SEARCH

SEARCH BY CITATION

Keywords:

  • dense water;
  • polynyas;
  • shelf processes;
  • eddies;
  • halocline layer;
  • Chukchi Sea;
  • Bering Strait;
  • Arctic Ocean

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Ocean Model Description
  5. 3. Ice Production and Buoyancy Fluxes
  6. 4. Dense Water Production
  7. 5. Halocline Water Formation
  8. 6. Discussion and Summary
  9. Acknowledgments
  10. References

[1] Dense water formation from coastal polynyas on the Chukchi Shelf is examined using a primitive-equation ocean model forced by surface buoyancy fluxes from a time-dependent polynya model for the winters of the 1978–1998 period. The model is forced by either meteorological observations or National Center for Environmental Prediction (NCEP) reanalysis data. During the 21-year period the surface forcing and the dense water production vary by a factor of 2. Using meteorological observations, most winters produce substantial amounts of water with a density anomaly of at least 1.2 kg m−3 (salinity increase of 1.5 psu), while using NCEP forcing, density anomalies typically reach 0.8 kg m−3 (1.0 psu) but are rarely as large as 1.2 kg m−3. With meteorological forcing, the production of dense water is fairly uniform throughout the entire period, whereas with NCEP forcing, nearly all of the water with density anomaly greater than 1.2 kg m−3 is formed during the early part of the modeled period (1978–1984), with the density anomaly rarely reaching 1 kg m−3 in the later part (1987 and onwards). Interannual variability is high with large differences between successive winters in both ice production and dense water production. Most of the observed variability can be explained by varying wind fields, with offshore winds creating polynyas between 30 and 75% of the time during November to April. Using a climatologically based mean initial salinity of 31.6 psu, we show that maximum salinities produced rarely exceed 33.5 psu. Furthermore, on the basis of moored observations in Bering Strait, we conclude that the interannual variability of the initial salinity is of the same magnitude as the interannual variability in dense water formation, and thus both are equally important in determining whether or not winter water is dense enough to contribute to the cold halocline layer of the Arctic Ocean. Winters with high fractions of offshore winds can produce anomalies up to 1.8 kg m−3 (2.2 psu). This, together with the varying initial salinity (density) fields, can achieve waters with maximum salinities up to 35.4 psu. Finally, we find that the derived ice volumes and dense water productions are highly sensitive to the forcing (meteorological vs. NCEP), and comparison with in situ observations is highly recommended.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Ocean Model Description
  5. 3. Ice Production and Buoyancy Fluxes
  6. 4. Dense Water Production
  7. 5. Halocline Water Formation
  8. 6. Discussion and Summary
  9. Acknowledgments
  10. References

[2] The formation and maintenance of the Arctic Ocean halocline have received considerable attention [e.g., Coachman and Barnes, 1962; Treshnikov and Baranov, 1976; Aagaard et al., 1981; Melling and Lewis, 1982; Steele et al., 1995; Rudels et al., 1996; Steele and Boyd, 1998]. The cold halocline, between depths of 50 and 200 m, acts as an effective barrier between the cold and relatively fresh, well-mixed surface layer and the warmer and saltier Atlantic layer. In the absence of a halocline, warm water could reach the surface and melt the ice cover from below. One principal halocline formation mechanism, first suggested by Nansen [1906] and then revisited by Aagaard et al. [1981, 1985], is ice formation and accompanying brine rejection over the shallow continental shelves, mainly from leads and polynyas, which create cold and salty waters that can find their way across the shelf edge and eventually renew both halocline and deep waters of the Arctic Ocean. Polynyas are particularly interesting because they potentially may produce very dense water, depending on the atmospheric conditions and the initial salinity of the underlying water. Aagaard et al. [1981] argued that most shelves are too broad or too stratified by river runoff to produce dense halocline water by ice growth alone. They concluded that the Chukchi Shelf and the area around Novaya Zemlya are the two most promising areas for producing dense shelf waters.

[3] Despite the sound logic and conceptual appeal of this mechanism, many supporting details have been slow to emerge, and some basic aspects remain unclear. In particular, estimates of the amount of dense shelf water needed to maintain the halocline are rather rough and uncertain. Aagaard et al. [1981] combined a simple volume estimate of the halocline layer with a residence time to estimate the need for an annual rate of formation of dense shelf water of 2.5 Sv with a salinity exceeding 34 psu (1 Sv ≡ 106 m3 s−1). Björk [1989, 1990] studied a one-dimensional box model of the vertical structure of the Arctic basins and found that a production of 1.2 Sv of dense shelf water was needed in order to obtain close agreement between modeled and observed T-S profiles. These steady state estimates are uncertain because they involve many assumptions with few supporting data and are also highly dependent on the residence time of Arctic waters that is relatively unknown.

[4] Of equal importance (and perhaps uncertainty) is the question of how much dense halocline water is actually produced on continental shelves. Direct estimates are nearly impossible at this time because there are surprisingly few observations of salty and cold waters that can be definitively attributed to water masses formed in coastal polynyas. Melling [1993] analyzed observations from the Beaufort Shelf and found that the ice growth on the shelf by itself was not strong enough to ventilate the halocline. He concluded that favorable initial conditions must prevail, where the fresh water layer is forced offshore and replaced by upwelled Atlantic Water. This then cools and thus raises the initial density before the onset of ice growth. Schauer [1995] presented data from hydrographic moorings just outside Storfjord at Svalbard, where she found relatively salty and cold waters flowing out of the fjord as a result of intense ice growth from polynyas inside the fjord. From moorings she further found a salinity increase of ∼0.8 psu during the 1991/1992 winter. Perhaps the most compelling examples were reported by Weingartner et al. [1998], who analyzed hydrographic measurements from moorings at several locations on the Chukchi Shelf and in Bering Strait from 1991 and 1992. They found outbreaks of cold and hypersaline water capable of ventilating the halocline as well as the deep Canada Basin, which were attributed to ice formation in polynyas. Between these events they also found evidence of upwelling of warm and salty waters.

[5] Several attempts have been made to estimate the total production of halocline waters over Arctic continental shelves. Martin and Cavalieri [1989] and Cavalieri and Martin [1994] used passive microwave satellite data to estimate polynya sizes and ice production rates from various coastal regions around the Arctic Ocean. Cavalieri and Martin [1994] found a possible contribution of dense water production rates of 0.7–1.2 Sv from coastal polynyas to the halocline layer. Winsor and Björk [2000] extended the work of Cavalieri and Martin [1994] over a longer time period (1958 to 1997) using a numerical polynya model forced by the National Center for Environmental Predictions (NCEP) reanalysis data to estimate polynya sizes and ice production rates. They found large interannual variability in dense water formed in coastal polynyas around the Arctic perimeter, and much smaller ice and salt productions (∼30%) than those found by Cavalieri and Martin [1994]. The most productive area found in the Arctic was the Chukchi Shelf, which Winsor and Björk [2000] argued was the only area with potential to form water dense enough to contribute to both the halocline and deeper layers. Winsor and Björk [2000] estimated a mean production of 0.2 Sv of halocline water from polynyas in the Arctic, and concluded that the cold and salty waters produced in polynyas are less likely to be the main process contributing to the halocline layer than previously thought. Using a different approach, Goldner [1999] constructed an inverse box model of the Arctic Ocean in which observed mass, heat, salt, and δ18O are conserved within measurement uncertainties. He found that the requirements of steady state are best satisfied when 0.2 ± 0.1 Sv of dense water flows from the shelves into the Canada Basin, about the same as the Winsor and Björk [2000] estimate for the entire Arctic Ocean.

[6] One of the most obvious deficiencies in all of the above estimates of dense water production on Arctic continental shelves is the absence of a dynamical ocean model. Recent studies have begun to examine the dynamics of the ocean response to coastal polynyas. For example, Gawarkiewicz and Chapman [1995], using a three-dimensional primitive equation model forced by a constant negative surface buoyancy flux in a limited region at the surface, found that the dominant process transporting dense water across the shelf is small-scale (15–30 km) eddies formed by baroclinic instability around the edge of the polynya. Chapman and Gawarkiewicz [1997] showed that the production of dense water and the maximum density anomaly possible are limited by the offshore eddy flux, and they provided simple algebraic formulas relating the dense water production to polynya geometry and ice formation rates. Chapman [1999] analyzed the ocean response to a time-dependent polynya using periodically varying air temperature and wind speed, and found that the density anomaly and volume flux produced are nearly independent of polynya width and atmospheric temperature but strongly dependent on offshore wind speed. Gawarkiewicz [2000] showed that the eddies can transport dense water off the shelf and onto the continental slope and thus can ventilate deeper layers.

[7] Consideration of these ocean processes should improve our understanding of dense water production on Arctic shelves. Toward this end, we have constructed a model of dense water production on the Chukchi Shelf, between Cape Lisburne and Pt. Barrow (see Figure 1). We chose this region because, as noted above, previous studies have identified the Chukchi Shelf as one of the most productive areas for cold and saline shelf waters because of the frequent polynyas [e.g., Aagaard et al., 1981; Wallace et al., 1987; Björk, 1989; Cavalieri and Martin, 1994; Jones et al., 1995]. Using a time-dependent polynya model to force a three-dimensional ocean model, we compute the ocean response to the highly intermittent and varying surface buoyancy forcing each winter over a 21-year period from 1978 to 1998. Our main objectives are twofold. First, we analyze the volume distributions of water properties produced and their relation to maintenance of the halocline layer and deep-water renewal of the Arctic Ocean. Second, we investigate the interannual variability in both forcing and dense water production. In addition, we examine the sensitivity of our results to the meteorological inputs by comparing the results using two different forcing data sets.

image

Figure 1. Map of the Chukchi Shelf area between Cape Lisburne and Pt. Barrow. Also shown are the mooring site in Bering Strait (dot marked M), the NCEP grid point offshore of Barrow and within Barrow Canyon (dot marked N), and the offshore mooring site (dot marked HS).

Download figure to PowerPoint

[8] The ocean model is described in section 2. Forcing for the ocean model is provided by a polynya model following Winsor and Björk [2000], and the typical behavior of the polynya forcing, i.e., ice production and the resulting cross-shelf variations in surface buoyancy flux, are shown in section 3. The ocean response to the computed surface buoyancy flux is then presented in section 4, including interannual variability and the volume distributions of dense water. In section 5, we combine our results with the (varying) initial density (salinity) field to examine possible halocline water formation. Finally, results and implications are discussed and summarized in section 6.

2. Ocean Model Description

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Ocean Model Description
  5. 3. Ice Production and Buoyancy Fluxes
  6. 4. Dense Water Production
  7. 5. Halocline Water Formation
  8. 6. Discussion and Summary
  9. Acknowledgments
  10. References

[9] The ocean model is the s-coordinate primitive-equation model (SPEM5.1), developed by D. Haidvogel's group at Rutgers University, which solves the standard hydrostatic and Boussinesq momentum, density and continuity equations. This model is similar to the original SPEM [Haidvogel et al., 1991], except that it uses finite differences in the vertical and a generalized stretched vertical coordinate [Song and Haidvogel, 1994]. We apply the model in nearly the same way as Chapman and Gawarkiewicz [1997] and Chapman [1999], so the details are not repeated here. The model domain is a highly simplified representation of the Chukchi Shelf; a straight channel with periodic boundaries at the open ends and constant depth of 40 m. The offshore wall is placed 150 km away from the coast where it does not influence dense water production or transport. In order to resolve the small scales of the ocean response, the grid spacing is uniform at 2 km in each horizontal direction and 2 m in the vertical. Given the length and number of our calculations (180 days for each of 21 winters), we use a shortened channel length of 100 km and then multiply the dense water volumes by 4.5 to obtain estimates for the entire 450 km coastline between Cape Lisburne and Pt. Barrow.

[10] Standard dynamical assumptions are made: rigid lid, no flow or density flux through solid boundaries, linearized bottom stress with a bottom friction coefficient of 10−4 m s−1, the Mellor-Yamada level 2 turbulence closure scheme to estimate the vertical mixing coefficients (constrained between 10−5 and 10−3 m2 s−1), convective adjustment to mix statically unstable density distributions, and small horizontal Laplacian subgrid scale mixing with viscosity of 20 m2 s−1 and diffusivity of 5 m2 s−1. Rotation is uniform with Coriolis parameter of 1.38 × 10−4 s−1.

[11] We have neglected several potentially important dynamical effects, both in keeping with the simplified nature of the polynya model and because this is the first step in including any ocean dynamics in realistic estimates of total dense water production on a continental shelf. For example, we use a straight coastline because the polynya model is one-dimensional with no alongshelf variations. We do not include an explicit ice model because the polynya model already parameterizes the most important ice movements, and furthermore this eliminates possible problems with ice encountering the offshore wall. We ignore surface wind stresses because their primary effect would be to mix the water column beneath the polynya, but static instability caused by brine rejection already accomplishes this. Surface wind stresses could alter the motion of eddies within the polynya region, but we expect this influence to be secondary. We have ignored the influence of alongshelf currents known to occur in the polynya regions along the north coast of Alaska [e.g., Weingartner et al., 1998]. Chapman [2000] has shown that alongshelf currents can dramatically alter dense water production if the advective timescale of the alongshelf current traveling through the polynya is shorter than the time required for dense water eddies to form and produce a quasi-equilibrium within the polynya. In the present numerical calculations, the periodic channel basically represents an infinite coastline, so an alongshelf current would have virtually no effect on dense water production. Furthermore, as shown below (Figure 5), the quasi-equilibrium is typically reached within 20–30 days of the onset of winter, which is shorter than the advective timescale of about 45 days between Cape Lisburne and Pt. Barrow [Weingartner et al., 1998], so an alongshelf current should not substantially affect our results. Clearly, these and other idealizations can (and should) be relaxed in more realistic future calculations, but we feel they are appropriate for the present purposes.

[12] Each calculation begins from rest with a homogeneous ocean. At time t = 0, the effects of brine rejection from a coastal polynya are represented by applying a surface buoyancy flux B0 within a strip along the length of the channel, adjacent to the coast and extending a distance b offshore. The values of B0 and the polynya width b vary in time as produced by the polynya model (section 3). The surface buoyancy flux decreases sharply to zero offshore of b. The model time step is 432 s.

[13] The ocean model calculates density anomalies, whereas most previous studies discuss changes in salinity. Therefore, throughout the paper we use a simple conversion from density to salinity, where the change in density with salinity is 0.81 kg m−3 psu−1.

3. Ice Production and Buoyancy Fluxes

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Ocean Model Description
  5. 3. Ice Production and Buoyancy Fluxes
  6. 4. Dense Water Production
  7. 5. Halocline Water Formation
  8. 6. Discussion and Summary
  9. Acknowledgments
  10. References

[14] The surface buoyancy flux and polynya width that drive the ocean model are estimated using the time-dependent polynya model developed by Winsor and Björk [2000]. The model is forced by either meteorological observations at Barrow (WMO station 70026), or by the National Center for Environmental Prediction (NCEP) reanalysis project 10-m wind components and air temperature on a 6-hourly basis together with specific humidity. Cloudiness and solar radiation are represented by an annual cycle of monthly means; thus there are no interannual variations in these forcing functions. The solar radiation comes from Maykut [1982] and the cloud cover data are 8-year means from the International Satellite Cloud Climatology Project (ISCCP). The polynya width b is computed as

  • equation image

where hi is the effective collection thickness of frazil ice deposited at the downwind ice edge, un is the ice velocity normal to the coast taken to be positive in the offshore direction, and P is the (area-averaged) ice production given by

  • equation image

where Qnet is the net heat loss (estimated from standard bulk formulas), ρi is the density of ice and Lh the latent heat of fusion. The negative buoyancy supply B0 associated with the ice growth P is then given by

  • equation image

[15] Here Δρs is the increase in density with salinity and ρ0 is a reference density (1025 kg m−3). For details regarding the model computations, see Winsor and Björk [2000]. Both the polynya and ocean models are solved each winter for the period from 1 November to 30 April, which is designated by the autumn year, e.g., 1982 means the period 1 November 1982 to 30 April 1983.

[16] Statistics from the polynya model (forced by meteorological observations) are summarized in Table 1, while a typical example of the time histories of B0 and b are shown in Figures 5a and 5b for the 1987 winter. There are typically between 20 and 30 polynyas each winter, with as many as 40 in some winters. A typical polynya in this area is open 2–3 days, but there are events with polynyas that stay open for as long as one month. The mean polynya width is 13 km, well above the estimated overall mean width of typical Arctic polynyas of ∼5 km [Winsor and Björk, 2000]. Accumulated ice growth over the winters varies between 9 and 15 m, with a mean of 12 m. A typical winter accumulated ice growth on the Arctic shelves is ∼1.5 m, which shows that ice growth within polynyas can be 10 times as intense. The ice volume (accumulated ice growth times polynya width times 450 km) varies by a factor of 2.3 during the 21-year period. From accumulated ice growth we compute the nominal salinity increase dS that would occur if all salt rejected from the growing ice is evenly distributed through the water column beneath the polynya (without any mixing of surrounding waters), and this varies between 5.0 and 8.2 psu, with a mean of 6.4 psu (last column in Table 1).

Table 1. Winter Statistics From the Polynya Model Forced by Meteorological Observations at Barrowa
WinterMean/Maximum Width, kmAccumulated Ice Growth, m, equation imageIce Volume, km3, equation imageSalinity Increase equation image
  • a

    See Figure 1. Asterisk indicates L is the coastline length, 450 km.

197810/3811496.2
197914/7110555.6
198011/3812596.3
19819/3811485.9
198212/4715797.9
19839/2710395.5
198410/3413607.1
198510/3512526.4
198611/3912546.5
198711/3813747.2
198812/519465.0
198910/3312566.3
19908/3711466.3
199113/6112746.8
199211/3912546.3
199314/3714927.4
199412/5012676.6
199511/479485.0
199612/4212676.6
199714/6015948.2
199815/479595.0

[17] Figure 2 shows the ice volume produced for each winter using meteorological forcing (solid line) and NCEP data (dashed line). The ice volume from meteorological forcing lies between 39 and 94 km3 with a mean value of 61 km3, varying by a factor of more than two during the 21-year period. Interannual variability is high, e.g., the 1993 winter had an ice production of 92 km3 compared to the previous winter's production of 54 km3, a difference of 38 km3. The ice volume obtained using NCEP forcing is quite different, lying between 18 and 88 km3 with a mean value of 42 km3, on average about 70% of the meteorologically forced mean ice volume. The variability is much higher with a factor of 5 between maximum and minimum winter ice volumes. The long-term trend also differs; the meteorological forcing gives an increasing trend for the 21-year period, with the largest volumes produced at the end of the period, whereas the NCEP forcing gives the opposite with a long-term decreasing trend, and maximum volumes in the beginning of the period.

image

Figure 2. Ice volumes produced in polynyas using meteorological (solid line), and NCEP (dashed line) forcing for each winter. The 21-year mean ice volumes are 61 km3 and 42 km3 using meteorological and NCEP forcing, respectively.

Download figure to PowerPoint

[18] The cross-shelf distribution of the averaged buoyancy flux for each winter is shown in Figure 3, forced by meteorological observations (solid line) and NCEP data (dashed line). Most winters show a similar pattern with the largest buoyancy flux at the coast and then decreasing to zero about 20 km offshore, with most of the buoyancy loss occurring within 10 km of the coast. Some winters break the pattern; most noticeably 1980, which produced high buoyancy fluxes well past 40 km offshore for the NCEP data. The 1982 and 1997 winters produced the highest flux, with B0 =1.7 × 10−7 m2 s−3 near the coast and still greater than 1 × 10−7 m2 s−3 some 10 km offshore. In contrast, some winters show very modest buoyancy forcing, e.g., in 1988 the nearshore buoyancy flux is only 0.9 × 10−7 m2 s−3 and 0.5 × 10−7 m2 s−3 for meteorological and NCEP forcing, respectively. The distribution and strength of buoyancy flux across the shelf is important when considering water mass formation, and we will return to this issue in the next section.

image

Figure 3. Cross-shelf distribution of surface buoyancy fluxes (in m2 s−3), averaged over each winter, using meteorological observations (solid line) and NCEP data (dashed line).

Download figure to PowerPoint

4. Dense Water Production

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Ocean Model Description
  5. 3. Ice Production and Buoyancy Fluxes
  6. 4. Dense Water Production
  7. 5. Halocline Water Formation
  8. 6. Discussion and Summary
  9. Acknowledgments
  10. References

[19] The ocean response to buoyancy removed by ice growth is relatively unknown. The general response in this investigation is similar to that found in earlier studies [e.g., Gawarkiewicz and Chapman, 1995; Chapman and Gawarkiewicz, 1997; Chapman, 1999; Kikuchi et al., 1999; Gawarkiewicz, 2000], with a complicated density field in which baroclinic eddies form and transport dense water away from the source region. For example, Figure 4 shows plan views of the bottom density anomaly at four selected times during the 1987 winter, using the meteorological forcing. Both the polynya forcing and the ocean response in 1987 are typical of most winters. After 17.5 days (Figure 4a) there is an increase in density near the coast, with small instabilities growing along the edge of the polynya region, about 20–35 km offshore. After 25 days (Figure 4b) the increase in density near the coast is ∼1.0 kg m−3 (1.2 psu), while the densified water extends offshore a distance of ∼45 km. The instabilities have grown into larger eddies that are visible along the entire model domain. After 32.5 days (Figure 4c) the ocean response has become increasingly complicated with numerous eddies evolving and carrying the dense water farther offshore, and at 40 days (Figure 4d) the offshore extent is some 70–80 km, with a nearshore density increase of about 1.1 kg m−3 (1.3 psu). This gradual offshore movement of dense water corresponds to a velocity of roughly 2 cm s−1.

image

Figure 4. Plan views of bottom density anomaly from the ocean model (using meteorological forcing) during the 1987 winter after (a) 17.5, (b) 25, (c) 32.5, and (d) 40 days. The scale for density anomalies is shown on the right in kg m−3. The uniformly black offshore region represents ambient water with zero density anomaly.

Download figure to PowerPoint

[20] The temporal evolution of the polynya forcing (polynya width b, buoyancy flux B0, polynya width times buoyancy flux bB0) and the density increase near the coast is shown in Figure 5 for the 1987 winter. The dashed curve in Figure 5d is the density increase that would occur in the absence of eddies, i.e. with no ocean currents. As discussed by Chapman [1999], the ocean tends to integrate the effects of individual polynya events, generating a response that is much smoother than the forcing time series. The density increases with each polynya event until eddies start to carry the dense water away from the coast (about day 20). After this time, only the largest polynya events with the strongest buoyancy fluxes are able to change the density substantially. Despite continued polynya activity, the density increases only slightly owing to continuing offshore transport by eddies. In fact, during longer periods with no polynya activity the offshore transport of dense water and consequent onshore flux of less-dense ambient water causes a slight decrease in nearshore density. From about day 130 and onwards there is a continuous decrease in density, despite some polynya activity, as the nearshore denser water is replaced by ambient water. Clearly the eddies severely limit the maximum density increase produced during a winter, a result that is typical of all of our calculations.

image

Figure 5. Temporal evolution of (a) polynya width b, (b) buoyancy supply B0, (c) bB0, and (d) density anomaly ρ for the 1987 winter, using meteorological forcing. The dashed line in Figure 5d shows the nominal density increase without an ocean model, which reaches 5.6 kg m−3 for this particular winter (not visible in Figure 5d) compared to 1.5 kg m−3 from the ocean model.

Download figure to PowerPoint

[21] Figure 6 shows the distribution of volumes of water within various density anomaly ranges, averaged over the 21-year period, using both meteorological and NCEP forcing. Dense water with anomalies above 1.4 kg m−3 (1.7 psu) are produced in very modest volumes (<40 km3) on average, whereas anomalies in each density class up to 0.8 kg m−3 (1.0 psu) are produced at relatively large volumes (>230 km3). The primary difference in dense water formation between using meteorological and NCEP forcing lies in the 0.8–1.4 kg m−3 (1.0–1.7 psu) interval where the meteorological forcing produces much larger volumes. In fact, for density anomalies between 1.0 and 1.4 kg m−3 (1.2–1.7 psu), the meteorological forcing produces about three times as much water as the NCEP forcing does. The average total volume of dense water produced each winter by polynyas in this area is ∼1430 km3 (with meteorological forcing), equivalent to 0.05 Sv of water with anomalies between 0.2–1.8 kg m−3 (0.25–2.2 psu).

image

Figure 6. Distribution of 21-year mean volumes of water produced in the ocean model within different density anomaly ranges, using meteorological (black bars) and NCEP (gray bars) forcing.

Download figure to PowerPoint

[22] The interannual variability in dense water production is high. Figure 7a shows the distribution of volumes of water in various density anomaly ranges produced each winter using meteorological forcing, and Figure 7b the corresponding results for NCEP forcing. Looking at the three highest density anomaly ranges (1.2–1.8 kg m−3 corresponding to salinity increase between 1.5–2.2 psu) in Figure 7a, the variability is large. These densest water masses form in 14 out of 21 winters where some winters produce volumes exceeding 600 km3. The production of these water masses is spread out over the 21-year period, with slightly more being produced at the end of the period (after 1992). The largest and smallest maximum density anomalies and their volumes for the entire 21-year period are 1.6–1.8 and 0.8–1.0 kg m−3 with volumes 110 and 357 km3 for the 1997 and 1988 winters, respectively. Turning to Figure 7b (forced by NCEP data), the result is very different with the highest three density anomaly ranges occurring only in 5 out of 21 winters, and all in the first part of the period (1978–1986). From 1988 onwards, both anomalies and volumes are relatively small, with maximum anomalies in the 0.8–1.2 interval, while some winters, most noticeably 1988, produced a maximum anomaly of only 0.6 kg m−3 (0.7 psu). The interannual variability is higher than that obtained using meteorological forcing with large differences between successive winters. Such variability will have a large impact on the ability to form dense water that can ventilate deeper layers. We return to the cause of the large difference in density anomalies using meteorological versus NCEP forcing in section 6.

image

Figure 7. Distribution of volumes of water produced in the ocean model within different density anomaly ranges for each winter using (a) meteorological and (b) NCEP forcing. Density anomalies are binned in increments from 0.2 to 2 kg m−3 by 0.2 kg m−3 (scale on right side).

Download figure to PowerPoint

5. Halocline Water Formation

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Ocean Model Description
  5. 3. Ice Production and Buoyancy Fluxes
  6. 4. Dense Water Production
  7. 5. Halocline Water Formation
  8. 6. Discussion and Summary
  9. Acknowledgments
  10. References

[23] Here we analyze the typical maximum salinities that are produced by the end of the winter and how these might contribute to intermediate and deep waters in the Canada Basin. For this purpose we use a long-term climatologically based initial salinity for this area of 31.6 psu, determined from the mean surface salinity for October and November [Environmental Working Group, 1998]. In addition, we use observations of salinity from moorings situated in Bering Strait for the 1990–1997 period (K. Aagaard, personal communication, 2000) in order to capture the typical variability in the initial salinity field. The moorings included current meters equipped with thermistors and conductivity cells, and separate temperature/conductivity recorders, moored at 5–10 m above the seabed. For details regarding these current, temperature, and salinity data, see Weingartner et al. [1998]. Table 2 shows monthly mean salinities from September and October for each year from these moorings (for location of the moorings see Figure 1). The September salinities vary between 30.8 and 33.2 psu whereas the October salinities vary from 31.5 to 32.4 psu. Weingartner et al. [1998] found a typical advection time from Bering Strait to the Chukchi Shelf of about 2 months, leading us to use the September observations as an indication of initial salinity. Furthermore, we note that the climatologically based initial salinity of 31.6 psu falls within the midrange of these salinities.

Table 2. Monthly Mean Autumn Salinities From Moorings in Bering Straita
YearSeptemberOctober
  • a

    Salinities are in psu.

199033.1732.36
199131.8732.36
199230.84
199332.3031.46
199431.7031.97
199532.1432.02
199630.76
199732.1832.41

[24] Figure 8a8 shows the expected maximum salinities produced each winter, computed from the density anomaly added to the mean climatological initial salinity of 31.6 psu. The thick solid line shows salinities using meteorological forcing, and the dashed line shows the corresponding salinities using NCEP forcing. Upper and lower thin solid lines show the estimated salinity using the maximum and minimum initial salinity from the moorings in Bering Strait (see Table 2) added to the computed salinity increase using meteorological observations. The meteorological forcing produces 5 winters with salinities exceeding 33.5 psu (lower limit of the halocline [Melling, 1998]), and a mean maximum salinity of 33.3 psu, while the NCEP forcing salinity rarely exceeds 33.5 psu (only for 3 winters), with a mean maximum salinity of 33.0 psu. The large effect of varying initial salinities is obvious from this plot. The initial salinity (from the moorings in the Bering Strait) varies between 30.8 to 33.2 psu, a range of 2.4 psu. When using the maximum initial salinity, all winters produce water with salinities capable of ventilating the lower halocline (>34 psu). Maximum salinities may reach 35.4 psu. When using the lowest initial salinity of 30.8 psu (bottom thin line in Figure 8.), the maximum salinity produced is only ∼33 psu, at the lower limit of that needed to contribute to the halocline in the Canada Basin [Melling, 1998], and comparable to the maximum initial salinity.

image

Figure 8. (a) Maximum shelf salinities for the 21-year period from 1978 to 1998. Salinities are computed from the maximum density anomaly produced each winter, added to the mean initial salinity of 31.6 psu based on climatology, using meteorological (thick solid line) and NCEP (dashed line) forcing. Upper and lower thin solid lines show the estimated salinity using the maximum and minimum initial salinity from the moorings in Bering Strait (see Table 2) rather than the climatological mean, using meteorological observations. (b) Maximum shelf salinities achieved for the 1990–1997 period using initial salinities observed each year from moorings in Bering Strait (Table 2), using meteorological (bold line) and NCEP (dashed line) forcing. The shaded area shows typical salinities of the halocline layer in the Canada Basin (∼33.5–34.5 psu).

Download figure to PowerPoint

image

Figure 8. (continued)

Download figure to PowerPoint

[25] Figure 8b shows the autumn salinities from the moorings (dotted line) together with the modeled maximum shelf salinities for the 1990–1997 period (solid line shows meteorological forcing and dashed line NCEP forcing). The shaded area shows typical halocline salinities in the Canada Basin (∼33.5–34.5 psu). The results using meteorological forcing suggest that contributions to the halocline from polynyas on the Chukchi Shelf occur regularly with, in this particular case, 6 out of 8 winters producing waters with salinities exceeding 33.5 psu. Using NCEP forcing, no lower halocline water (∼34.5 psu) was produced, and only 3 out of 8 winters produced water with salinities exceeding 33.5 psu. It is apparent that the influence of the initial conditions on the ability to form halocline water is comparable to the influence of coastal polynyas. That is, the salinity increase from the ocean model is typically about 1 psu during this period, whereas the initial salinity itself varies by about 2 psu. Thus the production of really salty (dense) water requires a combination of high initial salinities and polynyas that create unusually large density anomalies, e.g. 1997.

6. Discussion and Summary

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Ocean Model Description
  5. 3. Ice Production and Buoyancy Fluxes
  6. 4. Dense Water Production
  7. 5. Halocline Water Formation
  8. 6. Discussion and Summary
  9. Acknowledgments
  10. References

6.1. Comparison of Model Estimates With Observations

[26] A brief comparison of our modeled density anomalies with observations from the Chukchi Shelf is instructive. Weingartner et al. [1998] present mooring data from the Chukchi Shelf and Bering Strait for the 1991 winter, together with open water estimates for this region based on SSM/I satellite images. In particular, they show observations of cold, hypersaline water attributed to waters formed in coastal polynyas (their Figure 6). Observations from a mooring at the head of Barrow Canyon in the very eastern part of our model domain (Figure 1), show that the ambient salinity before the observed salinity increase (which occurred around 1 February 1991) was about 32.5 psu and then increased to ∼34 psu (for a short period nearly up to 35 psu). The salinity increase was then 1.5–2 psu, which is equivalent to a density anomaly of 1.2–1.6 kg m−3. Observations farther out on the shelf at a water depth of about 50 m (mooring HS3 in their Figure 6, marked HS in our Figure 1) show a more moderate salinity increase of about 1.2–1.5 psu (1.0–1.2 kg m−3). Maximum density anomalies from our model calculations for this period are in the 1.2–1.4 kg m−3 range, in concert with the observations. Furthermore, Figure 8b shows the maximum shelf salinity to be about 33.5 psu for 1991 using meteorological forcing. Thus, the model produces water masses that fall within the range of the observations.

[27] Aagaard et al. [1981] presented observations of cold and salty waters during February to March, 1977 offshore of Cape Lisburne. The mean near-bottom salinity was greater than 34.2 psu over the entire 250-km-long section (see their Figure 14) with maximum salinities in excess of 34.7 psu within 40 km of the coast. Although this specific winter is not covered by our investigation it shows clearly that dense water is formed in this area. In contrast, Aagaard and Roach [1990] found little evidence of dense water in Barrow Canyon during the 1986–1987 winter. Instead they found warm and salty events indicating upwelling from the Atlantic layer. This is consistent with Figure 7, which shows a marked decrease in dense water production for the 1986 winter for both meteorological and NCEP forcing. These observations provide evidence of the large interannual variability of dense water in this area, which is supported by the present analysis.

[28] Björk [1989, 1990] compared model-generated T-S profiles with observations from the Arctic basins, and found that 1.2 Sv of cold and salty shelf water was needed to get close agreement. Furthermore, this shelf contribution needed to have a salinity distribution much like the distributions found in the present results. Cavalieri and Martin [1994] computed ice production and salt release from Arctic polynyas based on estimates of open water area derived from SSM/I satellite images. For the 1978–1986 period they estimated a mean ice production of 165 km3 for the same area under investigation here (see their Table 5). Our modeled ice production for the same period (using meteorological forcing) is 55 km3, about one third of their estimate. This result was also found by Winsor and Björk [2000], using NCEP forcing, who concluded that the coarse resolution of the SSMR and SSM/I images (25 × 25 km) probably overestimates the open water area.

[29] Winsor and Björk [2000] used NCEP data to force the same polynya model as used here to compute statistics of polynya sizes and ice volumes for the entire Arctic Basin. Their study did not include an ocean model, but they estimated the potential annual winter shelf salinities and volumes by applying a constant cross-shelf throughflow of 2 cm s−1 across the polynya whose salinity was equal to the ambient water (i.e., initial salinity). Not surprisingly, the ice volumes found in their study agree well with those found here using NCEP data, with the same decreasing long-term trend shown in Figure 2. Figure 9 shows density anomalies and volumes produced each winter during the 1979–1996 period from this study (solid lines) along with those from Winsor and Björk [2000] for the same area as investigated here (dashed lines). For the present study, we plot volumes for the largest density anomaly range in which at least 100 km3 of water was produced. This choice is somewhat arbitrary, but 100 km3 is a significant volume, and it leaves out only a few of the absolute highest density anomalies from the distributions in Figure 7. The results found here typically have lower density anomalies but greater volumes than those found by Winsor and Björk [2000]. The mean density anomalies for the two studies are 1.6 and 1.0 kg m−3, respectively, while the maximum density anomalies in 1982 are 2.2 and 1.8 kg m−3, respectively. The present calculations produced a mean volume of 250 km3, with high variability, whereas the Winsor and Björk [2000] mean volume is 153 km3, with less variability and a long-term decreasing trend. The inclusion of an ocean model in the present study provides shelf salinity estimates that are closer to observations, and the full distribution of water masses produced during a winter as opposed to the single salinity and volume estimate in the work of Winsor and Björk [2000].

image

Figure 9. Winter estimates of (a) density anomalies and (b) volumes from the present study (solid lines), and Winsor and Björk [2000] (dashed lines). See text for details.

Download figure to PowerPoint

6.2. Meteorological Observations and NCEP Reanalysis Data

[30] What is the reason behind the large differences in ice volume and dense water production found using meteorological observations and NCEP forcing? Basically, the NCEP forcing does not produce as much negative buoyancy as the observations from Barrow. Figure 10 shows monthly mean air temperatures, wind speeds, and offshore wind speeds for the entire 21-year period from observations at Barrow (thick line) and NCEP reanalysis data (thin line). Two main features stand out in this figure. First, the NCEP data show a large anomaly in both air temperature, wind speeds, and offshore wind speeds for the 1979–1981 period, with air temperatures that are about 10°C warmer and wind speeds that are 30–50% higher than the meteorological observations. To be sure that this is an artificial anomaly in the NCEP data, we compared the observations at Barrow with those from another meteorological station located close to Barrow (Wainwright, WMO station 70030, see Figure 1). The observations from both stations agree very well (not shown), so we are confident that the anomaly in the NCEP data is not real. Warmer air temperatures reduce the surface heat flux, while stronger winds increase the surface heat flux. Therefore, we might expect the effects of the anomalies in the NCEP data during the 1979–1981 period to cancel in the polynya model, and this happens in 1979 and 1981 (see Figure 3). However, the anomalous offshore wind during 1980 (Figure 10c) creates exceptionally wide polynyas which leads to the large offshore extent in buoyancy supply for that winter (>40 km). The anomalous wind speeds during 1979–1981 are also responsible for the large ice volumes and the corresponding dense water anomalies obtained during this period using the NCEP forcing (see Figures 2 and 7b, respectively) because ice production is proportional to B0b, which depends on wind speed but is nearly independent of air temperature [e.g., Chapman, 1999]. Apart from the anomaly in the NCEP data the two air temperature time series agree well, with a correlation coefficient r = 0.78 (N = 122).

image

Figure 10. Monthly mean (a) air temperatures, (b) wind speeds, and (c) offshore/onshore wind speeds (offshore winds being positive) for the 21-year period from 1978 to 1998, based on meteorological observations (thick line) and NCEP data (thin line).

Download figure to PowerPoint

[31] The second important feature is that the NCEP wind speeds in Figure 10b show a general decreasing trend from 1986 onward with a mean wind speed for the 1986–1998 period of 4.7 m s−1 compared to 5.7 m s−1 for the meteorological observations, and with a difference in monthly mean wind speed up to 2.5 m s−1. The overall agreement is less than that for air temperature (r = 0.37, N = 122). This decline in wind speed is the main cause of the decreasing trend in ice volume and dense water production compared to the results using meteorological forcing. This is also evident in the offshore winds speeds in Figure 10c which show some months with offshore winds (positive values) in the meteorological data and onshore winds in the NCEP data. The correlation between these two curves is 0.75. The effect of the trend can be seen in Figure 3 as a decrease in the average NCEP buoyancy fluxes relative to the meteorologically based average fluxes after 1986. An analysis of the reasons behind the behavior of the NCEP forcing is outside the scope of the present paper, but it shows that one should be careful when using such products, especially if they are not also compared to in situ observations.

6.3. Dense Water Formation

[32] Most of the variability seen in dense water formation can be traced to variations in polynya activity that are caused by variations in the fraction of offshore winds. To demonstrate, Figure 1111 shows the fraction of polynyas and the fraction of offshore winds for each winter using meteorological and NCEP forcing. The meteorological forcing contains a higher percentage of offshore winds and more polynyas for all 21 winters (mean of 61%). The interannual variability is high, e.g., 1997 with polynyas 75% of the time and the following winter (1998) with only 54%. NCEP forcing produces fewer polynyas with an overall mean of 46%, and a fraction of only 34% for the 1988 winter. Consider the two extreme winters of 1988 and 1997; 1997 produced the densest water with anomalies up to 1.8 kg m−3, while 1988 produced the least, with maximum anomalies in the 0.8–1.0 kg m−3 range. The air temperatures and wind speeds did not differ much between the two winters. The most important difference is in the frequency of occurrence of offshore winds and, consequently, the number of days with polynyas; the 1997 winter had polynyas ∼75% of the time whereas the 1988 winter only had polynyas for ∼49% of the time. Finally, we point out that a visual correlation between ice volume and either polynya activity or offshore winds is striking (compare Figures 2 and 11). This is consistent with Chapman's [1999] conclusion that offshore winds are more important in dense water production than air temperature. Thus, for this region, the location of atmospheric low-pressure systems (and the Beaufort High) is crucial for forming dense water from coastal polynyas. That is, a shift in the general atmospheric circulation over this region can produce a large change in dense water production.

image

Figure 11. Statistics for each winter of (a) percent occurrence of polynyas and (b) percent offshore winds, using meteorological (solid line) and NCEP (dashed line) forcing.

Download figure to PowerPoint

image

Figure 11. (continued)

Download figure to PowerPoint

[33] The results presented thus far suggest that there may be a simple relationship between the computed winter ice volume and the dense water distribution computed by the ocean model. To investigate, we have compared correlation coefficients between the dense water volume of each density class in Figure 7a and the winter ice volume in Figure 2. The results are somewhat surprising; nearly all high correlations were found for the top four density classes (anomalies between 1.0 and 1.8 kg m−3), with very high correlations (r ≥ 0.9) for anomalies between 1.2 and 1.8 kg m−3. The correlations for anomalies between 0.2 and 1.0 kg m−3 were all very low (r < 0.2). This suggests that we can estimate the volumes of the densest water from a simple linear relationship with the ice volume produced by polynyas for any given winter. Also noticeable is the linear cross-shelf distribution of buoyancy (Figure 3) that could be used to form an analytical relationship between the supplied buoyancy and dense water production. Such relationships can be useful tools for estimating dense water production from forcing alone, without the need to run a complicated ocean model, but we leave this for future investigations.

6.4. Model Sensitivity and Uncertainties

[34] We have already noted (in section 2) that numerous assumptions and simplifications were made in designing the ocean model calculations. We have to conducted a number of additional model calculations (not presented) to examine the sensitivity of our results to a few of these choices; e.g., bottom depth, sloping bathymetry, model resolution and turbulence closure scheme. Not surprisingly, the density anomalies achieved are most sensitive to the water depth used; a shallower water depth produces denser water but in smaller volumes. Sensitivity tests show that the volume fluxes and density anomalies are relatively insensitive to a sloping bottom, finer grid resolution and the choice of turbulence closure scheme. As most offshore transport of dense water is controlled by baroclinic eddy formation, relatively high resolution is needed. The model was found to be insensitive to increased model resolution, implying that the 2 km resolution used here is sufficient to resolve the baroclinic eddies (the internal Rossby radius being ∼5 km).

[35] Another uncertainty lies in the computed buoyancy supply. The polynya model assumes free drift of ice, whereas the sea ice in the Chukchi Sea may not respond to offshore wind events as easily, especially so at the end of winter when the sea ice cover has grown in thickness, with ridging and overall ice stress being present (see Winsor and Björk [2000] for a more detailed discussion of the polynya model). The buoyancy supplied by the polynya model is most likely an estimate of the upper limit of the buoyancy that can be produced by polynyas in this particular area, so our results for ice volume and dense water production should probably be regarded as upper limits as well.

6.5. Summary

[36] We have presented a first attempt to model the formation and distribution of dense water on the Chukchi Shelf over an extended period of time; 1978–1998. Cold and salty water produced by coastal polynyas may form density anomalies ranging from 0.6 to 1.8 kg m−3, equivalent to a salinity increase of 0.7 to 2.2 psu. Large amounts of water with density anomalies of at least 1.0 kg m−3 are produced nearly every winter. However, anomalies exceeding 1.4 kg m−3 (1.7 psu) are rare and volumes involved are small during the 21-year period investigated. In terms of halocline water production, variations in the initial salinity appear to be as important as the anomalies produced in polynyas, contributing to the intermittent nature of this process. Results also show that the ice volume and dense water produced in the polynyas are quite different when forcing the model with the NCEP reanalysis product instead of the meteorological observations. This shows that comparison to in situ observations should be done for any work using such products in the Arctic Basin.

[37] Although these results are suggestive, we feel that much further work is needed to produce truly accurate estimates of dense water production on Arctic shelves. Present estimates of intermediate and deepwater production are highly uncertain. Our understanding of the processes that actually couple the dense shelf water production with halocline maintenance is primitive, at best. Our knowledge of the required dense water flux and the initial conditions as well as the coupling of the atmosphere, ice and ocean in this polynya region is equally sketchy. Many more detailed observations and modeling studies are needed to increase our knowledge of this intriguing system.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Ocean Model Description
  5. 3. Ice Production and Buoyancy Fluxes
  6. 4. Dense Water Production
  7. 5. Halocline Water Formation
  8. 6. Discussion and Summary
  9. Acknowledgments
  10. References

[38] We are grateful to Knut Aagaard, Tom Weingartner, and Rebecca Woodgate for generously sharing their salinity data from Bering Strait. Mark Ortmeyer helped with the meteorological forcing data. We thank Göran Björk for his helpful comments on the manuscript. DC received financial support for this research from the High-Latitude Dynamics Program of the Office of Naval Research under grant N00014-99-1-0333 as part of the western Shelf-Basin Interactions (SBI) program. Woods Hole Oceanographic Institution contribution 10592.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Ocean Model Description
  5. 3. Ice Production and Buoyancy Fluxes
  6. 4. Dense Water Production
  7. 5. Halocline Water Formation
  8. 6. Discussion and Summary
  9. Acknowledgments
  10. References