a. Three-dimensional mapping

From 14 June to 1 July 1999, hydrographic and bio–optical measurements were made from the towed undulating vehicle SeaSoar (Pollard 1986). SeaSoar was towed using a hydrographic cable and profiled from the sea surface to 110 m over deep water and to within 10 m of the bottom over the bank. Cycle time in deep water was about 6.5 minutes resulting in surface points being separated by 1.3 km at the ship's typical 7 kt (3.6 m s⁻¹) tow speed. Horizontal separation between profiles at middepth is half this value. Cycles over the shallow bank, 0–70–0 m, took 2 minutes, and so surface points were separated by 420 m. Bottom avoidance was accomplished by using R/V Oceanus' 3.5-kHz Knudsen echosounder as input to the SeaSoar flight control software.

The SeaSoar was equipped with a Seabird 911+ conductivity–temperature–depth (CTD) instrument with pumped, dual temperature–conductivity sensors pointing forward through the nose of SeaSoar. Details of how the CTD data are processed can be found in Barth et al. (2000b). Fluorescence was measured with a WETLabs FlashPak fluorometer (460-nm excitation, 695-nm emission) mounted on the upper tail fin of SeaSoar. For making horizontal maps and vertical sections, the measurements were first averaged vertically to 2-dbar bins and horizontally to 1.25 km. A complete set of the SeaSoar maps and vertical sections can be found in Barth et al. (2000a).

Horizontal water velocity was measured using a 150-kHz narrowband RDI shipboard acoustic Doppler current profiler and is reported in Ullman et al. (2000). The ensemble averaging interval was 150 s and the pulse and bin lengths were both set to 8 m. Absolute velocity was calculated using ship's velocity derived from bottom tracking when available or from GPS P-Code navigation data. An Ashtech 3DF attitudinal GPS system was used to correct for gyrocompass variations.

SeaSoar was towed across the southern flank of Georges Bank in several basic patterns. The first was a “radiator” pattern consisting of 45-km-long legs oriented roughly perpendicular to the local isobaths (Fig. 1). Cross-bank lines were separated by 7.5 km for a total along-bank coverage of 30 km. Each cross-bank line took approximately 3.4 h to complete depending on the phase of the tide. This pattern was centered near the 95-m isobath and sampled the shelfbreak front. After defining the local hydrography, bio-optical, and velocity fields around the shelfbreak front, a subsurface isopycnal float and a 15-m drogued drifter (see section 2c) were released. The subsurface float was tracked acoustically while towing SeaSoar on a butterfly pattern extending 20–30 km across the bank and 10 km along-bank. The pattern was repeated in place or shifted to the west depending on the location of the float. The butterfly pattern took approximately 6.5–8 h to complete depending on what phase of the tide it was started on and was repeated seven times before recovering the drifter and float.

b. Advection with mean and tidal flow

To create synoptic maps, the 2-dbar, 1.25-km averaged vertical profiles from a SeaSoar “radiator” pattern were advected to a common time using a time-independent, horizontally and vertically dependent mean flow and a time- and horizontally varying, depth-independent tidal velocity field. The mean velocity field was obtained by referencing each cross-bank section to the location of the shelf break, here defined as the 130-m isobath, and then averaging the sections together. The tidal model consisted of a least squares fit of the steady plus two semidiurnal (M₂ and S₂) and two diurnal (K₁ and O₁) tidal constituents to the vertically averaged shipboard ADCP velocities collected during the 16 days the ship was on site and from a 15-day cruise on the R/V Endeavor (16–30 June 1999) in the same region. In addition, moored velocity data from an array in the middle of the study region (R. Schlitz 2002, personal communication) were also used in the analysis. The spatial fit, following Candela et al. (1992), used a second-order polynomial for the steady velocity and first-order polynomials (i.e., planes) for the four tidal constituents. The model fit is very similar to an analysis using all available shipboard ADCP data from southern flank Georges Bank Global Ocean Ecosystem Dynamics (GLOBEC) cruises (C. Flagg 2002, also see online at http://www.oasdpo.bnl.gov/mosaic/globec), but we choose to use our fit because it better represents the tidal velocities in the smaller geographic area of interest here. When individual SeaSoar cross-bank vertical sections are compared with subsurface float positions, the 2-dbar, 1.25-km averaged profiles making up that section are all advected to the time of the float fix using the tidal fit. Using a root-mean-square difference between the observed depth-averaged velocity and the tidal fit over the period of the first float deployment of 0.05 m s⁻¹, and the maximum time interval between the closest SeaSoar profile and the float of 1000 s, we estimate that the error in repositioning the SeaSoar profiles is less than about 50 m. This is far less than the error in positioning the float acoustically (see next section).

c. Subsurface float and surface drifter

A satellite-tracked surface drifter and an acoustically tracked subsurface float were used to track fluid parcel trajectories in the vicinity of the shelfbreak front on the southern flank of Georges Bank. The surface drifter (Brightwaters Instrument Corporation Model 115) had a 10-m-long drogue centered at a depth of 15 m. This drifter obtains its GPS location every 30 min and transmits the last seven positions via Argos every 90 s. On the ship, a GONIO Argos receiver was used to intercept the Argos messages when the drifter was within range and a computer attached to the receiver displayed the drifter positions. Argos messages from the drifter were also transmitted to the ship via twice-daily e-mail transfers. The surface drifter was equipped with a nighttime flasher to aid their recovery. Self-contained ONSET temperature loggers were placed at nominal depths of 0, 5, 10, 13.3, 16.7, and 20 m on the drifter. The loggers recorded temperature every 30 s with a resolution of approximately 0.15°C. The time constant of the logger is approximately 1 min.

The subsurface float, the Coastal Ocean Lagrangian (COOL) float (Prater et al. 2003, unpublished manuscript, hereinafter PHRF) was developed to measure diapycnal velocities (a measure of diapycnal mixing) in the coastal ocean. The COOL float incorporates the features of the f/h float (Rossby et al. 1994) and an isobaric vertical current meter (Voorhis 1968) that has been referred to as the VCM. Glass was used for the isopycnal float since it has a very small thermal expansion coefficient. Thus, the float will remain on the same density surface even if the temperature and salinity of the water changes but its density does not. If a water parcel is displaced vertically, it will either expand or contract because of the change in pressure and not change its potential density. Since the glass float is less compressible than seawater, it will not follow this water parcel. Therefore, a compressee is added to the float to make the float's compressibility be that of seawater.

The VCMs are isobaric floats equipped with vanes. Water flowing past these vanes rotates the float and, based on this rotation rate, the vertical velocity of the water can be calculated. Due to the isobaric nature of these floats, they mainly measure the vertical velocity of internal waves. The COOL float, however, being isopycnal, will move up and down with the internal waves. An example of this behavior will be shown later and is detailed in PHRF. The COOL float's pressure case is 2.2 m long and 0.095 m in diameter. Midway along its axis, eight vanes extending 0.34 m from the float's axis are oriented 15° from the horizontal. The rotation of the float is measured with a Precision Navigation Vector 2× compass. A vertical velocity measured from such an isopycnal platform (i.e., by rotation of the float) is a diapycnal velocity. For further discussion of diapycnal velocity, see McDougall (1984). Note that a diapycnal velocity is used to effect diapycnal mixing in isopycnal-coordinate circulation models (Bleck et al. 1992).

The COOL float samples temperature and pressure every 64 s. There was no conductivity sensor on this version of the COOL float. Since the equation of state for the float is known, as is its temperature and pressure, the salinity of the water the float is embedded in can be determined (Boebel et al. 1995). During the initial sinking phase of the float, the compass heading is recorded every 8 s; after this period, it is recorded every 64 s. The small onboard memory of the present float limited missions to less than 2.5 days with these sampling rates. At the end of the mission, the float released its drop weight, the compressee, and surfaced. The float was recovered and the data downloaded.

Tests of a prototype (Rajamony et al. 1996) and several short deployments of the COOL float (Hebert et al. 1997) showed that the rotation rate of the float was linearly proportional to the vertical velocity past the float. The vanes of the float appear to act like the threads of a screw; the lower the pitch, the more rotation of the float for the same vertical velocity. Since there are no moving parts, there is almost no lower velocity limit that the float can measure. However, with a compass accuracy of 2°, samples collected every 64 s limited the accuracy of vertical velocities to approximately 0.07 mm s⁻¹. In determining the vertical velocity and water displacement, the absolute rotation of the float must be determined. For most of the time series, the change in compass heading between adjacent samples was less than 90°. On occasion, the change was greater than 120°. For those cases, the adjacent headings were used to determine which way to rotate the float. In general, it was chosen to rotate the float in the direction of smallest rotation. Therefore, it is possible that the vertical velocity is underestimated (a 180° change in 64 s gives a vertical velocity of 6 mm s⁻¹) at that time and in the opposite direction.

The COOL float has a 12-kHz pinger that is triggered precisely every 8 s. This allow us to track the float through triangulation from the ship. The acoustic signal was received by the ship's hydrophone and displayed on a line scan recorder (LSR). A second precision sound source located on the ship was also recorded by the LSR. The range of the float from the ship was determined by the time interval between the two sources and assuming a sound speed of 1500 m s⁻¹. The position of the float was determined at each closest approach of the float by the ship. The position of the ship and the range to the float at this time and several minutes before and after were used to determine the float's position. There is ambiguity in the possible float position from this calculation. However, with multiple passes in several hours, it is possible to choose the right position. The error in horizontal float positioning is estimated as 200 m. Internal waves contribute “noise” of about 5 m in the vertical, the measured peak-to-peak wave amplitude, to an instantaneous float fix. For plotting float trajectories, the raw subsurface float fluxes are fit using a model consisting of a slowly varying current and a time-varying M₂ tidal flow, both of which are spatially independent.

a. Mean field and internal tide solibores

During the 17-day cruise to the southern flank of Georges Bank in June 1999 the winds were generally weak with speeds rarely exceeding 10 m s⁻¹ and more typically 5 m s⁻¹ or less (Hebert et al. 1999). A satellite sea surface temperature image from 20 June 1999 showed typical cold water over Georges Bank and warm water over and seaward of the continental slope (Fig. 1). A warm-core Gulf Stream ring was located to the southeast of the study region and was clearly interacting with water on the southern flank of Georges Bank as evident by a streamer of cold, bank water being pulled off and around the anticyclonic warm-core ring. A series of three-dimensional mapping activities and Lagrangian drifter and float tracking experiments were carried out both at the location of the tidal mixing front, around the 60-m isobath, and at the shelfbreak front centered around the 90–100-m isobaths. In this paper, only results from the shelfbreak front will be reported.

Before releasing a subsurface float in the vicinity of the shelfbreak front, a 2-day “radiator” grid survey was conducted from 20 to 21 June during a period of low wind speed (<5 m s⁻¹). As described previously, each 2-dbar, 1.25-km, averaged vertical profile was advected to the start time of the grid using both the mean flow and modeled tidal velocity field. Because the mean velocity field consists of a 0.30 m s⁻¹ surface intensified jet centered on the 100-m isobath (not shown), the advected measurement locations are compressed in the along-bank direction (Fig. 2). Line E was slightly delayed relative to line D because of a needed SeaSoar repair. The tide is relatively weak (0.4 m s⁻¹ or less) at this location during neap tides as compared with tidal velocities on the northern flank of Georges Bank, which distort SeaSoar grids more significantly (Dale et al. 2003).

The temperature, salinity, and density anomaly (hereinafter, “density”) fields at 39 m obtained from this synoptic map and rotated 30° clockwise from an east–north coordinate system are shown in Fig. 2. The SeaSoar profile in the southwest corner is arbitrarily chosen as the origin of the rotated coordinate system. A shelfbreak density front separating cold, fresh bank water from warm, salty slope water runs from NE to SW crossing the local isobaths (a front following the isobaths would parallel the x axis in this rotated coordinate system). The warm, salty, and light anomalies onbank from the shelfbreak front, the best example of which is located at 15 km in the x direction and 20 km in the y direction, are created by strong local depressions in the upper layer depth caused by onbank propagation of the internal tide (discussed in more detail below).

A typical cross-bank section from the center of the survey grid (line C) is shown in Fig. 3. A warm, salty layer in the upper 20 m is found over the center part of the cross-bank section consistent with the satellite temperature field (Fig. 1). The presence of relatively cool and fresh upper-layer water offshore of about the 120-m isobath is due to mesoscale variability introduced by the presence of a warm-core ring to the southeast of the study region. At depth, the shelfbreak front separating cold, fresh shelf water from warm, salty slope water is readily apparent. The foot of the front, approximated by the 26.15 kg m⁻³ density contour, is between the 90-m and 100-m isobaths. The “cold pool” is located within 60 m of the bottom and onbank of the shelfbreak front. A plot of buoyancy frequency (Fig. 3f) shows a strong pycnocline at 20–25 m and the shelfbreak front. Along-bank velocity (Fig. 3e) shows a 0.4 m s⁻¹ jet to the southwest centered near the 100-m isobath (cross-bank velocity is shown in Fig. 4c). High chlorophyll fluorescence (Fig. 3d) is found following and just beneath the shallow pycnocline.

The hydrographic and bio-optical fields show a rapid deepening of the base of the surface layer from 25 to 40 m near 40.77°N, with a gradual return to the more typical 25-m depth over about 15 km in the offbank direction (best seen in density, Fig. 3c). This is the manifestation of an internal tide “solibore” (Henyey and Hoering 1997) propagating onbank from its formation region near the shelf break (Lamb 1994). Henyey and Hoering (1997) define a solibore using Baines' (1986) description: “the larger-amplitude wave trains which are of tidal origin and are generated at the shelf break or other topographic features such as ridges and banks.” Henyey and Hoering (1997) go on to say that a solibore will begin life more borelike, but as the energy supply is reduced it will end up more wavelike. We see evidence for both borelike structures and packets of individual waves.

An even stronger signal of the internal tide solibore is evident in the line D SeaSoar/ADCP cross-bank section (Fig. 4, right panels). The thermocline is depressed over 30 m near 40.80°N and the leading edge—the solibore is propagating to the northwest (to the left in Figs. 3 and 4)—is steeper than the trailing edge. The temperature anomaly is formed by subtracting the mean temperature section over the five cross-bank sections (Fig. 4a). On line C (Fig. 4, left panels) there is 2°C warm anomaly at 20–30-m depth near 5 km from the shelf break. The solibore is more evident on line D (Fig. 4, right panels) with a 5°C warm anomaly at 30–40-m depth near 40.80°N and a cold anomaly of up to 6°C near 40.74°N (Fig. 4e). The cold anomaly, or elevation of the thermocline trailing the solibore, is due to on-bank tidal flow encountering the bank edge. The rapid depression followed by a slow elevation trailing the internal solibore is captured in the numerical modeling experiments of Lamb (1994). Cross-bank velocity fields show a mode one vertical structure, onbank flow above the pycnocline and offbank flow beneath, associated with the internal tide solibore on lines C and D (Figs. 4c,f).

The 1.25-km spatially averaged SeaSoar data do not show individual waves within the internal solibore depression. A set of rank-ordered nonlinear internal waves is often observed within the internal tidal bore packet (e.g., Apel et al. 1975; Liu 1988). In contrast, individual waves do not appear in the solibore depression as it propagates up to 35 km onbank from the shelfbreak in the numerical solutions of Lamb (1994). Although the horizontal resolution of the SeaSoar measurements—300 m at middepth for the unaveraged profiles—may not be sufficient to resolve individual waves, we speculate that there may not be individual waves in the depression because these observations are near the generation site of the solibores. That is, rank-ordered individual waves within the packet would be expected farther away from the generation site as nonlinearity and dispersion allow the shelfbreak depression to evolve. A set of internal solitons within a packet was measured by SeaSoar during summer 1996 in 65 m of water over 50 km from the shelfbreak near 70.5°W, to the west of the present study site but with a similar hydrographic and physical setting (J. Barth 2002, unpublished data). Their wavelength was estimated as 650 m—the distance between SeaSoar profiles in water 65 m deep is 150 m—in the range of wavelengths estimated using measurements from a nearby mooring (Levine and Boyd 1998). Given these wavelengths, the present SeaSoar measurements could have marginally resolved individual waves, but we suspect that they are absent because of the proximity to the generation point for the internal solibores.

The internal solibore is also apparent in the 39-m-depth property fields (Fig. 2), although the sharp horizontal fronts are smoothed somewhat by the contouring algorithm. Low-density water at 39 m brought down by the propagating internal tide solibore can be seen at different cross-bank locations on successive sections. Specifically, low density at y = 5 km on line B, at y = 12.5 km on line C and at y = 20 km on line D are manifestations of the same internal tide solibore propagating onbank. Low-density water at y = 32 km on line A is from a solibore formed during the previous tidal cycle to the one that formed the solibore observed on lines B–D.

The propagation speed of the internal tide solibore is estimated as 0.30–0.45 m s⁻¹ by comparing the time and location of the feature on successive sections and accounting for the presence of a cross-bank barotropic tidal flow. Note that because of the presence of the along-bank mean flow, the sections near the middle of the map are nearly collocated when advected to a common time. Therefore, estimates of the internal tide solibore speed are essentially from the same cross-bank section. The observed internal tide solibore phase speed can be compared with theoretical mode-1, semidiurnal internal wave propagation speeds estimated using density profiles from three locations across the southern flank of the bank (see locations indicated in Fig. 3). The estimated theoretical phase speeds are 0.40, 0.46, and 0.67 m s⁻¹ at 40.95°, 40.81°, and 40.68°N, respectively. The phase speed is largest in the deep, highly stratified offbank region and slows as the shallow, more weakly stratified water column on the bank is approached. The observed propagation speed is closer to the slow theoretical phase speeds consistent with the fact that the features were observed onbank of the shelfbreak front. The solibore apparent in the repeated sections is formed near the shelf break and propagates onbank when the on–offbank tidal velocity becomes slack and begins to flow onbank (Lamb 1994; Loder et al. 1992; Dale et al. 2003). The influence of the internal solibore on the behavior of a subsurface float will be described below.

b. First subsurface float deployment

After the shelfbreak frontal region was mapped using SeaSoar and shipboard ADCP, a subsurface COOL float was released on the onbank side of the front. A target density of 25.8 kg m⁻³ was chosen and the float was released at 0008 UTC 22 June 1999 at 40.805°N, 67.089°W. Immediately thereafter a surface drifter equipped with temperature sensors on its subsurface drogue was released at the same location. A series of “butterfly” sampling patterns were then occupied using SeaSoar and shipboard ADCP. The butterfly patterns were designed to give a useful cross-bank representation of the hydrographic and velocity fields, to allow acoustic tracking of the subsurface float through triangulation and to be about 0.5–0.7 M₂ tidal cycles long (Fig. 5). The butterfly patterns were shifted westward as needed as the float moved with the mean flow. The density at 55 m from all the SeaSoar butterfly sampling (Fig. 5a) shows a fairly linear mean shelfbreak front, again running from NE to SW slightly crossing the isobaths. The many repeated crossings of the front that go into this map serve to average out the effects of tides and internal solibores. The 25.8 kg m⁻³ density surface rises from 70 m onbank of the shelfbreak front to about 25 m in the offbank direction (Fig. 5b). Last, a strong temperature gradient exists along the 25.8 kg m⁻³ density surface separating cold shelf water from warm slope water (Fig. 5c).

The surface drifter moved off to the west with the mean flow (average speed of 0.18 m s⁻¹) modulated by the M₂ tidal flow (Fig. 6). In contrast, the subsurface float translated more slowly westward (mean speed of 0.09 m s⁻¹) and moved about 15 km in the offbank (cross-front) direction. As will be demonstrated further, the subsurface float moved up the density surface (Fig. 5b) as it transited in the offbank direction.

A time series of pressure and temperature reveals the behavior of the float during this 50-h mission (Fig. 7a). The float descended through the cold pool, sampling water less than 8°C, before reaching its target density of 25.8 kg m⁻³ at a depth of 80 m (13 m above bottom) and a temperature of 8.5°C.

As described in section 2b, cross-bank SeaSoar hydrographic sections from butterfly lines closest in time to a float fix were advected using a tidal velocity model to the time of the float fix (Figs. 8 and 9). Errors in float position are indicated by the size of the filled black circles and the 25.8 kg m⁻³ density contour is dashed. The first SeaSoar section after the float has reached its target density and has begun to rise is shown in Figs. 8a and 9a. The series of sections shows how the float moved up the frontal isopycnals from a region of cold to warm water.

To determine the degree to which the float was isopycnal and its actual density, the SeaSoar observation closest to the float's location both in depth and in horizontal location can be examined. Details of this comparison can be found in PHRF. The mean density over 14 observations (with a temperature and pressure range of 9°C and 35 dbar) is 25.84 kg m⁻³ with a standard deviation of 0.10 kg m⁻³, confirming the float's target density and that it was isopycnal given the error in positioning the float and in interpolating the SeaSoar observations.

The float record (Fig. 7) contains evidence for the mean vertical velocity over the float's mission, internal solibores, isothermal rebound of the density surfaces after the passage of the internal tide, along-isopycnal diffusion, and diapycnal mixing. We describe each of these in detail in the following sections.

5) Diapycnal velocity

Near day 173.55, the float records a rise in temperature accompanied by a period of significant diapycnal velocity downward past the float (Fig. 7b). The water in which the float was located has undergone heating via a diapycnal flux in the presence of a diapycnal temperature gradient as evident in a vertical section of temperature from this time (Fig. 9b). By integrating w_ω over time, the float is displaced about 13 m relative to the water in which it was released during this diapycnal mixing event (Fig. 7c).

The diapycnal warming event from day 173.47 to 173.55 (Fig. 11) may be examined further in terms of a one-dimensional heat balance. An estimate of the potential influence of horizontal diffusion will be made following this calculation. Since the float is isopycnal, the time rate-of-change of temperature measured by the float should balance the diapycnal velocity times the temperature gradient

View Expanded

From (1) an expression for the diapycnal velocity is

View Expanded

The float heats up by 2.6°C from 173.53 to 173.55, an interval of 1728 s (Fig. 11a). From the closest SeaSoar profile to the float's location, ∂T/∂z = 0.59°C m⁻¹ in the depth range 40–45 m (Fig. 12). The depth range of the mean density plus and minus one standard deviation (25.84 ± 0.10) is 37–43 m. Calculating w_ω from (2) yields an estimate of −2.6 × 10⁻³ m s⁻¹ or −2.6 mm s⁻¹. This is close to the float's average measured w_ω during this isopycnal warming event (173.47–173.55) (Fig. 11b). Given the uncertainty in estimating ∂T/∂z from a nearby SeaSoar profile, we conclude that the measured w_ω is consistent with the physics described by (1).

The float measures a smaller increase in temperature near the beginning of the diapycnal event (173.47–173.49), but then remains isothermal near 9°C even in the presence of nonzero measured w_ω from 173.49 to 173.53. The SeaSoar measurements do not have sufficient horizontal resolution to resolve the temperature gradients to fine enough scale to apply (1) continuously to the float record. However, the lack of temperature change during this period is likely due to the absence of a diapycnal temperature gradient at this temperature (Fig. 12). This cusp in the vertical profile of temperature results from the presence of the cold pool on the outer shelf (Fig. 9).

Returning to the one-dimensional balance used in estimating w_ω from (1), a temperature change of 3.6 × 10⁻⁴ °C s⁻¹ due to horizontal diffusion—our estimate earlier of along-isopycnal diffusion in the frontal region—represents 20% of the observed 1.5 × 10⁻³ °C s⁻¹ temperature change during this event. Thus, there is at least a 20% error in estimates of w_ω based on assuming a one-dimensional balance.

Since the isopycnal float makes a direct measurement of the flow past the float due to diapycnal mixing, we can estimate κ_T (m² s⁻¹), the diapycnal thermal diffusivity as follows. Assuming again a one-dimensional heat balance following the float

View Expanded

equating (1) and (3), assuming that w_ω is independent of z, and integrating between z₁ and z₂ yields

View Expanded

For the diapycnal warming event during the first float deployment (Figs. 11 and 12), using values of w_ω = −2.6 × 10⁻³ m s⁻¹, [T(z₁) − T(z₂)] = 2.6°C, and ∂T/∂z = 0.86 and 0.29 °C m⁻¹ at z₁ = 42 m and z₂ = 46 m, respectively, (see Fig. 12) gives κ_T = 1.2 × 10⁻² m² s⁻¹.

Another period of warming near day 174.525 after isothermal rebound is again accompanied by a negative diapycnal velocity (heat flux down from above). A vertical section just prior to this warming shows the float position relative to the vertical temperature gradient (Figs. 8e and 9e).

c. Second subsurface float deployment

A second deployment of a subsurface float and drogued surface drifter was carried out from 25–28 June 1999. During this period the shelfbreak front rose to about 40 m starting from around 70 m near the foot on the bank (Fig. 13). In contrast to the first float deployment period though, the front deepened again to the south (Fig. 13). This doming of the isopycnals and a more convoluted front (cf. Figs. 5 and 13) are likely due to the mesoscale variability introduced by the presence of a warm core ring offshore. The subsurface float was advected by the semidiurnal tide, but moved less than 5 km from its release point (Fig. 14). The surface drogued drifter also moved with the tide, but was advected initially to the southwest before moving offbank and eastward. This cyclonic movement is consistent with the observed ridge in the density field (Fig. 13) via geostrophic balance.

A time series from the float shows an overall descent from around 40 m at deployment to 60 m at recovery for a net 10 m day⁻¹ downwelling (Fig. 15). Cross-bank sections of density advected with the tide to the time of selected float fixes show that isopycnals in the shelfbreak front bend down on the offbank end of the sections (Fig. 16). The float rides on top of cold pool except at 178.439 59 (Fig. 17d) when the float is located at a local temperature minimum such that warm water lies both above and below. This section is from just after the cooling and warming events described below which rely on the presence of an inflection point in the vertical profile of temperature.

A comparison of the float record with nearby SeaSoar profiles reveals that this float was also isopycnal within the error of the estimates (PHRF). Over 19 comparative points, the mean (standard deviation) is 25.88 (0.05) kg m⁻³ versus 25.84 (0.10) kg m⁻³ for the first release.

Signatures of internal tide solibores are present (e.g., 177.35, Fig. 15), but are not as evident as during the first float deployment because the background density field is more complex. This complexity affects both formation and propagation of the internal tide. Note, however, that during the solibore passage at 177.35 downward velocities reach −0.01 m s⁻¹, the float remains isothermal and there is relatively little measured flow past the float. As during solibore passage during the first deployment, this response is essentially adiabatic. Evidence for an internal tide solibore is present at 5 km on-bank from the shelf break at 178.439 59 (Fig. 16e).

Over the entire record, water moves up past the float such that the float ends up about 13 m below the water in which it was originally released (Fig. 15c). The overall movement of water past the float occurs in a few periods of significant positive diapycnal velocity, for example, 177.85, 178.05, and 178.3–178.5. Cooling occurs during the first and last of these events, so colder water must lie below the float. In fact, the float is riding on top of the cold pool (Fig. 17).

The event from 178.3 to 178.5 accounts for about half of the total relative vertical displacement of water past the float (Fig. 18). Cooling occurs at the start of this event (178.3–178.323; gray vertical bar labeled A in Fig. 18) when the float lies above the cold pool. A cooling of 0.42°C in 2160 s in the presence of a vertical temperature gradient from a nearby SeaSoar profile (not shown) of 0.12°C m⁻¹ and using (2) yields a diapycnal velocity of 1.62 × 10⁻³ m s⁻¹, consistent with the observed motion past the float. Immediately after the float cools because of water moving up from below, the float warms even in the presence of continued motion of water from below (B: 178.323–178.342). This is consistent with the existence of warm water below the float, i.e. below the cold pool (Fig. 17d). The float repeats this behavior once more, cooling (C: 178.342–178.355) then warming (D: 178.355–178.375) with water movement up past the float in the presence of the inflection in the ∂T/∂z profile. Then the float undergoes two cycles of downward movement of water past the float accompanied by cooling (E, G), followed by upward movement of water past the float and warming (F, H). This pattern is a consequence of the float now being located beneath the ∂T/∂z inflection point so that colder water lies above and warmer below.