIT originating from PIOMAS and SIA originating from ERA5, but SIT loss and SIA loss remain inconsistent with one another even when we compute SIA using PIOMAS (not shown). One possible explanation for this is that dynamic processes such as sea ice divergence have a more significant role than thermodynamic processes in enhancing local SIA loss in PIOMAS following August cyclones, resulting in reductions in ice concentration without noticeable reductions in ice thickness (Clancy et al. 2022).
Overall, Fig. 4 indicates that sea ice loss has increased significantly in the recent decade following the absence of cyclones in June and the presence of cyclones in August. This suggests that changes in the surface energy forcing, sea ice cover, and/or upper-ocean properties in recent decades have made relatively quiescent conditions in the Arctic more destructive to sea ice at the beginning of melt season, and stormy conditions more destructive to sea ice later in the melt season.
To determine whether the enhanced ice loss in June and August originates within specific regions of the Arctic, we repeat the above analysis but only for cyclones in the Eurasian or Amerasian sectors outlined in Fig. 1. Figure 5 shows the average SIA and SIT changes following only cyclone and noncyclone cases within the Eurasian sector (0°–140°E). The accelerated ice loss following noncyclones in June is even more apparent in the Eurasian sector (Figs. 5a,c) than in the whole domain (Figs. 4a,c). In August, however, the average change in SIA is similar following cyclones and noncyclones (Fig. 5b). August cyclones in the Eurasian sector slightly accelerate SIT loss in the early decade (not significant) and decelerate SIT loss in the recent decade (significant at 1–3-day lag; Fig. 5d).
As in Fig. 4, but for the Eurasian sector (0°–140°E).
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
The large difference in the number of Eurasian-sector cyclones between June and August could affect the comparison of sea ice impacts. Therefore, we recomputed the average SIA and SIT changes for five randomly selected sets of 40 Eurasian-sector cyclone and noncyclone cases from each month to determine whether comparing smaller but equal-sized samples of cases changes the results (Fig. S1 in the online supplemental material). In June, we consistently find that noncyclones are followed by more SIA loss than cyclones in the recent decade. Differences between the averages are statistically significant for at least four lag times in all of the randomized trials. In the early decade, the SIA loss following June noncyclones in the Eurasian sector is indistinguishable from that following cyclones at most lag times in all of the randomized trials. In August, the differences between sea ice loss following cyclones and noncyclones in either decade are much smaller than in June and are rarely significant. Therefore, the basic results from Fig. 5 seem to be robust to the different sample sizes in the early versus recent decades and in June versus August.
When we restrict the analysis instead to the Amerasian sector of the Arctic (140°E–120°W), we find that ice loss (SIA or SIT) in June is similar following cyclones and noncyclones in both decades (Figs. 6a,c). In August, however, significantly more SIA loss follows cyclones than noncyclones in the recent decade (Fig. 6b). Upon repeating the analysis with five random subsets of 40 cyclones and noncyclones in each month and in each decade, we consistently find more ice loss following August cyclones in the recent decade than noncyclones, with statistically significant differences evident in four out of the five trials (Fig. S2 in the online supplemental material). This indicates that this result is likely robust to the different sample sizes. SIT in the recent decade also decreases by a larger percentage following cyclones than noncyclones in August, but the difference is only significant at 5–7-day lags (Fig. 6d). In comparing Figs. 5d and 6d, it is apparent that August cyclones in the Eurasian sector have become less destructive to sea ice in terms of PIOMAS SIT, while those in the Amerasian sector have become more destructive. There is not an obvious explanation for this divergent SIT response between the two subregions. Nevertheless, both the SIT and SIA responses in the recent decade support the basic finding that August cyclones are becoming more destructive to sea ice primarily within the Amerasian sector of the Arctic.
As in Fig. 4, but for the Amerasian sector (140°E–120°W).
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
To summarize, we find evidence that recent climate change in the Arctic has increased the tendency for noncyclone conditions in June to enhance the local, 1–7-day loss of sea ice relative to cyclone conditions. By contrast, cyclone conditions in August have become more likely to enhance local sea ice loss relative to noncyclone conditions. These changes in the sea ice responses are geographically distinct in that the enhanced ice loss following the absence of a cyclone in June is confined to the Eurasian side of the Arctic, while enhanced ice loss following cyclones in August is confined to the Amerasian side of the Arctic. The following sections examine the surface atmospheric energy forcing and upper-ocean characteristics that explain these geographically distinct changes in the sea ice response.
c. Atmospheric contributions to local sea ice loss in the Eurasian sector (0°–140°E)
We begin by comparing atmospheric forcing on sea ice between the early and recent decades in the Eurasian sector of the Arctic. Figure 7 shows the surface energy fluxes from the atmosphere in June and in August computed from ERA5 (colors are as in Figs. 2–6). Atmospheric fluxes in ERA5 are positive when they are directed downward toward the surface. Recall that in June in the Eurasian sector, the local ice loss following noncyclones in the recent decade has become significantly larger than that following cyclones (Figs. 5a,c). According to Fig. 7a, differences in the net radiative flux at the surface may explain why this has occurred. The total surface flux (NET) is larger for noncyclones than for cyclones in both decades, which makes sense in June because noncyclone conditions are more likely to be associated with clearer skies and a larger amount of shortwave (SW) radiation reaching the surface. However, the noncyclone net flux has increased from being 6 W m−2 larger than for cyclones in the early decade, to 16 W m−2 larger in the recent decade. All components of the surface energy budget (not just the SW flux) contribute to this larger relative increase in the average energy flux for noncyclones in June.
Distributions of atmospheric terms in the surface energy budget computed from ERA5 averaged within 500 km of cyclones (“C”) and noncyclones (“N”) in each decade in the Eurasian sector (0°–140°E). The terms are (from left to right) the total surface energy flux from the atmosphere (NET), the net shortwave radiative flux (SW), the net longwave radiative flux (LW), and the surface heat flux (Heat), which is the sum of the sensible and latent heat flux. All terms are averaged over 1200 and 1800 UTC on the previous day and 0000 and 0600 UTC on the cyclone/noncyclone day to reduce the effect of the diurnal cycle on the results. Positive values are downward (into the surface). Large dots are the averages over all cases, and errors bars represent the 95% confidence interval.
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
In August (Fig. 7b), surface fluxes from the atmosphere in the Eurasian sector are more similar between cyclones and noncyclones and between the early and recent decades than they are in June. Part of the reason for this is that the difference in the amount of SW radiation reaching the surface in clear-sky versus cloudy conditions is smaller in August than in June (Finocchio and Doyle 2021). As will be shown below, cyclones are accompanied by a marked increase in cloud cover relative to noncyclones. This increased cloud cover associated with cyclones reduces the average SW flux relative to noncyclones by about 15 W m−2 in June in both decades, but by no more than 6 W m−2 in August.
The differences in surface fluxes between June and August and between the early and recent decades in the Eurasian sector can be better understood by examining the average surface properties and cloud cover characteristics within this region of the Arctic. Figure 8 shows the distributions of SIC, SIT, midlevel cloud fraction, and surface albedo averaged within 500 km of cyclone and noncyclone cases in the Eurasian sector. We focus on midlevel clouds in ERA5 (∼800–450 hPa) because midlevel clouds have a substantial radiative impact at the surface during summer (Curry et al. 1993) and because cyclones are associated with larger relative increases in midlevel cloud fraction than low-level cloud fraction, which is consistently high across the Arctic during summer (Eastman and Warren 2010). In June (Fig. 8a), the average SIC decreases by 5.3% for cyclones and by 12.2% for noncyclones from the early to the recent decade. The average SIT correspondingly decreases by 5.1 cm for cyclones and 7.2 cm for noncyclones between the two decades. As a result of these significant declines in ice concentration and thickness, the area-averaged surface albedo in June decreases from 52% to 48% for cyclones, and from 53% to 45% for noncyclones between the two decades. These reductions bring the surface albedo within the Eurasian sector well below the June “break-even albedo” of 54% (Perovich 2018), resulting in a regime where more energy is absorbed at the surface under clear skies than under cloudy skies. Figure 8a shows that the average midlevel cloud fraction for noncyclones (∼33%) is about half of that for cyclones (∼64%) in both decades. The significantly lower surface albedo in the recent decade therefore enables more of the abundant downward SW radiation during noncyclone conditions in June to be absorbed at the surface, likely resulting in greater ice melt and upper-ocean heating (Perovich et al. 2007).
As in Fig. 7, but the depicted terms are (from left to right) sea ice concentration (SIC), sea ice thickness (SIT; dm), midlevel (800–450 hPa) cloud fraction (Mid Cld), and surface albedo (Sfc Alb). All terms are computed at 0000 UTC on the cyclone/noncyclone date using ERA5 data except for SIT, which is computed from PIOMAS output (Zhang and Rothrock 2003).
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
In contrast to June, the average surface albedo in August in the Eurasian sector has not changed significantly between the early and recent decades for either cyclones or noncyclones (Fig. 8b). This is due in part to smaller reductions in August SIC and SIT from the early to the recent decade than in June. The similar surface albedo between the early and recent decade, coupled with the weaker cloud shortwave radiative forcing in August, likely explains why the absorbed SW flux at the surface in Fig. 7b has remained fairly consistent in each decade, and by extension, why the August sea ice response has not changed significantly under cyclone and noncyclone conditions (Fig. 5b).
d. Atmospheric contributions to local sea ice loss in the Amerasian sector (140°E–120°W)
In the Amerasian sector, June SIA reductions are similar following cyclones and noncyclones (Fig. 6a). Moreover, June SIA reductions in this sector have not changed considerably from the early to the recent decade. One explanation for this is that the surface flux differences, both between cyclones and noncyclones and between the early and recent decades, are much smaller in this part of the Arctic. The average surface flux in June for noncyclones only exceeds the average cyclone flux by 3–5 W m−2 in the Amerasian sector as compared with 6–16 W m−2 in the Eurasian sector, and the magnitude of these differences has actually decreased slightly in the recent decade (Fig. 9a). Furthermore, the average June SIC is about 10% higher in the Amerasian sector than in the Eurasian sector, which translates to an average surface albedo that is between 5% and 10% higher (Fig. 10a vs Fig. 8a). The higher average albedo is above the June “break-even” value of 54% (Perovich 2018)—even in the recent decade—meaning that more energy is absorbed at the surface under cloudy conditions than under clear-sky conditions. Unlike in the Eurasian sector where both SIC and surface albedo have decreased significantly from the early to the recent decade (Fig. 8a), SIC and surface albedo in the Amerasian sector have remained fairly constant between the two decades (Fig. 10a). This may explain why noncyclone conditions in June have not become significantly more destructive to sea ice in the recent decade in the Amerasian sector like they have in the Eurasian sector.
As in Fig. 7, but for the Amerasian sector (140°E–120°W).
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
As in Fig. 8, but for the Amerasian sector (140°E–120°W).
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
In August, substantial reductions in late-summer ice cover in the Amerasian sector over the past few decades have resulted in average SIC, SIT, and surface albedo values that are similar to the Eurasian sector (Fig. 10b vs Fig. 8b). As a result, the average energy flux from the atmosphere in the Amerasian sector has also become similar to that in the Eurasian sector in August in the recent decade (Fig. 9b vs Fig. 7b). In the Eurasian sector, we attributed the similarity in sea ice loss following cyclones and noncyclones to the similar surface energy forcing from the atmosphere. If the atmospheric energy forcing were the only factor governing sea ice loss, then we would also expect to find similar amounts of sea ice loss following August cyclones and noncyclones in the recent decade in the Amerasian sector, but this is not the case (Fig. 6b). Factors other than the thermodynamic forcing from the atmosphere must therefore be responsible for the relative increase in sea ice loss following August cyclones when compared with noncyclones in the Amerasian sector during the recent decade.
The dynamic (wind) forcing on sea ice can be particularly important for driving local sea ice changes in August, when there is more low-concentration sea ice that is susceptible to wind-induced deformation (Spreen et al. 2011; Kwok et al. 2013; Itkin et al. 2017; Clancy et al. 2022). Figure 11 shows distributions of 10-m wind speed, sea ice speed, and sea ice divergence averaged within 500 km of August cyclones and noncyclones in the Eurasian sector (Fig. 11a) and the Amerasian sector (Fig. 11b). Sea ice divergence and speed are both computed from the National Snow and Ice Data Center (NSIDC) sea ice motion vector dataset (Tschudi et al. 2020). The average ice speed during cyclones has increased significantly in the Amerasian sector, from 7.6 cm s−1 in the early decade to 9.2 cm s−1 in the recent decade, but it has remained fairly constant in the Eurasian sector (6.7 vs 6.8 cm s−1). The faster ice speed in the Amerasian sector in recent years is not a result of faster surface wind speeds during cyclones, which have actually decreased slightly in the recent decade according to Fig. 11b. The ice drift speed as a percentage of the average wind speed has increased between the two decades, from 1.1% to 1.4% in the Amerasian sector and from 1.0% to 1.1% in the Eurasian sector. The larger increase in this percentage in the Amerasian sector has consequences for local sea ice divergence. Despite considerable spread in this quantity, we find that the average sea ice divergence during August cyclones in the recent decade is larger in the Amerasian sector (∼5 × 10−3 s−1) than in the Eurasian sector (∼3 × 10−3 s−1). The disparity between the average sea ice divergence under cyclone and noncyclone conditions is also larger in the Amerasian sector. Therefore, the tendency for August cyclones in the Amerasian sector to result in increasingly rapid sea ice loss in recent years could be partly due to sea ice moving at a greater fraction of the surface wind speed, resulting in larger local sea ice divergence during Arctic cyclones.
Distributions of ERA5 10-m wind speed, as well as the sea ice speed and divergence computed from the NSIDC ice motion vector dataset (Tschudi et al. 2020) for cyclones (“C”) and noncyclones (“N”) in (a) the Eurasian sector (0°–140°E) and (b) the Amerasian sector (140°E–120°W). All quantities are averaged within 500 km of cyclones/noncyclones in each decade using wind and ice motion fields at 0000 UTC. Large dots are the averages over all cases, and errors bars represent the 95% confidence interval.
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
e. Oceanic contributions to local sea ice loss
Steele et al. (2010) found that late-summer ice volume loss in the Pacific Arctic is primarily due to ice bottom melt, implying that ocean processes play a critical role in sea ice loss during the melt season. Most of the upper-ocean heating that takes place in the Arctic during summer is a result of absorbed solar radiation, which forms a layer of relatively warm seawater at about 25–35-m depth that is trapped beneath the seasonal halocline. This warm layer is referred to as the near-surface temperature maximum (NSTM; Jackson et al. 2010). High winds from Arctic storms are capable of mixing this relatively warm seawater upward via turbulent mixing and Ekman pumping, which can enhance ice bottom melt (e.g., Zhang et al. 2013; Smith et al. 2018; Stern et al. 2020). Upper-ocean warming from the early to the recent decade may therefore contribute to the changing short-term impacts of Arctic cyclones on sea ice by providing a warmer reservoir of seawater to be mixed upward under high wind conditions to melt ice more rapidly. In this section, we utilize the GOFS3.1 and GLORYS ocean reanalyses described in section 2 to examine recent changes in the upper ocean and the relevance to the changing short-term impacts of cyclones on sea ice documented in this study.
Figure 12 shows the upper-ocean temperature profiles from GOFS3.1 averaged within 500 km of cyclones in each month and in each sector of the Arctic. The smaller cyclone sample size is due to GOFS3.1 data only being available from 1994 to 2015, which excludes three years of cyclone cases from each decade. In the Eurasian sector in June (Fig. 12a), the average upper-ocean temperature increases with depth below a deep near-isothermal layer extending from the surface to about 40-m depth. The warmer water at depth likely indicates Atlantic inflows into the Barents Sea and through the Fram Strait. By August, an NSTM is evident in the Eurasian sector just below a shallower near-isothermal layer extending from the surface to about 10-m depth in both decades (Fig. 12b). Notably, the average temperatures at the depths corresponding to the NSTM in the Eurasian sector have not changed significantly from the early to the recent decades. In the Amerasian sector, the average upper-ocean temperature profile in June also exhibits a deep isothermal layer (Fig. 12c). However, unlike in the Eurasian sector, temperature does not increase as quickly with depth below this layer. Furthermore, temperatures below the isothermal layer are significantly warmer in the recent decade than in the early decade in the Amerasian sector. The heating of the upper ocean by atmospheric fluxes only extends down to about 60-m depth because of attenuation of SW radiation (Steele et al. 2010). Therefore, the significant warming below ∼50-m depth in the Amerasian sector is more likely to be a result of increased lateral heat flux convergence within this sector in recent years, possibly due to warmer inflows of intermediate Atlantic waters (Polyakov et al. 2012).
Ocean in situ temperature profiles from 0- to 90-m depth from the GOFS3.1 reanalysis averaged over cyclone cases in each decade in (left) June and (right) August in (a),(b) the Eurasian sector (0°–140°E) and (c),(d) the Amerasian sector (140°E–120°W). Shaded regions denote the 95% confidence interval. Faint lines indicate ocean profiles for individual cyclone cases in each decade. Large dots are drawn at depths where the average temperatures differ between the early and recent decades with ≥95% confidence, according to a Mann–Whitney U test. The cyclone samples are smaller than those used for the atmospheric analysis because GOFS3.1 data are only available from 1994 to 2015.
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
By August in the Amerasian sector, there are clear and significant differences in the upper-ocean temperature profiles above 50-m depth between the early and recent decades (Fig. 12d). In the early decade, a weak NSTM is evident at about 25-m depth. In the recent decade, temperatures are significantly warmer than in the early decade at all depths shown in Fig. 12d. However, the most warming is evident at the NSTM, where the average temperature has increased by 0.8°C from the early to the recent decade. The Amerasian sector includes the Canada Basin, where both Steele et al. (2011) and Jackson et al. (2011) also find substantial warming at the depth of the NSTM in recent years. We also compute the change in the GOFS3.1 average ocean temperature profiles 48 h after August cyclones in the Amerasian sector (Fig. S3 in the online supplemental material) and find significantly more cooling at the top of the NSTM (12–20-m depth) in the recent decade than in the early decade. This suggests that the substantial warming of the NSTM in recent decades has also increased the extent to which cyclones are capable of mixing warm seawater upward from beneath the summer halocline to melt sea ice.
The ocean temperature profiles computed from the GLORYS ocean reanalysis (Fig. 13) exhibit notable differences from GOFS3.1. In June for example, instead of the nearly isothermal layer that extends from the surface to 40-m depth in GOFS3.1 (Figs. 12a,c), the average ocean temperature in GLORYS steadily increases from the surface to 90-m depth in both the Eurasian and Amerasian sectors (Figs. 13a,c). GLORYS also exhibits statistically significant warming in June between the early and recent decades at more depths than in GOFS3.1. The most obvious difference between GOFS3.1 and GLORYS, though, is in the Amerasian sector in August (Fig. 13d vs Fig. 12d), where the NSTM from GLORYS is much less apparent than in GOFS3.1. Most notably, the average temperature at the depth of the NSTM increases by only 0.2°C from the early to the recent period in GLORYS (as compared with 0.8°C in GOFS3.1). This large discrepancy raises the question of which of these two ocean reanalyses provides a more accurate representation of the upper-ocean temperature profile, particularly in the more recent period.
As in Fig. 12, except profiles are ocean potential temperature from the GLORYS reanalysis.
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
We estimate the accuracy of the upper-ocean profiles from each reanalysis in the recent period by comparing with observations from seven ITPs (Toole et al. 2011; Krishfield et al. 2008) that were operating within the Canada Basin during the summers of 2013–15 (Fig. 14a). First, we take daily averages of upper-ocean temperature profiles from each of the selected ITPs. We then compare these daily averaged ITP profiles with temperature profiles obtained from each reanalysis at the grid point nearest to the daily averaged position of each ITP. Figures 14b and 14c show the resulting profiles of average temperature errors in GOFS3.1 (magenta) and GLORYS (cyan) relative to ITP data in June (Fig. 14b) and August (Fig. 14c). Both reanalyses exhibit substantial cold biases in June and August between 40- and 80-m depth, which is due to a poor representation of the relatively warm Pacific Summer Water that resides at these depths within the Canada Basin (not shown). Figure 14c also shows that, around the depth of the NSTM, GOFS3.1 has a large warm bias in August of about 2°C while GLORYS has a slight cold bias of −0.2°C. These biases indicate that, within the Amerasian sector, GOFS3.1 greatly overestimates the temperature of the NSTM in recent years while GLORYS slightly underestimates the temperature of the NSTM.
(a) Positions of seven ITPs from June to August of the years listed in the legend for June (thin colored segments), July (thin gray segments), and August (thick colored segments). Also shown are profiles of the average ocean temperature error from GOFS3.1 (magenta) and GLORYS (cyan) reanalysis relative to ITP observations in (b) June and (c) August. The number of daily-averaged ITP profiles used to compute the average error profiles is printed on top of (b) and (c).
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
Sea ice cover strongly controls the amount of SW radiation that is absorbed in the upper ocean throughout the melt season. Therefore, SIC and SIT biases may contribute to the upper-ocean temperature biases shown in Figs. 14b and 14c. Figure 15 shows distributions of SIC (Figs. 15a,b) and SIT (Figs. 15c,d) from GOFS3.1 (magenta), GLORYS (cyan), and reasonably accurate benchmark datasets (orange; ERA5 for SIC and PIOMAS for SIT). In June, the average SIC in GLORYS is about 2% higher than in ERA5 (significant at >95% confidence), while the average SIC in GOFS is 1% lower than in ERA5 (not significant). In August, the average SIC in GLORYS is 5% higher than in ERA5 while the average SIC in GOFS3.1 is 3% lower than in ERA5 (all differences significant at >95% confidence). The sign of the SIT biases relative to PIOMAS is consistent with that of the SIC biases relative to ERA5. The average SIT in GOFS3.1 is 0.5 m thinner than PIOMAS in June and 0.8 m thinner in August. The underestimation of ice thickness in GOFS3.1 is even larger when we restrict the analysis to the recent period (2009–15; not shown). In GLORYS, however, the average SIT is 0.3 m thicker than PIOMAS in June and 0.7 m thicker in August. These biases strongly suggest that the ocean temperature biases near the NSTM in August are related to biases in the representation of sea ice in each reanalysis; the overly thin summer sea ice in GOFS3.1 in particular may be responsible for the large positive biases in August ocean temperature near the NSTM shown in Fig. 14c.
(a),(b) Distributions of SIC in GOFS3.1 (magenta), GLORYS (cyan), and ERA5 (orange), and (c),(d) distributions of SIT in GOFS3.1 (magenta), GLORYS (cyan), and PIOMAS (orange). Vertical lines indicate the average values from each dataset, which are printed in each panel. The colors of asterisks next to each value denote the dataset(s) from which the printed average significantly differs (p < 0.05) according to a paired Wilcoxon test.
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
We can only conclude from this analysis that the true extent of upper-ocean warming that has occurred from the early to the recent decade is somewhere between what GOFS3.1 and GLORYS depicts. Nevertheless, both reanalyses indicate statistically significant warming of the August NSTM within the Amerasian sector between the two decades examined in this study, and even the smaller amount of upper-ocean warming in GLORYS is probably sufficient to accelerate sea ice loss following late-summer cyclones in recent years. Moreover, the average August SIT within the Amerasian sector from PIOMAS has decreased by 94 cm from the early period (1994–2000) to the recent period (2009–15), which is roughly consistent with the ∼60 cm decade−1 rate of ice thickness reduction that Lindsay et al. (2009) estimated for September sea ice in the Arctic Ocean. This substantial ice thinning that has occurred in recent years not only makes August sea ice within the Amerasian sector more susceptible to wind-induced drift and divergence during Arctic cyclones in the recent decade (Fig. 11b), but also facilitates further upper-ocean warming within the Amerasian sector by allowing for greater absorption of solar radiation during the summer months.
4. Conclusions
We compared the 1–7-day changes in sea ice area and thickness following Arctic cyclones from two different decades (1991–2000 and 2009–18) to quantify how recent climate change has affected the short-term impacts of cyclones on sea ice. In both decades, we find that days without cyclones (i.e., noncyclone days) in June are followed by more sea ice area loss than cyclone days. The opposite is the case in August, when cyclone days are followed by more sea ice loss than noncyclone days. The key result from this study, however, is that recent climate change has made the absence of a cyclone in June and the presence of a cyclone in August even more destructive to sea ice on short time scales.
Upon restricting the analysis to two different regions of the Arctic, we find that the recent increase in sea ice loss following noncyclone conditions in June primarily occurs in the Eurasian sector (0°–140°E). Figure 16a schematically illustrates the changes within this part of the Arctic that have enabled noncyclone conditions in June to become more destructive to sea ice in recent years. A large portion of the retreating sea ice edge in June resides within the Barents and Kara Seas, which has caused ice concentration within the Eurasian sector to decline between the early and recent decades and, in turn, has reduced the average surface albedo. Because noncyclone conditions are characterized by about one-half as much cloud cover on average as cyclone conditions (Fig. 8a), the lower surface albedo within the Eurasian sector in the recent decade has allowed for more of the abundant insolation on noncyclone days in June to be absorbed at the surface to melt ice and warm the upper ocean (Perovich et al. 2007; Perovich 2018). We believe this explains why June noncyclone conditions in the Eurasian sector are followed by significantly more sea ice area and thickness loss in the recent decade than the cloudier cyclone conditions.
Schematic diagram illustrating the hypothesized pathways to enhanced sea ice loss from (middle) the early decade (1991–2000) to (right) the recent decade (2009–18) during (a) June noncyclone conditions and (b) August cyclone conditions, with (left) maps outlining the relevant geographic areas. Only the upper ∼60 m of the ocean is shown, where color shading denotes ocean temperature and the dashed black line approximates mixed layer depth. In June [(a)], reduced thickness and extent of Eurasian sea ice (white blocks) has decreased surface albedo in the recent decade, allowing more of the abundant solar radiation on noncyclone days (downward orange arrow) to melt ice and heat the upper ocean. In August [(b)], the thinner Amerasian sea ice in the recent decade drifts faster (white arrows) for the same wind forcing (light blue arrows), leading to greater ice divergence. Warmer seawater trapped beneath the oceanic mixed layer (red shading) may also enhance ice bottom melt via wind-induced upper-ocean mixing (curved arrows).
Citation: Journal of Climate 35, 23; 10.1175/JCLI-D-22-0315.1
In August, we find that cyclones within the Amerasian sector of the Arctic (140°E–120°W) have become more destructive to sea ice in the recent decade. Differences in the average energy flux from the atmosphere between the two decades are too small within this sector to explain the greater sea ice loss following August cyclones in the recent decade. Figure 16b illustrates the two processes that we believe have led to increased ice loss following August cyclones: increased sea ice divergence and turbulent mixing of warmer seawater from beneath the oceanic mixed layer. Sea ice within this sector drifts at a larger fraction of the surface wind speed in the recent decade than in the early decade—likely related to the almost 1-m reduction in the average ice thickness between the two decades. As a result, sea ice divergence has increased in response to strong winds from August cyclones. Moreover, August ocean temperatures just beneath the mixed layer have warmed significantly from the early to the recent decade within the Amerasian sector. Our estimates of this warming range from 0.2°C (GLORYS reanalysis) to 0.8°C (GOFS3.1 reanalysis). A comparison of each ocean reanalysis with upper-ocean profiling observations suggests that the smaller amount of warming in GLORYS is probably a more accurate estimate, although it is worth noting that point observations in the Canada Basin indicate a 1.5°C warming of the NSTM between 1993 and 2009 (Jackson et al. 2011), which is in better agreement with GOFS3.1. In any case, even the smaller amount of upper-ocean warming in the GLORYS reanalysis is likely sufficient to enhance ice bottom melt in response to cyclone-induced ocean mixing. In addition, the faster sea ice drift velocity in response to surface wind forcing during cyclones further promotes upward mixing of the warmer subsurface seawater. We hypothesize that these significant changes in the upper-ocean temperature profile and sea ice drift over the last 30 years have conspired to make August cyclones in the Amerasian sector more destructive to sea ice. Notably, the surface winds are similar between the early and recent decades for both the cyclone and noncyclone cases examined in this study, ruling out stronger wind forcing as a cause of accelerated sea ice loss in the recent decade.
Our hypothesis that warmer seawater beneath the oceanic mixed layer increases the ability of late-summer cyclones to melt sea ice in recent years assumes that these cyclones mix the warmer seawater upward toward the ice bottom. However, we do not provide direct evidence in this study of greater ice bottom melt due to ocean mixing in the recent decade. More research utilizing fully coupled simulations and/or several years of coincident ice-mass balance and upper-ocean profiling observations in high wind conditions is needed to actually demonstrate stronger and more destructive upper-ocean mixing during late-summer cyclones in recent years. Future work should also explore the possibility, implied in the results of this study, that unusually clear conditions at the beginning of the melt season followed by stormy conditions at the end of the melt season could lead to the first ice-free summer in the Arctic.
Acknowledgments.
This research was supported by the Office of Naval Research Arctic Cyclone DRI (Program Element 0601153N). We gratefully acknowledge the feedback from two anonymous reviewers, which greatly improved this paper. We also thank Michael Sprenger (ETH-Zürich) and Matt Fearon (SAIC) for creating the ERA5 cyclone database, and Jon Christophersen (NRC) for his help with the ocean reanalysis products.
Data availability statement.
All of the datasets used in this study are freely available on public repositories, except for the filtered cyclone database, which is available upon request from the authors. ERA5 data are available on the Copernicus Climate Change Service Climate Data Store (https://cds.climate.copernicus.eu). PIOMAS sea ice thickness data are available from the University of Washington Applied Physics Laboratory Polar Science Center (http://psc.apl.uw.edu/research/projects/arctic-sea-ice-volume-anomaly/data/). The polar pathfinder sea-ice motion vector data are available from the NSIDC (https://nsidc.org/data/nsidc-0116). The GOFS3.1 ocean reanalysis data are available through the HYCOM consortium (https://www.hycom.org/dataserver/gofs-3pt1/reanalysis). GLORYS ocean reanalysis is provided by E.U. Copernicus Marine Service Information (https://doi.org/10.48670/moi-00021). The Ice-Tethered Profiler data were collected and made available by the Ice-Tethered Profiler Program (Toole et al. 2011; Krishfield et al. 2008) based at the Woods Hole Oceanographic Institution (https://www.whoi.edu/itp).
REFERENCES
Brasseur, P., and J. Verron, 2006: The SEEK filter method for data assimilation in oceanography: A synthesis. Ocean Dyn., 56, 650–661, https://doi.org/10.1007/s10236-006-0080-3.
Choi, N., K.-M. Kim, Y.-K. Lim, and M.-I. Lee, 2019: Decadal changes in the leading patterns of sea level pressure in the Arctic and their impacts on the sea ice variability in boreal summer. Cryosphere, 13, 3007–3021, https://doi.org/10.5194/tc-13-3007-2019.
Clancy, R., C. M. Bitz, E. Blanchard-Wrigglesworth, M. C. McGraw, and S. M. Cavallo, 2022: A cyclone-centered perspective on the drivers of asymmetric patterns in the atmosphere and sea ice during Arctic cyclones. J. Climate, https://doi.org/10.1175/JCLI-D-21-0093.1, in press.
Crawford, A. D., and M. C. Serreze, 2016: Does the summer Arctic frontal zone influence Arctic Ocean cyclone activity? J. Climate, 29, 4977–4993, https://doi.org/10.1175/JCLI-D-15-0755.1.
Crawford, A. D., J. V. Lukovich, M. R. McCrystall, J. C. Stroeve, and D. G. Barber, 2022: Reduced sea ice enhances intensification of winter storms over the Arctic Ocean. J. Climate, 35, 3353–3370, https://doi.org/10.1175/JCLI-D-21-0747.1.
Cummings, J. A., 2005: Operational multivariate ocean data assimilation. Quart. J. Roy. Meteor. Soc., 131, 3583–3604, https://doi.org/10.1256/qj.05.105.
Cummings, J. A., and O. M. Smedstad, 2013: Variational data assimilation for the global ocean. Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications, Vol. II, S. K. Park and L. Xu, Eds., Springer, 303–343.
Curry, J. A., J. L. Schramm, and E. E. Ebert, 1993: Impact of clouds on the surface radiation balance of the Arctic Ocean. Meteor. Atmos. Phys., 51, 197–217, https://doi.org/10.1007/BF01030494.
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.
Deser, C., and H. Teng, 2008: Recent trends in Arctic sea ice and the evolving role of atmospheric circulation forcing, 1979–2007. Arctic Sea Ice Decline: Observations, Projections, Mechanisms, and Implications, Geophys. Monogr., Vol. 180, Amer. Geophys. Union, 7–26, https://doi.org/10.1029/180GM03.
Eastman, R., and S. G. Warren, 2010: Interannual variations of Arctic cloud types in relation to sea ice. J. Climate, 23, 4216–4232, https://doi.org/10.1175/2010JCLI3492.1.
Fearon, M. G., J. D. Doyle, D. R. Ryglicki, P. M. Finocchio, and M. Sprenger, 2021: The role of cyclones in moisture transport into the Arctic. Geophys. Res. Lett., 48, e2020GL090353, https://doi.org/10.1029/2020GL090353.
Finocchio, P. M., and J. D. Doyle, 2021: Summer cyclones and their association with short-term sea ice variability in the Pacific sector of the Arctic. Front. Earth Sci., 9, 738497, https://doi.org/10.3389/feart.2021.738497.
Finocchio, P. M., J. D. Doyle, D. P. Stern, and M. G. Fearon, 2020: Short-term impacts of Arctic summer cyclones on sea ice extent in the marginal ice zone. Geophys. Res. Lett., 47, e2020GL088338, https://doi.org/10.1029/2020GL088338.
Heo, E.-S., M.-K. Sung, S.-I. An, and Y.-M. Yang, 2021: Decadal phase shift of summertime Arctic dipole pattern and its nonlinear effect on sea ice extent. Int. J. Climatol., 41, 4732–4742, https://doi.org/10.1002/joc.7097.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Hirahara, S., M. A. Balmaseda, E. de Boisseson, and H. Hersbach, 2016: Sea surface temperature and sea ice concentration for ERA5. ERA Rep. 26, 27 pp., https://www.ecmwf.int/en/elibrary/16555-sea-surface-temperature-and-sea-ice-concentration-era5.
Hunke, E. C., W. H. Lipscomb, A. K. Turner, N. Jeffery, and S. Elliott, 2015: CICE: The Los Alamos sea ice model documentation and software user’s manual version 5.1. Tech. Rep. LA-CC-06-012, 116 pp., http://www.ccpo.odu.edu/∼klinck/Reprints/PDF/cicedoc2015.pdf.
Itkin, P., and Coauthors, 2017: Thin ice and storms: Sea ice deformation from buoy arrays deployed during N-ICE2015. J. Geophys. Res. Oceans, 122, 4661–4674, https://doi.org/10.1002/2016JC012403.
Jackson, J. M., E. C. Carmack, F. A. McLaughlin, S. E. Allen, and R. G. Ingram, 2010: Identification, characterization, and change of the near-surface temperature maximum in the Canada Basin, 1993–2008. J. Geophys. Res., 115, C05021, https://doi.org/10.1029/2009JC005265.
Jackson, J. M., S. E. Allen, F. A. McLaughlin, R. A. Woodgate, and E. C. Carmack, 2011: Changes to the near-surface waters in the Canada Basin, Arctic Ocean from 1993–2009: A basin in transition. J. Geophys. Res., 116, C10008, https://doi.org/10.1029/2011JC007069.
Kay, J. E., T. L’Ecuyer, A. Gettelman, G. Stephens, and C. O’Dell, 2008: The contribution of cloud and radiation anomalies to the 2007 Arctic sea ice extent minimum. Geophys. Res. Lett., 35, L08503, https://doi.org/10.1029/2008GL033451.
Koyama, T., J. Stroeve, J. Cassano, and A. Crawford, 2017: Sea ice loss and Arctic cyclone activity from 1979 to 2014. J. Climate, 30, 4735–4754, https://doi.org/10.1175/JCLI-D-16-0542.1.
Krishfield, R., J. Toole, A. Proshutinsky, and M.-L. Timmermans, 2008: Automated ice-tethered profilers for seawater observations under pack ice in all seasons. J. Atmos. Oceanic Technol., 25, 2091–2105, https://doi.org/10.1175/2008JTECHO587.1.
Kwok, R., G. Spreen, and S. Pang, 2013: Arctic sea ice circulation and drift speed: Decadal trends and ocean currents. J. Geophys. Res. Oceans, 118, 2408–2425, https://doi.org/10.1002/jgrc.20191.
Lei, R., D. Gui, J. K. Hutchings, P. Heil, and N. Li, 2020: Annual cycles of sea ice motion and deformation derived from buoy measurements in the western Arctic Ocean over two ice seasons. J. Geophys. Res. Oceans, 125, e2019JC01531, https://doi.org/10.1029/2019JC015310.
Lellouche, J.-M., and Coauthors, 2021: The Copernicus global 1/12° oceanic and sea ice GLORYS12 reanalysis. Front. Earth Sci., 9, 698876, https://doi.org/10.3389/feart.2021.698876.
Lindsay, R. W., and A. Schweiger, 2015: Arctic sea ice thickness loss determined using subsurface, aircraft, and satellite observations. Cryosphere, 9, 269–283, https://doi.org/10.5194/tc-9-269-2015.
Lindsay, R. W., J. Zhang, A. Schweiger, M. Steele, and H. Stern, 2009: Arctic sea ice retreat in 2007 follows thinning trend. J. Climate, 22, 165–176, https://doi.org/10.1175/2008JCLI2521.1.
Liu, Z., and A. Schweiger, 2017: Synoptic conditions, clouds, and sea ice melt onset in the Beaufort and Chukchi seasonal ice zone. J. Climate, 30, 6999–7016, https://doi.org/10.1175/JCLI-D-16-0887.1.
Lukovich, J. V., J. C. Stroeve, A. Crawford, L. Hamilton, M. Tsamados, H. Heorton, and F. Massonnet, 2021: Summer extreme cyclone impacts on Arctic sea ice. J. Climate, 34, 4817–4834, https://doi.org/10.1175/JCLI-D-19-0925.1.
Madec, G., and Coauthors, 2008: NEMO ocean engine (Note du Pole de modélisation 27). Institut Pierre-Simon Laplace (IPSL), 300 pp., http://eprints.soton.ac.uk/id/eprint/64324.
Maksimovich, E., and T. Vihma, 2012: The effect of surface heat fluxes on interannual variability in the spring onset of snow melt in the central Arctic Ocean. J. Geophys. Res., 117, C07012, https://doi.org/10.1029/2011JC007220.
Metzger, E. J., and Coauthors, 2017: Global Ocean Forecast System 3.1 validation testing. Naval Research Laboratory Tech. Memo. NRL/MR/7320--17-9722, 60 pp., https://apps.dtic.mil/sti/pdfs/AD1034517.pdf.
Ogi, M., and I. G. Rigor, 2013: Trends in Arctic sea ice and the role of atmospheric circulation. Atmos. Sci. Lett., 14, 97–101, https://doi.org/10.1002/asl2.423.
Overland, J. E., J. A. Francis, E. Hanna, and M. Wang, 2012: The recent shift in early summer Arctic atmospheric circulation. Geophys. Res. Lett., 39, L19804, https://doi.org/10.1029/2012GL053268.
Park, H.-S., S. Lee, S.-W. Son, S. B. Feldstein, and Y. Kosaka, 2015: 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.
Peng, L., X. Zhang, J.-H. Kim, K.-H. Cho, B.-M. Kim, Z. Wang, and H. Tang, 2021: Role of intense Arctic storm in accelerating summer sea ice melt: An in situ observational study. Geophys. Res. Lett., 48, e2021GL092714, https://doi.org/10.1029/2021GL092714.
Perovich, D. K., 2018: Sunlight, clouds, sea ice, albedo, and the radiative budget: The umbrella versus the blanket. Cryosphere, 12, 2159–2165, https://doi.org/10.5194/tc-12-2159-2018.
Perovich, D. K., B. Light, H. Eicken, K. F. Jones, K. Runciman, and S. V. Nghiem, 2007: Increasing solar heating of the Arctic Ocean and adjacent seas, 1979–2005: Attribution and role in the ice-albedo feedback. Geophys. Res. Lett., 34, L19505, https://doi.org/10.1029/2007GL031480.
Persson, P. O. G., 2012: Onset and end of the summer melt season over sea ice: Thermal structure and surface energy perspective from SHEBA. Climate Dyn., 39, 1349–1371, https://doi.org/10.1007/s00382-011-1196-9.
Polyakov, I. V., A. V. Pnyushkov, and L. A. Timokhov, 2012: Warming of the intermediate Atlantic Water of the Arctic Ocean in the 2000s. J. Climate, 25, 8362–8370, https://doi.org/10.1175/JCLI-D-12-00266.1.
Schreiber, E. A. P., and M. C. Serreze, 2020: Impacts of synoptic-scale cyclones on Arctic sea-ice concentration: A systematic analysis. Ann. Glaciol., 61, 139–153, https://doi.org/10.1017/aog.2020.23.
Schweiger, A., R. Lindsay, J. Zhang, M. Steele, H. Stern, and R. Kwok, 2011: Uncertainty in modeled Arctic sea ice volume. J. Geophys. Res., 116, C00D06, https://doi.org/10.1029/2011JC007084.
Smith, M., and Coauthors, 2018: Episodic reversal of autumn ice advance caused by release of ocean heat in the Beaufort Sea. J. Geophys. Res. Oceans, 123, 3164–3185, https://doi.org/10.1002/2018JC013764.
Spreen, G., R. Kwok, and D. Menemenlis, 2011: Trends in Arctic sea ice drift and role of wind forcing: 1992–2009. Geophys. Res. Lett., 38, L19501, https://doi.org/10.1029/2011GL048970.
Spreen, G., L. de Steur, D. Divine, S. Gerland, E. Hansen, and R. Kwok, 2020: Arctic sea ice volume export through Fram Strait from 1992 to 2014. J. Geophys. Res. Oceans, 125, e2019JC016039, https://doi.org/10.1029/2019JC016039.
Sprenger, M., and Coauthors, 2017: Global climatologies of Eulerian and Lagrangian flow features based on ERA-Interim. Bull. Amer. Meteor. Soc., 98, 1739–1748, https://doi.org/10.1175/BAMS-D-15-00299.1.
Steele, M., J. Zhang, and W. Ermold, 2010: Mechanisms of summertime upper Arctic Ocean warming and the effect on sea ice melt. J. Geophys. Res., 115, C11004, https://doi.org/10.1029/2009JC005849.
Steele, M., W. Ermold, and J. Zhang, 2011: Modeling the formation and fate of the near-surface temperature maximum in the Canadian Basin of the Arctic Ocean. J. Geophys. Res., 116, C11015, https://doi.org/10.1029/2010JC006803.
Stern, D. P., J. D. Doyle, N. P. Barton, P. M. Finocchio, W. A. Komaromi, and E. J. Metzger, 2020: The impact of an intense cyclone on short-term sea ice loss in a fully coupled atmosphere–ocean–ice model. Geophys. Res. Lett., 47, e2019GL085580, https://doi.org/10.1029/2019GL085580.
Timmermans, M.-L., 2015: The impact of stored solar heat on Arctic sea ice growth. Geophys. Res. Lett., 42, 6399–6406, https://doi.org/10.1002/2015GL064541.
Toole, J. M., R. A. Krishfield, M.-L. Timmermans, and A. Proshutinsky, 2011: The ice-tethered profiler: Argo of the Arctic. Oceanography, 24, 126–135, https://doi.org/10.5670/oceanog.2011.64.
Tschudi, M. A., W. N. Meier, and J. S. Stewart, 2020: An enhancement to sea ice motion and age products at the National Snow and Ice Data Center (NSIDC). Cryosphere, 14, 1519–1536, https://doi.org/10.5194/tc-14-1519-2020.
Vessey, A. F., K. I. Hodges, L. C. Shaffrey, and J. J. Day, 2020: An inter-comparison of Arctic synoptic scale storms between four global reanalysis datasets. Climate Dyn., 54, 2777–2795, https://doi.org/10.1007/s00382-020-05142-4.
Wang, Z., J. Walsh, S. Szymborski, and M. Peng, 2020: Rapid Arctic sea ice loss on the synoptic time scale and related atmospheric circulation anomalies. J. Climate, 33, 1597–1617, https://doi.org/10.1175/JCLI-D-19-0528.1.
Wernli, H., and L. Papritz, 2018: Role of polar anticyclones and mid-latitude cyclones for Arctic summertime sea-ice melting. Nat. Geosci., 11, 108–113, https://doi.org/10.1038/s41561-017-0041-0.
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.
Zahn, M., M. Akperov, A. Rinke, F. Feser, and I. I. Mokhov, 2018: Trends of cyclone characteristics in the Arctic and their patterns from different reanalysis data. J. Geophys. Res. Atmos., 123, 2737–2751, https://doi.org/10.1002/2017JD027439.
Zhang, J., and D. A. Rothrock, 2003: Modeling global sea ice with a thickness and enthalpy distribution model in generalized curvilinear coordinates. Mon. Wea. Rev., 131, 845–861, https://doi.org/10.1175/1520-0493(2003)131<0845:MGSIWA>2.0.CO;2.
Zhang, J., R. Lindsay, A. Schweiger, and M. Steele, 2013: The impact of an intense summer cyclone on 2012 Arctic sea ice retreat. Geophys. Res. Lett., 40, 720–726, https://doi.org/10.1002/grl.50190.
Zhang, X., A. Sorteberg, J. Zhang, R. Gerdes, and J. C. Comiso, 2008: Recent radical shifts of atmospheric circulations and rapid changes in Arctic climate system. Geophys. Res. Lett., 35, L22701, https://doi.org/10.1029/2008GL035607.
