Mesoscale eddies in the South Atlantic Bight and the Gulf Stream Recirculation region: Vertical structure

Authors


Abstract

Sea level anomalies from altimeters are combined with decade-long potential temperature and salinity profiles from Argo floats to investigate the vertical structure of mesoscale eddies in the South Atlantic Bight (SAB) and the Gulf Stream Recirculation region. Eddy detection and eddy tracking algorithms are applied to the satellite observations, and hydrography profiles from floats that surfaced inside eddies are used to construct three-dimensional composites of cyclones and anticyclones. Eddies are characterized by large temperature and salinity anomalies at 500–1000 m depth and near the surface, and by small anomalies at 200–400 m below the surface at the depth of the North Atlantic Subtropical Mode Water. Anomalies associated with anticyclones are generally larger and found deeper in the water column compared to those due to the presence of cyclones. Geostrophic velocities around eddies generally exceed their translation speed in the top 1000 m of the water column. As such, these eddies can trap water in their interior as they propagate westward. Combining the volume of water inside eddies above their trapping depths with the number of eddies that propagate into the SAB each year, it is estimated that cyclones and anticyclones transport 3.5 ± 0.9 Sv and 4.1 ± 1.7 Sv onshore toward the Gulf Stream, respectively. The total volume transport of 7.6 ± 2.2 Sv represents an important fraction of previous estimates of the onshore transport in the Gulf Stream Recirculation gyre. Since eddies are characterized by large temperature and salinity anomalies, they also contribute significantly to the onshore transport of heat and salt.

1 Introduction

Mesoscale eddies are ubiquitous features in the global oceans, having long been recognized to influence the distribution of physical [e.g., Atkinson and Blanton, 1986; Lagerloef, 1992], chemical [e.g., Lee et al., 1991], and biogeochemical [e.g., Oschlies and Garçon, 1998] properties in the marine environment. They have also been shown to contribute to shelf-slope exchange [e.g., Washburn et al., 1993] and to strongly influence biological processes, either by inducing vertical fluxes of nutrients that can support new production [e.g., McGillicudy et al., 1998] or by passively advecting phytoplankton [e.g., Chelton et al., 2011a].

Much has been learned about eddy characteristics and propagation properties over the last few years. In particular, recent studies have shown that mesoscale variability in the ocean occurs primarily as nonlinear eddies, which in contrast to linear waves can transport momentum, heat, mass, and chemical constituents present in the seawater as they propagate [Chelton et al., 2007, 2011b]. Many studies have used satellite observations of sea level anomalies (SLA) or results from numerical models to describe the horizontal structure and kinematic properties of eddies in diverse oceanic settings, including at and around Eastern [e.g., Chaigneau et al., 2009; Kurczyn et al., 2012; Hormazabal et al., 2013] and Western [e.g., Qiu and Chen, 2005, 2010; Castelao and He, 2013; Kang and Curchitser, 2013; Yang et al., 2013] Boundary Currents and in marginal seas [e.g., Chen et al., 2011, 2012; Hu et al., 2012]. Satellite altimetry data have been further blended with profiles from Argo floats [Le Traon, 2013] to provide three-dimensional reconstructions of composite cyclonic and anticyclonic eddies [e.g., Chaigneau et al., 2011; Chen et al., 2012; Liu et al., 2012a; Yang et al., 2013], allowing for descriptions of their mean vertical structure. Those studies have shown that eddies can potentially transport large amounts of water, heat, and salt trapped in their interior as they propagate.

Off the U.S. Southeast coast at the South Atlantic Bight (SAB; we refer to SAB as the area between Cape Canaveral and Cape Hatteras to the west of 75.5°W; Figure 1), strong eddy activity is observed [Castelao and He, 2013]. The region farther offshore is generally characterized by westward flow as part of the southern portion of the Gulf Stream Recirculation region [Kwon and Riser, 2005]. In addition to locally generated eddies [Castelao and He, 2013], the SAB is influenced by eddies that propagate westward after they are generated either farther east [Chelton et al., 2011b] or farther north after the Gulf Stream separates from the coast at Cape Hatteras [Kang and Curchitser, 2013]. Observations that eddies are frequently encountered south of the Gulf Stream exist for over 70 years [Iselin, 1940; Iselin and Fuglister, 1948]. Some of these eddies can propagate southwestward as far south as 25°N [Parker, 1971]. They can also be strongly baroclinic, with at least order-of-magnitude vertical variations in their horizontal velocity [Warren, 1967]. Eddies in the region have been shown to be highly nonlinear [Castelao and He, 2013], with the potential to transport water trapped in their interior toward the SAB. However, much remains to be learned about their vertical structure. If eddies are characterized by large temperature and salinity anomalies in their cores, it is possible that they can play an important role setting the hydrography of the SAB offshore of the Gulf Stream. Since SAB water offshore of the Gulf Stream may ultimately be entrained into the Gulf Stream itself [e.g., Schmitz and McCartney, 1993], the water transported into the SAB by nonlinear eddies can be exported back into the North Atlantic when the Gulf Stream separates from the coast at Cape Hatteras [Kwon and Riser, 2005] or even be upwelled onto the SAB shelf when Gulf Stream waters move onshore as bottom intrusions [Lee et al., 1991]. Here we apply recently developed eddy detection and eddy tracking algorithms to high-resolution SLA fields in combination with decade-long profiles of potential temperature and salinity from Argo floats to investigate the vertical structure of mesoscale eddies in the South Atlantic Bight and in the southern portion of the Gulf Stream Recirculation region. We also quantify the potential for mesoscale eddies to transport water, heat and salt into the South Atlantic Bight as they propagate westward.

Figure 1.

Distribution of Argo floats that surfaced inside cyclonic (blue) and anticyclonic (red) eddies in the South Atlantic Bight and the southern portion of the Gulf Stream Recirculation region. Floats that surfaced outside eddies are shown in gray. Numbers of Argo floats in the study region are shown in the top left corner using the same color code. Thick black line is a contour of mean dynamic topography, which shows the long-term average position of the core of the Gulf Stream. The 60, 200, and 800 m isobaths are shown in light gray.

2 Methods

Gridded sea level anomalies (SLA) analyzed here were constructed by SSALTO/DUACS at 1 day interval using measurements from simultaneously operating altimeters. The observations are distributed by AVISO (Archiving, Validation, and Interpretation of Satellite Oceanographic data) as the “Reference Series,” which is built by combining data from a constant number (i.e., two) of altimeters over the entire time period. It is well known, however, that satellite surface elevation observations near the coast are less reliable because of intrinsic difficulties in the various corrections applied to altimeter data in those regions [e.g., Volkov et al., 2007; Saraceno et al., 2008], and that uncertainties in satellite-derived geostrophic velocities increase with proximity to shore [Liu et al., 2012b]. The width of the region near the coast where observations should be discarded was determined for the South Atlantic Bight by Castelao and He [2013] following Saraceno et al. [2008]. Castelao and He [2013] showed that, in the study region, along-track satellite altimeter observations are often missing inshore of 40 km from the coast. Offshore of 40 km from the coast, on the other hand, few observations (∼10%) are missing and altimeter data are highly correlated with tide gauge observations throughout the region. Therefore, all data from the AVISO SLA gridded product that fall within 40 km from the coast were eliminated before analyses.

We applied the Chelton et al. [2011b] eddy detection and eddy tracking algorithms to gridded SLA observations from September 2000 to January 2011, as in Castelao and He [2013]. Cyclonic and anticyclonic eddies are identified separately. Briefly, for each map of SLA, we look for regions (i.e., a set of connected pixels) that contain a local maximum (minimum) in SLA, and where all of the pixels in the region are above (below) a given SLA threshold for anticyclonic (cyclonic) eddies. In order to make the algorithm threshold-free and avoid having to choose an arbitrary value, the SLA field is partitioned using a range of thresholds from −100 cm to +100 cm, proceeding upward in increments of 1 cm for anticyclonic eddies and downward for cyclonic eddies until a closed contour of SLA that satisfies the criteria above is found. This also makes the algorithm capable of handling situations where an eddy is embedded in large-scale background sea surface height variations [Chelton et al., 2011b]. Once an eddy is detected, its amplitude is computed as the difference between the maximum (minimum) SLA within the anticyclonic (cyclonic) eddy and the average SLA around the outermost closed contour of SLA that defines the eddy perimeter. The eddy radius is defined as the radius of the circle that has the same area as the region within the eddy perimeter. After eddies were identified in each SLA map, the Chelton et al. [2011b] eddy tracking algorithm was used to determine the trajectory of each eddy. Specifically, for each eddy identified at time step k, the eddies identified at the next time step k + 1 are searched to find the closest eddy lying within a circle with radius of 21.5 km and with amplitude and area that fall between 0.89 and 1.21 times those of the reference eddy being considered. These are the same radius and thresholds used in Chelton et al. [2011b] and Castelao and He [2013], but modified to take into account the fact that a daily data set is used here (in contrast to the data sets used in Chelton et al. [2011b] and Castelao and He [2013], where observations were available every 7 days). If such an eddy is found, it is associated with the trajectory of the reference eddy at time k. Additional details of the eddy detection and eddy tracking algorithms can be found in Chelton et al. [2011b]. Eddies that could only be tracked for less than four consecutive weeks were discarded from subsequent analyses to reduce the risk of spurious eddies arising from noise in the SLA fields. The 4 week lifetime is commensurate with the 35 day e-folding time scale of the Gaussian covariance function in the objective analysis procedure used to construct the SLA fields of the AVISO Reference Series [Chelton et al., 2011b].

In order to investigate the vertical structure of mesoscale eddies in the region, we use autonomous CTD profiling floats from the Argo program from the same time period of the altimeter observations (September 2000 to January 2011) following Chaigneau et al. [2011] and Yang et al., [2013]. Argo data are automatically processed and quality controlled by the Argo data center [Wong et al., 2003; Böhme and Send, 2005; Owens and Wong, 2009], and only data flagged as good were retained in the analyses. Profiles for which the deepest data acquisition is shallower than 1000 m were eliminated. We further eliminated “suspicious” profiles using multiple criteria to flag hydrographic profiles as unreliable following Lentz [2003], Castelao [2011], Chaigneau et al. [2011], and Yang et al. [2013]. Last, every remaining profile was visually inspected, and those with suspicious temperature-salinity diagram were discarded. The remaining 3360 temperature and salinity profiles were then linearly interpolated into 201 regularly spaced vertical levels from the surface to 2000 m with an interval of 10 m. Observations in all profiles extended to at least 1000 m (since shallower profiles were eliminated), and 70% and 35% extended to at least 1500 m and 1800 m, respectively. At each vertical level, the dynamic height is calculated assuming a level of no motion at 1500 m. A deeper level of no motion was not used because of the limited vertical extend of Argo observations. Tests using a level of no motion of 2000 m produce qualitatively similar results. The dynamic height was not calculated for profiles that did not extend to at least 1500 m below the surface.

For each profile, anomalies of every property (potential temperature, salinity, potential density, and geopotential anomaly) were computed. For that, a climatological average of each variable was first built for each profile by averaging all available Argo profiles collected within a radius of 100 km and no more than 30 days apart of the day of the year in which the original profile was collected. For a profile collected on 15 October of a given year, for example, all profiles within a radius of 100 km obtained between 15 September and 14 November for all years would be used to compute the climatological averages. Once the climatology for each profile was build, anomalies were computed by removing the climatological averages from the respective quantity. Looking at anomalies is useful to identify how eddies can modify the background temperature and salinity fields.

Anomaly observations from Argo profiles were combined with altimetry data to reconstruct the vertical structure of eddies. The approach of combining satellite data and in situ observations from the Argo program has been widely used with great success over the last few years [e.g., Qiu and Chen, 2005, 2010; Chaigneau et al., 2011; Chen et al., 2011; Hu et al., 2012; Liu et al., 2012a; Yang et al., 2013]. All profiles in which a float surfaces inside cyclonic or anticyclonic eddies (as determined by the altimeter observations; Figure 1) were transformed into an eddy-centered coordinate system normalized by the eddy radius. Therefore, a profile surfacing at the eddy center will have coordinates inline image. For a profile surfacing at the boundary of an eddy, inline image. Anomalies for each depth were then objectively interpolated assuming an isotropic Gaussian covariance decorrelation scale of one eddy radius following Chaigneau et al. [2011] to obtain the three-dimensional structure of the composite eddy. Except when noted otherwise, only profiles located between 75°W–66°W and 27°N–34°N are used in the objective interpolation, since that region is characterized by a relatively higher concentration of floats surfacing within eddies. In order to determine the horizontal extent of the composite eddy as a function of depth, we average the geostrophic velocity along closed contours of geopotential anomaly embedding the eddy center. The closed geopotential anomaly contour associated with the strongest average geostrophic velocity was set as the edge of the composite eddy, since that is the region where fluid is more likely to be trapped in the eddy interior as it propagates (see below).

Nonlinear eddies can transport momentum, heat, mass, and the chemical constituents of seawater and therefore contribute to water mass distributions and ocean biology [Robinson, 1983]. An eddy is considered nonlinear here when the rotational speed around an eddy exceeds its propagation speed [Samelson and Wiggins, 2006; Chelton et al., 2011b], effectively trapping water inside its boundary. In order to determine the trapping depths [Chaigneau et al., 2011] of the composite eddies, we compare the rotational speed at each depth (i.e., the average geostrophic velocity along the edge of the composite eddy) to the mean translation speed of all long-lived eddies in the region based on the altimeter observations (3.5 cm s−1) [Castelao and He, 2013].

3 Results

The mean potential temperature-salinity (θ/S) diagram constructed using the Argo observations reveals the dominant water masses in the region (Figure 2). The diagram reveals a tendency for decreased near-surface salinities toward the west, consistent with observations reported by Schmitz et al. [1993]. θ/S diagrams for the different subregions are quite similar below 250–300 m below the surface. One of the dominant features in the region is the presence of the North Atlantic Subtropical Mode Water (the Eighteen Degree Water) [Worthington, 1959], an approximately isothermal and isohaline layer about 300 m thick with temperature of about 18°C and salinity of ∼36.5. The layer can be more easily seen at about 300 m below the surface as the region of widely spaced isotherms and isohalines in Figure 10. Subtropical mode water is found throughout the northwestern part of the subtropical gyre [Hanawa and Talley, 2001].

Figure 2.

Mean potential temperature versus salinity diagrams for four subregions extending from 27°N to 34°N. Longitudinal limits for each subregion are labeled. Dashed lines are potential density contours. Blue squares are plotted at 500 m intervals.

Potential temperature, salinity, and potential density anomalies in a cross section passing through the center of the composite eddies (at inline image) are shown in Figure 3. The average cyclonic eddy is characterized by large anomalies of up to −1.6°C at its center, peaking at 750 m below the surface. Anomalies colder than −1°C extend from 500 to 1000 m below the surface. A secondary core with anomaly of up to −1°C is found at 50–100 depth. A clear minimum in the magnitude of the temperature anomaly occurs between those two cores (∼200–400 m), roughly coinciding with the depth of the Eighteen Degree Water described by Worthington [1959]. The magnitudes of temperature anomalies associated with anticyclonic eddies are larger than those associated with cyclones. Peak anomaly is also located slightly deeper in the water column. Temperature anomalies in excess of 1°C can be found from 510 to 1100 m below the surface, peaking at 2.4°C at 810 m below the surface. A secondary peak at about 100 m below the surface is also found. Maximum anomalies of individual cyclonic and anticyclonic eddies can be larger than those observed in the composite eddies, but the magnitude of the vast majority of them is smaller than 4°C (Figure 4).

Figure 3.

Vertical sections of (left) potential temperature (θ; °C), (middle) salinity (S), and (right) potential density (σθ; kg m−3) anomalies across (top) composite cyclonic and (bottom) anticyclonic eddies (subscripts C and A, respectively) at Δy = 0. Eddy edges are denoted by black lines.

Figure 4.

Histograms of maximum potential (left) temperature and (right) salinity anomalies from all profiles from floats that surfaced inside cyclonic (blue) and anticyclonic (red) eddies.

The cyclonic and anticyclonic composite eddies are also characterized by large deep salinity anomalies (Figure 3). In both cases, anomalies are located at shallower depths compared to maximum temperature anomalies. Peak salinity anomalies are found at 650 m below the surface for cyclonic eddies and at 750 below the surface for anticyclonic eddies. As was the case for potential temperature, salinity anomalies are larger for anticyclonic eddies, by up to 60%. Anomalies are small at depths where the Eighteen Degree Water is found. In contrast to temperature anomaly profiles, no clear secondary intensification in the magnitude of the salinity anomaly is observed near the surface. Salinity anomalies at the top 100 m of the water column are actually positive for cyclonic eddies, although values are generally quite small. This occurs because of the salinity maximum observed at 90–130 m below the surface (Figure 2). Vertical displacements associated with cyclonic eddies pull the high salinity water upward, leading to a positive (but small) salinity anomaly near the surface (Figure 3). Maximum salinity anomalies at depth for individual eddies can reach up to ∼1 in a few instances (Figure 4).

The anomaly and climatological profiles can be used to obtain a rough estimate of the vertical displacement associated with eddies for each depth (Z) as in Richman et al. [1977]:

display math(1)

where inline image(z) is the average potential temperature or salinity anomaly at each depth from all floats surfacing inside cyclonic or anticyclonic eddies, and inline image is the climatological vertical potential temperature or salinity gradient based on profiles from floats surfacing outside eddies. We only estimate Z(z) for the top 1000 m of the water column, since inline image is quite small at depth. For both cyclones and anticyclones, estimates of the vertical displacement associated with eddies based on potential temperature or salinity profiles are similar to each other, suggesting that the estimate is robust (Figure 5). Estimates based on potential density yield quantitatively similar results (not shown). For cyclonic eddies, the displacement is of about 50–60 m between 400 and 1000 m below the surface, decreasing to about 25–30 m at 100 m depth. For anticyclones, displacements at 200 m depth are about −50 m, strengthening to approximately −100 m at 1000 m below the surface. It is important to emphasize that errors in the estimates at depth are likely higher, since vertical gradients (the denominator in equation (1)) are reduced there. Although the calculations above assuming that temperature and salinity anomalies within eddies are solely due to vertical displacements of the average fields represent a great simplification, they suggest that those vertical displacements are significant over a large fraction of the water column, including the region where potential temperature and salinity anomalies are small at about 300 m below the surface.

Figure 5.

Approximate vertical displacements (m) associated with the presence of cyclonic and anticyclonic eddies (subscripts C and A, respectively). Vertical displacements were estimated according to equation (1) using potential temperature (θC and θA) or salinity (SC and SA) profiles.

Since cyclonic eddies are associated with cold and fresh anomalies at depth, and anticyclones are characterized by warm and salty anomalies, their effects on the density field are partially cancelled out. Since the compensation is only partial, however, density anomalies are still relatively large (Figure 3). Potential density anomalies are positive for cyclonic eddies and negative for anticyclones, indicating that temperature plays a larger role controlling density variations. Peak density anomaly occurs deeper in the water column (i.e., at depths where salinity anomalies are relatively small and, therefore, have a small effect cancelling out the temperature-induced density anomalies), at 830 and 850 m below the surface for cyclonic and anticyclonic eddies, respectively. Interestingly, the largest potential density anomaly for cyclonic eddies occurs near the surface. Between 30 and 80 m below the surface, the potential density anomaly is twice as large as at depth, even though potential temperature anomalies are comparatively smaller at those shallow depths (see Figure 3, top left). This occurs because salinity anomalies at the surface are slightly positive, and therefore do not act as to cancel out the temperature-induced density anomaly. For anticyclonic eddies, potential density anomalies at the surface are roughly of the same magnitude as at depth. In both cases, a strong reduction in the magnitude of the anomalies is observed at the depth where the Eighteen Degree Water is found. Despite those differences, the vertical structure of cyclonic and anticyclonic eddies in the region are qualitatively similar. Most of the differences are restricted to the magnitude of the anomalies, or to relatively small vertical shifts in the depths of peak anomalies.

Lateral fields of potential temperature anomaly at the depths of the two peaks shown in Figure 3 (left) reveal the horizontal structure of the composite cyclonic and anticyclonic eddies (Figure 6; lateral fields of potential density anomalies at the same depths show qualitatively similar structures). At both depths, maximum perturbations are found approximately at the center of the composite anticyclonic eddy (center of eddies were determined based on satellite altimetry), suggesting little tilting of its vertical structure. The center of the composite cyclonic eddy at depth is slightly shifted toward the southeast. Geostrophic velocities at the level of maximum potential temperature anomaly relative to 1500 m are of the order of 15 cm/s, which is substantially higher than the average translation speed of eddies in the region as determined by altimetry observations (∼3.5 cm/s) [Castelao and He, 2013].

Figure 6.

Objective interpolation of potential temperature (θ; °C) anomalies and geostrophic velocities (V; m s−1) for (top) cyclonic and (bottom) anticyclonic eddies (subscripts C and A, respectively). Potential temperature anomalies are shown (left) at 80 m and 90 m below the surface and (middle) at 750 m and 810 m below the surface for cyclonic and anticyclonic eddies, respectively. Black contour shows maximum (minimum) potential temperature anomaly inside the anticyclone (cyclone) times one e-folding scale. Solid dots in left and middle plots represent anomalies estimated from Argo profiles in the eddy-centered referential. Geostrophic velocities around cyclonic and anticyclonic eddies at 750 and 810 m below the surface, respectively, are shown on the right. Colors are magnitude of velocity vectors.

Consistent with the vertical distribution of potential temperature and salinity anomalies, cross sections of meridional geostrophic velocity anomalies at inline image are qualitatively similar for the composite cyclonic and anticyclonic eddies (Figure 7). Velocities are stronger and reach deeper in the water column for the anticyclone, which is consistent with the larger magnitude of the potential density anomalies observed in that case at depth (Figure 3). In both cases, maximum surface velocities are about 35–40 cm/s. Vertical profiles of the maximum rotational speed along closed contours of geopotential anomaly (Figure 7, right) reveal a similar picture, with the swirl velocity of anticyclonic eddies slightly larger than that of cyclones. We note that surface rotational speeds of 35–40 cm/s are somewhat larger than rotational speeds estimated by satellite altimetry (average of ∼26 cm/s). This difference may be at least partially related to smoothing of the gridded SLA altimetry fields, since eddies with e-folding scales smaller than 0.6° are attenuated by the objective analyses procedure implemented by SSALTO/DUACS [Chelton et al., 2011b]. The advective nonlinearity parameter (Figure 7), defined as the ratio of the rotational speed to the average translation speed of eddies in the region, is used to determine the trapping depth for cyclones and anticyclones. The nonlinearity parameter is larger than 1 in the top 1050 m for cyclonic eddies, and in the top 1150 m for anticyclonic eddies. Eddies in the region are therefore highly nonlinear (consistent with Castelao and He [2013]), with a large potential for transporting material in their interior as they propagate predominantly toward the west.

Figure 7.

Vertical sections of the meridional geostrophic velocity (m s−1) relative to 1500 m of the (left) composite cyclonic and (middle) anticyclonic eddies at Δy = 0. Eddy edges are denoted by black lines. Maximum swirl velocities averaged over the composite cyclonic (blue) and anticyclonic (red) eddy edges as a function of depth are shown on the right. The advective nonlinearity parameter is computed as the ratio between the swirl velocity and the average translation of eddies in the region obtained by Castelao and He [2013]. The trapping depth for each composite eddy is the vertical extent of the fluid above the depth where the nonlinearity parameter is 1.

The analyses presented so far have relied in combining multiple CTD profiles inside eddies to obtain a composite picture of their three-dimensional structure. Comparison between the potential temperature or salinity anomalies and the magnitude of the sea level anomaly associated with an eddy for each individual profile reveals that those quantities are significantly correlated (at the 95% confidence level) for most of the water column (Figure 8). Correlations are largest between 400 and 1000 m below the surface, ranging from 0.45 to 0.65. Near the surface, where other processes (e.g., heating and cooling) are expected to influence the temperature and salinity distributions, correlations are reduced and not statistically significant. We note that correlations between sea level and the full temperature and salinity profiles (instead of anomalies) may be higher in the surface layer, as demonstrated for other regions [Liu and Weisberg, 2012]. With the exception of the correlation between sea level and temperature anomalies inside anticyclonic eddies, correlations are either strongly reduced or not statistically significant below about 1000 m. At depths where the magnitudes of the correlations are high and significant, a linear regression of the form inline image (where inline image is either temperature or salinity anomalies and inline image is the magnitude of the sea level anomaly) can be used to quantify the relation between sea level and temperature or salinity anomalies. Profiles of the slope of the regression between 400 and 1000 m below the surface, where correlations are strongest (see Figure 8), indicate that cyclonic and anticyclonic eddies with a sea level anomaly of 10 cm are associated with variations in potential temperature anomalies of about −1°C and over 2°C, respectively (Figure 9). Salinity anomalies are also about twice as large for an anticyclonic eddy of the same amplitude compared to a cyclone. The analysis also reveals that the largest variations in potential temperature or salinity anomaly associated with changes in eddy amplitude occur at considerably shallower levels for cyclonic eddies. The offset of the regression analysis suggests that cold and fresh anomalies as determined by the Argo observations would still be observed inside a cyclonic eddy of “zero amplitude” as determined by altimeters, while warm and salty anomalies would be observed inside an anticyclonic eddy of “zero amplitude.” This bias is consistent with the idea that at least some of the eddies identified with the satellite observations are attenuated during the processing by SSALTO/DUACS (see Appendix in Chelton et al. [2011b] for details) and, therefore, eddy amplitudes for the region possibly represent lower bound estimates.

Figure 8.

Vertical profiles of correlation coefficients between magnitude of (left) sea level anomalies and potential temperature or (right) salinity anomalies for cyclonic (blue) and anticyclonic (red) eddies. Solid lines denote coefficient is significant at the 95% confidence level.

Figure 9.

Vertical profiles of slope (solid) and offset (dashed) from regression analysis between magnitude of (left) sea level anomalies and potential temperature or (right) salinity anomalies for cyclonic (blue) and anticyclonic (red) eddies.

Figure 10.

Average amplitude of (top left) cyclonic and (top right) anticyclonic eddies as a function of longitude from altimeter observations. Vertical sections of (middle) mean potential temperature (θ; °C) and (bottom) salinity (S) anomalies for (left) cyclonic and (right) anticyclonic eddies (subscripts C and A, respectively) are shown in colors. Contours in middle and bottom plots show respective climatological values. Contour interval for potential temperature plots is 1°C, with the 10°C and 20°C isotherms in bold. Contour interval for salinity plots is 0.1, with the 35.5, 36, and 36.5 isohalines in bold.

Castelao and He [2013] analyses of 19 years of satellite altimetry observations revealed a great deal of spatial variability in eddy amplitude in the region (their study area was restricted to the west of 73°W). The significant correlations observed between eddy sea level and potential temperature or salinity anomalies primarily between 400 and 1000 m below the surface (Figure 8) suggest that it is possible that the temperature and salinity anomalies will also vary spatially. This can be better illustrated in longitude-depth plots of potential temperature or salinity anomalies inside eddies. In this case, anomalies from all individual profiles from floats that surfaced inside eddies were bin-averaged in 1° longitudinal bands, between 27°N and 34°N. Largest anomalies are observed between ∼73°W and 75°W–76°W, a region where the amplitude of the eddies is also enhanced (Figure 10), especially in the case of cyclones [Castelao and He, 2013]. Overlaying the anomalies with the background potential temperature and salinity fields (i.e., averages based on profiles located outside eddies) provides a clear picture of why the anomalies are enhanced near the surface and at depth, reaching a minimum at about 300 m (see also Figure 3). Peak anomalies are found at the midthermocline [McDonagh et al., 2010] and at the near-surface thermocline, and are small at the depth of the North Atlantic Subtropical Mode Water, where vertical temperature and salinity gradients are small.

Castelao and He [2013] showed that there is a strong tendency for eddies in the region to propagate westward, consistent with observations elsewhere [e.g., Lee and Mellor, 2003; Morrow et al., 2004; Chelton et al., 2007, 2011b]. Since the eddies are highly nonlinear (Figure 7), they can provide an effective mechanism to transport water, heat, and salt above their trapping depths [Flierl, 1981] into the South Atlantic Bight. To estimate the mean transport associated with the eddies, we divide the volume, heat, or salt anomaly of one eddy by 1 year, and multiply that by the number of eddies per year that propagate into the South Atlantic Bight [e.g., Gordon and Haxby, 1990; van Ballegooyen et al., 1994; Doglioli et al., 2007; Chaigneau et al., 2011]. Using the average radius for cyclonic and anticyclonic eddies estimated by satellite observations, its vertical variation (Figure 7, left and middle), and the trapping depth in each case (Figure 7, right), the volume of fluid trapped inside an eddy that can be transported as it propagates is 2.8 × 1013 m3 and 4.1 × 1013 m3 for a cyclone and an anticyclone, respectively. If only one cyclonic or anticyclonic eddy propagates into the SAB over the course of 1 year, that would amount to a transport of 0.9 Sv or 1.3 Sv, respectively. Using a longer time series of eddy propagation spanning 19 years [Castelao and He, 2013], an average of 3.9 long-lived (i.e., lasting ≥ 4 weeks) cyclonic eddies and 3.16 long-lived anticyclones propagate westward in each year across 75.5°W, the approximate offshore boundary of the SAB (of course, many more eddies are generated to the west of 75.5°W) [Castelao and He, 2013]. There is considerable variability in the number of eddies that propagate into the SAB in each year, however, ranging from 2 to 5 for cyclones and from 1 to 6 for anticyclones. Therefore, the total volume transport due to the westward propagation of nonlinear eddies is estimated to be 3.5 ± 0.9 Sv in the case of cyclones and 4.1 ± 1.7 Sv in the case of anticyclones (Table 1), for a total transport of 7.6 ± 2.2 Sv. If the transport is scaled by the number of eddies that propagate into the SAB in each individual year, total transport estimates range from 4.5 to 12 Sv (Figure 11). The total heat transport into the South Atlantic Bight due to nonlinear eddies, estimated as in Chaigneau et al. [2011] by integrating the available heat anomaly inside the composite eddy above the trapping depth and taking into account the average number of nonlinear long-lived eddies per year that propagate into the region, is −0.93 × 1013 W and 1.54 × 1013 W for cyclones and anticyclones, respectively (Table 1). The total salt transport associated with the westward propagation of nonlinear long-lived cyclonic and anticyclonic eddies is −2.1 × 105 kg/s and 4.6 × 105 kg/s, respectively (Table 1). Although the heat and salt transports associated with cyclonic and anticyclonic eddies are of opposite sign and therefore partially cancel out, the interannual variability in the number of cyclonic and anticyclonic eddies that propagate into the SAB can lead to relatively large net heat and salt transports (Figure 11).

Table 1. Transports Associated With Eddies That Propagate Into the South Atlantic Bight
 CyclonesAnticyclones
Volume transport (Sv)3.54.1
Heat transport (×1013 W)−0.931.54
Salt transport (×105 kg s−1)−2.14.6
Figure 11.

Time series of (top) volume, (middle) heat, and (bottom) salt transports by long-lived nonlinear mesoscale cyclonic (blue) and anticyclonic (red) eddies across 75.5°W, the approximate offshore boundary of the South Atlantic Bight. Total transports are shown in black.

4 Discussion and Conclusions

Sea level anomaly (SLA) altimeter observations were combined with decade-long in situ measurements from the Argo program to identify the vertical structure of mesoscale eddies at and eastward of the South Atlantic Bight (SAB) in the southern portion of the Gulf Stream Recirculation region (Figure 1). Eddies were detected and tracked in the altimeter data set using recently developed algorithms [Chelton et al., 2011b], and collocated CTD profiles from Argo floats were used to form a composite picture of the three-dimensional structure of cyclonic and anticyclonic eddies in the region. In particular, potential temperature and salinity profiles obtained by floats that surfaced inside eddies were compared to average profiles from floats that surfaced outside eddies to quantify temperature and salinity anomalies associated with the presence of cyclones and anticyclones. The analyses reveal that the background stratification plays an important role determining the influence of eddies generating temperature and salinity anomalies. Largest anomalies occur at depth, between 500 and 1000 m below the surface (Figure 3), where the background temperature and salinity vertical gradients are strong (Figure 10). A secondary peak in potential temperature anomaly occurs near the surface. Between these two peaks, any eddy-induced vertical displacements leads only to small temperature and salinity anomalies because of the presence of the North Atlantic Subtropical Mode Water (Figure 10), a relatively isothermal and isohaline layer centered at about 300 m below the surface [Worthington, 1959] characterized by weak background stratification and a local minimum in buoyancy frequency [Taft et al., 1986]. Observations from a meridional hydrographic section at 66°W from the World Ocean Circulation Experiment [Joyce et al., 2001] reveal that strong vertical gradients in nitrate and phosphate concentrations occur in the region at 600–800 m below the surface (see their Plate 1). This suggests that large anomalies in nutrient concentrations are likely to occur at those depths associated with the passage of cyclonic and anticyclonic eddies. Although weaker than at depth, vertical gradients in nutrient concentrations are also observed above 600 m, especially for nitrate and silica. Since significant vertical displacements are estimated to occur above 600 m below the surface associated with eddies of either polarity (Figure 5), this suggests that the effect of eddy-induced vertical nutrient fluxes can extend into shallower depths, and therefore be important for biological processes [e.g., McGillicudy and Robinson, 1997; McGillicuddy et al., 1998].

Using observations from an area to the north of the study region centered along the mean position of the Gulf Stream after its separation at Cape Hatteras (but still including the northern portion of the domain shown in Figure 1), Carnes et al. [1990] showed that the deeper structure of the Gulf Stream and eddies can be well reproduced with synthetic temperature profiles computed from altimeter-derived sea surface heights. Our results show a consistent picture, with temperature and salinity anomalies significantly correlated with the sea surface displacement due to the presence of eddies, especially between 400 and 1000 m below the surface (Figure 8). A linear regression analysis suggests that, for a given sea surface displacement associated with an eddy, a larger potential temperature or salinity anomaly is generated in response to the passage of an anticyclone compared to a cyclonic eddy (Figure 9). The analysis also suggests that anomalies due to anticyclonic eddies are generally generated deeper in the water column. Comparing the distribution of the amplitude of all eddies observed in the region from altimeter observations with the distribution of the sea level anomalies of the eddies sampled by Argo floats reveal that floats sampled a relatively large number of anticyclones of large amplitude, especially to the east of 74°W–75°W (Figure 12). In the case of cyclonic eddies, however, few Argo floats surfaced near the center of large eddies, especially in the region between 73°W and 68°W. It is possible, therefore, that the regression analysis (Figure 9) provides a better representation of anticyclonic eddies, and that the composite cyclonic eddy is biased toward the characteristics of low-amplitude eddies. Regardless of eddy polarity, the offset of the regression was found to be substantially different than zero at the depths where anomalies are intensified (Figure 9), which may be indicative of a possible bias. One possible scenario is that the analysis underestimates the amplitude of the smaller eddies in the region detected in the altimeter observations. In that case, future satellite missions such as the Surface Water and Ocean Topography (SWOT) mission [Fu et al., 2010], which will provide higher resolution sea level measurements, can help improve the understanding of the coupling between surface and subsurface variability within eddies.

Figure 12.

Amplitude of cyclonic and anticyclonic eddies as a function of longitude as determined by altimeter observations (gray circles). Sea level anomalies at locations of surfacing of Argo floats inside cyclonic and anticyclonic eddies are shown by blue and red circles, respectively.

Potential temperature and salinity anomalies associated with cyclonic and anticyclonic eddies in the region (Figure 3) seem to be much higher than in Eastern Boundary Currents [e.g., Chaigneau et al., 2011], by a factor of 2 or 3. They also reach deeper in the water column. However, they are comparable with anomalies observed in the northwestern subtropical Pacific Ocean[Yang et al., 2013]. As in the present study, Yang et al. [2013] observed that temperature anomalies associated with anticyclonic eddies are often larger than those associated with cyclones, especially in the western part of the basin. They attributed part of the difference to systematical bias in the climatology they used to estimate the anomalies (CSIRO Atlas of Regional Seas 2009—CARS), which under represents the presence of warm tropical water transported by the North Equatorial Current and the Kuroshio. The Argo-climatology systematical bias was estimated to be up to +0.2°C. This is substantially smaller than the difference in potential temperature anomalies between cyclones and anticyclones in the western subtropical North Pacific, however, suggesting that at least part of the difference may be real. Yang et al. [2013] also observed a double-core structure in the potential temperature anomalies associated with eddies, separated in the vertical by a region of low anomalies coinciding with the presence of North Pacific Subtropical Mode Water. In contrast to the southern portion of the Gulf Stream Recirculation region, near-surface anomalies in the northwestern subtropical Pacific Ocean are larger than anomalies at depth. The deep anomaly core is also located shallower in the water column, between 300 and 700 m below the surface.

The large temperature and salinity anomalies associated with cyclones and anticyclones in the Gulf Stream Recirculation region indicate that strong geostrophic velocities around the eddy (U) can be found quite deep in the water column (Figure 6). The average translation speed (c) of eddies in the region is relatively small. This leads to large U/c ratios, exceeding 1 in the top ∼1000 m of the water column (Figure 7). When the advective nonlinearity parameter [Chelton et al., 2011b] U/c is larger than 1, eddies can effectively trap water inside and transport water properties along their tracks [Samelson and Wiggins, 2006]. The trapping depth (i.e., the depth where U/c > 1) [Chaigneau et al., 2011] in the region is substantially larger than the trapping depth estimated for eddies in the northwestern subtropical Pacific Ocean by Yang et al. [2013]. The difference is partially because of the larger propagation speed of eddies in their study region, which is located at lower latitudes (18°N–26°N) and therefore characterized by higher phase speed for first-mode baroclinic Rossby waves [Chelton et al., 1998]. The swirl velocities around eddies are also slightly smaller, leading to trapping depths of 300–400 m. Off the coast of Chile, Chaigneau et al. [2011] also estimated trapping depths smaller than in the Gulf Stream Recirculation region. In their case, trapping depths between 20°S and 30°S were 280 m for cyclones and 570 m for anticyclones. Although the propagation speeds of eddies in that latitudinal band are comparable to those of eddies in the Gulf Stream Recirculation region, eddies off Chile are characterized by comparatively small temperature and salinity anomalies, leading to smaller rotational speeds. Compared to other regions, therefore, eddies in the southern portion of the Gulf Stream Recirculation region seem to be highly nonlinear, with a large potential for trapping fluid in their interior in a substantial fraction of the water column. This leads to a large transport of water by eddies into the South Atlantic Bight, despite the fact that only a relatively small number of long-lived eddies propagate from offshore into the SAB each year. The average volume transported by eddies into the SAB is estimated to be 7.6 ± 2.2 Sv (3.5 ± 0.9 Sv for cyclones and 4.1 ± 1.7 Sv for anticyclones), but total transport can be in excess of 10 Sv in years of high eddy activity (Figure 11).

The general circulation of the subtropical western North Atlantic was described by Kwon and Riser [2005] using hydrography and subsurface velocity measurements from 71 profiling floats during a 5 year period. The baroclinic transport in the Gulf Stream, assuming a level of no motion at 900 decibars, was found to be 40–50 Sv. The onshore volume transport in the Gulf Stream Recirculation across 75.5°W (taken here as the offshore boundary of the SAB) was 10–20 Sv. The Gulf Stream transport in the SAB is known to increase from about 30 Sv in the Florida Current through the Straits of Florida off Miami [Schmitz and Richardson, 1968; Richardson et al., 1969] to about 55–65 Sv near Cape Hatteras [Schmitz and McCartney, 1993], implying an onshore transport in the recirculation region of a few tens of Sverdrups (see Figure 8 in Schmitz and McCartney [1993] for a cartoon of the circulation). This suggests that the ∼7.6 Sv reported here as being transported on average by nonlinear eddies as they propagate westward into the SAB represent an important component of the total transport in the Gulf Stream Recirculation region. Heat and salt transport by nonlinear eddies are also quite large (Figure 11), potentially playing an important role setting the hydrographic conditions in the South Atlantic Bight offshore of the Gulf Stream. Since that water can be entrained in the Gulf Stream, heat, salt, and volume transport by nonlinear eddies may ultimately influence the temperature and salinity characteristics of the Gulf Stream itself. Onshore surface Ekman flow during southward winds [e.g., Atkinson et al., 1989] or upwelling associated with Gulf Stream meanders [e.g., Lee et al., 1991] can further transport the Gulf Stream water onshore onto the shelf [Castelao, 2011].

It is important to point out that a level of no motion of 1500 m was used here because of the limited vertical extension of the Argo observations. Using a deeper level of no motion would presumably lead to larger rotational speeds, and therefore to a deeper trapping depth in the eddies. We note, however, that the vertical shear of the rotational speeds around the composite cyclonic and anticyclonic eddies is substantially reduced below 1000 m (Figure 7), suggesting that using a level of no motion of 1500 m allows for capturing the portion of the eddy characterized by the strongest swirl velocities, and, therefore, with the strongest potential for transporting trapped fluid in its interior as it propagates.

It is also important to point out that the time series of volume, heat and salt transport shown in Figure 11 represent a simplification, since they are constructed using the volume and available heat and salt content anomalies from the composite cyclonic or anticyclonic eddies spread over a year and taking into account the number of eddies in each year that propagate into the SAB. In reality, the transports associated with each individual eddy will likely vary with its amplitude, since potential temperature and salinity anomalies vary to some degree as a function of the sea level displacement associated with the eddy (Figures 8 and 9). The average rotational speed around the eddy, and therefore its trapping depth, will likely also change as a function of eddy amplitude. Last, eddy amplitude will itself change as it propagates [e.g., Samelson et al., 2014], potentially leading to changes in temperature and salinity anomalies and in the trapping depth. These make computing time series of transport by nonlinear eddies very challenging with the sparse data set used here. Thus, the time series shown in Figure 11 represent broad measures of the transports demonstrating the potential for large interannual variability, rather than precise quantitative estimates of the transports in each given year.

The analysis presented here suggests that potential temperature and salinity anomalies inside eddies to the west of 75.5°W in the SAB are reduced compared to offshore (Figure 10). Vertical profiles of anomalies inside composite cyclonic and anticyclonic eddies generated using only observations to the west of 75.5°W indeed show reduced anomalies, especially at depth (Figure 13, thick lines). To the extent that temperature and salinity anomalies are related to sea level anomalies (Figure 8), it is possible that this can be at least partially explained by the decrease in the amplitude of eddies to the west of that longitude (top plot in Figure 10, see also Castelao and He [2013]). It is also possible, however, that this is mostly a result of a sampling bias. First, very few floats surfaced inside eddies to the west of 75.5°W (Figure 1), resulting in a small number of profiles available for analysis. Additionally, there is a clear bias toward floats surfacing either inside small-amplitude eddies or near the boundary of large-amplitude eddies (Figure 12). While to the west of 75.5°W 80% of all cyclonic eddies identified in the altimeter observations have amplitude larger than 10 cm, only three Argo floats have surfaced inside cyclones with sea surface displacement larger than 10 cm. The bias seems to be even larger for anticyclones. In that case, the maximum sea surface displacement at the location of the surfacing of Argo floats was 5 cm. For comparison, 60% of all anticyclones identified in the altimeter data set to the west of 75.5°W have amplitude larger than 10 cm. Therefore, it is possible that the small anomalies observed in the SAB (Figures 10 and 13, thick lines) are mostly a result of the low number of Argo floats to the west of 75.5°W. As a result, large eddies, which are expected to produce the largest temperature and salinity anomalies, were not adequately sampled. Indeed, potential temperature and salinity anomalies are much larger if they are computed using only floats surfacing inside eddies to the east of 75.5°W (where the number of floats is larger, see Figure 1) but that later on propagate westward across that longitude line into the SAB (Figure 13, thin lines). As more altimeter and Argo float observations accumulate with time, the likelihood of better sampling the large eddies found in the SAB will increase, allowing for more accurate representation of the temperature and salinity anomalies in the region.

Figure 13.

Profiles of (left) potential temperature and (right) salinity anomalies averaged inside composite cyclonic (blue) and anticyclonic (red) eddies. Thick lines are for composite eddies constructed using only profiles from floats surfacing inside eddies located to the west of 75.5°W. Thin lines are for composite eddies constructed using only profiles from floats surfacing inside eddies located to the east of 75.5°W and that later on propagate westward across that longitude.

It would be interesting to repeat the analyses presented here to the region between the Gulf Stream after separation at Cape Hatteras and the Middle Atlantic Bight (MAB). Several studies have shown that warm core rings can be formed at the Gulf Stream and subsequently drift westward toward the MAB [e.g., Joyce et al., 1992; Gawarkiewicz et al., 2001; Lee and Brink, 2010, among many others]. The region is characterized by strong eddy activity [Kang and Curchitser, 2013] and high mean eddy amplitude (see Figure 10 in Chelton et al. [2011b]). If eddy amplitude is correlated with potential temperature and salinity anomalies at depth, one would expect large anomalies and strong rotational speeds in eddies found between the Gulf Stream and the MAB. Additionally, propagation speeds for those eddies are likely to be small at those relatively high latitudes, since they are characterized by low phase speed for first-mode baroclinic Rossby waves [Chelton et al., 1998]. It is likely, therefore, that eddies will be characterized by large advective nonlinearity parameters and large trapping depths, implying a great potential for transporting mass, heat and salt as they propagate toward the MAB.

In summary, observations from multiple altimeters and Argo floats were combined to investigate the three-dimensional structure of mesoscale eddies in the South Atlantic Bight and the southern portion of the Gulf Stream Recirculation region. Cyclonic and anticyclonic eddies are characterized by large potential temperature and salinity anomalies between 500 and 1000 m below the surface. Anomalies are also intensified near the surface, above the depth of the North Atlantic Subtropical Mode Water. Eddies are highly nonlinear, transporting on average ∼7.6 Sv of water trapped in their interior into the South Atlantic Bight as they propagate westward. They are also characterized by relatively large heat and salt transports. As such, nonlinear mesoscale eddies are potentially an important component of the onshore transport in the Gulf Stream Recirculation gyre.

Acknowledgments

Support by NASA Ocean Surface Topography Mission through grant NNX13AD80G is gratefully acknowledged. We thank CLS/AVISO for distributing the SSALTO/DUACS altimeter observations, and the Argo program for collecting and making the float data freely available online. We also thank Cori Pegliasco and Alexis Chaigneau at the Laboratoire d'Etudes en Géophysique et Océanographie Spatiales (LEGOS), France for suggesting the use of the Argo float data to compute climatological temperature and salinity profiles. All observations used here are available at http://www.aviso.altimetry.fr/en/ and http://www.argo.ucsd.edu/.