Ivanov and Watanabe, 2013

December 2013

TRAJECTORY OF THE ARCTIC

Ecological Applications, 23(8), 2013, pp. 1765–1777 Ó 2013 by the Ecological Society of America

Does Arctic sea ice reduction foster shelf–basin exchange? VLADIMIR IVANOV1,2,4

AND

EIJI WATANABE1,3

1

International Arctic Research Center, University of Alaska, Fairbanks, Alaska 99775 USA 2 Arctic and Antarctic Research Institute, St. Petersburg, Russia 3 Japan Agency for Marine-Earth Science and Technology, Kanagawa, Japan

Abstract. The recent shift in Arctic ice conditions from prevailing multi-year ice to firstyear ice will presumably intensify fall–winter sea ice freezing and the associated salt flux to the underlying water column. Here, we conduct a dual modeling study whose results suggest that the predicted catastrophic consequences for the global thermohaline circulation (THC), as a result of the disappearance of Arctic sea ice, may not necessarily occur. In a warmer climate, the substantial fraction of dense water feeding the Greenland–Scotland overflow may form on Arctic shelves and cascade to the deep basin, thus replenishing dense water, which currently forms through open ocean convection in the sub-Arctic seas. We have used a simplified model for estimating how increased ice production influences shelf–basin exchange associated with dense water cascading. We have carried out case studies in two regions of the Arctic Ocean where cascading was observed in the past. The baseline range of buoyancy-forcing derived from the columnar ice formation was calculated as part of a 30-year experiment of the panArctic coupled ice–ocean general circulation model (GCM). The GCM results indicate that mechanical sea ice divergence associated with lateral advection accounts for a significant part of the interannual variations in sea ice thermal production in the coastal polynya regions. This forcing was then rectified by taking into account sub-grid processes and used in a regional model with analytically prescribed bottom topography and vertical stratification in order to examine specific cascading conditions in the Pacific and Atlantic sectors of the Arctic Ocean. Our results demonstrate that the consequences of enhanced ice formation depend on geographical location and shelf–basin bathymetry. In the Pacific sector, strong density stratification in slope waters impedes noticeable deepening of shelf-origin water, even for the strongest forcing applied. In the Atlantic sector, a 1.53 increase of salt flux leads to a threefold increase of shelf-slope volume flux below the warm core of Atlantic water. This threefold increase would be a sufficient substitute for a similar amount of dense water that currently forms in the Greenland, Iceland, and Norwegian (GIN) seas but is expected to decrease in a warming climate. Key words: Arctic Ocean; Arctic sea ice; cascading; climate change; convection; dense water formation; ocean–air energy exchange; ocean circulation model; polynya; shelf–slope water exchange.

INTRODUCTION Record-low sea ice concentrations in the Arctic Ocean observed in the summers of 2007–2012 suggest that the reduction in Arctic sea ice is happening much faster than was predicted in the early 2000s (Stroeve et al. 2007, Wang et al. 2009). This depletion began in the 1970s (Rothrock et al. 1999, Stroeve at al. 2007), and during recent decades, mean ice thickness has decreased by 1.3 m, while summer ice extent has decreased by about 40% in the Arctic Ocean (Comiso et al. 2008). One of the expected consequences of ice melting is an increase in freshwater outflow from the Arctic Ocean to lower Manuscript received 13 June 2011; revised 24 September 2012; accepted 3 December 2012. Corresponding Editor: M. Sturm. For reprints of this Invited Feature, see footnote 1, p. 1743. 4 E-mail: [email protected]

latitudes. Freshening of the upper water layers in the Greenland, Iceland, and Norwegian (GIN) seas is predicted to cause substantial weakening of the global thermohaline circulation (THC) by cutting off its sinking sub-Arctic branch (Stocker and Wright 1991). In the present climate, this sinking branch is maintained mainly by dense water formation via deep convection in the GIN seas and to a lesser degree by dense water formation on the Arctic shelves (Rudels and Quadfasel 1991). Formation of dense water on the Arctic shelves results from sea ice freezing and subsequent brine ejection into the underlying water column. A larger amount of newly formed ice provides, in principle, a larger amount of dense water. This suggests that production of dense water on the Arctic shelves is less sensitive to varying climate conditions, since winter ice formation will not cease, even in a warmer climate. Rather, the transition of processes in the Arctic shelf

1765

1766

Ecological Applications Vol. 23, No. 8

INVITED FEATURE

seas to the sub-Arctic type, currently observed in the Barents Sea (Ivanov and Shapiro 2005), may enhance dense water formation. Intensified outflow of dense water from the Arctic shelves may compensate for weakening dense water production in sub-Arctic seas. Existing estimates (Anderson et al. 1999) suggest that, during the 1990s, about 30% of the dense water overflow across the Greenland–Scotland Ridge originated on the Arctic shelves. The ecological response to sea ice reduction is also an important issue. Phytoplankton productivity in the Arctic basin area is supported by continuous shelf-tobasin transport of nutrient and organic matters (Watanabe et al. 2012). Retreat of the summer ice edge enhances shelf–basin exchange via upwelling events (Carmack and Chapman 2003) and baroclinic eddy activities (Watanabe and Hasumi 2009). We hypothesize that a similar effect is produced by dense water cascading in winter. Given that climate predictions using general circulation models (GCMs) are assumed to produce a plausible future scenario (i.e., that in a few decades, the Arctic Ocean will be ice-free in summer), we can regard present conditions as a transition stage, characterized by gradual ice depletion. The changes in sea ice are substantial enough to prompt the overall question: how does observed sea ice reduction affect dense water formation on the Arctic shelves? To study this question and its specific implications, we have employed a dual modeling approach. We use a coupled sea-ice–ocean model, the Center for Climate System Research Ocean Component Model (COCO), version 3.4 (Hasumi 2006, Watanabe 2011), for estimating interannual variations in ice formation/ salt flux on cascading sites in the Pacific and Atlantic sectors of the Arctic Ocean. It should be remarked that some features of polynya-type ice formation are also reproduced, even by a relatively coarse resolution model, and that the thermodynamic/dynamic relationship of sea ice production with mechanical divergence of sea ice volume due to wind forcing and ocean current may be addressed. To assess the range of variation in shelf–basin heat–salt exchange associated with cascading, we use another ocean model, the Proudman Oceanographic Laboratory Coastal-Ocean Modeling System (POLCOMS) (Holt and James 2001). In Specific features of Arctic dense water cascading, existing knowledge on dense water formation and cascading in the Arctic Ocean is presented. In Modeling of dense water formation, ice formation/salt flux variability demonstrated in the 30-year experiment using the COCO model is presented. In Response of dense water cascading to variable buoyancy forcing, we apply the POLCOMS to carry out case studies on dense water cascading at two specific sites, which characterize typical conditions in the Pacific and Atlantic sectors of the Arctic Ocean. In Discussion and conclusions, we discuss model results in the context of presently observed and predicted changes in climate conditions in the Arctic Ocean and GIN Seas.

SPECIFIC FEATURES

OF

ARCTIC DENSE WATER CASCADING

In the Arctic Ocean, the physical mechanism producing dense water is linked with sea ice thermodynamics (Shapiro et al. 2003). If ice cover is persistent yearround, the temperature of the surface water is nearly constant and close to the freezing point. Heat fluxes from ocean to atmosphere do not result in any further cooling of water. The dependence of density change on water temperature in the surface layer is very small, first due to limited temperature variability and second due to low values of the thermal expansion coefficient of sea water at low temperatures. Heat fluxes result in new ice formation and release of brine into the water; the salinity of new ice is only about one-third of the salinity of sea water (Maykut 1986), hence producing significant brine rejection. The newly formed dense water begins to play a significant role in climate regimes only if it is transported off the productive shelves. Dense water transport off the shelves occurs via a bottom-trapped gravity current and baroclinic eddies (termed ‘‘cascading’’). Cascading is a specific buoyancy-driven motion, in which dense water descends down the continental slope to a greater depth (Ivanov et al. 2004). The Arctic shelf water potentially attains density excess over its deeper counterpart because of the spatial irregularity in the ice growth rate, which depends primarily on ice concentration and thickness (Holland et al. 1999, Maqueda et al. 2004). Under similar meteorological conditions, heat exchange through the ice-free surface and thin young ice is about two orders of magnitude higher than that through the surrounding pack ice (e.g., Smith et al. 1990). As a result, extremely dense water forms during winter inside flaw polynyas, wind-induced openings in the ice cover that occasionally separate fast and drifting ice in coastal areas (polynya-type of dense water formation). In the seasonally ice-covered seas (e.g., the Barents Sea), dense water forms in the winter in marginal ice zones (MIZ) and under rapidly growing young ice (MIZ-type of dense water formation). This type of dense water formation is quite efficient, because there is no obstacle for the ice floes, which drift from the shelf to the warm deeper water and melt there, additionally increasing the horizontal density contrast and associated pressure gradient (Ivanov and Shapiro 2005). Both types of dense water formation are found in the marginal Arctic seas. Ivanov et al. (2004: Fig. 32) identified 10 sites in the Arctic Ocean where dense water cascading was occasionally observed. Spatial distribution of these sites is essentially non-uniform, with five regions located in the Barents Sea in the seasonal ice cover zone. These primarily represent the locations where the MIZ-type of dense water formation takes place. Other sites are concentrated around the Severnaya Zemlya Archipelago and in the eastern Chukchi Sea. These cascading sites are associated with the polynya-type of dense water formation (Winsor and Bjork 2000, Dethleff 2010).

December 2013


Kwok et al. (2009) reported that between 2003 and 2008, seasonal ice became the dominant ice type in the Arctic Ocean. The area covered by multiyear ice (just over 50% in winter 2003) decreased to 34% in winter 2008. The area covered by seasonal ice (first-year ice; sea ice of no more than one winter’s growth) increased accordingly to 64%. Another important conclusion reached by that study is that the mean thickness of the first-year ice did not change within the standard deviation. These two points suggest general intensification of seasonal ice formation over the areas earlier occupied by multiyear ice. First-year ice reaches maximum thickness by the end of winter. In late fall and early winter, this newly forming ice is thinner than multiyear ice would be in a similar location. This results in substantially higher accretion rate under developing first-year ice than under multiyear ice (Maykut 1986). A high accretion rate leads to intensive brine ejection and large salt input in the under-ice water column, favoring dense water formation (Holland et al. 1999). In order to examine the sensitivity of Arctic dense water cascades to the increasing seasonality of the sea ice cover, we selected two cascading sites, one each in the Pacific and Atlantic sectors of the Arctic Ocean (Fig. 1). We chose these regions for the following reasons. First, both are located inside the boundary of the 1979–2000 median winter ice extent. We could be sure, therefore, that cascading events documented here in the past developed as polynya-type events. As a result of Arctic sea ice cover shrinking, ice conditions shifted toward more pronounced seasonality with longer periods of low, or even absent, ice concentration (data available online).5 The decline of summer ice concentration prompts intensive ice formation during the next winter. Second, each of these regions is characterized by specific background conditions, including bottom bathymetry (Fig. 1) and vertical stratification (Fig. 1b). In the Alaskan coastal region of the Chukchi Sea (hereafter referred to as AC), dense water forms over a broad, shallow shelf (30–40 m in depth and more than 120 km in mean width) between Cape Lisburne and Point Barrow (Ivanov et al. 2004: Fig. 27). A vertical temperature profile in the adjacent deep water shows a minimum at 100 m depth and a maximum at 450–500 m, representing Pacific winter water and the Atlantic Water core, respectively. Salinity distribution is generally characterized by a thick halocline layer from the bottom of the seasonal mixed layer down to a depth of 250 m. Cascading events down to 180 m depth were documented around the AC region in earlier studies (Weingartner et al. 1998, Ivanov et al. 2004). Modeling of dense water formation in the AC region demonstrated that under typical ‘‘pre-warming’’ forcing (1978–1998), maximum salinity on the AC shelf reached 33.8 psu (Winsor and Chapman 2002: Fig. 8, which can induce 5

http://www.nsidc.org/arcticseaicenews/

1767

cascading down to a depth of 160–180 m. Formation of dense water on the northwestern Severnaya Zemlya shelf (hereafter referred to as SZ) occurs on shallow banks (Ivanov et al. 2004: Fig. 23). The narrow (40–50 km width) shelf ends with a steep continental slope, with a mean inclination of about 0.03 between 200 and 1000 m depth contours. The vertical temperature profile is typical for the Atlantic sector of the Nansen Basin with a temperature maximum at 300–400 m, representing the intermediate Atlantic water. The temperature minimum at 50 m shows the depth of winter convection. The characteristic feature of the salinity profile is a strong vertical gradient in the upper 50-m layer (summer feature), which transforms into a moderate permanent gradient in the thermocline. Cascading of dense water down to the 400 m depth in the SZ region was captured by observations and simulated in the model (Ivanov and Golovin 2007). MODELING

OF

DENSE WATER FORMATION

To estimate the baseline range of buoyancy removed by columnar ice growth (columnar ice is the ice that grows under existing ice cover) in the AC and SZ regions, we employed a pan-Arctic Ocean model (COCO). COCO has reproduced reasonable performances for Arctic basin-scale sea ice variation and freshwater dynamics (Watanabe and Hasumi 2005, Proshutinsky et al. 2011). The 30-year model integration was designed to study interannual variations in ice formation/salt flux at selected cascading sites. Although the spatial resolution of about 25 km is insufficient to resolve polynyas and flaw leads, which are the major producers of brine (e.g., Winsor and Bjork 2000), the model is capable of providing information on general circulation patterns, seasonal and interannual variations in thermohaline structure, and ice conditions in the target regions. Description of numerical experiments The model and experimental design of the basin-scale numerical experiment are described below. The sea ice component adopts the zero-layer thermodynamic formulation of Semtner (1976) and an elastic-viscousplastic rheology (Hunke and Dukowicz 1997). The ocean component is a free-surface ocean general circulation model (OGCM). The model domain contains the entire Arctic Ocean, the GIN Seas, and the northern part of the North Atlantic. The bathymetry is constructed from the merged product of the International Bathymetric Chart of the Arctic Ocean (IBCAO) and the Earth Topography Five-Minute Gridded Elevation Data Set (ETOPO5; available online).6 The horizontal resolution is about 25 km, and there are 28 hybrid sigma-z vertical levels. The sigma coordinate composed of three levels is applied in the upper 10 m. The vertical 6

http://www.whoi.edu/page.do?pid¼30630

1768

INVITED FEATURE

grid width varies from 2 m at the top level to 500 m at the bottom level. The atmospheric forcing components are constructed from the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) daily reanalysis data set (Kalnay et al. 1996). The Pacific water inflow with seasonal cycle is provided at the Bering Strait. The annual mean inflow and salinity of Pacific water at the Bering Strait are set to be 0.8 Sv (1 Sv [ 106 m3/s) and 31 psu, respectively. The experiment is initiated from the ice-free ocean at rest and temperature, salinity fields of the Polar Science Center Hydrographic Climatology (PHC; Steele et al. 2001). The spin-up is run for ten years under the atmospheric forcing of 1979. The main experiment is then performed using daily forcing from 1979 to 2008. The simulated sea ice and ocean properties capture well-known features. The total sea ice in the Arctic Ocean gradually decreases after the 1990s and reaches the record minimum in September 2007 (Wang et al. 2009). The sea ice margin retreats in the Chukchi and Siberian shelves during summer, as detected by standard products derived from passive microwave measurements, including the Special Sensor Microwave/Imager (SSM/I) and the Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E; Cavalieri et al. 1996, Kwok 2008). Winter sea ice covers most of the Chukchi Sea, and a thinner ice area emerges along the northwestern coast of Alaska, where a wind-driven coastal polynya is frequently formed (Martin et al. 2004). In the Severnaya Zemlya region, the coastal polynya location varies around the island depending on surface wind direction during winter. The model represents major basin-scale geostrophic currents, such as the Beaufort Gyre and the Transpolar Drift, and the spatial maximum of freshwater content references a salinity of 34.8 psu (Proshutinsky et al. 2005) throughout the integration period. The remarkable surface freshening in the Canada Basin in the late 2000s (Proshutinsky et al. 2009, Yamamoto-Kawai et al. 2009) is also reproduced. Interannual variations in sea ice production and mean salinity In this section, we analyze sea ice thermal production, associated salt flux, and mean salinity in the AC and SZ regions simulated by COCO. Yearly values are defined by the total from September in each corresponding year through August of the next year. Annual sea ice production is calculated as the sum during the freezing period, when the monthly mean of net sea ice production shows positive (cf. negative values indicate sea ice melting). The volume change arising from lateral sea ice advection is not included in the production values. Salt flux due to sea ice formation was calculated following formula 1 of Dethleff (2010). Sea ice salinity was fixed to 4 psu in this experiment.


Alaskan coastal region.—We define the AC dense water production region as the area between Cape Lisburne and Point Barrow (Fig. 1), where the depth is shallower than 30 m. Annual sea ice production averaged in the AC region is presented in Fig. 2a. The long-term mean from 1979–2007 is 4.25 m, and the standard deviation of annual anomalies is 0.79 m. The total volume, which is obtained by multiplying the coastal area of 23 363 km2 by the area mean production, covers from 70 to 146 km3. This amount broadly overlaps with satellite-based estimates (Martin et al. 2004, Tamura and Ohshima 2011). Winter sea ice production shows interannual variation with no significant long-term trend. An annual anomaly greater (or lower) than one standard deviation is recorded in 1982, 2000, and 2003 (1980, 1994, and 2005). The positive (or negative) peak of these anomalies in 2000 (1994) is also consistent with Martin et al. (2004). The earliest month of sea ice freezing varies between September (6 years), October (22 years), and November (1 year; Fig. 3a). The latest month of freezing is April in all years, so this seems to be independent of the magnitude of annual production. The freezing period is seven months on average. The seasonal maximum is recorded in December and January in most years, but ranges from November to February. Interannual variation in annual salt ejection ranges from 1.68 to 3.76 3 1012 kg, which can be converted to salt flux at the ocean surface of 3.47–8.86 3 103 gm2s1 in the AC region. The potential increase in area mean salinity, which is defined as the annual salt ejection divided by the entire water volume in the AC region, is 2.41–5.54 psu. This range is somewhat larger than the climatology provided by Dethleff (2010), probably because the exact target area differs. Dethleff (2010) focuses on flaw/lead and polynya regions, while our estimate includes the columnar sea ice formation from first-year ice. The simulated seasonal salinity change during the freezing period averaged in the AC region is shown in Fig. 2e. The long-term mean is 2.22 psu and the standard deviation is 0.52 psu. As shown in the sea ice production, there are significant interannual variations in the salinity change, with no significant long-term trend. The positive (or negative) peaks of annual anomalies in 2000 and 2003 (2005) are in phase with sea ice production. The correlation coefficient between sea ice production and salinity change is relatively high (0.64), despite a few discrepant cases (e.g., 1994). On the other hand, the scatter plot of potential and simulated salinity increase reveals no simple proportional relationship between these values (Fig. 4). Actually, salinity in a target region is determined by lateral water-mass exchange, with the surrounding areas in addition to surface salt flux. In particular, the formed dense water is preferably transported to deep basins without remaining on the coastal shelf, as discussed in AC and SZ comparison.

December 2013


1769

FIG. 1. (a) Bathymetry of the Arctic Ocean (Jakobsson et al. 2000). Location of the Alaskan Coastal (AC) and Severnaya Zemlya (SZ) regions is shown by red polygons. (b) Vertical distribution of spatially averaged temperature (8C) and salinity in September in the AC and SZ regions (calculated from Steele et al. 2001).

Severnaya Zemlya Region.—We define the SZ dense water production region with a 100-m isobath northeast of the Severnaya Zemlya Archipelago (Fig. 1) because the SZ shelf slope is steeper than the AC region. The annual sea ice production of 3.60 6 0.60 m averaged in the SZ region is smaller than in the AC region and has no significant decadal trend (Fig. 2b). The total volume, which is obtained by multiplying the coastal area of 9927

km2 by the area’s mean production, ranges from 30 to 42 km3. Annual anomalies of greater (or lower) than one standard deviation appear in 1983, 1988, 1994, and 2004 (1985, 1992, 1997, 1998, and 2003). The earliest month of sea ice freezing is September for 27 years and October for 2 years (Fig. 3b). The latest month is April for 28 years and May for one year. The average freezing period

1770

INVITED FEATURE


FIG. 2. Interannual variations in (a, b) thermal sea ice production, (c, d) net mechanical divergence of sea ice volume converted to equivalent changes in mean ice thickness, and (e, f) actual salinity change during the freezing period averaged in (a, c, e) AC and (b, d, f) SZ regions. See Modeling of dense water formation: Interannual variations in sea ice production and mean salinity for exact definition of area averages. Solid and dashed gray lines show 30-year mean and standard deviations, respectively.

is about eight months. The seasonal maximum is widely recorded from October to March. The seasonal change of area mean salinity during the freezing period is shown in Fig. 2f. The total amount of 1.20 6 0.37 psu is smaller than in the AC region. The annual anomaly exceeds (or falls below) the standard deviation in 1981, 1993, 2001, and 2004 (1984, 1985, 1995, 1999, 2000, and 2005). The salt rejection of 0.62– 1.25 3 1012 kg is close to the estimate by Dethleff (2010) in the same region. The converted salt flux of 2.99–6.07 3 103 gm2s1 is consistent with the salt flux due to columnar ice growth in Ivanov and Golovin (2007). It should be noted that the correlation coefficient between sea ice production and salinity change is just 0.29, although both the potential and simulated salinity increases are within 0.5–2.5 psu (Fig. 4). AC and SZ comparison.—According to model results, sea ice production in both coastal regions is closely related to the mechanical divergence/convergence of sea

ice during the freezing period. The correlation coefficient between net divergence and thermal production of ice is 0.70 and 0.72 in the AC and SZ regions, respectively. The net divergence of ice volume is converted to an equivalent change in mean ice thickness in the same region (Fig. 2c, d). This shows positive values in all years with marked expansion in the open water area, except for 1997 and 2003 in the SZ region. The longterm mean of net ice divergence is higher in the AC region (4.76 6 2.00 m) than in the SZ region (3.03 6 2.13 m). The positive/negative peaks of net ice divergence in 1982, 2000, and 2003 (AC region) and 1994, 1997, and 2003 (SZ region) are in phase with those of sea ice production. This coherence points to sea ice divergence as accounting for a significant part of the interannual variations in sea ice production. The magnitude of sea ice divergence/convergence is in turn controlled by surface wind forcing, ocean currents, and sea ice internal stress.

December 2013


1771

FIG. 3. Interannual variations in ice-freezing period in (a) AC and (b) SZ regions. Maximum ice production months in each year are shown by crosses.

Generally, accumulation of salt, which is ejected during ice formation, may not continue throughout the entire freezing season. At some point in time, salty/dense water starts moving/cascading under the influence of gravity, as described in Specific features of arctic dense water cascading. Cascading is controlled by multiple external factors, including bottom topography, background stratification, wind, and currents (Shapiro et al. 2003). In the strongly stratified Chukchi Sea, with a gentle bottom slope, dense water is expected to be transported by large scale currents rather than downward cascading. Strong barotropic currents following major features of bottom topography, including the Alaskan Coastal Current, would transport dense water toward the northern edge of the Chukchi Sea continental slope (Woodgate and Aagaard 2005). Moreover, the winter mean atmospheric regime shows a dominant easterly wind accompanied by intensified high sea level pressure in the Beaufort Sea, which can push coastal water offshoreward. In the SZ region, moderate stratification over steep slopes allows dense water to cascade more deeply toward the basin interior (Ivanov and Golovin 2007). These ocean transports prevent local salt accumulation from occurring, so that the actual

FIG. 4. Scatter plot of potential and simulated salinity increase during the freezing period of each year in the AC and SZ regions.

1772


INVITED FEATURE

salinity change is expectedly smaller than the potential increase due to ice production. RESPONSE OF DENSE WATER CASCADING TO VARIABLE BUOYANCY FORCING In this section, we examine how parameters of dense water cascading in the AC and SZ regions would change under different external forcing (salt flux at the ocean surface). We use the numerical model POLCOMS to carry out these case studies (model available online).7 A detailed model description was given in earlier publications (cf. Holt and James 2001). In the present experiments, boundary conditions at the free surface include only salt flux associated with ice freezing. On the side walls and bottom of the model domain, the normal scalar gradients were set to zero. The bottom boundary condition for turbulent energy reflects a steady-state balance between shear and dissipation. At open boundaries, the model uses flux/radiation boundary condition and the depth mean normal velocities for the barotropic component. Scalar variables in the boundary points are advected from interior points in case of outflow and from zero normal gradient in case of inflow. Experimental setup The model domain in both target regions (AC and SZ) is a ;242 3 242 km trapezium-shaped area representing a fragment of shelf and slope. A circular flat bank with a depth of 50 m and a diameter of 40 km is located in the center of the grid area. The bank ends on a linear slope that reaches the flat seabed in the deep ocean basin. The difference in seabed topography between AC and SZ is featured in a slope inclination (0.003 in the AC region and 0.03 in the SZ region). The spatial grid width along longitude is 2 km, which resolves the baroclinic Rossby radius for Arctic Ocean shelves deeper than 50 m (Aksenov et al. 2011). The spatial step along latitude is the same at the southern edge and decreases with increased latitude to 1.85 km at the northern edge. There are 36 terrain-following levels (so-called s-levels) in the vertical. The distance between s-levels varies from 1.4 m in the shallowest layer to about 100 m in the deepest part of the model domain. The maximum distance between slevels inside surface and bottom boundary layers is 2.5 m. The barotropic and baroclinic time steps are 4 and 480 seconds, respectively. The horizontal viscosity coefficient in the Laplacian mixing term equals 20 m2/ s, following Winsor and Chapman (2002). Constant salt flux at the free surface induces a salinity/buoyancy change in the uppermost grid cell. The water column is then mixed down to the convection depth. We prescribed initial temperature and salinity profiles in a piecewise linear shape, which approximated mean September profiles, presented in Fig. 1b. Initial thermohaline properties are horizontally uniform. The upper 7 http://cobs.pol.ac.uk/modl/metfcst/POLCOMS_ DOCUMENTATION/node2.html

50-m layer was assigned vertical homogeneity, with temperature at the freezing point and salinity equal to mean salinity in the 50-m water column mixed by thermohaline convection (31.91 psu for the AC region, 34.20 psu for SZ). Permanent salt flux was set over the bank area throughout the duration of each experiment (50 days). We used the range of salt flux obtained in COCO model experiments to estimate the range of actual salt flux at the surface, which includes an additional component for ice formation in polynyas. As was shown by Ivanov and Golovin (2007), the relative contribution of polynya events to the total salinity increase on the northwestern Laptev Sea shelf is about 70%, while 30% is attributed to columnar ice growth. Applying this proportion to the mean range of salt flux variation for both regions from COCO model results, 3.23–7.47 3 103 gm2s1, we get the range of total salt flux (after rounding): 0.01–0.03 gm2s1. In POLCOMS experiments, we used this range for salt flux, prescribing a constant value (Chapman 1999) within this range in each individual experiment. The duration of each experiment was 50 days. Longer runs did not principally change the results described in the next subsection. Case studies Evolution of density (buoyancy) and velocity fields in the course of numerical experiments is generally in line with multiple model studies of dense water cascading (e.g., Gawarkiewitz and Chapman 1995, Jiang and Garwood 1998, Ivanov and Golovin 2007). After initial salt accumulation in the forcing region and geostrophic adjustment, dense water moves down and across the slope, primarily in the shape of baroclinic eddies. Compensating upwelling transports deep water upslope (Fig. 5). Within some time interval (;25–30 days), lateral salt flux comes into dynamic equilibrium with applied salt flux at the surface, and despite continuous salt input, salinity in the source region stabilizes at some ‘‘saturation’’ level. This level depends on the magnitude of surface buoyancy forcing and steepness of the slope (Shapiro et al. 2003). Heat and salt fluxes associated with cascading/upwelling circulation determine the magnitude of the impact of cascading on water masses in the deep ocean. We calculated these fluxes in accordance with Ivanov and Golovin (2007: formula 2). Results of these calculations for AC and SZ are presented in the next two subsections. Alaskan coastal region.—Strong salinity stratification in the ambient water is the major factor preventing deep descent of shelf-origin dense water in the AC region. Maximum salinity increase on the shelf is high and varies between 0.5 and 1 psu for the given forcing range. However, this increase makes the shelf water salinity equal to that in the ambient water at 100–130 m. The contribution of lower temperature water on the shelf to overall density is rather small, thus constraining the cascading depth to 125 m. The flat bottom is less

December 2013


1773

FIG. 5. Calculated POLCOMS model mean vertical sections of (a, c) temperature (8C) and (b, d) salinity in (a, b) AC and (c, d) SZ regions for surface forcing of 0.03 gm2s1. Values were obtained using spatial averaging over equal distances around the bank and time averaging over the last 25 days of calculation.

restrictive. The theoretical estimation of Nof’s speed (Ivanov et al. 2004) for slope inclination 0.003 and 1 psu salinity contrast between shelf and slope water at 718N is 65 cm/s, or about 13 cm/s of maximum cross-slope velocity. This value is quite large, and it provides fast descent of dense water down slope in the case of neutral stratification in the ambient water. These features of cascading in the AC region are visible at mean crossslope sections (see Fig. 5a, b) and are reflected in shelf–

slope heat/salt fluxes (Fig. 6a, b). Even for the weakest applied forcing (0.01 gm2s1), shelf–slope salt flux is high (up to 0.8 gm2s1). However, this flux is limited by the upper 50 m (i.e., almost horizontal). The corresponding heat flux is negative and small (up to 1.0 3 104 MJm2s1). Increased forcing from 0.01 to 0.02 gm2s1 increases the depth of dense water descent from 100 to 125 m. The head of the cascade reaches saltier water, and the salt flux turns negative in the lower

1774


INVITED FEATURE

50 m. Since deep water is also warmer than shelf water, heat flux increases by an order of magnitude. A further increase in surface forcing leads to a nonlinear response in fluxes. This happens due to initiation of upwelling of water from the deep. Compensating upwelling is an inseparable feature that accompanies cascading (Ivanov and Golovin 2007). In the Arctic Ocean, where temperature increases with depth in the upper 200–300 m, the effect of upwelling is particularly significant, allowing warming and salinization of the upper waters in the course of cascading. Superposition of cascading-induced negative fluxes with upwelling-induced positive fluxes results in a decrease in total flux in the deepest 50-m bin, despite increased forcing at the surface. In general, three-times variation in forcing does not radically affect cascading parameters in the AC region. In the upper 50 m, salt fluxes increase roughly linearly with increased forcing. Ventilation of the upper 125 m intensifies due to cascading-upwelling circulation. However, an increase in forcing does not substantially affect the depth of dense water penetration, which changes by only 20–30 m as a result of the damping effect of the ambient stratification. Severnaya Zemlya region.—Earlier studies of cascading in the SZ region (Ivanov and Golovin 2007) revealed that the maximum depth of dense water penetration is ;400 m. This finding was obtained for surface forcing of 0.017 gm2s1, which is slightly less than that used in the present model. The new results show that variation in surface salt flux leads to substantial changes in cascading penetration. The maximum depth of dense water penetration increases from 300 m (for forcing of 0.01 gm2s1) to the base of the slope at 500 m (for 0.03 gm2s1 forcing; Fig. 6c, d). In the real ocean, where the base of the slope is at ;3500 m, 500 m is likely shallower than the actual termination depth. Similar to the AC region, vertical distribution of salt fluxes (Fig. 6d) demonstrates the change of sign. In the upper layers, salt flux is positive, but starting from the depth at which dense water reaches a saltier environment, it reverses to the negative. This depth increases with increased surface forcing from ;200 to ;300 m. As a result, the net effect of dense water cascading on deep water masses is freshening. In the shallow layers (0–200 m), positive salt flux causes water salinization. The nonlinear increase of this salinization for the strongest forcing is also explained by the intensified upwelling of warm and salty water from the deep. Heat fluxes are negative throughout the water column, except for the uppermost layer. Absolute values of salt fluxes in the SZ region are several times less than in the AC region for similar surface forcing. This is due to the fact that in the SZ region, incoming salt is redistributed within a larger depth range due to the higher average salinity of the ambient water. On the contrary, the average heat flux is substantially larger in the SZ region (;102 MJm2s1). This is also linked with dense water descent down to the level of warmer ambient water.

AC and SZ comparison.—The foregoing case studies have elucidated specific features of dense water formation and associated cascading in two geographical locations with different ice and bathymetry conditions, while under similar external forcing. In the AC region, relatively fresh shallow water is not able to acquire a sufficient amount of salt and to become dense enough for deep cascading through a stratified environment. Maximum depth of dense water penetration is limited to 125 m, which is several hundred meters shallower than the warm core of Atlantic water in this region. Increased forcing makes this surface layer saltier and denser but does not change the overall circulation pattern. In the SZ region, an increase in surface salt flux significantly changes cascading parameters. For weak and moderate forcing, which characterizes pre-warming ice conditions, dense water reaches the core of the Atlantic water. However, under increased salt forcing, dense water passes through this Atlantic water, efficiently ventilating the water mass and underlying deep waters. In addition, intensive upwelling brings the deep water onto the shelf, contributing to effective shelf–slope exchange of properties. DISCUSSION

AND

CONCLUSIONS

This study was motivated by recent dramatic changes in the Arctic sea ice cover, a transition from dominance of multi-year ice to dominance of first-year ice. This change is generally in line with the GCM-based prediction of a shift toward a seasonal Arctic ice cover. This transition to seasonal ice may affect hydrographic conditions in the Arctic Ocean on a decadal timescale. Seasonal ice grows fast, thus producing large amounts of brine and making the underlying water saltier and denser. There is observational evidence that this may already be the case in the Atlantic sector of the Arctic Ocean, where recent changes in the state of the Arctic sea ice cover were accompanied by an increase of salinity in the upper layer in the late winter season (McPhee et al. 2009). The excess fresh water from the seasonal ice melt is currently accumulating in the Beaufort Gyre (Morrison et al. 2012). It is anticipated that changes in the prevailing regime of large-scale atmospheric circulation over the Arctic Ocean may trigger a rapid release of huge amounts of fresh water from the Beaufort Gyre to the GIN seas with negative consequences for global THC (Mauritzen 2012). We hypothesize that a transition to seasonal ice cover may shift the major region of dense water formation from the GIN seas to the Arctic shelves, thus preventing cessation of THC due to excess freshening in the GIN Seas. In the present climate, the contribution of dense water originating on Arctic shelves to the THC is estimated to be about 30%. In a seasonally ice-covered Arctic Ocean, the rate of ice formation in winter is expected to increase proportionally to the extension of ice-free area in summer. Intensified sea ice freezing on Arctic shelves will lead to enhanced dense water

December 2013


1775

FIG. 6. (a, c) Heat and (b, d) salt fluxes across the bathymetry contours in the (a, b) AC and (c, d) SZ regions for different forcing (surface salt flux): green, 0.01 gm2s1; blue, 0.02 gm2s1; red, 0.03 gm2s1. Values were averaged over 50-m depth bins and over the last 25 days of calculation.

production. We have used a simplified model setting to estimate how increased ice production may influence the shelf–slope exchange associated with dense water overflows from the Arctic shelves. We have carried out case studies in two regions of the Arctic Ocean where cascading has been observed in the past. The range of forcing caused by increased ice formation was calculated in the 30-year run of the large-scale coupled ice–ocean GCM COCO. Statistical analyses of COCO results indicate that mechanical divergence of sea ice volume accounts for a significant part of the interannual variations in sea ice thermal production in these coastal regions. This forcing was then used in a regional model with analytically prescribed bottom topography and vertical stratification, typical for specific cascading sites in the Pacific and Atlantic sectors of the Arctic Ocean.

Our results demonstrate that enhanced ice formation may substantially intensify shelf–slope exchange. However, this intensification does not change the structure of lateral fluxes in the Pacific sector, where strong density stratification in slope waters impedes noticeable deepening of shelf-origin water, even for the strongest forcing applied. In the Atlantic sector, where salinity on shelves is high in comparison with adjacent slope waters, a threefold increase in surface forcing (which is physically realistic, according to COCO results) pushes shelf-origin dense water below the Atlantic water layer, ventilating all water masses en route. An increase in forcing from 0.02 to 0.03 gm2s1 leads to a threefold increase in shelf–slope volume flux, below the warm core of Atlantic water (from 0.04 to 0.12 Sv). Taking into account the aforementioned estimation of the fraction of water from

1776


INVITED FEATURE

Arctic shelves contributing to the THC, this threefold increase is sufficient to replace a similar amount of dense water currently forming in the GIN Seas. The nutrient-rich dense shelf water that interacted with sea-bottom sediments has been suggested as a source of biological activity in the lower and higher trophic levels in the basin interior (Nishino et al. 2005, Llinas 2009). Thus, the cascading depth is important to the polar marine ecosystem, as shelf-water intrusion into the deeper layer of the basin area hardly enforces primary phytoplankton production in the surface euphotic zone. The findings obtained in this study indicate that the primary productivity would become restricted according to the increase in cascading depth. This result looks rather encouraging as a preliminary estimation. It shows the principal possibility that Arctic cascades will maintain the THC if the predicted freshening in the GIN Seas occurs. However, we must note that our analysis used several simplifications, which may have affected the results. Warm water that moves up along the slope and reaches the surface in the production zone must be cooled down to the freezing point before new ice formation may occur. This may take some time, during which salt flux at the surface is on hold. In the case of intensive ocean-air heat exchange, this time is estimated on the order of hours, which is not very restrictive. On the other hand, warm water from the deep melts the ice, making it thinner and therefore more mobile. We have not considered background ocean currents here, which may move dense water out of the production zone before strong density contrast is built up. This may be particularly important if the shallow shelf, where dense water forms, is on the pathway of persistent and fast flow. In the SZ region, the core of boundary current is located off-shelf (Rudels et al. 2000). Therefore, this current may only turn dense water on the slope downstream from its undisturbed route. In the AC region, the Alaska Coastal Current (ACC) is likely an important means of dense water transportation, accelerating the transport of dense water toward the shelf edge (e.g., Weingartner et al. 1998). However, as follows from our analysis, dense water in the AC region cannot acquire sufficient density for deep descent, even in the absence of the ACC. Therefore, additional advection would not make things much different. Finally, we used very simplified bottom topography and did not consider specific features of the seabed, such as, for example, canyons. Model studies of dense water cascading in the presence of small-scale irregularities of bottom topography point out that these features normally facilitate shelf-water exchange (e.g., Kampf 2005). Therefore, using a regular bottom slope may not cause overestimation of shelf–slope exchange. To take into account all mentioned factors, application of a pan-Arctic coupled ice–ocean model with spatial resolution of ;2 km may be recommended for future studies of cascading-related shelf–slope exchange in a warming climate.

ACKNOWLEDGMENTS This study was carried out at IARC with financial support from NSF Research Grant 1249133, JAMSTEC, JAXA, and OSL. E. Watanabe appreciates the Grants-in-Aid for Scientific Research (S) of the Japan Society for the Promotion of Science (JSPS) JFY2010-2014, No. 22221003, ‘‘Catastrophic reduction of sea ice in the Arctic Ocean: its impact on the marine ecosystems in the polar region.’’ Parts of pan-Arctic Ocean modeling were executed using JAMSTEC Earth Simulator version 2. The authors are grateful to anonymous reviewers for their useful comments and suggestions. LITERATURE CITED Aksenov, Y., V. V. Ivanov, A. J. G. Nurser, S. Bacon, I. V. Polyakov, A. C. Coward, A. C. Naveira-Garabato, and A. Beszczynska-Moeller. 2011. The Arctic Circumpolar Boundary Current. Journal of Geophysical Research 116:C09017. Anderson, L. G., E. P. Jones, and B. Rudels. 1999. Ventilation of the Arctic Ocean estimated by a plume entrainment model constrained by CFCs. Journal of Geophysical Research 104 C6:13423–13429. Carmack, E., and D. C. Chapman. 2003. Wind-driven shelf/ basin exchange on an Arctic shelf: The joint roles of ice cover extent and shelf-break bathymetry. Geophysical Research Letters 30(14):1778. Cavalieri, D., C. Parkinson, P. Gloersen, and H. J. Zwally. 1996. Sea ice concentrations from Nimbus-7 SMMR and DMSP SSM/I passive microwave data. National Snow and Ice Data Center, Boulder, Colorado, USA. Chapman, D. C. 1999. Dense water formation beneath a timedependent coastal polynya. Journal of Physical Oceanography 29:807–820. Comiso, J. C., C. L. Parkinson, R. Gersten, and L. Stock. 2008. Accelerated decline in the Arctic sea ice cover. Geophysical Research Letters 35:L01703, 6. Dethleff, D. 2010. Linear model estimates of potential salt rejection and theoretical salinity increase in a standardized water column of recurrent Arctic flaw leads and polynyas. Cold Regions Science and Technology 61:82–89. Gawarkiewitz, G., and D. Chapman. 1995. A numerical study of dense water formation and transport on a shallow, sloping continental shelf. Journal of Geophysical Research 100(C3):4489–4507. Hasumi, H. 2006. CCSR Ocean Component Model (COCO) version 4.0. Center for Climate System Research Report. University of Tokyo, Tokyo, Japan. Holland, M. M., J. A. Curry, and J. L. Schramm. 1999. Modelling thermodynamics of sea ice thickness distribution. 2. Sea ice/ocean interactions. Journal of Geophysical Research 102(C10):23093–23107. Holt, J. T., and I. D. James. 2001. An s-coordinate density evolving model for the northwest European continental shelf 1, model description and density structure. Journal of Geophysical Research 106:14015–14034. Hunke, E. C., and J. K. Dukowicz. 1997. An elastic-viscousplastic model for sea ice dynamics. Journal of Physical Oceanography 27:1849–1867. Ivanov, V. V., and P. N. Golovin. 2007. Observations and modeling of dense water cascading from the Laptev Sea shelf. Journal of Geophysical Research 112:C09003. Ivanov, V. V., and G. I. Shapiro. 2005. Formation of dense water cascade in the marginal ice zone in the Barents Sea. Deep-Sea Research Part I: Oceanographic Research Papers 52:1699–1717. Ivanov, V. V., G. I. Shapiro, J. M. Huthnance, D. L. Aleynik, and P. N. Golovin. 2004. Dense water cascades around the World Ocean. Progress in Oceanography 60:47–98. Jakobsson, M., N. Z. Cherkis, J. Woodward, R. Macnab, and B. Coackley. 2000. New grid of Arctic bathymetry aids

December 2013


scientists and mapmakers. Eos, Transactions American Geophysical Union 81(9):89–96. Jiang, L., and R. W. Garwood. 1998. Effects of topographic steering and ambient stratification on overflows on continental slopes: a model study. Journal of Geophysical Research 103(C3):5459–5476. Kalnay, E., et al. 1996. The NCEP/NCAR 40-year reanalysis project. Bulletin of the American Meteorological Society 77:437–471. Kampf, J. 2005. Cascading-driven upwelling in submarine canyons in high latitudes. Journal of Geophysical Research 110:C02007. Kwok, R. 2008. Summer sea ice motion from the 18 GHz channel of AMSR-E and the exchange of sea ice between the Pacific and Atlantic sectors. Geophysical Research Letters 35:L03504. Kwok, R., C. F. Cunningham, M. Wesnahan, I. Rigor, H. J. Zwally, and D. Yi. 2009. Thinning and volume loss of the Arctic Ocean ice cover: 2003–2008. Journal of Geophysical Research 114:C07005. Llinas, L. 2009. Zooplankton inside an Arctic Ocean cold-core eddy: probable origin and fate. Deep-Sea Research Part II: Topical Studies in Oceanography 56:1290–1304. Maqueda, M. A., A. J. Willmott, and N. R. T. Biggs. 2004. Polynya dynamics: a review of observations and modeling. Reviews of Geophysics 42:RG1004. Martin, S., R. Drucker, R. Kwok, and B. Holt. 2004. Estimation of the thin ice thickness and heat flux for the Chukchi Sea Alaskan coast polynya from Special Sensor Microwave/Imager data. Journal of Geophysical Research 109:C10012. Mauritzen, C. 2012. Arctic freshwater. Nature Geoscience 5:162–164. Maykut, G. A. 1986. The surface heat and salt balance. Pages 395–463 in N. Untersteiner, editor. The geophysics if sea ice. Plenum, New York, New York, USA. McPhee, M. G., A. Proshutinsky, J. H. Morison, M. Steele, and M. B. Alkire. 2009. Rapid change in freshwater content of the Arctic Ocean. Geophysical Research Letters 36:L10602. Morrison, J., R. Kwok, C. Perralta-Ferris, M. Alkire, I. Rigor, and M. Steele. 2012. Changing Arctic Ocean freshwater pathways. Nature 481:66–70. Nishino, S., K. Shimada, and M. Itoh. 2005. Use of ammonium and other nitrogen tracers to investigate the spreading of shelf waters in the western Arctic halocline. Journal of Geophysical Research 110:C10005. Proshutinsky, A., et al. 2011. Recent advances in Arctic Ocean studies employing models from the Arctic Ocean Model Intercomparison Project. Oceanography 24:102–113. Proshutinsky, A., R. Krishfield, M.-L. Timmermans, J. Toole, E. Carmack, F. McLaughlin, W. J. Williams, S. Zimmermann, M. Itoh, and K. Shimada. 2009. Beaufort Gyre freshwater reservoir: State and variability from observations. Journal of Geophysical Research 114:C00A10. Proshutinsky, A., et al. 2005. Arctic Ocean study: synthesis of model results and observations. Eos, Transactions American Geophysical Union 86:378–381. Rothrock, D. A., Y. Yu, and G. A. Maykut. 1999. Thinning of the Arctic sea-ice cover. Geophysical Research Letters 26:3469–3472. Rudels, B., R. D. Muench, J. Gunn, U. Schauer, and H. J. Friedrich. 2000. Evolution of the Arctic Ocean boundary

1777

current north of the Siberian shelves. Journal of Marine Systems 25:77–99. Rudels, B., and D. Quadfasel. 1991. Convection and deep water formation in the Arctic Ocean–Greenland Sea system. Journal of Marine Systems 2:435–450. Semtner, A. J. 1976. A model for the thermodynamic growth of sea ice in numerical investigations of climate. Journal of Physical Oceanography 6:379–389. Shapiro, G. I., J. M. Huthnance, and V. V. Ivanov. 2003. Dense water cascading off the continental shelf. Journal of Geophysical Research 108(C12):3390. Smith, S. D., R. D. Muench, and C. H. Pease. 1990. Polynyas and leads: an overview of physical processes and environment. Journal of Geophysical Research 95(C6):9461–9479. Steele, M., R. Morley, and W. Ermold. 2001. PHC: a global ocean hydrography with a high-quality Arctic Ocean. Journal of Climate 14:2079–2087. Stocker, T. F., and D. G. Wright. 1991. Rapid transitions of the ocean’s deep circulation induced by changes in surface water fluxes. Nature 351:729–732. Stroeve, J., M. M. Holland, W. Meier, T. Scambos, and M. Serreze. 2007. Arctic sea ice decline: faster than forecast. Geophysical Research Letters 34:L09501. Tamura, T., and K. I. Ohshima. 2011. Mapping of sea ice production in the Arctic coastal polynyas. Geophysical Research Letters 116:C07030. Wang, J., J. Zhang, E. Watanabe, M. Ikeda, K. Mizobata, J. E. Walsh, X. Bai, and B. Wu. 2009. Is the dipole anomaly a major driver to record lows in Arctic summer sea ice extent? Geophysical Research Letters 36:L05706. Watanabe, E. 2011. Beaufort shelfbreak eddies and shelf-basin exchange of Pacific summer water in the western Arctic Ocean detected by satellite and modeling analyses. Journal of Geophysical Research 116:C08034. Watanabe, E., and H. Hasumi. 2005. Arctic sea ice response to wind stress variations. Journal of Geophysical Research 110:C11007. Watanabe, E., and H. Hasumi. 2009. Pacific water transport in the western Arctic Ocean simulated by an eddy-resolving coupled sea ice-ocean model. Journal of Physical Oceanography 39:2194–2211. Watanabe, E., M. J. Kishi, A. Ishida, and M. N. Aita. 2012. Western Arctic primary productivity regulated by shelf-break warm eddies. Journal of Oceanography 68:703–718. Weingartner, T., D. Cavalieri, K. Aagaard, and Y. Sasaki. 1998. Circulation, dense water formation and outflow on the northeast Chukchi shelf. Journal of Geophysical Research 103:7647–7661. Winsor, P., and G. Bjork. 2000. Polynya activity in the Arctic Ocean from 1958 to 1997. Journal of Geophysical Research 105(C4):8789–8803. Winsor, P., and D. C. Chapman. 2002. Distribution and interannual variability of dense water production from coastal polynyas on the Chukchi Shelf. Journal of Geophysical Research 107 C7:3079. Woodgate, R. A., and K. Aagaard. 2005. Revising the Bering Strait freshwater flux into the Arctic Ocean. Geophysical Research Letters 32:L02602. Yamamoto-Kawai, M., F. A. McLaughlin, E. C. Carmack, S. Nishino, K. Shimada, and N. Kurita. 2009. Surface freshening of the Canada Basin, 2003–2007: river runoff versus sea ice meltwater. Journal of Geophysical Research 114:C00A05.