8170 JOURNAL OF VOLUME 27

On the Link between Arctic Decline and the Freshwater Content of the Beaufort Gyre: Insights from a Simple Process Model

PETER E. D. DAVIS,CAMILLE LIQUE, AND HELEN L. JOHNSON Department of Earth Sciences, University of Oxford, Oxford, United Kingdom

(Manuscript received 30 January 2014, in final form 1 August 2014)

ABSTRACT

Recent satellite and hydrographic observations have shown that the rate of freshwater accumulation in the Beaufort Gyre of the has accelerated over the past decade. This acceleration has coincided with the dramatic decline observed in Arctic sea ice cover, which is expected to modify the efficiency of momentum transfer into the upper ocean. Here, a simple process model is used to investigate the dynamical response of the Beaufort Gyre to the changing efficiency of momentum transfer, and its link with the enhanced accu- mulation of freshwater. A linear relationship is found between the annual mean momentum flux and the amount of freshwater accumulated in the Beaufort Gyre. In the model, both the response time scale and the total quantity of freshwater accumulated are determined by a balance between Ekman pumping and an - induced volume flux toward the boundary, highlighting the importance of eddies in the adjustment of the Arctic Ocean to a change in forcing. When the seasonal cycle in the efficiency of momentum transfer is modified (but the annual mean momentum flux is held constant), it has no effect on the accumulation of freshwater, although it does impact the timing and amplitude of the annual cycle in Beaufort Gyre freshwater content. This suggests that the decline in Arctic sea ice cover may have an impact on the magnitude and seasonality of the freshwater export into the North Atlantic.

1. Introduction et al. 1988; Aagaard and Carmack 1989; Vellinga and Wood 2002; Jones and Anderson 2008). The Arctic Ocean receives an input of freshwater from Within the Arctic Ocean itself, more than 70 000 km3 Eurasian and North American river runoff, net pre- of freshwater is stored within a very fresh surface layer cipitation over evaporation (positive P 2 E), and sea ice (Aagaard and Carmack 1989; Serreze et al. 2006; Fig. 1), melt. The Atlantic contributes relatively salty water via which is separated by a strong halocline from the rela- Fram Strait and the Barents Sea, and the Pacific relatively tively warm and saline Atlantic-derived layer beneath freshwater via Bering Strait. After significant watermass (Aagaard et al. 1981). Over the past few decades, ob- transformation, freshwater is exported in both liquid and servations have shown that the freshwater content has solid (ice) form into the Nordic and Labrador Seas, been increasing (Proshutinsky et al. 2009; McPhee et al. through Fram Strait and the Canadian Arctic Archipelago, 2009; Rabe et al. 2011; Krishfield et al. 2014), with the respectively.(Coachman and Aagaard 1974; Aagaard and largest freshening observed in the Beaufort Gyre in the Carmack 1989; Carmack 2000; Serreze et al. 2006; Dickson Canada Basin (Fig. 1). Recently, Giles et al. (2012) have et al. 2007). A moderate freshening of the surface waters in suggested that the rate of freshening in the Beaufort these exit regions may impact the formation of North Gyre may be accelerating, and have proposed that Atlantic Deep Water, and thus affect both the global a dynamical adjustment in response to the decline in and the global climate (Dickson Arctic sea ice cover may be the cause. Understanding this dynamical response is the focus of this study. The Beaufort Gyre is a permanent anticyclonic circu- lation driven by the winds associated with the atmo- Corresponding author address: Peter Davis, Department of Earth Sciences, University of Oxford, South Parks Road, Oxford, OX1 spheric Beaufort high (Proshutinsky et al. 2009). These 3AN, United Kingdom. winds cause water to converge in the center of the gyre, E-mail: [email protected] and the resulting downwelling (Ekman pumping) leads to

DOI: 10.1175/JCLI-D-14-00090.1

Ó 2014 American Meteorological Society Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 1NOVEMBER 2014 D A V I S E T A L . 8171

the Beaufort Gyre has gained 8400 6 2000 km3 of freshwater between the periods 1992–99 and 2006–08, and Giles et al. (2012) used satellite observations of sea surface height between 1995 and 2012 to infer a similar increase of 8000 6 2000 km3. The results of Giles et al. (2012) show a clear spatial correlation between the positive trend in the sea surface height associated with the Beaufort Gyre, and the nega- tive trend in the curl of the wind field (i.e., a trend toward more anticyclonic winds). However, despite this clear spatial correlation, the temporal correlation is less clear. While the trend in the curl of the wind field remained constant over the whole period, the trend in the sea sur- face height changed sign after 2002. Consequently, Giles et al. (2012) have suggested that since 2002 the winds may FIG. 1. Climatology of Arctic freshwater content integrated have become more effective at driving an accumulation vertically from the surface to the depth of the 34.8 isohaline from of freshwater within the Beaufort Gyre, because of an the PHC (Steele et al. 2001). The Beaufort Gyre region as defined by Proshutinsky et al. (2009) and Giles et al. (2012) is highlighted increase in the efficiency of momentum transfer into the by the black box, and clearly stands out as the region of greatest upper ocean. On the other hand, Morison et al. (2012) freshwater content in the Arctic. argue that a shift in the advection pathways of Eurasian river water, associated with a positive phase of the Arctic an accumulation of freshwater through the mechanical Oscillation during 2005–08, can account for the increased deformation of the salinity field (Proshutinsky et al. 2002; accumulation of freshwater, with no role played by the Yang 2009; Proshutinsky et al. 2009). On annual time Beaufort Gyre wind-driven circulation. In reality, it is scales, stronger anticyclonic winds in winter result in an likely that both processes are important (Mauritzen accumulation of freshwater, whereas weaker (or possibly 2012). cyclonic) winds in summer relax the salinity gradient re- In the absence of sea ice, the magnitude of the mo- sulting in a release of freshwater (Proshutinsky et al. mentum transfer into the upper ocean (i.e., the ocean 2002). In addition, ice growth in winter and melt in surface stress) is determined simply by the surface wind summer will decrease or increase the freshwater content stress. However, in the partly ice-covered Arctic, the of the Beaufort Gyre respectively, and any changes in the magnitude of the momentum transfer is determined by advection of surface Pacific and Atlantic waters are also both the surface (tAir-Water) and an ice- likely to have an impact (Proshutinsky et al. 2009). water stress component (tIce-Water), with their relative Consequently, the mean annual cycle in freshwater con- magnitudes scaled by the ice concentration (a; Yang tent reflects the complex interplay among several differ- 2009): ent forcings. t 5 at 1 2 a t On interannual time scales, Proshutinsky and Johnson OceanSurface Ice-Water (1 ) Air-Water . (1) (1997) and Proshutinsky et al. (2002) have suggested that two different wind regimes exist within the Arctic: In the past, the thick and extensive Arctic sea ice cover cyclonic and anticyclonic. During the anticyclonic re- has acted to reduce the momentum transfer, due to the gime, freshwater is accumulated within the Beaufort large internal ice stresses reducing the ice-water stress Gyre over several years because of a strengthened at- component, and shielding of the ocean from direct wind mospheric Beaufort high. In contrast, during the cy- forcing. However, as the sea ice cover has begun to break clonic regime the atmospheric Beaufort high weakens, up and retreat farther and for longer each year (Stroeve and freshwater is released to the shelves where it may be et al. 2007; Maslanik et al. 2007; Cavalieri and Parkinson exported into the North Atlantic. As a result, there is 2012), it has become both weaker and thinner (Kwok and a strong linear relationship between the freshwater Rothrock 2009; Zhang et al. 2012; Laxon et al. 2013), and content of the Beaufort Gyre and the wind stress curl on the number of leads, melt ponds, and ice floe edges has interannual time scales (Proshutinsky et al. 2002, 2009). increased, changing the shape of the ice pack (Flocco Since 1997, the Arctic has been in the longest anticy- et al. 2012; Tsamados et al. 2014). Consequently, the ef- clonic regime on record, leading to a large excess accu- ficiency of momentum transfer into the upper ocean has mulation of freshwater within the Beaufort Gyre. Using increased, as a thinner and weaker sea ice cover is more hydrographic data, Rabe et al. (2011) have shown that easily forced by the winds, and the changing shape of the

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 8172 JOURNAL OF CLIMATE VOLUME 27 ice pack is providing more and more near-vertical faces investigate the dynamical response of the Beaufort Gyre for the wind to push against (Andreas et al. 2010). As to the changing efficiency of momentum transfer, and its a result, not only is the annual mean ocean surface stress link to the accelerated accumulation of freshwater, by

(tOceanSurface) increasing (i.e., the net forcing), but its idealistically perturbing the annual cycle in ocean sur- seasonality is also changing. For example, Martin et al. face stress through a wide parameter space. We do not (2014) have shown using the Pan-Arctic Ice–Ocean explicitly represent sea ice in the model, but rather ac- Modeling and Assimilation System (PIOMAS) model count for the effect of its decline via perturbations made that the Arctic-wide annual mean ocean surface stress has to the magnitude and seasonality of the ocean surface 2 2 increased by 0.006 N m 2 decade 1 between 2000 and stress. In section 2 we describe the model setup, before 2012, corresponding to an increase of approximately presenting the results of a control run in section 3.In 2 9% decade 1 based on the long-term mean stress of sections 4 and 5 the setup and results of two different 2 0.064 N m 2. Furthermore, Tsamados et al. (2014) have sets of idealized experiments designed to investigate the shown using the Los Alamos sea ice model (CICE) and response of the Beaufort Gyre to the decline in Arctic a new parameterization for form drag, that the drag co- sea ice cover are presented, with a discussion of the re- efficient (which can be taken as a measure of the effi- sults and their implications in section 6. Our conclusions ciency of momentum transfer into the upper ocean) are presented in section 7. exhibits a small positive trend over the Beaufort Gyre in summer between 1990 and 2012. 2. Model setup The accelerated accumulation of freshwater in the Beaufort Gyre has coincided with the dramatic decline A nonlinear 1.5-layer reduced gravity model (e.g., observed in Arctic sea ice cover. Has the corresponding Johnson and Marshall 2002; Allison et al. 2011) is used increase in the annual mean ocean surface stress, or its to simulate the surface halocline layer of the Canada changing seasonality, contributed to the accelerated Basin in the Arctic Ocean. The model is governed by the accumulation and thus modified the linear relationship following equations: observed between the freshwater content and the wind ›u t 1 (f 1 j)k 3 u 1 $B 5 A=2u 1 , and (2) stress curl (Proshutinsky et al. 2002, 2009)? The chang- ›t r h ing seasonality may have important implications for the 0 adjustment of the Arctic Ocean to a change in forcing. If ›h 1 $ (hu) 5 $ (k$h), (3) the dominant adjustment time scale is on the order of ›t a season, modifications to the length of each season, or an asymmetry between spinup and spindown time scales where u is the velocity, t is time, f is the Coriolis pa- over the annual cycle, may result in a multiyear trend in rameter equal to 2V, where V is the angular velocity of freshwater accumulation. the earth (i.e., the model is an f plane centered on the The dynamical response of the Beaufort Gyre to the pole), j 5 (›y/›x) 2 (›u/›y) is the relative vorticity, B 5 thinning, weakening, and changing shape of the Arctic g0h 1 (u2 1 y2)/2 is the Bernoulli potential, g0 is the re- 2 sea ice cover will depend upon exactly how much more duced gravity of 0.02 m s 2, A is the lateral friction co- 2 stress is transferred from the surface of the ice pack to efficient of 150 m2 s 1, t is the ocean surface stress [i.e., the ocean below (i.e., the change to the annual mean tOceanSurface in Eq. (1)], r0 is the average density of 2 ocean surface stress) and how the seasonal distribution 1026 kg m 3, h is the surface layer thickness, $ (k$h)is of this transfer may change (i.e., the change to the sea- an advective term arising from the Gent–McWilliams sonality in the ocean surface stress). However, both re- eddy parameterization (Gent 2011), and k is the eddy main poorly constrained and understood, and many diffusivity. The equations are discretized on a C grid different processes such as stratification, atmospheric with a resolution of 15 km 3 15 km. boundary layer stability, ocean circulation, and sea ice Following the idealized approach of Spall (2003, 2004, conditions are important in determining the transfer 2013), the Canada Basin is represented as a circular do- of momentum through sea ice (McPhee 2012). Indeed, main approximately 1500 km in diameter, connected to two state-of-the-art studies by Martin et al. (2014) and a sponge region by a channel approximately 150 km wide Tsamados et al. (2014), which were designed to investigate (Fig. 2). The sponge region is constantly restored to the how the magnitude and seasonality of the ocean surface initial layer thickness over the period of a day, and is stress have changed with the recent decline in Arctic sea designed to absorb any waves propagating out of the ice cover, show seasonal cycles in ocean surface stress domain, as well as allow an inflow or outflow transport to that are 1808 out of phase. Here we adopt the simplest develop through the channel in response to changes in possible approach and use an idealized process model to layer thickness inside the domain. In the vertical, the

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 1NOVEMBER 2014 D A V I S E T A L . 8173

FIG. 2. Reduced gravity model of the Canada Basin with an active surface halocline layer of density r 2Dr initialized to be 400 m thick, overlying a stationary (u 5 0) Atlantic layer of density r. The model domain is ;1500 km in diameter and is connected to a sponge region by a channel ;150 km wide. The sponge region is designed to absorb any waves propagating out of the domain, and to allow an inflow or outflow transport to develop in response to changes in layer thickness within the domain. The model is forced with an anticyclonic ocean surface stress centered over the domain. The magnitude of the curl of the stress field (which is proportional to the strength of the Ekman pumping) is at a maximum in the center of the domain, and is zero at the boundaries and in the outflow. model consists of an active halocline layer overlying a sta- where x and y are the distances to each grid point from tionary Atlantic layer. This approach is similar to that of the center of the domain along the x and y axes, re- Proshutinsky and Johnson (1997), who used a barotropic spectively, and C is a scale factor to ensure that the curl model to investigate the wind-driven dynamics of the upper of the stress field is continuous at the domain/channel layer of the Arctic Ocean. The halocline layer is initialized boundary. The stress fields are normalized between with a thickness of 400 m, matching the average depth of 0 and 1, and during each model run are multiplied by an the 34.8 isohaline outside of the Beaufort Gyre in the Polar idealized annual cycle in ocean surface stress to set the Science Center Hydrographic Climatology (PHC; Steele magnitude of the forcing. This stress formulation en- 2 et al. 2001). The reduced gravity of 0.02 m s 2 is based on sures that the curl of the ocean stress field reaches an average salinity of 32.3 in the halocline layer and 35.0 in a maximum in the center of the domain, and decreases the Atlantic layer from the PHC, resulting in an internal to zero at the boundaries and in the outflow region Rossby deformation radius of approximately 20 km. (Fig. 2). The model is forced with an anticyclonic ocean sur- To balance the input of vorticity from the winds (thus face stress [tOceanSurface in Eq. (1)] centered over the allowing the model to reach a steady state, or to spin domain (Fig. 2). Along any diameter within the domain, down if the forcing is reduced to zero), the effect of the ocean surface stress in the x and y direction is de- eddies and diapycnal mixing have been incorporated scribed by ð into the model using the Gent–McWilliams parameter- 1 ization (e.g., Gent 2011). Following the approach of t(x) 5 sinu r cos2rdr and (4) r Allison et al. (2011), the magnitude of the eddy diffu- ð sivity k can be determined by considering the total 1 t(y) 52cosu r cos2rdr , (5) transport T across any closed layer thickness contour r around the domain. driven by the an- ticyclonic ocean surface stress will cause water to accu- respectively, where u is the angle that lines connecting mulate in the center of the domain, steepening the each grid point with the center of the domain make with pressure gradient and driving an anticyclonic geo- the positive x axis, and r is the radial distance between strophic current. Baroclinic instability associated with 2 each grid point and the center of the domain (i.e., r 5 this current will result in an eddy-induced bolus trans- 2 2 x 1 y ). Within the channel and sponge region, the port toward the boundary of the domain. Consequently, stresses are described by the total transport across any closed thickness contour is the residual between the Ekman and eddy-induced (x) y t 5 C and (6) T 5 T 2 T ) and can be ob- r2 transport velocities ( Ek Eddies tained by integrating the cross-contour components of 2x the eddy (hy ) and Ekman (hy ) transport veloci- t(y) 5 C , (7) Eddies Ek r2 ties along the entire thickness contour:

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 8174 JOURNAL OF CLIMATE VOLUME 27 þ þ þ t(l) ›h than by the sea ice concentration, and therefore the an- T ’ h y 1 y dl ’2 dl 2 k dl ( Ek Eddies) r › , (8) nual cycle in ocean surface stress used to force the control 0 f r run reflects only the changing wind conditions, with 8 l where l is the distance in the azimuthal direction, t( ) is months of strong anticyclonic winds or high stress 22 the ocean surface stress aligned parallel to the thickness (0.023 75 N m ) followed by 4 months of weak anticy- 22 contour, and r is the radial direction; it has been assumed clonic winds or low stress (0.0125 N m ; Fig. 3c). Be- that the thickness contours form concentric circles inside cause of the difficulties in estimating the ocean surface the domain. Integrating Eq. (8) between the boundary stress directly from wind speed in the ice-covered Arctic, and the center of the domain and dividing by the radius the value of the ocean surface stress is taken from of the domain gives the Nucleus for European Modelling of the Ocean ðð þ (NEMO)–Louvain-la-Neuve Sea Ice Model (LIM) 1 t(l) 1 global coupled sea ice–ocean model, which has been used T ’2 dA 2 k(h 2 h ) dl, (9) r r f r c b in previous Arctic studies (Lique et al. 2009, 2010; Lique 0 2 and Steele 2012). Using a drag coefficient of 2.25 3 10 3 where hc and hb are the layer thickness in the center and (Timmermann et al. 2005), the model annual mean ice– 2 at the boundary of the domain respectively. If we as- ocean stress of 0.018 N m 2 between 1979 and 1990 cor- l 2 2 sume that at steady state T 5 0, t( ) 5 0.02 N m 2 (dis- responds to an ‘‘effective’’ wind speed of 2.0 m s 1. This is 21 cussed in more detail later) and hc 2 hb 5 60 m based on lower than the ERA-Interim annual mean of 2.5 m s the deepening of the 34.8 isohaline across the Beaufort over the same period, because of the sea ice cover re- Gyre in the PHC, then the eddy diffusivity (k) must ducing the transfer of momentum into the upper ocean. 2 equal approximately 1300 m2 s 1. Figure 4a shows the mean layer thickness and velocity While eddies are likely to be the first-order process fields over the last 5 years of the 40-yr control run, while responsible for balancing the Ekman pumping (Marshall Fig. 4b shows a cross section of the mean layer thickness et al. 2002), other processes such as lateral friction against through the center of the domain. As expected from the the Chukchi Cap may play a role, and in our model their scaling of the eddy diffusivity, the halocline layer has effect has been incorporated into the strength of the eddy thickened on average by 60 m across the domain, with the flux. Consequently, it must be noted that the strength of maximum thickening in the center. Around the bound- the eddy flux may be an overestimate. aries and in the outflow region little thickening has oc- curred, which makes us confident that any forcing in the outflow region is not affecting the response within 3. Control run the circular domain. Associated with the deepening of the Initially the model is run for 40 years from rest in order halocline layer is an anticyclonic geostrophic circulation to reach an idealized equilibrium state. The magnitude of centered over the domain. The time-averaged maximum 2 the annual cycle in ocean surface stress used to force this velocity of approximately 2 cm s 1 is consistent with the control run (Fig. 3c) is based on the average conditions climatological values reported in McPhee et al. (2009) and that existed between 1979 and 1990 (i.e., before the dra- McPhee (2013). Figures 4c and 4d show the annual cycles matic decline in Arctic sea ice cover; Stroeve et al. 2012). in layer thickness in the center of the domain, and the Figures 3a and 3b show the summer (4 months: July– transport through the channel over the last 5 years of the October) and winter (8 months: November–June) aver- control run. The annual cycle in layer thickness, which can ages of the mean sea level pressure from the monthly be interpreted as a proxy for the annual cycle in freshwater European Centre for Medium-Range Weather Forecasts content, oscillates around a mean of 458.7 m and has (ECMWF) Interim Re-Analysis (ERA-Interim; Dee a peak-to-peak amplitude of 3.4 m, consistent with the et al. 2011) and sea ice concentration from the Special results of Proshutinsky et al. (2009). The annual cycle in Sensor Microwave Imager (SSM/I) satellite instrument transport through the channel has a peak-to-peak ampli- 2 (Maslanik and Stroeve 1999) over this period. The winds tude of 0.2 Sv (1 Sv [ 106 m3 s 1), oscillating around zero. associated with the atmospheric Beaufort high are The maximum freshwater content and maximum inflow stronger and more anticyclonic during winter than during into the domain occurs at the end of our winter period, summer, and there is a decline in the sea ice concentra- whereas the minimum freshwater content and maximum tion over the Beaufort Gyre during the four months of outflow occurs at the end of our summer period, reflecting summer (although this seasonal change is much smaller the integrated forcing throughout the year. than that seen in more recent observations). The results Note that sensitivity analysis has shown that the re- of Yang (2009) suggest that over this period the ocean sults in the following sections are not dependent upon surface stress is dominated by atmospheric forcing rather the exact magnitude of the annual cycle in ocean surface

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 1NOVEMBER 2014 D A V I S E T A L . 8175

FIG. 3. Summer (July–October) and winter (November–June) averages of (a) mean sea level pressure from the ERA-Interim reanalysis and (b) sea ice concentration from the SSM/I sat- ellite instrument over the Arctic Ocean between 1979 and 1990. (c) The annual cycle in ocean surface stress used to force the control run is based on the occurrence of stronger and more anticyclonic winds over the Beaufort Gyre during the 8 months of winter (green) compared to the four months of summer (red). The reduction in sea ice cover between July and October defines our summer period. stress used to force the control run or on the realism of changing. Each may have an effect on the accumulation of the steady state. freshwater in the Beaufort Gyre by triggering a different dynamical response, and therefore two sets of experi- 4. Experiment 1: Modified seasonality ments were designed in order to investigate their effects Because the Arctic sea ice cover has begun to break up separately. and retreat farther and for longer each year, the annual In the first set of experiments, the effect of modifying mean ocean surface stress (i.e., the net forcing) is in- the seasonality in ocean surface stress while holding the creasing and also its seasonal cycle (i.e., the seasonality) is net forcing constant was investigated. Figures 5a and 5b

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 8176 JOURNAL OF CLIMATE VOLUME 27

FIG. 4. (a) Mean layer thickness (colors) and velocity field (vectors), (b) cross section of the mean layer thickness through the center of the domain (i.e., along the white line), (c) the annual cycle in layer thickness in the center of the domain (i.e., at the white dot), and (d) the annual cycle in transport through the channel (i.e., across the red line) all calculated over the last 5 years of the control run. Positive transports indicate an inflow and negative transports indicate an outflow. Integer year numbers correspond to the end of our summer period. show the idealized modifications made to the annual the domain, and the transport through the channel over cycle in ocean surface stress. Over a number of model the last 5 years of each model run. The key result is that runs, the ocean surface stress during the 4 months of if the net forcing is held constant, the changing season- summer was incrementally increased with respect to the ality in ocean surface stress has no effect on the fresh- control run (Fig. 5a), idealistically representing an in- water content of the Beaufort Gyre. The annual cycle in creased efficiency of momentum transfer into the ocean. layer thickness (and thus the freshwater content) in each of For high levels of stress in summer, the period over the model runs oscillates around the control run mean of which this stress was applied was then incrementally 458.7 m (Figs. 6b,e), and when the annual cycle in transport increased from 4 to 6 months (Fig. 5b), representing through the channel is integrated over the full 40 years of a longer melt season and therefore a longer period of each model run, it is clear that there is no net inflow into enhanced momentum transfer. At the same time, how- the domain (Figs. 6c,f). ever, the ocean surface stress in winter for each model Nevertheless, by changing the seasonality in ocean sur- run was reduced, to ensure that the net forcing remained face stress we have significantly altered the shape of the 2 constant at 0.02 N m 2 (gray dotted line in Figs. 5a and annual cycle in both the layer thickness in the center of 5b). Each model run was initialized from the layer the domain, and the transport through the channel. With thickness and velocity fields at the end of the control run respect to the control run, as the ocean surface stress is and then run for an additional 40 years. Note that, while steadily increased in summer (cf. the blue line with the we have used the terminology ‘‘summer’’ and ‘‘winter’’ dashed, dotted, and dot-dashed red lines in Fig. 6a), the here, our seasons are defined entirely in terms of periods phase of the annual cycle in both the layer thickness and of different ocean surface stress and need not relate to the transport through the channel eventually reverses, as specific times of the year. the stress in summer becomes greater than the stress in Figure 6 shows the results of these experiments in terms winter, and the amplitude increases, reflecting the increased of both the annual cycle in layer thickness in the center of amplitude of the annual cycle in ocean surface stress. The

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 1NOVEMBER 2014 D A V I S E T A L . 8177

FIG. 5. Modifications made to the annual cycle in ocean surface stress with respect to the control run (thick blue line) for (a),(b) the modified seasonality experiments and (c),(d) the modified net forcing experiments. In (a),(c), the ocean surface stress is incrementally increased in summer, representing the increased efficiency of momentum transfer. In (b),(d), the period over which high values of summer stress is applied is then incrementally increased, representing a longer melt season. For just the modified seasonality experiments in (a),(b), the stress in winter in each model run is reduced, to ensure that when integrated over the entire year the net forcing remains constant at 2 0.02 N m 2 (gray dashed line). peak-to-peak amplitudes increase from 3.4 m and 0.2 Sv to run was again initialized from the layer thickness and a maximum (over these model runs) of 5.7 m and 0.3 Sv for velocity fields at the end of the control run and then in- the layer thickness in the center of the domain (dot-dashed tegrated for an additional 40 years. red line in Fig. 6b) and the transport through the channel Figure 7a shows that the mean layer thickness in the (dot-dashed red line in Fig. 6c), respectively. center of the domain, averaged over the last 5 years of When the length of the melt season is increased for each model run, increases linearly with respect to the high values of ocean surface stress in summer (cf. the increased ocean surface stress. The linear relationship is blue line with the dot-dashed, dashed, and dotted green described by the equation Dh 5 2937:5Dt,whereDh and lines in Fig. 6d), the amplitudes of the annual cycle in Dt are the increase in mean layer thickness and net layer thickness and the transport through the channel forcing, respectively, from the control run. To estimate increase further to 8.6 m and 0.5 Sv, respectively (dotted the magnitude of the corresponding freshwater accumu- lines in Figs. 6e and 6f). The timing of the minimum lation, the annual cycle in transport through the channel freshwater content also occurs earlier because of the can be integrated in time to produce a cumulative volume shorter winter period, whereas the timing of maximum transport into the domain over the full 40 years of each freshwater content remains unchanged. model run. This cumulative volume transport can then be multiplied by a salinity anomaly:

5. Experiment 2: Modified net forcing S 2 S 5 ref SAnomaly , (10) To explore the dynamical response of the Beaufort Sref Gyre to an increased net forcing, and its effect on the freshwater content, a second set of experiments was where Sref is the reference freshwater salinity and S is the performed. We applied the same modifications to the in situ salinity, to calculate the total freshwater accumu- annual cycle in ocean surface stress as seen in the modi- lation. As the model only solves the dynamical equations fied seasonality experiments (Figs. 5a,b) but, in this case, and thus there is no salinity variable, S has been set to the ocean surface stress in winter was not reduced to a salinity of 27.7 in accordance with Giles et al. (2012). offset the increase in summer (Figs. 5c,d). Consequently, This value is based on the salinity measured in the mixed the net forcing increased from the control run value of layer by an Ice Tethered Profiler deployed in the Canada 2 2 0.02 N m 2 to a maximum of 0.028 N m 2.Eachmodel Basin during April 2007 (Toole et al. 2010), and as such

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 8178 JOURNAL OF CLIMATE VOLUME 27

FIG. 6. The annual cycle in (a),(d) ocean surface stress, (b),(e) the layer thickness in the center of the domain (white dot in Fig. 4a), and (c),(f) the transport through the channel (red line in Fig. 4a) from the last 5 years of each of the modified seasonality experiments. Positive transports indicate an inflow and negative transports indicate an outflow. Note that (a)–(c) are for model runs where the ocean surface stress in summer has been increased, while (d)–(f) are for model runs where the melt season length has been increased. Line styles and colors correspond to those used in Figs. 5a and 5b. The thick blue line represents the control run. Integer year numbers correspond to the end of our summer period.

FW the freshwater anomaly estimate that results must be where V(t) is the cumulative freshwater transport into FW viewed as an upper bound. The reference value (Sref) the domain as a function of time t, VTot is the total equals 34.7 [again in accordance with Giles et al. (2012)]. freshwater accumulated in the domain for each model Figure 7b shows the cumulative volume and freshwater run, and k is the growth constant that reveals an e-folding transports into the domain as a function of both time and time scale of approximately 3.3 yr for the accumulation the increase in net forcing. All model runs show a net flux of freshwater in the domain. This is slightly longer than of freshwater into the domain with respect to the control the e-folding time scale of approximately 2.7 yr for the run. The lack of seasonal variation when Dt is approxi- increase in the mean layer thickness in the center of the 2 2 mately 4 3 10 3 Nm 2 is due to the lack of seasonal domain. As with the mean layer thickness in the center variability in the forcing (i.e., the ocean surface stress in of the domain, the total accumulation of freshwater summer has been increased to a point where it equals the depends linearly on the increase in net forcing, and for the ocean surface stress in winter). Fitting an asymptotic greatest increase in net forcing in our suite of model 2 exponential growth model to each of the model runs runs (0.008 N m 2), the total freshwater accumulated in the domain is approximately 4000 km3. The re- FW 5 FW 2 2kt V(t) VTot (1 e ), (11) sults of Martin et al. (2014) suggest that between 2000

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 1NOVEMBER 2014 D A V I S E T A L . 8179

section we will discuss some of the broader implications of our results and their relevance to the wider Arctic system.

a. Adjustment timescales The results of both the modified seasonality experi- ments and the modified net forcing experiments shed some light on the time scales over which the Arctic Ocean adjusts to a change in forcing, as well as on which adjustment processes are responsible. Figures 6b and 6e show that over each season the layer thickness in the center of the domain continues to increase or decrease in response to the change in ocean surface stress, and never fully adjusts to a new equilibrium state. In con- trast, Figs. 6c and 6f suggest that the magnitude of the transport through the channel adjusts more quickly to the seasonal changes in ocean surface stress, asymptot- ing toward a new equilibrium value over the period of each season. Consequently, it appears the processes that dominate the adjustment to a change in forcing around the boundary of the domain are very different from those that determine the adjustment in the center. The response time scale (e-folding time scale) of the trans- port through the channel is approximately 1 month, which is in good agreement with the time required for a boundary trapped Kelvin wavepffiffiffiffiffiffiffi to propagate around the domain at a speed given by g0h. This suggests that FIG. 7. Results of the modified net forcing experiments. (a) The boundary waves are the dominant process behind the increase in mean layer thickness in the center of the domain plotted fast adjustment on the boundary. Waves propagating as a function of the increase in net forcing. The blue dot represents out of the sponge region (where we apply a strong re- the control run, the red dots represent model runs with an in- storing to the initial layer thickness of 400 m) effectively creased ocean surface stress in summer, and the green dots rep- resent model runs with a longer melt season (as per the colors in fix the layer thickness around the boundary in our model Figs. 5c and 5d). (b) The cumulative volume and freshwater on time scales longer than 1 month. transports into the domain over the full 40 years of each model run Away from the boundaries, however, Kelvin waves as a function of the increase in net forcing. have little effect. Here, the layer thickness as a function of time is determined by the balance between wind- and 2012, the Arctic-wide annual mean ocean surface induced Ekman pumping and the eddy-induced volume 22 stress has increased by 0.007 N m , which corresponds to flux toward the boundary [i.e., Eq. (8)]: an accumulation of freshwater in our model of ap- ðð ðð l þ proximately 3500 km3. This is around 50% of the in- › 1 t( ) 1 hdA52 dA 2 k h 2 h dl › r ( c b) , (12) crease observed by Giles et al. (2012) over broadly the t r 0 f r same period. where the first term on the right-hand side represents wind-induced Ekman pumping and the second term rep- 6. Discussion resents the relaxation effect of the eddy flux. Scale analysis Using a simple process model and idealized perturba- suggests that the relaxation time scale due to the eddy term tions to the ocean surface stress, we have demonstrated is given by rABG/klBG,whereABG and lBG are the area that the dynamical response of the Beaufort Gyre to the and circumference of the Beaufort Gyre respectively, and decline in Arctic sea ice cover, and the related increase in is approximately equal to 14 yr. As a result, we can con- the efficiency of momentum transfer into the upper clude that eddies play a limited role in balancing Ekman ocean, is a plausible mechanism behind the accelerated pumping on the seasonal time scale. Consequently, the accumulation of freshwater inferred by Giles et al. (2012). layer thickness in the center of the domain continually Despite the simple nature of our process model, in this adjusts to the change in ocean surface stress over each

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 8180 JOURNAL OF CLIMATE VOLUME 27 season, and does not tend toward a new equilibrium state. (where form drag is important), the drag coefficient peaks Furthermore, this long adjustment time scale ensures that at an ice concentration of approximately 50%. the annual cycle in layer thickness in the center of the Irrespective of the exact details, the Arctic is currently domain (e.g., 3.4 m in the control run) is only a small undergoing a transition from the presence of a thick and fraction of the change in mean layer thickness across the extensive sea ice cover year round to a seasonally ice Beaufort Gyre (60 m), as the annual cycle represents only free state. If the concept of an optimal ice concentration the very beginning of a much longer-term adjustment to for the efficiency of momentum transfer into the upper the changing ocean surface stress. Indeed, when the model ocean holds, then the current positive trend in the mo- is run from rest to a fully adjusted equilibrium state with mentum flux will begin to slow and eventually reverse, as a constant ocean surface stress of either 0.023 75 or the length of time when the sea ice concentration is 2 0.0125 N m 2 (i.e., the winter or summer value of ocean below the optimum value increases (Martin et al. 2014). surface stress from the control run, respectively), the dif- At this point, the accelerated accumulation of freshwa- ference in equilibrium layer thickness between the two ter in the Beaufort Gyre will stop, and the excess runs is approximately 33 m. freshwater that has built up will no longer be supported In the modified net forcing experiments, however, the by Ekman pumping. It is unclear how long it will take for eddy relaxation term eventually balances the wind- the positive trend in the momentum flux to reverse, and induced Ekman pumping, and it is the eddy relaxation this may have important implications for the future time scale that determines the total quantity of fresh- evolution of the Arctic’s freshwater budget and for water that is accumulated in the domain (see Fig. 7b). freshwater release into the North Atlantic. This highlights the importance of eddies in the Arctic Throughout this study, we have assumed in the con- Ocean’s adjustment to a change in forcing, and provides struction of our idealized annual cycles that the momen- a strong motivation to use eddy-resolving models when tum transfer will increase during summer. It should be investigating the transient response of the Arctic Ocean noted, however, that the seasons have been arbitrarily to future changes in the forcing. defined, and the increased momentum transfer in summer could equally be interpreted as an increased momentum b. Effect of sea ice concentration transfer in winter. Furthermore, given the results of the The results of both Martin et al. (2014) and Tsamados modified seasonality experiments, it is clear that whether et al. (2014) show a long-term positive trend in the an- the momentum transfer is increased in winter or summer nually averaged momentum flux into the upper ocean in is not important for the accelerated accumulation of response to the decline in Arctic sea ice. However, Martin freshwater in the Beaufort Gyre, as all that matters is the et al. (2014) find that the trend is not positive over all integrated net forcing over the entire year. seasons. Instead, their model suggests that the efficiency c. Relevance of our simple model of momentum transfer into the upper ocean is at an op- timum when the sea ice concentration is approximately Despite having no effect on the freshwater content of the 80% (i.e., in fall and spring). Above this point, a thick and Beaufort Gyre, modifications made to the seasonality in extensive sea ice cover damps the transfer of momentum ocean surface stress while the net forcing is held constant do due to the large internal ice stresses and shielding of the impact the annual cycle in freshwater content and transport ocean from direct wind forcing. Below this point, the through the channel in our model (see Figs. 6b,c,e,f). momentum transfer into the upper ocean decreases, as Given the limitations of our simple process model the drag associated with drifting sea ice is greater than (including, but not limited to, the parameterized effects that of open water. Consequently, between the 1980s and of eddies and diapycnal mixing, and the lack of vertical 2000s the ocean surface stress in summer decreases by 7% resolution and a sea ice component), it is possible there in their model, reflecting the increasing length of time are processes missing from our model that may be im- that the sea ice concentration is below 80%. It must be portant in determining the response of the Beaufort Gyre noted however, that this negative trend in summer was to a change in the seasonality in ocean surface stress. To not observed in the results of Tsamados et al. (2014).Itis explore this further, we examine the driving mechanisms possible that the form drag parameterization used by behind the annual cycle of freshwater content in the Tsamados et al., which was not implemented in the model Beaufort Gyre using Arctic climatological datasets and used by Martin et al., allows the ice concentration to data collected in the Beaufort Gyre. decline further before the efficiency of momentum Averaged over the Beaufort Gyre region, Fig. 8 shows transfer decreases due to the fraction of open water. For the annual cycle in liquid freshwater content anomaly example, the observations of Andreas et al. (2010) show integrated between both the surface and the 34.8 iso- that over summer sea ice and the marginal ice zone haline (blue solid line) and between the surface and 25 m

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 1NOVEMBER 2014 D A V I S E T A L . 8181

FIG. 8. The blue lines show the mean annual cycle in liquid freshwater content anomaly integrated between the surface and 25 m (dotted line) and between the surface and the depth of the 34.8 isohaline (solid line) over the Beaufort Gyre region from the MIMOC. The red line shows the mean annual cycle in the solid (ice) freshwater content anomaly from upward- looking sonar deployed as part of the Beaufort Gyre Exploration Project. The black line is the mean annual cycle in Ekman pumping velocity at the base of the Ekman layer over the Beaufort Gyre region, calculated from the wind stress curl. Positive velocities indicate a downwelling.

(blue dashed line) from the Monthly Isopycnal/Mixed- process model does resolve the important dynamical Layer Ocean Climatology (MIMOC; Schmidtko et al. processes behind the response of the Beaufort Gyre to 2013), the annual cycle in solid (ice) freshwater content a change in forcing, and that the decline in Arctic sea ice anomaly (red line) from upward-looking sonar deployed cover will affect the seasonality of the Beaufort Gyre and as part of the Beaufort Gyre Exploration Project, and its freshwater content in a manner similar to the results of the magnitude of the annual cycle in Ekman pumping our model. We would expect the amplitude of the annual (black line) from the ERA-Interim reanalysis. We have cycle in freshwater content (solid blue line in Fig. 8)to used MIMOC instead of PHC because it resolves the increase as the amplitude of the annual cycle in ocean annual cycle and has been updated with the most recent surface stress increases, and the timing of the peaks in the surface data collected by ice-tethered profilers. In the annual cycle to change as the efficiency of momentum upper 25 m of the water column, the annual cycle in transfer increases and decreases at different times liquid freshwater content (blue dashed line) is domi- throughout the year. nated by thermal forcing (sea ice formation and melt; d. Implications for the Arctic freshwater budget red line). It reaches a minimum in April–May because of brine rejection from sea ice formation, and a maximum in In the modified seasonality experiments, the ampli- October–November because of sea ice melt. On the other tude of the annual cycle in transport through the channel hand, over the full depth range of the Beaufort Gyre (i.e., averages 0.3–0.4 Sv. Compared to the studies of de Steur from the surface to the depth of the 34.8 isohaline), the et al. (2009) and Curry et al. (2014), this is smaller than annual variation in Ekman pumping (black line) appears the amplitude of the annual cycle in the volume export to be the dominant forcing behind the annual cycle in of surface layer waters through both Fram Strait (;4 Sv) total liquid freshwater content (solid blue line), with and Davis Strait (;0.6–0.8 Sv). The small amplitude in thermal forcing having little effect. The liquid freshwater our model directly reflects the limited annual variability content peaks in December–January due to stronger seen in the layer thickness, which is both consistent with Ekman pumping during winter (although there appears observational data (e.g., Proshutinsky et al. 2009) and is to be a short time lag) and is at a minimum during a direct consequence of the long time scale over which August–September due to weaker Ekman pumping and the model adjusts to the change in forcing. Conse- a relaxation of the salinity field. An intermediary peak is quently, while having some impact, the annual vari- observed in May, which is most likely associated with ability in the export of freshwater from the Beaufort a seasonal reintensification of the atmospheric Beaufort Gyre cannot be the main driver behind the annual var- high (Yang 2006, 2009). iability seen in the exports to either side of Greenland, Together, these results suggest that Ekman pumping is and variability driven by local conditions is likely to be the dominant process in setting the annual cycle in liquid more important. For example, Lique et al. (2009) have freshwater content within the Beaufort Gyre (although shown that the variability in the liquid freshwater flux locally the balance may be different; Proshutinsky et al. through Fram Strait depends upon variability in both the 2009). Consequently, we can conclude that our simple volume flux and the in situ salinity, which varies strongly

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 8182 JOURNAL OF CLIMATE VOLUME 27 because of ice formation and melt north of Greenland. summer may be reinforced by an enhanced wind forcing Given the far-reaching consequences that changes in the over the Beaufort Gyre region. export of freshwater to either side of Greenland may have on the circulation of the , future 7. Conclusions studies should aim to quantify the contribution that the seasonal release of freshwater from the Beaufort Gyre The overall aim of this study was to investigate the has on the total freshwater export from the Arctic dynamical response of the Beaufort Gyre to the changing Ocean. efficiency of momentum transfer associated with the de- Independent of any accumulation of freshwater due to cline in Arctic sea ice cover, and its link to the accelerated sea ice decline, the freshwater content of the Beaufort accumulation of freshwater observed by Giles et al. Gyre has been steadily increasing since the mid-1990s due (2012). From two sets of experiments where the annual to the large-scale anticyclonic wind regime (Proshutinsky cycle in ocean surface stress was idealistically perturbed et al. 2009). If, in the future, the wind regime becomes to reflect the effect of diminishing Arctic sea ice, we have cyclonic, then the excess freshwater stored in the Beau- shown that when the annual mean stress (i.e., the net fort Gyre will be released to the shelves (Häkkinen and forcing) is held constant, changes to its seasonal cycle Proshutinsky 2004; Condron et al. 2009; Jahn et al. 2010), (i.e., the seasonality) have no net effect on the freshwater where it may be exported into the North Atlantic content of the Beaufort Gyre in our model. We find that (Stewart and Haine 2013). Such a release of freshwater the Beaufort Gyre continually adjusts to the seasonal has been proposed as a mechanism to explain some of the changes in ocean surface stress since the adjustment time Great Salinity Anomaly (GSA) events, during which the scale, which is set by eddies, is considerably longer than subpolar North Atlantic and Nordic seas underwent de- seasonal. cadal periods of freshening (e.g., Dickson et al. 1988; On the other hand, we find a linear relationship between Belkin et al. 1998; Belkin 2004), with associated effects on the increase in the annual mean ocean surface stress and the formation of North Atlantic Deep Water (Dickson the amount of freshwater accumulated in the Beaufort et al. 2000; Jahn et al. 2010). Consequently, the acceler- Gyre, regardless of the seasonal cycle in the forcing. This ated accumulation of freshwater in the Beaufort Gyre linear relationship is similar to that found by Proshutinsky may exacerbate the effects of a switch to a cyclonic wind et al. (2002, 2009) between the freshwater content of the regime, by making the Arctic even more anomalously Beaufort Gyre and the wind stress curl over the region, fresh beforehand, and thus increasing the quantity of and it suggests that as the Arctic sea ice cover continues to freshwater that may be exported into the North Atlantic. decline, the linear accumulation of freshwater in the Beaufort Gyre for a given wind stress curl will increase e. Effect of changing meteorological conditions (e.g., Giles et al. 2012), because of the enhancement in the Throughout this study, we have focused on the effect of magnitude of the stress that is transferred from the surface changing sea ice conditions on the freshwater content of of the ice pack to the ocean below. This confirms that the the Beaufort Gyre, without considering the potential im- increased efficiency of momentum transfer into the upper pacts of changes in the wind field. However, as the strength ocean associated with the decline in Arctic sea ice cover is of the atmospheric Beaufort high projects onto several a plausible mechanism to explain the accelerated accu- different modes of atmospheric variability (Serreze and mulation of freshwater in the Beaufort Gyre. The total Barrett 2011; Mauritzen 2012), large-scale changes in at- quantity of freshwater accumulated for a given change in mospheric conditions may lead to a further accumulation ocean surface stress depends on the eddy diffusivity (i.e., or release of freshwater from the Beaufort Gyre that has on the length of time it takes for the eddy field to balance not been accounted for in our model. For example, since the change in Ekman pumping). The key dynamical pro- the mid-1990s a negative trend in the phase of the Arctic cess at play is as follows: as Arctic sea ice is becoming Oscillation has strengthened the atmospheric Beaufort weaker, thinner, and more broken up, the annually aver- high in winter (i.e., the anticyclonic wind regime) and, aged momentum flux into the upper ocean is increasing for more recently, a positive trend in the strength of the the same wind speed, resulting in an accelerated linear Arctic dipole between 2007 and 2012 may have acted to accumulation of freshwater through an enhanced me- strengthen the Beaufort high in summer (Overland et al. chanical deformation of the salinity field. 2012; Ogi and Wallace 2012). Furthermore, the study of Our results have implications for both the net amount Simmonds and Keay (2009) has shown that the strength of of freshwater stored in the Beaufort Gyre and the an- Arctic summer cyclones over the Eurasian Arctic has nual cycle in its freshwater content. This may have an been increasing over the past 30 years, and if this were to effect on the annual cycle in freshwater export from the continue, then the momentum flux into the ocean during Arctic Ocean, and future observational studies aimed at

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 1NOVEMBER 2014 D A V I S E T A L . 8183 detecting changes in the seasonality of the Beaufort Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Gyre freshwater content would be worthwhile. At the Configuration and performance of the data assimilation system. same time, our results have highlighted the importance Quart. J. Roy. Meteor. Soc., 137, 553–597, doi:10.1002/qj.828. de Steur, L., E. Hansen, R. Gerdes, M. Karcher, E. Fahrbach, and of eddies in the transient response of the Arctic to J. Holfort, 2009: Freshwater fluxes in the East Greenland a change in forcing, and more should be done to in- Current: A decade of observations. Geophys. Res. Lett., 36, vestigate the role they play using eddy resolving models. L23611, doi:10.1029/2009GL041278. Dickson, R. R., J. Meincke, S.-A. Malmberg, and A. J. Lee, 1988: Acknowledgments. We thank Michel Tsamados, The ‘‘Great Salinity Anomaly’’ in the northern North At- lantic 1968–1982. Prog. Oceanogr., 20, 103–151, doi:10.1016/ Daniel Feltham, and David Marshall for useful discus- 0079-6611(88)90049-3. sions during this study. The upward-looking sonar data ——, and Coauthors, 2000: The Arctic Ocean response to the North were collected and made available by the Beaufort Gyre Atlantic Oscillation. J. Climate, 13, 2671–2696, doi:10.1175/ Exploration Project based at the Woods Hole Oceano- 1520-0442(2000)013,2671:TAORTT.2.0.CO;2. graphic Institution (http://www.whoi.edu/beaufortgyre) ——, B. Rudels, S. Dye, M. Karcher, J. Meincke, and I. Yashayaev, 2007: Current estimates of freshwater flux through Arctic and in collaboration with researchers from Fisheries and subarctic seas. Prog. Oceanogr., 73, 210–230, doi:10.1016/ Oceans Canada at the Institute of Ocean Sciences. The j.pocean.2006.12.003. comments and suggestions of three anonymous re- Flocco, D., D. Schroeder, D. L. Feltham, and E. C. Hunke, 2012: viewers are also gratefully acknowledged. This work was Impact of melt ponds on Arctic sea ice simulations from 1990 funded by the Natural Environment Research Council to 2007. J. Geophys. Res., 117, C09032, doi:10.1029/ 2012JC008195. (NERC). Gent, P. R., 2011: The Gent–McWilliams parameterization: 20/20 hindsight. Ocean Modell., 39, 2–9, doi:10.1016/ REFERENCES j.ocemod.2010.08.002. Giles, K. A., S. W. Laxon, A. L. Ridout, D. J. Wingham, and Aagaard, K., and E. C. Carmack, 1989: The role of sea ice and other S. Bacon, 2012: Western Arctic Ocean freshwater storage in- fresh water in the Arctic circulation. J. Geophys. Res., 94, creased by wind-driven spin-up of the Beaufort Gyre. Nat. 14 485–14 498, doi:10.1029/JC094iC10p14485. Geosci., 5, 194–197, doi:10.1038/ngeo1379. ——, L. K. Coachman, and E. Carmack, 1981: On the halocline of Häkkinen, S., and A. Proshutinsky, 2004: Freshwater content var- the Arctic Ocean. Deep-Sea Res., 28A, 529–545, doi:10.1016/ iability in the Arctic Ocean. J. Geophys. Res., 109, C03051, 0198-0149(81)90115-1. doi:10.1029/2003JC001940. Allison, L. C., H. L. Johnson, and D. P. Marshall, 2011: Spin-up Jahn, A., B. Tremblay, L. A. Mysak, and R. Newton, 2010: Effect of and adjustment of the Antarctic Circumpolar Current and the large-scale on the variability of the global pycnocline. J. Mar. Res., 69, 167–189, doi:10.1357/ Arctic Ocean freshwater export. Climate Dyn., 34, 201–222, 002224011798765330. doi:10.1007/s00382-009-0558-z. Andreas, E. L, T. W. Horst, A. A. Grachev, P. O. G. Persson, C. W. Johnson, H. L., and D. P. Marshall, 2002: A theory for the sur- Fairall, P. S. Guest, and R. E. Jordan, 2010: Parametrizing tur- face Atlantic response to thermohaline variability. J. Phys. bulent exchange over summer sea ice and the marginal ice zone. Oceanogr., 32, 1121–1132, doi:10.1175/1520-0485(2002)032,1121: Quart. J. Roy. Meteor. Soc., 136, 927–943, doi:10.1002/qj.618. ATFTSA.2.0.CO;2. Belkin, I. M., 2004: Propagation of the ‘‘Great Salinity Anomaly’’ Jones, E. P., and L. G. Anderson, 2008: Is the global conveyor belt of the 1990s around the northern North Atlantic. Geophys. threatened by Arctic Ocean fresh water outflow. Arctic–Subarctic Res. Lett., 31, L08306, doi:10.1029/2003GL019334. Ocean Fluxes: Defining the Role of the Northern Seas in Cli- ——, S. Levitus, J. Antonov, and S.-A. Malmberg, 1998: ‘‘Great mate, R. R. Dickson, J. Meincke, and P. Rhines, Eds., Salinity Anomalies’’ in the North Atlantic. Prog. Oceanogr., 41, Springer, 385–404. 1–68, doi:10.1016/S0079-6611(98)00015-9. Krishfield, R. A., A. Proshutinsky, K. Tateyama, W. J. Williams, Carmack, E. C., 2000: The Arctic Ocean’s freshwater budget: Sour- E. C. Carmack, F. A. McLaughlin, and M. L. Timmermans, 2014: ces, storage, and export. The Freshwater Budget of the Arctic Deterioration of perennial sea ice in the Beaufort Gyre from Ocean, E. L. Lewis et al., Eds., Springer, 91–126. 2003 to 2012 and its impact on the oceanic freshwater cycle. Cavalieri, D. J., and C. L. Parkinson, 2012: Arctic sea ice variability J. Geophys. Res., 119, 1271–1305, doi:10.1002/2013JC008999. and trends, 1979–2010. Cryosphere, 6, 881–889, doi:10.5194/ Kwok, R., and D. A. Rothrock, 2009: Decline in Arctic sea ice tc-6-881-2012. thickness from submarine and ICESat records: 1958–2008. Coachman, L. K., and K. Aagaard, 1974: Physical of Geophys. Res. Lett., 36, L15501, doi:10.1029/2009GL039035. Arctic and subarctic seas. Marine Geology and Oceanography Laxon, S. W., and Coauthors, 2013: CryoSat-2 estimates of Arctic of the Arctic Seas, Y. Herman, Ed., Springer, 1–72. sea ice thickness and volume. Geophys. Res. Lett., 40, 732–737, Condron, A., P. Winsor, C. Hill, and D. Menemenlis, 2009: Simu- doi:10.1002/grl.50193. lated response of the Arctic freshwater budget to extreme Lique, C., and M. Steele, 2012: Where can we find a seasonal cycle NAO wind forcing. J. Climate, 22, 2422–2437, doi:10.1175/ of the Atlantic water temperature within the Arctic basin? 2008JCLI2626.1. J. Geophys. Res., 117, C03026, doi:10.1029/2011JC007612. Curry, B., C. M. Lee, B. Petrie, R. E. Moritz, and R. Kwok, 2014: ——, A. M. Treguier, M. Scheinert, and T. Penduff, 2009: A model- Multiyear volume, liquid freshwater, and sea ice transports based study of ice and freshwater transport variability along through Davis Strait, 2004–10. J. Phys. Oceanogr., 44, 1244– both sides of Greenland. Climate Dyn., 33, 685–705, 1266, doi:10.1175/JPO-D-13-0177.1. doi:10.1007/s00382-008-0510-7.

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC 8184 JOURNAL OF CLIMATE VOLUME 27

——, ——, B. Blanke, and N. Grima, 2010: On the origins of water ——, and Coauthors, 2006: The large-scale freshwater cycle of the masses exported along both sides of Greenland: A Lagrangian Arctic. J. Geophys. Res., 111, C11010, doi:10.1029/ model analysis. J. Geophys. Res., 115, C05019, doi:10.1029/ 2005JC003424. 2009JC005316. Simmonds, I., and K. Keay, 2009: Extraordinary September Arctic Marshall, J., H. Jones, R. Karsten, and R. Wardle, 2002: Can eddies sea ice reductions and their relationships with storm behavior set ocean stratification? J. Phys. Oceanogr., 32, 26–38, over 1979–2008. Geophys. Res. Lett., 36, L19715, doi:10.1029/ doi:10.1175/1520-0485(2002)032,0026:CESOS.2.0.CO;2. 2009GL039810. Martin, T., M. Steele, and J. Zhang, 2014: Seasonality and long- Spall, M. A., 2003: On the thermohaline circulation in flat bottom term trend of Arctic Ocean surface stress in a model. J. Geo- marginal seas. J. Mar. Res., 61, 1–25, doi:10.1357/ phys. Res., 119, 1723–1738, doi:10.1002/2013JC009425. 002224003321586390. Maslanik, J., and J. Stroeve, 1999 (updated daily): Near-real-time ——, 2004: Boundary currents and watermass transformation in DMSP SSM/I-SSMIS daily polar gridded sea ice concentra- marginal seas. J. Phys. Oceanogr., 34, 1197–1213, doi:10.1175/ tions. National Snow and Ice Data Center, Boulder, CO, digital 1520-0485(2004)034,1197:BCAWTI.2.0.CO;2. media. [Available online at http://nsidc.org/data/nsidc-0081.] ——, 2013: On the circulation of Atlantic Water in the Arctic Ocean. ——, C. Fowler, J. Stroeve, S. Drobot, J. Zwally, D. Yi, and J. Phys. Oceanogr., 43, 2352–2371, doi:10.1175/JPO-D-13-079.1. W. Emery, 2007: A younger, thinner Arctic ice cover: Increased Steele, M., R. Morley, and W. Ermold, 2001: PHC: A global ocean potential for rapid, extensive sea-ice loss. Geophys. Res. Lett., hydrography with a high-quality Arctic Ocean. J. Climate, 34, L24501, doi:10.1029/2007GL032043. 14, 2079–2087, doi:10.1175/1520-0442(2001)014,2079: Mauritzen, C., 2012: Arctic freshwater. Nat. Geosci., 5, 162–164, PAGOHW.2.0.CO;2. doi:10.1038/ngeo1409. Stewart, K. D., and T. W. N. Haine, 2013: Wind-driven Arctic McPhee, M. G., 2012: Advances in understanding ice–ocean stress freshwater anomalies. Geophys. Res. Lett., 40, 6196–6201, during and since AIDJEX. Cold Reg. Sci. Technol., 76-77, 24– doi:10.1002/2013GL058247. 36, doi:10.1016/j.coldregions.2011.05.001. Stroeve, J. C., M. M. Holland, W. Meier, T. Scambos, and ——, 2013: Intensification of geostrophic currents in the Canada M. Serreze, 2007: Arctic sea ice decline: Faster than forecast. Basin, Arctic Ocean. J. Climate, 26, 3130–3138, doi:10.1175/ Geophys. Res. Lett., 34, L09501, doi:10.1029/2007GL029703. JCLI-D-12-00289.1. ——, M. C. Serreze, M. M. Holland, J. E. Kay, J. Malanik, and A. P. ——, A. Proshutinsky, J. H. Morison, M. Steele, and M. B. Alkire, Barrett, 2012: The Arctic’s rapidly shrinking sea ice cover: A 2009: Rapid change in freshwater content of the Arctic Ocean. research synthesis. Climatic Change, 110, 1005–1027, Geophys. Res. Lett., 36, L10602, doi:10.1029/2009GL037525. doi:10.1007/s10584-011-0101-1. Morison, J., R. Kwok, C. Peralta-Ferriz, M. Alkire, I. Rigor, Timmermann,R.,H.Goosse,G.Madec,T.Fichefet,C.Ethe, R. Andersen, and M. Steele, 2012: Changing Arctic Ocean fresh- and V. Dulière, 2005: On the representation of high lati- water pathways. Nature, 481, 66–70, doi:10.1038/nature10705. tude processes in the ORCA-LIM global coupled sea ice– Ogi, M., and J. M. Wallace, 2012: The role of summer surface wind ocean model. Ocean Modell., 8, 175–201, doi:10.1016/ anomalies in the summer Arctic sea ice extent in 2010 and 2011. j.ocemod.2003.12.009. Geophys. Res. Lett., 39, L09704, doi:10.1029/2012GL051330. Toole, J. M., M. L. Timmermans, D. K. Perovich, R. A. Krishfield, Overland, J. E., J. A. Francis, E. Hanna, and M. Wang, 2012: The A. Proshutinsky, and J. A. Richter-Menge, 2010: Influences of recent shift in early summer Arctic atmospheric circulation. the ocean surface mixed layer and thermohaline stratification Geophys. Res. Lett., 39, L19804, doi:10.1029/2012GL053268. on Arctic Sea ice in the central Canada Basin. J. Geophys. Proshutinsky, A., and M. A. Johnson, 1997: Two circulation re- Res., 115, C10018, doi:10.1029/2009JC005660. gimes of the wind-driven Arctic Ocean. J. Geophys. Res., 102, Tsamados, M., D. L. Feltham, D. Schroeder, D. Flocco, S. L. Farrell, 12 493–12 514, doi:10.1029/97JC00738. N. T. Kurtz, S. W. Laxon, and S. Bacon, 2014: Impact of var- ——, R. H. Bourke, and F. A. McLaughlin, 2002: The role of the iable atmospheric and oceanic form drag on simulations of Beaufort Gyre in Arctic climate variability: Seasonal to de- Arctic sea ice. J. Phys. Oceanogr., 44, 1329–1353, doi:10.1175/ cadal climate scales. Geophys. Res. Lett., 29, 2100, doi:10.1029/ JPO-D-13-0215.1. 2002GL015847. Vellinga, M., and R. A. Wood, 2002: Global climatic impacts of ——, and Coauthors, 2009: Beaufort Gyre freshwater reservoir: a collapse of the Atlantic thermohaline circulation. Climatic State and variability from observations. J. Geophys. Res., 114, Change, 54, 251–267, doi:10.1023/A:1016168827653. C00A10, doi:10.1029/2008JC005104. Yang, J., 2006: The seasonal variability of the Arctic Ocean Ekman Rabe, B., and Coauthors, 2011: An assessment of Arctic Ocean transport and its role in the mixed layer heat and salt fluxes. freshwater content changes from the 1990s to the 2006–2008 pe- J. Climate, 19, 5366–5387, doi:10.1175/JCLI3892.1. riod. Deep-Sea Res. I, 58, 173–185, doi:10.1016/j.dsr.2010.12.002. ——, 2009: Seasonal and interannual variability of downwelling in Schmidtko, S., G. C. Johnson, and J. M. Lyman, 2013: MIMOC: A the . J. Geophys. Res., 114, C00A14, doi:10.1029/ global monthly isopycnal upper-ocean climatology with mixed 2008JC005084. layers. J. Geophys. Res., 118, 1658–1672, doi:10.1002/jgrc.20122. Zhang, J., R. Lindsay, A. Schweiger, and I. Rigor, 2012: Recent Serreze, M. C., and A. P. Barrett, 2011: Characteristics of the changes in the dynamic properties of declining Arctic sea ice: Beaufort Sea high. J. Climate, 24, 159–182, doi:10.1175/ A model study. Geophys. Res. Lett., 39, L20503, doi:10.1029/ 2010JCLI3636.1. 2012GL053545.

Unauthenticated | Downloaded 09/28/21 01:47 PM UTC