a. Map 1: 1–2 August

Sampling for map 1 was conducted from 2305 UTC 1 August through 2050 2 August. The detided velocity field (Fig. 4) reveals very high alongshelf gradients and large across-shelf currents. Indeed, the across-shelf component was comparable to the alongshelf component, that is, δ = O(1). We initially interpreted this field as resulting from a malfunction of the shipboard ADCP. However, careful comparison of the ADCP data with other available current measurements taken simultaneously proved their good quality.

Despite the high horizontal gradients, the current field during map 1 had vertical coherence. The currents shown in Fig. 4a were centered at 4.5-m depth within the upper mixed layer (UML). Figure 4b shows currents at 9.5-m depth, which is approximately at the bottom of the UML (although the UML depth exhibited substantial variations within the study area). The currents at 4.5 and 9.5 m were similar and generally lacked significant vertical shear. The only exception occurred near the coast where currents at 9.5 m are within the bottom boundary layer. At 14.5 m the currents decreased (Fig. 4c) and veered counterclockwise.

These highly variable currents suggest that the temporal variability of the flow over a period of observations could be significant and caused some temporal aliasing. However, there were also spatial features in the flow field, which will be described next and which can be determined based on the comparison with other available current data.

In order to distinguish these spatial structures, we will compare the currents from shipboard ADCP data with those at the moorings. The results are shown in Fig. 5 (alongshelf component) and Fig. 6 (across-shelf component). We chose the ADCP bin centered at 9.5-m depth because it is the uppermost level where the current data are available from all three offshore moorings: S3, C3, and N3 (Fig. 2). The shipboard detided currents of map 1 are plotted against time as a pseudo time series;ship track waypoints (wp) are also shown on the plot corresponding to times when the ship occupied those positions.

The foundation of our case is the very good agreement between the current data at the moorings and those from the ADCP measured near the mooring locations. Early on 2 August (0000 UTC) the upshelf flow at mooring S3 exceeded 20 cm s⁻¹ (Fig. 5, bottom). This upshelf current was also registered by the shipboard ADCP in the vicinity of mooring S3 (Fig. 5, middle). At the same time the alongshelf current at mooring C3 (central location) was zero while at mooring N3 (northern location) it was downshelf (5–7 cm s⁻¹). Thus, there was a strong convergence (approximately 25 cm s⁻¹) in alongshelf velocity between southern and northern parts of the study area. This convergence was recorded in the shipboard ADCP currents, as well. The ADCP upshelf current quickly decreased as the ship advanced toward the coast (wp B) and then upshelf (wp C, D: Fig. 5, middle). On transect CD (approximately 0430–0630), the upshelf velocity component was only 3–5 cm s⁻¹. At the same time the upshelf flow at mooring S3 remained 20 cm s⁻¹ while at C3 it was downshelf with increasing velocity from 10 to almost 20 cm s⁻¹. As a result, convergence between southern and central moorings remained strong. Farther north, the alongshelf velocity component was negative (directed downshelf) in the second half of 2 August both according to shipboard ADCP and N3 mooring data.

The mooring data also show some temporal variability, not surprising because of the transient wind event. We calculated the lagged cross-correlation coefficients between alongshelf wind stress at Tuckerton and detided alongshelf current component at the S3 and N3 moorings for the two-day records from 0000 1 August through 0000 UTC 3 August when the wind forcing occurred. The maximum cross-correlation coefficient at S3 was 0.996 while at N3 it was 0.857; time lag was between 9 and 10 h. Clearly, the alongshelf current temporal fluctuations were driven by the local wind forcing.

Nevertheless, the direction of the mean current at the northern location (N3 in Fig. 5, bottom) was opposite to the wind forcing. The opposition is readily explained by the intrusion of buoyant water that is evident in the surface salinity field (Fig. 4a). The intruding water was the leading edge of the Hudson coastal current.

The across-shelf currents revealed by the ADCP data are highly variable alongshelf. Yet, where this component could be checked against mooring data, agreement resulted. As the ship passed mooring S3 early on 2 August, the across-shelf currents recorded by both the ADCP and at the mooring were weak (Fig. 6). As the ship passed mooring C3 near wp F ten hours later, the flow was onshore at about 10 cm s⁻¹. Between these positions the ADCP recorded coherent offshore flow along CD and DE with speeds up to 40 cm s⁻¹ (Fig. 6, middle). This flow was not related to temporal variability. Indeed, both S3 and C3 current data show that temporal variations of the across-shelf currents from 0430 to 0830 (time interval of the measurements along the CD and DE lines) were within the range of 10–15 cm s⁻¹ (Fig. 6, bottom). Accordingly, we interpret this strong offshore flow as a spatial feature. The offshore flow occurred at the same location where the alongshelf component was close to zero (i.e., it changed sign from upshelf to downshelf)— compare Figs. 5 and 6. We conclude that this strong and spatially localized offshore flow was driven by the convergence of the upshelf wind-driven flow from the south and the downshelf buoyancy-driven flow from the north, which are evident in Fig. 5.

Farther upshelf, along FG and GH, the ADCP recorded coherent onshore flow with speeds up to 20 cm s⁻¹. The maximum onshore current occurred along the leading edge of the buoyant intrusion from the north (transect GH). Corroboration of this onshore flow along transects FG and GH is obtained from the simultaneous detided Ocean Surface Current Radar (OSCR) system data available for the northern part of the study domain. The OSCR data showed consistently good agreement with the ADCP measurements through the period of observations (Chant and Münchow 2000). We present these detided OSCR data for two times on 2 August, 1996: at 1000 (Fig. 7a) when the ship was approaching wp F along transect EF and at 1700 (Fig. 7b) when the ship was in the middle of transect IJ. This is the time interval when onshore flow was observed in the shipboard ADCP data. The detided OSCR data (Fig. 7) clearly show a corresponding onshore flow pattern in the surface currents for both snapshots even though the temporal changes were also essential. The width of this flow feature slightly exceeded 10 km.

Currents along transect JK were highly variable. The across-shelf component was toward the coast nearshore while toward the ocean some distance offshore. The spatial location of maximum offshore velocity changed in time over the period of shipboard measurements (1800–2100) according to OSCR data (not shown). This explains some discrepancy between the shipboard ADCP and the N3 mooring data (the former are instantaneous while the latter are averaged over 1 hour).

Thus, while there were essential temporal variations (evident in the mooring and OSCR data), persistent horizontal structure of the currents in the upper layer was present: upshelf and offshore flow in the southern part of the study domain, downshelf and onshore flow in the central part, and downshelf and offshore flow in the northern part. The strongest across-shelf currents tended to be spatially localized with an alongshelf scale O(10 km). The horizontal resolution of the moorings was insufficient to reveal these enhanced flows. However, mooring data were consistent with the shipboard ADCP data measured near the mooring locations. Across-shelf currents strongly depended on the distance from the coast with the lowest values observed at the coastal waypoints (Figs. 4a and 6).

The enhanced across-shelf flows that we observed also exhibited strong vertical shear (cf. Figs. 4b and 4c). Our study area represents the dynamically complicated inner shelf region where near-geostrophic motions interact with surface and bottom friction layers. Accordingly, there are two sources for the vertical shear of the horizontal velocity: (i) surface or bottom stresses and turbulent friction and (ii) horizontal baroclinic pressure gradient. On the inner shelf, friction layers merge and the friction effects produce a continuous change of the velocity vector in the vertical on the scale of the water depth. Under the observed conditions the stratification could reduce the turbulent friction coefficient in the pycnocline. However, the resulting velocity shear still should be almost uniform through the whole water column. This feature was illustrated by Lentz (1995) in his example of the exponential viscosity profile with close to zero value in the middle of water column. On the other hand, the water in the study area was seasonally stratified with a clear two-layer structure, a thin and very intense pycnocline separating surface and bottom mixed layers. The only exception occurred nearshore at wp J where the downwelling regime prevailed with the continuous stratification. Thus, we expect that the baroclinic shear for most of the domain would have a local maximum in the vertical within the pycnocline while the frictional shear would be almost uniform.

We analyzed the horizontal distribution of the mean vertical shear (averaged over the whole water column) and the maximum vertical shear measured by shipboard ADCP during map 1. The mean vertical shear varied between 0.025 and 0.04 s⁻¹ and did not show substantial horizontal changes, while the maximum vertical shear had greater range of horizontal variability with several spatial maxima: in the vicinity of wp C–E (0.15–0.20 s⁻¹), wp G (>0.15 s⁻¹), and in the middle of transect JK (>0.15 s⁻¹). In order to assess the relative importance of the baroclinic vertical shear, we present the horizontal distribution of the maximum vertical shear normalized by the local mean vertical shear (Fig. 8). For the areas with high values of this normalized maximum shear (say, greater than 4) baroclinic vertical shear prevails while for the areas with the lower numbers (2–3) frictional shear is more significant. In Fig. 8, we plotted only contours 4 and higher in order to emphasize the flows with baroclinic shear. There are two absolute maxima of the normalized shear: one is in the vicinity of wp C (>7) and the other at wp G (>6). These are also the two locations of maximum detided across-shelf velocities: offshore at wp C (almost 40 cm s⁻¹) and onshore at wp G (about 25 cm s⁻¹)&mdash≃e Figs. 4a,b and Fig. 6.

In the description of the velocity field we pointed out upshelf flow in the southern part and downshelf flow in the northern part. The vertical density structure along transects AB (southern part) and IJ (northern part) demonstrate essentially different dynamical regimes (Fig. 9). In the southern part of the study area the wind stress generated upshelf flow and established an upwelling regime with the pycnocline reaching the surface at the coast on transect AB. Along transect IJ isopycnals instead tilted toward the bottom near the coast. Buoyant water (density less than 1021 kg m⁻³) remained in contact with the bottom near the coast. Thus, the duration and strength of the transient upwelling-favorable wind event was insufficient to detach the advancing buoyant intrusion from the coast and to reverse the buoyancy-driven southward flow.

In Fig. 9 we also plot the vertical distribution of the Richardson number Ri:

View Expanded

where ρ is the water density, u and υ are the horizontal velocity components, g is the gravitational acceleration;subscript z denotes partial differentiation in the vertical (with positive direction upward). We estimated Ri for every vertical bin except those closest to the surface and bottom. We applied central differences to approximate vertical derivatives; thus, each Ri value was estimated over a 2-m depth interval. The minimum Ri value contoured in this plot is 10, because values below 10 turn out to represent the background level. High values (Ri > 10) imply the effective limit for penetration of turbulent mixing caused by the vertical shear of velocity.

Transect AB had the typical vertical structure for upwelling: the pycnocline approached the surface at the coast, both surface and bottom mixed layers were well pronounced, and high values of Ri were observed within the thin, intense pycnocline. Transect IJ had different features: at the coast the isopycnals tilted down toward the bottom and did not form a sharp, thin pycnocline. Instead, the density changed from the surface to the bottom rather gradually. These features are typical for a buoyancy driven coastal current or downwelling regime, when lighter water is driven offshore in the bottom boundary layer, thus mixing with the overlying water and reducing the stratification (Chapman and Lentz 1994). Within 10 km of the coast there were few high values of Ri, implying significant vertical mixing. Farther offshore along transect IJ, the pycnocline structure corresponded to upwelling-favorable wind forcing. Indeed, Fig. 4a shows offshore flow of the surface currents along transects IJ and JK some distance offshore. As a result, the pycnocline was deformed into a domelike shape. The dome was induced by the UML divergence caused, in turn, by the joint action of wind-driven offshore Ekman transport near the surface and the buoyancy-driven nearshore downwelling regime. This latter one drove offshore flow at the bottom with compensating onshore flow at the surface. The center of the dome was at longitude 74.12°W (Fig. 9, lower panel). This dome was not exposed to wind-induced mixing. Instead, the lighter, buoyant water was driven offshore above the “main” pycnocline (associated with the seasonal thermocline, isopycnals 1022–1024). Note the high values of Ri on the top of this dome (Fig. 9).

The baroclinic Rossby radius (R_i) is an appropriate across-shelf scale for baroclinic motions trapped near the coast. For a two-layer approximation, R_i is

View Expanded

where h₁ and h₂ are the thicknesses of the surface and bottom layers, f is the Coriolis parameter and g′ is the reduced gravity of the internal interface, g′ = gΔρ/ρ₀, Δρ is the density difference between the layers, and ρ₀ is the reference density. Taking the total depth to be 20–25 m with interface in the middle of the water column, Δρ = 3–4 kg m⁻³ with the reference density 1022 kg m⁻³ and f = 0.92 × 10⁻⁴ s⁻¹, the estimation for R_i is 4.2–5.4 km. From the velocity field shown in Fig. 4a, the baroclinic Rossby radius seems to be an appropriate scale for the alongshelf adjustment of the observed mesoscale flows.

b. Maps 2 and 4: 3–6 August

After map 1 was completed, the wind slackened so that map 2 was conducted under conditions of light wind forcing from 0304 3 August through 0300 4 August (Fig. 3). Figure 10 shows the detided velocity field in the UML derived from the ADCP shipboard measurements. Compared to the previous map the currents were weaker except in the northern part of the domain where the buoyant intrusion was advancing. The across-shelf component of the flow field is reminiscent of the previous map with offshore flow within the advancing edge of the buoyant intrusion, onshore flow in the central part (transect EF), and offshore flow in the southern part (wps A, C) of the study domain. These flow features moved southward relative to map 1 but they retained their position relative to the advancing buoyant intrusion. Indeed, the surface salinity distributions indicate that the intrusion in the northern part was gradually moving downshelf (cf. Figs. 4a and 10). At this time the leading edge reached transect FG. As a result, the system of mesoscale currents associated with this plume also moved downshelf.

The most unusual and interesting feature of the velocity field is the cyclonic circulation within the buoyant intrusion. Most theoretical and modeling studies predict an anticyclonic regime of circulation around the bulge of buoyant water near the plume source: Garvine (1987), Chao (1988a), Yankovsky and Chapman (1997). However, all these papers deal with plumes unforced by wind.

The vertical transect of detided alongshelf velocity and density along JK is presented in Fig. 11a. The density field was averaged (“soothed”) in order to remove smaller scale features, which are typically associated with internal waves. These features are evident, for example, in Fig. 9 and can degrade the estimation of the geostrophic velocity component. The isopycnal surfaces are dome shaped. The formation of this dome structure was evident earlier, during map 1 (Fig. 9); now the dome’s center extended seaward to longitude 74.0°W. Moreover, both positive (upshelf) and negative (downshelf) velocities are consistent with this density structure in the manner of geostrophic adjustment. Maximum velocity and vertical shear occurred precisely where the maximum horizontal density gradient was observed, that is, on both sides of the dome (Fig. 11a). The alongshelf velocity shear was approximately 60 cm s⁻¹ over 22 km. Thus, relative vorticity of this cyclonic feature is about 0.3f, where f is the planetary vorticity.

We estimated the geostrophic vertical shear of the alongshelf current based on thermal wind balance (Gill 1982) and show the results in Fig. 11b. As a reference level, we used the observed alongshelf velocity at 6.5-m depth and calculated the geostrophic velocity downward. The geostrophic velocity gives good qualitative agreement with the observed velocity. Moreover, even absolute values of the geostrophic shear are close to those observed. The only exception occurs at 73.9°W where the geostrophic vertical shear substantially overestimates the observed. Here a very intense pycnocline (3 kg m⁻³ over 5 m in vertical) leads to unrealistically high thermal wind shear. The important feature is that the actual velocity shear nearshore (depth 15–20 m, where the downwelling regime exists) is in good agreement with the geostrophic estimation, showing no indication of bottom stress influence. This resembles the dynamics of the slippery boundary layer (Garrett et al. 1993; Chapman and Lentz 1997), when the combination of density gradient and sloping bottom significantly reduces the bottom stress.

The dome of the pycnocline was formed due to divergence in the surface Ekman layer associated with the competition between two opposing forcings: wind and buoyancy. The upwelling-favorable wind produced offshore Ekman transport in the surface boundary layer while buoyancy forcing still maintained a downwelling regime near the coast (Fig. 9). This downwelling circulation consisted of offshore transport in the bottom boundary layer with compensating onshore flow at the surface. The divergence in the surface Ekman layer caused the rise of the pycnocline at this location. The divergence point can be seen in Fig. 4a on transect JK (approximately halfway between wp J and mooring N3). The geostrophic adjustment of the flow field to this density structure resulted in the observed cyclonic current within the intrusion. Even though this cyclonic feature was formed on the inner shelf where turbulent friction dominates, it seems to be a robust feature. This dome was shielded from erosion by wind-induced mixing by the buoyant water on top of it that was spread offshore by the upwelling-favorable wind. At the same time, the actual alongshelf velocity at the bottom (Fig. 11a) was close to zero due to strong vertical shear. This cyclonic flow, thus, was not exposed to strong bottom friction.

The velocity field observed during map 3 (4–5 August) was qualitatively similar to map 2. The buoyant intrusion advanced farther downshelf and the cyclonic circulation within it persisted. The final map, map 4, was conducted from 0630 5 August through 0335 6 August. During the second half of this map, the upwelling-favorable wind slightly increased (Fig. 3), which could have amplified cyclonic flow within the intrusion. As in the two previous maps, cyclonic flow remained the most prominent feature of the velocity field in the UML (Fig. 12). Below the pycnocline (15 m and deeper) this feature could hardly be seen, revealing the essentially baroclinic nature of this cyclonic flow. By this time, the buoyant intrusion had advanced to transect DE (Fig. 12). The offshore extension of the intrusion did not change from that of map 2 due to the lack of upwelling favorable winds between 3 and 5 August (Fig. 3). There is good agreement between the cyclonic pattern of UML currents and the configuration of the intrusion. Moreover, there is evidence of cyclonic circulation in the surface salinity distribution: a tongue of fresher water detached from the coast and turned offshore near wp F (Fig. 12). At wp E the current flowed toward the coast and then turned in the downshelf direction along transect DE. This might be evidence of the anticyclonic flow surrounding the bulge of buoyant water, as suggested by numerous theoretical and modeling studies.

The vertical structure of the density field reflected the different dynamical regimes in the northern, central and southern parts of the study area. Along transect AB, the “standard” water structure was observed: a flat and sharp pycnocline at the depth of 10–15 m separated surface and bottom mixed layers (Fig. 13a). Along transect EF flow convergence was evident in the across-shelf velocity component (Fig. 12) that resulted in continuously stratified water structure from surface to bottom (Fig. 13b). Finally, isopycnals on transect IJ (Fig. 13c) again revealed the dome structure associated with cyclonic flow within the plume. Over the top of this dome there was a thin layer of fresher buoyant water. These three very different across-shelf density sections well illustrate the striking alongshelf variability. The three sections of Fig. 13 are separated alongshelf by only about 20 km.