Abstract

This analysis evaluates the thermal state of the intermediate (depth range of 150–900 m) Atlantic Water (AW) of the Arctic Ocean, beginning in the 1950s and with particular focus on the transition from the 1990s to the 2000s and on changes during the 2000s. Using an extensive array of observations, the authors document AW warming trends across various time scales and demonstrate that the 2000s were exceptionally warm, with no analogy since the 1950s or probably in the history of instrumental observations in the Arctic Ocean. Warming in the recent decade was dominated by a warm AW pulse in addition to the underlying trend. Since 1997, the Canadian Basin experienced a faster warming rate compared with the Eurasian Basin. The relative role of the AW warmth in setting the net energy flux and mass balance of the Arctic sea ice is still under debate. Additional carefully orchestrated field experiments are required in order to address this question of ongoing Arctic climate change.

1. Introduction

This analysis has been motivated by the recent study of Bourgain and Gascard (2012), who analyzed an extensive collection of measurements of intermediate (depth range of ~150–900 m) Atlantic Water (AW; Fig. 1) temperatures from 1997 to 2008 and found no AW warming trend in the Eurasian Basin of the Arctic Ocean. AW supplies vast quantities of heat to the Arctic Ocean, and it is still debatable how much of this heat penetrates upward through the stable Arctic halocline and reaches sea ice (for discussion, see, e.g., Polyakov et al. 2012). The finding of Bourgain and Gascard (2012) is of interest because recent Fram Strait measurements have suggested that, starting from 1999, the Eurasian Basin received increasing amounts of heat through Fram Strait (Schauer et al. 2004, 2008) and, according to Polyakov et al. (2005, 2010, 2011), this additional supply of warmth resulted in increased AW temperatures along the AW pathway over the Siberian continental slope into the Arctic Ocean interior. These inconsistencies motivated us to evaluate the AW thermal state starting from the 1950s, with particular focus on the transition from the 1990s to the 2000s and changes during the 2000s. Using an extensive array of historical and modern observations, we document AW warming trends over various time scales and demonstrate that the 2000s were exceptionally warm, with no analogy in the history of Arctic Ocean instrumental observations.

Fig. 1.

Circulation of surface water (blue) and intermediate AW (red and pink) of the Arctic Ocean. The submarine Lomonosov Ridge, running from Siberia to Greenland, separates the Makarov and Eurasian Basins. A deeper midocean ridge, the Nansen–Gakkel Ridge, roughly parallel to the Lomonosov Ridge, divides the Eurasian Basin into the Nansen Basin (near the Barents Sea) and the Amundsen Basin (along the Lomonosov Ridge); neither is marked here. Yellow lines show position of cross sections used for estimates presented in Fig. 5. This figure is updated from Polyakov et al. (2011).

Fig. 1.

Circulation of surface water (blue) and intermediate AW (red and pink) of the Arctic Ocean. The submarine Lomonosov Ridge, running from Siberia to Greenland, separates the Makarov and Eurasian Basins. A deeper midocean ridge, the Nansen–Gakkel Ridge, roughly parallel to the Lomonosov Ridge, divides the Eurasian Basin into the Nansen Basin (near the Barents Sea) and the Amundsen Basin (along the Lomonosov Ridge); neither is marked here. Yellow lines show position of cross sections used for estimates presented in Fig. 5. This figure is updated from Polyakov et al. (2011).

2. Data and methods

In this study, we used Arctic Ocean temperature observations collected from 1950 to 2011. We have substantially updated the dataset used in our previous studies of long-term changes in the AW (Polyakov et al. 2004) and freshwater (Polyakov et al. 2008) content of the Arctic Ocean. These updates include several thousand oceanographic stations for the deep basin, mostly from the 2000s [cf., e.g., the number of stations used for composite time series of the AW temperature shown in Fig. 2 of Polyakov et al. (2004) and in Fig. 3]. Spatial data coverage is shown in Fig. 2.

Fig. 2.

Maps showing data coverage of oceanographic profiles (red dots). Depth is shown by 500- and 2000-m contours.

Fig. 2.

Maps showing data coverage of oceanographic profiles (red dots). Depth is shown by 500- and 2000-m contours.

The 1950s became the first decade during which the entire Arctic Ocean was covered by observations based on systematic winter (March–May) aircraft expeditions and year-round drifting stations. The aircraft-based program peaked in the 1970s, when seven pan-Arctic surveys comprising more than 1000 oceanographic stations were carried out. In the 1990s, icebreakers and submarines provided high-quality measurements covering vast areas of the central Arctic Ocean (Fig. 2). A significant increase the oceanographic observations was achieved in the 2000s, culminating during the International Polar Year 2007/08. Ship-based summer measurements in the 2000s were complemented by Ice-Tethered Profiler (ITP; http://www.whoi.edu/itp) drifters, providing year-round extensive conductivity–temperature–depth (CTD) measurements in the upper ~800 m. Not all data shown in Fig. 2 were used in our analysis. For example, in order to reduce excessive spatial density of the ITP observations, every other ITP profile was used. The AW is not present on shelves. That is why only deep-basin observations were utilized. The total number of measurements used in this analysis is denoted by the numbers at the bottom of Fig. 3. The number of stations used for estimates for specific time intervals is provided in the text.

Fig. 3.

Long-term variability in AWCT. Gray numbers represent normalized AWCT anomalies for the 10 regions outlined in Fig. 1 of Polyakov et al. (2004). Annual values of normalized AWCT anomalies are shown by the red solid line, and red dotted lines show 98% confidence intervals. Blue horizontal lines show decadal-mean anomalies. Numbers at the bottom denote the 5-yr-averaged number of stations used in the data analysis. This figure is adapted and updated from Polyakov et al. (2004).

Fig. 3.

Long-term variability in AWCT. Gray numbers represent normalized AWCT anomalies for the 10 regions outlined in Fig. 1 of Polyakov et al. (2004). Annual values of normalized AWCT anomalies are shown by the red solid line, and red dotted lines show 98% confidence intervals. Blue horizontal lines show decadal-mean anomalies. Numbers at the bottom denote the 5-yr-averaged number of stations used in the data analysis. This figure is adapted and updated from Polyakov et al. (2004).

Most observations prior to the mid-1980s were made using Nansen bottles. Typical measurement errors are 0.01°C for temperature and 0.02 for titrated salinity. Polyakov et al. (2003) found that these data define the AW core temperature (AWCT; the maximum AW temperature in the profile) quite accurately. For example, the course-resolution data allows the AWCT to be defined with a 0.98 correlation between time series of coarse- and fine-resolution temperature profiles. More recent oceanographic measurements were made using CTD instruments, which have increased vertical density and accuracy of temperature (0.001°C) and salinity (0.003 psu) measurements.

Our analysis utilized observations from two moorings. One of them was located in Fram Strait (78°50′N, 8°20′E), the gateway of AW inflow into the Arctic Ocean. These mooring observations began in 1997 (Fahrbach et al. 2001). A McLane moored profiler (MMP) has been used at the central Laptev Sea slope, moored 2002–11 (M1 mooring: 78°27′N, 125°40′E). The records from these two moorings were used in our previous publications (e.g., Polyakov et al. 2011). In this analysis, we use these records updated by data collected in 2009–11. The accuracy of Aanderaa instruments used for temperature measurements in Fram Strait is ~0.05°C, whereas the accuracy of the MMP temperature registration is 0.002°C.

The pan-Arctic time series of the normalized AWCT is composed using a technique similar to the method used for successful analysis of long-term AW and freshwater content variability (Polyakov et al. 2004, 2008). Using this method, the Arctic Ocean is divided into 10 boxes of approximately equal areas, and individual (snapshot) AWCT anomalies in these boxes were averaged within a given year and box to produce 10 regional time series. Each spatially and annually averaged value was then reduced to an anomaly relative to its regional mean and normalized by regional standard deviation. The resulting normalized regional time series for each box are averaged again, taking into account the size of each box, to obtain a single pan-Arctic time series. This technique provides an accurate spatial representation of area-averaged indices, since these results are less skewed by nonhomogeneity of spatial data coverage. Normalization is used in order to avoid problems caused by data gaps, particularly in the early parts of the records, when regional time series showing strong differences in the magnitude of variations are averaged in order to produce the pan-Arctic time series. For example, during the cold 1960s, there were only a few observations in the Canada Basin, whereas data coverage in the eastern Arctic, particularly in the vicinity of Fram Strait, was better (Fig. 2). Thus, averaging the anomalous nonnormalized regional AWCTs from the eastern Arctic, where relatively high variability was found, would result in exaggerated negative pan-Arctic anomalies in the 1960s. The method described above is designed to avoid such anomalies.

3. Results

Composite time series of AWCT shows that warming began in the late 1970s (Fig. 3). The linear trend evaluated by the least squares best-fit method has been 0.48 ± 0.05 per decade since 1970. The decadal means shown by blue horizontal lines in Fig. 3, based on simple averaging of all available normalized AWCT anomalies within each decade, capture this warming tendency as well. For example, the coldest decade was the 1970s (decadal mean = −0.62 ± 0.15; all statistical confidence intervals discussed in the text are provided for a 95% level). The warming that began in the second half of the 1980s made this decade warmer than the 1970s (decadal mean = −0.42 ± 0.09). This warming continued into the 1990s, showing the steepest temperature increase (decadal mean = +0.46 ± 0.15). The 2000s were even warmer than the 1990s representing the warmest decade in the history of instrumental AWCT observations since 1950 (decadal mean = +0.90 ± 0.25). These statistical estimates thus provide solid evidence for steady warming of the Arctic Ocean interior since the 1970s.

Comparing 1997–2008 observations, Bourgain and Gascard (2012) argued that there is no AWCT warming trend. This conclusion is inconsistent with our analysis of the long-term AWCT trend and decadal means and findings from previous studies (e.g., McLaughlin et al. 2009; Polyakov et al. 2005, 2010, 2011). This is why we augment our analysis using comparison of AW warming in the 1990s and 2000s. Figure 4 shows decadal-mean AWCTs and their anomalies for the 1970s, 1990s, and 2000s. In preparation of this figure (as well as Fig. 6), we have intentionally presented “raw” (i.e., without use of interpolation; as in Bourgain and Gascard 2012) AW temperatures and their anomalies in recent decades as colored circles, thus avoiding additional errors associated with spatial interpolation. Furthermore, comparison of AWCTs between different time periods is made using pairs of closest stations found within a search radius Rsearch (Figs. 4f, 6h,i). With few exceptions (probably due to synoptic-scale features like eddies), AW in all Arctic Ocean regions was warmer in the 1990s and 2000s, relative to the 1970s (Figs. 4d,e). Anomalies were particularly strong in the Eurasian Basin, with a somewhat warmer Nansen Basin and cooler Amundsen Basin in the 2000s compared with the 1990s. Figure 4f corroborates this conclusion, showing a generally warmer western Arctic Ocean and Nansen Basin (except for several locations in the northern Laptev Sea) and a cooler Siberian part of the Makarov Basin and eastern and central Amundsen Basin during the last decade. Despite evident spatial heterogeneity, point-to-point comparison demonstrates that the 2000s were warmer by ~0.13°C than the 1990s, thus providing a statistically significant estimate of pan-Arctic warming (Fig. 4f).

Fig. 4.

(a) AWCT (°C) averaged over the 1970s (color) and standard errors (black lines; 0.05°, 0.1°, 0.2°, and 0.3°C isolines are shown); (b),(c) AWCT (°C) from the 1990s and 2000s; (d),(e) 1990s and 2000s AWCT anomalies (°C) relative to the 1970s; (f) AWCT differences between the closest pairs of stations within a 100-km search radius from the 2000s and 1990s [the location of each point in (f) is the center between each pair of stations]. Statistically significant estimates of anomalies (at 95% confidence level) in (d),(e) are marked by larger circles (there are only a few stations with statistically insignificant anomalies).

Fig. 4.

(a) AWCT (°C) averaged over the 1970s (color) and standard errors (black lines; 0.05°, 0.1°, 0.2°, and 0.3°C isolines are shown); (b),(c) AWCT (°C) from the 1990s and 2000s; (d),(e) 1990s and 2000s AWCT anomalies (°C) relative to the 1970s; (f) AWCT differences between the closest pairs of stations within a 100-km search radius from the 2000s and 1990s [the location of each point in (f) is the center between each pair of stations]. Statistically significant estimates of anomalies (at 95% confidence level) in (d),(e) are marked by larger circles (there are only a few stations with statistically insignificant anomalies).

This estimate is statistically robust. For example, we tested its sensitivity to the choice of Rsearch and found that a change in Rsearch from 50 to 150 km with a 25-km increment led to statistically indistinguishable changes in the AWCT difference. The best comparison is provided when Rsearch is minimal. However, small Rsearch eliminates a good deal of measurements from comparison, leading to poor spatial coverage and thus reducing confidence of the statistical estimates. This is why we have used estimates derived for Rsearch = 100 km. We also evaluated the impact of data coverage on the computed estimate. For Rsearch = 100 km, we recalculated the temperature difference, randomly selecting approximately half of the originally used stations (so that the number of pairs used for the comparison dropped from 4811 to 2395). The estimate remained insensitive to this procedure (cf. 0.133° ± 0.005°C for all available pairs and 0.135° ± 0.007°C for the halved dataset), suggesting that the decadal data coverage was more than adequate for this analysis. Thus, these estimates provide solid statistical ground for our conclusion that the interior Arctic Ocean was warmer in the 2000s than in the 1990s.

We will now evaluate AWCT evolution in the 2000s. This decade was dominated by a warm AW pulse, which entered the Nansen Basin in 1999 (Schauer et al. 2004; Fig. 5). Further observations showed the propagation of this anomaly into the polar basin interior, following the Eurasian Basin bathymetry in a cyclonic sense so that this pulse of warm AW water reached the eastern Eurasian Basin in 2004 (Polyakov et al. 2005; Dmitrenko et al. 2008; Fig. 5). The temperature reached its maximum along the AW path in the Eurasian Basin in 2006–08 [see Fig. 2 of Polyakov et al. (2011) and Fig. 5]. The composite pan-Arctic time series suggests that 2006 was, overall, the warmest year (Fig. 3). Point-to-point comparison of AWCT measurements from the 2000s also suggests that 2006 was warmer than 2005 by ~0.15°C and warmer than 2007 by ~0.14°C. There is, however, a great deal of spatial heterogeneity. For example, Fig. 2 of Polyakov et al. (2011) and Fig. 5 here demonstrate that 2006 was the warmest year in the western Nansen Basin, whereas the maximum warming in the eastern Eurasian Basin was in 2007; local warm patches were found in 2008 as well (e.g., the temperature peak at ~125°E in Fig. 5).

Fig. 5.

Time series of AWCT anomalies (°C) relative to the time series means from oceanographic sections (blue) and mooring observations (red) from (top) Laptev Sea slope (~125°E) and (middle) Fram Strait. Dashed green lines show linear trends based on mooring records. (bottom) Normalized 3-yr running mean AWCT anomalies for the whole Arctic Ocean and Eurasian and Canadian Basins. Trends are shown for 1997–2010.

Fig. 5.

Time series of AWCT anomalies (°C) relative to the time series means from oceanographic sections (blue) and mooring observations (red) from (top) Laptev Sea slope (~125°E) and (middle) Fram Strait. Dashed green lines show linear trends based on mooring records. (bottom) Normalized 3-yr running mean AWCT anomalies for the whole Arctic Ocean and Eurasian and Canadian Basins. Trends are shown for 1997–2010.

Based on these considerations, for our further analysis of AWCT anomalies in the 2000s we selected three periods: 2000–05 (i.e., years prior to the AW pulse peak), 2006–07 (peak years), and 2008–10 (postpeak years). All three periods were warmer when compared to the “climatology” of the 1970s (Fig. 6). Despite strong spatial heterogeneity, Figs. 6h,i provide evidence that AWCT during the peak years of 2006–07 was higher than during the adjacent periods. The estimates of pan-Arctic temperature differences between the peak years and prepeak and postpeak years shown in Figs. 6h,i are robust, experiencing low sensitivity to the choice of Rsearch, with statistically indistinguishable estimates for Rsearch varying from 50 to 150 km. We also found that a random reduction by half in the number of pairs used in the comparative analysis does not lead to statistically significant changes. Thus, our estimates show that, in 2006–07, the AWCT was the highest of any year in the last decade.

Fig. 6.

(a) AWCT (°C) averaged over the 1970s (color) and standard errors (black lines shown are 0.05°, 0.1°, 0.2°, and 0.3°C isolines); AWCT (°C) for (b) 2000–05, (c) 2006–07, and (d) 2008–10; AWCT anomalies (°C) over (e) 2000–05, (f) 2006–07, and (g) 2008–10, relative to the 1970s; AWCT differences (h) between 2006–07 and 2000–05 and (i) between 2006–07 and 2008–10, between the closest pairs of stations within a 100-km search radius (the location of each point is the center between each pair of stations). Statistically significant estimates of anomalies (at 95% confidence level) shown in (e)–(g) are marked by larger circles (there are only a few stations with statistically insignificant anomalies).

Fig. 6.

(a) AWCT (°C) averaged over the 1970s (color) and standard errors (black lines shown are 0.05°, 0.1°, 0.2°, and 0.3°C isolines); AWCT (°C) for (b) 2000–05, (c) 2006–07, and (d) 2008–10; AWCT anomalies (°C) over (e) 2000–05, (f) 2006–07, and (g) 2008–10, relative to the 1970s; AWCT differences (h) between 2006–07 and 2000–05 and (i) between 2006–07 and 2008–10, between the closest pairs of stations within a 100-km search radius (the location of each point is the center between each pair of stations). Statistically significant estimates of anomalies (at 95% confidence level) shown in (e)–(g) are marked by larger circles (there are only a few stations with statistically insignificant anomalies).

Finally, we note a strong correlation between AWCT and AW heat content Q (J m−2), which is defined as a deviation in potential temperature from the freezing point multiplied by water density and the specific heat of water and integrated over a selected depth range. For example, using ~13 000 oceanographic stations, we found that AWCT and Q defined for the layer limited by 0°C from the top and 400 m from the bottom are correlated at R = 0.88. This estimate is robust, changing to R = 0.87 when a subset of pre-1990 data is used (thus eliminating more than 9000 modern stations); to R = 0.85 when the bottom limit of the layer is defined at 500 m; or to R = 0.84 when only Eurasian Basin data are used. The regression relationship between Q and AWCT is (J m−2).

Here, Q may serve as a useful measure of AW thermal state providing a broader perspective on the documented AW changes. AWCT anomalies expressed in terms of changes of QQ) suggest that the upper part of the AW (<400 m) gained about 196 MJ m−2 of heat since the 1990s. Increase of the layer thickness by choosing 700 m as the lower-layer bound provides a ~22% increase of ΔQ. Interannual changes of Q in the 2000s were of comparable magnitude. This anomalous Q integrated over the area of the deep Arctic Ocean (~5.2 × 106 km2) is equivalent to 1.0 × 1021 J, which is roughly 50% of variations (measured by standard deviation) of the advective annual horizontal atmospheric heat transport through 60°N (Trenberth and Caron 2001). It is also comparable to the 0–700-m global ocean heat gain since the early 1990s estimated as ~303 MJ m−2. The latter estimate was derived from Fig. 3.7 of Levy (2009), dividing ~1.1 × 1023 J by the area of the global ocean (~361 × 106 km2).

4. Discussion and conclusions

Analysis of available 1950–2010 data shows a steady increase of AW temperature since the late 1970s. Furthermore, there are solid statistical grounds supporting the statement that the 2000s was the decade with the warmest AW since the 1950s. Polyakov et al. (2004), analyzing a century-long record of AWCT, found that the last 10 yr of their record (limited by 2003) were the warmest. Thus, keeping in mind all potential caveats when using measurements from the first half of the twentieth century, with limited spatiotemporal coverage, there are reasons to believe that AW during the last decade was probably the warmest in the history of Arctic Ocean instrumental observations, initiated by Nansen in the 1890s.

Can the spatial distribution of AWCT differences in the 2000s (Figs. 6h,i) provide a hint on mechanisms governing the observed changes? In the Canada Basin, the spatial pattern of AWCT differences suggests that the warm anomaly was separated from the Chukchi Plateau slope in the early 2000s and then crossed the central basin in the southeast direction (Figs. 6h,i). This circulation pattern is consistent with the previous findings of McLaughlin et al. (2009). An almost synchronous cooling of the Nansen Basin after ~2007 is probably a manifestation of nonadvective processes modulating AWCT changes in this region. Polyakov et al. (2011) argued that the observed cooling cannot be explained by the influx of colder AW, because this would require an unrealistically rapid propagation of water from the Fram Strait along the slope downstream. They suggested that enhanced shelf–basin interactions seem to fit the observed AW cooling pattern. In the Amundsen and Makarov basins, we found no clear pattern of changes, with patches of cool and warm anomalies spread seemingly at random across the basins’ interiors.

Spatial heterogeneity of the AWCT anomalies in the central Arctic Ocean in the 2000s was most likely one of the reasons for the conclusion made by Bourgain and Gascard (2012) regarding the lack of an AW warming trend from 1997 to 2008. The authors, however, did not substantiate their conclusion, inferred from visual inspection of AW temperature maps by statistical analysis. Furthermore, their analysis underscores the inherent difficulty in differentiating between trends and strong fluctuations. During the last decade, AWCT changes were dominated by the decadal-scale warm AW pulse, in addition to the underlying trend (Fig. 5). Selecting periods for their composite AWCT maps, Bourgain and Gascard (2012) separated the two warmest years, 2006 and 2007, between two periods, thus contaminating their analysis of trends with the signal coming from decadal-scale variations. Despite problems with separating trends from strong variability, our local and composite pan-Arctic, Eurasian, and Canadian Basin AWCT time series do show warming since 1997, with a somewhat weaker trend in the Eurasian Basin (Fig. 5).

The analysis of long-term and recent AW changes presented here is subject to certain limitations. It assumed, for example, that seasonal AW variations are small and do not affect our results. This is probably true for vast areas of the Canadian and Amundsen Basins (Lique and Steele 2012). In the Nansen Basin, the seasonal signal is strong, particularly over the slope off Svalbard (Ivanov et al. 2009; Dmitrenko et al. 2009). We argue that measurements over the Nansen Basin slope were made in the summer, so the monthly temperature difference is on the order of 0.05°C; in the vicinity of Fram Strait this may be not true, however, and our estimates may be affected by the seasonal signal. Recent years provided unprecedented spatial data coverage; however, even in the 2000s AWCT maps bear numerous gaps so that spatial sampling is not often adequate for addressing important questions about the nature of spatial AWCT anomalies. Low sensitivity of our estimates to substantial (~50%) reductions in observations may be partially caused by highly correlated profiles taken in limited areas with densely populated measurements. At the same time, large areas like the Makarov Basin remained poorly covered by observations even during the successful 2000s. Thus, for climate monitoring, spatially distributed measurements may be preferable.

One of the fundamental questions of modern polar oceanography regards the relative role of oceanic heat fluxes in setting the net energy flux and the mass balance of Arctic sea ice. Lack of observation places serious limitations on our ability to understand and evaluate this potentially important thermodynamic link, and carefully orchestrated field experiment is needed to address this key question of ongoing Arctic climate change.

Acknowledgments

This study was supported by JAMSTEC (IP) and JAXA and NASA (IP and AP) grants. We thank M. McPhee and two anonymous reviewers for useful comments.

REFERENCES

REFERENCES
Bourgain
,
P.
, and
J. C.
Gascard
,
2012
:
The Atlantic and summer Pacific waters variability in the Arctic Ocean from 1997 to 2008
.
Geophys. Res. Lett.
,
39
,
L05603
,
doi:10.1029/2012GL051045
.
Dmitrenko
,
I. A.
, and
Coauthors
,
2008
:
Toward a warmer Arctic Ocean: Spreading of the early 21st century Atlantic Water warm anomaly along the Eurasian Basin margins
.
J. Geophys. Res.
,
113
,
C05023
,
doi:10.1029/2007JC004158
.
Dmitrenko
,
I. A.
, and
Coauthors
,
2009
:
Seasonal modification of the Arctic Ocean intermediate water layer off the eastern Laptev Sea continental shelf break
.
J. Geophys. Res.
,
114
,
C06010
,
doi:10.1029/2008JC005229
.
Fahrbach
,
E.
,
J.
Meincke
,
S.
Østerhus
,
G.
Rohardt
,
U.
Schauer
,
V.
Tverberg
, and
J.
Verduin
,
2001
:
Direct measurements of volume transports through Fram Strait
.
Polar Res.
,
20
(2
217
224
,
doi:10.1111/j.1751-8369.2001.tb00059.x
.
Ivanov
,
V. V.
, and
Coauthors
,
2009
:
Seasonal oceanic variability off Svalbard in 2004-06
.
Deep-Sea Res. I
,
56
,
1
14
.
Levy
,
J. M.
, Ed.,
2009
:
Global oceans [in “State of the Climate in 2008”]
.
Bull. Amer. Meteor. Soc.
,
90
,
S1
S196
.
Lique
,
C.
, and
M.
Steele
,
2012
:
Where can we find a seasonal cycle of the Atlantic Water temperature within the Arctic basin?
J. Geophys. Res.
,
117
,
C03026
,
doi:10.1029/2011JC007612
.
McLaughlin
,
F. A.
,
E. C.
Carmack
,
W. J.
Williams
,
S.
Zimmermann
,
K.
Shimada
, and
M.
Itoh
,
2009
:
Joint effects of boundary currents and thermohaline intrusions on the warming of Atlantic Water in the Canada Basin, 1993–2007
.
J. Geophys. Res.
,
114
,
C00A12
,
doi:10.1029/2008JC005001
.
Polyakov
,
I. V.
,
D.
Walsh
,
I.
Dmitrenko
,
R. L.
Colony
, and
L. A.
Timokhov
,
2003
:
Arctic Ocean variability derived from historical observations
.
Geophys. Res. Lett.
,
30
,
1298
,
doi:10.1029/2002GL016441
.
Polyakov
,
I. V.
, and
Coauthors
,
2004
:
Variability of the intermediate Atlantic Water of the Arctic Ocean over the last 100 years
.
J. Climate
,
17
,
4485
4497
.
Polyakov
,
I. V.
, and
Coauthors
,
2005
:
One more step toward a warmer Arctic
.
Geophys. Res. Lett.
,
32
,
L17605
,
doi:10.1029/2005GL023740
.
Polyakov
,
I. V.
, and
Coauthors
,
2008
:
Arctic Ocean freshwater changes over the past 100 years and their causes
.
J. Climate
,
21
,
364
384
.
Polyakov
,
I. V.
, and
Coauthors
,
2010
:
Arctic Ocean warming reduces polar ice cap
.
J. Phys. Oceanogr.
,
40
,
2743
2756
.
Polyakov
,
I. V.
, and
Coauthors
,
2011
:
Fate of early-2000s Arctic warm water pulse
.
Bull. Amer. Meteor. Soc.
,
92
,
561
566
.
Polyakov
,
I. V.
,
R.
Kwok
, and
J. E.
Walsh
,
2012
:
Recent changes of arctic multiyear sea-ice coverage and their likely causes
.
Bull. Amer. Meteor. Soc.
,
93
,
145
151
.
Schauer
,
U.
,
E.
Fahrbach
,
S.
Osterhus
, and
G.
Rohardt
,
2004
:
Arctic warming through the Fram Strait: Oceanic heat transport from 3 years of measurements
.
J. Geophys. Res.
,
109
,
C06026
,
doi:10.1029/2003JC001823
.
Schauer
,
U.
,
A.
Beszczynska-Möller
,
W.
Walczowski
,
E.
Fahrbach
,
J.
Piechura
, and
E.
Hansen
,
2008
:
Variation of measured heat flow through the Fram Strait between 1997 and 2006
.
Arctic-Subarctic Ocean Fluxes: Defining the Role of the Northern Seas in Climate
,
R. R. Dickson, J. Meincke, and P. Rhines, Eds., Springer Science, 65–85
.
Trenberth
,
K. E.
, and
J. M.
Caron
,
2001
:
Estimates of meridional atmosphere and ocean heat transport
.
J. Climate
,
14
,
3433
3443
.