Monitoring the greater San Pedro Bay region using autonomous underwater gliders during fall of 2006
Abstract
[1] Glider surveys of the greater San Pedro Bay region in the Southern California Bight during the fall of 2006 demonstrated the utility of autonomous underwater gliders in a coastal region with complex flow and significant anthropogenic inputs. Three Spray gliders repeatedly surveyed between Santa Catalina Island and the coast of Southern California collecting profiles of temperature, salinity, and chlorophyll fluorescence and estimates of vertically averaged currents. These observations provided context for shelf transport studies during the Huntington Beach 2006 experiment and showed the transition from summer to winter conditions. Vertically averaged currents were predominantly poleward following topography with horizontal scales of approximately 20 km. The gliders surveyed a small cyclonic eddy near Santa Catalina Island and provided a unique view of the structure of the eddy. Nitrate concentration within the euphotic zone was estimated to be 19% greater within the eddy and led to significantly elevated chlorophyll concentrations at the subsurface maximum. Glider observations of salinity reliably detected the distinctly fresh signature of the effluent plume from an ocean outfall near Huntington Beach, California. The salinity anomaly caused by the plume was used to track the spread of the plume as it was advected poleward and away from the coast while remaining subsurface.
1. Introduction
[2] The Southern California Bight is notable for its eddy-rich circulation and for being adjacent to a highly developed coastal region. Complex bathymetry and offshore islands further complicate flow within the bight, and proximity to urban centers leads to substantial anthropogenic input into the coastal ocean from sources such as ocean outfalls. The purpose of this study is to demonstrate the utility of autonomous underwater gliders in such a region. Using glider surveys within the greater San Pedro Bay region of the Southern California Bight during the second half of 2006, we show the physical and biological properties of a small eddy and track the effluent plume from a large ocean outfall. Glider observations provide a unique view of the three-dimensional structure of the eddy. The freshwater signature of the effluent plume is useful for tracking its spread and can be detected reliably by glider salinity measurements.
[3] The greater San Pedro Bay (SPB) region includes the San Pedro Shelf off of Orange County and southern Los Angeles County and the relatively deep San Pedro Basin between the shelf and Santa Catalina Island (Figure 1). Flow within SPB is usually poleward except in spring [Lynn and Simpson, 1987] and generally follows the bathymetry [Hickey, 1992]. Circulation and temperature distributions within SPB have been observed from mooring arrays, but there is a significant lack of salinity measurements in the region. Lynn and Simpson [1987] used CalCOFI data to characterize the seasonal variability of temperature, salinity, and currents within the California Current System, but only three CalCOFI stations (90.28, 90.30, and 90.35) fall within SPB.

[4] Small eddies are common within SPB and throughout the Southern California Bight. Remote sensing of the ocean surface using synthetic aperture radar reveals numerous eddies smaller than 50 km in diameter that are predominately cyclonic [DiGiacomo and Holt, 2001]. Modeling studies suggest that eddies within SPB are largely the result of alongshore flow impinging upon Santa Catalina Island [Dong and McWilliams, 2007]. Despite the ubiquity of these small eddies, in situ observations of their vertical structure have not been presented. Autonomous underwater gliders are well suited to collect these observations because of their ability to collect vertical profiles with horizontal spacing of a few kilometers or less and temporal resolution of a few hours.
[5] The Orange County Sanitation District (OCSD) discharges on the order of 1 × 106 m3 d−1 of treated wastewater through an outfall diffuser on the San Pedro Shelf off Huntington Beach, California [Boehm et al., 2002; OCSD, 2007]. The 3 m diameter outfall pipe extends 7.5 km offshore and discharges wastewater along the 60-m isobath (Figure 1) through a series of diffuser ports that are designed to dilute effluent by at least 100 fold upon initial release [Wu et al., 1994; Boehm et al., 2002]. Several previous studies have focused on the effluent plume from the OCSD outfall. No studies have shown the outfall to be a source of contamination to Orange County beaches, but tidal or diurnal processes may lead to cross-shelf transport of diluted effluent that could bring it to the surface nearshore [Boehm et al., 2002; Hamilton et al., 2004]. Autonomous gliders have the potential to provide spatially and temporally broad monitoring of the fate of the effluent plume if it can be reliably identified using available glider data.
[6] Some of the glider observations used here were collected as part of the Huntington Beach 2006 (HB06) experiment. The intensive field campaign during HB06 was designed to measure processes that transport and disperse sediment, biota, and contaminants in the nearshore ocean. HB06 focused on the San Pedro Shelf near Huntington Beach, California during October 2006 (HB06 Website, http://www.sccoos.org/projects/hb06).
[7] This paper focuses on the observations collected by Spray gliders in SPB during the fall of 2006. The spatial and temporal coverage of the gliders allows us to characterize the oceanographic conditions throughout SPB and provide a regional context for the smaller region of immediate focus during HB06. Glider measurements of salinity reveal structure not previously discussed in this region, and salinity proves to be a reliable tracer for oceanic discharge. The remainder of this paper is organized as follows: section 2 describes the Spray glider deployments and other observations used in this analysis; section 3.1 describes the physical and biological state of SPB and the transition from summer to winter conditions; section 3.2 presents observations of a small cyclonic eddy and an estimate of the nutrient flux to the euphotic zone in that eddy; section 3.3 describes the identification and tracking of the effluent plume from the OCSD outfall using salinity anomalies; and section 4 summarizes the results. Three appendices detail the correction of fluorescence measurements, calculation of current estimates from glider data, and comparisons between glider and mooring measurements.
2. Data and Methods
[8] We primarily use data from three Spray gliders [Sherman et al., 2001; Davis et al., 2002; Rudnick et al., 2004] deployed in SPB during the fall of 2006. We supplement those data with mooring data over the San Pedro Shelf and CalCOFI station data. The Spray deployments are described in section 2.1, and the mooring and CalCOFI data are described in section 2.2.
2.1. Glider Deployments
[9] Spray gliders are buoyancy-powered autonomous underwater vehicles that, in this application, profiled from the surface to 500 m with horizontal resolution of approximately 3 km [Sherman et al., 2001]. Two Spray gliders were deployed in SPB as part of the HB06 experiment. Just before recovery of the two HB06 gliders, a third glider was deployed along CalCOFI Line 90. The two HB06 gliders were deployed near Dana Point, California on 23 September 2006 and recovered near the same location on 20–21 October 2006, and the Line 90 (L90) glider was deployed from Dana Point on 19 October 2006 and recovered on 19 January 2007. All three gliders measured temperature, salinity, density, chlorophyll fluorescence, and acoustic backscatter at 750 kHz. We corrected chlorophyll fluorescence data and calculated current estimates as detailed in the appendices. Deployment statistics are given in Table 1.
Glider | SPS | SCI | L90 |
---|---|---|---|
Deployment date | 23 September | 23 September | 19 October |
Recovery date | 21 October | 20 October | 19 January |
Number of days | 28 | 27 | 92 |
Number of Profiles | 716 | 289 | 846 |
Spatial resolution (km) | 0.74 ± 0.86 | 1.94 ± 0.87 | 2.43 ± 0.82 |
Temporal resolution (hours) | 0.90 ± 0.81 | 2.25 ± 0.68 | 2.59 ± 0.57 |
- a Spatial and temporal resolutions are given as the mean value for each deployment plus or minus the standard deviation for each deployment.
[10] One of the HB06 gliders (referred to as the San Pedro Shelf (SPS) glider) completed four circuits of a survey pattern over the San Pedro Shelf in the vicinity of the OCSD ocean outfall, southward along the coast and along the inshore edge of the San Pedro Channel (Figure 1). Most of the 716 profiles collected by the SPS glider were in water shallower than 100 m, but the glider made dives to 500 m when in deep water. In many cases, several shallow profiles were completed without the glider surfacing to obtain a GPS fix. Consequently, the times and locations of some profiles were interpolated, and there were fewer estimates of vertically averaged velocity than profiles from the SPS glider. Because of the shallow profiles, horizontal (temporal) resolution was highest for the SPS glider (Table 1).
[11] The second HB06 glider (referred to as the Santa Catalina Island (SCI) glider) surveyed primarily over the San Pedro Basin just inshore of Santa Catalina Island (Figure 1). The glider completed four sections along a zig-zag pattern near Santa Catalina Island, as well as two complete crossings of the San Pedro Channel and several sections along the southeastern segment of the SPS glider's survey pattern. The SCI glider completed 289 profiles, most of which were to near 500 m depth. The glider was generally kept clear of the shipping lane to avoid collisions with vessels during surface intervals; the combined survey patterns of the SCI and SPS gliders covered most of SPB.
[12] The L90 glider completed four surveys along the roughly 500 km length of CalCOFI Line 90 and collected 846 profiles during its 3-month deployment. For this analysis we considered only the portions of the L90 glider's surveys while it was within SPB (Figure 1). By 23 October, it had progressed offshore of Santa Catalina Island and beyond the area surveyed by the other gliders; it returned to the region between 30 November and 5 December and again from 12–14 January 2007.
2.2. Mooring Data and CalCOFI Stations
[13] Numerous surface- and bottom-mounted moorings were deployed over the San Pedro Shelf during the fall of 2006 as part of HB06. While the Spray gliders rarely surveyed inshore of the 50-m isobath (Figure 1), the mooring observations extended from the 60-m isobath to inshore of the 10-m isobath near Huntington Beach (Figure 2). Surface moorings generally contained temperature loggers with 2−10 m vertical resolution throughout the water column and conductivity sensors at selected depths. Velocity profiles were available at some mooring locations from surface- or bottom-mounted RDI ADCPs. Preliminary data from seven moorings (MA-MG) deployed along the OCSD outfall pipe (Figure 2) were considered in this analysis, though only data from the MA site are shown. The moorings used here were deployed by OCSD and the United States Geological Survey. The third appendix compares the glider measurements with concurrent mooring measurements of salinity and currents.

[14] Three CalCOFI stations (90.28, 90.30, and 90.35) within SPB were sampled by R/V Roger Revelle on 27 October. Cast data from those stations are available online (http://www.calcofi.org/data/CTD/). Station 90.28 was very near the coast in 72 m of water, 90.30 was in the middle of the San Pedro Channel in 617 m of water, and 90.35 was southeast of Santa Catalina Island in 341 m of water (Figure 1). We use the downcast profiles of nitrate and potential density from these three stations. A high-resolution profile of nitrate was not available for Station 90.35, so we use nitrate concentrations from bottle samples for that station.
3. Results and Discussion
3.1. Regional Conditions
[15] Most of the observational effort during HB06 focused on the inner shelf, and the Spray gliders were the only tools for observing subsurface processes over the greater San Pedro Bay region. The glider observations showed the physical (section 3.1.1) and biological (section 3.1.2) conditions throughout SPB during HB06 and provided context for the eddy and effluent plume observations of sections 3.2 and 3.3.
3.1.1. Physical Features
[16] Vertically averaged currents were generally northwestward during September and October 2006 (Figure 3). The mean vertically averaged current speeds were 0.055 m s−1 and 0.050 m s−1 during the deployments of SPS and SCI gliders, respectively. These average currents are not directly comparable since vertical averaging occurs over the depth of each glider profile and most dives by the SCI glider were to 500 m depth while the SPS glider made many more shallow dives (Table 1). Vertically averaged currents rarely exceeded 0.2 m s−1, allowing the gliders to navigate well throughout the survey period.

[17] To estimate the mean circulation in SPB during the concurrent deployments of the SPS and SCI gliders, we objectively mapped the observed vertically averaged currents. We estimated autocorrelations for the east and north components of the vertically averaged velocity using along-track separation so that observations near the same locations but separated by large times were given less weight. These empirical autocorrelations showed an abrupt decrease in correlation at very small scales that we attributed to noise and were otherwise well modeled by a Gaussian with a characteristic scale of 17.6 to 26.2 km. We used a characteristic length scale of 20 km for the objective map. We constrained the mapped vertically averaged currents to be nondivergent and included noise with a noise-to-signal ratio of 0.2. Bathymetry was not explicitly considered in the objective map, but we excluded current measurements from profiles shallower than 50 m and masked the resulting map to exclude areas shallower than 50 m. We also masked mapped currents offshore of Santa Catalina Island where there were no data. The resulting mean flow (Figure 3, black vectors) was generally poleward and closely followed the bathymetry [Hickey, 1992; Hickey et al., 2003]. In particular, the mean flow turned nearly due west where the San Pedro Shelf widens near Huntington Beach. The cyclonic signature of the eddy southeast of Santa Catalina Island (section 3.2) was apparent even in the mean currents.
[18] Glider surveys revealed a subsurface salinity minimum (SSM) as a dominant feature in the region (Figure 4). The SSM was a remnant of subarctic water from the core of the California Current that was advected eastward and northward, consistent with the circulation of the Southern California Eddy [Hickey, 1979; Lynn and Simpson, 1987]. For the average of all profiles by the SCI glider, a minimum salinity of 33.27 psu occurred at a depth of 35 m. The SSM was found near the σθ = 24.4 kg m−3 isopycnal throughout the glider surveys. Over the San Pedro Shelf, isopycnals shoaled noticeably, so the depth of the SSM was less than 25 m. Moored measurements of salinity showed the SSM over the San Pedro Shelf to depths as shallow as 8 m (mooring MG), but the strength of the minimum was somewhat reduced over the inner shelf. Thus it seems that recirculation of California Current waters was predominantly over the deeper parts of SPB.

[19] Repeated glider surveys allowed us to characterize the variability of the SSM within San Pedro Bay. Four repeated surveys of each of the patterns flown by the SPS and SCI gliders provided approximately weekly coverage of those patterns from late September to mid-October while four crossings of the San Pedro Channel by the SCI and L90 gliders provided approximately monthly repetition from late September to mid-January 2007. Changes in position and strength of the SSM were relatively small during the surveys of the SPS and SCI gliders; the SSM strengthened by 0.05 psu in the western part of SPB while it weakened by 0.05 psu along the San Pedro Shelf (Figure 4). Monthly surveys across SPB showed that the SSM weakened substantially in late fall. By mid-January, salinity at the SSM increased to 33.45 psu at a depth of 55 m (Figure 4). The observed weakening of the SSM may have resulted from reduced recirculation of California Current waters within the Southern California Bight following the summertime intensification of the Southern California Eddy [Lynn and Simpson, 1987]. Vertically averaged currents during the latter part of the deployment of the L90 glider were frequently onshore or equatorward (Figure 3), consistent with reduced recirculation and diminishment of the SSM.
[20] Overlying the subsurface salinity minimum was a surface mixed layer with thickness generally less than 20 m (13.2 ± 5.3 m for the SCI glider) with the mixed layer thickness defined as a density difference from the surface of 0.1 kg m−3 [Rudnick and Ferrari, 1999]. Hickey et al. [2003] reported a similar mixed layer depth for early fall using moored temperature and current measurements within the Southern California Bight during much of 1988. The repeated surveys of the SPS and SCI gliders showed that the mixed layer cooled by 1.5–2.1°C and freshened by 0.07–0.1 psu during September and October (Figure 4). Cooling and freshening were both greater for the SPS pattern, but surveys of that pattern also took 8 days longer than the offshore SCI pattern. Between the second and third surveys there was more substantial cooling (inshore and offshore surveys) and freshening of the mixed layer (inshore survey). This abrupt change in the mixed layer was coincident with the weakening of the SSM noted above. From late September to early December, the mixed layer cooled a total of 4°C, but little additional freshening occurred between late October and early December. While cooling continued into early winter, the mixed layer grew saltier and significantly deeper between the final two surveys by the L90 glider (Figure 4). The increase in salinity in both the mixed layer and SSM in late fall further suggests a change in horizontal advection rather than vertical exchange. On the basis of the depth of the mixed layer in mid-January, strong mixing extended from the surface to about 40 m depth. This mixing did not extend below the SSM to deeper saltier water, so saltier water must have been advected horizontally into the region near the surface.
[21] While significant cooling occurred in the mixed layer, the water column warmed below the pycnocline. The warming was particularly evident in the monthly repeats of the San Pedro Channel crossing where temperatures at 100 m were 1.0°C warmer during the third occupation on 30 November to 5 December than in the preceding surveys and warmed slightly more by mid-January. This warming below the pycnocline is consistent with the results of Hickey et al. [2003]. During this warming, water below 100 m first freshened, then grew saltier again by mid-January (Figure 4).
3.1.2. Subsurface Chlorophyll Maximum and Zooplankton
[22] A subsurface chlorophyll maximum (SCM) was observed throughout the region (Figure 4). The SCM was located slightly below, but generally within, the salinity minimum. Throughout the glider surveys, the SCM was found between the 24.5 kg m−3 and 25.2 kg m−3 isopycnals, though patches of high chlorophyll throughout the mixed layer occurred over the shelf. Like the SSM, the SCM was found shallower over the San Pedro Shelf where associated isopycnals were found closer to the surface (10–30 m over the shelf and 30–50 m over the San Pedro Basin). This shoaling of the SCM over the shelf may have contributed to the slightly enhanced fluorescence in the surveys by the SPS glider close to the San Pedro Shelf (Figure 4c).
[23] During the inshore surveys of the SPS glider, the depth of the SCM decreased by 20 m, while the depth of the SCM in the offshore surveys of the SCI glider remained virtually constant during the month of observations. By late November, the SCM over the San Pedro Basin had shoaled by more than 10 m to a depth 25 m. The first and last surveys of the offshore pattern by the SCI glider and the first two surveys across the San Pedro Channel by the SCI and L90 gliders showed elevated chlorophyll fluorescence compared to the other surveys of those regions (Figure 4). Portions of those surveys were within a cyclonic eddy that produced locally elevated chlorophyll levels (section 3.2).
[24] Acoustic backscatter measurements throughout SPB revealed a strong diel signal with high backscatter just below the mixed layer at night and deeper than 200 m during the day (Figure 5 shows a typical section). A similar signal has been noted in other glider surveys in Southern California [Davis et al., 2008] and is attributed to vertically migrating zooplankton feeding in shallow waters at night and retreating to deep water to avoid predation during daylight. Indeed, the observed maximum in backscatter was colocated with the SCM during night, suggesting that the migrating zooplankton were feeding on the abundant phytoplankton within the SCM.

3.2. Eddy Observations
[25] A small but persistent cyclonic eddy was sampled twice by the SCI glider and a third time by the L90 glider during the beginning of its deployment on CalCOFI Line 90. Vertically averaged currents during three passes of the gliders through the region to the southeast of Santa Catalina Island showed clear cyclonic rotation centered near 33.3°N, 118.1°W (Figure 6). The two passes by the SCI glider through this region were the only periods in which southward currents were observed consistently by that glider. The observations were made on 24–26 September, 11–14 October, and 20–22 October. Although it was not possible from the available observations to determine definitively that the three sets of observations were of the same eddy, the collocation of cyclonic currents suggested that this was the case.

[26] If the observed eddy remained stationary for at least one month, then it was somewhat different than most small eddies reported or modeled in the region. The common eddy evolution scenario involves eddies shed from instabilities in island wakes that then propagated downstream [Caldeira et al., 2005; Dong and McWilliams, 2007]. Not only did the eddy observed here not propagate, but it was also located near the upstream end of Santa Catalina Island. The proximity to the island and nearby bathymetric features suggests that the presence of Santa Catalina Island was key to the eddy's formation and persistence, but we do not have a dynamic explanation for the eddy based on our observations. Caldeira et al. [2005] generated a cyclonic eddy near the same location using the ROMS model, and other authors [Owen, 1980; DiGiacomo and Holt, 2001] have reported small eddies to the southeast of Santa Catalina Island in the fall, but none have suggested specific formation mechanisms.
[27] Whereas most previous observations of small eddies in this region showed only surface signatures of the eddies [DiGiacomo and Holt, 2001], glider observations provided information on both the vertical and horizontal structure of the eddy. Doming isopycnals were evident in each pass through the eddy, with vertical displacements of 15–20 m below the mixed layer over a horizontal scale of 30–50 km (Figure 7 is representative of the passes). The doming of these isopycnals generated geostrophic velocities (referenced to the observed vertically averaged currents and smoothed over 10 km) that were consistent with a cyclonic eddy extending to 80–120 m depth (Figure 7).

[28] On the basis of the cyclonic signature in the vertically averaged currents and domed isopycnals, we identified particular dives as being within the eddy. Dives 18–44 and 187–226 of the SCI glider, and dives 16–35 of the L90 glider were selected based on these criteria. We calculated an average chlorophyll profile for each group of dives and compared those with the mean chlorophyll profile for all observations within SPB that were away from the San Pedro Shelf (west of 117.9°W) and not within the eddy. Confidence intervals about these mean profiles were constructed assuming a t distribution and using the means and standard deviations of the groups of profiles. The observations showed chlorophyll concentrations at the SCM within the eddy to be significantly greater (at the 90% significance level) than over the entire region (Figure 8). The eddy had little effect on surface chlorophyll concentrations.

[29] Nutrient availability within the euphotic zone is often the limiting factor in phytoplankton abundance. Nitrate is one such nutrient for which measurements are readily available. Shoaling isopycnals are expected to bring increased nutrients into the euphotic zone [McGillicuddy et al., 1998], and to bring isopycnal-following phytoplankton into better illuminated waters where their light-dependent growth rates increase. Both of these effects should stimulate phytoplankton growth and increase phytoplankton abundance.


[31] We calculated nitrate profiles for each of the selected dives using the observed density profiles and (1). We created a mean nitrate profile outside of the eddy from the mean density profile from all dives of the SCI glider outside of the eddy and away from the shelf. The nitrate anomaly inside the eddy was defined relative to this mean nitrate profile. Integrating the nitrate anomaly between the surface and 100 m and averaging over all profiles within the eddy gave an estimated nitrate input due to doming isopycnals within the eddy of 0.15 mol m−2. Integration over the upper 100 m was chosen to approximate the euphotic zone based on chlorophyll levels. McGillicuddy et al. [1998] performed a similar estimate of nitrate input to the euphotic zone for large open ocean eddies moving past a site in the oligotrophic Sargasso Sea. Their estimate of the annual input of nitrate to the euphotic zone was 0.19 mol m−2. Located near a nutrient-rich coastal upwelling zone, the eddy near Santa Catalina Island was responsible for a similar nitrate flux over a period of a few months and despite being much smaller. The estimated increase in nitrate within the observed eddy represented more than 19% of the total nitrate in the euphotic zone. The doming isopycnals within the eddy increased nutrient inputs to the euphotic zone and allowed for increased local phytoplankton abundance.
3.3. Effluent Plume Tracking and Characterization
[32] The survey pattern of the SPS glider took it in the vicinity of the OCSD ocean outfall. Hamilton et al. [2004] showed that the effluent plume from the OCSD outfall was characterized by a salinity minimum at temperatures below 14°C that stood out from the background temperature-salinity profile. Salinity sections near the OCSD outfall showed two distinct salinity minima (Figure 10). The shallower of these minima (near the σθ = 24.4 kg m−3 isopycnal) was the shallow salinity minimum discussed above. Near the σθ = 25.0 kg m−3 isopycnal was the salinity anomaly that characterized the effluent plume.

[33] Near the 200 km point along the glider track (Figure 10), the salinity minimum that indicated the plume extended at least 0.1 kg m−3 deeper than the σθ = 25.0 kg m−3 and beyond the glider's profiling depth. That portion of the glider's track was nearly on top of the OCSD outfall diffuser, so we suspect that this may be observational evidence of the buoyant rise of the effluent plume [see Roberts, 1999]. Examination of individual profiles from that portion of the survey also revealed some small density inversions that may have been due to buoyant water rising from the outfall.
3.3.1. Manual Detection
[34] Previous authors [Jones et al., 2002; Hamilton et al., 2004] used salinities below an empirical threshold combined with detectable levels of fecal indicator bacteria to identify the effluent plume from the OCSD outfall. The SPS glider did not measure an independent indicator of the effluent plume, and the proximity of the SSM to the effluent plume made application of a simple threshold salinity unreliable, so we used anomalies in salinity to detect the plume. The variability in vertical structure between profiles over the entire deployment led to ambiguity in identifying the plume based on fixed salinity thresholds [Wu et al., 1994], so we considered profiles individually rather than in bulk as in previous studies [Jones et al., 2002; Hamilton et al., 2004]. We identified any portion of a θ-S profile with (1) a local minimum in salinity, (2) potential temperature less than 14°C, (3) potential density anomaly, σθ, greater than or equal to 24.6 kg m−3, and (4) a deviation from the background θ-S profile that was greater in magnitude than the small-scale variability in the remainder of the profile as indicative of the effluent plume from the OCSD outfall. The profile from dive 90 of the SPS glider (Figure 11) showed this anomaly particularly well.

[35] Manual evaluation of these criteria for each of the 716 profiles collected by the SPS glider found that 199 profiles indicated the presence of the effluent plume. Although location of a profile was not a criterion for identifying the effluent plume, nearly all of the dives that showed evidence of the plume were located near the OCSD outfall (Figure 12). Only 8 out of the 199 profiles were more than 6 km from the along-isobath portion of the outfall pipe. The proximity of the selected dives to the outfall pipe suggested that the criteria used in detecting the plume were reliable.

3.3.2. Automatic Detection
[36] Manual evaluation of the detection criteria given above is tedious and lacks objectivity. To automate the process of plume detection, we developed a filtering method to detect the effluent plume in individual profiles. With this automated method, salinity anomalies remained the primary indicators of the effluent plume. Profiles were analyzed in potential density space since the effluent plume was observed to follow isopycnals more than isobars. Measured salinity profiles were mapped onto uniformly spaced densities (Δσ = 0.01 kg m−3) using linear interpolation. Salinities were intentionally oversampled by the interpolation to avoid loss of variability in the original signal.
[37] In order to effectively identify salinity anomalies due to the effluent plume, the background profile of salinity with density had to be removed from each measured profile. How to accomplish this was the primary difficulty in developing an automatic detection scheme. Since the distribution of salinity as a function of density was reasonably uniform across SPB, we used all measured profiles by the SCI glider to construct a typical salinity versus density profile that did not include the effects of the effluent plume. We subtracted this typical profile from each individual profile measured by the SPS glider. We then filtered the resulting profiles to accentuate variability at the scales typical of the effluent plume. A band-pass filter that admitted variability with scales between 0.15 and 2 kg m−3 worked well. Any filtered profile with a salinity anomaly less than −0.040 psu for 24.7 < σθ < 25.2 kg m−3 was then considered to be indicative of the effluent plume. Figure 13 illustrates this detection scheme using the profile from dive 90 of the SPS glider.

[38] The automatic detection method identified portions of 197 profiles as indicative of the effluent plume. Of those, 126 were also identified by the manual detection. Most of the remaining profiles detected automatically were located in close proximity to both the OCSD outfall and profiles that were identified manually. Some profiles that were relatively far from the outfall were identified (Figure 12), and increasing the selectivity of the algorithm (by changing the band-pass parameters and cutoff values) eliminated many of those profiles from detection. Some profiles near the OCSD outfall were also eliminated when the selection criteria were altered. On the basis of the overall agreement between the manual and automatic detection methods, we believe that the parameters presented for the automatic method were an appropriate compromise between generating false-positive and false-negative results for this outfall.
3.3.3. Plume Transport
[39] The fate of the effluent plume is critical to beach water quality. Under conditions of alongshore flow in May 2000, Boehm et al. [2002] used measurements of E. coli to show that the effluent plume remained offshore and subsurface. The effluent plume from the OCSD outfall was similarly isolated from the nearshore and surface during fall of 2006. The effluent plume settled near the σθ = 25.0 kg m−3 isopycnal. At no point during the Spray surveys did this isopycnal reach the surface, so the effluent plume most likely remained subsurface within the survey region. Salinity measurements from the mooring array indicated that the effluent plume was not transported into the nearshore. The only mooring to show the salinity anomaly characterizing the effluent plume was MA, the mooring furthest from shore. Anomalously low salinities at MA were apparent at 45 m depth, but not at 25 m depth (Figure 11), so the effluent plume was trapped well below the surface.
[40] Most indications of the effluent plume were to the north and west of the OCSD outfall (Figure 12). Although our observations to the south and east of the outfall were sparse, the location of the plume to the northwest of the outfall was consistent with advection of the plume by the mean vertically averaged currents (Figure 3). Profiles that indicated the presence of the effluent plume to the south and southeast of the OCSD outfall may have been the result of the following: the transport of the effluent by currents in the opposite direction of the mean flow; our selection criteria admitting some profiles that were not truly indicative of the effluent plume; or the influence of other, smaller ocean outfalls in the region. Vertically averaged current estimates from the SPS glider on 6–8 October showed southeastward currents at the northern end of its survey pattern. During the same time, ADCP measurements at the MA mooring site indicated that the alongshore component of the current near the depth of the plume (approximately 40 m) switched directions (Figure 14). Thus it is plausible that the effluent plume was advected southeastward along the coast at some times during our observation period.

4. Conclusion
[41] Glider observations in the greater San Pedro Bay region during the fall of 2006 showed the physical and biological conditions during HB06 and the weekly to monthly variability during the transition from summer to winter conditions. The gliders surveyed most of SPB and collected measurements on spatial and temporal scales that have not been previously reported in the region. Moreover, our observations highlighted the importance of salinity structure in this portion of the Southern California Bight. The influence of subarctic waters recirculated in the bight appeared as a subsurface salinity minimum throughout the surveyed region. The strength of the SSM diminished in late fall, likely resulting from reduced recirculation within SPB.
[42] A subsurface chlorophyll maximum was present throughout San Pedro Bay during the fall of 2006, and zooplankton migrated upward to the chlorophyll maximum at night, presumably to feed. The strength and depth of the SCM varied throughout the surveys, notably shoaling over the San Pedro Shelf compared to over deeper water. The SCM was generally stronger over the San Pedro Shelf and strengthened in early winter.
[43] Three glider sections through a small cyclonic eddy to the southeast of Santa Catalina Island provided a novel view of an eddy of this size in the Southern California Bight. The stationary eddy extended from the surface to approximately 100 m depth with doming isopycnals supporting geostrophically balanced cyclonic currents. Consistent with theory, nitrate content in the euphotic zone was estimated to be 19% higher within the eddy, and chlorophyll concentrations were significantly greater within the eddy.
[44] Glider surveys in the vicinity of the OCSD ocean outfall were used to identify and track an effluent plume based on its characteristic low salinity anomaly. Separate manual and automatic detection schemes agreed well, and showed that the effluent plume settled near the 25.0 kg m−3 isopycnal and was advected primarily to the north and west of the outfall by observed currents. No observations showed the effluent plume near the surface, and comparison to mooring-based measurements of salinity over the inner San Pedro Shelf did not show evidence of onshore transport of the effluent plume during September-October 2006.
Acknowledgments
[58] The Spray glider deployments would not have been possible without the support of the Instrument Development Group at the Scripps Institution of Oceanography. Mooring data were collected by George Robertson at OCSD and Marlene Noble at USGS and distributed by SAIC. We thank all those involved in collecting the CalCOFI data used in this analysis. This work was supported by the California Coastal Conservancy's Coastal Ocean Current Mapping Program (COCMP), NOAA through grant NA17RJ1231, and a National Defense Science and Engineering Graduate (NDSEG) Fellowship to R.E. Todd.
Appendix A:: Correcting for Nonphotochemical Quenching
[45] As with all fluorometric measurements of chlorophyll, our sensors measured fluorescence at wavelengths characteristic of chlorophyll a. Voltages recorded by the fluorometers were converted to chlorophyll-like units by scaling using the factory calibrations. As the gliders were unable to collect in situ measurements of chlorophyll concentration, we regarded the chlorophyll concentrations presented here as relative values only.
[46] Because of the relatively slow movement of the gliders, a diel cycle in fluorescence is often observed in shallow waters with reduced fluorescence during daylight [Davis et al., 2008]. This cycle is largely the result of nonphotochemical quenching [e.g., Kiefer, 1973] of fluorescence, and therefore it is not representative of phytoplankton abundance, the quantity in which we are ultimately interested.






[49] It is relatively simple to show that the same result can be obtained by solving for A(z) in (A1) using least squares techniques. Thus we exploit the computational efficiency of the least squares approach in actually calculating the corrected fluorescence signal. The linear regression is calculated over sliding groups of 32 dives (approximately 4 days) and the correction is applied from the surface to the depth at which the correlation of surface light with fluorescence becomes greater than −0.2. That is, we apply the correction until the correlation (which should be negative for nonphotochemical quenching) becomes sufficiently small. This method is similar to that of Davis et al. [2008], but we do not make the assumption that the decay rate of the covariance of measured fluorescence with surface light, 〈(z, t)ϕ(0, t)〉, is equal to the attenuation rate of light.
[50] Because of the large number of dives that did not reach the surface during the deployment of the SPS glider, the technique could not be applied to the data from that glider. Thus only the chlorophyll measurements from the SCI and L90 gliders were corrected for nonphotochemical quenching in this analysis.
Appendix B:: Current Estimates
[51] The Spray gliders obtain a GPS fix upon reaching the surface after each dive and before leaving the surface for the next dive. These GPS fixes provide a measure of the total displacement of the glider while underwater. Measurements of pressure, pitch and heading allow us to determine the glider's displacement relative to the water during a dive. The difference between the measured displacement and the displacement relative to the water divided by the duration of the dive provides a measure of the vertically averaged current during each dive. We correct for surface drift and the slower ascent and decent rates of the glider near the top and bottom of the profiles.
[52] We calculate geostrophic shear from horizontal density gradients between pairs of adjacent profiles in the usual manner. Since we desire spatial gradients rather than temporal gradients, we disregard gradients calculated from those pairs of profiles that are too closely spaced in the horizontal relative to their temporal spacing. Under optimal conditions, the ratio of the horizontal separation of two profiles and the times of those profiles will be on the order of 0.25 m s−1, the speed of the glider through the water. We disregard gradients from pairs of dives in which this ratio is less than 0.125 m s−1.
[53] Integrating the geostrophic shear downward from the surface to a given depth gives the geostophic velocity at that depth relative to the surface. A vertically constant velocity is then added to the resulting velocity profile so that its vertical average is equal to the mean across-track vertically averaged current for each dive pair. As noted in the study by Davis et al. [2008], these geostrophic velocity estimates do not account for Ekman transport.
[54] Profiles of the resulting across-track geostrophic velocities must then be smoothed to eliminate small-scale effects that are not in geostrophic balance. A balance of terms in the equations of motion suggest that the internal Rossby radius is approximately the horizontal scale at which geostrophy is valid. For SPB, the internal Rossby radius is on the order of 10 km, so the profiles are binned to 10 km.
Appendix C:: Comparisons Between Glider and Mooring Measurements
[55] The SPS glider passed within a few hundred meters of the MA mooring during three circuits of its survey pattern. These passes provided an opportunity for comparing glider measurements of salinity and currents to similar measurements from the mooring. Twenty-one glider profiles of temperature and salinity within 0.5 km of the MA mooring and 26 estimates of vertically averaged current within 1.5 km of the mooring were made. A larger region around the mooring was required to obtain a sufficient number of estimates of vertically averaged currents because the glider made multiple profiles without surfacing while near the mooring.
[56] Temperature and salinity were measured by two Sea-Bird Electronics MicroCATs on the mooring at 25 and 45 m depths. We compared salinities along isotherms during the duration of each glider profile. Salinity measurements at 45 m on the mooring were always fresher than the corresponding glider measurements by an average of 0.10 psu. Glider measurements were fresher than mooring measurements at 25 m for all but two profiles with an average difference of 0.06 psu. The mooring measurements bracketed the glider measurements. Without other data, we could not determine which of the three sensors had the most absolute accuracy, but since the glider used a single sensor to collect a salinity profile, any offset in the glider measurements would not have affected the shape of the resulting salinity profile and our analysis in section 3.3 is not affected. The θ-S relationships were similar for the glider and mooring measurements (Figure 11).

