Repeat sampling on hourly time scales using an airborne synthetic aperture radar (SAR) is used to investigate the occurrence and evolving characteristics of spiral-shaped slick patterns, commonly presumed to be indicators of submesoscale ocean eddies, in the area around Santa Catalina Island, California (∼33.4°N, 118.4°W). Simultaneous SAR imagery and boat survey data are examined over two ∼5 h long periods spaced 3 days apart in April 2003. The SAR imagery reveals several spiral-like patterns, roughly 5 km in diameter, occurring downstream of the western end of Catalina. We believe that the most likely formation mechanism for these patterns is current-wake instability related to the flow of the Southern California Countercurrent along the north shore of Catalina. In one case, there is an observed cold-core eddy and vortex sheet attached to the tip of the island, similar to island-wake simulations done by Dong and McWilliams (2007). In another case, the SAR imagery shows a series of slick patterns that, at least initially, resemble spiral eddies, but the data show no clear evidence of actual ocean eddies being present either at depth or through a rotating surface expression. A speculation is that such features signify island-wake eddies that are relatively weak and dissipate quickly. An unexpected finding was how quickly a spiral slick pattern could deteriorate, suggesting a time scale for the surface feature of the order of only several hours. An implication of this result is that care is needed when interpreting a single satellite SAR imagery for evidence of active submesoscale eddies. Recommendations are made for future field studies.
 High-resolution imaging radar systems, such as synthetic aperture radar (SAR), can be used to study and monitor a range of ocean phenomena. For an overview of applications see the Synthetic Aperture Radar (SAR) Marine User's Manual [Jackson and Apel, 2004]. In this paper we report on the use of repeat-sampling SAR measurements to gain insight into the dynamics of small-scale ocean eddies, that is, those having diameters ∼O(10 km) and comparatively short lifetimes. Eddies of this size represent a manifestation of a submesoscale oceanography associated with upper-ocean stirring [Munk et al., 2000], and as such have important implications for dispersal of pollutants, transport of larvae, distribution of nutrients and plankton, and so on [e.g., DiGiacomo and Holt, 2001; Bassin et al., 2005; Caldeira et al., 2005].
 In this work, an airborne SAR is used to investigate eddy patterns in the Southern California Bight (Figure 1). A previous study by DiGiacomo and Holt  used satellite SAR imagery (ERS-1/2 imagery from the period 1992–1998) and other data as available to describe for the first time in detail the characteristics of small-scale eddies in this area. Eddies appeared predominantly as cyclonic spirals having diameters of 10 km or less. About half of the eddies occurred in the vicinity of the irregularly shaped island of Santa Catalina (hereafter, Catalina), which suggested some form of topographic generation. Caldeira et al.  and Dong and McWilliams  proposed that the northwestward flowing Southern California Countercurrent induces a wake downstream of Catalina, consisting in part of “spiral current wake eddies.” Some support for this is that DiGiacomo and Holt  found an increased occurrence of eddies around Catalina during winter, when the countercurrent is typically strongest; on the other hand, eddies were detected both upstream and downstream of the island (relative to the countercurrent), suggesting that eddies in the area may be created through more than one mechanism.
 To investigate the physical characteristics of the small-scale eddies and to explore possible mechanisms for their formation, we conducted a field campaign in April 2003 that combined repeated SAR sampling on hourly time scales with coincident shipboard measurements. Details of the approach are given in the next section, and results are presented in section 3. While the measurements were limited to only a few “good-visualization” days having joint aircraft-boat operations (16 and 19 April), the data do provide new perspectives into the evolution and lifetimes of eddy patterns. The observations are compared with previous studies in section 4; in particular, the “island current wake” hypothesis of Dong and McWilliams  is used to further investigate and assess the physical characteristics found in our data.
 SAR data were collected using the multifrequency polarimetric Airborne Synthetic Aperture Radar (AIRSAR) system flown aboard a NASA DC-8 aircraft [van Zyl et al., 1992]. Only data collected at L band (radar wavelength of ∼24 cm) vertical polarization are analyzed in the present paper. The L band images were found to have better signal-to-noise across the range of available incidence angles (22°–62°). The full polarization L band signatures of slick features in our data set are discussed by Schuler and Lee , and the variation of spiral-eddy slicks with radar frequency band and sea surface elasticity is examined theoretically by Cooper et al. . Flight lines, or passes, were done on reciprocal headings of 121°T and 301°T (approximately east-southeastward and north-northwestward) to align with the long axis of Catalina (see Figure 1). The coverage area included most of the Santa Monica Basin and San Pedro Channel, with each pass ∼100 km long, extending from south of Catalina to the channel islands of Santa Cruz and Anacapa in the north. Most of the observed eddy features occurred near Catalina (matching results of DiGiacomo and Holt ), and this area is the focus of the present study (Figure 1). Because the width of sea surface (or swath) imaged on each aircraft pass was 11.8 km, it took three adjacent passes (identified as 301–3, 121–4, 301–4) and 46 min to sample the study area. Additional flight lines, using the same orientations but done both closer inshore and farther offshore, were also worked into the sampling to provide larger spatial context. In practice, each of the primary passes was repeated, on average, at an interval of ∼1.5 h, although a ∼2 km wide strip of overlap on adjacent passes was typically resampled in about 25 min.
 Postflight analysis of AIRSAR data consisted of survey image processing and precision image processing. Survey imagery, which shows entire flight lines processed to a low-resolution uncalibrated browse product, are available for all our flights from the Jet Propulsion Laboratory data library (http://airsar.jpl.nasa.gov) under mission name “Southern California Coastal Waters, Ocean Eddies Study” and file names “SanMon” (for Santa Monica Basin) followed by a pass identification number (see Table 1). A subset of the precision-processed, calibrated AIRSAR data is available from the Alaska Satellite Facility (https://ursa.asfdaac.alaska.edu/cgi-bin/login/guest/). We found the survey imagery, after being converted to ground-range coordinates, useful both for preliminary analysis and in selecting subscenes for precision processing. An example of the survey imagery is shown in later Figure 3. Subsequent figures show data that has been precision processed. All images are oriented with their horizontal (across-page) axis aligned along 121°T and 301°T.
Table 1. Summary of Synthetic Aperture Radar Imagery Shown in This Papera
Thirteen passes were made on each of the 2 days listed. An a, b, or c in the pass identification (ID) indicates a second, third, or fourth pass done over the same area, respectively.
 In situ measurements were made using the R/V UCLA Sea World. Profiles of temperature and salinity were made underway to ∼35 m depth using a “minibat” system (M. J. Molemaker, Preliminary evaluation of spiral eddy experiment April 2003, Inst. of Geophys. and Planet. Phys., Univ. of Calif., Los Angeles, unpublished manuscript, 2003). Additional measurements were made at 0.6 m depth using a semisubmerged platform towed outside the wake of the boat [Marmorino and Trump, 2000]. Current profiles were measured using a hull-mounted 150 kHz acoustic Doppler current profiler (ADCP) and also a 600 kHz ADCP deployed on the towed platform. Not all measurement systems operated at all times. Typically, the boat made an initial transect from Marina del Rey, near Santa Monica (Figure 1), across San Pedro Channel, and toward the west end of Catalina Island. Coordination between the boat and aircraft relied on a real-time SAR processor on board the aircraft, which provided a display of the L band imagery. This allowed an assessment of candidate eddy slick patterns, whose locations were then relayed via radio to the research boat.
 The primary data collection period was 15–19 April 2003. A series of weather systems passed through the study area, resulting in periods of high wind speed (>7 m/s) on each day (Figure 2). One result of the unsettled weather was that advanced very high resolution radiometer imagery and quick-look RADARSAT products were of little help in guiding the aircraft sampling. Flights made on 16 and 19 April, when the wind speed was relatively low (<3 m/s) over about half of the day and conditions were suitable for visualization, are the focus of this paper. Boat data were collected on each of these days as well. On both 16 and 19 April a total of 13 flight lines were run, corresponding to a total sampling time of nearly 5 h for each day and three or four repeat measurements over a given location. These two sample periods are indicated in Figure 2. A flight planned for 17 April, another day having relatively low winds, was unfortunately canceled because of an aircraft mechanical problem. Flights done under higher winds on 15 and 18 April showed no distinct slick patterns near Catalina or elsewhere in the offshore waters, but they did reveal groups of slick bands occurring inshore where winds were lower, presumably indicating packets of internal waves [e.g., Noble and Xu, 2003]. One additional low-wind flight was done later in the month (on 28 April), but no eddy patterns near Catalina were observed.
 In this section we describe the SAR observations from 16 and 19 April, together with in situ measurements, satellite sea surface temperature maps, and available simulations of oceanographic conditions using a fine-resolution ocean model.
3.1. Case of 16 April
 An overview of the study area on 16 April can be provided by a mosaic of the survey imagery from three SAR passes (Figure 3a). Figure 3b shows an estimate of the surface current field based on changes between the mosaic shown and a repeat mosaic (not shown) completed 1.6 h later. Each surface current vector in Figure 3b was derived by estimating the time displacement of a small-scale slick feature as described by Lyzenga and Marmorino . Vectors derived from adjacent passes (∼25 min apart) yielded similar current vectors but over only narrow data swaths. The SAR-derived flow pattern resembles that shown in Figure 4a, which is a simulation of conditions on 16 April as calculated by the Regional Ocean Modeling System (ROMS) [Li et al., 2008; Dong et al., 2009]. Both the SAR and ROMS results show relatively strong westward flow north of the island, corresponding to the Southern California Countercurrent, and weaker, spatially varying flow to the south. The thermal expression of the countercurrent appears in a sea surface temperature (SST) image for 16 April (Figure 5a), which shows warmer water extending northward into the San Pedro Channel along both the mainland and north of the island. The SAR and ROMS flow fields also show a cyclonic eddy occurring upstream (off the east end) of the island. In the simulation this eddy corresponds to a topographic low. The eddy also appears in the SST image as a relatively cold area, consistent with a doming of isotherms [Todd et al., 2009, Figure 7]. A similar upstream eddy occurs on 19 April in the ROMS simulation (Figure 4b) and SAR data (not shown), but there is no clear SST signature in that case. No in situ data are available east of the island for either 16 or 19 April. The area just off the west end of the island (see highlighted area in Figure 3), where we do have data from a boat survey and which shows an additional area of cyclonic circulation, is discussed below.
3.1.2. West End of Catalina
Figure 6 shows a time sequence of imagery collected on 16 April of the area just downstream of the island. In three of the four scenes (Figures 6b–6d) the research boat is visible, its position at the time of image acquisition being highlighted by a circle. Vectors shown along the boat track indicate the ADCP-measured current vectors at 4 m depth. There is considerable variability between scenes, as well as some perplexing initial structure nearer the island (e.g., Figures 6a and 6b), some of which may be associated with a weak atmospheric front that moved southwestward through the study area between 1600 and 1800 UT, which encompasses the period shown in Figure 6a and 6b. The front resulted in change in wind direction of 100° during the sampling period (Figure 2), although the measured wind speed remained low (<2 m/s).
 Two features of the SAR imagery in Figure 6 are of primary interest. These are a relatively bright, curvilinear signature that extends off the tip of the island (clearest in Figures 6c and 6d, but segments also visible in Figures 6a and 6b) and a set of slicks that curve around the end of the bright SAR feature (clearest in Figure 6c). These slicks have widths of about 100 m and spacings in the range of 0.5–1 km. To shed some light on these, the image pair of Figures 6b and 6c (sample interval of 1.7 h) was used to derive a surface current field on a finer scale than that shown in Figure 3b. The result (Figure 7) shows a small cyclonic eddy, having a diameter of the order of 5 km, a swirl velocity of ∼0.1 m/s, and centered about 2 km below the end of the bright SAR feature, which itself is a current front along which there is both surface convergence and (cyclonic) shear. Interactions between the frontal current gradients and surface gravity waves enhance the small-scale surface roughness (including small-scale wave breaking) and increase radar backscatter along the front, making it brighter than the background [Lyzenga, 1998; Johannessen et al., 2005]. The surface waves mostly involved in this interaction are locally wind driven and of intermediate scale, having wavelengths of the order of 1 m, and not the long ocean swell (wavelength ∼150 m) that is apparent in the SAR images. The cyclonic flow around the end of the front appears to be traced out by the slicks. A close examination of the image sequence (Figure 6) shows, however, that the slicks cannot be traced to the north side of the SAR current front; rather, an apparently independent set of slicks lies just north of and approximately parallel to the front. (This is clearest in Figure 6c.) The slicks are presumed to disappear at the front because wave breaking disrupts the surface surfactant film. Slicks that terminate at a front cannot wind up into a spiral. Away from the front, a number of slick features are relatively unchanged over the entire sampling period; an example is highlighted (by a plus sign) in Figure 6.
3.1.3. In Situ Data
 Based on the initial features seen using the real-time SAR processor aboard the aircraft, the boat was prompted to make a circuit of the area off the west end of the island. The resulting 4 m depth ADCP vectors (Figures 6b–6d) show a flow pattern qualitatively consistent with the SAR-derived flow pattern. There is evidence of near-surface convergence and shear in the initial part of leg AB, then downstream flow on the remainder of leg AB and on leg BC, and a reversing flow (back toward the island) along leg CD. Temperature and salinity measurements made at 0.6 m depth along the three legs of track ABCD are shown in Figure 8a. (The minibat salinity data were contaminated on this day.) There is no significant variation over the track, indicating weak gradients over the large scale of the survey, nor was there a change in the initial part of leg AB. Over the entire survey track, the surface mixed-layer depth (from minibat temperature data not shown) ranged from about 10 to 20 m.
 Conditions along leg EF (made ∼2 h after the SAR data in Figure 6d) are shown separately in Figure 9. The current vectors in this case have been averaged over depths of ∼3 to 25 m to reduce “noise” on scales ∼O(100 m). Spatial variability on those scales may be associated with internal waves, and Figures 6b and 6c show small groupings of dark lines spaced 100–200 m apart that may be internal wave signatures. The vector rotation in Figure 9 is consistent with the track slicing across a ∼5 km diameter cyclonic eddy at some small distance from the eddy's center. Near the point where the vectors are smallest and changing direction most rapidly the near-surface temperature decreases by about 1°C and there is a small increase in the surface salinity. This point can be chosen for a nominal center location of the eddy.
3.2. Case of 19 April
3.2.1. Synthetic Aperture Radar (SAR) Features Downstream of Catalina
 SAR imagery of the area downstream from the west end of Catalina collected on 19 April is shown in Figure 10. The figure is a mosaic of two passes made 24 min apart and so presents a nearly synoptic view. Vectors show the surface current field estimated from the displacement of distinctive slick features, as before, from a second view obtained 1.6 h later. A well-defined area of low radar backscatter (hence, no vectors) extends southwestward from the island and corresponds to an area of low wind in the lee of a mountain range on the western end of Catalina. Island-induced wind wakes are common in the Southern California Bight and have their own unique impact on the sea surface [Caldeira et al., 2005]. As on 16 April, the wind direction changed over the sampling period, in this case by about 200°, while the wind speed remained low (≤3 m/s). The current vector field shows an overall flow toward the northwest, but there is a transition to a more northward flow in the lower part of the area (beneath the dashed line in Figure 10), resulting in an area of cyclonic shear. There is no bright frontal signature extending off the west end of the island as in the 16 April data.
 Three slick features, α, β, and γ, are highlighted in Figure 10. Feature β, being partly sampled by the research boat (see below), is of primary interest. It has the appearance of a cyclonic, ∼5 km diameter spiral eddy, with slicks ∼50–200 m wide and spaced ∼0.5–2 km apart. Spirals with similar characteristics were found near Catalina by DiGiacomo and Holt [2001, Figure 8b]. Feature α is another spiral pattern, similar in size but somewhat less distinct; it lay too far west to have been sampled by the boat. Feature γ, while suggesting cyclonic rotation, has no set of concentric slicks (nor does it develop such slicks over time). The approximately equal spacing of these three features and the pattern of seemingly interconnecting slicks gives at least the impression that the features are linked and show a common dynamical evolution rather than being patterns arising purely by chance. The location of each feature over time, based on four repeat SAR images (points 1–4), is shown in Figure 11. Feature trajectories show a mean motion toward the northwest and a small clockwise rotation that is consistent with surface wind drift following a clockwise rotation of the wind during the latter half of the sampling period (Figure 2).
 All four available views of feature β, covering a period of about 5 h, are shown in Figure 12. There is an unexpected distortion of the feature over time, so much so that the initial spiral pattern becomes nearly unrecognizable in the final view (Figure 12d); nevertheless, smaller slick features that comprise the overall pattern do persist (e.g., the white plus signs in Figure 12). Feature α also exhibits a pattern disintegration over time (noted on Figure 11 by the presence of only two location data points). The image sequence shows no evidence of a rotation of the overall pattern. In order to see if there might be a small eddy embedded within the larger-scale flow field, we used the image pair of Figures 12a and 12b to derive a residual vector field (Figure 13). The residual field shows considerable spatial variability with a speed of ∼5 cm/s, or only ∼10% of the original background flow. If there is any evidence for an eddy, it comes from a very small core region (lying just to the upper right of the white plus sign in Figure 13), but this is tenuous at best. A similar analysis for feature α also shows no evidence of overall rotation, although in that case the innermost part of the spiral had such low backscatter as to be opaque to slick-based feature tracking. Over the larger scale in Figure 13, the residual flow of the spiral is toward the left on the left-hand edge and toward the right on the right-hand edge. This is consistent with the stretching deformation of the spiral pattern apparent in the Figure 12 sequence. Also, there being no indication of a converging current in the residual velocity field of Figure 13 is consistent with an absence of a bright frontal feature. An additional feature of Figure 12 is an increase in number of short dark streaks toward the end of the sampling period (Figure 12d). As on 16 April, these have a spacing of the order of 100 m, too large to be Langmuir circulation “windrows” (expected spacing of ∼2.5 times a mixed-layer depth of ∼15 m) but consistent with the wavelength of transient internal wave packets observed near the west end of Catalina in the 19 April imagery (see Schuler and Lee's  analysis of pass 121–4b).
3.2.2. In Situ Data
 The rapid distortion of feature β, which was being witnessed on board the aircraft, made it difficult to provide an optimal sampling strategy to the research boat. Nevertheless, the boat was directed to make an initial southwestward transect (leg AB), which sampled the area east of the spiral (Figure 12c), and then a northwestward transect (leg BC), which resulted in a cut approximately through the left-center side of the residual spiral pattern (Figure 12d). As before, the location of the boat is highlighted where it appears in the imagery (Figures 12c and 12d only). The 4 m ADCP current vectors show little variation except near the southern ends of the two transects (near B), where the current abruptly turns more northward. This change locally coincides with the dashed line in Figure 10. The relationship (if any) of this “southern” front to the three SAR features is unclear. Water lying north of the southern front is warmer (Figure 8b), and there is a continual increase in temperature northward along leg BC. A warming to the north is consistent with the SST map for 19 April (Figure 5b), although the SST data (acquired at 0258 local time) show overall lower temperatures compared to the in situ data as well as to Figure 5a. This is probably because the in situ measurements (made from 1140 to 1355 local time) include the effect of diurnal warming of the near-surface layer.
 Vertical sections made along leg BC are shown in Figure 14. This leg has its closest approach to the approximate center of the spiral pattern at a distance of about 10 km along the track. The density section shows no obvious doming of the isopycnals, which is the expected signature of a cyclonic eddy. Nor is there an expression of an eddy in the velocity sections, which is consistent with the absence of an eddy in the SAR-derived residual surface current field (Figure 13). The in situ data thus provide no clear evidence for an ocean eddy occurring over the spatial scale of the spiral pattern. This result is consistent with the lack of a vorticity signature in the SAR-derived velocity field. We also note that none of the many ambient or background slicks, as they advect toward the northwest, show any evidence of a cyclonic winding up during the sampling period. In particular, as well as we could judge, slick lines passing very near the west end of the island were not twisted or deformed in any systematic way. Hence, there is no evidence for active generation of spirals. This supports the idea that the spiral signatures we do see for this case are remnant features, with either the eddy having dissipated or the spiral slicks somehow having become separated from the eddy that created them. A final piece of evidence comes from examining boat wakes that cross some of the slicks. Previous work shows that spiral eddies are associated with slick bands that exhibit strong cyclonic shear across them [e.g., Munk et al., 2000; Eldevik and Dysthe, 2002]. However, vessel wakes in Figures 12b and 12c that cross slicks comprising the spiral pattern show no shear distortion [cf. Munk et al., 2000, Figure 7], nor is there horizontal shear in the ADCP current on leg AB where the boat crosses the prominent slick band extending between features β and γ.
4.1. Island Current-Wake Hypothesis
 In the two cases we have examined, SAR features occur near to and approximately downstream of the island. This suggests the current-wake scenario mentioned in section 1 as a possible generation mechanism. Numerical simulations show that current wakes are formed when the Southern California Countercurrent flows northward around Catalina [Caldeira et al., 2005; Dong and McWilliams, 2007]. That this condition was met is shown by the SST maps collected closest in time to the SAR data (Figure 5), as these show evidence of warm water flowing northward into the San Pedro Channel and along the north shore of the island. This is the signature of the warm countercurrent, which is also indicated in the ROMS simulations (Figure 4), in the SAR-derived flow fields (Figure 3), and in the boat data (Figure 8b).
 An example of the formation and evolution of a Catalina wake eddy is shown by Dong and McWilliams [2007, Figure 11]. The process begins when a vortex sheet, having positive (cyclonic) vorticity, is formed along the northern coast of the island. The vortex sheet separates from the western tip of the island, and within 1 day a separated cyclonic eddy forms. Primarily during the eddy generation phase, cold water upwells and cools the surface by up to 0.5°C at the center of the eddy. After moving away from the island (and away from the effects of a weak anticyclonic eddy), the eddy becomes circular and has a diameter (based on its vorticity distribution) of about 6 km. The cycle of eddy creation has a period of between 2.5 and 3.0 days in the simulations.
4.2. Are the SAR Features Current-Wake Eddies?
 Based on the results of Dong and McWilliams , key observables for an island current-wake eddy include a cyclonic vortex sheet, a cyclonic eddy, surface cooling, and an ∼3 day cycle time. Our data sets can be compared with these criteria. We begin with the 16 April data.
 An enlargement of the SST map for 16 April is shown in Figure 15 (left). We have overlain the approximate center position of the 16 April eddy based on the final leg of boat sampling (Figure 9), as well as the approximate location of the bright frontal signature from the final SAR image (Figure 6d). The various data sets are approximately synoptic, as the SAR image precedes the SST data by ∼1.5 h, while the boat data lag the SST data by ∼40 min. It can be seen that the spot (the red plus sign in Figure 15, left) where the boat measured cold water at 0.6 m depth lies along the western edge of an isolated patch of relatively cold surface water in the SST map. The patch has an approximate diameter of about three pixels (2.7 km) and a temperature of 14.8°C, about 1.0°C colder than the surrounding surface water and comparable to the boat measurement. The northern edge of the cold patch lies very nearly along the western half of the bright SAR signature. As that signature corresponds to a zone of cyclonic current shear, it approximates a cyclonic vortex sheet extending from the island. As the cold patch lies approximately within the 5 km diameter cyclonic eddy found in the SAR and boat data sets, it is consistent with upwelling of cold water within the eddy.
 It is illuminating to compare the SAR results directly with the simulations done by Dong and McWilliams , who show the vorticity distribution near Catalina at 0.5 day intervals over a particular 4 day period in March 2002 (although without any data assimilation being used). Figure 16 reproduces the simulated vorticity distribution at day 3.5, which corresponds to the stage where part of the vortex sheet has separated from the western tip of the island to form an attached, asymmetrical eddy. Figure 16b shows that the observed 16 April eddy pattern (slicks and surface flow field) and the SAR frontal signature coincide approximately with the location of the simulated eddy and vortex sheet. This is, of course, not proof that the observations and the simulations are of the same physical phenomena, but the similarities of length scale and position so close to the island are suggestive.
 Similar comparisons can be made for 19 April. Even though the available SAR and boat data do not show the presence of active eddies, we can address whether the SAR signatures might be consistent with eddies that, for some reason, have either dissipated or never formed in a robust, coherent manner. Figure 15 (right) shows an enlargement of the SST map for 19 April with an overlay of the initial positions of the three SAR features, α, β, and γ, found on that day. In this case, the SAR and boat survey data were collected ∼6 h after the SST data. SAR features α and β fall approximately along the southern (cyclonic) edge of the warm countercurrent, which can be seen to extend northwestward from along the north coast of the island. Feature γ appears to fall more in the interior of the warm water and north of the tip of the island. The locations of features α and β are not inconsistent with the simulated eddy positions shown by Dong and McWilliams ; for example, feature β lies very near their simulated eddy on day 1.5, but this alone is not very revealing because the trajectories of both simulated and real eddies may coincidentally cross. More significant would be finding in addition a similarity in the separation between successive eddies. In the simulations this separation is of the order of 30 km, or about five eddy diameters. In the observations, the separation between features α and β and between β and γ is L ∼15 km. If we base an eddy diameter on feature β's initial ∼5 km diameter spiral pattern (Figure 12, left), then L is about three eddy diameters, but if we assume the vorticity distribution had a smaller diameter, say ∼3 km, then L matches the simulated spacing. The eddy separation can be recast in terms of an eddy-shedding period Tp = L/U0 ∼1.2 days, where U0 ∼ 0.15 m/s is a typical upper-layer current speed downstream of the island (Figure 4b). This value of Tp is, however, only about half the value predicted by Dong and McWilliams  for eddy generation at the west end of the island.
 It is interesting that the SST map (Figure 15, right) shows an isolated cold area (about 5 km in diameter) near feature α's location. Could this cold area and feature α (the westernmost feature observed on 19 April) be remnants of the 16 April eddy? Is it possible any cyclonic surface circulation associated with the SST signature dissipated in the ∼6 h before the SAR observations? We have very limited data to address the history of the 16 April eddy and its surface signatures. SST data from 15 April are contaminated by cloud cover. AIRSAR and satellite SAR imagery collected on 18 April show no slicks, because of high winds. However, examination of an SST map for 2145 UT 17 April (not shown) reveals a very similar cold patch to that in Figure 15 (left). If we assume that cold patch also corresponds to the 16 April eddy, then the “dwell” time of the eddy near the tip of the island was 1 day. This is not inconsistent with the simulations, as they show that it takes a developing eddy about 1 day before it separates completely from the island. If the cold patch/eddy separated sometime on 17 April and advected to feature α's location on 19 April, then its translation speed would have been of the order of 25 km in 2 days, or about 0.12 m/s, and comparable to U0.
 In summary, considering the available evidence, we tentatively conclude that the cyclonic feature observed on 16 April is an example of an island current-wake eddy in the early stage of formation. Also, although we cannot discount the possibility that both features derive from some other mechanism entirely, we suspect that features α and β are dissipated current-wake eddies, weaker for some unknown reason than the simulated eddies. For example, moderate to strong westerly winds (opposing the Southern California Countercurrent) on 15 April and 17–19 April might have played some role in modulating and weakening the eddies.
 Our uncertainty in interpreting the data is illustrative of a range of complicating factors that make it difficult to identify unambiguously an island wake signature in SAR, SST, or other data. First there is the generally high level of complexity in the circulation patterns in the area. For example, as Dong and McWilliams  point out, the flow upstream of an island may vary with time and even contain eddy structures generated elsewhere, and the downstream wake may also interact with other features in the flow. They also mention coastline irregularity, such as a headland or peninsula, as being sources for eddy generation. Catalina itself has headlands along the northern coast, which may possibly be source regions for wake eddies; possible examples of this include the eddy in Figure 8 of DiGiacomo and Holt  as well as others closely associated with the northern Catalina coastline seen in the map in their Figure 13. Caldeira et al.  suggest that a small change in the angle of attack of the approaching current can dramatically change the nature of the island current wake. Also, candidate spiral eddies previously identified in the Southern California Bight often do not have a signature in the closest available SST data. Examples include a cyclonic spiral found 10 km from the west end of Catalina by Caldeira et al. [2005, Figure 2] and multiple cyclonic spirals near Catalina as shown by Holt [2004, Figure 2.15].
 Airborne SAR measurements repeated on hourly time scales along with in situ measurements have provided a glimpse of the characteristics, evolution, and lifetimes of eddy-like slick patterns occurring in the vicinity of Santa Catalina Island in the Southern California Bight. The study was limited to relatively short sampling periods on only two “good visualization” days, and we had limited success coordinating our sampling and obtaining a clear view of eddy properties at depth. In this regard, the study should be considered a pilot study that helps to determine the spatial and temporal scales to be sampled in subsequent investigations. We have made some speculations as to the origin of the slick patterns we observed. We believe that the most likely formation mechanism is current-wake instability related to the flow of the Southern California Countercurrent along the north shore of Catalina. The evidence includes several similarities to the Catalina eddies simulated by Dong and McWilliams . In one case, there is observed upwelling near the center of an eddy and a cyclonic circulation pattern and trailing vortex sheet (via both in situ measurements and deductions from SAR-based feature tracking), all features predicted by the simulations. In the second case, three days later the SAR imagery shows a series of “cyclonic spiral” slick patterns that at least initially resemble spiral eddies, but the data show no evidence of actual ocean eddies being present either at depth or through a rotating surface expression. A speculation is that such features arise from weak island-wake eddies that dissipate quickly. This would need to be tested in future work.
 An unexpected finding was how quickly a spiral slick pattern could deteriorate. We now have, at least for one example, an indication of the time scale of the surface feature, being on the order of only several hours. This does not necessarily mean that the underlying structure of a spiral eddy has such a short life span, but it does appear that at least the surface signature can be fairly fragile and short lived. Some care is needed, therefore, in interpreting spiral patterns in single-look satellite SAR imagery as direct evidence of submesoscale eddies. Of the features we observed that most closely resemble spirals (that on 16 April and features α and β on 19 April), only one of three appears associated with cyclonic near-surface currents at the time of observation. Extrapolating from these results, we suspect some of the small-scale eddies in the census of DiGiacomo and Holt , in particular those of a similar dimension occurring near the west end of Catalina, may not have been active eddies when they were observed; nevertheless, DiGiacomo and Holt's conclusions regarding the spatial and temporal statistics of eddies should still be valid. Additional research is warranted on the apparent decay of some eddies and whether the subsurface distribution of organisms and other constituents might yet remain coherent for some time after an eddy becomes inactive and its surface expression deteriorates.
 The results raise the question about how one might do a better job of sampling spirals in a future field study. Given the rapid evolution that is possible, a recommendation is to use SAR resample times of no more than about 20 min. A short resample time would be expected to yield a more accurate, spatially dense, SAR-derived flow field, and this would be useful to have in dynamic areas, for example, near the front observed on 16 April (Figure 7), and over smaller areas, for example, the possible “cyclonic core” in the 19 April data (Figure 13). Also, in hindsight, it would have been better to have considerably reduced the length of our SAR passes, allowing more repeat sampling over the west end of Catalina. Another improvement would be to augment the necessarily duration-limited airborne SAR imagery with additional observational techniques such as “rapid repeat” satellite remote sensing [Holt and Hilland, 2000], a shore-based HF Doppler radar [e.g., Bassin et al., 2005], or even a mountain-top Web cam on Catalina. If a small research boat is again used, it should be more locally deployed from day to day or be considerably faster, to avoid long-duration transects and to be accessible for the kind of adaptive sampling needed to sample possibly rapidly evolving features. Deploying underwater gliders (either remote controlled or preprogrammed) is another approach to doing repetitive sampling over an area suspected to contain eddies [e.g., Todd et al., 2009]. In addition, one can deploy and redeploy satellite-tracked surface drifting buoys with drogues at varying depths from 2, 5, and 10 m. Real-time model simulations should also be included in preparation and execution of the field campaign.
 Future airborne missions might also include an infrared imager, as slicks have high contrast in thermal imagery and can be detected thermally under wind speeds too low for SAR slick detection [Marmorino et al., 2008]. Also, a thermal imager would reveal whether a spiral has a cold core, and it would be helpful in testing new dynamic models for the creation of slicks, such as the “cold filamentary intensification” model of McWilliams et al. . An airborne thermal survey done using a 2.5 km swath (∼4 megapixels using a high-resolution camera) and 185 km/h aircraft speed could cover a 10 by 10 km area off the west end of Catalina with a revisit interval of about 20 min. It is interesting to note that, in perhaps the earliest report of high-resolution airborne infrared ocean imagery, McAlister and McLeish  showed what they suggested was a small spiral eddy formed off the north shore of Catalina.
 This work was funded by the Office of Naval Research through Naval Research Laboratory Work Unit 72-8179 (G.O.M. and M.A.S.) and by the National Aeronautics and Space Administration through a contract with the Jet Propulsion Laboratory, California Institute of Technology (B.H. and P.M.D. plus UCLA boat). We thank JPL scientists Charles Morris (AIRSAR mission planning) and Bruce Chapman and William Fiechter (AIRSAR processing), Walter Klein of NASA Drydent Flight Research Center, who was mission manager for the AIRSAR flights, and Dave Foley of the NOAA Coastwatch Program for providing AVHRR data. We also thank Yi Chao of the Jet Propulsion Laboratory and Jei-Kook Choi of the Naval Oceanographic Office for the use of the ROMS output used in Figure 4. Final preparation of the manuscript was funded in part by NRL Work Unit 72-9201. We thank two anonymous reviewers for their insightful and helpful comments. The manuscript contents are solely those of the authors and do not constitute a statement of policy, decision, or position on behalf of NASA, the U.S. Navy, NOAA, or the U.S. Government. NRL contribution JA/7230-09-0260.