Nonseasonal fluctuations of the Arctic Ocean mass observed by the GRACE satellites
ABSTRACT Time variable gravity observations from the GRACE satellites reveal strong non-seasonal fluctuations of bottom pressure in the Arctic Ocean on the time scales from 2 to 6 months and a record-high bottom pressure anomaly in February of 2011. Here, we examine the nature and driving forces behind those fluctuations. We find that the non-seasonal variability of the Arctic Ocean mass is strongly coupled to wind forcing. The zonal wind pattern is correlated with a di-pole pattern of Arctic Ocean mass changes. Westerly wind intensification over the North Atlantic at about 60°N as well as over the Russian Arctic continental shelf break cause the ocean mass to decrease in the Nordic seas and in the central Arctic, and to increase over the Russian Arctic shelf. Basin-wide Arctic Ocean mass fluctuations are correlated with northward wind anomalies over the northeastern North Atlantic and Nordic seas, and over the Bering Sea. We show that positive (negative) Arctic Ocean mass anomalies are associated with anticyclonic (cyclonic) anomalies of the large-scale ocean circulation pattern. Based on ocean model simulations, we conclude that the observed non-seasonal Arctic Ocean mass variability is mostly explained by the net horizontal wind-driven transports, and the contribution of fresh water fluxes is negligible. We demonstrate that transport anomalies across both the Atlantic and Pacific gateways were equally important for generating large Arctic Ocean mass anomalies in 2011.
- Journal of Climate - J CLIMATE. 01/1992; 5(6):541-560.
- [show abstract] [hide abstract]
ABSTRACT: The impact of river run-off on global ocean mass redistribution is analysed by means of simulations with the baroclinic general circulation model OMCT driven by real-time atmospheric forcing fields from the European Centre for Medium Range Weather Forecasts (ECMWF). River run-off data have been deduced from a Hydrological Discharge Model (HDM) forced with ECMWF data as well. While submonthly mass variability is generally insignificant for GRACE de-aliasing purposes in most oceanic regions, monthly mean mass signals of up to 2 hPa occur in the Arctic Ocean during the melt season. Additionally, from freshwater fluxes due to precipitation, evaporation and river run-off the seasonal variations of total ocean mass are calculated. Correspondence with observed mass variations deduced from monthly GRACE gravity solutions indicates that a combination of ECMWF, HDM and OMCT allows a consistent prognostic simulation of mass exchanges among the atmosphere, ocean and continental hydrosphere. Thus, interpretations of GRACE based mass anomalies should account for both regional and global river run-off effects.Geophysical Journal International 01/2007; 168(2):527 - 532. · 2.85 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: The latest release of GRACE (Gravity Recovery and Climate Experiment) gravity field coefficients (Release-05, or RL05) are evaluated for ocean applications. Data have been processed using the current methodology for Release-04 (RL04) coefficients, and have been compared to output from two different ocean models. Results indicate that RL05 data from the three Science Data Centers - the Center for Space Research (CSR), GeoForschungsZentrum (GFZ), and Jet Propulsion Laboratory (JPL) - are more consistent among themselves than the previous RL04 data. Moreover, the variance of residuals with the output of an ocean model is 50-60% lower for RL05 data than for RL04 data. A more optimized destriping algorithm is also tested, which improves the results slightly. By comparing the GRACE maps with two different ocean models, we can better estimate the uncertainty in the RL05 maps. We find the standard error to be about 1 cm (equivalent water thickness) in the low- and mid-latitudes, and between 1.5 and 2 cm in the polar and subpolar oceans, which is comparable to estimated uncertainty for the output from the ocean models.Ocean Science. 10/2012; 8(5):859-868.
Non-seasonal fluctuations of the Arctic Ocean mass observed by the GRACE
Denis L. Volkov1 and Felix W. Landerer2
1Cooperative Institute for Marine and Atmospheric Studies, University of Miami, and Atlantic
Oceanographic and Meteorological Laboratory, NOAA, Miami, Florida.
2Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California.
Corresponding author contact details:
Denis L. Volkov
NOAA / AOML
4301 Rickenbacker Causeway
Miami, FL 33149
Tel: (305) 361-4344
Submitted to Journal of Geophysical Research – Oceans (9 August 2013)
Revised – 3 November 2013
Accepted – 6 November 2013
Time variable gravity observations from the GRACE satellites reveal strong non-seasonal
fluctuations of bottom pressure in the Arctic Ocean on the time scales from 2 to 6 months and a
record-high bottom pressure anomaly in February of 2011. Here, we examine the nature and
driving forces behind those fluctuations. We find that the non-seasonal variability of the Arctic
Ocean mass is strongly coupled to wind forcing. The zonal wind pattern is correlated with a di-
pole pattern of Arctic Ocean mass changes. Westerly wind intensification over the North Atlantic
at about 60°N as well as over the Russian Arctic continental shelf break cause the ocean mass to
decrease in the Nordic seas and in the central Arctic, and to increase over the Russian Arctic
shelf. Basin-wide Arctic Ocean mass fluctuations are correlated with northward wind anomalies
over the northeastern North Atlantic and Nordic seas, and over the Bering Sea. We show that
positive (negative) Arctic Ocean mass anomalies are associated with anticyclonic (cyclonic)
anomalies of the large-scale ocean circulation pattern. Based on ocean model simulations, we
conclude that the observed non-seasonal Arctic Ocean mass variability is mostly explained by
the net horizontal wind-driven transports, and the contribution of fresh water fluxes is negligible.
We demonstrate that transport anomalies across both the Atlantic and Pacific gateways were
equally important for generating large Arctic Ocean mass anomalies in 2011.
Continuous monitoring of the Arctic Ocean sea surface height (SSH) with satellite altimetry
is inhibited by sea ice. Several recent studies used altimetry and hydrographic data to examine
the variability of sea level in the ice-free sub-Arctic regions [e.g. Mork and Skagseth, 2005; Steel
and Ermold, 2007; Volkov and Pujol, 2012; Volkov et al., 2013a, 2013b]. Tide gauges have been
used to study coastal Arctic Ocean sea level [Proshutinsky et al., 2004, 2007; Richter et al.,
2012; Calafat et al., 2013], but only a few gauges have sufficiently long records over the last two
decades and their geographical distribution is uneven.
Non-tidal sea level variability is due to (i) changes in thermohaline properties of seawater
(steric changes) from variations in buoyancy fluxes and heat and salt advection by ocean
currents, and (ii) changes in the mass of a water column, caused by the wind-driven
redistribution of water within the ocean and the exchange of water between the ocean,
atmosphere, and land. Several bottom pressure recorders have been used to study the mass-
related component of sea level variability in the central Arctic [e.g. Morison et al., 2007;
Peralta-Ferriz et al., 2011]. While there is no continuous, basin-wide monitoring of the steric
component of the Arctic Ocean SSH, the Gravity Recovery and Climate Experiment (GRACE,
Tapley et al. ) satellite mission has been providing observations of the monthly ocean
mass (OcM) variations since 2002.
It has been shown that GRACE can reliably observe Arctic OcM changes. Good agreement
between GRACE OcM and in-situ bottom pressure recorders has been found near the North
Pole, in the Beaufort Sea, and in the Fram Strait (just west of Spitsbergen) [Morison et al., 2007;
Peralta-Ferriz and Morrison, 2010; Chambers and Bonin, 2012; Volkov et al., 2013b]. Morison
et al.  found a negative trend from 2002-2006 in GRACE bottom pressures in the central
Arctic that they attributed to upper-ocean freshening. Annual Arctic OcM variations of about 2
cm have been associated with seasonal river runoff and precipitation minus evaporation [Ponte et
al., 2007; Peralta-Ferriz and Morison, 2010]. Using bottom pressure measurements at the North
Pole and in the Beaufort Sea, Peralta-Ferriz et al.  analyzed fast sub-monthly oscillations
of the Arctic OcM and linked them to the meridional winds over the Nordic seas that modulate
the associated geostrophic slope current.
In February 2011, GRACE observed a record-high anomaly of the Arctic OcM with a peak
magnitude exceeding 4 cm above the 2003-2012 average. Here, we analyze the non-seasonal
variability of the Arctic OcM, obtained by subtracting the monthly mean climatology from
monthly GRACE measurements. Thus, our study is mostly focused on GRACE signals with
periods from 2 to about 6 months that dominate the non-seasonal variability of the Arctic OcM.
We investigate the driving mechanisms for the observed non-seasonal fluctuations. In particular,
we evaluate the relative contributions of wind forcing and fresh water fluxes and identify the
sources that contributed to the record-high Arctic OcM anomaly in February 2011.
A recent study by Landerer and Volkov  explored similar non-seasonal OcM
fluctuations in the Mediterranean Sea and attributed them to concurrent wind stress anomalies in
the subtropical North Atlantic. The Arctic Ocean has a number of oceanographic features
reminiscent of the Mediterranean Sea. Albeit larger, the Artic Ocean is also a semi-enclosed
basin with a narrow and shallow gateway to the Pacific Ocean at the Bering Strait, and with a far
less restricted exchange with the Atlantic Ocean through the Fram Strait and the Barents Sea.
Therefore, we hypothesize that forcing mechanisms driving the non-seasonal OcM fluctuations
in the Arctic Ocean are possibly similar to the Mediterranean.
2. Data and Methods
2.1 Ocean mass from GRACE
The GRACE twin-satellites have observed time-variations of the Earth’s gravity field since
May 2002 and provide unique measurements of large-scale ocean mass changes. We use the
GRACE Release-05 monthly bottom pressure anomalies based on spherical harmonics from the
Geoforschungzentrum Potsdam (GFZ RL 5.0), which we refer to as OcM. The GRACE data are
mapped on a 1°×1° grid and distributed via GRACE Tellus web resource
(http://grace.jpl.nasa.gov). Details on data processing can be found in Chambers and Bonin
. While the results presented in this manuscript are based on GFZ RL 5.0 product, similar
results are obtained with the University of Texas Center for Space Research (CSR RL 5.0)
The Arctic OcM data are used over the geographical domain bounded by 65°N in the south,
so that the data include the Arctic Ocean and the adjacent basins of the North Atlantic (Baffin
Bay, Norwegian and Greenland seas). Using the Wahr et al.  approach, which essentially
assumes that short-period signals are noise, we obtain a 1-sigma error estimate of 8-9 mm for
one monthly GRACE OcM value. This is an upper limit for the error, especially in the Arctic,
where as will be shown below, much of the non-seasonal signal is real and not noise.
On monthly and longer time scales, local OcM changes as observed with GRACE are not
affected by atmospheric pressure due to the compensating inverted barometer effect on sea level.
However, OcM needs to be corrected for the average atmospheric pressure over the global ocean,
as water is incompressible. Therefore, we subtract the global mean OcM, which in addition to the
global atmospheric pressure effect also removes any net ocean mass change from net global
freshwater fluxes. Furthermore, we subtract a monthly mean climatology and a linear trend to
focus on the non-seasonal time scales. From these OcM anomalies, we calculate horizontal
geostrophic velocity anomalies to analyze associated changes in the large-scale ocean
circulation. Note that due to recent battery management issues since 2011, the GRACE
instruments are periodically turned off for periods of up to 4 weeks at approximately 6-month
intervals. For the following comparisons of OcM to wind stress, we linearly interpolate the
missing GRACE months. This means that rapid OcM fluctuations that may occur during missing
months may not be recovered properly.
2.2. Atmospheric data
The observed OcM variability is coupled to monthly mean surface wind stress data obtained
from the ERA-Interim re-analysis [www.ecmwf.net, Dee et al., 2011].?? We use wind stress data
over the circumpolar domain bounded by 50°N in the south. This is larger than the domain used
for OcM data, because the Arctic OcM can be correlated with wind forcing beyond the Arctic
Ocean boundary. To focus on the non-seasonal fluctuations, the monthly mean climatology was
subtracted, as was done for the OcM data. In addition, we use the monthly Arctic Oscillation
(AO) index distributed by NOAA’s National Weather Service Climate Prediction Center
2.3. ECCO2 ocean data synthesis
To compare with and analyze the observed non-seasonal OcM variability, we use a high-
resolution (18 km grid spacing) ECCO2 (Estimating the Circulation and Climate of the Ocean,
Phase II) ocean data synthesis product (www.ecco2.org), which is obtained by a least-squares fit
of a global full-depth-ocean and sea-ice configuration of the Massachusetts Institute of
Technology general circulation model (MITgcm) [Marshall et al., 1997] to selected satellite and
in-situ data. The MITgcm includes a dynamic/thermodynamic sea ice model [Losch et al., 2010].
The simulations are forced by the Japanese 25-year Re-Analysis (JRA-25)
[http://jra.kishou.go.jp/JRA-25/, Onogi et al., 2007] and constrained by observations using the
model Green’s function to adjust a series of empirical control parameters [Menemenlis et al.,
2005]. Observational constraints include satellite altimetry SSH, sea surface temperature, vertical
temperature and salinity profiles, and sea ice concentrations, motion, and thickness. The control
parameters include initial hydrography, atmospheric boundary conditions, and background
vertical diffusivity. The ECCO2 solutions are computed on a cube-sphere grid with 18-km
horizontal grid spacing and 50 vertical levels with thicknesses ranging from 10 m at the surface
to 456 m near the bottom.
We use the monthly averages of the ECCO2-simulated sea level, bottom pressure, and wind
stress from January 2003 to December 2012 and 3-day averages of bottom pressure, horizontal
velocities, and fresh water fluxes over the 2010-2011 (2 years) time interval. The 3-day averages
are used for the comparison of the time derivative of OcM with net lateral transports and fresh
water fluxes into the study domain. From the monthly fields we remove the monthly mean
climatology and a linear trend. From the 3-day fields we subtract the annual and semi-annual
signals, computed by fitting harmonic functions with corresponding frequencies in a least-
squares sense. The net lateral transports are computed from horizontal velocities across the 65°N
circumpolar section and across its segments in the North Atlantic and Bering Strait.
2.4. Identification of coupled fields
To evaluate the temporally co-varying, but not necessarily spatially co-located signals in
OcM and wind stress fields, we use a coupled Empirical Orthogonal Functions (EOF) analysis
[e.g., Bretherton et al., 1992; von Storch and Zwiers, 1999]. The coupled EOF decomposition of
two variables extracts only those spatial patterns that explain most of the covariance between the
two fields. The temporal evolution of these patterns is demonstrated by the principal component
(PC) time series. Here, we consider only the leading coupled modes of variability and present the
spatial patterns in the form of heterogeneous correlation maps (i.e., correlation between the PC
time series of wind stress and OcM field and vice versa). The 2003-2012 time series of OcM and
wind stress have 120 independent measurements. Accounting for the autocorrelation of the time
series, the number of effective degrees of freedom is about 60. This means that correlation
coefficients above 0.25 are significant at 95% confidence level.
3.1. Arctic Ocean mass changes from GRACE and ECCO2
The analysis of the ECCO2-simulated SSH and OcM indicates that the non-seasonal SSH
variability in the Arctic Ocean is mostly mass related (Figure 1). The non-seasonal SSH variance
explained by the non-seasonal OcM reaches 80-100% over the Russian shelf seas, the Canadian
Basin, and in the interior of the Nordic seas. The contribution of the non-seasonal OcM is
smaller in the advective regions, where steric effects are large, in particular along the Norwegian
and East Greenland Currents, Transpolar Drift, and over the Siberian and Alaskan continental
shelf breaks. The Transpolar drift is supported by a front between the Pacific-derived and
Atlantic-derived upper-ocean water that swings across the Makarov Basin depending on the
Arctic Oscillation (AO) phase and is characterized by a large standard deviation of salinity
[Morison et al., 2000].
The area-averaged non-seasonal OcM from GRACE exhibits variability with a standard
deviation of about 1 cm (Figure 2). This value is greater than the annual OcM amplitude (about
0.7 cm), estimated by a harmonic analysis. There is a good agreement between the GRACE-
observed and ECCO2-simulated non-seasonal OcM in the Arctic Ocean with correlation of 0.73
(Figure 2). Most short-term OcM fluctuations are well reproduced by the model. What is
important for the objectives of this study is that the observed high anomaly in February 2011 is
well simulated by ECCO2. The timing and the amplitudes of the observed and simulated
anomalies are almost identical, indicating that ECCO2 can be used to investigate the mechanisms
responsible for the observed non-seasonal OcM fluctuations.
The largest discrepancy between GRACE and ECCO2 OcM was in November 2011. This
discrepancy can be explained by a partial GRACE instrument outage between November 17 and
December 12. In addition, the November 2011 GRACE monthly solution is actually derived
from observations between October 16 and November 16. Introducing the gap and resampling
the 3-day ECCO2 OcM time series at the GRACE epoch reduces the apparent misfit between the
model and observations for this particular event by about 2 cm (not shown). Therefore, the
November 2011 peak simulated by ECCO2 is likely to be a real OcM anomaly.
3.2. Ocean mass changes in relation to wind forcing
The coupled EOFs between the non-seasonal OcM and the zonal (τx) and meridional (τy)
wind stress components establish the statistical relationship between the Arctic OcM and wind
stress. The first coupled τx and OcM EOF mode (Figure 3) explains 56% of the covariance and
the correlation between the corresponding PC time series is 0.57 (Figure 3c). The first coupled τy
and OcM EOF mode (Figure 4) explains 73% of the covariance and the correlation between the
corresponding PC time series is 0.68 (Figure 4c).
The coupled OcM/τx mode reveals a di-pole oscillation pattern of the Arctic OcM (Figure
3a), with one center located in the Arctic interior and the other over the Russian continental
shelf. The absolute correlation between the PC-1 of τx and OcM in these areas exceeds 0.5. The
corresponding wind stress pattern (Figure 3b) has two maxima, located over the northeastern
North Atlantic and over the Kara, Laptev, and East-Siberian seas with absolute correlations
above 0.4. The time evolution of these spatial patterns (Figure 3c) indicates that when the zonal
wind between about 70°N–85°N and 60°E– 210°E intensifies/reduces, OcM rises/falls over the
Russian shelf seas and falls/rises in the central Arctic, which is consistent with Ekman dynamics.
An intensification/reduction of westerly winds around 60°N in the North Atlantic also leads to an
increased/decreased southward Ekman transport that is correlated with a decrease/increase of
OcM in the Nordic seas and in the central Arctic. The di-pole oscillation pattern in Figure 3a
corresponds to the second EOF of bottom pressure from Peralta-Ferriz  and Peralta-
Ferriz et al. [accepted for publication in J. Climate, 2013], forced by changes in atmospheric
pressure over the central Arctic and east Greenland that drive zonal winds. The zonal wind
anomalies over the Arctic Ocean are related to the Arctic Oscillation (AO) (Figure 3c). The
correlation coefficient between the PC-1 of τx and the AO index is 0.52, while the correlation
between the PC-1 of OcM and the AO index is 0.28. The positive/negative AO index
corresponds to lower/higher than average sea level pressure at the North Pole and, in turn, to
stronger/weaker westerly winds. The low AO index in 2010 was associated with an OcM
decrease over the Russian shelf seas and an OcM increase over the central Arctic.
The coupled OcM/τy mode reveals a basin-wide coherent pattern of OcM changes, with the
exception of the shallow areas of the Kara, Laptev, and East Siberian seas (Figure 4a). The
correlation between the PC-1 of τy and OcM over most of the Arctic Ocean is well above 0.5.
The PC-1 time series (Figure 4c) demonstrate that this is the mode that describes the basin-
averaged fluctuations of OcM shown in Figure 2. The?? correlation?? between?? the?? PC-‐1?? of?? OcM?? 205??
and?? τy?? is?? 0.96,?? while?? the?? correlation?? between?? the?? PC-‐1?? of?? OcM?? and?? the?? basin-‐averaged?? OcM?? 206??
is?? 0.65.?? The record high OcM value observed in February 2011 (Figure 2) is well represented by
the coupled PC-1 of OcM and τy. The basin-coherent mode corresponds to the first EOF of
bottom pressure forced by an atmospheric pressure dipole that straddles the Nordic seas from
Peralta-Ferriz  and Peralta-Ferriz et al. [accepted for publication in J. Climate, 2013].
We note that the basin-coherent OcM fluctuations are correlated with the meridional wind over
the northeastern North Atlantic, the Nordic seas, and over the Bering Sea (Figure 4b). Thus, an
intensification/reduction of the meridional wind over these areas is associated with a basin-wide
increase/decrease of the Arctic OcM. This result complements and extends the established
relationship between the sub-monthly fluctuations of the Arctic OcM measured by bottom
pressure gauges and winds over the Nordic seas [Peralta-Ferriz et al., 2011]. Unlike the zonal
wind, the first coupled OcM and τy EOF modes are not correlated with the AO index (r=0.004
and r=−0.15, respectively).
3.3. Atmospheric and oceanic circulation patterns
To reveal the characteristic atmospheric and oceanic circulation patterns associated with the
observed large OcM anomalies, we plot the 10-m wind speed anomaly, wind stress curl anomaly,
and oceanic geostrophic velocity anomaly vectors averaged over the periods when the basin-
mean OcM anomalies (in Figure 2) are larger than ±1 cm. The southward/northward wind
anomalies over the Nordic seas and the Bering Sea occur during the low/high OcM in the Arctic
(Figure 5). The associated basin-wide oceanic circulation pattern during the low/high Arctic
OcM is cyclonic/anticyclonic with intensified/reduced Fram Strait outflow from the Arctic
Ocean, and stronger/weaker Norwegian Current and Barents Sea throughflow (Figure 6).
The GRACE-derived geostrophic velocity anomalies are very small, on the order of 0.1 cm/s,
but simple estimates indicate that these anomalies are realistic. For example, velocity anomalies
across the Fram Strait are about 0.1 cm/s. For an average depth of 1000 m and width of 600 km
this yields a volume transport anomaly of about 0.6 Sv, which is close to in situ observations
[e.g., Schauer et al., 2004]. At Bering Strait, when it is ice free, the along-strait wind is likely to
modulate the inflow into the Arctic Ocean, similar to how wind affects transport across the Strait
of Gibraltar and hence the Mediterranean sea level [Fukumori et al., 2007; Landerer and Volkov,
The southward/northward wind anomalies over the Nordic seas during the low/high Arctic
OcM anomalies (Figure 5) favor the westward/eastward Ekman transport anomalies and, hence,
southward/northward anomalies of the Ekman slope current across the region and through the
Fram Strait. The observed strengthening/weakening of the East-Greenland Current and to a
lesser extent the weakening/strengthening of the West-Spitsbergen Current during the low/high
Arctic OcM anomalies is consistent with Ekman dynamics (Figure 6). Morison  showed
that the seasonal variation of the West Spitsbergen Current at one location in 1982-1985 was
largely due to Sverdrup transport. However, the relevance of Sverdrup dynamics in the Nordic
seas non-seasonal transport is questionable, because the meridional gradient of planetary
vorticity (beta) is weak at high latitudes and topographic effects are particularly important [Nøst
and Isachsen, 2003]. The results of our study demonstrate that during the low/high Arctic OcM
anomalies the wind stress curl anomalies are positive/negative over most areas of the Nordic seas
and over the Fram Strait, while they are negative/positive along the East Greenland shelf and
over the Denmark Strait (Figure 5, color). At the same time, there are strong
southward/northward transport anomalies in the Fram Strait and in the Greenland Sea (Figure 6).
This indicates that Sverdrup dynamics in this area do not explain the corresponding fluctuations
of the Arctic OcM.
On the other hand, the low/high Arctic OcM anomalies are associated with negative/positive
wind stress curl anomalies over the North Atlantic, just south of 65°N (Figure 5, color). By
virtue of Sverdrup dynamics, the negative/positive wind stress curl drives the
southward/northward mass transport of wind driven currents. Therefore, during the low/high
Arctic OcM anomalies there should be wind driven southward/northward transport anomalies
over this part of the North Atlantic that hypothetically can lead to outflow/inflow anomalies
from/to the Nordic seas. The likelihood and efficacy of this mechanism needs to be explored in a
3.4. Where does the water come from?
We have presented evidence that the non-seasonal variations of the Arctic OcM are related to
wind forcing over the basin itself (zonal wind) and over the adjacent regions (meridional wind).
Similar to the Mediterranean Sea, winds can force water in or out of the basin and change the
OcM [Fukumori et al., 2007; Landerer and Volkov, 2013]. However, it remains unclear how
strong the contribution of wind forcing is, and whether fresh water fluxes contribute significantly
as they do for the seasonal variations [Ponte et al., 2007; Dobslaw and Thomas, 2007; Perralta-
Ferriz and Morison, 2010]. What, in particular, caused the large anomalies in February of 2011
and in November 2011?
The time change of the basin-averaged OcM is compensated by the lateral volume transport
and fresh water fluxes (precipitation, evaporation, and river runoff) (Figure 7a). The rate of OcM
change due to fresh water fluxes (red curve) is mostly positive. Any pressure buildup in the
Arctic Ocean due to fresh water input would quickly propagate away to the rest of the World
Ocean. Although wind forcing may act to retain fresh water in the basin by reducing the
compensating outflow, as occurs at the seasonal time scale [Dobslaw and Thomas, 2007;
Perralta –Ferriz and Morison, 2010], the non-seasonal OcM variability is not significantly
affected by fresh water retained in the basin. The variability of the basin-averaged non-seasonal
OcM is mostly determined by the net transport across 65°N (Figure 7a, blue curve).
The observed record high anomaly at the beginning of 2011 is the result of the net transport
anomalies across 65°N. There were several relatively prolonged events with positive transport
anomalies into the Arctic in winter 2010/2011 (Figure 7a). Using the model, we separate the net
transport across 65°N in the Atlantic Ocean and through the Bering Strait to estimate their
relative contributions (Figure 7b). In 2010-2011, the time mean transport across 65°N in the
Atlantic Ocean is -0.9 Sv (southward), while the time mean transport through the Bering Strait is
0.7 Sv (into the Arctic Ocean). The remaining input of 0.2 Sv into the Arctic is due to river
runoff and precipitation minus evaporation. These estimates are very close to observations [e.g.
Beszczynska-Möller et al., 2011; Serreze et al., 2006]. The Atlantic sector and Bering Strait
transports partly compensate for each other and the correlation between them is -0.49. The
variability of the total transport is mostly determined by the variability of the Atlantic sector
transport (Figure 7b). The correlation between the total transport and the Atlantic sector transport
is 0.77, while the correlation between the total transport and the Bering Strait transport is not
To investigate the relative contributions to the record-high OcM anomalies in February and
November 2011, we compute the de-trended cumulative sums of the net Atlantic sector and
Bering Strait transports, scale them by the Arctic Ocean area and multiply by the model time
interval (3 days) to obtain the equivalent OcM time series (Figure 8, note that the tick marks of
x-axis correspond to mid-month dates). The OcM anomaly in February 2011 was mostly due to a
positive anomaly of the Atlantic sector transport. At the same time, the Atlantic transport
anomaly was not compensated by a sizable reduction in the Bering Strait inflow. On the
contrary, the Bering Strait inflow increased about two weeks later, leading to a further Arctic
OcM increase. In fact, the Arctic OcM reached its maximum value at the end of February with
approximately equal contributions from Atlantic and Pacific inflows (Figure 8). The OcM
anomaly in November 2011 was initiated by a stronger than average inflow through the Bering
Strait in the second half of October followed by an inflow anomaly from the Atlantic Ocean.
Thus, although the variability of the Arctic OcM is mostly due to the Atlantic sector transport
anomalies, both the Atlantic and Pacific gateways can be equally important for generating large
Arctic OcM anomalies.
The ECCO2 atmospheric circulation patterns associated with February and November 2011
OcM anomalies (Figure 9) are characterized by northward anomalies over the Nordic and
Barents seas and positive wind stress curl anomalies in the North Atlantic south of 65°N (similar
to Figure 5b). Unlike the Fram Strait, where sea ice is relatively free to move, near the narrow
and shallow Bering Strait, winter sea ice is constrained by interaction with land and bottom and
can attenuate the transfer of momentum to the water. However, strong northward wind anomalies
over the Bering Sea were able to force water into the Arctic Ocean in February 2011, when the
area with the concentration of sea ice over 50% extended south to about 60°N (blue curve in
Figure 9, left). The Bering Strait throughflow increased in the second half of February and
remained high through March, but in March it was already compensated by an outflow through
the Atlantic sector (Figure 8). Because of the possible effect of sea ice in the northern part of the
Bering Sea, the mechanism here is therefore different from direct forcing by the along-strait
wind as occurs in the Mediterranean Sea [Fukumori et al., 2007; Landerer and Volkov, 2013].
We suggest that a likely mechanism is related to Ekman dynamics: the northward wind