The Pacific sector of the Arctic Ocean (PA, hereafter) is a region sensitive
to climate change. Given the alarming changes in sea ice cover during recent
years, knowledge of sea ice loss with respect to ice advection and melting
processes has become critical. With satellite-derived products from the
National Snow and Ice Center (NSIDC), a 38-year record (1979–2016) of the
loss in sea ice area in summer within the Pacific-Arctic (PA) sector due to
the two processes is obtained. The average sea ice outflow from the PA to the
Atlantic-Arctic (AA) Ocean during the summer season (June–September) reaches
As the Arctic climate warms (Comiso, 2010; Overland et al., 2010; Graham
et al., 2017), a wide range of researchers and the public show compelling
interest in topics associated with the drop in sea ice coverage (Kay and
Gettelman, 2009; Spreen et al., 2009, 2011; Polyakov et al., 2010; Zhang et al.,
2010; Woods et al., 2013; Tjernström et al., 2015;
Notz and Stroeve, 2016; Screen and Francis, 2016; Koyama et al., 2017;
Smedsrud et al., 2017; Stroeve et al., 2017; Niederdrenk and Notz, 2018).
For the period since the late 1970s, sea ice extent has been decreasing as
revealed from a series of satellite microwave observations, ranging from
Comiso (2011) reported a more negative trend in the multiyear ice (MYI)
coverage, approximately
Much attention is also paid to the decline in sea ice thickness in the
Arctic Ocean. The sea ice extent covered by MYI in March decreased from
approximately 75 % (mid-1980s) to 45 % (2011), and coverage fraction of
the oldest ice (no less than a fifth of a year) dropped from 50 % of the
multiyear ice cover to 10 % (Comiso, 2011). Such a large decline in
coverage of the thicker and older component of the MYI ice cover means a
decrease in the mean ice thickness in the Arctic Ocean. For example, placing
sea ice thickness derived from ICESat (2003–2008) in the context of a 43-year
submarine record (1958–2000), the overall mean winter thickness in a
sizable portion of the central Arctic Ocean shows a decline of 1.75 m in
thickness from 3.64 m in 1980 to 1.75 m from the ICESat record (Kwok
and Rothrock, 2009). Moored sonars in the Fram Strait also observed a
decrease in the annual mean sea ice thickness, from earlier 3.0 m (1990s) to
recent 2.2 m (2008–2011) (Hansen et al., 2013). Due to Arctic sea
ice age changes, satellite measurements from 2003 to 2008 (ICESat) and
CryoSat-2 (2011–2015) reveal that there are net reductions in Arctic ice
volume, of approximately
Although we are familiar with the fact that the sea ice mass balance is closely related to the dynamic (advection) and thermodynamic (melting) processes, quantitative knowledge about their contributions to the sea ice area changes within the Arctic Ocean is scarce. In addition the utilization of satellite observations, modeling studies have been commonly used to diagnose the dynamic and thermodynamic forcing (e.g., Lindsay et al., 2009). In this study, we examine the contributions of these two processes to the sea ice depletion within the PA side where sea ice loss is the most pronounced during summer (Cavalieri and Parkinson, 2012; Kawaguchi et al., 2014; Lynch et al., 2016; Comiso et al., 2017). In a preceding investigation, Kwok (2008a) examined the sea ice retreat within the PA sector due to melting and advection in the summers of 2003–2007. However, the record of 5 years is too short to draw any robust conclusion about the variability and trend in the sea ice area changes due to the two processes. Kwok (2008b) investigated the sea ice transport between the PA and Atlantic-Arctic (AA) sectors for the period 1979–2007. However, the contribution of ice melting to sea ice decline in PA is beyond the scope of their study. Taking advantage of a longer and updated version of sea ice motion data provided by the NSIDC, this study attempts to quantify the contributions of the two processes (advection and melting) to the retreat of sea ice in summer within the PA sector over the period 1979–2016. Moreover, the possible causes for their variability and trends are examined by highlighting the role of large-scale atmospheric circulation.
We organize this paper as follows. The data and method are summarized in Sect. 2. The estimates of sea ice outflow and melting are presented in Sect. 3. The connection between sea ice retreat within the PA sector and typical large-scale atmospheric circulation is analyzed in Sect. 4 by exploring the connection between the PA sea ice loss and the Arctic Oscillation (AO), North Atlantic Oscillation (NAO), and dipole anomaly (DA). Section 5 reiterates the key findings and concludes this study.
A gridded sea ice motion (SIM) product is provided by the National Snow and
Ice Data Center (NSIDC) (
Satellite-derived daily SIC records (1978–2017)
(
Following Kwok (2008a), the Arctic Ocean is divided into the PA and AA sectors (Fig. 1). The division is defined by a line linking the easternmost tip of Severnaya Zemlya and the southwestern tip of Banks Island. With a length of 2840 km, the line serves as the gateway through which the sea ice area flux between the two sides of the Arctic Ocean is calculated.
The fluxgate (red line) that is used to estimate sea ice area flux.
The blue lines represent the boundaries of the PA and AA regimes. The
endpoints in the North American and the Eurasian sides are (125.1
Sea ice area flux is taken as the integral product of gate-perpendicular SIM
and SIC for all grids across the fluxgate. The daily field of sea ice area
flux (
In our sign convention, sea ice transport from the PA to the AA sector is
referred to as a positive flux (i.e., outflow) and the reverse direction is
taken as a negative flux (i.e., inflow). Supposing that the errors in the
SIM grid samples are unbiased, additive, uncorrelated, and normally
distributed (Kwok, 2008b), the uncertainty of the daily fields can be
expressed as follows.
The uncertainty of the seasonal (
Expected mean uncertainties in seasonal and annual sea ice area flux
(10
According to Kwok (2008a), the melting sea ice area in PA is the
difference between the total sea ice area loss in PA and the sea ice area
flux from PA to AA. A possible discrepancy is caused by deformation
(divergence and convergence) processes, which can lead to reduction in sea ice
area that may be misclassified as sea ice loss due to melt or export. The
coarse resolution of satellite observations does not allow us to accurately
quantify the sea ice loss in relation to sea ice deformation. However, in
winter, sea ice area change due to deformation is expected to be negligible
due to solid pack ice (approximately 1 %–2 %)
(Kwok et al., 1999). In summer, the deformation may be larger. A
larger (smaller) convergence (divergence) is hypothesized north (south) of
80
The atmosphere circulation modes screened in this study for possible linkages
with sea ice area changes in PA include the AO (leading mode of sea level
pressure (SLP) north of 20
The changes in SLP can have significant impacts on winds and hence SIM
through their perturbations of other climate variables, such as surface air
temperature (SAT) and precipitable water (PW). All these data are obtained
from NCEP/NCAR in NOAA (Kalnay et al., 1996), with a grid size of
Comparison of regional sea ice exchanges between the PA and AA
sectors in
To give credence to our estimates, we compared the estimates with the results
reported by Kwok (2008b). He made use of the SIM data retrieved from the
37 GHz channels of the combined 29-year SMMR and SSM/I time series between
1979 and 2007. Overall, the two estimates agree well with respect to summer
and winter sea ice area fluxes (Fig. 2). As shown in Fig. 2, data pairs are
distributed close to the
The normalized monthly anomaly data have been adopted for studying the
monthly variability. As a result, the seasonal variability is removed and,
therefore, the distinct variations in monthly estimates are clearer and a
direct comparison of variability among different months is feasible. The
normalized or standardized procedure applied for the monthly anomaly fields
(
Normalized monthly anomaly fields of sea ice area flux. The anomaly field is calculated as the difference between the monthly estimate and the mean value of the same month computed over the period of interest. Then, the normalized anomaly field is obtained by dividing the monthly anomaly field by the standard deviation (SD) of the corresponding month over the 38-year period. The bottom row denotes the annual mean value of the normalized anomaly.
Figure 3 displays the variations in anomaly fields for monthly sea ice flux between the two subsectors in the Arctic Ocean. Overall, the temporal variability is high (Fig. 3). The first decade (1979–1988) is characterized by the occurrence of frequent negative anomaly fields, while for the remaining periods the emergence of positive anomalies seems to be more common (Fig. 3). In particular, negative anomalies are observed in nearly all months (except for February) in 1985, yet almost all months experience positive ice flow anomalies in 1995, 2007, and 2008. The frequently observed positive anomaly in recent periods compared to the early period (1979–1987) is further reflected in the annual mean normalized anomaly (bottom row in Fig. 3).
Frequency distribution histograms for months with different
normalized anomalies in sea ice area flux in different decades. The fraction
depicted as the percentage for each period, for example, (
The temporal variability in the monthly sea ice area flux fields is further emphasized in Fig. 4, where the frequency distribution histogram for months with different normalized anomaly amplitudes are displayed for different decades (P1: 1979–1988; P2: 1989–1998; P3: 1999–2008, P4: 2009–2016). The dominance of anomalous low sea ice area fluxes during the first decade is reflected as an asymmetric distribution pattern in frequency (Fig. 4a), with a larger number of months decreasing to the negative anomaly side. Comparatively, for the following three periods (Fig. 4b–d), the distribution pattern begins to become more symmetric, mainly because of the growing number of months with positive anomaly fields.
To depict the decadal evolution in frequency distribution, the individual
months for each period are binned into four groups with different anomaly
amplitude ranges (
Sea ice motion trend in
Monthly trends for the sea ice area flux between PA and AA. The
trends for the SIM and SIC fields over the fluxgate are also provided (% yr
Note:
The monthly trends in the sea ice area flux are listed in Table 2. The trend
in percentage for each month denotes the fraction of the monthly trend
estimate relative to the mean sea ice area flux for the corresponding month.
Basically, all months (except for January) show positive trends, with a mean
value of 2.52 % yr
Figure 6 shows the estimates of the sea ice area transport for different
seasons and years between the two sectors. The average annual sea ice export
of
Annual and seasonal sea ice area fluxes between the PA and AA
sectors over the period 1979–2016. The annual ice flow (annual cycle) is
shown by the green line. Summer (from June through September) and winter
(from October through May) fluxes are denoted with blue and red lines,
respectively. The dashed lines representing the linearly fitted trends are
obtained through the application of the least-squares method. The labels
Over the 38-year period, significant positive trends are observed for the sea
ice area fluxes during the summer and winter seasons (Fig. 6). Over the
38-year period, the winter sea ice export exhibits a positive trend of
SIM trends during
Cross-gate SIM climatology (black solid line) and trends (arrows)
for
Due to the extensive coverage in longitudes and latitudes across the
investigated fluxgate (2840 km), regional variations in the trend of the
across-gate sea ice area flux fields are expected. Broadly, the Arctic sea
ice circulation is characterized by the Beaufort Gyre (BG) and transpolar
drift stream (TDS). How does the SIM trend vary in these two regimes? To
answer this question, we not only present the overall pattern of spatial
distribution of SIM trends over the entire Arctic Ocean (Fig. 7), but also
depict the details of the cross-gate SIM trends in fields (Fig. 8). Figure 7 shows that SIM increases in the BG and TDS regimes during both winter
(Fig. 7a) and summer (Fig. 7b) seasons. In particular, the increasing
SIM trends in the narrow southern arm of the BG regime appear in winter
(Fig. 7a) and summer (Fig. 7b) The SIM trends during winter (summer) of
approximately 0.12 cm s
Sea ice melting is a sensitive indicator reflecting the warmer Arctic
climate. An estimate of melting area of sea ice (dotted–dashed line in Fig. 9) is obtained as the difference between the observed sea ice area loss in
the PA (red line in Fig. 9) and the sea ice area flux through the fluxgate
(blue line in Fig. 9) during the summer season (June–September). Over the
38-year period, the mean melting area is
Selected daily changes in sea ice areas within the PA (red) and AA (green) during summer (June–September) for the period 1979–2016. The horizontal dashed line represents the sea ice area in the PA on 1 June, which is used as a benchmark to measure the sea ice area changes due to melting. The cumulative daily sea ice area flux, with reference to the top dashed line, is shown as a blue line. The melting sea ice area, denoted by the vertical dotted–dashed line, is taken as the difference between the total decline in sea ice area within the PA and the accumulated flux through the gate.
With the sea ice area budget shown in Fig. 9, we quantify the relative
contribution to sea ice area changes in summer due to the advection and
melting processes over the period of 1979–2016 (i.e., melting (
Fractions of sea ice area loss in the PA sector that are related to the sea ice melting and advection processes.
The trend of melting sea ice area within the PA is apparent over the 38-year
period (Fig. 11), with an overall positive trend of 3.20 % yr
Time series for melting sea ice areas within PA.
Additionally, note that the day of year (DOY) of the minimum sea ice area in
the PA sector displays an overall positive trend of 0.29 d yr
Day of year (DOY) of the annual minimum sea ice area in PA.
Wind forcing has been significant in modulating the sea ice variability in summer (Ogi et al., 2016). As an example, sea ice depletion induced by the summer melting process is connected to the wind forcing, which can help to cause a warmer Arctic climate through the advection of warmer and moister air from the south (Wang et al., 2005; Zhang et al., 2013; Lee et al., 2017). The Arctic-wide wind forcing is linked to large-scale atmospheric circulation patterns (Zhang et al., 2008; Overland and Wang, 2010; Stroeve et al., 2011). Hence, the connection between the sea ice area loss in the PA sector and three typical atmospheric indices (AO, NAO, and DA) is assessed here. NAO and AO represent the dominant atmospheric circulation modes in guiding sea ice movement and Fram Strait export before 1994 (Rigor et al., 2002; Nakamura et al., 2015), while the DA seems to play a leading role over the latter period after 1995 (Wang et al., 2005). Here our objective is to examine how the sea ice variability due to advection and melting processes is quantitatively related to the interannual and decadal changes in these atmospheric modes. Furthermore, the potential impacts on sea ice loss in related climatic variables (SAT, SLP, and PW) coupled with different atmospheric circulation patterns are highlighted.
Variations in and trends of the mean atmospheric indices in summer
(June–September), including
Overall, the interannual variability for the three indices is large, and two
indices, NAO and DA, reveal significant trends during summer (Fig. 13).
Over the investigated 38-year period, the sea ice area reduction in summer
within the PA seems to have been slightly influenced by AO fluctuations,
with a low correlation coefficient (
Correlations between summer mean atmospheric index and total sea ice area loss, and sea ice area decline due to melting and advection processes over the period 1979–2016.
The influence of NAO-associated atmospheric circulation in winter on sea ice
changes in summer has been broadly documented (Kwok, 2000; Jung and
Hilmer, 2001; Parkinson, 2008), which is especially clear for the period
from the late 1980s to early 1990s (Jung and Hilmer, 2001) when NAO was
in its peak positive phase. In this study, the summer NAO is negatively
correlated with the summer retreat of sea ice area within the PA sector (
The changes in typical climate variables (SAT, SLP, and PW) over the
1979–2016 period (the first column, panels
As there is a clear negative trend in the summer mean NAO index (Fig. 13b), one may wonder how the summer NAO index trend is related to the
increasing loss of sea ice within the PA sector. As shown in Fig. 14g, the
NAO-associated SLP distribution pattern, with a greater SLP in the western
Arctic near the Canadian Archipelago, is favorable for a pattern of
anticyclonic atmospheric circulation within the PA. Such a clockwise pattern
of atmospheric circulation is hypothesized to push more ice from the eastern
and northern Beaufort Sea to the western and southern Chukchi Sea (Figs. 7b and 8b), where extensive sea ice melting has been commonly observed
(Kwok, 2008a). The moderate negative correlation between NAO and melting
sea ice area (
The overall association between sea ice area variability and the DA index is
comparatively robust (
Associated with the DA-associated SLP distribution pattern, the emergence of
enhanced BG circulation is identified in the Beaufort Sea (Fig. 14f). In
particular, the sea ice drifts in the southern arm of the BG regime are
largely strengthened (Fig. 7b), which contributes to the large inflow of
sea ice to the southern Beaufort Sea (Fig. 8b), where sea ice undergoes
dramatic melting processes in summer as indicated by the remarkable SAT
increase there (Fig. 14a). This coupled mechanism between dynamic
(advection) and thermodynamic (melting) processes resembles that caused by
the NAO. As a consequence, the depletions in sea ice in summer due to both
melting and advection are relatively strongly correlated to the DA index of
summer, with
Correlations between summer mean atmospheric index and sea ice minimum area in the PA sector for the different decades and the whole 38-year period.
Since AO shows a negligible trend (Fig. 13a), the 38-year climatic changes
related to AO are insignificant (Fig. 14d, h, and i) and smaller in
magnitude compared to the DA- and NAO-associated changes. The NAO pattern
is conventionally deemed a regional index, representing parts of the
broader AO pattern. However, NAO-associated SLP changes (Fig. 14g) show a
stronger gradient across the fluxgate than that of AO-associated SLP (Fig. 14h), which would favor more sea ice outflow from the PA to AA sectors. In
comparison with NAO (Fig. 14g), the AO-associated SLP distribution shows a
much weaker gradient across the Arctic Ocean (Fig. 14h), although it may
contribute to the sea ice advection from the Pacific side to the Atlantic
side. As a result, lower correlations between AO and sea ice melting and
advection processes are expected, with
The temporally varying association between atmospheric circulation and sea ice drift reflects Arctic climate changes. How does the linkage vary with time? Does it remain stable? To answer these questions, correlations between different summer mean atmospheric indices and minimum sea ice areas in the PA sector over different decades are obtained (Table 4). The AO effects are relatively unstable, imposing clear and decreasing positive effects over the first 3 decades but reversing to a negative moderate impact in the last period (P4) (Table 4). A clear association with DA arises during the latter two periods: P3 (1999–2008) and P4 (2009–2016). In contrast, the NAO index appears to have a significant impact only during the last period (2009–2016) (Table 4).
PW serves as an important indicator of Arctic climatic conditions. In view
of the distribution of PW changes (Fig. 14i), we find it broadly agrees
with that of SAT in the PA sector (Fig. 14a). For example, over the
Siberian marginal seas, warmer SATs accompany more PWs. If the PW
constituents drop to the surface as rain during a warming summer, they may
benefit the melting process of the local ice/snow surface by lowering the
surface albedo of ice/snow. Compared with the NAO-associated changes (Fig. 14k), the spatial distributions of DA-associated SAT (Fig. 14b) and PW
changes (Fig. 14j) are, to a greater degree, consistent with those of the
summer SIC trends (Fig. 15), particularly over the area throughout the
marginal seas in the PA sector, such as the Beaufort Sea, Chukchi Sea, East
Siberian Sea, and Laptev Sea where the rates of sea ice area decline
approach 2.0 % yr
Summer SIC trends over the period 1979–2016.
To summarize this section, we found different atmospheric forcing patterns
exert varying influences on summer sea ice variability. Overall, the
connections are relatively strong between sea ice loss and DA (
Using the new version (v3.0) of NSIDC products (SIM and SIC), we quantify
the contributions of the advection and melting processes to the sea ice
retreat within the PA sector over the period 1979–2016. A synoptic view of
their 38-year variability and trends on different timescales is presented.
Over this period, the annual (October to following September) mean sea ice
export is
The linkage between sea ice loss and wind forcing associated with different large-scale atmospheric circulations during the summer season is explored. Overall, the AO is weakly connected to the sea ice loss on the PA sides, while the NAO is moderately correlated with the sea ice decline caused by melting. In contrast, the DA shows a more robust connection with the sea ice decrease in the PA through influence on both sea ice melting and advection. Dynamically, the DA-associated SLP conveys heat and moist air from the south to the north, resulting in the increasing SAT and PW over the marginal seas of the PA sector and contributing to the significant sea ice retreat. In addition, the positive trend in the DA induces stronger meridional wind forcing over the transpolar stream, leading to increased sea ice outflow and more ice decline in the PA. Thermodynamically, both the DA and NAO indices are associated with a strengthened anticyclonic SLP pattern over the southern Beaufort Sea. This feature will promote the westward transport of sea ice from the Canadian Basin to the south of the Beaufort Sea and Chukchi Sea where extensive melting sea ice has been detected in summers during the recent decade. By contrast, AO-associated sea ice changes due to melting and advection processes are not distinct, although temporally robust correlation is expected (Table 4).
The significant sea ice retreat also plays a crucial role in triggering regional responsive feedback in the atmosphere (Overland and Wang, 2010), as indicated by the warmer SAT (Fig. 13a) and decreased SLP (Fig. 13d) observed on the broad Siberian Arctic Ocean side. Therefore, if the current distinct sea ice loss on the Pacific-Arctic Ocean persists, the diminishing SLP in areas over the Laptev Sea, one of the centers of action of the DA, will further enhance the positive DA trend. Consequently, stronger dynamic (advection) and thermodynamic (melting) effects associated with the DA on sea ice retreat within the PA sector are probably foreseen in the predictable future.
NSIDC sea ice motion is available at
HB and QY led the analysis and integrated the data. XL, LZ, YW, and YL process the sea ice motion and sea ice concentration data. HB, XL, and LZ performed the area flux calculations. HB, QY, and HH drafted the paper. All authors discussed the results and commented on the paper.
The authors declare that they have no conflict of interest.
We thank for the following organizations for providing the data used in this study. NSIDC provided the satellite-derived ice motion and concentration data, and the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) provided the reanalysis product.
This research has been supported by the National Natural Science Foundation of China (grant no. 41406215) and the NSFC Shandong Joint Fund for Marine Science Research Centers (grant no. U1606401).
This paper was edited by John Yackel and reviewed by two anonymous referees.