1. Introduction
Uncertainties in the timing of a seasonally ice-free Arctic come from model errors, missing or unresolved physics, uncertainties in the forcings, and natural variability of the Arctic climate system. Mechanisms for long-term variability in Arctic sea ice extent (SIE) can be linked with the ocean, with the atmosphere, and with local feedbacks amplifying the sea ice response (e.g., Kashiwase et al. 2017). For instance, the Atlantic meridional overturning circulation (AMOC) and the Atlantic multidecadal variability are significantly correlated with long-term variability in SIE in global climate models (Day et al. 2012; Zhang et al. 2019; Zhang 2015; Yu et al. 2017), and the associated poleward ocean transport anomalies drive decadal changes (dominant time scale of 14 years) in sea ice extent in the Barents Sea (Årthun et al. 2012; Årthun and Schrum 2010). The strength of the AMOC however is still underestimated across climate models (Yan et al. 2018). Earlier work (e.g., Proshutinsky and Johnson 1997) showed the existence of two different regimes related to large-scale circulation patterns in the Arctic: one with cyclonic and one with anticyclonic circulation anomalies in the central Arctic. These two regimes, related to the location and the intensity of the Icelandic low and the Siberian/Arctic high, affect the surface currents and sea ice extent in the Arctic Ocean with a period of 10–15 years. Rigor et al. (2002) associated these modes of variability with the Arctic Oscillation (AO; Thompson and Wallace 1998) and proposed a link between the AO, coastal ice divergence along the Eurasian coastline, and local sea ice condition at the end of the melt season (see also Nikolaeva and Shesterikov 1970; Krumpen et al. 2013; Kwok et al. 2013; Williams et al. 2016; Brunette et al. 2019). Another possible link between AO variability and minimum sea ice extent is through the cloud cover and phase as well as the longwave radiative forcing, which affects the thickness of the sea ice at the onset of the melt season in regions of generally thin sea ice cover (e.g., along the Eurasian coastline; Letterly et al. 2016), long-term changes in sudden stratospheric warming (SSW), which is a precursor to spring changes in AO phase and associated coastal divergence (Smith et al. 2018), and wind and surface temperature anomalies in the eastern and western Greenland Sea (Deser et al. 2000).
On the atmospheric side, the Pacific decadal oscillation (PDO), defined as the leading EOF of monthly mean sea surface temperature (SST) anomalies over the North Pacific, affects SIE mainly in the Pacific sector and shows decadal variability at a time scale of approximately 15–25 years (Mantua and Hare 2002). The importance of SST variability in the Pacific is also consistent with modeling work of Dong et al. (2019) and England et al. (2020). Furthermore, variability in the Pacific Ocean has been shown to influence the timing of a seasonally ice-free Arctic (Screen and Deser 2019). The mechanism for this influence includes anomalous surface wind forcing causing coastal divergence which triggers the ice albedo feedback (Kashiwase et al. 2017; Hutchings and Perovich 2015) or changes in ocean heat transport through the Bering Strait (via changes in sea surface height or local winds; Woodgate et al. 2010, 2012; Serreze et al. 2019). Another low-frequency mode of atmospheric variability is linked with anomalous surface air temperature and summer Arctic ice melt (Ding et al. 2019). This coupled mode of variability between the atmosphere and the ice is present in reanalyses and observational dataset but not yet well captured in current global climate models (Ding et al. 2019).
The Community Earth System Model version 1 (CESM1) when forced with the representative concentration pathway 8.5 (RCP8.5) simulates a seasonally ice-free Arctic around 2050 (Kay et al. 2015; Jahn et al. 2016). The current and projected trajectories of the sea ice cover in the Arctic depend on both the forcing (e.g., greenhouse gas emissions) and natural climate variability (Jahn et al. 2016; Notz and Stroeve 2018). While we can hope to reduce the uncertainty in the timing of an ice-free Arctic (with better physics, higher-resolution models, and accurate forcing scenarios), there is an inherent uncertainty associated with long-term (decadal to multidecadal) climate variability. For instance, Jahn et al. (2016) found an uncertainty of two decades in the timing of an ice-free Arctic in the CESM-LE. Long-term climate variability of SIE in the Arctic—and validation of global climate models (GCMs) at long time scales—is difficult to assess from observations due to the length of reliable, consistent long-term pan-Arctic observations from satellites (1979 onward; about four decades). There exist however long-term, reliable, regional time series of SIE, each with different record length, derived from ship-log records from fishing vessels and other local economic activities in the Atlantic, Eurasian, and Pacific sectors of the Arctic. These have been compiled into the Gridded Monthly Sea Ice Extent and Concentration, 1850 Onward, version 1.1, a pan-Arctic sea ice concentration dataset covering the 1850–2013 time period (Walsh et al. 2015, updated 2016, referred to herein as “Sea Ice Back to 1850” or SIBT1850). This dataset provides a unique opportunity to assess the realism of longer-term variability simulated by climate models regionally (Walsh et al. 2016).
Fortuitously, the regions where longer time series exist are collocated with regions of known sources of decadal variability in sea ice extent. Variability in the Atlantic sector is mainly associated with the AMOC and anomalous ocean heat transport through the Barents Sea Opening (Muilwijk et al. 2018; Bitz et al. 2005; Auclair and Tremblay 2018). Variability in the Eurasian sector of the Arctic is mainly associated with the atmosphere modes of variability (NAO, PDO, AO), which govern variability in coastal divergence along the Eurasian coastline, and with surface radiative forcing (Brunette et al. 2019; Williams et al. 2016; Rigor et al. 2002; Letterly et al. 2016). Finally, variability in the Pacific sector is associated with the ocean heat transport (Woodgate et al. 2010; Serreze et al. 2019; Maslowski et al. 2004). This gives hope that we can assess the realism of the simulated longer-term variability in these regions of the Arctic. A 40-member ensemble of the Community Earth System Model (CESM-LE) was produced (Kay et al. 2015). CESM1 has a realistic representation of the Arctic climate, in particular, ice thickness, concentration, and seasonality (Jahn et al. 2016). We use SIBT1850 to assess the realism of the simulated long-term variability in the CESM-LE and quantify the uncertainty associated with natural climate variability on the timing of an ice-free Arctic.
The paper is structured as follow: the data from observations and model are described in section 2; a detailed description of the statistical techniques used to analyze the variability in different frequency bands is presented in section 3; a validation of the SIBT1850 dataset against an independent Russian dataset is presented in section 4; the main findings are discussed in sections 5 and 6; and the main conclusions drawn from the study are summarized in section 7.
2. Data
a. Sea ice extent
We use SIBT1850 to assess the regional variability in the September sea ice extent within the Arctic Ocean (Walsh et al. 2016, 2015, updated 2016). In the following, we use September and minimum sea ice extent interchangeably. A previous study has shown that sea ice concentration in the Barents and Kara Seas is more realistic in SIBT1850 than in HadISST1 (Wang et al. 2018). The spatial and temporal resolutions of SIBT1850 are 0.25° × 0.25° latitude–longitude and monthly, respectively. We define sea ice extent as the sum of all grid cells where the sea ice concentration exceeds 15%. The different sectors of the Arctic used in this study are defined in Fig. 1.
Map of the Arctic Ocean, including definition of the Atlantic, Eurasian, and Pacific sectors.
Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0561.1
The data sources in SIBT1850 mainly include maps of ice edge positions, information from newspapers, ship and aircraft observations, whaling ship logbook entries, and digitized ice charts, which are all converted to ice concentration (see Table 1 for a list of observational sources and time period, ranked in order of reliability). When only the ice edge position was recorded, a marginal sea ice zone was included where the sea ice concentration transitions from 0% to 100% over a length scale derived from the satellite record. Point observations, such as logbook entries from whaling vessels, are assumed to apply over a radius of 10–20 km, depending on the likelihood (probability) of finding sea ice in a given grid cell based on the previous and following 5 years.
When a sea ice observation is missing for a specific month and grid cell, the concentration is interpolated from the previous and following months (when available). If the temporal gap is longer, the ice concentration is calculated from the average of the three years with the highest spatial correlation with the given month. Both methods are referred to as analog filling of spatial gaps. When analog filling of spatial gaps is not possible, analog filling of temporal gaps is used instead. In this case, the ice concentration is calculated from the average of the three years with the highest spatial correlation with the nearest month where observation is available (Walsh et al. 2016). In general, the data coverage is better in the peripheral seas than in the central Arctic. Analog filling of spatial gaps is often used for months prior to 1950 and analog filling of temporal gaps is mostly used for winter months (not used in the current study). In general, the analog filling method (mainly used for the 1850–1900 time period) offers a significant improvement in the retrieval of sea ice extent variability with values similar to that of the recent records (see Figs. 3a–c). The reader is referred to Walsh et al. (2015, updated 2016) for a detailed description of the analog filling methods.
SIE digitized from the book Climate Change in Eurasian Arctic Shelf Seas (Frolov et al. 2009, hereafter F2009) are also used to validate SIBT1850 in the Chukchi, East Siberian, Laptev, Kara, Barents, and Greenland Seas. The geographical location of the peripheral seas in F2009 differs slightly from the one accepted by the International Hydrographic Organization (Chukchi and East Siberian Seas extend farther north in F2009 and the border between the Greenland Sea and the Norwegian Sea is different). F2009 uses observations from different sources to infer August SIE and provides a continuous record for 1900–2003. These include airborne campaigns and satellite observations by the Arctic and Antarctic Research Institute (AARI) starting in 1940 (F2009) and shipborne and airborne compilations from Vize (1944) for the period 1924–39. For the period 1900–23, ice edge positions were reconstructed based on observations from navigation charts and from trade, commercial and expedition ships such as the compilations from Nansen (1915) and Lesgaft (1913). Of the above, the airborne campaigns from AARI are considered the most reliable.
SIBT1850 does not include sea ice extent after 2013. We complete the time series with the NOAA/NSIDC Climate Data Record of Passive Microwave Sea Ice Concentration, version 3 (Peng et al. 2013; Meier et al. 2017a) and with the Near-Real-Time NOAA/NSIDC Climate Data Record of Passive Microwave Sea Ice Concentration, version 1 for year 2019 (Meier et al. 2017b).
The length of the SIBT1850 dataset makes it particularly suitable for low-frequency climate variability studies. The longest time series are for the Chukchi, Barents, and Greenland Seas (see Fig. 2) where whaling ships activities (in the Chukchi Sea) as well as shipping and exploration activities from the Danish Meteorological Institute (in the Greenland and Barents Seas) were abundant. In all peripheral seas, ice concentrations estimated from ice edge positions are widely used after 1901 when observations of a high order of reliability were unavailable. The quality of the data is better after 1930 and earlier estimates must be used with caution (see section 5a for a validation of the observations).
Fraction of each Arctic peripheral sea area where observed August or September sea ice concentrations exist in the Gridded Monthly Sea ice Extent and Concentration, 1850 Onward, version 1.1 dataset (Walsh et al. 2016). The analog filling method was used in the pale yellow area.
Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0561.1
b. CESM-LE
The Community Earth System Model, version 1 (CESM1) is a fully coupled general circulation model that includes atmosphere, ocean, land, and sea ice components. The atmospheric component is the Community Atmospheric Model, version 5.2 (CAM5.2), which includes improved parameterizations for aerosols, boundary layers, and radiation compared to its predecessor CAM4 (Meehl et al. 2013). CAM5.2 has a 1° horizontal resolution and 30 vertical levels and includes aerosol indirect effect. The land model is the Community Land Model, version 4 (CLM4), which notably includes an improved representation of the snow cover and a more accurate representation of global river discharge (Lawrence et al. 2011). The ocean model is the Parallel Ocean Program, version 2 (POP2; Gent et al. 2011; Kay et al. 2015). Compared to previous versions, POP2 simulates a sharper thermocline, an improved equatorial current structure, and reduced sea surface temperature and salinity biases in the North Atlantic (Danabasoglu et al. 2012). POP2, however, underestimates the multidecadal variability of the Atlantic meridional overturning circulation (AMOC) when compared to observations, possibly due to a weak low frequency variability of the North Atlantic Oscillation (NAO; Kim et al. 2018). The sea ice component is the Community Ice Code, version 4 (CICE4), a recent version of the Los Alamos sea ice model (Hunke et al. 2008). The thermodynamic component takes into account brine pocket dynamics (Bitz and Lipscomb 1999) and the dynamic component is based on the elasto-viscous-plastic (EVP) rheology (Hunke and Dukowicz 2002). POP2 and CICE4 have a 1° resolution and use a curvilinear coordinate system with the North Pole located over Greenland to avoid the singularity at the North Pole.
The CESM–Large Ensemble (CESM-LE) includes 40 ensemble members for the 1920–2100 period differing by round-off level perturbations in their initial conditions (Kay et al. 2015) and following the representative concentration pathway 8.5 (RCP8.5) for the 2006–2100 period (Meinshausen et al. 2011). By the end of the twenty-first century, all the ensemble members reach a global warming of approximately 5 K and the Arctic is seasonally ice-free in all ensemble members. The large number of ensemble members, the forcing scenario, and the length of the simulations make this model suitable for studying low-frequency internal climate variability and climate change (Kay et al. 2015; Jahn et al. 2016).
3. Method
In the following, the CWT is computed over normalized sea ice extent anomaly time series. The anomaly time series are constructed by removing the ensemble mean and dividing by the mean standard deviation of all ensemble members. For the observations, we remove the linear trend from two distinct periods separately because of the recent acceleration in the decline of sea ice extent. The break point is chosen such as to minimize the root-mean-square error (χ2) between the observed SIE and the linear approximation. The chosen year varies across the peripheral seas but is located in the 1990s in most of the cases. We determine the significance level and reliability of the CWT using the cone of influence (COI). In a CWT, the time series is padded with zeros at the beginning and end to bring its total length to the next power of 2. The zeros are removed afterward, but the discontinuities remain. The COI delimits the region of the spectrum where the impact of the discontinuity is negligible (i.e., when the wavelet power has dropped by a factor of e−2; Torrence and Compo 1998). The significance of the results are assessed using the 95% confidence levels from a red-noise spectrum, meaning that the null hypothesis is assumed to be a mean power spectrum corresponding to a lag-1 autocorrelation (Gilman et al. 1963). Results are considered significant if their time period is inside the COI and if the frequency identified in the time series is significant at the 95% level.
We also examine the variations of power over a range of scales using the scale-averaged wavelet power. The scale-averaged wavelet power corresponds to the weighted sum of the wavelet spectrum for scales within a specific frequency interval (decadal time scale in the following, corresponding to the 8–16-yr band). This provides the temporal evolution of the variability for a certain frequency range. This method is relevant in our study since it allows for the comparison of many time series precisely for decadal variability. Finally, we assess the spatial pattern of decadal variability from maps of wavelet power for a specific frequency band. To this end, we compute the CWT of the sea ice concentration anomaly, select a specific scale range of the wavelet spectrum, and compute a value for the average wavelet power in that range. These maps are not shown on a log scale to allow for the localization of hot spots of variability more easily.
4. Validation of the observations
We use F2009—a dataset with broad coverage, both temporally and spatially, particularly in the Eurasian Arctic—to validate SIBT1850 for the period 1900–2003 in two specific regions: the Greenland, Barents, and Kara Seas (GBK) and the Laptev, East Siberian, and Chukchi Seas (LEC). F2009 is mostly independent from SIBT1850, with SIBT1850 using some AARI ice charts starting in 1933. F2009 only has observations for the month of August and for the GBK and LEC regions. For this reason, we use August SIE (instead of September) for this comparison and use a slightly different definition of the peripheral seas following F2009 for consistency. The correlation between August and September SIE is high (R2 = 0.95), justifying the use of the August SIE for the validation of the September SBT1850 data. Overall, the August SIE and interannual variability in SIBT1850 and F2009 are in good agreement in both GBK and LEC (R2 = 0.72 and R2 = 0.74 respectively; see Fig. 3).
August SIE from SIBT1850 and F2009 (solid lines) and running standard deviation over 10-yr segments (dashed lines) for the (a),(b) Greenland, Barents, and Kara (GBK) Seas and (c),(d) Laptev, East Siberian, and Chukchi (LEC) seas. The burgundy lines are constructed by replacing the missing SIE values in the 1850–1900 SIBT1850 record with the climatology from the satellite record on a point-by-point basis, rather than by the 3 years with the highest spatial correlation with the given month as done in Walsh et al. (2015, updated 2016).
Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0561.1
The SIEs in the Greenland–Barents–Kara region from SIBT1850 and F2009 are in good agreement except during the time periods 1910–25 and 1950–65 when the SIEs from SIBT1850 are on average 0.35 and 0.38 × 106 km2 higher than that of F2009, respectively. These biases result in slightly larger decadal changes in the interannual variability of SIE during those decades when compared with F2009 (see Fig. 3a). The interannual variability in SIE in both datasets is in good agreement except in the 1940s and the late 1960s and early 1970s when the variability in SIBT1850 is higher (see Fig. 3b).
In the Laptev–East Siberian–Chukchi region, SIE and SIE variability between SIBT1850 and F2009 agree very well for the 1938–2019 period. In the 1900–38 time period, there is a significant bias in SIBT1850 SIE and a much reduced SIE variability when compared to the post-1938 record, the pre-1900 record, and that of F2009. This suggests an issue with the data quality during that time period. Whaling ship records were more seldom in the Chukchi Sea after 1900 and SIBT1850 relied more heavily on the analog filling method, leading to lower interannual variability. Still, the analog filling method is an improvement compared with simply filling missing values with a climatology (see Figs. 3a–c, burgundy lines). In 1938, there is an abrupt drop in SIE of approximately 0.46 × 106 km2. This coincides with the increase in the area covered by maps from the Danish Meteorological Institute and the increased use of ice charts from the Russian Arctic and Antarctic Research Institute. Spatial maps of sea ice concentration from SIBT1850 show that this decrease is associated with a sudden northward displacement of the ice edge in the Laptev Sea and partly in the East Siberian Sea (not shown). This retreat is likely associated with the inclusion of new data sources and may not be realistic.
Finally, we tested the sensitivity of the results to changes in the spatial resolution of the dataset, by averaging four grid cells together in SIBT1850 in order to match the spatial resolution of CESM-LE. The results presented in the paper are robust to this change in resolution (result not shown). In conclusion, this comparison suggests that the interannual and more importantly the decadal variability is well represented in SIBT1850 post-1938, and that caution should be exercised in the interpretation of the results for the earlier part of the record due to errors associated with changes in data sources.
5. Results
a. Interannual variability and trends
The September SIE significantly decreases in the twenty-first century in both the pan-Arctic and the three regions of interest (Atlantic, Eurasian, and Pacific sectors; see Figs. 4 and 5). In the pan-Arctic, the rate of decline of sea ice extent starts to increase around 1990 [see Fig. 4a herein and Comiso et al. (2008)]. The observed min SIE mostly lies within the envelope of simulated minimum SIE, except for the two recent record low min SIE in 2007 and 2012. The observed minimum SIE anomalies lie mainly within the three standard deviation (σ) range (99% likelihood) from the CESM-LE except for six anomalous years in the presatellite era (Fig. 4b). In the transition period to a seasonally ice-free Arctic, the interannual variability increases in the simulated minimum SIE anomalies. This is in accord with the twentieth- and twenty-first-century simulations of the Community Climate System Model, version 3 (Holland et al. 2008). Contrary to Holland et al. (2008), the large positive and negative anomalies around the ensemble mean are not distributed symmetrically in time, with a narrow peak of negative anomalies (2010–40) occurring earlier in the transition and a broader peak of positive anomalies (2010–60) extending longer during the transition (see Fig. 4b). This is due to the presence of the coastline earlier in the transition, which decreases the likelihood of a large positive anomaly and low mean sea ice extent later in the transition which decreases the likelihood of an important negative anomaly. We refer to the period between 2010 and 2040 as the transition period from a perennial to a seasonal sea ice cover characterized by enhanced interannual variability and variability across EMs. In the observational record, we see a hint of increased interannual variability starting in the 1990s, but nothing that was not observed before (e.g., see Fig. 4c, 1952–62 time period).
(a) Arctic minimum SIE from the 40 ensemble members (EMs) of the CESM-LE (thin gray lines), the ensemble mean (thick black line), and the observation from SIBT1850 (red line). (b) Minimum SIE anomaly from the 40 ensemble members of the CESM-LE (thin gray lines) plus or minus three standard deviations σ (thick black lines) and from SIBT1850 (red). (c) Running standard deviation over 10-yr segments from the 40 ensemble members of the CESM-LE (thin gray lines), the average running standard deviation (thick black line), and running standard deviation over 10-yr segments from SIBT1850 (red line). The first and last years, ±1σ, when an EM becomes seasonally ice-free are shown as solid and dotted vertical lines, respectively.
Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0561.1
(a),(d),(g) Minimum SIE from the 40 ensemble members of the CESM-LE (thin gray lines), the ensemble mean (thick black line), and the observations from SIBT1850 (thick red line). (b),(e),(h) Minimum SIE anomaly from the 40 ensemble members of the CESM-LE (thin gray lines) plus or minus three standard deviations σ (thick black lines) and from SIBT1850 (red line). (c),(f),(i) Running standard deviation over 10-yr segments from the 40 ensemble members of the CESM-LE (thin gray lines), the average running standard deviation (thick black line), and running standard deviation over 10-yr segments from SIBT1850. Columns are for the (left) Atlantic, (center) Eurasian, and (right) Pacific sectors. The first and last years, ±1σ, when an EM becomes seasonally ice-free are shown as solid and dotted vertical lines, respectively.
Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0561.1
In the Atlantic sector, the observed SIE and SIE anomalies fall mostly within the range of the ensemble members, except for two anomalous years around 1960 (see Figs. 5a–c). The interannual variability is the largest of all three sectors and the standard deviation over 10-yr segments is significantly larger than that of SIBT1850, except around the 1960s and 1990s when the variability is comparable (see Fig. 5g). In this region, the ice extent is influenced by ocean heat transport (Bitz et al. 2005; Auclair and Tremblay 2018), local winds (Onarheim et al. 2018), and the strength of the Atlantic meridional overturning circulation (AMOC; Mahajan et al. 2011). CESM shows an increase in variability in the Atlantic sector at the beginning of the transition to a seasonal ice cover, contrary to the observational record that shows no sign of this increase in the last two decades.
In the Eurasian sector, the minimum sea ice extent starts to decrease around 1990 and has the fastest rate of decline of all three sectors (see Fig. 5). The observed SIEs falls in large part within the range of the ensemble members except for several years spread throughout the record and the rate of decline is similar to the ensemble member with the fastest decline in CESM-LE. The ensemble mean variance is similar to that of the observations, and also increases at the beginning of the sea ice decline in line with observations that show a similar increase in the mid-1980s (see Figs. 5d–f). Indeed, the running standard deviation over 10-yr segments nearly doubles in the CESM-LE and observations at the beginning of the transition period to a seasonally ice-free Arctic. The observed variance has remained in the range of the model until now.
In the Pacific sector, the decrease in sea ice extent is more gradual. The observed SIEs fall mainly within the range of the ensemble members, except for a few years, particularly in the early 2000s when large negative anomalies are observed (see Figs. 5g,h). The interannual variability is the second largest after the Atlantic and before the Eurasian sectors, and is slightly smaller than 30% of the total SIE (see Figs. 5c,f). There is an increase in the variance in CESM-LE during the transition to an ice-free Arctic as observed, although the increase lags the sea ice decline by approximately two and three decades in the Pacific and Eurasian sectors of the Arctic contrary to observations where it is synchronous. Notwithstanding this lag, the running standard deviations over 10-yr segments in the observations lie within the envelope of the CESM-LE.
In summary, the variability increases during the period of transition in the first half of the twenty-first century with larger negative anomalies earlier in the transition and larger positive anomalies later in the transition in the pan-Arctic and all sectors of the Arctic. The large simulated negative anomalies in the earlier part of the transition are in accord with the observational record in the Eurasian and Pacific sector of the Arctic (see Fig. 5). The increase in SIE variability observed in the early 1980s and 2000s in the Eurasian and Pacific sectors (respectively) are mostly in line with the simulated increased variability, although the simulated increase in the Pacific sector occurs somewhat later. The simulated interannual variability in the Atlantic sector of the Arctic is higher than that in the observations except around the 1960s and 1990s when it is comparable.
b. Wavelet analysis: 1920–2019 time period
In the following, we assess the realism of the decadal variability of the pan-Arctic September sea ice extent anomalies in CESM-LE using wavelet analysis. In particular, we present wavelet power spectra of the minimum SIE anomaly for a typical ensemble member (EM-28) of the CESM-LE for the 1920–2019 period and compare with that of the SIBT1850 (see Fig. 6). This spectra shows the distribution of power at different frequencies (or period; y axis) as a function of time (x axis). The conclusions drawn from this analysis are robust with respect to the exact choice of ensemble member. The observed wavelet power spectrum shows significant high-frequency variability for most years, especially after 1930 when the data quality and number used to produce the dataset are higher. The CESM shows variability at decadal (8–10 years), and multidecadal (20–30 years) time scales (in agreement with observations), with higher power in the early twenty-first century for most ensemble members including EM-28, but the signal is generally not statistically significant and/or is located within the cone of influence.
(top) Normalized Arctic SIE anomaly (unitless) and (bottom) its wavelet power spectrum for (a) SIBT1850 and (b),(c) EM-28 of CESM-LE 1920–2019 and 1920–2060 time periods, respectively. The period (y axis in bottom panels) is a function of the wavelet transform scale. The black lines are the 95% confidence level. The hatched regions are the cones of influence (COIs) where edge effects are important. The wavelet power (color bar) is a function of the correlation coefficient between SIE and a sinusoidal signal of the corresponding period. The solid vertical line in (c) indicates the year when EM-28 becomes ice-free.
Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0561.1
In the Atlantic sector, most of the ensemble members of CESM-LE show higher multiannual (6–10 years), decadal (16–25 years), and multidecadal (30–64 years; nonsignificant or within the COI) variability and similar interannual variability when compared with SIBT1850 (see Figs. 7a,b). There exist however, some ensemble members (6 out of 40 EMs) with a similar pattern and strength of variability at decadal to multidecadal time scales when compared to SIBT1850 (e.g., Fig. 8). This implies that the uncertainty in the timing of a seasonally ice-free Arctic associated with sea ice loss in the Atlantic sector may be realistic.
(top) Normalized SIE anomaly (unitless) and (bottom) its wavelet power spectrum for the (a)–(c) Atlantic, (d)–(f) Eurasian, and (g)–(i) Pacific sectors of the Arctic: (left) SIBT1850 for 1920–2019, and (center), (right) EM-28 of CESM-LE 1920–2019 and 1920–2060, respectively. The period (y axis of all maps) is a function of the wavelet transform scale. The horizontal axis is the time. The black lines are the 95% confidence level. The hatched regions are COIs (the regions influenced by edge effects). The wavelet power (color bar) is a function of the correlation coefficient between SIE and a sinusoidal signal of the corresponding period.
Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0561.1
(top) Normalized SIE anomaly and (bottom) its wavelet power spectrum for the Atlantic sector of the Arctic: (a) SIBT1850 for 1920–2019, and (b),(c) EM-22 of CESM-LE for 1920–2019 and 1920–2060, respectively. The period in the bottom panels is a function of the wavelet transform scale. The black lines are the 95% confidence level. The hatched regions are the COIs (the regions influenced by edge effects). The wavelet power (color bar) is a function of the correlation coefficient between SIE and a sinusoidal signal of the corresponding period. The solid vertical line in (c) indicates the year when EM-22 becomes ice-free.
Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0561.1
In the Eurasian sector, different ensemble members show very different wavelet spectra in the twentieth century but generally show weak variability at interannual time scale, in agreement with SIBT1850. At longer time scales and in early part of the twenty-first century, CESM shows higher power at decadal (8–16 years; significant) and interdecadal (30–60 years: nonsignificant) time scales compared with that of SIBT1850, which shows higher (significant) power at multidecadal time scale (15–20 years; see Fig. 7d)—the time scale of the Arctic Ocean Oscillation reported by Proshutinsky et al. (2015).
In the Pacific sector, CESM-LE generally shows weak decadal (8–16 years) and multidecadal variability (32–64 years; inside of the COI) that is in line with the observations. The model shows weak but significant multiannual variability (2–4 years) for the early and mid-twentieth century with an increase in multiannual (4 years), decadal (16 years), and multidecadal (32–64 years) variability in recent years, in agreement with SIBT1850 (although partially within the cone of influence). CESM-LE captures a persistent, but not significant, signal across the time series at multidecadal time scales (30–40 years), which is also present in SIBT1850 and is much weaker in EM-28 than in most EMs.
c. Wavelet analysis: 1920–2060 time period
In the full 1920–2060 time period, the wavelet power spectrum shows significant low-frequency variability (8–12 and 16–32 years) between 2010 and 2040 (significant at the 95% confidence level). The most common pattern across ensemble members is a significant increase in multiannual, decadal, and multidecadal variability during the transition to a seasonally ice-free Arctic, followed by a decrease at the end of the time series when the sea ice cover is seasonal. This increase in interannual variability is in line with observations (see Fig. 6b); early signs of decadal to multidecadal variability are also present in the observations but located in the cone of influence. The high power at low frequency is also apparent in the average wavelet power spectrum of all EMs, with a broader multidecadal signal that starts earlier (results not shown). A spread in the signal at multidecadal time scales could also be related to a longer decorrelation time in sea ice variability for low frequencies (Torrence and Compo 1998).
In all sections of the Arctic, we see increased power at multiannual (2–4 years), decadal (8–10 years), and multidecadal (16–32 years; within the COI) in the twenty-first century with the largest increase at decadal scales in the Eurasian sector during the transition to an ice-free Arctic (2010–40, consistent across EMs), and Pacific sector approximately 20 years later (see Fig. 7i). The Pacific sector includes the region north of the Canadian Arctic Archipelago and Greenland, where thick and persistent sea ice has a longer residence time [~30 years based on ice flux through the archipelago from Kwok et al. (2013)], which may explain the late increase in variability (WWF 2014).
d. Wavelet power at decadal (8–16 years) time scale
1) Time series
The scale-averaged wavelet power of the 40 ensemble members for CESM-LE and SIBT1850 allows for a direct comparison of the model simulated variability for the 1920–2060 period with the observational record. At decadal time scales (8–16 years), the observed variability in SIBT1850 is mainly within the range of the CESM scale-averaged wavelet power across the Arctic, except for the 1930–60 period where the decadal variability is underestimated (see Fig. 9). This peak in observed decadal variability is due to a period of low ice years in the Eurasian and to a lesser extent in the Pacific sector, and warmer atmospheric temperature associated with internal atmospheric variability (see Figs. 4a, 9, and 10; Overland et al. 2004; Solomon et al. 2007). The decadal variability in SIBT1850 is not significant at the 95% level even during the 1935–45 period when it is significant at the 90% level. Similarly, most of the ensemble members do not show significant decadal variability until well into the transition to an ice-free Arctic. Decadal variability in SIBT1850 shows an increase in recent years; however this increase does not stand out when compared with that of the 1940s and is affected by edge effect (i.e., in the COI). The uncertainty in the timing of a seasonally ice-free Arctic mainly stems from the large power at decadal time scales. The observations do not show significant signs yet of increased variability at decadal time scale.
SIE scale-averaged wavelet power over the 8–16-yr band for the CESM-LE ensemble-mean wavelet power (thick black line) and 10th–90th percentiles (gray shading), and for the SIBT1850 (thick red line). The dotted and dashed black lines are the 95% confidence level for the model and observations, respectively. The hatched regions indicate the COIs.
Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0561.1
SIE scale-averaged wavelet power over the 8–16-yr band for the CESM-LE ensemble-mean wavelet power (thick black line) and 10th–90th percentiles (gray shading), and for the SIBT1850 (thick red line) for the (a) Atlantic, (b) Eurasian, and (c) Pacific sectors of the Arctic. The dotted and dashed black lines are the 95% confidence level for the model and observations, respectively. The hatched regions indicate the COIs.
Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0561.1
Regionally, we observe that decadal variability in SIBT1850 lies within the range of the CESM over most of the time series and in all sectors of the Arctic except in the Eurasian sector around 1940 when the decadal variability is underestimated. In the late twentieth century CESM-LE significantly underestimates decadal variability present in SIBT1850 in the Eurasian and Pacific sectors, although some of the discrepancies are partly in the COI. A more plausible explanation is that the simulated magnitude of decadal variability is realistic but delayed by 20–30 years in the Eurasian and Pacific sectors of the Arctic (see Fig. 10). This is in line with other work showing that GCMs tend to underestimate the rate of decline of sea ice in the Arctic (Melia et al. 2016; Stroeve et al. 2007, 2012). More precisely for the CESM, Stroeve et al. (2012) showed that the observed minimum SIE lies in the lower part of the envelope defined by the CESM. In the Atlantic sector, SIBT1850 decadal variability is consistently lower than CESM-LE but still lies within the range of the CESM for all times. In the Eurasian sector of the Arctic, decadal variability in the SIBT1850 is consistently higher than the mean of CESM-LE and outside the ensemble range in the 1940s (see Fig. 10b).
2) Spatial maps
The spatial maps of 8–16-yr wavelet power presented below focus on the 1930–2019 time period when the data coverage is sufficiently high to realistically assess decadal variability on a regional basis from CESM-LE and SIBT1850 (see Fig. 11). When we subsample the September sea ice concentration from the satellite era to obtain a data coverage similar to that of SIBT1850 prior to 1930, we greatly weaken decadal variability in the Greenland, Chukchi, and Barents Seas and no decadal variability is seen in other peripheral seas (see Fig. 12). When we subsample the September sea ice concentration from the satellite era to obtain data coverage similar to that of any post-1930 30-yr period of the SIBT1850, the spatial patterns and magnitude of decadal variability are the same as that of the satellite record, showing decadal variability in all peripheral seas of the Arctic. Note that it is the temporal (rather than spatial) resolution of the dataset that has the largest impact on the reconstructed pattern and magnitude of decadal variability; that is, a coarser spatial resolution of the same SIBT1850 SIC dataset does not change the spatial pattern or the intensity of decadal variability (result not shown). Results show that the observed 8–16-yr variability is located almost exclusively in the Marginal Ice Zone (MIZ; defined here as the zone between 15% and 80% ice concentration, Fig. 11d). In the ensemble mean of CESM-LE, the high wavelet power at 8–16 years is mostly located on the outer edge of the MIZ in agreement with SIBT1850 (although weaker). Decadal variability observed in the East Siberian Sea is outside of the 2σ range of CESM-LE (see Figs. 11b,c).
Spatial maps of SIE wavelet power over the 8–16-yr band from (a) the ensemble-mean CESM-LE, (b) plus and (c) minus two standard deviations; (d) from SIBT1850; and (e) from EM-35 for the common time period 1930–2019. The green contours are the average 80% and 15% ice cover (lines are not shown in the Canadian Arctic Archipelago).
Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0561.1
Spatial maps of SIE wavelet power over the 8–16-yr band for the (a) subsampled satellite record using the 1880–1910 data coverage from SIBT1850, (b) subsampled satellite record using the 1925–55 data coverage from SIBT1850, and (c) 1980–2010 satellite record.
Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0561.1
In the Atlantic sector, decadal variability for the period 1930–2019 is generally larger in the CESM-LE than in SIBT1850, especially in the southern Greenland Sea, close to the 15% ice concentration contours (see Figs. 11a,d). In fact, none of the 40 EMs accurately represents the magnitude of decadal variability in the Greenland Sea and the spatial pattern is shifted to the south (results not shown). Since the exact location of decadal variability varies across the EMs, the pattern is weaker in the ensemble mean (see Fig. 11a). While a dozen ensemble members show a similar pattern in wavelet power in the 8–16-yr band over the East Siberian and Laptev Seas when compared with observations (EMs 1, 2, 9, 14, 16, 29, 35, 37, and 39), the intensity is slightly lower than observation in every ensemble member (see EM-35 in Fig. 11e as an example). In the Pacific sector, decadal variability is generally similar in CESM-LE and SIBT1850, and some ensemble members (eight) show similar patterns to those in SIBT1850. Decadal variability in the observations is important north of Bering Strait, suggesting an oceanic source (Mantua and Hare 2002). This increased variability north of Bering Strait is also seen in some ensemble members (15 out of 40). Ocean heat fluxes through the Bering Strait influence sea ice on the broad shallow shelf and have been increasing in the last decades (Woodgate et al. 2010, 2012). The observed sea ice decadal variability might be linked to a possible under representation of the PDO and the Bering Strait inflow, which are important factors for an accurate representation of the changing SIE in the Arctic (Screen and Deser 2019).
A map of 8–16-yr wavelet power in SIBT1850 for the satellite record (1980–2019) shows an increase in variability in the East Siberian and Laptev Seas compared to the 1930–2019 period (cf. Figs. 11 and 13 and Fig. 9). The sea ice extents from recent decades also show a clear rise in decadal variability in the Atlantic sector of the Arctic (Greenland and Barents Seas) compared to the 1930–2019 period. For the 1980–2019 period in the CESM-LE, only four ensemble members (compared to a dozen for the 1930–2019 period) show a similar pattern of decadal variability in the East Siberian and Laptev Seas in the same time period. The intensity, however, remains lower in the CESM-LE and, of the four EMs in agreement with SIBT1850, none also accurately represents the variability in the Greenland and Barents Seas. CESM hot spots of decadal variability in the Eurasian sector of the Arctic are sometimes displaced to the west, or variability is too strong in the Pacific sector of the Arctic. While the satellite era only includes two to four cycles of 8–16 years, the observed increase of decadal variability in the wavelet power map is consistent with the one observed in the wavelet power spectrum outside of the COI. Finally, differences in magnitude and spatial patterns across CESM ensemble members are greater in the 1980–2019 period than in the 1930–2019 period (Figs. 13b,c).
Spatial maps of SIE wavelet power over the 8–16-yr band for the (a) ensemble-mean CESM-LE, (b) plus and (c) minus two standard deviations; (d) from SIBT1850; and (e) from EM-35 for the common time period 1980–2019. The green contours are the average 80% and 15% ice cover (lines are not shown in the Canadian Arctic Archipelago).
Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0561.1
6. Discussion
a. Atlantic sector
Results from the wavelet analysis show that decadal variability in sea ice extent in CESM-LE is well represented in the Barents Sea and overestimated in the Greenland Sea, two regions influenced by the partitioning of the North Atlantic Drift Current (Orvik and Skagseth 2003). The overestimated decadal variability in the Greenland Sea occurs despite a realistic AMOC variability at a decadal time scale (Kim et al. 2018). Other modeling and observational studies have shown large variability at a 5–10-yr period in the ocean heat transport and more precisely in the Atlantic sector of the Arctic (e.g., Muilwijk et al. 2018; Mercier et al. 2015). A comparison of the CESM-LE OHT in the Fram Strait, Barents Sea Opening (BSO), and Bering Strait with observations shows good agreement in the BSO and a significant underestimation in the Fram and Bering Straits (Auclair and Tremblay 2018). OHT covaries with sea ice area change in the Barents Sea (and Kara Sea) and has also been linked with significant decrease in sea ice extent in the CESM-LE and other GCMs (Årthun et al. 2012; Auclair and Tremblay 2018; Holland et al. 2006; Muilwijk et al. 2018). Analysis of sea surface temperature also shows a 8–12-yr variability in the northern North Atlantic (Delworth et al. 2013), which could be linked to sea ice extent variability. The ocean heat transport in the Atlantic sector has been linked with the Arctic or North Atlantic Oscillation (Årthun and Schrum 2010; Muilwijk et al. 2018) which also show variability at decadal time scale and CESM-LE has been shown to produce a realistic NAO (Deser et al. 2017). Finally, low-frequency NAO variability, which is underestimated in the CESM-LE (Kim et al. 2018), can influence the storm activity in the Atlantic sector of the Arctic (Hurrell et al. 2003) and therefore influence sea ice conditions (Simpkins 2018).
b. Eurasian sector
Results from the wavelet analysis show that decadal (8–16 years) variability in CESM-LE is underestimated in the Eurasian sector of the Arctic, particularly in the East Siberian Sea. A key mode of variability at that time scale, termed the Arctic Ocean Oscillation, is linked with changes in wind forcing and the strength of the Arctic high, in a manner similar to the seasonal changes in the Arctic sea level pressure field (Proshutinsky and Johnson 1997). In this cycle, the cyclonic regime corresponds to offshore winds, surface ocean currents and ice drift perpendicular to the Eurasian coastline, leaving behind a thinner sea ice cover. Rigor et al. (2002) suggested that variability in the strength of the Arctic high (or Arctic Oscillation) was responsible for variability in sea ice extent in this region (see also Dewdney 1979). Williams et al. (2016), Itkin and Krumpen (2017), and Brunette et al. (2019) demonstrated this link between the AO, late winter coastal divergence, and summer sea ice extent and thickness using a Lagrangian model forced with buoy- and satellite-derived sea ice drifts. The underestimation of decadal variability in the CESM-LE sea ice extent in the Eurasian Arctic could therefore be related to a bias in the position of the Arctic high as suggested by DeRepentigny et al. (2016), which leads to sea ice drift parallel to the coast as opposed to away from the coast, or could be linked to a weaker than observed decadal variability in the Arctic Oscillation. We note that the largest trend in sea ice loss over the last decades is observed in September in the East Siberian Sea (Onarheim et al. 2018) and only 10% of September sea ice loss in this region is explained by internal variability (England et al. 2019).
c. Pacific sector
Results from the wavelet analysis show that decadal (8–16 years) variability in the Pacific sector of the Arctic from model and observations increases at the onset of the sea ice decline with a delayed response by a few decades in CESM when compared to SIBT1850. Oceanic heat enters the Arctic through the Bering Strait over the broad Eurasian shelf and interact with sea ice locally (Woodgate et al. 2010). Variability in sea ice extent in the Chukchi Sea is driven by both surface winds and thermodynamics processes (Frey et al. 2015) and as the sea ice cover retreats, storminess activity over the Arctic Ocean increased as well, enhancing heat exchange (Maslowski et al. 2000). Woodgate et al. (2012) showed that the observed Bering Strait inflow increased by 50% in recent decades, leading to a significant increase in heat transport. The mean and the magnitude of these changes, however, are largely underestimated in the CESM-LE (Auclair and Tremblay 2018). The strength of the Aleutian low and the Arctic high sets the sea surface height difference across the Bering Strait and therefore the volume flux across the strait. The wind in the Bering Strait is typically southward and reduces the magnitude of the volume flux trough the Bering strait. Therefore, variability in both the AO and the PDO can have a large influence on the Bering Strait OHT and on the sea conditions in the Chukchi and East Siberian Seas. The Pacific decadal oscillation varies on a 15–25-yr time scale (Mantua and Hare 2002) and changes in the PDO and AO can cause flushing of some of the oldest ice out of the Arctic Basin (Lindsay and Zhang 2005). The spatial pattern in PDO simulated by the CESM is in good agreement with observations, something that many CMIP5 models fail to represent (Wei et al. 2017). It has been shown that knowing the phase of the PDO can improve the skill of decadal-scale predictability of an ice-free Arctic by up to 7 years (Screen and Francis 2016).
d. Future evolution
Interannual and decadal variability is increasing in every region of the Arctic and is projected to continue to increase in the future as the pack ice transitions to a seasonally ice-free Arctic (Jahn et al. 2016; Mioduszewski et al. 2019). In the CESM-LE, we observe a larger standard deviation (4.3 vs 3.3 years) in the timing of an ice-free (less than 1 million km2) Arctic in ensemble members that have more power at decadal time scale (larger than ± one standard deviation from the mean) for the 1980–2019 period. This difference, however, is not significant at the 95% level; a much larger number of EMs would be required in order to assess the link between uncertainty and power at decadal time scales. CESM-LE projects this variability to peak in the first half of the century (around 2030) and decrease afterward due to the absence of a significant summer sea ice cover, in accord with previous studies (e.g., Goosse et al. 2009). Across all ensemble members, the simulated decadal variability increases earlier in the Eurasian sector, followed by the Atlantic sector and later the Pacific sector. Observations from the Eurasian and the Pacific sectors of the Arctic show early signs of increased decadal variability, approximately 10–20 years before the CESM—although they are located partly in the cone of influence. If this tendency is maintained, this suggests a delayed response in sea ice changes in CESM. Ocean heat flux is a key factor in determining the ice edge positions (Bitz et al. 2005). However, as the perennial ice continues to retreat past the continental shelves of the Arctic, rapid changes in September sea ice extent are expected to be less correlated with ocean transport and more with variability in the atmosphere–ice heat turbulent and radiative fluxes (Auclair and Tremblay 2018). The spatial location of hot spots of decadal variability varies across ensemble members as we reach the transition period to a seasonally ice-free Arctic, but the most common scenario predicted by CESM is an increase in decadal variability in the regions of long-lasting ice, namely the central Arctic and north of Greenland and Ellesmere Island (results not shown). Overall, the observed increase in variability at decadal time scale is well represented in CESM-LE (observations are systematically above the mean but within the envelope of different ensemble members), except in the Atlantic sector where CESM overestimates decadal variability and except for the timing of the increase in the Eurasian sector of the Arctic. Given that the sea ice retreat in the CESM-LE is Pacific-centric (DeRepentigny et al. 2016) and that decadal variability has been constantly increasing in this area in the last decades, the analysis suggests that a seasonally ice-free Arctic could happen earlier than projected by CESM (by 10–20 years) but supports the estimate of the uncertainty of two decades in the exact timing of an ice-free Arctic reported in Jahn et al. (2016).
7. Conclusions
We present a regional analysis of the low-frequency variability in Arctic minimum sea ice extent simulated by the Community Earth System Model–Large Ensemble, taking advantage of the fact that different peripheral seas of the Arctic have different observational record length. The observational record is the recent Gridded Monthly Sea Ice Extent and Concentration, 1850 Onward, version 1.1 Walsh et al. (2015, updated 2016). We show that sea ice extent decadal variability is underestimated in the CESM-LE in the Eurasian sector of the Arctic, more precisely in the East Siberian Sea, and overestimated in the Greenland Sea. Decadal variability from the CESM-LE is in line with observations in the Pacific sector of the Arctic (although the timing of the increase in variability lags by 10–20 years when compared with observations) and in the Barents Sea. Our current understanding of sea ice variability in the Eurasian sector of the Arctic leads us to believe that mechanisms of coastal divergence in the East Siberian Sea are not well represented in the CESM-LE, likely due to biases in the position of the climatologic Arctic high pressure system. Decadal variability in the Arctic Oscillation appear to be underestimated in the CESM-LE. Presumably, a more realistic variability at decadal time scale of the Arctic Oscillation would increase sea ice extent variability in the Eurasian sector of the Arctic leading to more uncertainties in future projections of the sea ice cover.
In the Greenland Sea, the decadal variability in sea ice extent is overestimated, likely due to issues related to the partitioning of ocean heat transport between the Barents Sea Opening, the Fram Strait, and recirculation of waters within the Greenland Sea (Årthun and Schrum 2010; Mahlstein and Knutti 2011; Årthun et al. 2012; Auclair and Tremblay 2018; Muilwijk et al. 2018). A more realistic representation of variability at decadal time scale would reduce the uncertainties in the projection of an Atlantic-centric retreat of sea ice.
Minimum simulated SIE decadal variability is in line with observations in the Pacific sector of the Arctic, where the Bering ocean heat transport (OHT) has a larger impact on SIE decline. A Pacific-centric sea ice decline is currently underway in the Arctic (see also DeRepentigny et al. 2016), in line with the hypothesis that Bering Strait OHT anomalies have a larger impact on minimum SIE when it enters on a shallow, broad shelf (Auclair and Tremblay 2018). Given that the magnitude of change in decadal variability in the Pacific sector of the Arctic is realistic but delayed by 10–20 years, the uncertainty in the timing of a projected SIE decline may be realistic although it could occur earlier in the real system. Uncertainties in the timing of an ice-free Arctic associated with better physics, model resolution, and forcing scenarios cannot be assessed using a large ensemble of only one model and is beyond the scope of the present paper. The 21-yr uncertainty in the timing of an ice-free Arctic reported by Jahn et al. (2016) appears to be realistic but could even be higher given that the low-frequency variability in the Eurasian sector of the Arctic is underestimated.
Finally, results suggest the presence of a transition period between an ice-covered and a seasonally ice-free Arctic characterized by increased variability at all time scales. We suggest that we are already in the transition period. Mechanisms of decadal variability that were understood for the last decades might not influence Arctic sea ice extent the same way in a transient climate. Other climate diagnostics (e.g., ice thickness) need to be studied to understand climate in the Arctic and its complex interaction with the ocean and atmosphere. Regional studies focusing on mechanisms specific for the regions might be key for understanding the changing response of the sea ice to atmospheric and oceanic processes.
Acknowledgments
We acknowledge the CESM Large Ensemble Community Project and supercomputing resources provided by NSF/CISL/Yellowstone. The Gridded Monthly Sea-Ice Extent and Concentration, 1850 Onward, version 1, the NOAA/NSIDC Climate Data Record of Passive Microwave sea-ice concentration, version 3, and the Near-Real-Time NOAA/NSIDC Climate Data Record of Passive Microwave Sea-Ice Concentration, version 1 were obtained from the National Snow and Ice Data Center NSIDC. Amélie Desmarais acknowledges funding from the Fonds de recherche du Québec–Nature et Technologies (FRQNT). This work is a contribution to the National Science and Engineering Research Council–Discovery Program, the Canadian Sea Ice and Snow Evolution Network (CanSISE), funded by NSERC Climate Change and Atmospheric Research Program, the NSF Office of Polar Program Grants 1504023, 1603350, and 1928126, NASA Grant 80NSSC20K1259, Québec-Océan, and ArcTrain Canada.
REFERENCES
Årthun, M., and C. Schrum, 2010: Ocean surface heat flux variability in the Barents Sea. J. Mar. Syst., 83, 88–98, https://doi.org/10.1016/j.jmarsys.2010.07.003.
Årthun, M., T. Eldevik, L. H. Smedsrud, O. Skagseth, and R. B. Ingvaldsen, 2012: Quantifying the influence of Atlantic heat on Barents Sea ice variability and retreat. J. Climate, 25, 4736–4743, https://doi.org/10.1175/JCLI-D-11-00466.1.
Auclair, G., and L. B. Tremblay, 2018: The role of ocean heat transport in rapid sea ice declines in the Community Earth System Model large ensemble. J. Geophys. Res. Oceans, 123, 8941–8957, https://doi.org/10.1029/2018JC014525.
Bitz, C. M., and W. H. Lipscomb, 1999: An energy-conserving thermodynamic model of sea ice. J. Geophys. Res., 104, 15 669–15 677, https://doi.org/10.1029/1999JC900100.
Bitz, C. M., M. M. Holland, E. C. Hunke, and R. E. Moritz, 2005: Maintenance of the sea-ice edge. J. Climate, 18, 2903–2921, https://doi.org/10.1175/JCLI3428.1.
Brunette, C., B. Tremblay, and R. Newton, 2019: Winter coastal divergence as a predictor for the minimum sea ice extent in the Laptev Sea. J. Climate, 32, 1063–1080, https://doi.org/10.1175/JCLI-D-18-0169.1.
Comiso, J. C., C. L. Parkinson, R. Gersten, and L. Stock, 2008: Accelerated decline in the Arctic sea ice cover. Geophys. Res. Lett., 35, L01703, https://doi.org/10.1029/2007GL031972.
Danabasoglu, G., S. C. Bates, B. P. Briegleb, S. R. Jayne, M. Jochum, W. G. Large, S. Peacock, and S. G. Yeager, 2012: The CCSM4 ocean component. J. Climate, 25, 1361–1389, https://doi.org/10.1175/JCLI-D-11-00091.1.
Day, J. J., J. C. Hargreaves, J. D. Annan, and A. Abe-Ouchi, 2012: Sources of multi-decadal variability in Arctic sea ice extent. Environ. Res. Lett., 7, 034011, https://doi.org/10.1088/1748-9326/7/3/034011.
Delworth, T. L., R. Zhang, and M. E. Mann, 2013: Decadal to centennial variability of the Atlantic from observations and models. Past and Future Changes of the Ocean’s Meridional Overturning Circulation: Mechanisms and Impacts, Geophys. Monogr., Vol. 173, Amer. Geophys. Union, 131–148, https://doi.org/10.1029/173GM10.
DeRepentigny, P., L. B. Tremblay, R. Newton, and S. Pfirman, 2016: Patterns of sea ice retreat in the transition to a seasonally ice-free Arctic. J. Climate, 29, 6993–7008, https://doi.org/10.1175/JCLI-D-15-0733.1.
Deser, C., J. E. Walsh, and M. S. Timlin, 2000: Arctic sea ice variability in the context of recent atmospheric circulation trends. J. Climate, 13, 617–633, https://doi.org/10.1175/1520-0442(2000)013<0617:ASIVIT>2.0.CO;2.
Deser, C., J. W. Hurrell, and A. S. Phillips, 2017: The role of the North Atlantic Oscillation in European climate projections. Climate Dyn., 49, 3141–3157, https://doi.org/10.1007/s00382-016-3502-z.
Dewdney, J. C., 1979: Climate, soils, vegetation. A Geography of the Soviet Union. 3rd ed. Pergamon Press, 17–37, https://doi.org/10.1016/B978-0-08-023739-8.50014-6.
Ding, Q., and Coauthors, 2019: Fingerprints of internal drivers of Arctic sea ice loss in observations and model simulations. Nat. Geosci., 12, 28–33, https://doi.org/10.1038/s41561-018-0256-8.
Dong, Y., C. Proistosescu, K. C. Armour, and D. S. Battisti, 2019: Attributing historical and future evolution of radiative feedbacks to regional warming patterns using a Green’s function approach: The preeminence of the western Pacific. J. Climate, 32, 5471–5491, https://doi.org/10.1175/JCLI-D-18-0843.1.
England, M., A. Jahn, and L. Polvani, 2019: Nonuniform contribution of internal variability to recent Arctic sea ice loss. J. Climate, 32, 4039–4053, https://doi.org/10.1175/JCLI-D-18-0864.1.
England, M., L. M. Polvani, and L. Sun, 2020: Robust Arctic warming caused by projected Antarctic sea ice loss. Environ. Res. Lett., 15, 104005, https://doi.org/10.1088/1748-9326/abaada.
Frey, K. E., G. Moore, L. W. Cooper, and J. M. Grebmeier, 2015: Divergent patterns of recent sea ice cover across the Bering, Chukchi, and Beaufort Seas of the Pacific Arctic region. Prog. Oceanogr., 136, 32–49, https://doi.org/10.1016/j.pocean.2015.05.009.
Frolov, I., Z. Gudkovichand, V. Karlin, E. G. Kovalev, and V. Smolyanitsky, 2009: Climate Change in Eurasian Arctic Shelf Seas: Centennial Ice Cover Observations. Springer-Verlag, 166 pp.
Gent, P. R., and Coauthors, 2011: The Community Climate System Model version 4. J. Climate, 24, 4973–4991, https://doi.org/10.1175/2011JCLI4083.1.
Gilman, D. L., F. J. Fuglister, and J. M. Mitchell, 1963: On the power spectrum of red noise. J. Atmos. Sci., 20, 182–184, https://doi.org/10.1175/1520-0469(1963)020<0182:OTPSON>2.0.CO;2.
Goosse, H., O. Arzel, C. M. Bitz, A. de Montety, and M. Vancoppenolle, 2009: Increased variability of the Arctic summer ice extent in a warmer climate. Geophys. Res. Lett., 36, L23702, https://doi.org/10.1029/2009GL040546.
Holland, M., C. Bitz, and B. Tremblay, 2006: Future abrupt reductions in the summer Arctic sea ice. Geophys. Res. Lett., 33, L23503, https://doi.org/10.1029/2006GL028024.
Holland, M., C. Bitz, B. Tremblay, and D. Bailey, 2008: The role of natural versus forced change in future rapid summer Arctic ice loss. Arctic Sea Ice Decline: Observations, Projections, Mechanisms, and Implications, Geophys. Monogr., Vol. 180, Amer. Geophys. Union, 133–150, https://doi.org/10.1029/180GM10.
Hunke, E. C., and J. K. Dukowicz, 2002: The elastic–viscous–plastic sea ice dynamics model in general orthogonal curvilinear coordinates on a sphere—Incorporation of metric terms. Mon. Wea. Rev., 130, 1848–1865, https://doi.org/10.1175/1520-0493(2002)130<1848:TEVPSI>2.0.CO;2.
Hunke, E. C., and W. H. Lipscomb, A. K. Turner, N. Jefferey, and S. Elliot, 2008: CICE: The Los Alamos Sea ice model documentation and software user’s manual version 4.0 LA-CC-06-012. Tech. Rep. LA-CC-06-012, 500 pp.
Hurrell, J. W., Y. Kushnir, G. Ottersen, and M. Visbeck, 2003: An overview of the North Atlantic Oscillation. The North Atlantic Oscillation—Climate Significance and Environmental Impact, Geophys. Monogr., Vol. 134, Amer. Geophys. Union, 1–35, https://doi.org/10.1029/134GM01.
Hutchings, J. K., and D. K. Perovich, 2015: Preconditioning of the 2007 sea-ice melt in the eastern Beaufort Sea, Arctic Ocean. Ann. Glaciol., 56, 9498, https://doi.org/10.3189/2015AoG69A006.
Itkin, P., and T. Krumpen, 2017: Winter sea ice export from the Laptev Sea preconditions the local summer sea ice cover and fast ice decay. Cryosphere, 11, 2383–2391, https://doi.org/10.5194/tc-11-2383-2017.
Jahn, A., J. E. Kay, M. M. Holland, and D. M. Hall, 2016: How predictable is the timing of a summer ice-free Arctic? Geophys. Res. Lett., 43, 9113–9120, https://doi.org/10.1002/2016GL070067.
Kashiwase, H., K. I. Ohshima, S. Nihashi, and H. Eicken, 2017: Evidence for ice-ocean albedo feedback in the Arctic Ocean shifting to a seasonal ice zone. Sci. Rep., 7, 8170, https://doi.org/10.1038/s41598-017-08467-z.
Kay, J. E., and Coauthors, 2015: The Community Earth System Model (CESM) large ensemble project: A community resource for studying climate change in the presence of internal climate variability. Bull. Amer. Meteor. Soc., 96, 1333–1349, https://doi.org/10.1175/BAMS-D-13-00255.1.
Kelly, P. M., 1979: An Arctic sea ice data set, 1901–1956. Glaciological Data, Report GD-5: Workshop on Snow Cover and Sea Ice Data, National Snow and Ice Data Center, 101–110.
Kim, W. M., S. Yeager, P. Chang, and G. Danabasoglu, 2018: Low-frequency North Atlantic climate variability in the Community Earth System Model large ensemble. J. Climate, 31, 787–813, https://doi.org/10.1175/JCLI-D-17-0193.1.
Krumpen, T., M. Janout, K. I. Hodges, R. Gerdes, F. Girard-Ardhuin, J. A. Hölemann, and S. Willmes, 2013: Variability and trends in Laptev Sea ice outflow between 1992–2011. Cryosphere, 7, 349–363, https://doi.org/10.5194/tc-7-349-2013.
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.
Lawrence, D. M., and Coauthors, 2011: Parameterization improvements and functional and structural advances in version 4 of the Community Land Model. J. Adv. Model. Earth Syst., 3, M03001, https://doi.org/10.1029/2011MS00045.
Lesgaft, E., 1913: Arctic Ocean Ice and a Seaway from Europe to Siberia (in Russian). St. Petersburg, 237 pp.
Letterly, A., J. Key, and Y. Liu, 2016: The influence of winter cloud on summer sea ice in the Arctic, 1983–2013. J. Geophys. Res. Atmos., 121, 2178–2187, https://doi.org/10.1002/2015JD024316.
Lindsay, R. W., and J. Zhang, 2005: The thinning of Arctic sea ice, 1988–2003: Have we passed a tipping point? J. Climate, 18, 4879–4894, https://doi.org/10.1175/JCLI3587.1.
Mahajan, S., R. Zhang, and T. L. Delworth, 2011: Impact of the Atlantic meridional overturning circulation (AMOC) on Arctic surface air temperature and sea ice variability. J. Climate, 24, 6573–6581, https://doi.org/10.1175/2011JCLI4002.1.
Mahlstein, I., and R. Knutti, 2011: Ocean heat transport as a cause for model uncertainty in projected Arctic warming. J. Climate, 24, 1451–1460, https://doi.org/10.1175/2010JCLI3713.1.
Mantua, N. J., and S. R. Hare, 2002: The Pacific decadal oscillation. J. Oceanogr., 58, 35–44, https://doi.org/10.1023/A:1015820616384.
Maslowski, W., R. Newton, P. Schlosser, A. J. Semtner, and D. Martinson, 2000: Modeling recent climate variability in the Arctic Ocean. Geophys. Res. Lett., 27, 3743–3746, https://doi.org/10.1029/1999GL011227.
Maslowski, W., D. Marble, W. Walczowski, U. Schauer, J. L. Clement, and A. J. Semtner, 2004: On climatological mass, heat, and salt transports through the Barents Sea and Fram Strait from a pan-Arctic coupled ice-ocean model simulation. J. Geophys. Res., 109, C03032, https://doi.org/10.1029/2001JC001039.
Meehl, G. A., and Coauthors, 2013: Climate change projections in CESM1(CAM5) compared to CCSM4. J. Climate, 26, 6287–6308, https://doi.org/10.1175/JCLI-D-12-00572.1.
Meier, W. N., F. Fetterer, M. Savoie, S. Mallory, R. Duerr, and J. Stroeve, 2017a: NOAA/NSIDC climate data record of passive microwave sea ice concentration, version 3. National Snow and Ice Data Center, accessed 2019, https://doi.org/10.7265/N59P2ZTG.
Meier, W. N., F. Fetterer, and A. K. Windnagel, 2017b: Near-real-time NOAA/NSIDC climate data record of passive microwave sea ice concentration, version 1. National Snow and Ice Data Center, accessed 2019, https://doi.org/10.7265/N5FF3QJ6.
Meinshausen, M., and Coauthors, 2011: The RCP greenhouse gas concentrations and their extensions from 1765 to 2300. Climatic Change, 109, 213–241, https://doi.org/10.1007/s10584-011-0156-z.
Melia, N., K. Haines, and E. Hawkins, 2016: Sea ice decline and 21st century trans-Arctic shipping routes. Geophys. Res. Lett., 43, 9720–9728, https://doi.org/10.1002/2016GL069315.
Mercier, H., and Coauthors, 2015: Variability of the meridional overturning circulation at the Greenland–Portugal OVIDE section from 1993 to 2010. Prog. Oceanogr., 132, 250–261, https://doi.org/10.1016/j.pocean.2013.11.001.
Mioduszewski, J. R., S. Vavrus, M. Wang, M. Holland, and L. Landrum, 2019: Past and future interannual variability in Arctic sea ice in coupled climate models. Cryosphere, 13, 113–124, https://doi.org/10.5194/tc-13-113-2019.
Muilwijk, M., L. H. Smedsrud, M. Ilicak, and H. Drange, 2018: Atlantic water heat transport variability in the 20th century Arctic Ocean from a global ocean model and observations. J. Geophys. Res. Oceans, 123, 8159–8179, https://doi.org/10.1029/2018JC014327.
Nansen, F., 1915: To the Country of the Future (in Russian). Petrograd, 455 pp.
Nikolaeva, A., and N. Shesterikov, 1970: A technique for calculation of ice conditions (Laptev Sea, as example). AARI Proc., 292, 143–217.
Notz, D., and J. Stroeve, 2018: The trajectory towards a seasonally ice-free Arctic Ocean. Curr. Climate Change Rep., 4, 407–416, https://doi.org/10.1007/s40641-018-0113-2.
Onarheim, I. H., T. Eldevik, L. H. Smedsrud, and J. C. Stroeve, 2018: Seasonal and regional manifestation of Arctic sea ice loss. J. Climate, 31, 4917–4932, https://doi.org/10.1175/JCLI-D-17-0427.1.
Orvik, K. A., and Ø. Skagseth 2003: The impact of the wind stress curl in the North Atlantic on the Atlantic inflow to the Norwegian Sea toward the Arctic. Geophys. Res. Lett., 30, 1884, https://doi.org/10.1029/2003GL017932.
Overland, J. E., M. C. Spillane, D. B. Percival, M. Wang, and H. O. Mofjeld, 2004: Seasonal and regional variation of pan-Arctic surface air temperature over the instrumental record. J. Climate, 17, 3263–3282, https://doi.org/10.1175/1520-0442(2004)017<3263:SARVOP>2.0.CO;2.
Peng, G., W. N. Meier, D. J. Scott, and M. H. Savoie, 2013: A long-term and reproducible passive microwave sea ice concentration data record for climate studies and monitoring. Earth Syst. Sci. Data, 5, 311–318, https://doi.org/10.5194/essd-5-311-2013.
Proshutinsky, A. Y., and M. A. Johnson, 1997: Two circulation regimes of the wind-driven Arctic Ocean. J. Geophys. Res., 102, 12 493–12 514, https://doi.org/10.1029/97JC00738.
Proshutinsky, A. Y., D. Dukhovskoy, M.-L. Timmermans, R. Krishfield, and J. L. Bamber, 2015: Arctic circulation regimes. Philos. Trans. Roy. Soc., 373A, 20140160, https://doi.org/10.1098/rsta.2014.0160.
Rigor, I. G., J. M. Wallace, and R. L. Colony, 2002: Response of sea ice to the Arctic oscillation. J. Climate, 15, 2648–2663, https://doi.org/10.1175/1520-0442(2002)015<2648:ROSITT>2.0.CO;2.
Screen, J. A., and J. Francis, 2016: Contribution of sea-ice loss to Arctic amplification is regulated by Pacific Ocean decadal variability. Nat. Climate Change, 6, 856–860, https://doi.org/10.1038/nclimate3011.
Screen, J. A., and C. Deser, 2019: Pacific Ocean variability influences the time of emergence of a seasonally ice-free Arctic Ocean. Geophys. Res. Lett., 46, 2222–2231, https://doi.org/10.1029/2018GL081393.
Serreze, M. C., A. P. Barrett, A. D. Crawford, and R. A. Woodgate, 2019: Monthly variability in Bering Strait oceanic volume and heat transports, links to atmospheric circulation and ocean temperature, and implications for sea ice conditions. J. Geophys. Res. Oceans, 124, 9317–9337, https://doi.org/10.1029/2019JC015422.
Simpkins, G., 2018: Arctic storms. Nat. Climate Change, 8, 13, https://doi.org/10.1038/s41558-017-0050-4.
Smith, K. L., L. M. Polvani, and L. B. Tremblay, 2018: The impact of stratospheric circulation extremes on minimum Arctic sea ice extent. J. Climate, 31, 7169–7183, https://doi.org/10.1175/JCLI-D-17-0495.1.
Solomon, S., and Coauthors, 2007: Technical summary. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 19–91.
Stroeve, J. C., M. M. Holland, W. Meier, T. Scambos, and M. Serreze, 2007: Arctic sea ice decline: Faster than forecast. Geophys. Res. Lett., 34, L09501, https://doi.org/10.1029/2007GL029703.
Stroeve, J. C., V. Kattsov, A. Barrett, M. Serreze, T. Pavlova, M. Holland, and W. N. Meier, 2012: Trends in Arctic sea ice extent from CMIP5, CMIP3 and observations. Geophys. Res. Lett., 39, L16502, https://doi.org/10.1029/2012GL052676.
Thompson, D. W. J., and J. M. Wallace, 1998: The Arctic oscillation signature in the wintertime geopotential height and temperature fields. Geophys. Res. Lett., 25, 1297–1300, https://doi.org/10.1029/98GL00950.
Torrence, C., and G. P. Compo, 1998: A practical guide to wavelet analysis. Bull. Amer. Meteor. Soc., 79, 61–78, https://doi.org/10.1175/1520-0477(1998)079<0061:APGTWA>2.0.CO;2.
Vize, V. Y., 1944: Hydrological conditions in the ice edge area in the Arctic seas (in Russian). AARI Proc., 184, 125–151.
Walsh, J. E., 1978: A data set on Northern Hemisphere sea ice extent. Glaciological Data, Report GD-2, Part 1, National Snow and Ice Data Center, 49–51.
Walsh, J. E., and C. M. Johnson, 1979: An analysis of Arctic sea ice fluctuations, 1953–77. J. Phys. Oceanogr., 9, 580–591, https://doi.org/10.1175/1520-0485(1979)009<0580:AAOASI>2.0.CO;2.
Walsh, J. E., W. L. Chapman, and F. Fetterer, 2015, updated 2016: Gridded monthly sea ice extent and concentration, 1850 onward, version 1.1. National Snow and Ice Data Center, accessed 2019, https://doi.org/10.7265/N5833PZ5.
Walsh, J. E., F. Fetterer, J. Scott Stewart, and W. L. Chapman, 2016: A database for depicting Arctic sea ice variations back to 1850. Geogr. Rev., 107, 89–107, https://doi.org/10.1111/j.1931-0846.2016.12195.x.
Wang, R.-B., S. Li, and Z. Han, 2018: Evaluation of the HadISST1 and NSIDC 1850 onward sea ice datasets with a focus on the Barents-Kara Seas. Atmos. Oceanic Sci. Lett., 11, 388–395, https://doi.org/10.1080/16742834.2018.1504599.
Wei, Y., H. Yu, J. Huang, Y. He, B. Yang, X. Guan, and X. Liu, 2017: Comparison of the Pacific decadal oscillation in climate model simulations and observations. Int. J. Climatol., 38, e99–e118, https://doi.org/10.1002/joc.5355.
Williams, J., B. Tremblay, R. Newton, and R. Allard, 2016: Dynamic preconditioning of the minimum September sea-ice extent. J. Climate, 29, 5879–5891, https://doi.org/10.1175/JCLI-D-15-0515.1.
Woodgate, R. A., T. J. Weingartner, and R. Lindsay, 2010: The 2007 Bering Strait oceanic heat flux and anomalous Arctic sea-ice retreat. Geophys. Res. Lett., 37, L01602, https://doi.org/10.1029/2009GL041621.
Woodgate, R. A., T. J. Weingartner, and R. Lindsay, 2012: Observed increases in Bering Strait oceanic fluxes from the Pacific to the Arctic from 2001 to 2011 and their impacts on the Arctic Ocean water column. Geophys. Res. Lett., 39, L24603, https://doi.org/10.1029/2012GL054092.
WWF, 2014: Last ice area biophysical reader. World Wildlife Fund, https://arcticwwf.org/newsroom/publications/last-ice-area-biophysical-reader/.
Yan, X., R. Zhang, and T. R. Knutson, 2018: Underestimated AMOC variability and implications for AMV and predictability in CMIP models. Geophys. Res. Lett., 45, 4319–4328, https://doi.org/10.1029/2018GL077378.
Yu, L., S. Zhong, J. Winkler, M. Zhou, D. Lenschow, B. Li, X. Wang, and Q. Yang, 2017: Possible connections of the opposite trends in Arctic and Antarctic sea-ice cover. Sci. Rep., 7, 45804, https://doi.org/10.1038/srep45804.
Zhang, R., 2015: Mechanisms for low-frequency variability of summer Arctic sea ice extent. Proc. Natl. Acad. Sci. USA, 112, 4570–4575, https://doi.org/10.1073/pnas.1422296112.
Zhang, R., R. Sutton, G. Danabasoglu, Y.-O. Kwon, R. Marsh, S. G. Yeager, D. E. Amrhein, and C. M. Little, 2019: A review of the role of the Atlantic meridional overturning circulation in Atlantic multidecadal variability and associated climate impacts. Rev. Geophys., 57, 316–375, https://doi.org/10.1029/2019RG000644.
