E. Böhm1 L. J. Pietrafesa

As described by Csanady and Hamilton (1988), the flow regime over the slope of the southern Middle Atlantic Bight (MAB) includes a current reversal in which southwestward flow over the upper and middle slope becomes entrained in the northeastward current adjacent to the Gulf Stream. In this paper we use satellite-derived data to quantify how this current system is impacted by lateral motions of the Gulf Stream. In our analysis, the Gulf Stream’s thermal front is delineated using a 2-year long time series of sea surface temperature derived from NOAA/AVHRR satellite data. Lateral motions of the Gulf Stream are represented in terms of temporal variations of the area, east of 73^o W, between the Gulf Stream thermal front and the shelf edge. Variations of slope water flow within this area are represented by anomalies of geostrophic velocity as derived from sea level anomaly time series determined from TOPEX/POSEIDON satellite altimeter data. A strong statistical relationship is found between Gulf Stream displacements and parabathic flow over the continental slope. It is such that the southwestward flow over the slope is accelerated when the Gulf Stream is relatively far from the shelf edge, and is decelerated (and perhaps even reversed) when the Gulf Stream is close to the shelf edge. This relationship between GS displacements and parabathic flow is also observed in numerical simulations produced by the Miami Isopycnic Coordinate Model. In qualitative terms, it is consistent with the notion that when the Gulf Stream is closer to the 200-m isobath, it is capable of entraining a larger fraction of the local water masses. Alternatively, when the Gulf Stream is far from the shelf-break, more water is advected into the MAB slope region from the northeast. Analysis of the diabathic velocity component indicates that much of the cross-slope transport by which the southwestward flow entering the study region is transferred to the northeastward flow exiting the region occurs in a narrow band roughly centered at 36.75^o N, order 150 km north of Cape Hatteras. This transport, and thus the cyclonic circulation of the southern MAB, strengthens when the Gulf Stream is relatively close to the shelf edge, and weakens when the Gulf Stream is far from the shelf edge.

The oceanographic region near Cape Hatteras (CH), North Carolina (Figure 1) is highly complex, encompassing a number of water masses and current systems. Extending offshore from CH is an eastward oriented shoal, called Diamond Shoals, that serves as a partial barrier to along-shelf flow (Pietrafesa et al., 2002) and separates two oceanographically different regions: the Middle Atlantic Bight (MAB) and the South Atlantic Bight (SAB). Near Diamond Shoals, the Gulf Stream (GS) typically separates from the continental margin and flows into deeper water. In the region of this separation, lateral displacements of the GS tend to be small (Sun and Pietrafesa, 1995), with a standard deviation of order 10 km (Pickart and Watts, 1993). However, meanders grow quite rapidly downstream, and at 70° W the standard deviation of their lateral displacements is greater than 50 km (Halliwell and Mooers, 1983). The seasonal as well as inter-annual behavior of GS lateral movements has been quantitatively addressed with field measurements (Halkin and Rossby, 1985; Tracey and Watts, 1986; Watts et al., 1995) and with remotely sensed observations (Kelly and Gille, 1990; Lee and Cornillon, 1995, 1996a, 1996b). In general, the GS tends to be north (south) of its mean position in the fall (summer), a seasonal cycle that has been associated with large scale wind and heating cycles (Worthington, 1976, and Fu et al., 1987).

At the shorter spatial scales of lateral meandering (order 10 km), both the vertical and horizontal structure of the GS have been found to be dependent on the distance of the GS’s north wall from the steep bathymetry near CH (Pickart and Watts, 1993). As the GS migrates offshore or away from the coast, its core velocity increases while its transport remains essentially constant. As the GS migrates onshore, it deepens but flows less swiftly. This structural change has only been observed in the proximity of CH where the bottom slope is greater than the GS interfacial slope and has been attributed to potential vorticity conservation of the deep water columns as they migrate over a bottom steeper than the GS interface (Sun and Pietrafesa, 1995).

Another impact of GS meandering on conditions in the slope region is evidenced by discharges of water from the GS near CH (Churchill and Cornillon, 1991). This type of phenomenon is expected to affect the manner in which the volume between the GS and the shelf break is conserved.

GS meandering is also associated with vertical water mass movement. As its interface tilts in response to the lateral excursions, a retreating GS can induce upwelling. However, Halkin and Rossby’s (1985) Pegasus observations indicated a decrease of the magnitude of such structural GS fluctuations, and consequently less upwelling (or downwelling), downstream of CH. Their observations revealed that at 73° W the kinetic energy due to structural changes of the GS only accounts for a third of the total eddy kinetic energy (1500 cm² s^-2) and is only three times as large as mid-ocean values. Moreover, Watts et al. (1995) found that at a longitude of about 74° W, the GS meanders tend to occur without changing the interface tilt.

Lee and Cornillon (1996a) and Kontoyiannis and Watts (1994) show that in the region 74°-70° W, progressive GS meanders have a broad range of periods, from days to years, with wavelengths of 200 to 1100 km. Standing meanders are responsible for the standing wave pattern (in-phase GS displacement) between CH and 69° W that has a range of periods between 0.5 and 1.0 year. For yet longer meanders, the propagation is retrogressive (in the presence of planetary beta and amplitude-dependent effects). The low wave numbers thus contain stationary waves while the higher wave numbers are dominated by downstream propagating waves.

GS meanders and GS warm-core rings are likely responsible for most of the sea level variation over the MAB outer slope and continental rise. However, at the southwestern extreme of the MAB slope region warm-core rings are rare. From a census of GS warm-core ring activity, Auer (1987) found that only 13 of 115 warm-core rings observed over a 5-year period entered the longitude band between 72.6° and 75^o W.

The southern MAB slope region is thus an area marked by a transition between a topographically steered regime of the GS and an unstable (both barotropically and baroclinically) jet downstream of CH. The circulation of slope water in this region, and further to the east, was examined in a landmark paper by Csanady and Hamilton (1988). Their proposed circulation scheme contains a cyclonic gyre in the western Slope Sea, the oceanographic region between the GS’s northern edge and the continental shelf edge. Over the southern MAB slope, this gyre takes the form of a southwestward flow adjacent to the continental margin and a return northeastward flow along the northern GS edge. Subsequent drifter studies have confirmed this basic circulation pattern (Dragos et al., 1996; Berger et al., 1996; Gawarkiewicz and Linder, this issue), but offered little evidence of a permanently closed gyre in the western Slope Sea. Rather, observed recirculation of drifters in the Slope Sea has been intermittent, with a roughly 20 % occurrence (Berger et al., 1996; Gawarkiewicz and Linder, this issue).

Previous studies have shown that lateral displacements of the GS can significantly impact the circulation northeast of CH. Parker (1976) found a close correlation between changes in transport at an environmental buoy located over the MAB continental rise and lateral movements of the GS. In particular, the transport tended to increase as the GS moved onshore toward the buoy and decrease as the GS receded offshore. In a later study, Bane et al. (1988) examined the strength of currents measured over the upper continental slope of the southern MAB (at 36° 40΄ N and 73° W) in relation to GS distance from the shelf-edge. Their analysis indicated that the southwestward flow over the upper slope tended to intensify as the GS approached the shelf edge. This was evidenced by the behavior of monthly mean currents measured at an upper slope location. These were directed toward the southwest and of 30-40 cm s^-1 magnitude when the GS was within 150 km of the shelf edge, but were near zero when the GS was 300 km from the shelf edge. However, these findings are in conflict with the conclusions of a recent study by Dong and Kelly (2003) in which strength of the current over the MAB slope, as derived from analysis of TOPEX/POSEIDON (T/P) altimeter data, was related to the position of the GS. They found that the large-scale southwestward flow over the MAB and Scotian slopes tended to intensify as the large-scale position of the GS shifted offshore.

In this study, we focus on the impact of GS movements on the circulation over the slope of the southern MAB. Our operating hypothesis is that the lateral displacements of the GS alter sea level, and its gradient, and therefore the circulation over the slope. Moreover, we assume that these lateral displacements can be accurately represented by changes in the area between the GS and the 200-m isobath. The main objective is to quantitatively assess the slope water response, if there is one, to these displacements. In particular, we focus on the deformation of sea-level topography over the slope, and the corresponding geostrophic velocity, as it might adapt to lateral fluctuations of the GS. To test our hypothesis, we quantitatively compare the time varying GS thermal front location, derived from Advanced Very High Resolution Radiometer (AVHRR) sea surface temperature (SST) observations, with the sea-level anomaly (SLA) derived from the U.S. National Aeronautics and Space Administration and French T/P altimeter data. In Section 2, we describe the data sets used and the methods that are applied to them. The primary results are discussed in Section 3 and consist of: time series of spatially averaged velocities and SLA in the slope region, analysis of the spatial correlation of these velocities to the GS lateral fluctuations. In Section 4 these results are discussed and conclusions are put forth.

Our region of study is bounded by: the 200-m isobath to the west; the 73° W meridian to the east and the moving GS front to the south. The average size of this domain (shaded in Figure 1) is approximately 26,000 km². Because this domain is roughly triangular in shape, the area between the GS front and the shelf edge, rather than the distance between the two, was chosen to represent the time-dependent forcing created by GS displacements. Such an integral, scalar quantity has the advantage of being nearly independent of the smaller scale meanders that are known to affect the shape of the GS thermal front, but not its larger scale position.

Production of a GS frontal position time series was accomplished using AVHRR SST images. These images, obtained from the NOAA polar orbiting satellites, have a frequency of twice daily and a spatial resolution of 1.1 km at nadir. Due to obscuration by clouds, which in the vicinity of CH was found to be significant (order 80% of the time over the period of our study), it was necessary to use composite imagery. Compositing AVHRR imagery is a common practice for eliminating cloud cover, and typically encompasses images over periods of 1 to 7 days (Babin et al., 2004). The trade-off in selecting the period length is between reducing cloud cover and smearing oceanographic features of interest. In the present work, compositing was done by retaining the warmest pixels observed at each geographical location over a 2-day interval.

For our analyses, we produced a GS frontal position time series spanning the period from October 1992 through December 1994 with individual position tracks roughly corresponding to T/P passes through the study region (Figure 2). This was accomplished with two different methods. For cycles up to July 1993, the space-time interpolation used by Chin and Mariano (1997) was applied. For the period July 1993 through December 1994, we subjectively digitized the GS (thermal) front as no space time interpolated frontal positions were available from Chin and Mariano (1997). The digitization was done with the composite imagery by an operator who moved a cursor over the area of maximum temperature gradient associated with the GS front. The result was an array of latitude and longitude of frontal locations separated by spatial gaps due to the residual cloud cover. This array was transformed into a time series of frontal positions by manually combining the front realizations into continuous lines centered on the respective T/P pass times. It should be kept in mind that for visible portions of the GS front, the subjective digitization error is of the order of a few km. This error is less than the error induced by compositing images obtained at different times and then considering them to be concurrent.

From the resulting frontal position time series we generated a probability distribution of GS front location. This was done by counting the number of frontal positions in each 0.25^o x 0.25^o square of our region of interest, and then dividing the “number of frontal positions” field by its maximum value. The resulting probability distribution (Figure 1), exhibits a downstream widening about its axis (thick solid line) of highest probability.

It is important to note that our inspection of the SST time series revealed no GS warm-core rings entering our study area (west of 73^o W) during our study period (October 1992-December 1994). Warm-core rings did thus not directly impact sea level or currents examined in our study.

The T/P altimeter uses a dual-frequency radar system to determine the height of the ocean surface from an altitude of 1336 km. The satellite is in a 10-day repeat orbit. The data utilized in our analysis consisted of measurements of the height of the ocean surface interpolated to a 0.25^o resolution global grid (see Hendricks et al., 1996 for a description of the processing applied to the data set). They were retrieved in network common data format (netCDF) files from the Colorado Center for Astrodynamic Research (CCAR) web server (http://www-ccar.colorado.edu). They were relative to the JGM-3 gravity model, with tidal corrections computed using the Desai and Wahr (1995) tide model. Standard Geophysical Data Record (GDR) corrections were applied, including corrections for inverted barometer, wet and dry troposphere, ionosphere, and electromagnetic bias. The resulting data accuracy was better than 5 cm point wise. For the shelf regions this accuracy is an upper limit since the tidal correction is not as accurate there (see Dong and Kelly, 2003).

We specifically utilized a data set of SLA taken as the difference from a 2-year mean sea-level surface as computed from the 1993-1994 T/P altimetry. Interpolation of sea-level data, in the region in between tracks, was necessary since altimetric data were collected along tracks that were separated by large horizontal distances (Figure 2). For the T/P data set this interpolation was carried out using an iterative objective analysis scheme (Cressman, 1959) using each 10-day coverage as a snapshot of the ocean sea-level field. Visual inspection of the complete T/P time series indicated that the sampling afforded by two intersecting tracks was fortuitously well-suited for our region of interest. The tracks divided the active area into sub-areas that were of similar size, and for which the distance between the points along the T/P tracks, and the remaining points, was not very large compared to the interpolation spacing. The two tracks intersecting near the middle of our area were sampled within a day of each other (the 9^th day of each T/P cycle) and can thus be considered simultaneous. It is apparent, however, from the orientation of the tracks that the alongshore SLA was better sampled than the across-shelf SLA. We address this problem in more detail below.

2.3. Parabathic and diabathic velocity fields

From the SLA fields it is possible to generate a time series (sampled every 10 days by the T/P altimeter) of derived quantities, such as geostrophic velocity anomalies associated with the sea level departure from the two-year (1993-1994) mean. We focus on the parabathic (along bathymetry with the coast on the left, i.e., positive to the northeast) and diabathic (normal to the parabathic direction and positive down-slope) components, with the bathymetric orientation determined using the ETOPO5 (Earth Topography - 5 minute) digitized bathymetry. This topographic database is constructed by averaging, at 1/12^o horizontal and 1-m vertical resolutions, several uniformly gridded data bases.

To compute the parabathic and diabathic components of the residual geostrophic velocity field, a two-step procedure was followed. First, we computed the eastward and northward component of the geostrophic velocity anomaly for each of the complete 10-day cycles. This was done by taking the north and east derivative of sea-level and using the geostrophic relationships: u= -(g/f) dh/dy and v= (g/f) dh/dx, where g is the gravitational acceleration, f is the Coriolis parameter, h is sea-level, x is east coordinate, y is north coordinate, and u and v are the east and north components of velocity anomaly, respectively. In the second step, the velocity at each grid point was projected along the parabathic and diabathic directions. The diabath at each grid point was found as the direction of maximum negative slope (i.e., the free-fall line) of the sea floor from the meridional and zonal slopes of the bottom. The parabath was simply defined to be 90^o to the left of the diabath. Since the bathymetric data set ETOPO5 is gridded at a higher resolution (1/12°) than the altimeter data (1/4°), a bilinear interpolation was performed to subsample the bathymetry accordingly.

2.4. MICOM simulation data

The Miami Isopycnic Coordinate Model (MICOM, Bleck et al., 1995) was employed to explore the simulated response of the slope current regime to GS forcing. This step was important because, while the altimeter yields SLA only along its track and provides limited resolution of SLA in the onshore direction, the MICOM simulation is uniformly gridded (the generation of the simulated data set is described in Bleck et al., 1995). In particular, we were interested in a comparison of our observations with the much higher resolution (0.08° spatial and 0.25 days temporal) data available in the model. MICOM offered an advantage over the use of T/P sea level data in that we were able to directly determine the position of the GS front from the MICOM simulated sea level. A limitation of MICOM is that it is forced by monthly mean wind fields and energy budgets.

Exploiting the fact that the sea level is lower inshore of the GS, we used a histogram of sea level values for each simulated field to discriminate between the GS and the inshore region, thereby obtaining a border for each field. Since the thermal and sea level fronts are well correlated, and differ primarily by their geographical location (about 40 km according to Kelly and Gille, 1990), the discrepancy in the area estimates based on sea-level and those based upon thermal information is expected to be roughly constant, and thus have no impact on the rate of change of area.

3. RESULTS

3. 1. Gulf Stream frontal probability distribution

Since we are using simultaneous observations of the GS front and the SLA on the inshore side of the front, it is appropriate to first describe the location of the GS as it relates to SLA statistics.

The spatial probability distribution of the GS front (Figure 1) shows a widening envelope downstream from CH to 74° W. East of 73° W, the eastern limit of our study area, the axis of high probability veers from a northeastward to an eastward orientation. The distribution exhibits an asymmetry relative to the axis of high probability, with a greater widening of the envelope on the southern side of the axis. Given the steeper bathymetry north of the GS, this asymmetry is not surprising since topographic steering becomes weaker as the GS moves over a deeper ocean floor to the south and downstream.
3.2. Sea level anomaly

The two years of data used (1993-1994) in the analysis combining SLA and GS front edge information are presented in Figure 2. SLA is shown in color-coded form, from blue (low) to red (high). Also shown in each panel are the T/P tracks (dotted lines), and the GS thermal front edge (solid line) at the time of the T/P pass and the 200-m isobath (thick solid line). Note that starting on 16 July 1993, the digitized GS front is not as smooth as in the preceding cycles. This is due to a shift in the technique used in determining frontal location, as discussed in Section 2.1. For the earlier cycles, the space-time interpolation applied by Chin and Mariano (1997) filtered out some of the smaller scale meanders, giving a smoother appearance to the GS paths.

In interpreting Figure 2, it is important to keep in mind that the sign of the SLA (which can be positive or negative) is predominantly controlled by the deviation of the GS from its average position. For example, a southward displacement of the GS from its two-year average position will tend to depress sea-level and thus leave a negative (blue) SLA to the left of the mean GS frontal position. Conversely, a northward displacement will tend to raise sea-level producing a positive (orange-red) SLA to the right of the mean GS frontal position.

The SLA time series illustrates the annual cycle in sea-level (high in summer and low in winter) in the spatial domain between 35° and 39° N and 77° and 73° W. Also revealed is an inter-annual variation (1994 is higher than 1993) previously discussed by Pietrafesa et al. (1981).

The root mean square (RMS) SLA (Figure 3a) is a measure of the vertical oscillations of the sea surface. The RMS SLA indicates a sea-level variability that is enhanced over the shelf, partially as a result of the inaccurate tidal correction, and is also enhanced at about at 50 km south of the GS frontal axis (as defined by the most likely position of the GS front, dashed line). The reduced variability (<12.5 cm) in the slope region west of 74^o W is consistent with the relatively small GS lateral displacements between 75° and 74° W as indicated by the narrow probability distribution of the GS front in this longitude range (Figure 1). Further downstream, the amplitude of the RMS SLA increases to over 17.5 cm as the oscillations of the GS increase in amplitude.

The eddy kinetic energy (EKE) indicates the distribution of the eddy component of the geostrophic velocity associated with the flow. The EKE per unit mass is computed by taking ½ of the sum of the squares of the meridional and zonal components of the geostrophic velocity anomaly described in Section 2.3. In general, EKE (Figure 3b) tends to be highest where the slope of the RMS SLA is steepest. The EKE distribution shows a ridge of values greater that 400 cm²s^-2 along the GS front axis.

3.4. The circulation induced by GS displacements

In order to quantitatively address the dependence of the slope circulation (derived from the T/P SLA) on the GS lateral displacements (derived from the thermal front observations), we define the area between the GS and the 200-m isobath as the independent variable. The surface geostrophic velocity anomaly and the SLA are then two related dependent variables. To simplify the comparison, the surface geostrophic velocity anomaly field is averaged component-wise within the area. The resulting time series are presented in Figure 4. The SLA averaged within the GS-to-shelf edge area (Figure 4b) is characterized by a seasonal signal (high in summer, low in winter) due to the steric adjustment related to the Atlantic Ocean’s thermal expansion/contraction cycle in the mixed layer. Note, however, the generally higher positive sea levels in 1994 relative to 1993, seen also in the SLA time series (Figure 2, Section 3.2).

A seasonal signal is present in the first half of the area time series (Figure 4a), which exhibits a minimum in November 1992 and a maximum in April 1993. This is consistent with the expected seasonal variation in GS position (Lee and Cornillon, 1995) in which the GS at its northernmost position in fall (minimum area) and its southernmost position in spring (maximum area). However, such a behavior is not observed the following year, indicating the year-to-year variability.

The area-averaged diabathic and parabathic components of geostrophic velocity anomaly (Figure 4c,d) differ in that the parabathic component (also shown as a dashed line in Figure 4a) is strongly anti-correlated with the area time series, while the diabathic component does not show any obvious visually-inferred correlation. In quantitative terms, the correlations of area with the parabathic and diabathic velocity components are -0.79 and -0.31, respectively. Both are significant at the 95% confidence level. The high negative correlation between the parabathic velocity component and the area is an indication of negative parabathic flow anomaly (stronger flow to the SW) associated with periods when the GS is far from the shelf and positive parabathic flow anomaly (stronger flow to the NE) associated with periods when the GS is closer to the shelf. The lower, but statistically significant, negative correlation between area and the diabathic velocity component implies that diabathic transport to deeper water (off-slope) tends to be enhanced when the GS is relatively close to the shelf-edge.

In order to further clarify the observed temporal correlation between the velocity anomaly field and GS lateral displacements quantified by the area changes, we consider the spatial characteristics of the geostrophic response to the area changes. This is done with the aid of contour plots of the correlation between area and the two components of the geostrophic velocity anomaly (Figure 5a,b). The correlation distribution of the parabathic velocity anomaly component (Figure 5b) with area includes a broad region of strong negative correlation over the continental slope, and mostly to the north of the most likely position of the GS front. This broad region of negative correlation indicates that when the area between the GS and the 200-m isobath is smaller (greater) than average, the along-isobath geostrophic velocity is stronger (weaker) in the positive direction (to the northeast) than average. In more general terms, the correlation indicates that large areas between the GS front and the shelf edge correlate with enhanced inflow into the slope region, via a strong southwestward flow along the slope, while small areas correlate with either reduced inflow to, or outflow from, the slope region. This is consistent with the notion that when the GS is closer to the 200-m isobath, it is capable of entraining a larger fraction of local water masses along its edge and transporting these to the northeast. Alternatively, when the GS is far from the shelf edge, more water is advected into the region from the northeast. Episodes in which there is either minimum influx to, or outflow from, the slope region associated with a GS advance to the shelf edge are nicely illustrated by the spatially-averaged parabathic velocity anomaly time series (Figure 4a,d). One such episode occurs during Nov. 1992, when the GS is at its closest approach to the shelf edge (as indicated by the area time series) and the spatially-averaged parabathic velocity anomaly is in excess of 20 cm/s to the northeast.

In comparison with the correlation field relating the parabathic velocity component with area, correlation field relating area with the diabathic velocity component (Figure 5a) shows a much more complicated structure and much smaller correlation magnitudes. It does exhibit a band of relatively high negative correlations extending east of the shelf edge and roughly confined within the 36.5-37 ^oN latitude band. These high correlations identify this band as a potentially important zone of slope/shelf water flow to the GS. Their sign implies that when the GS is close to the shelf edge the eastward flow of slope/shelf water in this band is enhanced, and when the GS is far from the shelf edge the eastward flow in this band is diminished. In Section 3.5, we further examine the character of the diabathic flow over the upper slope, as revealed by SLA gradients, in relation to GS position.

We should note here that our interpretation of the results focuses on flow over the slope and largely ignores flow over the shelf. This is principally due to the limitation of the tidal model applied to the T/P data at depths shallower than 200 m (as previously noted by Dong and Kelly, 2003).

To further investigate the modes of response to the GS fluctuations, we constructed contour plots of the correlation between the rates of change of both the diabathic and parabathic velocity components (i.e., the accelerations) and the rate of change of GS-to-shelf edge area (Figure 6). The correlation of parabathic component of acceleration with area change (Figure 6b) is not qualitatively different from the correlation of its undifferentiated counterpart with area. It exhibits a large negative correlation in most of the slope region, indicating that an increasing area correlates with an increasing negative parabathic acceleration over the slope (i.e., an increase in the southwestward component of flow over the slope). On the contrary, the diabathic correlation (Figure 6a) exhibits two lobes of opposite signs separated by a region of very low correlation. This indicates a spatially coherent response to the varying GS that was not observed in the diabathic acceleration time series. It is interesting that the region between the above mentioned lobes of the diabathic correlation field roughly coincides with the change in orientation of the axis of maximum GS front probability (in the proximity of 37° N).

In an attempt to confirm the spatial correlation distributions discussed above, the same type of correlation contour plots (as done for the data derived from the AVHRR and T/P satellite data) were produced with simulated currents and GS frontal positions produced by MICOM (Figures 7 and 8). These correlation plots confirm a very strong negative correlation between area and the parabathic component of geostrophic velocity anomaly over the slope. Also confirmed is the weaker negative correlation between parabathic velocity acceleration and the rate of area change (dA/dt). The correlations between diabathic component of velocity anomaly and area (area time derivative) are much smaller, and the spatial structure of these correlations is again rife with shorter scale features.
3.5. Sea level response along the 750-m isobath

Under the assumption that sea level anomaly along a mid-slope isobath would be representative of the slope response to GS displacements, we constructed sea level anomaly profiles along the 750-m isobath, a depth contour that lies in the vicinity of a T/P track (Figure 2).

Three sets of profiles were constructed on the basis of the area between the GS and the 200-m isobath: a) with area more than half a standard deviation above its mean (i.e., GS “far”); b) with area more than half a standard deviation below its mean (GS “close”); and c) the remaining profiles (GS “average”). Averaging along-isobath, the sea level for each of the sets yields the results that clearly indicate different behaviors for the “close” and “far” cases (Figure 9). When the GS is “close”, the averaged SLA shows a deep depression centered at ~37.5 °N. On the southern side of the depression is a steep southwestward rise in averaged SLA, roughly contained within the 36.5-37^o N latitude band. Noting that such a sea level rise would correspond to an off-slope geostrophic flow, this identifies the 36.5-37 ^oN latitude band as an area of enhanced off-slope flow when the GS is close to the shelf-edge. This is consistent with the conclusion reached above based on the spatial correlation between GS-to-shelf edge area and diabathic velocity anomaly (Section 3.4). Such a band of enhanced off-slope flow is not apparent in the average SLA profile computed for times when the GS is far from the shelf edge (the dot-dashed curve in Figure 9). In contrast with the SLA profile computed for the GS close to the shelf edge condition, this shows no areas of enhanced off-slope flow over the upper slope.

To further distinguish the response of upper slope currents to displacements of the GS close to and far from the shelf edge, we computed correlations relating the GS-to-shelf edge area with the parabathic and diabathic velocity anomalies along the 750-m isobath, as a function of latitude, for the GS close and far cases (Figure 10). The correlations of area with both velocity components show clear differences between the GS close and far cases. The correlations of parabathic velocity with area (Figure 10a) show negative valleys for both the GS close and far cases. However, the maximum negative correlation for the GS close case, is greater and is nearly a full degree further south at (37.5 ^oN) than the maximum negative correlation of the GS far case. The correlations of area with diabathic velocity (Figure 10b) further confirm the small influence of GS displacement on upper slope diabathic flow when the GS is far from the shelf edge. However, the correlations also indicate that GS displacements close to the shelf edge have an impact on diabathic flow over the upper slope. This impact is most evident in the 36.5-37.5^o N band over which the velocities are strongly anti-correlated with GS-to-shelf edge area.

The relationship between geostrophic velocity and GS position indicated by our analysis, in which parabathic velocity over the slope is stronger to the SW when the GS is further from the shelf edge, matches the finding of Dong and Kelly (2003) noted in the Introduction. Because their analysis was applied over a much broader scale, encompassing the entire MAB and extending onto the Scotian slope, it is uncertain as to whether our similar correlations between parabathic slope flow and GS position are the result of identical processes, or simply fortuitous – the result of different processes operating on differing scales. Further complicating the matter is the analysis of Bane et al. (1988) that indicates an increase in southwestward flow over the slope with decreasing GS-to-shelf edge distance, opposite of the relationship indicated by our analysis and that of Dong and Kelly (2003). It should be noted that the analysis of Bane et al. differed significantly from ours in that they compared the velocity of a single current meter moored over the upper slope with the distance between the current meter and the GS thermal front directly offshore. Furthermore, their velocity time series was from the extreme northeastern edge of our study region (36° 40΄N; 73°W). Nevertheless, their findings coupled with our results and those of Dong and Kelly (2003) raise the possibility that GS lateral movements may impact parabathic flow over the slope over a range of spatial and/or inter-annual scales not fully captured by our study or that of Dong and Kelly.

A significant finding related to the diabathic velocity component is the indication of a region of enhanced off-slope flow, in response to a decreasing GS-to-shelf edge area, near 37 ^o N, more than 150 km north of CH. This is revealed both by the SLA field over the 750-m isobath (Figure 9) and by the correlation of GS-to-shelf edge area with diabathc velocity anomaly (Figures 5 and 10). The latter indicates that the enhancement of diabathc flow in response to GS movement is most significant when the GS is relatively close to the shelf edge (Figure 10). This result suggests that the southern terminus of the slope water circulation proposed by Csanady and Hamilton (1988) may be situated in the vicinity of 37^o N.

Off-slope diversion of the southwestward flow over the slope well north of CH is consistent with the recent analysis of drifter data by Gawarkiewicz and Linder (this issue). They examined the tracks of an ensemble of drifters that were carried into the southern MAB shelf and slope region by the prevailing southwestward circulation. They found that a large proportion of the drifters were extracted from this southwestward flow within the area over which our analysis indicates a strong diabathic flow to deeper water. Quantitatively, Gawarkiewicz and Linder document 43 drifters entering the MAB shelf/slope region south of 38^o N with only 11 continuing southwestward past 36^o N. The drifters lost from the southwestward flow were entrained into the GS or into the northwestward current adjacent to the GS; and most of these were drawn from the slope, as opposed to the shelf, region.

It is curious that the enhancement of offshore diabathic flow with decreasing GS distance to the shelf edge is not reproduced in the MICOM results, suggesting that the mechanisms responsible for this flow enhancement are not captured by the MICOM simulation. One such mechanism may be the expulsion of GS water into the southern MAB. Churchill and Cornillon (1991a,b) found that this often produces a band of discharged GS water over the southern MAB slope, and that the geostrophic current at the northern margin of this band commonly draws shelf and slope water eastward to the GS. This flow has been most commonly observed in the 36-38^o N latitude band (Churchill and Cornillon, 1991a,b; Churchill et al., 1993; Churchill and Berger, 1998). The full dynamics and impact of this and possibly other mechanisms responsible for the diversion of southwestward MAB flow to the GS current system await further study.
ACKNOWLEDGEMENTS

TOPEX/POSEIDON data were produced by J. R. Hendricks and obtained from the Colorado Center for Astrodynamic Research at Boulder, Colorado. The MICOM simulation data were obtained from Dr. A. Sawdey at the University of Minnesota. GS front data were partially obtained from Dr. A. Mariano at the University of Miami. Support for this study was provided by the National Oceanic & Atmospheric Administration, National Ocean Service via the NOAA Coastal Services Center under grant _# NOAA Grant No. NA16RP2543. Support was also provided by the US Department of Energy Grant DOE-9310689, the Office of Naval Research under Grant N00014-98, and the National Science Foundation under Grants OCE-98-18804, OCE-03-27249, OCE-03-272349. Our thanks go to two anonymous reviewers who offered valuable suggestions for improving the manuscript. The authors especially acknowledge Dr. Gabe Csanady who we admire as a mentor and colleague. His pioneering work on coastal dynamics and shelf-slope interaction helped guide this study.

Auer, S. J. (1987). Five-year climatological survey of the Gulf Stream system and its associated rings. Journal of Geophysical Research, 92, 11,709-11,726.

Babin, S .M., Carton, J. A., Dickey, T. D. and Wiggert J .D (2004). Satellite evidence of hurricane-induced phytoplankton blooms in an oceanic desert, Journal of Geophysical Research, 109, C03043, doi:10.1029/2003JC001938..

Bane, J. M., Brown, O. B., Evans R. H., & Hamilton, P. (1988). Gulf Stream remote forcing of shelfbreak currents in the Mid-Atlantic Bight. Geophysical Research Letters, 15, 405-407.

Berger, T. J., Donato, T. F., Dragos, P., & Redford, D. (1996). Features of the surface circulation in the Slope Sea as derived from SST imagery and drifter trajectories. Journal of Marine Environmental Engineering, 2, 43-65.

Bleck, R., Dean, S., & Sawdey, A. (1995). A comparison of data-parallel and message-passing versions of the Miami Isopycnic Coordinate Ocean Model (MICOM). Parallel Computing, 21, 1695-1720.

Churchill, J. H., & Cornillon, P. C. (1991a). Water Discharged From the Gulf Stream North of Cape Hatteras. Journal of Geophysical Research, 96, 22,227-22,237.

Cressman, G. P. (1959). An operational objective analysis system. Monthly Weather Review, 87, 367-374.

Csanady, G. T., & Hamilton, P. (1988). Circulation of slope water. Continental Shelf Research, 8, 565-624.

Desai, S. D. & Whar, J. M. (1995). Empirical ocean tide models estimated from TOPEX/POSEIDON altimetry. Journal of Geophysical Research, 100(25), 205-228.

Dragos, P., Aikman III, F., & Redford, D. (1996). Lagrangian statistics and kinematics from drifter observations pertaining to dispersion of sludge from the 106-Mile Site. Journal of Marine Environmental Engineering, 2, 21-41.

Fu, L., Vazquez, J., & Parke, M. E. (1987). Seasonal variability of the Gulf Stream from satellite altimetry. Journal of Geophysical Research, 92, 749-754.

Gawarkiewicz, G., & Linder, C. A. Lagrangian flow pattern north of Cape Hatteras using near-surface drifters, Progress in Oceanography, this issue.

Halkin, D., & Rossby, T. (1985). The structure and transport of the Gulf Stream at 73° W. Journal of Physical Oceanography, 15, 1439-1452.

Halliwell, G. R., Jr., & Moers, C. N. K. (1983). Meanders of the Gulf Stream downstream of Cape Hatteras 1975-1978. Journal of Physical Oceanography, 13, 1275-1292.

Kelly K. A., & Gille, S. T. (1990). Gulf Stream surface transport ant statistics at 69°W from Geosat altimeter. Journal of Geophysical Research, 95, 3149-3161.

Kontoyiannis, H., & Watts, D. R. (1994). Observations on the variability of the Gulf Stream path between 74° and 70° W. Journal of Physical Oceanography, 24, 1999-2013.

Lee, T., & Cornillon, P. (1995). Temporal variation of meander intensity and domain-wide lateral oscillations of the Gulf Stream. Journal of Geophysical Research, 100, 13,603-13,613.

Lee, T., & Cornilliion, P. (1996a). Propagation and growth of Gulf Stream meanders between 75° and 45° W. Journal of Physical Oceanography, 26, 205-224.

Lee, T., & Cornillion, P. (1996b). Propagation of Gulf Stream meanders between 74° and 70° W. Journal of Physical Oceanography, 26, 225-241.

Parker, C. E. (1976). Some effects of lateral shifts of the Gulf Stream on the circulation northeast of Cape Hatteras. Deep-Sea Research, 23, 795-803.

Pietrafesa, L. J., Chao, S., & Janowitz, G. S. (1981). The variability of sea level in the Carolina Capes. Working Paper 81-11 UNC Sea Grant College Program, 39 pp.

Pickart, R. S., & Watts, D. R. (1993). Gulf Stream meanders over steep topography. Journal of Geophysical Research, 98, 6895-6905.

Tracey, K.L., & Watts, D. R. (1986). On Gulf Stream meander characteristics near Cape Hatteras. Journal of Geophysical Research, 91, 7587-7602.

Watts D. R., Tracey, K. L., Bane, J. M., & Shay, T. J. (1995). Gulf Stream path and thermocline structure near 74°W and 68°W. Journal of Geophysical Research, 100, 18,291-18,312.

Worthington, V. (1976). On the North Atlantic circulation. Oceanographic Studies, John Hopkins University, Baltimore, MD, 6, 1-110

FIGURE CAPTIONS

Figure 1. Chart of the Gulf Stream (GS) scaled probability distribution front based on the digitized frontal edges shown in Figure 2. The distribution was determined by counting the number of frontal points in each 0.25^o x 0.25^o square and dividing the resulting field by its maximum value. Note the downstream widening of the GS front distribution around its axis (thick solid line) of highest probability. The 200- and 750-m depth contours are drawn as solid and dashed lines respectively. The area discussed in the text is shaded.

Figure 2. Sea-level anomaly (cm) for (a) 1993 and (b) 1994. Each panel represents a Topex/Poseidon cycle and shows: sea-level anomaly in color (red high, blue is low), dotted lines along which Topex/Poseidon data are collected, digitized SST front (solid line). Each row of panels represents data for a season from winter (top) to fall (bottom). Both seasonal cycle (high in spring and summer and low in fall and winter) as well as 1993 to 1994 change (1994 has a more positive anomaly than 1993) are visible in this figure.

Figure 3. (a) Contours of RMS Sea Level Anomaly (cm) for the entire observation period. (b) Contours of eddy kinetic energy in (cm² s^-2) of the residual geostrophic flow derived from the SLA. Superimposed on each plot are the 200-m depth contour and the axis of maximum probability of the Gulf Stream frontal distribution (see Figure 1).

Figure 4. Time series of: (a) surface area between GS and the 200-m isobath; (b) SLA averaged within the surface area; (c) diabathic velocity and (d) parabathic velocity averaged within the surface area. A seasonal signal is evident in (b) due to the thermal expansion/contraction cycle in the mixed layer. To illustrate its correlation with the area time series, the averaged parabathic velocity is also shown as a dashed line in (a) with the scale on the right.

Figure 5. Contours of spatial correlation of (a) diabathic and (b) parabathic components of the residual velocity (from T/P) with the area between the GS thermal front and the 200-m isobath. Superimposed on each plot are the 200-m depth contour and the axis of maximum probability of the Gulf Stream frontal distribution (see Figure 1).

Figure 6. Contours of spatial correlation of (a) diabathic and (b) parabathic components of the acceleration (from T/P) with the area time derivative, dA/dt.

Figure 7. Contours of spatial correlation of (a) diabathic and (b) parabathic components of the residual velocity (from MICOM simulation) with the area (defined in the text).

Figure 8. Contours of spatial correlation of (a) diabathic and (b) parabathic components of the acceleration (from MICOM simulation) with the area time derivative, dA/dt.

Figure 9. Averaged SLA for large (GS far, dot-dashed), averaged (solid) and small (dashed) area (GS close) along the 750-m isobath. Far (close) is defined as area greater (smaller) than 1/2 standard deviation above (below) the average area value (2.76 10⁴ km²).