a. Ocean surface current radar

The OSCR radar system was deployed for the ONR Duck94 experiment along the outer banks of North Carolina during October 1994. The system consisted of two HF radar transmit/receive stations operating at 25.4 MHz that sensed the electromagnetic signals scattered from surface gravity waves with wavelengths of 5.9 m. The HF radar system mapped the coastal ocean currents over a 25 km × 44 km domain at 20-min intervals with a horizontal resolution of 1.2 km at 700 grid points (Fig. 1). The radars were located at the Army Corps of Engineers Field Research Facility (FRF) in Duck (master), North Carolina, and an oceanfront site in Corolla (slave), North Carolina, which equated to a baseline distance of 24 km. Each site consisted of a 4-element transmit and 16-element receiving array oriented at an angle of 120° (southeast-northwest at Duck) and 210° (northeast-southwest at Corolla) over a distance of 85 m (Haus et al. 1995). The manufacturer’s cited accuracy of the radial and the vector speeds are 2 and 4 cm s⁻¹, respectively, and the directional resolution is 5°. These error estimates are a function of the spectral resolution, angle of intersection between the master and slave radial beams (Lipa and Barrick 1983), and the positioning accuracy of the antenna or boresight. Other effects such as atmospheric noise and sea echo may also induce additional uncertainity in the velocity field or limit range performance (Barrick 1980).

Over the 29-day experiment, only 69 surface current maps of the possible 2073 samples were not acquired. The data return approached 95% averaged over the OSCR domain, as shown in Fig. 2. The highest return (approaching 100%) occurred in the inner to midshelf region, decreasing seaward to a minimum of about 50% in the far field at 44 km, representing the theoretical limit of OSCR’s range. The decrease in the far-field return was predominantly in response to the nor’easter that occurred from 14 to 16 October [yearday (YD) 287 to 289] where the near-surface winds were 14 m s⁻¹, as shown in Fig. 3a from the 20-m CoOP array. The 40-h low-pass filtered winds and pressures indicated the passage of synoptic atmospheric events occurring at 5–7-day intervals. An abrupt change in the wind direction of 180° occurred on 18 October (YD 292). Subsequently, near-surface winds rotated clockwise over the next few days with speeds of 4–5 m s⁻¹, which was important for the current response. The barometric pressure decrease was about 25 mb over a 2-day interval during the nor’easter, as shown in Fig. 3b. This strong forcing event also excited the surface wave field where significant wave heights exceeded 6 m at the inner-shelf mooring, causing OSCR dropouts in the far field, as noted by Haus et al. (1995).

b. Chronology of surface current events

The acquisition of surface current observations began 1 October 1994 and ended 30 October 1994. Surface flows responded to the passage of fronts and storms, coastally trapped buoyant jets, and tidal currents (Table 1). At the start of the experiment (event 1, not shown), winds of about 10 m s⁻¹ forced surface flows directed toward the north over the outer 25 km of the measurement domain, whereas closer to the coast, these flows were directed in the offshore direction. The winds subsequently reversed direction and the surface current pattern indicated a boundary between the inner- (15 km) and midshelf (25 km) response. This current pattern then moved farther offshore over a 10-h period as surface currents turned from northward to a southeast direction. From 1000 to 1300 UTC 2 October, strong surface flows exceeded 40 cm s⁻¹ between 20 and 30 km offshore (Fig. 4a). By the end of this transitional period, nearly uniform velocities were directed toward the south to southwest until 4 October when a wavelike structure in the along-shelf flow was detected within 10 km of the coastline (Fig. 4b). Offshore of this structure, nearly uniform flows in a southward direction were observed at speeds exceeding 50 cm s⁻¹. These surface currents decreased substantially to 15 cm s⁻¹ over the next day, whereas inner-shelf currents remained between 25 and 30 cm s⁻¹.

Except for the brief pressure change of 10 mb when winds exceeded 10 m s⁻¹ on 10 October, atmospheric conditions from 7 to 12 October (event 3) were relatively quiescent as onshore winds ranged from 4 to 6 m s⁻¹ (Fig. 3a); surface velocities were generally less energetic over the inner- to midshelf regime. However, a surface current convergence structure was observed inshore of 25 km late on 7 October (Figs. 4c,d). Along the northern periphery of the domain, surface flows were 40 cm s⁻¹ directed toward the southwest. This current structure turned southward within 10 km of the coast, narrowing to a width of 5 km with magnitudes of about 25 cm s⁻¹. In the central portion of the inner- to midshelf, this surface current patch widened downstream to an approximate width of 10 km where a convergence zone was located. This inner-shelf flow again narrowed to within a few kilometers of the coast along the southwestern boundary of the domain. Given this predominant inner-shelf flow toward the south and a trapping scale of 5–10 km from the coast, this part of the circulation may have been due to a buoyant jet emerging from the Chesapeake and Delaware Bays farther upstream (Boicourt 1981; Munchow and Garvine 1993). These buoyant flows tend to be trapped within 1–2 deformation radii of the coast, which scales from 3 to 5 km in this region. In the absence of strong atmospheric forcing (Fig. 3a) and weak tidal velocities, this flow of 20 cm s⁻¹ is a manifestation of the downstream spreading of this buoyant jet.

In response to the nor’easter that struck the Outer Banks on 14 October 1994 (event 5), observed surface flows increased and turned more southwestward (Fig. 4e) due to near-surface winds of 14 m s⁻¹ (Jensen et al. 1995). Sea surface conditions of enhanced white-capping and wave-breaking reduced the ability of the OSCR system to resolve the velocities at the extremities of the domain (Haus et al. 1995). Strong surface currents flowing southward continued to increase until reaching a maximum speed exceeding 75 cm s⁻¹ on 16 October. These surface flows decreased by 18 October when coastal intensification occurred within 7 km of the coast, engulfing the 20-m CoOP mooring to form a boundary between stronger inner-shelf currents and weaker offshore flows located between 20 and 25 km from the coast.

On 30 October, surface flows were 25–30 cm s⁻¹ directed toward the north in the center of the domain (Fig. 4f). These stronger flows were confined to an area of about 8 km × 8 km, and as time evolved this submesoscale patch of current variability propagated toward the north to northeast at 1–3 grid cells (1.2–3.6 km) over a 3-h period. This rate of propagation ranged from 10 to 32 cm s⁻¹, equating to a deformation radii of 3–5 km that were consistent with previous studies (Shay 1996).

c. CoOP measurements

CoOP is a National Science Foundation-sponsored interdisciplinary investigation of the biological and physical coupling responsible for the cross-shelf larval transport (Butman 1994). Alessi et al. (1996) described observations from three Woods Hole Oceanographic Institution current meter moorings equipped with VMCM and Seacat sensors sampling currents, temperatures, and salinities at 4-min intervals in 12.5-, 20-, and 25-m water depths (see Table 2 for instrument summary at the 20- and 25-m arrays). The distribution of the VMCM and the Seacat sensors in the water column provided measurements to examine the vertical structural changes of the currents, temperatures, and salinities that were effective in the Coastal Ocean Dynamics Experiment (Lentz 1994). Notice that the positioning of these arrays was not optimal with respect to the OSCR domain since the moorings were located close to the southern periphery where the angle of intersection between the radial vectors is outside the optimal range of 30° to 150° (i.e., 12.5-m mooring). In addition, the moorings were displaced by about 1 to 2 km southward during the nor’easter but were still within OSCR’s footprint.

To facilitate direct comparisons between the CoOP mooring data at 4-min intervals and OSCR data at 20-min intervals, the mooring data were windowed using a five-point hanning window (Otnes and Enochson 1978); subsampled at the 20-min intervals; and combined to form a profiler time series of velocities, temperatures, salinities, and densities at several vertical levels. The currents were rotated into a bottom topography reference frame (340°T) to form cross-shelf (u) and along-shelf (υ) components. Note that kinetic energy was conserved in this 20° coordinate rotation.

d. Observed velocity time series

The ocean current time series from the 25-m mooring indicated agreement in the current structure within the water column (Fig. 5). The semidiurnal (i.e., M₂) and diurnal (i.e., K₁) tidal amplitudes dominated the daily signals in both velocity components. Several surface current events can be traced to the current and density structure, including atmospheric frontal passage at the beginning of the experimental period on 2 October (YD 275) and during the nor’easter on 14 October (YD 287) shown in Fig. 3a. Strong cross-shelf currents were directed in the offshore direction ranging from 40 to 45 cm s⁻¹ at the surface and 4 m that decreased to 10–15 cm s⁻¹ at 6 m. The cross-shelf currents of about 25 cm s⁻¹ reversed direction at 12 m, to flow in the onshore direction, suggesting a two-layer regime due to upper-layer divergence and lower-layer convergence. However, during the storm period, the flow was accelerated toward the south to southwest direction from 14 to 16 October (YD 287 to 289), which caused a surface convergence region toward the coast. The along-shelf current was about 70–75 cm s⁻¹ compared to 60 cm s⁻¹ at 4 m, and 50 cm s⁻¹ at 6 and 12 m. During the relaxation of the response, currents exponentially decreased over the next few days until 18 October (YD 292) when the wind abruptly changed direction. At this time, large current oscillations of 20 cm s⁻¹ were excited with periods close to the local inertial period (≈20.3 h). Similar results were also observed at the 20-m site except during the nor’easter when the surface along-shelf component exceeded 90 cm s⁻¹ toward the south—nearly 40 cm s⁻¹ more than the 4-m current.

e. Observed density time series

Temperatures, salinities, and densities at 25 m indicated periods when the ocean was weakly and strongly stratified between the upper and lower layers (Fig. 6). The buoyancy frequency [N⁻² = −(g/ρ_o)∂ρ/∂z] was estimated by first averaging the density at 2.2 and 7.6 m (defined as the upper layer) and 13.2 and 23.6 m (defined as the lower layer), and calculating the vertical density gradient (∂ρ/∂z), assuming a two-layered ocean. This assumption is justified given the results from the density structure observations from the CoOP conductivity–temperature-depth profiles at the current meter moorings (Waldorf et al. 1996). Based upon these transect data, this fresher water had an original salinity of about 28 psu north of the OSCR domain. However, as this buoyant plume moved southward, the water mass mixed with the surrounding shelf water thereby increasing its salinity (30–31 psu) prior to arriving along the southern part of the OSCR domain.

The density was governed by freshwater intrusions when salinity decreased to between 30 and 31 psu (Figs. 6b, c). At the beginning of the record, a two-layer regime was evident in the observations between 7.6 and 13.2 m due to this intrusion. On 3 October (YD 276), higher-frequency oscillations were present in the density at the 13.2-m level due to internal waves; however, there was little evidence of these signals at the 7.6- and 23.6-m levels. Two more episodes of freshwater intrusions (S ≈ 30.5 psu) occurred on 7 October (YD 280) and 10 October (YD 283) and were confined to the upper few meters. This period coincided with low winds and tidal forcing (Table 1), and when a southward-flowing buoyant jet was observed within 10 km of the coast (Figs. 4c, d). A subsequent salinity decrease to 32 psu and density of 1.0225 g cm⁻³ in the water column occurred between 10 and 11 October (YD 283 and 284); and as the nor’easter passed over the area from 14 to 16 October (YD 287 to 289), the water column was weakly stratified and cooled by 1.5°C. During the relaxation stage, two additional freshwater intrusions occurred in the upper 7.6 m, suggesting the presence of a buoyant jet, which is usually confined to the upper 5 m as it emerges from Chesapeake Bay (Boicourt 1981). During these intrusions, the buoyancy frequency reached a maximum of 12–14 cph with more typical values of 8 cph between the two layers. The buoyancy frequency decreased to 1–2 cph during periods of wind-induced mixing, causing the column to become well mixed. The phase speeds for this two-layer regime generally ranged from 10 cm s⁻¹ to about 40 cm s⁻¹ over the time series, which agreed with those inferred from the surface current data (Fig. 4).

a. Tidal analysis

The semidiurnal (M₂, S₂) and diurnal (K₁, O₁) tidal constituents were isolated from the surface and subsurface currents time series at the 20- and 25-m moorings by a least squares fit to each of these tidal constituents. As shown in Table 3, the dominant tidal components were the semidiurnal tidal constituent (M₂) and a weaker diurnal tide constituents (K₁, O₁) that were influenced by bottom frictional effects. The along-shelf component differences between the surface and 4-m semidiurnal tidal currents were a maximum of 1.4 cm s⁻¹ at the 20-m mooring with more typical differences of approximately 1 cm s⁻¹. The variance explained by a combination of the semidiurnal and diurnal tidal components ranged from 2% to 10% at the 20-m mooring where the maximum was located at 8.7 m. By contrast, the explained variance for the cross-shelf current component reached a maximum of 22% in the upper 6 m at the 25-m mooring. Although the predicted variance for the along-shelf component was comparable (≈10–20 cm² s⁻²), this tidal component accounted for up to 9% of the observed variance. The amplitudes of semidiurnal components of 8 cm s⁻¹ at the surface decreased to 4 cm s⁻¹ at the bottom, and tended to be a factor of two to four times larger than the diurnal component at this array. Moreover, the semidiurnal tidal phases were nearly uniform in the vertical indicative of a predominantly depth-independent or barotropic component. Thus, in addition to containing more energetic barotropic and to a lesser extent baroclinic components, the differences between surface and 4-m tidal components were less than those previously found (Prandle 1987; Shay et al. 1995).

b. Rotary spectra

A Tukey window was applied to the surface and subsurface current data prior to removing the mean in calculating the horizontal kinetic energy spectra. These data were transformed and spectrally averaged to decrease leakage of kinetic energy to adjacent frequency bands (Otnes and Enochson 1978). The Fourier coefficients were then decomposed into clockwise and counterclockwise rotating components for the frequency-dependent motions (Gonella 1972; Mooers 1973). Spectral energy levels at the low frequencies (<10⁻² cph) were fairly consistent at each level (Fig. 7). The inertial and diurnal tidal band peaks in the surface current spectra indicated a preference of clockwise-rotating energies, which is consistent with near-inertial motions and the tidal analyses. Spectral energy levels decreased to a minimum of 10² (cm² s²)/cph at 12-m depth (Fig. 7d), resulting in a decade less spectral energy content than found in the surface current spectra (Fig. 7a). The dominant semidiurnal tidal frequency (M₂) peak was nearly equal in energy throughout the upper 12 m, as noted in Table 3. Thus, the semidiurnal tidal component contained a significant barotropic component, whereas the spectral energy peak at the near-inertial frequency suggested intermittent, baroclinic oscillations. At higher frequencies, there were flows with frequencies of 0.2–0.3 cph, as found in the HIRES-2 data where horizontal wavelengths of 4–6 km were found (Shay 1996). This band represents an intermittent 3–5-h oscillation defined by 12–15 data points for a 20-min sample interval, indicating physical processes resolved by both HF radar and VMCM current measurements.

c. Rotary shear spectra

To examine the current variability in the upper 6 m, vertical current shear spectra were estimated between the surface and the 4- and 6-m depths. Vertical shear spectral energy levels (s⁻² cph⁻¹) suggested that the diurnal and semidiurnal tidal peaks were not as significant as in the current spectra (Fig. 8). As shown in Table 3, the K₁ and M₂ tidal current speeds at the 25-m mooring were 2.6 and 7.9 cm s⁻¹ at the surface, 2.5 and 7.3 cm s⁻¹ at 4 m, and 1.9 and 6.5 cm s⁻¹ at 6 m, respectively. Since the phases were nearly equal at these three levels, tidal current shears were negligibly small for these components. Current differences in this band were about 1 cm s⁻¹ over 400 cm, equating to shear variances of approximately 2.5 × 10⁻³ s⁻², suggesting the predominance of depth-independent flows. By contrast, the near-inertial current component (≈f) was baroclinic, rotating clockwise over time and depth as energy propagated downward (Leaman 1976). At the higher end of the spectra, shears were also evident in the band centered at 0.2 cph where spectral peaks were found in the current spectra (Fig. 7). These spectral peaks were not as significant compared to the near-inertial peak found in the shear spectra. The high-frequency falloff in the spectral energies and shears was not as rapid as in the subsurface current measurements due to the spatial averaging over the grid cell by the HF radar. However, spectral peaks and high-frequency falloff for both the current and shear spectral densities approached unity in variance at 1 cph, suggesting reliable observations from both OSCR and VMCM current measurements.

a. Time series

Current measurements at both the surface and 4-m levels showed marked agreement at the 25-m mooring (Fig. 9), including semidiurnal tidal current amplitudes of 7–8 cm s⁻¹ that were in phase. Superposed on this tidal flow were buoyant-jet intrusions and the wind-induced response where the surface and subsurface along-shelf current components were −75 and −60 cm s⁻¹ between 14 and 16 October (YD 287 to 289). An important indicator in the comparison was the bulk current shears (Fig. 9c). This vector representation of bulk current shear was a maximum of 6 × 10⁻² s⁻¹ toward the south during the nor’easter event, followed by large amplitude, rotary shear flows. These bulk shears were consistent with those found in HIRES-2 where the current meters were spaced 9 and 13 m beneath the surface. Following Kundu (1976), daily averaged complex correlation coefficients (Fig. 9d) were above 0.8 over the series except during the period when the wind changed direction on 18 October (YD 292) and excited near-inertial oscillations with amplitudes of about 20 cm s⁻¹ (Fig. 3a). The complex phases, defined as the counterclockwise angle of the subsurface current vector relative to the surface current vector, ranged between −42° and 27°. The time-averaged correlation coefficient and phase were 0.93° and 7°, indicative of reasonable results.

Current measurements at the 20-m mooring exhibited similar behavior prior to the onset of the nor’easter event, followed by significant differences subsequent to storm passage (Fig. 10). Note that the time series started on 4 October due to the loss of current meter mooring data during the first few days of October, yielding 1868 data points (Alessi et al. 1996). In addition to the tidal forcing, the nor’easter accelerated the surface along-shelf component to a maximum of about 90 cm s⁻¹, whereas the subsurface current at 4 m was about 40 cm s⁻¹ less due to the superposition of the wind-forced flow and the buoyant jet. The subsequent current oscillations were also out of phase with decreased amplitudes at depth, suggestive of baroclinic motions similar to those detected at the deeper mooring from 18 to 24 October (YD 292 to 298). The variability in the corresponding bulk current shears (Fig. 10c) decreased the correlation coefficients between 0.4 and 0.6 where phases ranged between −120° and 134° (Fig. 10d). The time-averaged correlation coefficient and phase were 0.88° and 5°, which is consistent with those found at the 25-m mooring.

b. Regression analyses

Using the 2073 samples of the surface and subsurface velocities at the 25-m mooring, the cross-shelf and along-shelf current components were regressed based on least squares (Fig. 11). At both depths, mean biases and slopes between these components ranged from 1 to 4 cm s⁻¹ and 0.94 to 1.07, respectively. The slopes imply that the subsurface currents at 4 and 6 m were typically within 10% of the surface currents, depending upon the current component. In previous experiments, subsurface currents intermittently exceeded the surface currents due to sheared baroclinic regimes associated with the Gulf Stream and Florida Current (Shay et al. 1995; Shay et al. 1998, manuscript submitted to J. Geophys. Res.). This intermittency in the current maximum was also reflected in the histograms of the velocity differences for each component where the data follow a theoretical Gaussian distribution. Approximately 95% of the differences lie within the ±1.96 s, where s is the standard deviation. Over 90% of the current differences was within the envelope of ±12 cm s⁻¹ with the peak skewed between −4 to −8 cm s⁻¹ at the 4- and 6-m levels, respectively. Root-mean-square differences at 4 m were about 7 cm s⁻¹ for the cross- and along-shelf components (Table 4), increasing to about 9 cm s⁻¹ at 6-m depth. Thus, over the 2.3-m separation distance, rms differences increased by 1–2 cm s⁻¹ rms per meter of depth. Not only were these differences 30% to 50% less than those previously found from moorings (Shay et al. 1995; Shay et al. 1998, manuscript submitted to J. Geophys. Res.) and shipboard ADCP (Chapman et al. 1997) but they approached the measurement uncertainty in the instruments.

Regression analyses at the 20-m mooring (1868 points) revealed consistent results in the along-shelf component with those observed from the 25-m mooring (Fig. 12). The biases ranged from 1 to 2 cm s⁻¹ and slopes were 0.96 and 0.89 for the cross-shelf and along-shelf components, respectively. The increased differences in the cross-shelf component were due in part to the coastally trapped buoyant jet that engulfed the 12.5-and 20-m CoOP moorings. There is clustering of points in the along-shelf component where surface currents ranged from 75 to 90 cm s⁻¹ toward the south, and the corresponding 4-m currents were 40–75 cm s⁻¹. These maximum currents corresponded to a wind-driven acceleration of the surface flow and a lagged subsurface response at 4 m at the onset of the storm. At 6 m, the slope for the cross-shelf component indicated slightly larger discrepancies to the surface current observations compared to those at the deeper mooring (Figs. 12c,d). In both records, rms differences were more consistent with the HIRES-2 results of 10 cm s⁻¹ for the cross-shelf component, and about 13 cm s⁻¹ for the along-shelf component at 4 and 6 m. Regression slopes at 4 m were again O(1) for both components as in the HIRES-2 observations; however, the regression slopes at 6 m were markedly less. One explanation is that the buoyant jet was confined to depths less than 6 m as found by Boicourt (1983) and shown in Fig. 6 where fairly sizeable changes in the vertical structure beneath the surface plume structure exist. The histograms were also consistent with those found at the 25-m mooring except that the peak was skewed toward negative values of 8–12 cm s⁻¹. Approximately 95% of the current differences were within ±16 and 24 cm s⁻¹ at 4 and 6 m, respectively. This envelope of variability was larger than observed at the 25-m mooring because the angle of intersection of the master and slave radials at the 20-m array was not as optimal as at the 25-m array (Haus et al. 1995). That is, the along-shelf component was resolved, whereas the cross-shelf component was not well resolved along the 20-m isobath, particularly after the nor’easter event when the mooring was displaced 1 km southward. However, these rms differences at the 20-m mooring remained less than previously found (Matthews et al. 1988; Shay et al. 1995; Shay et al. 1998, manuscript submitted to J. Geophys. Res.).

a. Time series

A key issue emerging from previous studies is what do these surface current measurements represent in terms of barotropic processes induced by free-surface slopes and baroclinic currents forced by horizontal density gradients. Given the current meter distribution (Table 2), the depth-averaged current was estimated by vertically averaging the current meter data over 4.5-, 2.5-, 3-, 5-, 5-, and 5-m bin widths for the current meters at 4-, 6-, 8-, 12-, 17-, and 23-m levels, respectively. This time-dependent current was compared directly to the surface currents, as shown in Figs. 13a and 13b. Notice that a large fraction of the diurnal and semidiurnal tidal variability was embedded in the depth-averaged velocity. This time-dependent signal suggests that the tides such as the M₂ component were barotropic consistent with the results in Table 3. During the storm period from 14 to 16 October (YD 287 to 289), the wind-driven surface signal was predominately barotropic based upon differences of about 15 cm s⁻¹ between the surface and depth-averaged flows, which is consistent with theory. That is, if winds persist in nearly the same direction for more than a pendulum day (i.e., Fig. 3), a large fraction of the wind-driven energy excites mean and barotropic currents (Veronis 1956). As shown in Table 4, rms differences between the surface current and the depth-averaged flow were 10–12 cm s⁻¹ except for the along-shelf component at 20 m, which was about 15 cm s⁻¹. Since these values were larger than rms differences between the observed surface and subsurface signals, the surface currents also contained baroclinic contributions. Mean differences were 6 cm s⁻¹ for the cross-shelf component, and 7 cm s⁻¹ for the along-shelf component. The correlation coefficient decreased from 0.93 for the observed flows to 0.83 for the depth-averaged currents with a 5° increase in phase compared to those from the observed time series.

To isolate baroclinic processes, the depth-averaged current was removed from the surface and 4-m currents (Figs. 13c,d). This approach accentuated the coastally trapped buoyant jet confined to the upper layer and internal waves such as near-inertial motions during the abrupt wind change. On YD 275, 283, 293, and 297, the along-shelf current of 20–25 cm s⁻¹ was southward and correlated to the salinity intrusions shown in Fig. 6. Since the jet was confined to the upper few meters, this feature was resolved at the 25-m mooring. Between 18 and 24 October (YD 292 to 298), near-inertial motions had amplitudes of 20 cm s⁻¹ that were in phase, suggesting high correlations between these two levels compared to the correlations associated with the depth-averaged flow and the surface current. During this period, the baroclinic correlation coefficients were a factor of two to five times larger than the barotropic correlations. Averaged over the time series, the baroclinic and barotropic components contributed about 57% and 43% to the observed surface current variability, respectively. Thus, not only did OSCR sense both baroclinic and barotropic components in the surface flow but removing the depth-averaged or barotropic current effectively isolated the internal waves and baroclinic events embedded within the records.

b. Regression analyses

The barotropic and baroclinic flows were regressed to determine any systematic bias in these flows (Fig. 14). The barotropic comparison revealed a bias of 2–3 cm s⁻¹ and the slope of the regression line was 1.68 between the surface cross-shelf current and the depth-averaged flow, where the correlation coefficient was 0.59. The correlation for surface and depth-averaged along-shelf components was 0.81 where the slope and bias were 0.94 and 7.8 cm s⁻¹, respectively. The histograms were fairly tight with the maximum difference skewed between 6 and 10 cm s⁻¹ for the velocity components. The baroclinic surface and subsurface (4 m) flows indicated less scatter and narrower distributions in the differences. Regression slopes were 1.3 and 0.98 with similar biases of 2.5 and 0.6 cm s⁻¹ (Figs. 14c,d). Since the range of the baroclinic components was significantly reduced compared to the barotropic currents, the extrema in the surface along-shelf current were associated with barotropic processes and the passage of strong atmospheric forcing events. Correlations for the individual baroclinic components were 0.75 and 0.63, which were improved over the barotropic components. The maximum in the histogram for the cross-shelf component was ±4 cm s⁻¹, whereas for the along-shelf component there was no distinguishable peak in the distribution. Thus, the surface currents from the HF radar detected both barotropic and baroclinic flows forced by winds, tides, and buoyancy forcing events in the coastal ocean.