560 JOURNAL OF PHYSICAL VOLUME 29

Lagrangian Exploration of the California Undercurrent, 1992±95

NEWELL GARFIELD,CURTIS A. COLLINS,ROBERT G. PAQUETTE, AND EVERETT CARTER* Department of Oceanography, Naval Postgraduate School, Monterey, California (Manuscript received 25 November 1996, in ®nal form 8 April 1998)

ABSTRACT During the period 1992±95, nineteen isobaric RAFOS ¯oats, placed in the California Undercurrent at inter- mediate depths (150±600 m) off Monterey and San Francisco, California, reveal a region of varying width of subsurface, poleward ¯ow adjacent to the . The ¯oat trajectories exhibit three patterns: pole- ward ¯ow in the undercurrent; reversing, but predominately alongshore, ¯ow adjacent to the continental margin; and, farther offshore, anticyclonic motion accompanied by slow westward drift. Flow continuity of the under- current exists between Pt. Reyes and at least Cape Mendocino with an average speed dependent on the ¯oat depth. Speeds were variable, but common features were acceleration occurring to the south of Pt. Arena and deceleration to the north of Cape Mendocino. An important mechanism for ¯oats, and water, to enter the interior from the undercurrent is through the formation of submesoscale coherent vortices.

1. Introduction the undercurrent. Measurements obtained with surface- tracked parachute drogues in the region off Monterey The California Undercurrent is found adjacent to the California (Reid 1962), Baja California (Wooster and continental slope in the subtropical northeastern Paci®c Jones 1970), and Oregon (Stevenson et al. 1969) con- Ocean. The undercurrent transports equatorial water tributed important information on the existence of the poleward such that the waters next to the slope are undercurrent, but all were short duration deployments warmer, saltier, and lower in oxygen than those found with limited spatial extent. RAFOS ¯oats, the latest gen- farther offshore (Hickey 1979). The undercurrent has eration of acoustically tracked Lagrangian ¯oats, pro- been identi®ed off Baja California (Wooster and Jones vide the ability to tag and track subsurface ¯ow for long 1970), Vancouver Island (Freeland et al. 1984), and at distances and long times at reasonable cost. In the east- many locations in between. The ¯ow continuity has not ern Paci®c the axis of the sound channel is relatively been observed directly (Mooers 1989), but is inferred shallow and it was anticipated that this would allow the from maps of hydrographic property distributions use of these ¯oats to study the relatively shallow un- (Hickey 1979; Chelton 1984; Lynn and Simpson 1987; dercurrent. Tisch et al. 1992) and from direct observations of pole- In 1992, we began a Lagrangian exploration of the ward ¯ow along the continental margin (Noble et al. undercurrent. The goals of our program were to measure 1987; Huyer et al. 1989; Largier et al. 1993; Collins et al. 1996b, Smith et al. 1996). Rischmiller (1993), using the path and continuity of the ¯ow and to determine the data from an acoustically tracked dropsonde, found that character of in the undercurrent. The primary the area of poleward ¯ow off Point Sur extended beyond onshore±offshore exchanges for surface and near-sur- the upper slope to a distance of 200 km from the coast face waters are thought to take place through processes and to a depth of 750±1000 m or greater. Smith et al. associated with dynamics and the offshore demonstrated the presence of the undercurrent through ¯owing ®laments that appear to be anchored to up- ADCP transections occupied between Pt. Conception, welling centers. Other coastal studies (Schwing et al. California, and Oregon. 1991; Rosenfeld et al. 1994; Parker 1996; Baltz 1997; There have been few Lagrangian measurements of Steger et al. 1998, manuscript submitted to Deep-Sea Res., hereafter SSC) have shown that there is an ener- getic ®eld along the continental margin and shelf. Our investigation sought to examine the role of the Cal- * Current af®liation: Taygeta Scienti®c, Monterey, California. ifornia Undercurrent in these continental margin mixing patterns. This report describes the results of our Lagrangian Corresponding author address: Dr. Newell Gar®eld, Department of Geosciences, San Francisco State University, 1600 Holloway Ave., exploration of the undercurrent from ¯oats deployed San Francisco, CA 94132. during the period from 1992 through 1995. Float de- E-mail: gar®[email protected] ployments and data processing are described in section

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC APRIL 1999 GARFIELD ET AL. 561

2. Individual ¯oat trajectories, and pressure and tem- perature data are presented in sections 3 and 4, respec- tively. Eddies are characterized in section 5. The dis- cussion and summary of ®ndings are in section 6. Pre- liminary statistics, including Lagrangian decorrelation scales, are summarized in the appendix.

2. Methods The RAFOS system is described in detail by Rossby et al. (1985, 1986). We describe below only those el- ements of the RAFOS system that were particular to our experiment. To distinguish our ¯oats, we identify them by number with the pre®x ``NPS'' (for Naval Post- graduate School). a. NPS RAFOS ¯oats Quasi-isobaric RAFOS ¯oats (¯oats without pressure compensating devices) were used throughout this por- tion of the experiment. An ideal isobaric ¯oat would settle to the speci®ed pressure surface and remain at that pressure, regardless of the surrounding water. However, RAFOS ¯oat density depends in part on the glass and end cap thermal expansions. Swift and Riser (1994) noted that the RAFOS operational surface will not co- incide with surfaces commonly quanti®ed in oceano- graphic literature. Goodman and Levine (1990) derived a ``response ratio'' to quantify how well a ¯oat follows FIG. 1. Northeast Paci®c between southern California and Wash- either isobaric (response ratio of one) or isopycnal (re- ington locating the four NPS sound sources (SS1 to SS4) and the sponse ratio of zero) surfaces. The response ratio, r ϭ NOAA PMEL sound source (V1). The isobaths for 200, 1000, 2000, Ϫ1 3000, and 4000 m are contoured. The primary launch site is indicated (1 Ϫ s)/(1 Ϫ s ϩ [N/cg ]), compares the ratio (s)of by the open circle. The two arrows show the orientation for rotating the ¯oat's compressibility to that of and the the ¯oat velocity into along- and across-shore directions Brunt±VaÈisaÈlaÈ frequency (N) normalized by cgϪ1 (ഠ3.8 cyc hϪ1). At the target depths for this study the NPS ¯oats had a compressibility ratio of s ഠ 0.4 and a re- resolution of 2.5 decibar (1 dbar ϭ 104 Pa, and a change sponse ratio of r ഠ 0.75±0.8. Most of the variation of of 1 dbar ഠ 1 m vertical change) and an accuracy of the response ratio came from changes in the normalized 1% full scale. buoyancy frequency [N/cgϪ1]. All but the ®rst two NPS ¯oats were ballasted at the b. Float navigation Deep Ocean Simulation Facility, Naval Civil Engineer- ing Laboratory, Port Hueneme, California. The 1.9-m An acoustic sound source array using Webb Research diameter pressure vessel is operated with ®ltered tap 183-dB instruments ensoni®ed the region between 34ЊN water rather than distilled water. Our initial ballasting (Pt. Conception, California), and 43ЊN (Cape Blanco, efforts yielded signi®cant variability from the target Oregon), and extending offshore to at least 132ЊW (Fig. pressures, which was probably due to the unknown tap 1). Four NPS sound sources (SS1 to SS4) were initially water density. We now compensate for this by deter- deployed in August 1992 and were serviced in August mining the tank water density before and after each 1994. The source just offshore from Pt. Arena, SS4, was ballast run using a carefully ballasted and weighted RA- eliminated in 1994 because of confused signal reception FOS hull as a hydrometer. by ¯oats and shadowing by Pt. Arena. In addition to In addition to recording the time of arrival of the the NPS sources, a sound source (V1) was installed by acoustic signals from the moored sound sources used the NOAA Paci®c Marine Environmental Laboratory for ¯oat navigation, our ¯oats recorded temperature and (PMEL) in August 1993, near the Juan de Fuca Ridge. pressure. Temperature and pressure were measured over The NPS sources originally transmitted acoustic signals a 1-s averaging time interval after each listening cycle, three times per day, with each source broadcast stag- which, depending on the mission parameters, meant one gered by a half hour. The schedule was changed in 1994 to three samples per day. Temperature had a resolution to twice per day. The PMEL source also transmits twice of 0.02ЊC and an accuracy of 0.05ЊC. Pressure had a per day, broadcasting one-half hour before the NPS se-

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC 562 JOURNAL OF VOLUME 29

(on the order of tens of meters) that each source remains near the sound channel axis. The ¯oat depths (see be- low) were generally above the sound channel axis. Es- timates of sound trajectories from a relatively small an- gular range indicate that timing errors from different ray paths were probably small. The sound sources are mon- itored through the NPS Pt. Sur Ocean Acoustic Obser- vatory array, a former navy underwater lis- tening station. The standard deviations, after removing linear trends, for arrival times recorded over a two-year period for sources 1, 2, and 3 were, respectively, 0.167 s, 0.162 s, and 0.183 s. Float clocks were calibrated prior to launch, but clock drift once the ¯oat reached thermal equilibrium at depth was not measured. Since ¯oats were typically not re- covered, the ¯oat clock drift was not known and had to be approximated. In determining the ¯oat position, errors from the un- known sound speed and clock drift cannot be easily distinguished, and a single editing approximation to the source clocks was done to account for both (Paquette 1996). After applying this correction, the absolute un- certainty in position is estimated to be less than 10 km, with the short-term relative error for each ¯oat being smaller (generally about 1 km). Data processing used NPS modi®ed versions of the University of Rhode Island programs. Details of the FIG. 2. Dilution of precision (DOP) for the eastern Paci®c computed modi®cations can be found in Paquette (1996). Follow- for the geometry of the three southern sound sources (SS 1±3). Units are kilometers. Note that the northern sources were used for navi- ing range and position determination, quality control gating ¯oats north of Cape Mendocino, not the sources diagramed and statistical analyses were done. Different procedures in this ®gure. were used to smooth the positions, depending upon the sampling schedule. For ¯oats that sampled multiple times every other day, the data were averaged to obtain quence. The RAFOS ¯oats were programmed to listen single ®xes every other day. These positions were then to all sources, allowing for ¯oat positioning two or three ®ltered once with a three-point moving average. For times per day. ¯oats with ®xes evenly spaced with respect to time, Navigational errors arise from several sources. A position data were ®ltered twice with a three-point mov- ``®xed'' error is due to the geometry of the solution and ing triangular ®lter. In order to develop a consistent the instrument precision. The resolution of the instru- dataset, all ¯oat data were then interpolated to daily ment acoustic signal arrival time is 0.3 s. This translates ®xes using a cubic spline interpolation, and velocity was to an error of ϳϮ0.5 km in any range estimate. The determined by differencing these daily positions. Fi- position accuracy factor resulting from this range error nally, the velocity estimate was also smoothed (once) and the site geometry, referred to as the dilution of with a moving three-point triangular ®lter. The inter- precision (dop), is a multiplier for other errors. Figure polated and smoothed data were used in this paper. 2 contours the dop for navigation using the three south- ern sound sources in the region where most of our ¯oats c. Experimental design operated. A small dop indicates that the solution ge- ometry introduces only a small navigation error. The Our sampling plan was to launch triads of RAFOS dop gives a lower bound on the navigation accuracy ¯oats in the California Undercurrent. The primary error that can be expected given the geometry of the launch site was west of San Francisco (Fig. 1). Here, solution for the different ranges. Other navigation error triads of ¯oats were launched in a triangular pattern 8 arises primarily from the unknown sound speed and km on a side. Two ¯oats were launched above the 1000- clock drift. m isobath, while the third was launched at the offshore The sound sources were all moored at the climato- apex of the equilateral triangle. logical mean depth of the deep sound channel (some- The ®rst two ¯oats (NPS1 and 2) were programmed times called the SOFAR channel) at each location, 510± for 30-day missions in order to check system operations. 590 m (Johnson and Norris 1968). Although the channel Subsequent subsurface mission lengths were increased, depth varies seasonally, the variability is small enough ®rst to 60 days, then to 100 or more days (Fig. 3a).

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC APRIL 1999 GARFIELD ET AL. 563

FIG. 4. Depth range (mean, median, and standard deviation) of the NPS RAFOS ¯oats. The NPS ¯oat number is in italics to the right of the mean depth, and the number of daily pressure values used to determine the statistics is below the variance bar. The depth for NPS14 was estimated from the temperature record. The depth variability is placed at the launch date for each ¯oat.

after we realized the effectiveness of the acoustic trans- missions, 275 dbar. Cruises were not dedicated to ¯oat deployments, so launches were conducted as opportunity arose during other research cruises. Our only criterion was that the time interval between ¯oat triad launches exceed the Lagrangian timescale. Since we had no Lagrangian mea- surements, we used the Eulerian decorrelation timescale of 18 days calculated from a current meter at 350 m over the upper slope off Pt. Sur (Collins et al. 1996b). In fact, the interval between launches was signi®cantly FIG. 3. (a) Deployment schedule for the 20 ¯oats used in this study. greater. Regular deployments began in July 1993 and The shading separates the ¯oat days into the three types of observed motion. Black are ¯oat days in the California Undercurrent, dark gray followed in September and November 1993, January, are ¯oat days in the interior region, and light gray are ¯oat days for April, May, and August 1994. There was a bias toward reversing marginal ¯ow. The white region in NPS6 represents period late summer to early winter as illustrated in Figs. 3b of unknown position. (b) Frequency distribution of the number of and 3c, where the white bars indicate the total ¯oat days ¯oat daily observations as a function of year and (c) month. White by year and month. bars in b and c indicate the total ¯oat days for all ¯oats.

d. Float performance Battery capacity determined the number of listening cy- Our ¯oats exhibited depth holding characteristics sim- cles per day. Short duration ¯oats were set to listen ilar to those seen by other RAFOS ¯oat users (Bower every 8 or 12 hours. The longer duration ¯oats listened 1994). The mean, median, and standard deviation of less frequently with mission lengths of 300 or more days daily interpolated pressure values are plotted in Fig. 4. listening every other day. The equilibrium level reached by four ¯oats deviated The choice of the ¯oat depth was a compromise be- substantially from the target pressure. NPS 2, 22, and tween the depth of the undercurrent core, about 100 m 33 were about 300 dbar too deep, while NPS5 was 200 (Rischmiller 1993), and the depth of the SOFAR channel dbar too shallow. axis, about 550 m (Johnson and Norris 1968). A depth Once on the equilibrium pressure level, three factors of 100 m falls within the , which we felt can cause the quasi-isobaric ¯oats to change that level: was too shallow for reliable acoustic listening. On the internal waves, water mass changes, and ¯oat mass other hand, a depth of 550 m would be too deep, pos- changes. Internal waves and internal have oscil- sibly below the undercurrent. Consequently, two prin- lation frequencies that are aliased in the data by the ¯oat cipal pressures were targeted, ®rst 350 dbar, then later, sampling schedule. The smaller expansion coef®cient

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC 564 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 29 for the ¯oat compared to that for water will tend to that ¯oats next to the coast moved parallel to the coast, reduce the oscillation amplitude. while those farther offshore moved away from the coast, Indications that a ¯oat sank during a mission are a primarily by transport in anticyclonic eddies. A more large standard deviation and a large difference between detailed examination shows that three general patterns the mean and median depths. NPS 9, 19, and 24 were are illustrated by the ¯oat trajectories: 1) steady pole- ``slow sinkers'' (Bower 1994), showing a constant slow ward ¯ow in the undercurrent, 2) reversing parallel ¯ow depth increase over the mission period. It is assumed that remained near the continental margin, and 3) west- that this sinking was caused by slow leaks in the hollow ward migration, usually in an eddy, in the ocean interior. ballast weight (a one gram change in ¯oat mass trans- These will be referred to as California Undercurrent, lates to approximately 40-m depth change). Four ¯oats marginal, and interior patterns. were ``fast sinkers,'' undergoing catastrophic sinking. Most ¯oats exhibited more than one of these ¯ow NPS20 and NPS23 sank immediately; NPS33 sank patterns (Fig. 3a). For example, NPS26 demonstrated quickly on day 293 of a planned 499-day mission. marginal reversing ¯ow for the ®rst 35 days, undercur- NPS19 initially acted as a slow sinker, then two days rent ¯ow for the next 50 days and interior ¯ow for the before the end of its scheduled 203-day mission it sank last 40 days of its subsurface mission. To examine the abruptly. Six ¯oats experienced sinking and four of them characteristics of each ¯ow type, the trajectories were terminated the subsurface mission when the ambient subjectively broken into segments corresponding to the pressure exceeded a preset allowable value. Other prob- ¯ow patterns. The general criterion used to distinguish lems were a malfunctioning pressure transducer on between marginal ¯ow and undercurrent ¯ow was to NPS14 (the temperature record suggests that it was at choose the period when there was continual poleward a depth of just less than 300 dbar) and intermittent pres- ¯ow and no subsequent southerly ¯ow. Interior ¯ow sure sensor failures on NPS21 and NPS24. was started when the ¯oat motion was no longer parallel For the other 13 ¯oats, the mean offset from the target to the local bathymetry or the ¯oat exhibited eddy mo- pressure was 18 dbar shallow with a standard deviation tion and the water depth exceeded 2000±3000 m. While of 30 dbar. These are the ¯oats in Fig. 4 that have nearly there was never much ambiguity about the three pat- coincident mean and median pressures and small stan- terns, the actual timing of the transition from one to dard deviations. another might be dif®cult to assign to a particular day. The observed depth changes and pressure±tempera- Thus, there may be an error of a few days in choosing ture behavior for the ¯oats that did not leak appear to the transition from one trajectory pattern to another. be real and are due to a ¯oat encountering waters of The mean ¯ow for the total ¯oat ensemble was di- different temperature and density along the route (Col- rected toward 310Њ with a vector average speed of 2.5 lins et al. 1996a). When isopycnals are inclined to the cm sϪ1 and a mean speed of 8.7 cm sϪ1 (Table 1). The isobaric surface, the buoyancy force on the ¯oat will principal axis of variance was oriented at 340Њ, with the change with changes in the water density and temper- ellipse of variance having axes of 7.5 and 6.6 cm sϪ1. ature (Swift and Riser 1994). The change in pressure Examination of the three ¯ow patterns reveals that the and the change in the thermal expansion coef®cients for segments along the continental margin had the variance water and glass affect the balance between the ¯oat and oriented with the major axis parallel to the continental the ambient pressure. When the density increases, the margin (330Њ). In the ocean interior, the variance was ¯oat will rise, and vice versa. Notwithstanding, the wa- much more isotropic: the variance axes were 7.7 and ter temperature record often ¯uctuates in phase with the 7.5 cm sϪ1. pressure record; a sinking ¯oat will have a temperature The remainder of this section will describe the ¯ow increase. That is because the equilibrium displacement patterns observed in the undercurrent, along the conti- of the ¯oat will be less than the isopycnal displacement nental margin and in the ocean interior. of the water causing the change in buoyancy. The tem- perature±pressure relationship for isobaric ¯oats can be a. California Undercurrent used to check whether or not the ¯oat density is chang- ing due to water mass change or to leakage. Normally The six ¯oats caught in the undercurrent had contin- the T±P scatter will be in a pattern moving orthogonal uous poleward ¯ow parallel to the coast for distances to the trend of the local T±P relationship. ranging from 290 to 480 km. The trajectories for these If the T±P scatter for a ¯oat trends with the vertical T±P ¯oat segments are shown in Fig. 6; Table 2 gives the trend of the water, the ¯oat is probably sinking rather dates, distance, mean depth, and speed of the six un- than adjusting to new conditions. dercurrent ¯oat segments. Four of the ¯oats, NPS 4, 5, 8, and 19 were imme- diately entrained in the undercurrent. NPS5 (this ¯oat 3. Flow patterns has been described in detail by Collins et al. 1996a) Nineteen ¯oats were successfully navigated for all or remained along the continental margin until near Pt. most of the subsurface mission. The ensemble spaghetti Saint George (41Њ50ЈN). Then it moved offshore and diagram for these ¯oats is given in Fig. 5. It is apparent became entrained in an eddy. During the undercurrent

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC APRIL 1999 GARFIELD ET AL. 565

FIG. 5. Ensemble spaghetti diagram showing the smoothed subsurface trajectories of all the NPS RAFOS ¯oats. The launch sites are indicated by circles, and the location where each ¯oat surfaced is indicated with an ``ϫ.'' The ¯oat number is placed at the surfacing location. Depth shading is keyed to isobaths at 200, 1000, 2000, and 4000 m. transit, the ¯oat generally remained above the 200-m Of the six triads and one pair of RAFOS ¯oat de- isobath and had a mean speed of 16.3 cm sϪ1. Along ployments, NPS 4 and 8 represent the only instance in the way, the ¯oat passed under the active upwelling which more than one ¯oat was entrained in the under- centers at Pt. Arena and Cape Mendocino. NPS5 rep- current at the time of launching. NPS8 initially traveled resents the shallowest ¯oat to date. It oscillated between faster until ϳ39Њ30ЈN, where it slowed down while 115 and 180 dbar with a mean depth of 135 dbar. Our NPS4 passed inshore of it. At Punta Gorda, the trajec- ability to track such a shallow ¯oat appears remarkable. tory of NPS4 curved offshore, passing over the Gorda

TABLE 1. Float velocity and speed statistics. Number U mean and std dev V mean and std dev Mean speed and std dev Type of ¯ow of ¯oats Pseudo¯oat (cm sϪ1) (cm sϪ1) (cm sϪ1) (a) determined from apportioning the ¯oats into the different types of observed ¯ow All 19 141 Ϫ1.7 Ϯ 1.8 0.9 Ϯ 2.6 8.7 Ϯ 4.1 undercurrent 6 14 Ϫ4.1 Ϯ 1.1 10.1 Ϯ 1.6 11.9 Ϯ 1.7 marginal 6 38 Ϫ0.02 Ϯ 1.5 Ϫ1.0 Ϯ 2.3 5.0 Ϯ 1.8 interior 13 75 Ϫ2.0 Ϯ 1.8 0.4 Ϯ 1.6 9.7 Ϯ 4.4 (b) (in a coordinate system rotated counterclockwise 32Њ to alongshore and across-shore axes) from apportioning the ¯oats into the different types of observed ¯ow

Ur mean and std dev Vr mean and std dev Mean speed and std dev All 19 141 Ϫ1.0 Ϯ 1.7 1.7 Ϯ 2.7 8.7 Ϯ 4.1 undercurrent 6 14 1.8 Ϯ 1.3 10.8 Ϯ 1.4 11.9 Ϯ 1.7 marginal 6 38 Ϫ0.6 Ϯ 0.5 Ϫ0.8 Ϯ 2.6 5.0 Ϯ 1.8 interior 13 75 Ϫ1.5 Ϯ 1.9 1.4 Ϯ 1.4 9.7 Ϯ 4.4

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC 566 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 29

FIG. 6. Ensemble and individual spaghetti plots for the California Undercurrent ¯ow regime. The undercurrent segment is drawn as a heavy line in the ensemble panel. In the individual

Escarpment and the Mendocino Ridge. NPS4 is the only Instead the ¯oat was entrained in an eddy that continued ¯oat to be entrained in the California Current jet ¯owing to translate north±northwest for the remainder of the offshore at Cape Mendocino. Seven days after NPS4, submerged mission. NPS8 passed northward around Cape Mendocino and NPS26 was launched alone ten days later, 22 August its trajectory continued more north-northwest, moving 1994, from a site between Pt. Reyes and Monterey Bay offshore rather than following the bathymetry to the (37Њ05ЈN, 123Њ08ЈW). The ¯oat moved equatorward north-northeast. West of Pt. St. George the ¯oat became along the continental margin for the ®rst 32 days, then entrained in an eddy that apparently was moving with swung inshore and began moving poleward in the un- the undercurrent. Just south of Cape Blanco (42Њ50ЈN), dercurrent at 35Њ10ЈN. West of Pt. Arena the ¯oat en- NPS8 left the eddy and began moving westward. The tered an eddy that broke away from the undercurrent. third ¯oat of this triad, NPS10, ®rst went southeast and The ¯oat was moving westward when it surfaced. This then offshore. ¯oat demonstrated ¯ow continuity from south of Pt. Sur NPS19, launched 25 April 1994, initially settled deep, north past Pt. Reyes. 420 dbar, but still was carried poleward in the under- The ¯oats transported in the undercurrent are con- current, remaining between the 200-m and 1000-m iso- centrated in the summer, or wind relaxation, period from bath for 50 days, until reaching ϳ41ЊN. There it started July to November, with the exception of NPS19, which gradually moving into deeper water (2000 m) while still was launched in April 1994. NPS5 was launched 7 July moving northward. At about 41.6ЊN the ¯oat stalled, 1993, while NPS4 and NPS8 were both launched 3 Sep- then began moving westward in an eddy. tember 1993. NPS 26 and 28 were launched 10 days Two ¯oats underwent marginal ¯ow before being en- and 60 km apart in August 1994. trained in the undercurrent. NPS28 remained near the The vector average ¯ow for these ¯oats was directed launch site for 25 days before becoming entrained. It toward 320Њ with onshore and alongshore ¯ows of 1.8 then was carried northward to west of Punta Gorda 40 Ϯ 1.3 and 10.8 Ϯ 1.4 cm sϪ1, respectively. The mean days later. There the ¯oat moved offshore and did not speed, 11.9 Ϯ 1.7 cm sϪ1, was nearly the same as that curve with the bathymetry north of Cape Mendocino. observed for interior ¯oats. The principle axis of vari-

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC APRIL 1999 GARFIELD ET AL. 567

FIG.6.(Continued) panels the daily positions are indicated by the red symbols. Isobaths for 200, 1000, 2000, and 4000 m are drawn in blue in the individual panels.

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC 568 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 29

TABLE 2. The transit dates, alongshore distance, and mean and standard deviation of the depth and speed for the six ¯oat segments within the undercurrent. Alongshore Days in distance Depth Speed Float Start date current (km) (dbar) (cm sϪ1) 4 5 Sep 1993 27 286 242 Ϯ 4 13.0 Ϯ 5.1 5 8 Jul 1993 34 435 135 Ϯ 14 16.3 Ϯ 5.2 8 5 Sep 1993 43 410 302 Ϯ 12 11.4 Ϯ 5.2 19 26 Apr 1994 55 454 430 Ϯ 7 10.1 Ϯ 4.4 26 28 Sep 1994 52 480 314 Ϯ 7 12.5 Ϯ 3.6 28 6 Sep 1994 50 457 357 Ϯ 9 11.6 Ϯ 4.8

ance was oriented along 336Њ, with 5.0 and 4.7 cm sϪ1 no. In contrast, entrainment into the undercurrent by the for the major and minor axes. The range of observed two ¯oats was not by eddy coalescence with the current. speeds was quite large. NPS5 had a range of daily mean speeds from 4.8 to 27.9 cm sϪ1. Note also that the shal- b. Marginal ¯ow lowest ¯oat (135 dbar) had the greatest speed, and the deepest ¯oat (420 dbar) the slowest. A linear regression The trajectories of the ¯oats that remained along the of mean ¯oat pressure (dbar) versus mean ¯oat speed continental margin and experienced reversing ¯ow with (cm sϪ1) for the six ¯oats gives, in this depth range, a little net movement are illustrated in Fig. 8. Although depth-related undercurrent speed of S(z) ϭϪ0.02 ´ z ϩ the depths of NPS2 and NPS33 were deep, between 600 18.4, where S is the speed (cm sϪ1) and z is the pressure and 700 dbar, the mean speeds of these two ¯oats were (dbar). The vertical shear, Ϫ0.02 cm sϪ1 dbarϪ1, agrees 2.5 and 4.6 cm sϪ1, respectively, with most of the move- well with velocity shear estimates of Ϫ0.016 cm sϪ1 ment parallel to the bathymetry. dbarϪ1 from the Pegasus acoustic dropsonde (Risch- The four other ¯oats, NPS3, 10, 26, and 28, were miller 1993) but appears to be less than geostrophic between 300 and 350 dbar and experienced extended shear estimated from the geostrophic velocity sections periods of reversing ¯ow. The mean speeds of these in Tisch et al. (1992). Collins et al. (1998, manuscript four ¯oats were in the range of 3.1 to 12.3 cm sϪ1. submitted to Deep-Sea Res., hereafter C98) investigate NPS3, on a 60-day mission starting 7 July 1993, ®rst the disparity between geostrophic shear and absolute moved northwest for 65 km, then onshore, stalled for shear in the undercurrent. 10 days and then moved 110 km slowly south for an- Collins et al. (1996a) noted that undercurrent models other 20 days. The ¯oat then reversed direction, moved predict that ¯oat speeds should show a maximum as- north, and surfaced 24 km from the launch point. sociated with the coastal promontories. To test this, ¯oat NPS10, launched 3 September 1993, did a small anti- speed and acceleration were correlated against a number cyclonic loop (diameter 20 km) around the launch site of parameters to examine the speed ¯uctuations. No during the ®rst 20 days, stalled, then moved south 60 signi®cant correlations were found for the ¯oat motion km, stalled again, and then moved into the interior. It with distance offshore, width or divergence of the shelf spent 50 days along the margin. NPS26 was launched or slope, and curvature of the coast or shelf margin. A south of the usual launch site and initially moved cy- slight trend of decreasing speed as a function of distance clonically offshore, then southeast to 35Њ12ЈN. It then offshore was the only possible correlation, but it was curved anticyclonically, entrained in the undercurrent not statistically meaningful. off Cape San Martin (35Њ50ЈN). NPS28, launched off Float speed in the undercurrent did not show any San Francisco, remained in the local area with little consistent pattern along the coast (Fig. 7) except for the movement for 24 days before becoming entrained in the acceleration north of Point Reyes followed by the de- undercurrent. celeration south of Pt. Arena. In all cases, the maximum The ¯oats that encountered reversing marginal ¯ow speed was reached south of Pt. Arena, and the ¯oats were launched between year days 190 and 250 (early were decelerating, or had reached a minimum speed, at July to early September), which is representative of the Pt. Arena. There does not appear to be a preferred mo- summer season of reduced upwelling and more variable tion at either Pt. Reyes or Cape Mendocino. The lowest winds. There is probably no statistical signi®cance to mean speed was at Pt. Reyes, and the highest at Cape this, since most ¯oat deployments occurred during the Mendocino, but the latter was in¯uenced by the high same period. The point to be made, however, is that speed of NPS5. none of the ¯oats with marginal reversing ¯ow were Five of the six undercurrent segments ended because launched during the winter period, the season ascribed the ¯oats became entrained in anticyclonic eddies. The to having reversing ¯ow (Chelton 1984). sixth ended when NPS4 became entrained in the off- By de®nition, the vector mean speed for marginal shore-¯owing California Current jet at Cape Mendoci- ¯oats was low (not signi®cantly different from zero)

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC APRIL 1999 GARFIELD ET AL. 569

FIG. 7. Plot of ¯oat speed (cm sϪ1) vs latitude for the NPS RAFOS ¯oats caught in the under- current. The thin solid lines show individual ¯oat daily speeds with ¯oat number is given at the start latitude. The thick dashed line is the mean speed. The asterisks give the number of ¯oats used to determine each mean. The major coastal promontories are labeled at the top of the ®gure and indicated by open circles at the top and bottom of the ®gure.

and the mean speed of 5.0 cm sϪ1 was only about half both entered anticyclonic eddies upon leaving the un- that of the other two ¯oat classes. The variance ellipse dercurrent and then apparently left the eddies. NPS8 was oriented along 326Њ, with 5.1 and 4.4 cm sϪ1 for moved westward, while NPS19 took a large cyclonic the axes. curve (130-km diameter) to the north for a period of The ¯oats undergoing marginal ¯ow were just west approximately 25 days followed by 20 days of straight of the ¯oats transported in the undercurrent. There ap- trajectory to the west-northwest. This was the only ex- pears to be a region directly adjacent to the undercurrent ample of sustained cyclonic ¯ow in any of the interior where the ¯ow was polarized alongshore, but with di- ¯oats. rection and strength variable. This ¯ow also appeared The mean speed computed for the interior ¯oat seg- more barotropic with a small vertical velocity gradient. ments was 9.7 Ϯ 4.1 cm sϪ1, while the mean u and ␷ velocities were Ϫ2.0 Ϯ 1.8 and 0.4 Ϯ 1.6 cm sϪ1, re- spectively. The mean ¯ow direction was 281Њ, and the c. Interior ¯ow vector average speed was 2 cm sϪ1. The direction was Figure 9 illustrates the interior eastern Paci®c trajec- primarily due to propagation of the eddy ®eld, which tories for the 14 ¯oats that were transported away from is discussed in section 5. Variability was nearly isotro- the continental margin. Seven of the ¯oats moved west- pic, with the axes being 7.7 and 7.4 cm sϪ1 in length. ward almost immediately after launch, never encoun- The predominance of the eddy motion and the distance tering the undercurrent. One ¯oat exhibiting marginal from the ocean boundary both contribute to the structure ¯ow, NPS10, eventually entered the interior. The other of the isotropic variability. ®ve ¯oats were ®rst entrained in the undercurrent. There did not appear to be a pattern to where or when these d. Other ¯oats ¯oats left the undercurrent. As noted previously, most eddy motion was anticy- Four other ¯oat deployments provide further evidence clonic; only one ¯oat (NPS9) never encountered an of the barotropic nature of the marginal ¯ow. Three short eddy. Ten of the 14 ¯oats exhibiting interior ¯ow sur- duration ¯oats (NPS 22, 24, and 30) were launched west faced while still entrained in an eddy. NPS 8 and 19 of Monterey Bay (centered about 36Њ26ЈN, 122Њ55ЈW)

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC 570 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 29

FIG. 8. As in Fig. 6 but for the continental margin ¯ow segment. on 19 May 1994 as part of a coastal tomography ex- warm water deformed anticyclonically. NPS7 was in a periment (Benson 1995). The ¯oats, spread between the warm anticyclonic feature west of San Francisco. depths of 300 and 600 dbar, had remarkably similar The ¯oats in triad NPS 4, 8, and 10 do not appear to motion, moving northwest (parallel to the coast), with have been in eddies that had surface expressions in the speeds between 7.9 and 10.2 cm sϪ1. Two months earlier temperature ®eld (Fig. 11). NPS4 was moving offshore, in March 1994, a modi®ed RAFOS ¯oat launched at along the north side of the Cape Mendocino jet. On the 37Њ23ЈN, 122Њ41ЈW, operated between 720 and 840 dbar day of the image, 15 November 1993, the ¯oat was in (Sanford et al. 1995). Like the tomography ¯oats, it a small cyclonic feature that is not apparent in the image. moved northwest with an average speed of 3.3 cm sϪ1. NPS8 had just completed anticyclonic motion in the area south of Cape Blanco. There is no indication of this e. Float trajectories and feature in the SST data. NPS10 was traveling slowly west-northwestward on that day and was located south Float trajectories for the ®rst two ¯oat triads were of another cold jet. The SST data suggest that there may superimposed on satellite AVHRR imagery to examine have been anticyclonic motion at the surface. whether there were surface temperature expressions of the motion experienced by the ¯oats. Lynn and Simpson (1987) noted that the California Current regime was 4. Pressure and temperature patterns dominated by anticyclonic eddies; therefore one might Spectral analyses of the temperature and pressure re- expect to see correspondence between the surface tem- cords were not pursued because of the severe temporal perature structure and the ¯oat subsurface motion. The aliasing. The three special tomography ¯oats that sam- ®rst triad, NPS 3, 5, and 7, are shown plotted on the pled temperature and pressure hourly showed mean SST image for 27 August 1993 (Fig. 10). The position spectral peaks at 12.6 h for both temperature and pres- of each ¯oat for the image day is indicated by an as- sure (Benson 1995). These hourly ®xes may have ali- terisk. On that day NPS3 was at the southern extreme ased the data since the Brunt±VaÈisaÈlaÈ frequency is about of its trajectory, just starting the return to the north. 30 min for the depths of interest. Due to their smaller NPS5 was in an anticyclonic eddy just south of Cape compressibility, RAFOS ¯oats oscillate faster than the Blanco, where there is a surface signature of cold and local water column. While not usable for spectral anal-

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC APRIL 1999 GARFIELD ET AL. 571 ) Continued .8.( IG F

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC 572 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 29

FIG. 9. As in Fig. 6 but for the interior Paci®c ¯ow segment. yses, the daily temperature and pressure records were showed any indication of depth change or water mass used for trend recognition and to identify speci®c events. change at the major promontories. The CUC-segment temperature records all showed a Lynn et al. (1982) used California Cooperative Fish- northward temperature decrease, while there were mixed eries Investigations (CalCOFI) hydrographic data to results in the pressure records (Fig. 12). The two shal- demonstrate the alongshore northward decrease in tem- lowest ¯oats, NPS 4 and 5, sank while traveling north- perature and increase in density at depths above 500 m. ward in the undercurrent but the three ¯oats between Their data show that the alongshore temperature gra- 300 and 360 dbar experienced a general rise in depth dient is largest at 300 m, decreasing toward both the during the undercurrent transit. The deepest undercur- surface and depth. The general rising of the ¯oats in rent ¯oat, NPS19, at 430 dbar, underwent a minimal this depth range may be a response to the larger tem- sinking of 0.02 dbar/day, or 1.2 dbar during the 58 days perature gradient. while it was in the undercurrent. NPS19 was identi®ed The two ¯oats that appear to have left eddies before as a slow sinker due to a leak. Taking this into account, the mission end, NPS 7 and 8, both rose and measured it appears that the ¯oats shallower than 250 dbar un- a temperature decrease as they ended anticyclonic mo- derwent a general depth increase, while those deeper than that pressure experienced a general depth decrease tion. This most probably represents the change of water while going northward in the undercurrent. mass and decreased temperatures upon expulsion from South of Pt. Arena NPS5 slowed down, moved in- the eddies. shore to near the 200-m isobath, rose 20 dbar, and reg- The pressure±temperature response of the marginal istered a 0.2ЊC temperature drop. The ¯oat may have reversing ¯ow and of the interior eddy segments did not been in¯uenced by the deep portion of the upwelling reveal any consistent patterns. Floats in eddies north of plume observed in the satellite imagery. However, at Cape Mendocino tended either to have no pressure or the stronger Cape Mendocino upwelling site the ¯oat temperature trends, or to cool slightly. The two ¯oats, accelerated and sank slightly. NPS4 showed no response NPS 4 and 31, in large eddies had no gradients of pres- at Pt. Arena, but rose in cooler water as it left the un- sure or temperature. The other ®ve eddy segments south dercurrent and moved offshore. None of the other ¯oats of Cape Mendocino revealed an even mix of rising±

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC APRIL 1999 GARFIELD ET AL. 573 ) Continued .9.( IG F

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC 574 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 29

FIG. 10. Satellite imagery for 27 August 1993, with ¯oat trajectories for NPS 3, 5, and 7 superimposed. The dots are the daily interpolat- ed ¯oat positions The ¯oat position on the day of the satellite overpass is marked by a large black dot. cooling, sinking±warming, and no pressure or temper- eraged eddy radius increased to a mean of 34.8 Ϯ 21.5 ature gradients. All gradients were small. km with an increased rotational period of 23.2 Ϯ 7.4 days. 5. Eddy characteristics NPS31 was the only ¯oat clearly in a mesoscale-sized eddy. It had a ¯oat motion radius of 71 km and a mean Sanderson's (1995) least squares technique was ap- period of 27 days. At the time it surfaced, NPS4 also plied to estimate the eddy kinematics from the smoothed appeared to be in a mesoscale eddy with radius 67 km ¯oat segments. The technique allows estimation of both and period 35 days. All the other ¯oat eddy motion rotational and translational eddy parameters. Float data occurred with radii less than 35 km, which is less than from 13 ¯oats were used to analyze 12 eddies (Table the internal Rossby radius, and generally exhibited 3). All the eddies were anticyclonic with negative vor- shorter rotational periods. ticity in the range Ϫ1.36 to Ϫ0.41 (ϫ10Ϫ5 sϪ1). Figure Eddy formation apparently occurred within the un- 13 shows the mean translation vector of each eddy. Ad- dercurrent ¯ow region. Neither of the ¯oats entrained jacent to the undercurrent, eddy motion was 3.0 cm sϪ1 in a direction slightly offshore from the poleward bathy- in the undercurrent were in eddies at the time of en- metric trend. The average radius was 32.8 Ϯ 10.2 km trainment. NPS 5, 8, 26, and 28 all demonstrated cy- with a mean rotational period of 16.5 Ϯ 5.1 days. Since cloidal motion while still moving northward with the there is no way to know the distance from the ¯oat to undercurrent. Once developed, the eddy trajectories the edge of the eddy, the radii obtained from these ¯oat showed a general westward movement. NPS 6, 11, and data represent minimums. For eddies away from the 13 illustrated either one eddy splitting or two eddies continental margin, mean eddy translation was west- colliding. Although all three ¯oats began in the same ward with a mean speed of 1.2 cm sϪ1. The ¯oat-av- eddy, near the end of the ¯oat missions, one ¯oat

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC APRIL 1999 GARFIELD ET AL. 575

FIG. 11. Satellite imagery for 15 November 1993, with ¯oat trajectories for NPS 4, 8, and 10 superimposed. The dots are the daily interpolated ¯oat positions The ¯oat position on the day of the satellite overpass is marked by a large black dot.

(NPS11) appeared to enter an adjacent eddy while the width on the order of 100 km. This distance coincides other two remained in the original feature. with the location of the dynamic topography trough de- The other ¯oats either entered eddies immediately termined from the CalCOFI hydrographic data (Schwing after being launched (NPS 11, 13, 31) or after leaving et al. 1994). the undercurrent and drifting for some time (NPS 6, 7, The embedded undercurrent is a distinct, continuous 10, 14, and 19). Once entrained, most ¯oats remained feature over the distance sampled, 35Њ±43ЊN. The un- in eddies for the duration of the submerged drift. The dercurrent has a sharp offshore boundary, and the un- exceptions were ¯oats NPS 7, 8, and 19, which appear dercurrent ¯ow is stronger and has a larger vertical bar- to have been ejected from their respective eddies prior oclinic shear than the surrounding marginal ¯ow. The to mission end. vertical current shear between 150 and 600 dbar is Ϫ0.02 cm sϪ1 dbarϪ1. The other characteristic of the 6. Discussion ¯ow is the development of small eddies that initially move with the ¯ow before breaking away and moving a. Undercurrent into the ocean interior. Within the present dataset, eddy The undercurrent and marginal ¯ow segments suggest formation appears to occur all along the sampled region. that the subsurface ¯ow along the northeast Paci®c mar- Some ¯oats were launched into eddies, others reveal gin can be characterized by a narrow jet, usually located eddy development while moving northward. right along the continental slope embedded in a larger Annual variability cannot be investigated with these region of variable ¯ow that moves parallel to the con- data. However, the present measurements showing a tinental margin. The width of the larger ¯ow regime is pronounced undercurrent cover the fall and winter pe- not known, but the ¯oat data and the analysis of closely riods, the time that Chelton (1984), using geostrophic spaced Pegasus data (Rischmiller 1993; C98) suggest a computations, found little evidence of the undercurrent.

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC 576 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 29

Hickey (1979) suggested a northward weakening of the undercurrent. The ¯oats experienced a general north- ward temperature decrease, but our results do not sup- port an alongshore speed gradient. Around Cape Men- docino the ¯ow became more variable, but the ¯oats that remained in the undercurrent continued with speeds similar to those farther south. The temperature decrease probably results from the gradual exchange of water in the undercurrent. Cooler water appears to be continually entrained, while the generally warmer undercurrent wa- ter is removed through eddy formation and break away. Off California, surface water is transported offshore in major ®laments, or jets, that appear to be anchored at coastal upwelling centers (Brink and Cowles 1991), and in small ®laments, or squirts (Ramp et al. 1991), as well as in eddies (Davis 1985; SSC). In contrast, offshore transport of subsurface undercurrent water ap- pears to occur primarily through the formation of small subsurface anticyclonic eddies: only one ¯oat moved offshore associated with the Cape Mendocino jet, while four others continued northward. As these ¯oats moved offshore, there was no distinct T/P ¯uctuation indicating a transition from equatorial to subarctic water. It has not yet been ascertained whether these undercurrent eddies form at ®xed geographical locations or more or less randomly along the coast. South of Point Reyes, in the Gulf of the Farallones, FIG. 12. Pressure (a) and temperature (b) plotted against latitude there have been a number of cruises that obtained ship- for the six ¯oat segments in the California Undercurrent. Float iden- board acoustic Doppler current pro®ler (ADCP) mea- ti®cation numbers are printed at the beginning and end of each seg- surements that included the continental margin (Ramp ment. The major coastal promontories indicated in Fig. 7 are again identi®ed by open circles on the latitude axes and labeled with initials. et al. 1995; Parker 1996; Baltz 1997; SSC). During 5 of the 14 surveys, strong anticyclonic subsurface eddies with equatorial water cores were found along the con- Recent analyses of hydrographic data (Tisch et al. 1992), tinental margin. Unfortunately, most of the surveys did Pegasus data (C98), and the results of this study suggest not extend far enough offshore to completely map the that there are signi®cant currents at depth along the undercurrent ¯ow. During 1993 there were three surveys continental margin. Perhaps the traditional use of large spaced a week apart that demonstrated that the eddies hydrographic station spacing and the 500-m reference are not stationary. The survey area was not large enough level for dynamic topography previously obscured the to tag and track each eddy, but it appears that the eddies existence of the current. were moving poleward in the undercurrent.

TABLE 3. Eddy kinematics for strongly rotational ¯oat segments in both the undercurrent and the ocean interior: %U, %V are the percentages of the U and V velocity components, respectively, accounted for by the combination of rotation and eddy translation.

u, ␷ Latitude Longitude Radius Period Vorticity Float Start date (ЊN) (ЊW) (cm sϪ1) (km) (days) (sϪ1 ϫ 105)%U %V 4 4 Dec 1993 39.1 130.3 Ϫ2.15 Ϫ0.19 69.8 35.7 Ϫ0.41 99.4 98.9 5 12 Aug 1993 42.2 125.3 Ϫ0.83 0.58 17.4 10.8 Ϫ1.36 90.1 88.9 6, 11, 13 3 Dec 1993 37.8 124.6 Ϫ0.86 Ϫ0.72 25.8 24.5 Ϫ0.61 83.6 72.8 11 10 Jan 1994 37.5 125.5 Ϫ0.98 Ϫ0.87 18.0 17.7 Ϫ0.84 96.9 97.4 7 27 Jul 1993 37.2 125.4 Ϫ2.40 Ϫ0.04 23.1 11.9 Ϫ1.23 92.6 93.4 8 19 Oct 1993 41.6 125.0 Ϫ0.61 3.05 29.5 21.0 Ϫ0.71 94.9 92.5 10 16 Dec 1993 38.2 125.7 Ϫ1.23 2.63 17.1 22.8 Ϫ0.64 99.3 99.3 14 30 Jan 1994 38.2 124.6 Ϫ0.99 0.47 26.5 24.2 Ϫ0.61 97.5 98.5 19 7 Jun 1994 41.6 126.4 Ϫ2.25 Ϫ0.68 26.2 16.6 Ϫ0.90 92.9 75.6 26 14 Nov 1994 39.0 124.3 Ϫ4.71 1.21 27.9 13.7 Ϫ1.07 84.0 83.1 28 23 Nov 1994 41.1 125.3 Ϫ2.49 3.39 33.8 15.6 Ϫ0.93 94.9 93.8 31 31 Aug 1994 38.7 126.3 Ϫ1.50 Ϫ1.53 73.2 27.3 Ϫ0.55 99.0 98.5

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC APRIL 1999 GARFIELD ET AL. 577

FIG. 13. The mean trajectory of each eddy sampled by the NPS RAFOS ¯oats. The number at the base of each vector is the ®rst ¯oat to encounter the eddy as listed in Table 3. The origin of each vector is the location where the eddy was ®rst encountered.

The moored current meter record from 350 m off Pt. not been observed, a diapycnal mixing process was the Sur (Collins et al. 1996b) shows a number of instances most plausible mechanism to consistently produce the of rotary ¯ow reversals with periods on the order of a anticyclonic SCVs. D'Asaro (1988) proposed another week or less (their Fig. 4). The strength and period of generation mechanism for SCVs found in the Beaufort these anticyclonic reversals suggest that they may be gyre of the Arctic Ocean. He suggested that anticyclonic eddies moving with the undercurrent. eddies are generated in trapped boundary currents when the frictional torque along the continental margin (in- shore side of the ¯ow) reduced the potential vorticity b. Eddies and ¯ow variability to zero. Irregular topography and the frictional effects The small size and anticyclonic rotational bias sug- appear to combine for eddy formation. gest that undercurrent-spawned eddies are another ex- The best recognized SCVs are the meddies (Medi- ample of a submesoscale coherent vortex (SCV) for- terranean eddies) found in the North Atlantic. Several mation (McWilliams 1985). SCVs are small monopoles investigators have studied the formation of Mediterra- typi®ed by horizontal scales less than the ®rst baroclinic nean Water eddies in the North Atlantic (McDowell and radius of deformation, an aspect ratio (vertical/horizon- Rossby 1978; Armi et al. 1989; Richardson et al. 1989; tal) of the same order as f/N (the ratio of the Coriolis Zenk et al. 1992; Pingree and Le Cann 1993; Bower et parameter to the Brunt±VaÈisaÈlaÈ frequency) (McWilliams al. 1997). These eddies provide a mechanism for trans- et al. 1983), anticyclonic motion with speeds sometimes porting Mediterranean Water far into the Atlantic in- exceeding geostrophic balance, and a relative minimum terior (McWilliams 1985). Meddies (McDowell and in the vertical density gradient. These eddies appear to Rossby) and smeddies (shallow meddies) (Pingree and have long lives and generally move with the mean re- Le Cann) appear to develop in a region of strong an- gional ¯ow, providing a mechanism for transporting wa- ticyclonic curvature of the bathymetry (viewed in the ter properties long distances from the source. Mc- direction of ¯oat motion). Williams proposed that, although eddy generation had Bower et al. (1995, 1997) suggested that frictional

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC 578 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 29 torque was responsible for forming Atlantic meddies at eddies had radii of 40 and 45 km, the size of the internal speci®c locations as the Mediterrean out¯ow moves Rossby radius, with estimated rotational periods of 9 along the continental slope. Their results suggest that and 14 days. Both appeared to be in solid body rotation. SCV formation occurs where the topography forms a In addition to the prominent eddies, Huyer et al. mea- subsurface promontory. They hypothesize that Cape St. sured a number of other eddies, one being the same Vincent and Estremadura Promontory are the two prin- eddy transporting NPS7. This latter eddy did not have cipal sites for meddy formation. the strong spiciness signature but was still comprised Three patterns of eddy motion were observed in the of water saltier and warmer than the surrounding water. northeastern Paci®c, although two of the patterns were There were a number of closely spaced eddies in the each observed by only single ¯oats. NPS4 demonstrated vicinity of this eddy. The dimensions of the eddies were instability along the edge of an offshore ¯owing ®lament similar to the radii of the ¯oat motions. after it left the undercurrent at Cape Mendocino and SCV formation is seen both as an eddy transported moved offshore in a relatively smooth trajectory. Com- with the undercurrent (NPS8 and NPS28) and as a fea- parison with the satellite AVHRR data for the time sug- ture encountered after the ¯ow has separated from the gests that this ¯ow coincides with the surface cold water continental margin (NPS5 and NPS19). The present data jet streaming offshore from Cape Mendocino. While combined with ADCP data (Ramp et al. 1995; Parker moving offshore, the ¯oat went through two short me- 1996; Baltz 1997) suggest that frictional torque is the anders, the latter of which was somewhat larger than dominant forcing mechanism for SCVs formed in the the former. This ¯oat apparently ended its mission in a undercurrent. Cycloidal motion during eddy formation mesoscale eddy. Clear satellite imagery was not avail- within the undercurrent might explain why some data able to determine surface conditions at the time of sur- suggest that the undercurrent comprises a series of ed- facing. dies moving poleward rather than of a single coherent The second type of motion was associated with me- current. The present data strongly suggest that the eddy soscale eddies. NPS31 became entrained in mesoscale features are created in and transported by the current. anticyclonic motion eight days after launch and re- In adding to the ``eddy zoo'' of eastern boundary formed mained in the large eddy for the remainder of its mis- eddies, we suggest that the undercurrent SCVs be re- sion. This was the only clear example of a ¯oat entrained ferred to as ``cuddies'' (California Undercurrent eddies). in a mesoscale eddy. Satellite imagery reveals that the Once formed, cuddy translation is markedly different eddy had a surface temperature expression, and a RA- than the surface ¯ow. The mixed layer responds to the FOS ¯oat at 1580 dbar suggests that the eddy penetrated local wind, and surface drifters demonstrate a generally to at least that depth. The scale of this eddy is similar meandering, but southerly, drift (Brink et al. 1991; to scales seen with surface ¯oats (Brink et al. 1991; Swenson and Niiler 1996). The small subsurface anti- Swenson and Niiler 1996). cyclonic eddies show a complete decoupling from the Ten ¯oats, exhibiting anticyclonic motion with a ro- surface. All cuddies, once detached from the undercur- tational radius in the range of 15±33 km, appear to be rent, moved primarily westward. The ¯oat trajectories other examples of submesoscale coherent vortices. The show these eddies being present to at least 131ЊW, and vorticity of these SCVs is smaller than the value for none of the ¯oats present evidence of eddy decay. Using typical meddys, but within the range of other SCVs the observed westward speed of 1.5 cm sϪ1, an eddy (D'Asaro et al. 1994). The surrounding vorticity ap- could translate about 240 km in six months. Some of proaches zero giving a vorticity gradient suf®ciently the long-duration ¯oats presently deployed may provide strong to inhibit water exchange. more insight on the eddy life spans. Huyer et al. (1998, manuscript submitted to Deep- Cuddy westward migration and long duration may Sea Res.) conducted two large, high-resolution upper- explain the offshore subsurface variability in water ocean surveys of the California Current region, one in types. Temperature±salinity scatterplots of grouped June and the other in August of 1993. Hydrographic CTD data often reveal increased variability between the parameters were measured in the upper 300 m while the 26 and 27 kg mϪ3 density anomaly surfaces (Tisch et ship was underway through the use of an undulating al. 1992). In the California Current region off central Seasoar vehicle. Velocity was measured simultaneously California, this corresponds to a depth range of about using a ship-mounted acoustic Doppler current pro®ler. 150±600 m. The variability may exist because the un- The presence of anticyclonic subsurface eddies were dercurrent anticyclonic eddies allow the equatorial Pa- prominent features in each cruise. The ``spiciness'' (Fla- ci®c water signature to remain intact for the lifetime of ment 1986) was used to show that the eddies contained the eddies. This provides a mechanism for delivering warm salty water and probably originated in the Cali- the equatorial Paci®c water offshore and could explain fornia Undercurrent at the latitudes of the observations. why the T±S variability remains strong and why this Huyer et al. estimated that the eddies had formed along water is often found embedded in southerly ¯owing Cal- the continental margin within the preceding 6±8 months. ifornia Current water. The observed eddies were subsurface features with The occurrence of undercurrent eddies transporting spiciness maximums at 150 m. The two most prominent equatorial Paci®c water into the eastern Paci®c interior

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC APRIL 1999 GARFIELD ET AL. 579 in a manner similar to meddy transport in the Atlantic ¯ow, but could be the result of the existence of the suggests that this is an important mechanism wherever boundary, which inhibits onshore ¯ow. midlatitude eastern boundary currents exist. Lagrangian Results from single- and two-particle statistical anal- studies in the south Atlantic, the south Paci®c, and In- yses are presented. The small size of the dataset and the dian will very likely encounter similar anticy- range in pressure mean that these estimates have large clonic subsurface eddies associated with eastern bound- uncertainties. These preliminary results are described ary currents. here in order to allow comparison with other published results. 7. Conclusions Diffusivity and dispersion The RAFOS dataset reveals two distinct features of the undercurrent ¯ow: the ¯ow continuity over a four- For purposes of computing both Eulerian and La- degree latitudinal range, and the distinct differences be- grangian statistics, NPS22 was not used because the tween surface and 300-m ¯ow in the interior eastern record length was less than 20 days. The number of Paci®c. The observed undercurrent ¯ow continuity was degrees of freedom for statistical analyses was increased anticipated, and the present ¯oats provide new data for by creating ``pseudo'' ¯oats. Every 10 days, the position examining the ¯ow speed along a curving continental of a ¯oat was used as the starting point for a ``new'' margin. The measured speeds do not conform to existing pseudo¯oat that then traveled as the parent ¯oat. In this theoretical models of subsurface ¯ow. Neither the pres- manner, a ¯oat with a 95-day record, for example, would ence of SCV eddies nor that of a mean westward interior be used to create a set of seven ¯oats, the original ¯oat ¯ow were anticipated at the start of this program. To record of 95 days plus six pseudo¯oats of 85, 75, ..., further verify the presence of SCV eddies, the historical 35 days. Minimum record length for the pseudo¯oats data need to be examined for the existence of distinct was 20 days. In this way the number of ``¯oats'' avail- subsurface signatures of equatorial water embedded in able for statistical analyses was increased from 19 to the subarctic waters of the offshore California Current. 141. Each ¯oat was also separated into groups corre- sponding to the three patterns of ¯ow with the statistics Acknowledgments. We thank Tarry Rago, who has then determined, again creating pseudo¯oats by restart- been directly involved with all aspects of this program. ing each ¯oat at 10-day intervals. Thus, the sum of the Kirk Kingsbury, NCEL pressure facility director and pseudo¯oats for the segments does not total the 141 operator, repeatedly went out of his way to ensure that pseudo¯oats for the entire record. The statistics were ballasting went smoothly and successfully. Mike Cook also computed with the coordinate system rotated 32Њ is thanked for his MATLAB tutoring. N.G. and C.A.C. counterclockwise, parallel to the coast so that velocity thank Jeff Paduan for many bene®cial discussions. components representing onshore (ur) and alongshore Frank Schwing ®rst suggested the name ``cuddy.'' The (␷ r) ¯ow could also be determined. Estimated ¯oat sta- suggestions of the two reviewers substantially improved tistics for the entire ¯oat ensemble and for the three this manuscript. We acknowledge support from the Na- patterns of ¯oat motion include mean speed, velocity, val Postgraduate School, the Oceanographer of the and velocity variability as well as the Lagrangian ve- Navy, and the Of®ce of Naval Research. locity correlation scales. Our data cannot resolve interannual or annual vari- ability, but these statistics can describe the character of APPENDIX the observed patterns. Float Statistics 1) ARRAY BIAS The mean currents in the interior and the principle axis of variance found from the RAFOS ¯oats are dif- Our project design of deploying the ¯oats near the ferent from results of surface drifters from this region. shore complicates estimating both Eulerian and La- Swenson and Niiler (1996) summarize most surface La- grangian statistics. Swenson and Niiler (1996) discuss grangian measurements obtained between 1985 and biasing problems present in determining Eulerian sta- 1990. In their most northerly box (40ЊN, 125ЊW) they tistics when, as in the present study, the deployment show (their Fig. 7) a mean south-southwest ¯ow with distribution is concentrated into one region. Floats orig- a vector speed of about 18 cm sϪ1. They also report inating from the eastern margin reach the interior from large asymmetry to the variance axes. The principle axis only one direction. There is no contribution of ¯oats is oriented at 40Њ. That contrasts with the subsurface arriving from the west. Spatial variability in the small- results found here, a mean ¯ow directed to 310Њ and a scale, eddy dispersive ®eld will bias Eulerian averages, vector mean speed of 2.8 cm sϪ1. One variance axis was and statistical reliability will be reduced in regions of oriented at 340Њ, with little asymmetry to the variance high velocity since ¯oats do not remain in those areas. magnitude. Davis (1985) noted that the offshore-di- Floats constrained to a depth, whether surface or sub- rected mean is not necessarily an indication of divergent surface, do not sample convergence effects. In addition

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC 580 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 29

TABLE A1. Estimation of the Lagrangian time, space, and diffusion scales. Timescale Length scale Diffusivity Number

of Pseudo U V U V Kuu K␷␷ Type ¯oats ¯oats (days) (days) (km) (km) (m2 sϪ1) (m2 sϪ1) (a) from the Lagrangian autocorrelation function All ¯ts 19 141 4.4 Ϯ 1.7 4.6 Ϯ 1.8 20.9 Ϯ 12.6 23.2 Ϯ 13.2 1369 Ϯ 1623 1505 Ϯ 1625 undercurrent 6 14 2.7 Ϯ 0.4 2.9 Ϯ 0.7 9.5 Ϯ 3.4 11.4 Ϯ 3.4 415 Ϯ 257 480 Ϯ 165 interior 13 75 4.4 Ϯ 1.8 4.0 Ϯ 1.3 25.5 Ϯ 15.0 23.1 Ϯ 15.1 1965 Ϯ 2007 1831 Ϯ 2028 marginal 6 38 5.0 Ϯ 1.2 5.6 Ϯ 2.2 12.6 Ϯ 3.6 18.4 Ϯ 9.1 367 Ϯ 280 629 Ϯ 360 (b) from the Lagrangian autocorrelation function with the coordinate system rotated 32Њ counterclockwise UV U V All ¯ts 19 141 4.4 Ϯ 1.7 4.6 Ϯ 2.0 21.2 Ϯ 15.4 23.1 Ϯ 13.0 1421 Ϯ 1730 1476 Ϯ 1590 undercurrent 6 14 2.7 Ϯ 0.5 2.7 Ϯ 0.6 9.6 Ϯ 3.3 10.9 Ϯ 3.2 409 Ϯ 263 482 Ϯ 181 interior 13 75 4.4 Ϯ 1.8 4.0 Ϯ 1.5 25.8 Ϯ 15.7 23.1 Ϯ 14.6 2016 Ϯ 2101 1789 Ϯ 1990 marginal 6 38 4.8 Ϯ 1.2 5.8 Ϯ 2.3 12.3 Ϯ 3.7 19.0 Ϯ 9.9 358 Ϯ 233 654 Ϯ 411 to those problems, our RAFOS array did not sample at A positive j is dominated by cyclonic rotation, while a a single depth; ¯oat depths ranged from 135 to 650 dbar. negative j indicates domination by anticyclonic rotation. Vertical shear is not estimated, but is treated as hori- In the mean the ¯ow is anticyclonic, most strongly so zontal shear. The noted depth dependence of the un- in the interior ¯ow where the motion is dominated by dercurrent speed emphasizes this last problem. Finally, the eddy ®eld. the small size of the present dataset does not allow Poulain and Niiler (1989) de®ne the Lagrangian in- robust statistical con®dence. Despite all these problems tegral time T L and space LL scales as the time and dis- and the large uncertainty to the computed values, the tance over which a drifter remembers its path comparison with other published values suggests that t the results are indicative of the subsurface diffusion. 1 T L ϵ R (␶) d␶ (A3) i 2 ii ͗uЈ͘iL͵ 0 2) SINGLE PARTICLE STATISTICS and

Single particle statistical methods (Taylor 1921) have t been extended for oceanographic application. We follow 1 LL ϵ R (␶) d␶ ϭ [͗uЈ͘21/2L] T , (A4) the formulations that appear in the recent work of, i [͗uЈ͘21/2] ͵ ii i L i i L among others, Poulain and Niiler (1989), Thomson et 0 al. (1990), Brink et al. (1991), Paduan and Niiler (1993), where ͗͘L represents the Lagrangian average. The au- and Poulain et al. (1996). Since a large number of in- tocorrelation functions are time dependent, and the usual dependent observations are necessary to recover reliable practice for estimating the time and space scales is to statistics, the practice of creating a larger set of ¯oats integrate the autocorrelation function from zero to the by starting pseudo ¯oats from each ¯oat after a set time time of the ®rst zero crossing. Estimates of the Lagrang- was used. ian space and Eulerian time scales based on the La- Decorrelation time and space scales, ¯ow polariza- grangian autocorrelation functions for these ¯oat tra- tion, and the rate of mixing, can be estimated from the jectories are given in Table A1. The Lagrangian time ¯ow variance, which is inferred from the sample La- scale varied between 2.7 days in the undercurrent to 5.6 grangian autocovariance function days for ¯oats showing reversing ¯ow along the con- T tinental margin. A 5-day decorrelation time indicates 1 R (␶, T ) ϭ uЈ(t)uЈ(t ϩ ␶) dt, (A1) that creating the pseudo¯oats by restarting the ¯oat tra- ij T ͵ ij 0 jectories every 10 days did not introduce a bias. The single particle diffusivities represent the time rate where uЈϭu Ϫ umean is the perturbation velocity at each time, ␶ is the time lag, and T is the length of the time of change of dispersion about the mean trajectory, or series (Paduan and Niiler 1993). The autocorrelation the rate at which the envelope of ¯oat paths diverges functions are obtained by normalizing the covariance from the mean. By design, the ¯oats were deployed in functions. The mean ¯oat polarization, or sense of ro- a region with asymmetric velocity patterns. The diffu- tation, is de®ned as sivities for the undercurrent and marginal ¯ow cases are not meaningful because those ¯ows are constrained by ␶ the boundary. The diffusivity is meaningful for the in- J(␶) ϭ (R (␶Ј) Ϫ R (␶Ј)) d␶Ј, (A2) ͵ u␷␷u terior ¯oats. The diffusivity was estimated from the La- 0 grangian autocovariance function (Paduan and Niiler where Rij is cross correlation (Poulain and Niiler 1989). 1993) as

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC APRIL 1999 GARFIELD ET AL. 581

TABLE A2. Comparison of interior single particle statistics with other eastern Paci®c Lagrangian estimates. Diffusivity Timescales Length scales

Kuu K␷␷ U V U V Author (m2 sϪ1) (m2 sϪ1) (days) (days) (km) (km) Subsurface present study 1970 1830 4.4 4.0 25 23 Thomson et al. (1970) 1700 2200 2.6 2.7 20 23 Surface Poulain and Niiler (1989) 3400 4300 4.2 4.7 40 48 Brink et al. (1991) 8620 2430 3.3 0.9 Paduan and Niiler (1993) 835 1524 2.8 3.8 13.9 22.0

t 3) TWO-PARTICLE STATISTICS Kii(t) ϭ Rii(␶) d␶. (A5) ͵ Triads of ¯oats were released in order to estimate 0 directly the dispersion from spreading particles. The

Here Kuu(t) should approach asymptotic values, Kϱ, for three triads and the ¯oat pairs NPS 9 and 14 and NPS long timescales (Davis 1991). The diffusivity given in 28 and 33 provide 11 particle-pair measurements orig-

Table A1 is an estimate of Kϱ. It was reached at a time inating from the launch point west of San Francisco. interval longer than the estimated integral timescale. The estimate from such a small number of pairs is ex- The cross-shore diffusion, 2015 m2 sϪ1 was slightly larg- tremely noisy. Many more ¯oat returns will be necessary er than the alongshore diffusion, 1788 m2 sϪ1. Subsur- before statistically robust estimates will be possible. face diffusion does not show a signi®cant bias in either With that understanding, we present these preliminary direction, probably due to the eddy ®eld dominance. results for comparison with the single particle estimates. The single particle diffusivities determined for the The rate at which two particles separate is the eddy ¯oats in the Paci®c interior can be compared with other diffusivity for the time-dependent spreading of par- estimates (Table A2). Thomson et al. (1990) analyzed ticles (Davis 1985). In a manner similar to the single the trajectories of a series of surface drifters that were particle diffusivity, the particle-pair diffusivity can be drogued at 100 to 120 m. Their study was in the eastern expressed as Paci®c west of Queen Charlotte Island, and the mean 1 d 1 trajectory of their ¯oats was eastward. Their mean dif- ␣ij ϭ sijЈs Ј , (A6) fusivities were about the same as those found in the 4 dt΂΃ N ͸ present study. The major difference is that their K ␷␷ where snЈ is the deviation of any particle pair from the component was larger than the K component, while uu mean separation of many pairs at that time step (snЈϭ the present result shows less directional variability. sn Ϫ s). Particle-pair dispersion was estimated for the The diffusivities at 300-m depth from the RAFOS triad deployments released from the primary launch site. ¯oats can also be contrasted with surface ¯oat single After an initial 3-day period of equal growth, the particle statistics derived from ¯oats that traversed the alongshore component of variance continues to grow same region. Brink et al. (1991) reported surface dif- while the cross-shore component begins to oscillate. fusivities that were larger than the subsurface values by This demonstrates the anisotropic nature of the ¯ow; a factor of 2 to 3. Unlike the present study, they found westward moving ¯oats encountered slow moving ed- Kuu to be larger than K␷␷ by a factor of 3.5. dies, while ¯oats in the undercurrent continued to move Results from drifter studies to the north, off Vancou- poleward. Northward variance is larger because of the ver Island (Paduan and Niiler 1993) and south of Pt. spreading between ¯oats in the undercurrent and ¯oats Conception (Poulain and Niiler 1989), had diffusivity entering the interior. Eastward variance remained small- values closer to the present study, but, like Thomson et er because there are no currents producing similar east- al. (1990), with K␷␷ larger than Kuu. Swenson and Niiler ward shear. (1996) demonstrated the areal variability of single par- The particle-pair diffusivities are extremely noisy, ticle diffusion and give a range of 2000±5500 and 1100± showing oscillatory behavior starting with day 10 and 8700 m2 sϪ1 for the zonal and meridional diffusivities. increasing in amplitude for both northward and eastward They did not calculate values for the region represented components. The development of the oscillatory be- in the present study. havior is most probably related to an increasing number The decorrelation time and length scales for both sur- of ¯oats becoming entrained in eddies after leaving the face and subsurface measurements all fall within a sim- undercurrent. In the mean, the alongshore diffusivity is ilar range of 2±5 days and 15±40 km. a factor of 6 greater than the cross-shore diffusivity.

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC 582 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 29

Our computed two-particle statistics are extremely Freeland, H. J., W. R. Crawford, and R. E. Thomson, 1984: Currents noisy. In part this is due to only having a limited number along the Paci®c coast of Canada. Atmos.-Ocean, 22, 151±172. Goodman, L., and E. R. Levine, 1990: Vertical motion of neutrally of ¯oats. It is also due to the ¯oat mission length being buoyant ¯oats. J. Atmos. Oceanic Technol., 7, 38±49. shorter than the lifespan of the SCV eddies. The eddies Hickey, B. M., 1979: The California Current SystemÐHypotheses appear to have limited diffusion during their existences, and facts. Progress in Oceanography, Vol. 12, Pergamon, 259± so the particle pair variance is a re¯ection of the eddy 284. rotation and slow drift rather than of particle diffusion Huyer, A., P.M. Kosro, S. Lentz, and R. C. Beardsley, 1989: Poleward ¯ow in the California Current System. Poleward Flows along in a more homogeneous ¯uid. Eastern Oceanic Boundaries, Coastal and Estuarine Studies No. 34, E. Neshyba, et al., Eds., Springer-Verlag, 142±156. , J. A. Barth, P. M. Kosro, R. K. Shearman, and R. L. Smith, REFERENCES 1998: Upper-ocean water mass characteristics of the California Current, summer 1993. Deep-Sea Res., in press. Armi, L., D. Hebert, N. Oakey, J. Price, P. L. Richardson, T. Rossby, Johnson, R. H., and R. A. Norris, 1968: Geographic variation of sofar and B. Ruddick, 1989: Two years in the life of a Mediterranean speed and axis depth in the Paci®c Ocean. J. Geophys. Res., salt lens. J. Phys. Oceanogr., 19, 354±370. 73(14), 4695±4700. Baltz, K. A., 1997: Ten years of hydrographic variability off central Largier, J. L., B. A. Magnell, and C. D. Winet, 1993: Subtidal cir- California during the upwelling season. M.S. thesis, Department culation over the Northern California Shelf. J. Geophys. Res., of Oceanography, Naval Postgraduate School, Monterey, CA, 98 (C10), 18 147±18 179. 319 pp. [Available from Department of Oceanography, Naval Lynn, R. J., and J. J. Simpson, 1987: The California Current System: Postgraduate School, 833 Dyer Rd., RM 328, Monterey, CA The seasonal variability of its physical characteristics. J. Geo- 93943.] phys. Res., 92, 12 947±12 966. Benson, K. R., 1995: High frequency subsurface Lagrangian mea- , K. A. Bliss, and L. E. Eber, 1982: Vertical and horizontal surements in the California Current with RAFOS ¯oats. M.S. distributions of seasonal mean temperature, salinity, sigma-T, thesis, Department of Oceanography, Naval Postgraduate stability, dynamic height, oxygen, and oxygen saturation in the School, Monterey, CA, 88 pp. [Available from Department of California Current, 1950±1978. CalCOFi Atlas, No. 50, South- Oceanography, Naval Postgraduate School, 833 Dyer Rd., RM west Fisheries Science Center. [Available from Southwest Fish- 328, Monterey, CA 93943.] eries Science Center, National Marine Fisheries Service, Bower, A. S., 1994: RAFOS Float Technology Workshop Proceed- FSWC1, 8604 La Jolla Shore Dr., La Jolla, CA 92038.] ings. Woods Hole, MA, Woods Hole Oceanographic Institution, McDowell, S. E., and H. T. Rossby, 1978: Mediterranean Water: An 115 pp. intense mesoscale eddy off the Bahamas. Science, 202, 1085± , L. Armi, and I. Ambar, 1995: Direct evidence of meddy for- 1087. mation off the southwestern coast of Portugal. Deep-Sea Res., McWilliams, J. C., 1985: Submesoscale, coherent vortices in the 42, 1621±1630. ocean. Rev. Geophys., 23, 165±182. , , and , 1997: Lagrangian observations of Meddy for- , E. D. Brown, H. L. Bryden, C. C. Ebbesmeyer, B. A. Elliott, mation during a Mediterranean Undercurrent seeding experi- R. H. Heinmiller, B. Lien Hua, K. D. Leaman, E. J. Lindstrom, ment. J. Phys. Oceanogr., 27, 2545±2575. J. R. Luyten, S. E. McDowell, W. B. Owens, H. Perkins, J. F. Brink, K. H., and T. J. Cowles, 1991: The Coastal Transition Zone Price, L. Regier, S. C. Riser, H. T. Rossby, T. B. Sanford, C. Y. Program. J. Geophys. Res., 96 (C8), 14 637±14 647. Shen, B. A. Taft, and J. C. van Leer, 1983: The local dynamics , R. C. Beardsley, P. P. Niiler, M. Abbott, A. Huyer, S. Ramp, T. Stanton, and D. Stuart, 1991: Statistical properties of near- of eddies in the Western North Atlantic. Eddies in Marine Sci- surface ¯ow in the California coastal transition zone. J. Geophys. ence, A. R. Robinson, Ed., Springer-Verlag, 92±113. Res., 96 (C8), 14 693±14 706. Mooers, C. N. K., 1989: Workshop Summary: Poleward ¯owÐOb- Chelton, D. B., 1984: Seasonal variability of alongshore geostrophic servational and theoretical issues. Poleward Flows along Eastern velocity off central California. J. Geophys. Res., 89, 3473±3486. Oceanic Boundaries, Coastal and Estuarine Studies No. 34, E. Collins, C. A., N. Gar®eld, R. G. Paquette, and E. Carter, 1996a: Neshyba, et al., Eds., Springer-Verlag, 3±16. Lagrangian measurement of subsurface poleward ¯ow between Noble, M., R. C. Beardsley, J. V. Gardner, and R. L. Smith, 1987: 38ЊN and 43ЊN along the West Coast of the United States during Observations of subtidal currents over the northern California summer, 1993. Geophys Res Lett., 23, 2461±2464. continental slope and adjacent basin: Some evidence for local , R. G. Paquette, and S. R. Ramp, 1996b: Annual variability of wind forcing. J. Geophys. Res, 92(C2), 1709±1720. ocean currents at 350m-depth over the continental slope off Point Paduan, J. D., and P. P. Niiler, 1993: Structure of velocity and tem- Sur, California. CalCOFI Reports, Vol. 37, 1±7. [Available from perature in the northeast Paci®c as measured with Lagrangian Southwest Fisheries Science Center, National Marine Fisheries drifters in fall 1987. J. Phys. Oceanogr., 23, 585±600. Service, FSWC1, 8604 La Jolla Shore Dr., La Jolla, CA 92038.] Paquette, R. G., 1996: RAFOS Float Manual. Naval Postgraduate , N. Gar®eld, T. A. Rago, F. W. Rischmiller, and E. Carter, 1998: School, Monterey, CA., 44 pp. [Available from Department of Mean structure of the inshore countercurrent and California Un- Oceanography, Naval Postgraduate School, 833 Dyer Rd., RM dercurrent off Point Sur, California. Deep-Sea Res., in press. 328, Monterey, CA 93943.] D'Asaro, E. A., 1988: Generation of submesoscale vortices: A new Parker, H. A., 1996: Variations in coastal circulation off central Cal- mechanism. J. Geophys. Res., 93 (C6), 6670±6694. ifornia, Spring±Summer of 1993, 1994, and 1995. M.S. thesis, , S. Walker, and E. T. Baker, 1994: Structure of two hydrothermal Department of Oceanography, Naval Postgraduate School, Mon- megaplumes. J. Geophys. Res., 99 (C10), 20 361±20 373. terey, CA, 71 pp. [Available from Department of Oceanography, Davis, R. E., 1985: Drifter observations of coastal surface currents Naval Postgraduate School, 833 Dyer Rd., RM 328, Monterey, during CODE: Method and descriptive view. J. Geophys. Res., CA 93943.] 90, 4741±4755. Pingree, R. D., and B. Le Cann, 1993: A shallow meddy (a Smeddy) , 1991: Observing the general circulation with ¯oats. Deep-Sea from the secondary Mediterranean salinity maximum. J. Geo- Res., 38(Suppl. 1), S531±S571. phys. Res., 98, 20 169±20 185. Flament, P., 1986: Finestructure and associated with up- Poulain, P.-M., and P.P.Niiler, 1989: Statistical analysis of the surface welling ®laments. Ph.D. dissertation, University of California, circulation in the California Current System using satellite- San Diego, 142 pp. [Available from Scripps Institute of Ocean- tracked drifters. J. Phys. Oceanogr., 19, 1588±1603. ography, University of California, La Jolla, CA 92093-0230.] , A. Warn-Varnas, and P. P. Niiler, 1996: Near-surface circulation

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC APRIL 1999 GARFIELD ET AL. 583

of the Nordic seas as measured by Lagrangian drifters. J. Geo- Measurement, S. P. Anderson, G. F. Appell, A. J. Williams III, phys. Res., 101(C8), 18 237±18 258. Eds. New York, NY, IEEE, 163±168. Ramp, S. R., P. F. Jessen, K. H. Brink, P. P. Niiler, F. L. Daggett, and Schwing, F. B., D. M. Husby, N. Gar®eld, and D. E. Tracy, 1991: J. S. Best, 1991: The physical structure of cold ®laments near Mesoscale oceanic response to wind events off central California Point Arena, California, during June 1987. J. Geophys. Res., in Spring 1989: CTD surveys and AVHRR imagery. CalCOFI 96(C8), 14 859±14 884. Reports, Vol. 32, 47±62. [Available from Southwest Fisheries , N. Gar®eld, C. A. Collins, L. K. Rosenfeld, and F. B. Schwing, Science Center, National Marine Fisheries Service, FSWC1, 1995: Circulation studies over the and slope 8604 La Jolla Shore Dr., La Jolla, CA 92038.] near the Farallon Islands, CA. Naval Postgraduate School Tech. Smith, R. L., S. D. Pierce, J. A. Barth, P. M. Kosro, and C. Wilson, Rep. NPS-OC-95-004, 22 pp. [Available from Department of 1996: Extensive observations of the poleward undercurrent along Oceanography, Naval Postgraduate School, 833 Dyer Rd., RM the continental margin off California, Oregon and Washington. Abstract only. Eos., Trans. Amer. Geophys. Union, 77(3), 328, Monterey, CA 93943.] (Suppl.), OS131. Reid, J. L., 1962: Measurements of the California Countercurrent at Steger, J., F. Schwing, C. A. Collins, L. Rosenfeld, N. Gar®eld, and a depth of 250 meters. J. Mar. Res., 20, 134±137. E. Gezgin, 1998: Circulation and water masses in the Gulf of Richardson, P. L., D. Walsh, L. Armi, M. Schroeder, and J. F. Price, the Farallones. Deep-Sea Res., in press. 1989: Tracking three meddies with SOFAR ¯oats. J. Phys. Stevenson, M. R., J. G. Pattullo, and B. Wyatt, 1969: Subsurface Oceanogr., 19, 371±383. currents off the Oregon coast as measured by Parachute Drogues. Rischmiller, F. W., 1993: Variability of the California Current System Deep-Sea Res., 16, 449±461. off Point Sur, California, from April 1988 to December 1990. Swenson, M. S., and P. P. Niiler, 1996: Statistical analysis of the M.S. thesis, Department of Oceanography, Naval Postgraduate surface circulation of the California Current. J. Geophys. Res., School, Monterey, CA, 159 pp. [Available from Department of 101 (C10), 22 631±22 645. Oceanography, Naval Postgraduate School, 833 Dyer Rd., RM Swift, D. D., and S. C. Riser, 1994: RAFOS Floats: De®ning and 328, Monterey, CA 93943.] targeting surfaces of neutral buoyancy. J. Atmos. Oceanic Tech- Rosenfeld, L. K., F. B. Schwing, N. Gar®eld, and D. E. Tracy, 1994: nol., 11, 1079±1092. Bifurcated ¯ow from an upwelling center: A cold water source Taylor, G. I., 1921: Diffusion by continuous movements. Proc. Lon- from Monterey Bay. Contin. Shelf Res., 14, 931±964. don Math. Soc., 20, 196±212. Rossby, H. T., E. R. Levine, and D. N. Connors, 1985: The Isopycnal Thomson, R. E., P. H. LeBlond, and W. J. Emery, 1990: Analysis of Swallow FloatÐA simple device for tracking water parcels in deep-drogued satellite-tracked drifter measurements in the north- the ocean. Progress in Oceanography, Vol. 14, Pergamon, 511± east Paci®c. Atmos.±Ocean, 24, 409±443. Tisch, T. D., S. R. Ramp, and C. A. Collins, 1992: Observations of 525. the and water mass characteristics off Point Rossby, T., D. Dorson, and J. Fontaine, 1986: The RAFOS System. Sur, California from May 1988 through November 1989. J. Geo- J. Atmos. Oceanic Technol., 3, 672±679. phys. Res., 97(C8), 12 535±12 555. Sanderson, B. G., 1995: Structure of an eddy measured with drifters. Wooster, W. S., and J. H. Jones, 1970: California Undercurrent of J. Geophys. Res., 100(C4), 6761±6776. northern Baja California. J. Mar. Res., 28, 253±260. Sanford, T. B., R. G. Deaver, and J. H. Dunlap, 1995: Barotropic Zenk, W., K. Schultz-Tokos, and O. Boebel, 1992: New observations ocean velocity observations from an Electric Field Float, a mod- of meddy movement south of the Tejo Plateau. Geophys. Res. i®ed RAFOS ¯oat. Proc. IEEE Fifth Working Conf. on Current Lett., 19, 2389±2392.

Unauthenticated | Downloaded 09/25/21 04:00 AM UTC