Corresponding author: R. Castro, Facultad de Ciencias Marinas, Universidad Autónoma de Baja California, Carretera Tijuana-Ensenada km 107, Ensenada, B. C. 22860, Mexico. (firstname.lastname@example.org)
 The hydrodynamic characteristics of mesoscale eddies of the southern Gulf of California are described from observations made in August 2004. Vertical profiles to ~1000 m were made with a conductivity-temperature-depth (CTD) probe with dissolved oxygen and fluorescence sensors, and a lowered acoustic Doppler profiler. The four sampled eddies were aligned along the Gulf axis and had alternating sense of rotation. The Okubo-Weiss parameter was used to calculate the eddy core radii, and it also revealed that the current field of almost all the sampled area was eddy dominated. The mean radii of the eddy cores varied between 32 and 36 km. Maximum surface velocities were 0.4–0.5 m/s. The anticyclonic eddies were 500–700 m deep, while the cyclones were 450–500 m deep. The depression of the isolines associated with the anticyclonic eddies was ~100 m at depths between 200 and 400 m; however, the thermocline/pycnocline, which spanned from ~20 to 75 m, was domed by ~10 m. The cyclonic eddies lifted the isolines by ~70 m at depths above 500 m, but the isolines were flat above ~60 m. This suggests that different dynamics operate in the layer above the strong and shallow pycnocline. In particular, the wind stress curl can affect circulation and pycnocline topography and, therefore, the patterns that appear in chlorophyll satellite images. In our case, the eddies were detectable in those images because of the chlorophyll-enhanced streamers along their edges and, for anticyclones, probably because the chlorophyll maximum found at 40–50 m depth was domed following the pycnocline.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The surface dynamics and thermodynamics of the Gulf of California (GC) (Figure 1) have been extensively studied in the seasonal time scale [Lavín and Marinone, 2003, and references therein]. While the seasonal circulation is important and well understood, satellite images and numerical models of the Gulf suggest that mesoscale eddies with diameters of ~70–120 km (GC width 120–160 km) are common in the southern Gulf; yet, they have not been directly studied in detail.
 The oldest direct observations of a mesoscale eddy in the Southern Gulf of California (SGC) (Figure 1) were made in August 1978 [Emilsson and Alatorre, 1997], using surface drifters, a lowered Aanderaa current meter, and a few conductivity-temperature-depth (CTD) casts; a cyclonic eddy of ~50 km radius and ~70 m depth was found over Farallón basin (see basin names in Figure 1). The only later hydrographic evidence of eddies in the SGC (by Fernández-Barajas et al.  from February 1992 and by Figueroa et al.  from March 1984) suggest a series of geostrophic eddies, 400–500 m deep, with alternating sense of rotation along the SGC; however, their stations were very wide apart or were from single along-gulf lines of stations, so that details were missing or no information was provided on the lateral structure.
 Satellite infrared and color images have shown that mesoscale structures are common features of the surface layer of the SGC [Badán-Dangon et al., 1985; Pegau et al., 2002; Navarro-Olache et al., 2004; Lavín and Marinone, 2003]. Using sea-viewing wide field-of-view sensor, ocean color images from early September 1999, Pegau et al.  suggested that the surface circulation during late summer in the SGC was dominated by eddies, with alternating sense of rotation, with radii of ~35–50 km. The presence of an eddy pair extending from Cabo Lobos (Figure 1) in three consecutive years suggested to these authors that the eddies were topographically locked. However, this study was based solely on satellite-derived chlorophyll pigment with no quantitative values, and there is no confirmation that the features they observed were actual deep eddies.
 The generation of the SGC eddies was investigated numerically by Zamudio et al.  with a 3-D 1/12° nested interannual-forced regional Gulf of California hybrid coordinate ocean model, with local and/or remote forcing, that was nested in basin-scale and global models. They found that remote forcing is essential for the generation of eddies in the SGC. The simulations show that the interaction of the poleward Mexican Coastal Current with specific topographic irregularities (capes at Topolobampo and Cabo Lobos, and the shelf ridge off the San Lorenzo River) could generate eddies by inducing baroclinic instabilities and that this mechanism is strengthened by the arrival of coastal-trapped waves of equatorial origin. They also concluded that local wind forcing was not essential for the generation of these eddies. In this model, the lifetime of the eddies is 2–3 months (August–October).
 The available evidence suggests that mesoscale eddies are a very important (if not the most important) component of the GC circulation during summer; yet, their physical characteristics have not been described in detail. It is likely that eddies have a role in the transport/retention of chemical properties and of planktonic organisms and in the redistribution and mixing of properties at different levels of the water column. Therefore, it is important to provide a quantitative description of relevant characteristics of the eddies (such as radii, depths, swirl velocities, Rossby numbers, vorticity, dynamic height anomalies, etc.) that can be used by numerical modelers and other observationalists. Tidal currents are weak (<0.05 m/s) in the SGC [Marinone and Lavín, 2005] and are not considered here.
 The objective of this study is to describe the hydrographic and kinematic characteristics of the mesoscale eddies of the Southern Gulf of California, based on highly detailed direct observations [CTD and lowered acoustic Doppler current profiler (LADCP)] made in August 2004. After describing data and methods in section 2, results (section 3) are presented of (a) the horizontal structure of eddy currents and hydrography, (b) the vertical structure of a representative anticyclone and cyclone, and (c) eddies and chlorophyll images. The discussion (section 4) deals with eddy dimensions and position, with the potential role of the wind stress curl on near-surface dynamics and chlorophyll images, with hydrographic structure, and with comparison of geostrophic velocity versus LADCP observations.
2 Data and Methods
 Hydrographic observations were made during cruise NAME-2, carried out from 7 to 20 August 2004, on board the R/V Francisco de Ulloa, covering a large part of the SGC with a total of 196 oceanographic stations in 11 across-gulf lines (Figure 1). The spacing between stations was ~10 km in the across-gulf transects, and the along-gulf distance between lines was between 30 and 40 km. This sampling scheme was a compromise between fine across-gulf sampling and ample along-gulf coverage; the sections crossing near eddy centers provided good individual characterizations, while the other lines provided between-eddies information that helped complete the description of the eddy field. The width of the Gulf decreases from 180 km at lines A and B to 130 km at line F (Figure 1).
 Temperature and conductivity profiles down to 1000 m (or near the bottom if shallower) were measured with a factory-calibrated CTD (SeaBird SBE-911 plus), with primary and secondary sensors and a sampling rate of 24 Hz. Dissolved oxygen (DO, mL/L) and fluorescence sensors (SBE43 and SeaPoint, respectively) were also included. The fluorescence sensor had a range of 0–150 mg/m3 of chlorophyll a (abbreviated chl-a; mg/m3) and a 1× gain. The data were processed and averaged to 1 db as documented by Castro et al. . Salinity was calculated with the Practical Salinity Scale (1978). Potential temperature (θ, °C) and potential density anomaly (γθ, kg/m3) were calculated according to UNESCO . Geostrophic velocity (VG, m/s) was computed using the deepest common depth as reference level, which was ~1000 m for most pairs of stations. Prior to geostrophic velocity calculation, the temperature and salinity cross sections were objectively mapped in order to remove internal waves and other small-scale variability. A standard objective mapping interpolation was used, with a classical Gaussian correlation function with relative errors of 0.1, with a 70 km horizontal length scale and a 30 m vertical scale. The chosen horizontal scale is about twice the baroclinic radius of deformation, which for our region is ~40 km [Chelton et al., 1998] (http://www-po.coas.oregonstate.edu/research/po/research/rossby_radius/index.html), thus ensuring that the smoothed sections resolve geostrophic flow. In addition, following Chereskin and Trunnell , we computed the covariance of temperature and density data from the cruise for the 0–10 and 250–300 m layers; a guess correlation of the covariance using a Gaussian function resulted in length scales of ~70 km. Objective mapping was also applied to the horizontal distributions, with scales 0.3° of latitude by 0.3° of longitude.
 Velocity profiles were measured with a broadband Teledyne-RDI 300 kHz lowered acoustic Doppler current profiler (LADCP) attached to the CTD protection frame. The absolute velocity profiles were obtained with the methods described by Visbeck . The sampling bins were 8 m thick. Objective analysis was also applied to the LADCP velocity components with horizontal scale of 70 km and 50 m in the vertical. The covariance at 8–48 and 248–304 layers was also computed for velocity components; the correlation length scale resulted ~30–40 km of the order of the Rossby radius of deformation. We also used 70 km to reduce ageostrophic noise in the LADCP observations [Chereskin and Trunnell, 1996].
 The horizontal distribution of LADCP velocities was used to calculate the vertical component of relative vorticity
and the Okubo-Weiss parameter W [Okubo, 1970; Weiss, 1991], which identifies regions where ζ dominates over strain
where Sn and Ss are the normal and shear components of the strain
 In areas where vorticity dominates the current field, W < 0, and in those where strain dominates, W > 0. When the flow field is partitioned with W, it is observed [Isern-Fontanet et al., 2004; Henson and Thomas, 2008] that eddies have a central area (the eddy core) where W is strongly negative, surrounded by an area (the eddy ring or circulation cell) where W > 0. The eddy core edge can be chosen as the connected curve where W = 0 or as the narrow closed strip area where −0.2σw ≤ W ≤ 0.2σw, where σw is the standard deviation of W in the area under observation. The outer eddy boundary is sometimes chosen as the line where ζ = 0 [e.g., Kurian et al., 2011]. Beyond the eddy boundary, the background flow is usually characterized by small positive and negative values of W (|W/σw| ≤ 0.2).
 The Rossby number in polar coordinates is Ro = |Centrifugal force/Coriolis force| = |V/fr|, where V is the tangential velocity, f is the Coriolis parameter, and r is radial distance from the eddy center [e.g., Takahashi et al., 2007]. Assuming that the eddy core rotates as a solid body, the swirl velocity and the angular frequency are related by V = ωr, and ω can be obtained from linear regression between V versus r for data points inside the eddy core. Then Ro = |ω/f|.
 Wind data with a horizontal resolution of 0.25° × 0.25° and sampled every 6 h were obtained from the product called “Cross-Calibrated Multiplatform (CCMP) Ocean Surface Wind Velocity Product for Meteorological and Oceanographic Applications” [Atlas et al., 2011;ftp://podaac-ftp.jpl.nasa.gov/OceanWinds/ccmp/L3.0]. Wind stress and wind stress curl were calculated from these data. Three-day mean satellite images (4 × 4 km) of surface chlorophyll concentration from the MODIS satellites were obtained for 7, 9, 11, and 17 August 2004 (http://oceancolor.gsfc.nasa.gov/cgi/level3.pl). For those dates, the MODIS K490 images were used to estimate the depth of the first optical depth, as the inverse of the diffuse attenuation coefficient at 490 nm wavelength.
 Surface currents were measured with two PacificGyre (http://www.pacificgyre.com) SVP ARGOS drifters equipped with holey socks centered at 15 m. The drifters were deployed from the R/V Francisco de Ulloa during a cruise (NAME-1) made from 5 to 18 June 2004.The data were cleaned and interpolated at 6 h intervals as described by Hansen and Poulain . The first drifter (No. 50021) sampled in the open Gulf from 19 July to 31 August 2004. The other drifter (No. 50012) sampled from 7 June to 19 August 2004.
3.1 Eddy Currents: Horizontal Structure
 The horizontal distribution of drifter and LADCP surface current vectors (8–48 m vertically averaged LADCP data) (Figure 2a) shows four eddies distributed along the southern Gulf, with alternating sense of rotation, namely (from the mouth to the head of the gulf): anticyclonic eddy A1, cyclonic eddy C1, anticyclonic eddy A2, and cyclonic eddy C2. Mean velocity magnitudes around the eddies (Figure 2b) were ~0.3 m/s, with maxima ~0.4–0.5 m/s in certain spots, especially where eddies interacted. The eddies were so close to each other that their respective current fields cannot be told apart by looking at Figure 2a.
 The surface distribution of W/σw (Figure 2c) calculated from the LADCP data shows that the main eddies were slightly elongated in the approximate direction of the gulf axis (NW-SE), except for A1, which was elongated in the N-S direction. The major and minor semiaxes for the eddy cores (where W = 0) are as follows: A1, 32 and 39 km; C1, 27 and 40 km; and A2, 27 and 38 km. There appeared another weak eddy southeast of A1, not evident in Figure 2a, but its W/σw and speed were lower than for the other eddies. Outside the eddy cores, W/σw was positive with maximum W/σw = 1 (where, in this case, σw = 1.17 × 10−10 s−2). This value of W/σw is slightly smaller than in typical areas surrounding eddy cores (1.5–2) [Isern-Fontanet et al., 2004]. Away from the eddies, |W| falls below 0.2σw in very few places (especially in the SE), which means that there was very little background area or that almost the entire sampled area was affected by the eddies. The surface relative vorticity field (Figure 2d), computed from the LADCP (8–48 m layer velocity field), defined very clearly eddies A1, C1, and A2 (not C2, because it was sampled by a single line of stations). The boundaries between eddies cannot be established with the velocity field nor with the distribution of W, but we can define them as the contour ζ = 0; the area between the contours W = 0 and ζ = 0 (Figure 2c) would be the “eddy ring” or the “circulation cell.” The core radii were not much smaller than those of the eddy rings, probably because the eddies were so close to each other.
 Eddy A1 (Figure 2) covered part of Pescadero basin. Its mean core radius was 36 km, and its highest speeds (~0.5 m/s) were at its eastern margin. Off the peninsula, speeds were between 0.2 and 0.3 m/s. The red drifter (50021) showed a slow (~0.13 m/s average) northward displacement across the position of A1, from 7 to 25 June 2004, which suggests that A1 was probably formed in July. In the surface layer (Figure 2d), A1 had the maximum negative relative vorticity (−2 × 10−5 s−1), while the mean value inside the W = 0 contour was −1.38 × 10−5 s−1.
 Cyclonic eddy C1 (Figures 2a and 2b) was located northeast of Bahía de La Paz, and it occupied much of Farallón basin. Its mean swirl speed was ~0.30 m/s along its edge (with a maximum over 0.40 m/s in the SW side), and its mean core radius was ~33 km. The drifter data suggest that eddy C1 was present at least since late June. The trajectory of the blue drifter between 26 and 29 June was toward the east in section D, and then it veered north, probably following the southern and eastern edges of the eddy. From 4 to 14 July, it completed a loop around C1 (~30 km radius) with a mean velocity of ~0.21 m/s. From 19 to 26 July, it made another smaller cyclonic loop inside C1 (~14 km radius, 0.20 m/s mean velocity). The red drifter, which on 20 July was outside La Paz Bay, roughly followed the southern and eastern edges of the eddy until 28 July, at ~0.28 m/s. C1 had a mean vorticity in the eddy core of 1.49 × 10−5 s−1 and a maximum positive vorticity of ~2 × 10−5 s−1.
 Eddy A2 was located around ~26°N (Figure 2), almost centered on Carmen basin. The highest speeds were found in the southern edge (where it interacted with C1), where they ranged between 0.3 and 0.4 m/s. The mean core radius was ~33 km. The presence of A2 was also revealed by the tracks of both drifters, which followed its western and northern edges. The blue drifter showed speeds of ~0.20 m/s (from 29 July to 4 August), and the red drifter (50021) had an average speed of ~0.30 m/s (from 28 July to 3 August). The vorticity of A2 was relatively lower than that of A1 (Figure 2d), with a mean value of −1.1 × 10−5 s−1 in the eddy core and a maximum of ~−1.5 × 10−5 s−1.
 Cyclonic eddy C2, located on Guaymas basin, was sampled with North-South section I (Figure 2a), which was a last-minute change in the cruise plan when the presence of this eddy became apparent in drifter tracks and satellite chlorophyll images. The radius of C2 was ~32 km, and the maximum (LADCP) velocities were ~0.4 m/s. This eddy was well sampled by the drifters. The blue drifter looped once around C2 with speed ~0.27 m/s, in a radius ~37 km, taking 12 days (7–19 August) to complete the loop. The red drifter, after drifting toward the NW (0.5–0.40 m/s; mean ~0.20 m/s) approximately parallel to the mainland coast, veered west off Cabo Lobos and made three loops around eddy C2 with ~30 km mean radius, between 7 August and 4 September, with mean speed ~0.33 m/s (range: 0.09–0.58 m/s) and rotation period between 6 and 9 days. Other drifters (not shown) trapped in this eddy showed that it lasted at least until October 2004.
 In general, the pattern of currents in the 248–304 m layer did not change much relative to the surface layer, although their speed decreased with depth (Figures 3a and 3b; note the change of scale relative to Figures 2a and 2b). The fastest currents (Figure 3b) were in the area of interaction between C1 and A1. The eddies had lower absolute values of W and vorticity (Figures 3c and 3d), but the cores were larger and more elongated than at the surface. The W = 0 and ζ =0 contours were quite close.
3.2 Horizontal Hydrographic Structure
 Figure 4 shows the horizontal distributions of potential temperature (θ), salinity (S), potential density anomaly (γθ), and DO concentration, averaged over the 0–10 m layer. The train of eddies described above based on current data was not apparent in the surface hydrographic variables. This is due to the relative homogeneity of the properties of the surface layer of the SGC. Surface temperature (Figure 4a) was lower close to the peninsula (~28°C –29°C) than that off the mainland coast (29.5°C–31°C). A nucleus with temperature lower than 30°C appeared around 24.5°N–109°W, but it does not seem to be associated to any cyclonic eddy. The salinity distribution showed an overall latitudinal gradient increasing from south to north (Figure 4b); low values (34.6–34.7) were observed at the center of the Gulf entrance (sections Z-X) and close to the mainland coast (34.7–34.8). The 35.0 isohaline apparently marked the boundary between Gulf of California Water and Pacific waters and was shaped like a tongue between sections C and A-X. The maximum values (35.3–35.4) were located in the northern part of the observed region, between sections H and I. γθ showed a gradient similar to that of θ but reversed (Figure 4c), with high values (~22 kg/m3) close to the peninsula and low values on the mainland coast (21.3–21.4 kg/m3). The surface DO concentration (Figure 4d) was homogeneous, at ~4.4 mL/L, with an increase to 4.6 mL/L off the mainland in the gulf mouth.
 In contrast to the lack of evidence of the eddies' presence in the surface properties of Figure 4, the horizontal distribution of hydrographic variables at 300 m (Figure 5) was clearly influenced by the eddy train, being characterized by well-defined cores with concentric circular isolines around relative maxima or minima. Anticyclonic eddies A1 and A2 were marked by cores that, relative to the surroundings, were warm (12°C and 12.5°C, respectively), salty (34.8 and 34.85), and with relatively high DO content (0.5 and 0.8 mL/L). The core of cyclonic eddy C1 was relatively cold (10.5°C), less saline (34.7) and with a lower DO content (~0.25 mL/L) than the surroundings; no concentric isolines of DO were apparent because the values of this variable were very low below the oxycline. The density anomaly contours (Figure 5c) were similar to those of temperature, with lower values (26.45–26.4 kg/m3) in the core of anticyclonic eddies and higher values (26.55 kg/m3) in cyclonic eddy C1.
 The descriptions above are congruent with the lifting of deep (deeper than ~300 m) isolines in cyclones and deepening in anticyclones. The magnitude of the vertical distortion of the deep isolines is illustrated by the topography of the 26.5 and 27.2 kg/m3 isopycnal surfaces (Figure 6). The vertical distortions of the 26.5 kg/m3 surface (Figure 6a) had amplitude ~120 m, with maximum depth (~350 m) in anticyclones A1 and A2 and minimum (~230 m) in cyclone C1. The distortions were smaller (~50 m) for the 27.2 kg/m3 isopycnal surface, with depths between 730 and 740 m for A1 and A2 and 680 m for C1. The center of eddies C1 and A2 seems to be shifted southward with depth (Figure 6b).
 To obtain an idea of the scale of the surface distortion associated to the eddies, we calculated the dynamic height (m) as geopotential anomaly/g and then estimated the anomalies by dividing by the average over the sampled area. This was done for the surface and for 100 m (to avoid near-surface effects), relative to 500 m. At 100 m, (Figure 7a), the maximum elevation at A1 was ~4 cm, C1 had a depression of −2 cm, and A2 had an elevation of 4 cm. The surface dynamic height anomaly relative to 500 m (Figure 7b) shows that the shape of C1 was similar to that in Figure 7a but with a maximum depression of 5–6 cm. Substantial changes of shape are apparent for anticyclones A1 (anomaly ~5–6 cm) and A2 (anomaly ~2 cm), compared to those in Figure 7a; this reflects the effect of the pycnocline topography, which is found above ~75 m, and is probably strongly affected by the wind, as will be discussed later.
3.3 Vertical Structure of Eddies
 Vertical sections across eddies were selected to describe their circulation and hydrography (see section positions in Figure 2): section C for anticyclonic eddy A1 (Figure 8), section H for eddy A2 (see Figure S1 in the Supporting Information), section F for cyclonic eddy C1 (Figure 9), and section I for eddy C2 (Figure S2). Only one anticyclone (A1) and one cyclone (C1) are described here; the graphs corresponding to A2 and C2 are found in the Supporting Information. Note the stretching of the vertical scale above 200 m in Figures 8c–8f and 9c–9f to better show the structure of the upper layers.
3.3.1 Anticyclonic Eddy A1 (Section C)
 The speed across section C (eddy A1) calculated by geostrophic balance (VG, Figure 8a) and measured with LADCP (VLADCP, Figure 8b) show similar structures, with inflow on the peninsula side and outflow spanning the central and eastern parts of the section. The VG inflow (Figure 8a) was ~50 km wide and ~700 m deep, with speed maximum (0.2 m/s) between 100 and 300 m. The VG outflow was ~65 km wide and ~700 m deep; the 0.2 m/s isotach reached ~300 m and the 0.25 m/s reached 200 m, with the maximum (~0.35 m/s) at the surface and at a core at ~100 m depth. The VLADCP inflow (Figure 8b) adjacent to the peninsula was faster (maximum ~0.35 m/s) than the VG inflow (maximum ~0.25 m/s); the core of maximum speed was between 200 and 350 m, while speeds ~0.10 m/s were found at ~600 m. The VLADCP outflow was weaker than the VG outflow; maximum VLADCP outflow speeds (~0. 25 m/s) were above 100 m. On the mainland shelf, a shallow inflow (above 50 m) was also observed, with VG speeds between 0.10 and 0.15 m/s.
 The vertical sections of potential temperature (θ), potential density anomaly (γθ), salinity, and DO across eddy A1 (Figures 8c–8f) show, below ~200 m, a central depression of the isolines. The depressions diminished with depth but were noticeable down to ~700 m. Almost no sideway shift with depth was exhibited by the center of the depression. The 11°C isotherm (Figure 8c) sank from ~290 m over the mainland slope to a depth of ~400 m at its maximum depth. The 26.5 kg/m3 isopycnal deepened from ~260 m over the slope to a depth of 350 m at its maximum depth (Figure 8d). The depression of the 1.5–0.5 mL/L DO isolines (Figure 8f) were located at a distance of 50 km from the peninsula and positioned on the west side of the section; the 1 mL/L isoline sank from ~140 to 240 m, whereas the 0.5 mL/L isoline reached ~310 m. The minimum oxygen zone (≤0.1 mL/L) was distributed between 400 and 850 m.
 In the upper 150 m of A1, isolines of temperature, salinity, density, and DO were not concave but slightly convex (Figures 8c–8f), especially in the top 70 m. Clear convexity was exhibited by the thermocline, that is, by isotherms 18°C to 28°C. The pycnocline (25–22 kg/m3, Figure 9d) and the oxycline (4.5–2.0 mL/L, Figure 8f) also showed slight doming. The near-surface dome shape of the pycnocline caused VG to diminish above ~75 m, producing the VG inflow maximum at 150–300 m depth (Figure 8a).
 The distribution of chlorophyll a (chl-a, color Figure 8d) presented a maximum (0.6–0.9 mg/m3) in the pycnocline and values <0.2 mg/m3 in the 20 m below the surface. The maximum values of chl-a in the thermocline were at the edges of the section (Figure 8d).
 Salinity above 200 m depth (Figure 8e) was structured in three layers: (1) a surface layer (<40 m) with the highest salinities (34.9–35.1); (2) a shallow relative minimum layer with salinity 34.7–34.8 between ~40 and 100 m depth [this layer was thicker (~85 m) and relatively fresher in the western half of the section]; and (3) a subsurface layer of relatively high salinity (34.9–35.0) between ~70 and 230 m, with a 35 km wide core of S = 35.0 between 110 and 150 m depth. Dissolved oxygen (Figure 8f) showed a homogeneous layer (4–4.5 mL/L) between 0 and 40 m and strong stratification (2.5–4 mL/L) between ~40 and 70 m (Figure 8d).
3.3.2 Cyclonic Eddy C1 (Section F)
 The VG distribution across cyclonic eddy C1 (Figure 9a) shows that it was ~550 m deep, was centered on the section, and had maximum speed ~0.25 m/s in the outflow and ~0.30 m/s in the inflow; both maxima were at or very close to the surface. The inflow shows an extension toward the mainland continental shelf (maximum VG ~0.45 m/s). The cyclonic flow pattern shown by VLADCP had a structure in the inflow similar to that of VG but deeper (~650 m), with maximum velocities (~0.25 m/s) between 0 and 300 m. The outflow maximum velocities (~0.20 m/s) occurred between 0 and 100 m.
 The isotherms and isopycnals in section F (Figures 9c and 9d) presented doming in the central part of the section from ~700 m and up to ~60 m. The maximum lifting of isotherms (~75 m) occurred between 150 and 350 m (Figure 9a). In the upper 50 m of the water column, the isotherms and isopycnals present a slight downward tilt toward the mainland. Chl-a (color, Figure 9d) was maximum (0.6–0.8 mg/m3) on the pycnocline at ~40 m; there was a surface layer with <0.2 mg/m3 of chl-a in most of the section, except in the mainland side, where it exceeded this value.
 The upper salinity distribution in section F was arranged in three layers, like in eddies A1 and A2 (Figures 9e and S1e). The maximum salinities (34.9–35.2) occurred in a 40 m deep surface layer. There followed the ~20 m thick shallow salinity minimum (<34.9), centered at 50 m depth. The >34.9 salinity layer was domed and reached ~150 m depth in the center of the section. The DO isolines (Figure 9e) between 350 and 100 m were domed. The depth of the 1.0 mL/L isoline ranged between 145 and 175 m, and the minimum oxygen layer was found between 300 and 800 m.
3.4 Eddies and MODIS Chlorophyll Images
 The along-Gulf sea surface temperature homogeneity in the SGC (Figure 4a) precluded satellite detection of the eddies with that variable, but MODIS CHL images did show them, in part due to the relatively high CHL filaments that emerged from coastal points. This is shown in Figure 10, which shows the 3 day mean CHL image for 9 August. To prove that the structures observed in the image correspond with the eddies described above, the Okubo-Weiss parameter W = 0 contours and the drifter velocities are overlaid on the image.
 The satellite CHL data at each CTD station were highly correlated with the corresponding near-surface CTD chlorophyll a (Figure 11) when vertically averaged over the top 20 and 30 m, which are depths of the order of the first optical depth estimated from the corresponding MODIS K490 image (not shown). Although the satellite CHL values are slightly higher than those from the CTD, the main peaks and lows coincide very well; the correlation coefficients were 0.85 and 0.82 for the 20 and 30 m averages, respectively.
 With the exception of the coastal band and an offshore filament off the mainland at the entrance to the GC (Figure 10) with the maximum surface CHL (~1 mg/m3), concentrations were low (0.1–0.25 mg/m3), but contrasts were sufficient to reveal the eddies and filaments. These structures were immersed in an overall surface CHL gradient increasing from ~0.1 mg/m3 at the Gulf entrance to 0.25 mg/m3 around the archipelago that contains the large islands of Tiburon and Angel de la Guarda, where tidal mixing is strong [Argote et al., 1995].
 On 9 August (Figure 10), relatively high CLH concentrations were observed over the north margins of C1 and C2, where thin filaments of relative high CLH appeared to arc cyclonically from the mainland coast at ~25.7ºN (close to Topolobampo) and ~27ºN (close to Cabo Lobos). In anticyclones A1 and A2, relatively high CHL concentrations were present in the center of each eddy, which were most evident in A1.
4.1 Structure and Position
 The structure of the Southern Gulf of California mesoscale eddies is described here for the first time based on detailed direct observations of currents and hydrographic properties, made in August 2004. This is a much more thorough description than those made previously from direct hydrographic observation [Fernández-Barajas et al., 1994; Figueroa et al., 2003], which did not have such detailed sampling. Four eddies were sampled, two anticyclones and two cyclones, arranged with alternating sense of rotation. Mean surface core radii were ~32–36 km (Figure 2 and Table 1); both anticyclones and cyclones had high surface velocities, with mean speeds ~0.3 m/s and maxima ~0.4–0.5 m/s. The depth of the eddies, given by the 0.05 m/s isotach, was 550–700 m for anticyclones and 400–500 m for cyclones (Figures 8, 9, S1, and S2; Table 1).
Table 1. Characteristics of Eddies in the Southern Gulf of California in the Summer of 2004, From CTD and LADCP Observationsa
Surface Semiaxes [km]
Range of Surface Speeds at Maximum [m/s]
Maximum Surface Vorticity [10−5 s−1]
Mean Surface Vorticity [10−5 s−1]
aThe surface radii are the minor and major radii of the zero-value closed curves of the Okubo-Weiss parameter, that is, the “eddy core.” The mean surface vorticity is the average inside the eddy core.
32 × 39
27 × 40
27 × 38
 Using the surface distributions of W = 0 and ζ = 0 (Figures 3c and 3d) to position the eddies relative to the bathymetry and coastal features (Figure 1), anticyclone A1 covered part of Pescadero basin, centered slightly to the west side of the basin, while A2 was located roughly over Carmen basin. Cyclone C1 occupied much of Farallón basin. Cyclonic eddy C2 was located south of Guaymas basin and north of Carmen basin.
 The data of Fernández-Barajas et al.  from a single line of stations along the longitudinal axis of the SGC during February 1992 showed geostrophic velocities of alternating sign reaching ~1000 m, suggesting the presence of eddies similar to those reported here. However, their cyclonic eddy at Pescadero basin had a radius of ~90–100 km, while the anticyclonic eddy between Carmen and Farallón basins was ~85 km in radius; both estimates are much larger than ours. Their maximum surface speed (0.95 m/s) was also much faster than in our observations, probably because they did not smooth the hydrographic data prior to calculating geostrophic velocity. The sequence of eddies described by Figueroa et al.  has characteristics similar to ours, maximum speeds ~0.5 m/s and ~500 m deep.
 The late summer eddies studied by Pegau et al.  from ocean color images of the SGC had radii ~35–50 km, and they estimated maximum velocities ~0.32 m/s, which is similar to our measurements. In general, the locations of those eddies (anticyclones in Guaymas, Farallón, and south Pescadero basins, and a cyclone in Carmen basin) are not in agreement with the ones reported here, which is not unexpected considering that Figueroa et al.  found no consistent patterns in the reported positions of mesoscale eddies in the SGC.
 Comparing our observations with the results of the numerical model of Zamudio et al.  for the dates of the survey (their Fig. 12e), the model showed an anticyclone covering much of Pescadero basin but was approximately two times larger than A1, which was located SW of that basin. The three remaining eddies were in roughly similar positions but had senses of rotation inverse of those observed directly. These discrepancies can be expected in a numerical model that did not include data assimilation. Compared with the observations, the model eddy radii (~33 km) were similar, the mean speed (>0.45 m/s) was slightly higher, and they were deeper (~1000 m).
 The origin of the SGC eddies is not fully understood, although it has been proposed that they may be caused by instabilities of the poleward Mexican Coastal Current (which outside the GC is ~400 m deep) [Lavín et al., 2006], especially during the passage of coastally trapped waves [Pegau et al., 2002; Zamudio et al., 2008].
4.2 Dynamics Above the Pycnocline
 One of the observed features of all the eddies reported here was that the isopycnals in the pycnocline (and other isolines in that zone) did not have the same concavity as the deeper ones. While the sense of rotation of the eddies was established by the concavity of the deeper (100–500/800 m) isopycnals, domed for cyclones, and depressed for anticyclones, the shallow pycnocline was domed in the anticyclones (Figures 8d and S1d) and flat in the cyclones (Figures 9d and S2d). This caused the maximum geostrophic velocity to tend to be under the pycnocline for anticyclones. The pycnocline in the SGC is very strong and shallow, covering from almost the surface to ~75 m depth (Figures 8, 9, S1, and S2). The effect of the topography of the strong and shallow pycnocline was also evident when comparing the dynamic height (Figure 7) at 100 m and the surface (both relative to 500 m)
 These results suggest that the dynamics of layers above and in the pycnocline may be different from that of the layers below. A possible mechanism for producing the observed pycnocline shape is the interaction of the wind and the eddy surface current. Since the wind stress depends on the wind velocity relative to the ocean surface velocity, even a homogeneous wind can cause a wind stress curl when blowing over eddies [Martin and Richards, 2001; McGillicuddy et al., 2007].
 This was investigated by calculating the wind stress (τ) with the formula
where ρa is the air density, CD is the drag coefficient as defined by Large and Pond  and modified by Trenberth et al. , W is the average of the CCMP daily wind during the period of observations and interpolated to the cruise stations by objective analysis (Figure 12a), and V is the surface current velocity. The time scale of the surface velocity field is mainly imposed by the geostrophic deep eddies, so that the 18 days used for averaging the wind should be appropriate for calculating the wind stress, including the effect of the surface current. The wind variability ellipses (Figure 12a), which were constructed from the standard deviation of the wind components during the 18 days, have smaller major semiaxes than the mean value, and the ellipses are approximately oriented in the direction of the average wind vector (N-NW). The wind stress curl assuming no water surface motion (V = 0) (Figure 12b) shows positive curl (upwelling) in the peninsular half and negative curl (downwelling) in the mainland half of the GC. Eddies A1 and A2 were in the upwelling area and C1 in a downwelling area. The distribution of wind stress curl taking V from the observed LADCP surface currents (V = VLADCP) (Figure 12c) shows a similar overall pattern, but it has a more marked mesoscale structure than with V = 0 (Figure 12b), and some of that structure reflects the effect of the eddies. This is clearer in Figure 13d, which shows the difference between the two wind stress curl distributions. The intense positive curl over anticyclonic eddies A1 and A2 will produce upward Ekman pumping and, therefore, a domed thermocline, as observed in Figures 8b and S2b. The negative curl over cyclonic eddy C1 would tend to depress or flatten the thermocline, as shown in Figures 9b and S2b. Therefore, the mesoscale features of the pycnocline topography may be affected by the wind.
 Beyond the possible effect of the wind on these particular eddies, this leads to a wider consideration of the wind as an important forcing agent for currents in the upper layers of the SGC, at least during summer when the thermocline is strong and shallow. In the numerical model results of Zamudio et al. , wind forcing was shown to be less important than remote forcing for the generation of the deep eddies; however, its importance was not negligible, and the model runs forced only by wind showed that it could have an effect to a depth of ~50 m and could generate weak and shallow eddies.
4.3 Eddies and CHL Satellite Images
 If the mesoscale features of the pycnocline were influenced by the wind, so would the CHL distribution observed from satellites. The chl-a values in images such as those in Figure 10 represent the CHL concentration averaged over the first optical depth, which in this area is around 20–30 m (estimated from the MODIS K490 images corresponding to those in Figure 10). In the GC during summer, there is always a subsurface chl-a maximum (Figures 8d, 9d, S1d, and S2d) inside the pycnocline, at ~50 m depth in this case. In contrast, the surface is usually very low in CHL. Therefore, what the images show is probably the average chlorophyll of the layers above the pycnocline (Figure 11), which means that the chl-a maximum at the pycnocline is undetected.
 The proposed effect of the wind may partly explain why the image of eddies in Figure 10 do not conform with the notion that cyclones have high chl-a in the center and anticyclones have it low. The uplifting of the pycnocline, and with it the chl-a structure, may explain why in the satellite images of Figure 10 anticyclones A1 and A2 contained high CHL in the center relative to their surroundings. In addition, the depression or flattening of the pycnocline may explain why cyclone C2 did not have a surface CHL maximum in its center relative to the surroundings.
 In addition to the potential effect of the wind on structures present in CHL satellite images, there are other processes that are also potentially very important. Satellite images often show high-chlorophyll jets starting as coastal features. It is very likely that ageostrophic currents also have influence; some trapping of surrounding water can occur, especially during eddy formation, and radial velocities can transport high chlorophyll water from eddies’ borders and surroundings toward the center. This subject merits further study.
4.4 Hydrographic Structure
 The temperature and salinity sections across the eddies (Figures 8, 9, S1, and S2) show that they contained the following water masses (using the classification of Lavín et al. [2009, Tables 1 and 4]): (a) Gulf of California Water (S > 34.9) in the top 200 m, with (b) a layer of Shallow Salinity Minimum Water (34.7 ≤ S < 34.9) centered at ~50 m, (c) Subtropical Subsurface Water (34.5 < S < 34.9, 9°C < T < 18°C) between ~200 and ~400 m, and (d) Pacific Intermediate Water, whose core is at ~700 m, in the lower part of the eddies.
 The across-Gulf homogeneity of the upper layer distribution contrasts with that found by Lavín et al.  in cruise NAME-1, made 2 months before the one reported here. Gulf of California Water occupied most of the western side of sections C-F, between surface and ~150 m. The remaining (eastern) part of the sections was occupied by an inflow of Pacific waters, with low salinities (S ≤ 34.7) between 40 and 150 m, and a shallow salinity minimum (34.5–34.6) between ~50 and 70 m. During that cruise, the eddies were not present, and they may have formed from the strong inflowing currents (up to 0.8 m/s) observed at the time, then redistributing the waters between both platforms and maintaining relatively low salinities values (34.7–34.8) in the layer where the shallow salinity minimum occurred.
4.5 VG and VLADCP
 The along-Gulf components of geostrophic (VG) and observed (VLADCP) velocities showed in general a similar structure (Figures 8a, 8b, 9a, 9b, S1a, S1b, S2a, and S2b), but there are some differences, which could in principle be attributable to the presence of ageostrophic flows. The Rossby number for these eddies is calculated as Ro = |ω/f|~ 0.1, where ω was calculated from the linear regression of VG versus r, the radial distance from the eddy center. This means that they were mostly in geostrophic balance, since the centrifugal force was only ~1/10 of Coriolis force, but the centrifugal force was not totally negligible. Assuming that VLADCP represents the gradient-wind velocity, the gradient-wind equation result that for cyclones VLADCP < VG and for anticyclones VLADCP > VG can be confirmed by visual inspection of Figures 8a, 8b, 9a, 9b, S1a, S1b, and S2a.
 The reference level for VG was the deepest common depth (~1000 m in most cast pairs), which implies a considerable decrease of VG in the deep layers. Although some observations indicate that there are significant currents below 1000 m [Roden, 1972; Collins et al., 1997; Mascarenhas et al., 2004], our LADCP observations show low speeds at 1000 m. The greatest difference between VG and VLADCP was in section I (Figures S2a and S2b): the 0.1 m/s isotach of VLADCP reached ~750 m depth, while the same isotach of VG did not go beyond 500 m.
 This article provides the first detailed description of the structure of the deep eddies of the Southern Gulf of California, as observed in August 2004.
 Stratification was very strong near the surface, with the thermocline (which dominated the pycnocline) between the surface and ~100 m showing a temperature difference of 12°C (from 18°C to 30°C). A maximum chlorophyll a layer was present throughout the Southern Gulf, centered at ~50 m, immersed in the pycnocline. The top 200 m were occupied by Gulf of California Water (S > 34.9), with the intrusion of a shallow salinity minimum layer at ~50 m. Between 200 and 400 m, there was Subtropical Subsurface Water, and Pacific Intermediate Water was found from 400 to the 1000 m sampled.
 Two cyclonic and two anticyclonic contiguous eddies were sampled, arranged with alternating sense of rotation. Eddy core mean surface radii were ~32–36 km, as measured with the Okubo-Weiss parameter. This parameter also showed that the eddies were very close together, and that the surface circulation in the sampled area was dominated by the eddies. Mean surface velocities for the observed eddies were ~0.3 m/s, with maxima ~0.5 m/s. This gives a Rossby number ~0.1, which means that they were not completely geostrophic. The depth of the eddies, given by the 0.05 m/s isotach, show that the anticyclones were deeper (550–700 m) than the cyclones (400–500 m).
 The sense of rotation of these eddies was imposed by the shape of the isopycnals in depths 200–700 m: in cyclones, these isopycnals were uplifted by ~70 m in their center, while those in anticyclones were depressed by ~100 m. By contrast, the pycnocline (0–100 m depth) together with the chlorophyll a maximum layer was concave over the anticyclones and flat over the cyclones. This may have caused the satellite CHL images to show high values inside anticyclones and low in one cyclone.
 We propose that because of the shallowness of the pycnocline, the wind stress curl can force those anomalies, possibly by interacting with the eddy surface velocity, and that in general the wind can affect the topography of the pycnocline in the Southern Gulf of California, which could explain the reported presence of shallow (~70 m deep) eddies in the region [Contreras-Catala et al., 2012]. It is not advisable to draw conclusions involving the eddies’ depth from satellite images. Although the surface velocities have the sense of rotation of the deep eddies, the surface dynamics of the gulf seems to be dominated by the wind stress and wind stress curl, which could affect surface processes such as primary production or the heat and salt balance.
 This is a product of project “The Role of Oceanic Processes on the Gulf of California SST Evolution during the North American Monsoon Experiment,” which was part of the North American Monsoon Experiment (NOAA contract GC04-219, P. I. Michael Douglas). This work was also supported by CONACyT (Mexico) projects D41881-F (P.I. MFL) and C01–25343 (P.I. RC), by UABC projects (P-0324 and P-0352), and by CICESE. M.F.L. was at SIO-UCSD as recipient of a UCMEXUS-CONACYT sabbatical scholarship, hosted by the late Prof. P. Niiler, while working on this article. Thanks to Mayra Pazos and the Drifter Data Assembly Center (GDP/NOAA) for handling drifter data. Technical support provided by A. Amador, C. Cabrera, R. Camacho, J. García, and C. Flores. We thank the support of the skipper and crew of the B/O Francisco de Ulloa.