• Clarke, R. A., , and J-C. Gascard, 1983: The formation of Labrador Sea Water. Part I: Large-scale processes. J. Phys. Oceanogr., 13 , 17641778.

    • Search Google Scholar
    • Export Citation
  • Davis, R. E., , J. T. Sherman, , and J. Dufour, 2001: Profiling ALACEs and other advances in autonomous subsurface floats. J. Atmos. Oceanic Technol., 18 , 982993.

    • Search Google Scholar
    • Export Citation
  • Dickson, R., , J. Lazier, , J. Meincke, , P. Rhines, , and J. Swift, 1996: Long-term coordinated changes in the convective activity of the North Atlantic. Progress in Oceanography, Vol. 38, Pergamon, 241–295.

    • Search Google Scholar
    • Export Citation
  • Gascard, J-C., 1978: Mediterranean deep water formation, baroclinic instability, and oceanic eddies. Oceanol. Acta, 1 , 315330.

  • Gascard, J-C., , and R. A. Clarke, 1983: The formation of Labrador Sea Water. Part II: Mesoscale and smaller-scale processes. J. Phys. Oceanogr., 13 , 17791797.

    • Search Google Scholar
    • Export Citation
  • Haine, T. W. N., , and J. Marshall, 1998: Gravitational, symmetric, and baroclinic instability of the ocean mixed layer. J. Phys. Oceanogr., 28 , 634658.

    • Search Google Scholar
    • Export Citation
  • Harcourt, R. R., 1999: Numerical simulation of deep convection and the response of drifters in the Labrador Sea. Ph.D. dissertation, University of California, Santa Cruz, 370 pp.

    • Search Google Scholar
    • Export Citation
  • Jones, H., , and J. Marshall, 1993: Convection with rotation in a neutral ocean: A study of open-ocean deep convection. J. Phys. Oceanogr., 23 , 10091039.

    • Search Google Scholar
    • Export Citation
  • Lab Sea Group, 1998: The Labrador Sea Deep Convection Experiment. Bull. Amer. Meteor. Soc., 79 , 20332058.

  • Lavender, K. L., , R. E. Davis, , and W. B. Owens, 2000: Mid-depth recirculation observed in the interior Labrador and Irminger Seas by direct velocity measurements. Nature, 407 , 6669.

    • Search Google Scholar
    • Export Citation
  • Lazier, J. R. N., 1973: The renewal of Labrador Sea Water. Deep-Sea Res., 20 , 341353.

  • Lilly, J. M., , P. B. Rhines, , M. Visbeck, , R. Davis, , J. R. N. Lazier, , F. Schott, , and D. Farmer, 1999: Observing deep convection in the Labrador Sea during winter 1994/95. J. Phys. Oceanogr., 29 , 20652098.

    • Search Google Scholar
    • Export Citation
  • Marshall, J., , and F. Schott, 1999: Open-ocean convection: Observations, theory, and models. Rev. Geophys., 37 , 164.

  • Renfrew, I. A., , G. W. K. Moore, , P. S. Guest, , and K. Bumke, 2002: A comparison of surface layer and surface turbulent flux observations over the Labrador Sea with ECMWF analyses and NCEP Reanalyses. J. Phys. Oceanogr., 32 , 383400.

    • Search Google Scholar
    • Export Citation
  • Schott, F., , and K. D. Leaman, 1991: Observations with moored acoustic Doppler current profilers in the convection regime in the Golfe du Lion. J. Phys. Oceanogr., 21 , 558574.

    • Search Google Scholar
    • Export Citation
  • Schott, F., , M. Visbeck, , and J. Fischer, . 1993: Observations of vertical currents and convection in the central Greenland Sea during the winter of 1988–1989. J. Geophys. Res., 98 , 1440114421.

    • Search Google Scholar
    • Export Citation
  • Schott, F., , M. Visbeck, , U. Send, , J. Fischer, , L. Stramma, , and Y. Desaubies, . 1996: Observations of deep convection in the Gulf of Lions, northern Mediterranean, during the winter of 1991/92. J. Phys. Oceanogr., 26 , 505524.

    • Search Google Scholar
    • Export Citation
  • Send, U., , and J. Marshall, 1995: Integral effects of deep convection. J. Phys. Oceanogr., 25 , 855872.

  • Stommel, H., 1972: Deep winter-time convection in the western Mediterranean Sea. Studies in Physical Oceanography: A Tribute to Georg Wüst on His 80th Birthday, A. L. Gordon, Ed., Vol. 2, Gordon and Breach, 207–218.

    • Search Google Scholar
    • Export Citation
  • Straneo, F., , M. Kawase, , and S. C. Riser, 2002: Idealized models of slantwise convection in a baroclinic flow. J. Phys. Oceanogr., 32 , 558572.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Positions of VCM float deployments in the Labrador Sea. Circles mark floats that drift at 400 m, squares at 700 m, and triangles at 1500 m. Filled symbols mark floats that were deployed in Feb–Mar 1997; open symbols mark floats that were deployed in Jan–Feb 1998. Contours mark isobaths in meters

  • View in gallery

    Locations of vertical profiles of temperature and salinity between Nov 1996 and Apr 1997, coded by color and shape to indicate the estimated depth of the mixed layer. The black box marks the “convection region,” where all mixed layers deeper than 800 m were measured. Note the mixed layers between 400 and 800 m deep that are located north of 60°N, and those located southwest of the tip of Greenland. Contours mark the 500-, 1000-, 2000-, 3000-, and 4000-m isobaths

  • View in gallery

    Average surface heat flux in winter (Jan–Apr) 1997 in the Labrador Sea from Moore–NCEP data. Negative values indicate heat loss from the ocean to the atmosphere. The contour interval is 50 W m−2, and the dashed box marks the convection region

  • View in gallery

    Mean winter circulation of the Labrador Sea, and locations of mixed layers deeper than 500 m in winter 1997. The mean winter circulation was computed by objectively mapping winter float trajectory data as in Lavender et al. (2000), and is plotted as contours (thick) of geostrophic pressure at 700 m. The contour interval is 1 cm and low pressure regions are shaded in gray. Circles mark locations of observed mixed layers between 500 and 800 m deep; stars mark observed mixed layers deeper than 800 m. Thin contours indicate the 500-, 1000-, and 1500-m isobaths

  • View in gallery

    PDF of vertical velocity at 400 m, normalized by bin width. Solid line indicates PDF of w between Feb and Apr 1997; dashed line indicates PDF of w between Jan and Apr 1998

  • View in gallery

    Locations of 88-h time series of vertical velocity and temperature at 400 m in the Labrador Sea in winter 1997. (a) Color-coded by w variance of each time series, 〈w2〉. (b) Color-coded by mean T of each time series, 〈T〉. Black box marks the convection region, and contours mark 500-, 1000-, 2000-, 3000-, and 4000-m isobaths

  • View in gallery

    Cumulative vertical displacement over 88 h at 400 m in winter 1997. Vertical displacement is computed by time-integrating each w time series

  • View in gallery

    Mean vertical velocity of each 88-h time series at 400 m in 1997, plotted vs float position in the mixed layer. Position in the mixed layer is defined as the float drift depth divided by the mixed layer depth. Thick line indicates linear least squares fit to the data points. The positive slope suggests a float sampling bias

  • View in gallery

    Mixed layer depth estimated from temperature profiles in the Labrador Sea plotted vs time, Nov 1996–Jun 1997. Open triangles highlight deep mixed layers observed later in the winter season than expected. Ticks mark the start of each month

  • View in gallery

    Low-pass filtered (with a 10-day running mean) time series of (a) vertical velocity and (b) temperature at 400 m in the convection region in winter 1997. Thick line indicates the mean; thin lines indicate the root-mean-square error about the mean

  • View in gallery

    All time series of vertical velocity measured at 400 m in 1997, with high surface heat loss events shaded. Moore–NCEP heat fluxes greater than 100 W m−2 are shaded in light gray; those greater than 200 W m−2 are shaded in dark gray

  • View in gallery

    Cumulative heat content at the sea surface in the convection region in winters 1997 and 1998. Heat content was computed by time-integrating Moore–NCEP heat fluxes in the convection region

  • View in gallery

    Low-pass filtered (with a 10-day running mean) time series of heat flux measurements in the convection region in winter 1997. Negative values indicate heat loss from the ocean to the atmosphere. Black lines indicate average surface heat flux from the original NCEP Reanalysis model (solid), and the calibrated Moore–NCEP data (dot–dashed). Blue lines indicate heat flux at the surface (solid) and at 400 m (dashed), computed from consecutive float temperature profiles. Red line indicates high-frequency vertical heat flux computed from time series measurements of w and T (see text for details)

  • View in gallery

    Examples of 88-h time series of vertical velocity and temperature measured in March 1997 in the convection region. Vertical velocity time series are in red; temperature time series are in blue. The estimated high-frequency vertical heat flux of the time series, −ρcpwT″〉, is in W m−2; the average pressure measured by the float during the 88-h drift period is in db

  • View in gallery

    Power spectra of (a) vertical velocity and (b) temperature. Spectra were computed from 450 time series at 400 m between Feb and Jun 1997

  • View in gallery

    Cross-spectral coherence and phase of vertical velocity and temperature. Cross-spectrum was computed from 450 time series at 400 m between Feb and Jun 1997. (a) Coherence. Horizontal line indicates the estimated level of statistical significance. (b) Phase in degrees

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 186 186 10
PDF Downloads 71 71 7

Observations of Open-Ocean Deep Convection in the Labrador Sea from Subsurface Floats

View More View Less
  • 1 Scripps Institution of Oceanography, University of California, San Diego, San Diego, California
  • | 2 Woods Hole Oceanographic Institution, Woods Hole, Massachusetts
© Get Permissions
Full access

Abstract

The occurrence and extent of deep convection in the Labrador Sea in winters 1996/97 and 1997/98 is investigated from measurements of over 200 neutrally buoyant subsurface Profiling Autonomous Lagrangian Circulation Explorer (PALACE) and Sounding Oceanographic Lagrangian Observer (SOLO) floats. In addition to providing drift velocity data and vertical profiles of temperature and salinity, 55 floats are equipped with vertical current meters (VCMs). Time series of vertical velocity (derived from measured pressure and vertical flow past the float) and temperature are obtained from the VCM floats. Mixed layer depths estimated from profile measurements indicate that convection reached depths greater than 1300 m in 1997, but no deeper than 1000 m in 1998. Deep mixed layers were concentrated in the western basin, although a number of deep mixed layers were observed southwest of Cape Farewell and also north of 60°N. The highest variance in vertical velocity and the lowest mean temperatures were found in the western basin, suggesting that this area is the main site of deep convection. Deep mixed layers and large vertical velocities were observed as late as April and May, despite the fact that surface forcing appears to have ceased. Estimates of mean vertical velocity appear to be affected by a float sampling bias, whereby floats preferentially sample convergent regions. The effect of this bias, which is dependent on the float depth within the convective layer, is to sample upward flow in early winter and downward flow in late winter when the convective layer has deepened. A one-dimensional heat balance model is examined, whereby the winter surface heat flux, estimated from temperature profiles, is balanced by the turbulent vertical heat flux associated with deep convection, estimated from time series measurements. The plume-scale vertical heat flux can only account for roughly −80 of −350 W m−2 measured at 400-m depth. The vertical heat flux at longer timescales is investigated, but cannot be resolved with this dataset. Failure to balance the surface heat flux by plume-scale motions, combined with an observed high variance of w and T at low frequencies, suggests that motion at these longer timescales contributes to the one-dimensional heat budget in winter.

Corresponding author address: Kara L. Lavender, 9500 Gilman Drive, Mail Code 0230, La Jolla, CA 92093-0230. Email: klavender@ucsd.edu

Abstract

The occurrence and extent of deep convection in the Labrador Sea in winters 1996/97 and 1997/98 is investigated from measurements of over 200 neutrally buoyant subsurface Profiling Autonomous Lagrangian Circulation Explorer (PALACE) and Sounding Oceanographic Lagrangian Observer (SOLO) floats. In addition to providing drift velocity data and vertical profiles of temperature and salinity, 55 floats are equipped with vertical current meters (VCMs). Time series of vertical velocity (derived from measured pressure and vertical flow past the float) and temperature are obtained from the VCM floats. Mixed layer depths estimated from profile measurements indicate that convection reached depths greater than 1300 m in 1997, but no deeper than 1000 m in 1998. Deep mixed layers were concentrated in the western basin, although a number of deep mixed layers were observed southwest of Cape Farewell and also north of 60°N. The highest variance in vertical velocity and the lowest mean temperatures were found in the western basin, suggesting that this area is the main site of deep convection. Deep mixed layers and large vertical velocities were observed as late as April and May, despite the fact that surface forcing appears to have ceased. Estimates of mean vertical velocity appear to be affected by a float sampling bias, whereby floats preferentially sample convergent regions. The effect of this bias, which is dependent on the float depth within the convective layer, is to sample upward flow in early winter and downward flow in late winter when the convective layer has deepened. A one-dimensional heat balance model is examined, whereby the winter surface heat flux, estimated from temperature profiles, is balanced by the turbulent vertical heat flux associated with deep convection, estimated from time series measurements. The plume-scale vertical heat flux can only account for roughly −80 of −350 W m−2 measured at 400-m depth. The vertical heat flux at longer timescales is investigated, but cannot be resolved with this dataset. Failure to balance the surface heat flux by plume-scale motions, combined with an observed high variance of w and T at low frequencies, suggests that motion at these longer timescales contributes to the one-dimensional heat budget in winter.

Corresponding author address: Kara L. Lavender, 9500 Gilman Drive, Mail Code 0230, La Jolla, CA 92093-0230. Email: klavender@ucsd.edu

1. Introduction

The Labrador Sea is one of the few regions in the World Ocean where an intermediate-depth water mass is formed through the process of open-ocean deep convection. Hydrographic measurements have documented the presence of well-mixed layers to depths down to 2300 m in the Labrador Sea (e.g., Dickson et al. 1996) and have identified a region in the western Labrador Sea where surface waters are denser than the surrounding area (Lazier 1973). Such evidence has been used to infer the occurrence of deep convection in winter, although few direct measurements of the active process in the Labrador Sea have been reported. Gascard and Clarke (1983) observed convection with three vertical current meter floats in the western Labrador Sea in late March 1976, and nearly two decades later a moored ADCP measured vertical velocity near Ocean Weather Station Bravo (Lilly et al. 1999). Both experiments measured large vertical velocities (up to 9 cm s−1) in winter, which were associated with periods of strong winds and cold air temperatures. However, both studies were limited spatially and could not resolve the extent of deep convection.

Most recently, the Labrador Sea Deep Convection Experiment (Lab Sea Group 1998) was undertaken to observe and model oceanic deep convection. The field component of this experiment, which took place in the winters of 1996/97 and 1997/98, consisted of a multitude of measurements of both the meteorology and the ocean state. In this paper we examine open-ocean deep convection with measurements from neutrally buoyant subsurface floats, including time series measurements of vertical velocity and temperature, vertical profiles of temperature and salinity, and measurements of horizontal drift velocity. After presenting the context of the problem and introducing the dataset used, the occurrence and extent of deep convection during the two winters is diagnosed, and the wintertime heat budget is examined.

2. Background

The process of open-ocean convection is generally described in terms of three phases: preconditioning, deep convection and lateral exchange (Marshall and Schott 1999). A localized [O(100's km)] region is considered to be “preconditioned” if it is more likely than surrounding regions to experience deep convection in response to buoyancy loss from strong atmospheric forcing. This state is generally marked by a weak stratification close to the surface, which can arise, for example, when a baroclinic cyclonic circulation causes isopycnals to dome upward. From hydrographic and point-measurement current meter data, Clarke and Gascard (1983) found some evidence of an O(200 km) cyclonic gyre in the western Labrador Sea. A more recent study of the mean circulation of the Labrador Sea found direct evidence for a localized cyclonic gyre in a region of weak flow in the western basin (Lavender et al. 2000). The weak flow allows water to remain in the region for an extended period of time, where it may be continually subjected to a large wintertime heat loss associated with strong offshore winds. This continued heat loss eventually triggers the second phase of the convection process, deep convection.

Deep convection refers to the deep vertical overturn initiated by a large buoyancy loss at the surface. Deep convection is thought to occur in plumes of O(1 km) in both horizontal and vertical scale, characterized by bursts of vertical velocity of O(10 cm s−1). Taken together, a collection of vertical plumes may act to form a dense, well-mixed patch in one of two ways. In the first, for which we use the historical term chimney, the plumes cause a net downward transport of dense water. In an alternate scenario, the downward motion of plumes is compensated by upward motion between the plumes, resulting in a zero mean vertical velocity over the patch. In this case, the dense patch is formed purely by mixing of the cold surface waters with warmer waters below.

Several attempts have been made to compute the mean vertical velocity, and hence the net vertical transport of dense water, during periods of strong atmospheric forcing in the Mediterranean Sea (Schott and Leaman 1991; Schott et al. 1996) and the Greenland Sea (Schott et al. 1993). In all three studies, moored ADCPs measured vertical velocity in one location at various depths. The studies all reported maximum downward vertical velocities as large as 5–10 cm s−1 during periods of strong surface cooling, and most reported a concurrent increase in vertical velocity variance. Weaker upward vertical velocities were observed in the times between the large downwelling events. Estimates of mean vertical velocity, computed over typically week-long periods of strong cooling, ranged from −1 cm s−1 (Schott and Leaman 1991) to +0.3 cm s−1 (Schott et al. 1996). While Schott and Leaman (1991) used their estimate of mean downwelling to infer a mean annual deep water renewal rate of 1 Sv (Sv ≡ 106 m3 s−1) in the Mediterranean Sea, the other authors argued that the small mean vertical velocities suggest that convective plumes act as mixing agents that do not transport mass vertically.

While this hypothesis cannot be confirmed without spatial information across the plume region, numerical experiments that model convection driven by a circular region of cooling over both a neutral and a weakly stratified water column (e.g., Jones and Marshall 1993) find that the upwelling regions between downward plumes nearly compensate the downwelling of fluid. Further physical arguments based on vorticity constraints (Send and Marshall 1995) suggest that a chimney with net downwelling is unphysical because of the unrealistically large sheared flows that would result from even small mean vertical velocities.

Whether or not there is a net mass transport by convection, one would expect to find a significant vertical heat flux associated with downward motion at the core of plumes, and the presumably more buoyant return flow outside. The result of these motions is a vertical heat flux that cools deeper water, leaving the signature of deep convection, and at least partially balances the heat loss to the atmosphere. No estimates of the vertical heat flux were made in earlier field studies, although Schott et al. (1993) noted that, during cooling events in the Greenland Sea, deep fluctuations in temperature tended to have longer timescales than deep fluctuations in vertical velocity. This raises interesting questions about the timescale of the vertical heat flux and the mechanism that carries it.

While a vertical heat flux is thought to occur at the plume scale, both vertical and horizontal heat fluxes may be associated with dense water sinking and spreading in the third phase of deep convection, lateral exchange. This phase begins once a region of dense water is formed and could occur during deep convection itself. In the theoretical chimney scenario where dense water is transported downward, the convergence and divergence of the flow in the water column induces a circulation that is cyclonic in the upper part of the water column and anticyclonic below. These flows become baroclinically unstable at the density front that surrounds the newly cooled water, generating geostrophic eddies with scales on the order of the local Rossby radius of deformation (Gascard 1978). The eddies advect dense water away from the region, and the remaining patch eventually spreads out at its neutral density while less dense fluid drawn in at the surface restratifies the upper water column (Stommel 1972).

On the other hand, if plumes do not cause a mean downwelling but act to distribute the surface buoyancy loss throughout the water column by mixing, then one would expect a uniform patch of water to form, the density of which is higher than the surrounding area. The large density gradient between this dense patch and its surroundings then induces, through the thermal wind relationship, a narrow rim current around the patch. Instability of the rim current forms cone-shaped eddies (cone-shaped due to the slanted isopycnals from rotational adjustment) that travel away from the convection region and eventually slump down and spread at their neutral density (Jones and Marshall 1993).

In either scenario, a horizontal exchange occurs through advection by eddies, which support a lateral heat flux in addition to the vertical heat flux. Such processes have been modeled numerically; however, heat fluxes have not been reported. Because the spatial resolution of the float measurements is not small enough to detect such eddies, we will not discuss the lateral exchange phase in detail. However, it is important to keep these theoretical ideas in mind when interpreting heat flux estimates from floats.

3. Dataset

The dataset used in this study consists primarily of measurements from 55 neutrally buoyant, subsurface, vertical current meter Profiling Autonomous Lagrangian Circulation Explorer (PALACE) and Sounding Oceanographic Lagrangian Observer (SOLO) floats (Davis et al. 2001). These floats, deployed as part of the Office of Naval Research (ONR) Labrador Sea Deep Convection Experiment, are a subset of a larger group of over 200 floats that were deployed throughout the North Atlantic as part of the World Ocean Circulation Experiment (WOCE). All PALACE and SOLO floats are programmed to cycle from the surface to a target depth where they drift with the subsurface flow for a preset time period. At the end of its cycle time, each float ascends to the surface where it remains for one day to transmit data via the ARGOS satellite system. The floats provide drift velocity data, as well as vertical profiles of temperature and salinity that are measured during ascent or descent.

In addition to pressure, temperature, and conductivity sensors, the vertical current meter floats (VCMs) are equipped with a propeller that rotates the float in response to vertical flow past it. One full vane rotation corresponds to 4 m relative water displacement. The vane rotation, measured every 8 seconds by an internal compass, is combined with the measured pressure, P, to obtain a time series of the absolute displacement of the water, η. Simultaneous time series of temperature, T, and in some floats conductivity, C, are also recorded by the VCMs at the drift depth, and half-hour averages of all quantities (P, η, T, and sometimes C) are reported to ARGOS.

Once the time series data are obtained from ARGOS, the displacement data are postcalibrated using the total rotations measured during the float's descent to its drift depth. The absolute vertical velocity, w, is then computed from the calibrated η. Davis et al. (2001) describe an asymmetry in the VCM response to upward and downward flows, which was corrected here by reducing the magnitude of upward measurements by 2.5%. With this correction the accuracy of the vertical velocity measurements is roughly 3 m day−1 (Davis et al. 2001). The accuracy of the temperature measurements is 0.005°C.

The VCMs were deployed in the Labrador Sea in January–February 1997, and January–February 1998 during hydrographic surveys carried out as part of the Deep Convection Experiment (Fig. 1). The nominal drift depth of each VCM is either 400 m (40 floats), 700 m (14 floats), or 1500 m (1 float), but all floats profile to near 1500-m depth. All VCMs that drift at 700 m were deployed in 1998, as were all floats that measure time series of conductivity. The VCMs operate under different duty cycles during the winter and spring (January–June) than during the summer and fall (July–December). During the first half of the year the VCMs record time series over their 88-h period at drift depth. The rest of the year the floats drift at depth for roughly nine days without recording time series. In both modes, vertical profiles of temperature and salinity are measured as the float descends to its drift depth.

VCM floats have target lifetimes of roughly 100 cycles, which is determined by the stored battery energy. Many floats, however, were unable to descend to depth as a result of losing one or more vanes of the propeller. This increases the buoyancy of the float, causing it to remain near the surface. Time series from floats with average drift depths less than 300 m, which includes those from floats with lost vanes, were discarded from this study. The dataset analyzed below consists of 483 (P, w, T) time series measured by 25 floats from January through June 1997 and 799 (P, w, T) time series measured by 20 floats from January through mid-July 1998. Of the time series measured in 1998, 494 also have time series of conductivity.

4. Deep convection in winters 1996/97 and 1997/98

The first question we address is whether deep convection occurred during the two winters of the experiment. Deep convection is often diagnosed by the presence of mixed layers deeper than a typical winter wind-mixed layer, indicating formation of intermediate or deep water masses. In this study the winter mixed layer is estimated from each vertical profile of temperature and salinity measured by floats in the Labrador Sea between November 1996 and April 1997 (Fig. 2). The mixed layer depth is defined by the break between a least squares fit of a straight line to the upper layer, and a second-order polynomial plus exponential fit to the lower layer of each temperature profile. Each temperature and salinity profile was examined to ensure that the defined mixed layer is well mixed to the surface and is not capped.

Mixed layers deeper than 400 m are observed across the basin, while mixed layers deeper than 800 m are clearly concentrated in the western Labrador Sea. Previous hydrographic measurements (Clarke and Gascard 1983) and those collected as part of the Deep Convection Experiment (Lab Sea Group 1998) have also shown evidence of deep convection in the western basin, although such ship surveys have limited spatial coverage because of harsh sea conditions and limited ship time.

The spatial coverage afforded by the floats allows the extent of the deepest mixed layers to be clearly identified. The region in the western Labrador Sea containing the deepest mixed layers is marked by the box in Fig. 2, and is dubbed the “convection region.” The mixed layers measured by floats also indicate convection to slightly shallower depths (400–800 m) to the north of 60°N and also southwest of the tip of Greenland. The occurrence of deep convection in these regions was unexpected, although neither region had previously been sampled during a year of strong convection.

Because deep convection is forced by a buoyancy loss at the sea surface, the location of strong winter surface forcing is a factor in determining the location of deep overturning. Buoyancy loss occurs through fluxes of heat and freshwater, which act to change the density of the ocean surface. Because shipboard measurements of these quantities are scarce at high latitudes, especially during the winter, models such as the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis model provide the best available estimates of basinwide, year-round meteorological fluxes. As part of the Deep Convection Experiment, outputs of the NCEP Reanalysis model have been calibrated from shipboard meteorological observations from the Labrador Sea, following the study of Renfrew et al. (2002) (G.W.K. Moore 2000, personal communication; hereafter referred to as Moore–NCEP data).

The heat flux component dominates the buoyancy flux in the Labrador Sea (Marshall and Schott 1999). Figure 3 depicts the average surface heat flux in the Labrador Sea in winter 1997 (January–April) from Moore–NCEP data. The highest mean heat flux is concentrated offshore of the Labrador coast in the northwest portion of the basin, where winter average values exceed 250 W m−2. This heat flux is the result of strong, cold, dry winds originating from the North American continent, and its offshore location is related to the winter sea ice cover that may extend hundreds of kilometers offshore.

The location of the maximum heat flux alone explains the deep overturning in the basin north of 60°N. Convection in the western basin and in the region southwest of Greenland, however, must be preconditioned in some way. Preconditioning is defined by a weak stratification so that a smaller surface buoyancy loss drives deeper overturning. This weak stratification may be caused, for example, by a dynamical upwelling associated with gyre spinup resulting in the upward tilting of isopycnals. Alternatively, the local circulation may act to retain convectively formed water in the same region year after year, where it is repeatedly subjected to large heat fluxes in the winter.

Figure 4 illustrates the mean winter circulation mapped from float drift velocity data, with the locations of deep mixed layers overlaid. There is a clear correspondence between the location of deep mixed layers and the cyclonic flows associated with low pressure regions, both in the western basin and southwest of Greenland. Evidence for a baroclinic cyclonic gyre in the western basin was described by Clarke and Gascard (1983) using hydrographic and current meter measurements, while Lavender et al. (2000) found that weak, largely barotropic flow in these regions traps water there for an extended period of time. In addition, estimates of autumn (October–December) stratification from temperature and salinity profiles find the most weakly stratified water in or near the regions of cyclonic flow in the western basin and southwest of Greenland. Thus, both regions appear to be preconditioned by this flow. The larger net winter heat flux in the western basin then drives convection to greater depths than in the region southwest of Greenland.

Mixed layer depth measurements describe the product of deep convection, while measurements of vertical velocity provide a description of the process itself. Time series measurements from 400 m only are presented here. Vertical velocity data from floats that drift deeper provide similar results, but are generally less energetic. Figure 5 shows the probability density function (PDF) of vertical velocity at 400 m from February to April 1997. Observed w ranges from −11.9 to +8.9 cm s−1, and the PDF is skewed negative (skewness = −0.66). The skewness is due to the higher number of large negative (w ≤ −5 cm s−1) velocities compared to large positive velocities. Large downwelling and large upwelling events are short-lived, with over 70% persisting for only one-half hour (the resolution of the time series), while the longest observed event lasted three hours. The magnitude and, surprisingly, the duration of these events as observed from floats are comparable to previous Eulerian observations of vertical velocity during convection (Schott and Leaman 1991; Schott et al. 1993, 1996; Lilly et al. 1999).

The spatial pattern of high variance in the vertical velocity field at 400 m is similar to the spatial pattern of deep mixed layers (Fig. 6a). Most examples of 88-h time series with large vertical velocity variance occur within the “convection region,” marked by the box in Fig. 6. This is the same region where the mean temperatures of the 400-m time series are clearly at a minimum (Fig. 6b). Unfortunately, in 1997 no VCMs were located north of 60°N or southwest of Greenland; thus the occurrence of active convection in these regions could not be assessed.

Despite the relative sparsity and short timescale of large w events, even small vertical velocities may result in large net vertical displacements over the 88-h period of a time series. Figure 7 illustrates the cumulative vertical displacement measured over each time series in winter 1997. Most net displacements are less than 200 m over 88 h (∼60 m day−1); however, a number of time series indicate vertical motion of up to 3 km upward and 2 km downward. Large upwelling and downwelling excursions are clearly visible in these time series and are attributed to convective activity.

An asymmetry is apparent in Fig. 7 in which more large positive displacements are observed than large negative displacements. This asymmetry is believed to be caused by a sampling bias in the float measurements. In an upright plume regime, horizontal convergences and divergences exist within the convective layer and tend to attract or expel isobaric floats. Shallow in the convective layer, downward plumes that grow with depth are convergent and draw floats into them. Toward the bottom of the layer, where plumes weaken with depth, the horizontally divergent plumes eject floats into upward motion on either side. The depth of the float within the convective layer then determines its sampling bias. Such behavior was found in numerical simulations of isobaric floats in a convection regime (Harcourt 1999). Harcourt found that model floats in the upper third of the mixed layer measure a more negative mean vertical velocity, relative to the Eulerian mean, while floats in the lower two-thirds of the mixed layer measure a more positive mean vertical velocity.

Figure 8 shows a scatterplot of the mean vertical velocity of each 88-h time series in the Labrador Sea plotted against float position in the mixed layer, as measured by the ratio of drift depth to mixed layer depth. Only 17% of floats at 400 m were located within the mixed layer, resulting in the small number of points in Fig. 8. However, despite the scatter about the best-fit line to the data, it appears that floats in the upper portion of the mixed layer tend to measure a negative mean vertical velocity, whereas floats near the bottom of the layer tend to measure a positive mean vertical velocity, as seen in the model by Harcourt (1999).

This sampling bias is time-dependent due to the increasing depth of the mixed layer as winter progresses (Fig. 9). The mean vertical velocity in the convection region is plotted as a function of time in Fig. 10a. This figure illustrates behavior that is consistent with the sampling bias described above, whereby floats tend to measure an upward mean velocity early in the winter when they drift near the bottom of the convective layer and a downward mean velocity when the mixed layer deepens and floats drift in the upper portion of the layer.

The seasonal evolution of the vertical velocity field is described in Fig. 11, where all time series of vertical velocity are plotted together. The gray shading indicates time periods during which the Moore–NCEP heat flux in the convection region is greater than 100 W m−2 (light) and 200 W m−2 (dark), indicative of strong atmospheric forcing. In midwinter, when the highest heat fluxes occur, the largest variance in w is observed. However, later in the season, in early April and as late as May, substantial w variance is observed when the surface heat flux indicates only a small heat loss or even a heat gain by the ocean.

Deep mixed layers are also observed during this late winter period. Figure 9 shows two “peaks” in mixed layer depth that are coincident with the periods of high vertical velocity activity in Fig. 11. The deepest mixed layer observed (1328 m deep) was measured on 10 April and was apparently formed by the w event in early April. It is conceivable that a mesoscale atmospheric event with large surface cooling occurred during this time that was not captured by the NCEP model; however there is no evidence for this.

The differences in deep convection observed in the winters of 1996/97 and 1997/98 are described in Table 1. Convection was much weaker in 1997/98 as a result of a lower net heat loss from the ocean to the atmosphere. The net heat loss in the convection region from January to April of 1998 was roughly 25% lower than in the previous winter (Fig. 12). As a result, shallower mixed layers, a warmer mean temperature at 400 m, and a less energetic vertical velocity field (Fig. 5) were observed in 1998 (Table 1).

5. One-dimensional heat budget

During the winter the ocean surface is cooled when heat is lost to the atmosphere during strong forcing events. This heat flux, negative indicating heat loss from the ocean, is carried downward into the ocean by vertical mixing from convective plumes and perhaps by other processes. We investigate the one-dimensional heat balance using two independent measures provided by the floats. First, vertical profiles of temperature are integrated in depth to estimate the heat content of the water column. A heat flux is then computed by temporally and spatially averaging the change in heat content between consecutive profiles. Second, an estimate of the vertical temperature flux, 〈wT〉, is computed from time series data. From these measurements the following one-dimensional heat balance is evaluated in the convection region in winter 1997:
i1520-0485-32-2-511-e1
where ρcp is the heat capacity of seawater, and the angle brackets denote a temporal and spatial filter that removes plume and eddy scales. Integrating from the maximum depth of the profiles to the 400-m depth of the time series, and assuming the heat flux below the base of the profiles is negligible, the terms in the following equation can be directly estimated from the float measurements:
i1520-0485-32-2-511-e2

Before evaluating the terms above, it is useful to compare the surface heat flux computed from profiles (0–1300 m) to the net surface heat flux from the NCEP model. Their difference is an estimate of the error in assuming zero heat flux below 1300-m depth and in neglecting lateral advection at scales unresolved by float drift. Figure 13 shows the surface heat flux in the convection region computed from the original NCEP Reanalysis product and from the calibrated Moore–NCEP data. The estimates differ most in midwinter, when the calibrated values are smaller by up to 250 W m−2. Also shown in Fig. 13 is the average surface heat flux computed from temperature profiles in the convection region. This curve is more similar in magnitude to the original NCEP curve than the Moore–NCEP curve, although the heat flux from profiles has much larger fluctuations. These fluctuations have a period of roughly 20 days, which closely corresponds to the period of NCEP fluctuations (the correlation between the two filtered time series is 0.87) and is likely related to the passage of storm systems. The heat flux measured by profiles, however, turns positive as early as March, when the model fluxes still indicate ocean cooling. We believe that this discrepancy results from a convergence of lateral eddy fluxes as suggested by Lilly et al. (1999).

The heat flux at 400 m, computed from temperature profiles for comparison with float 〈wT〉, is also shown in Fig. 13. The magnitude of the 400-m heat flux is significantly less than the surface heat flux in early winter, indicating that a large amount of heat is lost in initial cooling and deepening of the mixed layer above 400 m.

The vertical temperature flux can be broken down into a mean flux, wT, a low-frequency flux, 〈wT′〉, and a high-frequency flux, 〈wT″〉. The primes indicate the fluctuation about a mean, while the brackets represent the space–time mean of Eq. (1). The mean vertical velocity and temperature fields at 400 m in the convection region can be described by a spatially homogeneous, time-dependent mean (Fig. 10). Because our measurements of the mean vertical velocity are affected by a float sampling bias, we cannot comment on the contribution of wT to the heat budget.

Here the high-frequency components are taken as the fluctuations in an 88-h float record around the time mean of that record. Consequently, the low-frequency components are the fluctuations of 88-h averages about the temporal and spatial means w and T. The high-frequency heat flux, 〈wT″〉, can be thought of as the heat flux carried by vertical convective plumes, which are believed to have timescales of O(hours). If the ocean followed the simple picture of cold, downward plumes that are at least partially compensated by warm, upward motions, one would expect to see a high positive correlation between w″ and T″, as well as a large vertical heat flux. However, while 73% of time series are positively correlated, only 8% of time series have a correlation that is greater than 0.2. While causing some decorrelation, internal waves cannot explain this low correlation. In individual time series, plumelike behavior is observed in w, but the behavior of T appears unrelated, with frequent steplike features that are not seen in the w record.

Figure 14 illustrates four examples of time series of w and T measured near 400-m depth in March 1997. Figures 14a and 14b show examples of what might be considered “plumes,” where downward vertical velocities are associated with a decrease in temperature. These time series have very different behaviors, however. The time series in Fig. 14a show a fairly small vertical velocity (−2 cm s−1) clearly associated with a large change in temperature (1°C). This 2–3-h event is responsible for most of the high heat flux measured in the time series (−ρcpwT″〉 = −651 W m−2). Figure 14b, on the other hand, has examples of large w (5–10 cm s−1 downward) associated with much smaller, often steplike, decreases in temperature. These time series also measure a large heat flux (−ρcpwT″〉 = −528 W m−2), although in this case the downward events do not dominate the heat flux.

In both cases, although there are instances of plumelike behavior, the time series are not highly correlated. While the w time series exhibit high-frequency fluctuations, temperature tends to vary on longer timescales with steplike behaviors or trends. As indicated in Fig. 14b, even though w and T do not exhibit large simultaneous spikes, the vertical heat flux is the same order of magnitude as the largest surface heat fluxes. In this dataset, however, it is equally as common to find examples of time series with quasi-plumelike behavior that do not carry large heat fluxes.

In Fig. 14c, for example, there are clear instances of plume-scale velocities, but there is very little change in temperature over most of the time series. This results in a positive heat flux (−ρcpwT″〉 = 136 W m−2) over the 88-h record. Likewise, Fig. 14d illustrates an example of a time series whose measured heat flux is quite small (−ρcpwT″〉 = −62 W m−2). In this case the temperature drops by 0.3°C in a large step, but the vertical velocities are fairly small and the correlation between w and T is less than 0.04.

The mean vertical heat flux computed from 〈wT″〉 in the convection region (red line in Fig. 13) is clearly less than the heat flux at 400 m, indicating that the heat flux carried by vertical plumes cannot balance the winter surface heat loss. During the period of maximum cooling, 1 February–9 March, the profile-based heat flux at 400 m is −353 W m−2, whereas the vertical heat flux carried by plumes is only −82 W m−2. Errors due to measurement accuracy of w and T (3 m day−1 and 0.005°C, respectively), even if coherent over entire time series, lead to negligible errors (<1 W m−2) in the high-frequency heat flux. The error introduced by the sampling bias cannot be calculated from this dataset, although it would have to be surprisingly large to account for the discrepancy in the heat flux estimates.

If a one-dimensional heat balance is maintained, there must be a significant vertical heat flux carried at timescales longer than the plume scale, or in the 〈wT′〉 component. The power spectra of w and T (Fig. 15) are both very red, with the highest energy at the lowest nonzero frequency (88-h period). The cospectrum and quadrature spectrum (not shown) also peak at 88 h, but the coherence between w and T (Fig. 16) is not significantly different from zero at periods longer than 6 h. Thus, it is not possible to accurately measure the very small correlation between w and T at these timescales with this dataset. Because high variance is observed in both fields at the lowest frequencies, even a small correlation could result in a substantial vertical heat flux.

Estimates of 〈wT′〉 were computed from fluctuations of the 88-h record means around the large-scale, spatially homogeneous, time-dependent means, w and T. The resulting heat fluxes are roughly 300 W m−2; however, because of the low number of observations the standard deviation of the estimates is much larger than the mean. To remove variance related to the simplified representation of the mean field, estimates using a more complicated mean that is both spatially and temporally varying were also computed. While the spatial pattern of the residuals changed, there was no significant difference in the mean low-frequency heat flux. Although it is not possible to resolve the 〈wT′〉 signal with this dataset, the analysis is suggestive of a contribution to the vertical heat flux at timescales longer than O(days).

6. Summary and discussion

In this study we describe observations of deep convection in the Labrador Sea in the winters of 1996/97 and 1997/98 from subsurface float measurements. Subsurface floats are a valuable tool in providing year-round temporal coverage and widespread spatial coverage of the Labrador Sea, where the same harsh winter conditions that force deep convection make ship-based work difficult. From these floats a comprehensive set of wintertime measurements of temperature, salinity, and vertical velocity was obtained. Despite the limitations of the subsurface floats, namely measurements at only three depths and coarse horizontal resolution, the floats have offered a new look at the active process of intermediate water formation in the open ocean.

The float observations establish that deep convection did occur in both winters, although it was significantly weaker in 1997/98 due to a reduced surface heat loss. In both years the deepest mixed layers were observed in a roughly 300 km by 550 km region in the western Labrador Sea, coincident with previous observations of newly convected water. Deep mixed layers were also observed north of 60°N as well as southwest of the tip of Greenland, both regions not previously noted as locations of active deep convection. Judging from autumn profiles, both the western basin and the region southwest of Greenland are preconditioned for deep convection by weak flow in localized cyclonic gyres that retains weakly stratified water from the previous year. This same trapping characteristic improves the effectiveness of wintertime cooling in driving deep convection.

In the western basin, vertical velocity and temperature measurements reveal many of the same features of deep convection as observed in previous experiments. Maximum vertical velocities are O(10 cm s−1), although more large downward vertical velocities are observed than large upward velocities, and the vertical velocity variance peaks during periods of large surface heat fluxes. The products of deep convection, namely newly cooled water (<3°C at 400 m) and deep mixed layers, are concentrated in this region. Finally, estimates of the timescales of the convective plumes, characterized by large vertical velocities, are comparable to previous estimates. Most downwelling events persisted for less than 4 h.

The quasi-Lagrangian behavior of the nearly isobaric floats appears to bias low-frequency vertical velocity measurements because floats preferentially sample horizontally convergent regions. This bias precludes measurement of the mean vertical velocity field and leaves open the question of whether or not a collection of plumes transports mass vertically or whether the downwelling action is entirely compensated by upward motion resulting in a vertically mixed water column.

Although we do not have a reliable estimate of the mean vertical velocity in the convection region, it would be surprising if the mean w were greater than the O(0.1 cm s−1) observed from floats. Send and Marshall (1995) used a vorticity argument to argue that a mean downwelling of −1 cm s−1 would result in a shear of 3 m s−1 over a 10-km distance in the Mediterranean Sea. While this is certainly an unrealistically large shear, the same shear calculation computed with a 0.1 cm s−1 downwelling in the Labrador Sea gives a much lower value of 12 cm s−1 over 10 km. Such shear is not unrealistic, and does not eliminate the possibility of a mean transport associated with plumes. Without observational evidence, however, neither the theory of a downwelling chimney nor that of a mixed region can be ruled out.

In this study a one-dimensional heat budget model is evaluated in which the large winter heat loss at the ocean surface is balanced by a turbulent vertical heat flux associated with deep convection. Plume-scale fluctuations of vertical velocity and temperature from time series carry a heat flux of only O(−100 W m−2), while estimates from temperature profiles at the same depth are up to four times larger in midwinter. While the float sampling bias likely affects estimates of 〈wT″〉, it is not clear by how much. It would be surprising if this accounted for the discrepancy in heat flux estimates. Likewise, neither lateral advection nor lateral eddy fluxes are expected to close the heat budget. The slow cyclonic circulation advects newly mixed cold water within the convection region, whereas the lateral eddy fluxes would require a large upgradient flux of heat below 400 m, which is not expected.

Then the one-dimensional heat budget must contain a significant contribution at scales longer than O(days). Although the low-frequency vertical heat flux could not be resolved from this dataset, there is an indication of a large heat flux at these eddy timescales. One mechanism that could support such a vertical heat flux is baroclinic instability, which may form eddies that slump and bring newly convected water to its neutral density in the final stage of deep convection. Another possibility is the existence of slanted, as opposed to purely vertical, convective plumes. Such slantwise convection, investigated theoretically and in numerical models by Haine and Marshall (1998) and Straneo et al. (2002), could be forced by horizontal density gradients in the ocean or by horizontal gradients in the forcing field, both of which exist in the Labrador Sea. Slantwise convection not only fluxes heat laterally as well as vertically, but also develops convective layers that remain vertically stratified. This effect would make actively convective regions difficult to identify based on vertical profiles alone.

Clearly many open questions remain about the process of deep convection in the Labrador Sea. Some of these could be addressed with a detailed comparison between models and observations. The effect of a potential float sampling bias, for example, might be described from virtual floats in models of both upright and slantwise convection, or in models incorporating baroclinic instability effects. Comparisons of float measurements with model time series and vertical and lateral heat fluxes could also be used to evaluate the success of a given model.

Acknowledgments

We thank J. Dufour, J. Sherman, J. Valdes, R. Tavares, and B. Guest for technical development and preparation of the floats, D. Newton and C. Wooding for help with data processing, and the many scientists who deployed the floats at sea. NCEP reanalysis data is provided by the NOAA/CIRES Climate Diagnostics Center, Boulder, Colorado, from their Web site at http://www.cdc.noaa.gov. This work is supported by the Office of Naval Research and the National Science Foundation.

References

  • Clarke, R. A., , and J-C. Gascard, 1983: The formation of Labrador Sea Water. Part I: Large-scale processes. J. Phys. Oceanogr., 13 , 17641778.

    • Search Google Scholar
    • Export Citation
  • Davis, R. E., , J. T. Sherman, , and J. Dufour, 2001: Profiling ALACEs and other advances in autonomous subsurface floats. J. Atmos. Oceanic Technol., 18 , 982993.

    • Search Google Scholar
    • Export Citation
  • Dickson, R., , J. Lazier, , J. Meincke, , P. Rhines, , and J. Swift, 1996: Long-term coordinated changes in the convective activity of the North Atlantic. Progress in Oceanography, Vol. 38, Pergamon, 241–295.

    • Search Google Scholar
    • Export Citation
  • Gascard, J-C., 1978: Mediterranean deep water formation, baroclinic instability, and oceanic eddies. Oceanol. Acta, 1 , 315330.

  • Gascard, J-C., , and R. A. Clarke, 1983: The formation of Labrador Sea Water. Part II: Mesoscale and smaller-scale processes. J. Phys. Oceanogr., 13 , 17791797.

    • Search Google Scholar
    • Export Citation
  • Haine, T. W. N., , and J. Marshall, 1998: Gravitational, symmetric, and baroclinic instability of the ocean mixed layer. J. Phys. Oceanogr., 28 , 634658.

    • Search Google Scholar
    • Export Citation
  • Harcourt, R. R., 1999: Numerical simulation of deep convection and the response of drifters in the Labrador Sea. Ph.D. dissertation, University of California, Santa Cruz, 370 pp.

    • Search Google Scholar
    • Export Citation
  • Jones, H., , and J. Marshall, 1993: Convection with rotation in a neutral ocean: A study of open-ocean deep convection. J. Phys. Oceanogr., 23 , 10091039.

    • Search Google Scholar
    • Export Citation
  • Lab Sea Group, 1998: The Labrador Sea Deep Convection Experiment. Bull. Amer. Meteor. Soc., 79 , 20332058.

  • Lavender, K. L., , R. E. Davis, , and W. B. Owens, 2000: Mid-depth recirculation observed in the interior Labrador and Irminger Seas by direct velocity measurements. Nature, 407 , 6669.

    • Search Google Scholar
    • Export Citation
  • Lazier, J. R. N., 1973: The renewal of Labrador Sea Water. Deep-Sea Res., 20 , 341353.

  • Lilly, J. M., , P. B. Rhines, , M. Visbeck, , R. Davis, , J. R. N. Lazier, , F. Schott, , and D. Farmer, 1999: Observing deep convection in the Labrador Sea during winter 1994/95. J. Phys. Oceanogr., 29 , 20652098.

    • Search Google Scholar
    • Export Citation
  • Marshall, J., , and F. Schott, 1999: Open-ocean convection: Observations, theory, and models. Rev. Geophys., 37 , 164.

  • Renfrew, I. A., , G. W. K. Moore, , P. S. Guest, , and K. Bumke, 2002: A comparison of surface layer and surface turbulent flux observations over the Labrador Sea with ECMWF analyses and NCEP Reanalyses. J. Phys. Oceanogr., 32 , 383400.

    • Search Google Scholar
    • Export Citation
  • Schott, F., , and K. D. Leaman, 1991: Observations with moored acoustic Doppler current profilers in the convection regime in the Golfe du Lion. J. Phys. Oceanogr., 21 , 558574.

    • Search Google Scholar
    • Export Citation
  • Schott, F., , M. Visbeck, , and J. Fischer, . 1993: Observations of vertical currents and convection in the central Greenland Sea during the winter of 1988–1989. J. Geophys. Res., 98 , 1440114421.

    • Search Google Scholar
    • Export Citation
  • Schott, F., , M. Visbeck, , U. Send, , J. Fischer, , L. Stramma, , and Y. Desaubies, . 1996: Observations of deep convection in the Gulf of Lions, northern Mediterranean, during the winter of 1991/92. J. Phys. Oceanogr., 26 , 505524.

    • Search Google Scholar
    • Export Citation
  • Send, U., , and J. Marshall, 1995: Integral effects of deep convection. J. Phys. Oceanogr., 25 , 855872.

  • Stommel, H., 1972: Deep winter-time convection in the western Mediterranean Sea. Studies in Physical Oceanography: A Tribute to Georg Wüst on His 80th Birthday, A. L. Gordon, Ed., Vol. 2, Gordon and Breach, 207–218.

    • Search Google Scholar
    • Export Citation
  • Straneo, F., , M. Kawase, , and S. C. Riser, 2002: Idealized models of slantwise convection in a baroclinic flow. J. Phys. Oceanogr., 32 , 558572.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Positions of VCM float deployments in the Labrador Sea. Circles mark floats that drift at 400 m, squares at 700 m, and triangles at 1500 m. Filled symbols mark floats that were deployed in Feb–Mar 1997; open symbols mark floats that were deployed in Jan–Feb 1998. Contours mark isobaths in meters

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 2.
Fig. 2.

Locations of vertical profiles of temperature and salinity between Nov 1996 and Apr 1997, coded by color and shape to indicate the estimated depth of the mixed layer. The black box marks the “convection region,” where all mixed layers deeper than 800 m were measured. Note the mixed layers between 400 and 800 m deep that are located north of 60°N, and those located southwest of the tip of Greenland. Contours mark the 500-, 1000-, 2000-, 3000-, and 4000-m isobaths

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 3.
Fig. 3.

Average surface heat flux in winter (Jan–Apr) 1997 in the Labrador Sea from Moore–NCEP data. Negative values indicate heat loss from the ocean to the atmosphere. The contour interval is 50 W m−2, and the dashed box marks the convection region

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 4.
Fig. 4.

Mean winter circulation of the Labrador Sea, and locations of mixed layers deeper than 500 m in winter 1997. The mean winter circulation was computed by objectively mapping winter float trajectory data as in Lavender et al. (2000), and is plotted as contours (thick) of geostrophic pressure at 700 m. The contour interval is 1 cm and low pressure regions are shaded in gray. Circles mark locations of observed mixed layers between 500 and 800 m deep; stars mark observed mixed layers deeper than 800 m. Thin contours indicate the 500-, 1000-, and 1500-m isobaths

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 5.
Fig. 5.

PDF of vertical velocity at 400 m, normalized by bin width. Solid line indicates PDF of w between Feb and Apr 1997; dashed line indicates PDF of w between Jan and Apr 1998

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 6.
Fig. 6.

Locations of 88-h time series of vertical velocity and temperature at 400 m in the Labrador Sea in winter 1997. (a) Color-coded by w variance of each time series, 〈w2〉. (b) Color-coded by mean T of each time series, 〈T〉. Black box marks the convection region, and contours mark 500-, 1000-, 2000-, 3000-, and 4000-m isobaths

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 7.
Fig. 7.

Cumulative vertical displacement over 88 h at 400 m in winter 1997. Vertical displacement is computed by time-integrating each w time series

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 8.
Fig. 8.

Mean vertical velocity of each 88-h time series at 400 m in 1997, plotted vs float position in the mixed layer. Position in the mixed layer is defined as the float drift depth divided by the mixed layer depth. Thick line indicates linear least squares fit to the data points. The positive slope suggests a float sampling bias

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 9.
Fig. 9.

Mixed layer depth estimated from temperature profiles in the Labrador Sea plotted vs time, Nov 1996–Jun 1997. Open triangles highlight deep mixed layers observed later in the winter season than expected. Ticks mark the start of each month

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 10.
Fig. 10.

Low-pass filtered (with a 10-day running mean) time series of (a) vertical velocity and (b) temperature at 400 m in the convection region in winter 1997. Thick line indicates the mean; thin lines indicate the root-mean-square error about the mean

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 11.
Fig. 11.

All time series of vertical velocity measured at 400 m in 1997, with high surface heat loss events shaded. Moore–NCEP heat fluxes greater than 100 W m−2 are shaded in light gray; those greater than 200 W m−2 are shaded in dark gray

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 12.
Fig. 12.

Cumulative heat content at the sea surface in the convection region in winters 1997 and 1998. Heat content was computed by time-integrating Moore–NCEP heat fluxes in the convection region

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 13.
Fig. 13.

Low-pass filtered (with a 10-day running mean) time series of heat flux measurements in the convection region in winter 1997. Negative values indicate heat loss from the ocean to the atmosphere. Black lines indicate average surface heat flux from the original NCEP Reanalysis model (solid), and the calibrated Moore–NCEP data (dot–dashed). Blue lines indicate heat flux at the surface (solid) and at 400 m (dashed), computed from consecutive float temperature profiles. Red line indicates high-frequency vertical heat flux computed from time series measurements of w and T (see text for details)

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 14.
Fig. 14.

Examples of 88-h time series of vertical velocity and temperature measured in March 1997 in the convection region. Vertical velocity time series are in red; temperature time series are in blue. The estimated high-frequency vertical heat flux of the time series, −ρcpwT″〉, is in W m−2; the average pressure measured by the float during the 88-h drift period is in db

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 15.
Fig. 15.

Power spectra of (a) vertical velocity and (b) temperature. Spectra were computed from 450 time series at 400 m between Feb and Jun 1997

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Fig. 16.
Fig. 16.

Cross-spectral coherence and phase of vertical velocity and temperature. Cross-spectrum was computed from 450 time series at 400 m between Feb and Jun 1997. (a) Coherence. Horizontal line indicates the estimated level of statistical significance. (b) Phase in degrees

Citation: Journal of Physical Oceanography 32, 2; 10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2

Table 1.

Comparison of 1997 and 1998 time series data at 400 m. Heat flux estimates are from Moore-NCEP data; negative values indicate heat loss from ocean to atmosphere, and CR indicates convection region, defined as 55°–59°N, 51°–56°W.

Table 1.

*

Woods Hole Oceanographic Institution Contribution Number 10318.

Save