JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 93, NO. C6, PAGES 6695-6710, JUNE 15, 1988

Gulf Stream Bimodality and Variability Downstream of the Charleston Bump

JOHN M. BANE, JR. 1

Marine Sciences Program, University of North Carolina, Chapel Hill

WILLIAM K. DEWAR 2

Department of Oceanography, Florida State University, Tallahassee

Gulf Stream mean flow and variability have been studied in the Gulf Stream Deflection and Meander Energetics Experiment (DAMEX). This study was conducted between September 1981 and April 1982 in the area of the Charleston bump, off the coast of the southeastern United States. Using data from an array of current meters and satellite imagery, it has been shown that the seaward deflection of the Gulf Stream by the bump has a bimodal character and that the Gulf Stream meander field downstream of the bump differs between the two states of deflection. The two deflection conditions have been termed the weakly deflected state and the strongly deflected state in this paper. A weakly deflected Gulf Stream flows past the Charleston bump and close enough to the edge of the continental shelf that its shoreward sea surface temperature (SST) front remains inshore of the 600-m isobath between Charleston and Cape Hatteras. A strongly deflected Gulf Stream tums sharply eastward at the bump and flows seaward far enough to have its shoreward SST front offshore of the 600-m isobath for several tens of kilometers along the continental slope downstream of the bump. The transitions between deflection states may occur very rapidly, and once in a strongly deflected state the Gulf Stream may remain so for up to several months. Two transitions were observed during DAMEX, one from a weakly to a strongly deflected state and one in the other direction. The transition from weakly to strongly deflected state took only a few days at a given site, while the signatures of the other transition were not as definite in the data records. The strongly deflected state occurred during the fall and winter in DAMEX (October through January), and satellite observations from earlier years suggest that it is this time of year in which the strongly deflected state preferentially occurs. It was not clear from the DAMEX data alone how the transition from a weakly deflected to a strongly deflected state may be induced, but path or transport changes in the Gulf Stream as it approaches the bump and interactions between the current and rings or eddies are candidate mechanisms. The bimodal character of the stream's path affects the low-frequency variability of the current between the Charleston bump and Cape Hatteras in the following manner. Typical meander/frontal eddy patterns dominate the mesoscale fluctuation field during times of weak deflection, while larger-amplitude meanders of the Gulf Stream jet, which can have very sharp curvature and strong anticyclonic circulations in their crests, exist during times of strong deflection. Furthermore, these larger-amplitude meanders, which we call the DAMEX biweekly meanders, have generally slower phase propagation speeds than do the typical meander/frontal eddy events (about 20-25 km/day compared with 35-60 km/day) as well as longer periods (about 16 days compared to 1 week).

INTRODUCrION

Our understanding of the nature, causes, and effects of Gulf Stream variability has increased substantially within the past decade. Numerous observational studies have built a data base that has provided several derailed descriptions of fluctuation events and processes in the stream. In particular, wavelike meanders of the Gulf Stream and their attendant

frontal eddies are now recognized as a dominant form of mesoscale variability in the Gulf Stream within the South Atlantic Bight, the region between Cape Canaveral, Florida and Cape Hatteras, North Carolina (Figure 1). Early views of these features emerged from the pioneering studies of von Arx et al. [1955], who mapped the warm filaments of water

1Also at Departments of Geology and Physics, University of North Carolina, Chapel Hill.

2Also at Supercomputer Computations Research Institute, Florida State University, Tallahassee.

Copyright 1988 by the American Geophysical Union.

Paper number 8C0172. 0148-0227/88/008C-0172505.00

which trail southward away from Gulf Stream meander crests and over the outer continental shelf; and of Webster [1961a], who successfully observed the progression of a train of four meanders past Onslow Bay, North Carolina, during a 28-day period in 1958. The more recent reports by Maul et al. [1978], Vukovich and Crissrnan [1978], Legeckis [1979], Pietrafesa and Janowitz [1980], Bane et al. [1981], Brooks and Bane [1981], Lee et al. [1981], Dewar and Bane [1985], Li et al. [1985], and Osgood et al. [1987] and those contained in the May 30, 1983, special issue of the Journal of Geophysical Research have added considerably to the earlier studies and have suggested new questions about the dynamics of the Gulf Stream.

One striking feature of the Gulf Stream south of Cape Hatteras is the recurring seaward deflection of its path near 32øN latitude. Most of the investigators listed above have suggested that the deflection process plays a significant role in determining the nature of the Gulf Stream's variability. Legeckis [1976, 1979] has proposed that the deflection is caused by the flow of the stream over the "Charleston bump," a topographic feature located on the upper continental slope at about 31øN latitude offshore of Charleston, South Carolina (see Figure 1). Several studies

6695

6696 BkNE AND DEWAR: GULF STREAM BLMODALITY AND VARIABILITY

35 Q

80 • ?5 ø

/ HATTERAS

- "7// .. ::...."Z.4

_ :;• "CHARLESTON • DUMP'

JACKSONVILLE 30 •

25"

•O ø 75 ø

fluctuations [Osgood et al., 1987] observed during the experiment.

DAMEl(

DAMEX was composed of three observational components: (1) current meters and bottom pressure gauges were supported on seven moorings for the 7-month-long DAMEX period, which began in September 1981, (2) a conductivity-temperature-depth (CTD) survey was made in the deflection region during September 1981, and (3) four air-deployed expendable bathythermograph (AXBT) surveys were conducted along the Gulf Stream in an area which encompassed the deflection region in March 1982. The seven instrument moorings constituted the central DAMEX component, and they were grouped into two small three- mooring arrays and one single mooring (Figure 2). The southernmost array, E, contained three moorings with two current meters each and was located on the southwest

(upstream) flank of the bump, about 90 km from the center of the bump. The middle array, F, also consisted of three moorings with two current meters each, and it was located roughly 90 km northeast (downstream) of the bump. Bottom pressure gauges were mounted on two of the F moorings. A single mooring, G, with two current meters was positioned off Onslow Bay. Arrays E and F were moored in areas which had not previously been sampled, while the G mooring was placed at a site studied earlier, during the Gulf Stream

Fig. 1. The continental margin of the southeastern United States. Meanders Experiment in 1979 [cf. Brooks and Bane, 1981]. The 200-rn and 600-rn isobaths are shown, and they delineate the The three moorings in each of the E and F arrays were position of the "Charleston bump," the topographic irregularity placed in an 'L' configuration, with one mooring on the 300- near 31øN. The inset encloses the study area, which is shown in greater detail in Figure 2. m isobath and two on the 400-m isobath. In each array the

shallow mooring was identified as mooring 1, the one directly downslope from this as mooring 2, and the one alongslope to the north or northeast as mooring 3. The top

have shown that the stream's eddy variability increases (bottom) current meter at each of the seven moorings was downstream of the deflection [see Olson et al., 1983], and placed at a nominal depth of 210 m (270 m). We will refer Vukovich and Crissman [1978], Legeckis [1979], and Lee et to each instrument by a unique name which denotes its al. [1981] have documented individual meander/frontal eddy array, mooring, and instrument location. For example, E1T events propagating through the deflection region, (E1B) is the current meter at the top (bottom) position on apparently undergoing amplification while doing so. Dewar the E1 mooring. Table 1 gives mooring location and Bane [1985] have suggested from in situ data that the information. Gulf Stream is both barotropically and baroclinically The current meters measured horizontal velocity, unstable just downstream of the bump, supporting the temperature and conductivity at 30-min intervals from the notion that the bump topography is destabilizing the stream middle of September 1981 to the middle of April 1982. The and causing its meanders and other fluctuations to grow by data were low pass filtered to remove tides and internal feeding on the energy of the mean flow. Even with this waves by using a modified Lanczos filter with a quarter- observational background, the exact nature of the deflection power point at 40 hours. The resulting 40-hour low-passed process, how it varies in time, and how these variations (40 HRLP) time series averaged 200 days in length and had affect the enhanced Gulf Stream variability downstream of an effective sampling interval of 6 hours. This processing the Charleston bump have yet to be determined. follows that of our earlier experiments in the South Atlantic

During the period September 1981 through April 1982, Bight [cf. Brooks et al., 1981]. A full documentation of all the Gulf Stream Deflection and Meander Energetics DAMEX mooring data and processing techniques is given by Experiment (DAMEX) was conducted in the region of the Bane and Dewar [1983]. stream's seaward deflection. This experiment was designed to We use a rotated coordinate system in this presentation in study the deflection process and its relationship to Gulf order to facilitate our discussion of cross-stream and Stream variability. In this paper we describe the Gulf Stream downstream flow. Note in Figure 2 that some of the mean structure and events observed during DAMEX. In particular, flow vectors have significant offshore components. We we present data which suggest that the deflection of the have rotated the coordinate system at each of the three sites stream has a bimodal character and that the nature of the low- (E, F and G) to be approximately aligned with the mean frequency variability of the stream in the region from the current at that particular site. Coordinate orientations are Charleston bump to Cape Hatteras varies between the two such that the x (y) axis is positive in the offshore states of deflection. In related DAMEX papers we discuss the (downstream) direction. energetics of the stream [Dewar and Bane, 1985] and the The spectra which are discussed here use either the full dynamical balances and vertical velocities in Gulf Stream time series or some selected subset. The record-long time

BANE AND DEWAR: GULF STREAM BLMODAL1TY AND VARIABILITY 6697

4000

3000

CHAFtL.E $TON

SAVANNAH

2OO0

•ooo

•oo

I"[::AN FLC&/ VECTORS BOTTOPI CONTOURS IN I•t•TERS

Ill I I I ] I I i o 20 40 643 BO 0

cm/eec k'm

Fig. 2. The DAMEX study area. The seven instrument moorings are shown, which are grouped into two three-mooring arrays (E and F) and a single mooring (G). Arrays E and F were placed upstream and downstream of the Charleston bump, respectively. The 7-month-long mean current velocity measured at the 210-m level at each mooring is also shown (270-m level currents are shown for F2 and F3 because of instrument malfunctions at the 210-m level at each of those moorings). The location of the mean Gulf Stream shoreward surface thermal front as determined by Bane and Brooks [1979] and Olson et al. [1983] is indicated by the dashed line.

series were broken into pieces of 32 days, and the resulting spectral estimates were smoothed using a Parzen filter. Each of these spectra has 22.26 degrees of freedom and an effective bandwidth of 0.058 cpd. We also found it useful to

study selected periods within the complete time series. These subrecords were chosen to be 64 days in duration and for spectral calculations were broken into four pieces. Again, smoothing was done with a Parzen filter and resulted

6698 BANE AND DEWAR: GULF STREAM BIMODALITY AND VARIAB•

Table 1. DAMEX Mooring Summary

Array Mooring Location

E 1 31'14.7•, 79'40.7'W

Deployment Date, Recovery Date, Serial Depth 1981 1982 Meter Number m

,

Sept. 19 April 25 T 5705 201 B 3424 261

E 2 31'13.8•, 79'38.5'W Sept. 19 April 25 T 5707 229 B 5708 289

E 3 31'24.7•, 79'33.7'W Sept. 19 April 25 T 5706 219 B 3427 279

F* 1 32'25.5•q, 78'15.3'W Sept. 18 April 22 T 3337 206 B 3423 266

F* 2 32'16.9•q, 78'10.4'W Sept. 18 April 22 T 3425 2101 B 3344 2701

F 3 32'22.3•, 77'55.7'W Sept. 18 April 22 T 3426 212 B 3345 272

G -- 33'21.0•q, 76'40.7'W Sept. 17 April 21 T 3332 2101 B . 3343 270•'

*These moorings supported a bottom pressure gauge. l Nominal depths used.

in spectra with 14.84 degrees of freedom and an effective Our analysis indicates that these changes were associated bandwidth of 0.12 ClM. with differences in the degree of deflection of the Gulf

Stream at the Charleston bump. In the following discussion M}•AN FLow AND Low FREQUENCY FLUCTUATIONS we will describe in greater detail the nature of the

The Gulf Stream's Weakly and Strongly fluctuations and relate their changes to the deflection Deflected States process.

Fluctuations of the Gulf Stream changed in character twice We will refer to the configuration of the Gulf Stream prior during the 7-month DAMEX mooring period. The to mid-October and after January as the "weakly deflected" oscillations observed at each current meter mooring were state and to its configuration during the mid-October predominantly in the 4- to 8-day period band prior to mid- through January period as the "strongly deflected" state. October and after January, while during the mid-October Two satellite images of the sea surface temperature (SST) through January interval they were of distinctly lower field taken during DAMEX show the two deflection states frequency, with periods ranging approximately from 14 to quite nicely. Figure 3 presents redrawn versions of the 20 days. These changes were most noticeable at array F, advanced very high resolution radiometer (AVHRR) images although they are obvious at the other locations as well. shown in Plate 1 for December 3, 1981, and March 2, 1982,

'apparent' deflection

TED WEAKLY DEFLECTED

57;!! (03- DEC- 81) (02- MAR- 82)

..... 200 meier contour

----$O0 meter contOur

Fig. 3. Locations of the Gulf Stream on December 3, 1981, and March 2, 1982, during DAMEX. These retraced versions of the AVHRR SST images from those dates show the stream in a typical strongly deflected configuration (December 3) and weakly deflected configuration (March 2) in the region of the Charleston bump. (The original AVHRR images are shown as Plate 1 and can be found in the separate color section in this issue.) Note that the stream flows well seaward of the 600-m isobath just downstream of the bump when strongly deflected, but not when weakly deflected. Note also the train of meander/frontal eddy features downstream of the bump in the weakly deflected case.

BANE AN'D DEW•: GULF STREAM BIMODALITY AND VARIABK/q• 6699

which are typical of the strongly and weakly deflected The 40 HRLP time series data are discussed in the next states, respectively. (Plate 1 can be found in the separate few subsections. This presentation focuses on representative color section in this issue.) The primary difference to notice time series from one current meter at each of the E, F, and G between these two images is the offshore position of the locations. The time series for cross-stream velocity main body of the Gulf Stream just downstream of the component u, downstream velocity component v, Charleston bump. In the weakly deflected state the stream's temperature T, and salinity S are given below for E2T, F3B, shoreward SST front is everywhere shoreward of the 600-m and GT. Table 2 presents the first-order statistics for all isobath upstream of Cape Hatteras. In the strongly deflected instruments. state the shoreward SST front (and thus the entire width of the Gulf Stream) is seaward of the 600-m contour by at least Array E 50 km for that portion of its path just downstream of the Array E was located upstream of the center of the bump. We have found this to be a good "rule of thumb"; Charleston bump in order to monitor the stream as it namely, if the Gulf Stream's shoreward SST front in the first approached the deflection region from the south. Earlier several tens of kilometers just downstream of the Charleston studies have shown that the low-frequency flow variability bump is seaward of the 600-m isobath, then we consider the over the outer shelf (about 20-30 km upslope from array E) stream to be in its strongly deflected state. is usually dominated by northward propagating Gulf Stream

There is a noticeable difference in the stream's meander meanders and frontal eddies [Lee et al., 1981; Lee and patterns between the two states of deflection in Figure 3. Atkinson, 1983; Li et al., 1985]. No long-term direct Downstream of the bump, several of the typical measurements of subsurface currents had been made within meander/frontal eddy patterns and their associated warm the strearn's cyclonic frontal zone in this area prior to filaments may be seen in the weakly deflected example of DAMEX, however. March 2. In contrast, the path of the stream in that region The 7-month-long 40 HRLP time series from E2T are in the strongly deflected case of December 3 is a rather shown in Figure 4. Recall that a coordinate rotation has smooth arc. The stream leaves the deflection region flowing been made here to place y (and thus v) in the direction of eastward well past the 600-m isobath, and then it turns the mean flow, thereby giving little cross-stream mean. The gradually northward toward Cape Hatteras. Earlier mean velocities at all of the E instruments do exhibit a observations of the flow over the continental shelf and small cross-isobath component, however. Velocity upper slope in this region have shown that there is a quasi- fluctuations at array E were predominantly in the stationary, cyclonic circulation pattern (the so-called downstream direction. Typical rms amplitudes for the v "Charleston gyre") positioned shoreward of the Gulf fluctuations were near 20 cm/s, while those for the u Stream's arcuate path in this state [Singer et al., 1983]. fluctuations were about 4 cm/s. Good correlations between v Although none appear in this particular image, meanderlike and T fluctuations suggest that much of the variability at E fluctuations do occur in the Charleston to Cape Hatteras was due to lateral movements in the Gulf Stream. region during a period of strong deflection. Such meanders Fluctuations with periods ranging from about 20 days down of the Gulf Stream jet have lateral amplitudes greater and to 2 days are seen in the time series, with the lower phase speeds slower than do the more commonly observed frequencies generally being more energetic. During the early meander/frontal eddy patterns. Data presented below show and late portions of DAMEX, when the stream was in the this meander motion to result in a folded back, sometimes weakly deflected state, satellite imagery and current meter "sawtooth" pattern in the Gulf S tream's path, resulting in data indicate that the fluctuations were primarily very strong anticyclonic flow around each meander crest. We meander/frontal eddy events. Coherence estimates, to be wish to distinguish this type of meander motion from the discussed below, give a downstream propagation speed of 60 more commonly studied meander/frontal eddy motions of the km/day, consistent with the earlier observations of Lee et weakly deflected state, and so we refer herein to the larger- al. [1981] and Lee and Atkinson [1983], who found amplitude meanders observed downstream of the bump during meander/frontal eddy phase speeds ranging from 35 to 60 the strongly deflected condition of this study as the km/dayin this region. "DAMEX biweekly meanders" or just "biweekly meanders." The record-length spectrum for v at E2T (Figure 5) is This name reflects the fact that their observed periods are predominantly red and decays with roughly a-2 power law. close to 2 weeks, about double those of the meander/frontal Spectral values for u are about 2 orders of magnitude smaller eddies of the weakly deflected Gulf Stream. than those for v out to the 4-day period, consistent with the

The record-long mean velocity vectors in Figure 2 show time series in Figure 4. Comparable u and v energy levels the effect of the seaward deflection at the F moorings. The are not evident at any frequency; the fluctuations are always relatively low mean velocities there are due to the shift of more energetic in the downstream direction. The redness of the stream away from array F during the strongly deflected these spectra indicate that the lowest frequencies which may state. This greatly reduced the mean downstream speeds there have occurred at this location are not resolved by our 32-day for the mid-October through January period. Because these subrecords. Examination of the time series does not indicate moorings were sometimes in and sometimes out of the the presence of any trends. stream, the essentially eastward means at FiT and F2B may The transition to the strongly deflected state began at E2T not be reliable indicators of the longer-term mean path of about October 11, and was characterized by rapid increases the main stream in this area. (For comparison, the mean in v, T, and S (Figure 4). An offshore velocity fluctuation path of the shoreward SST front as indicated by Bane and that reached a peak of about 15 cm/s and lasted for about Brooks [1979], Legeckis [1979], and Olson et al. [1983] is 3-4 days accompanied these increases. During the strongly oriented near 60øT there.) The array F mean velocities are deflected period following October 11, v decreased and more suggestive of the cyclonic flow around the southern ultimately reversed for a short time near November 16. Both edge of the Charleston gyre. T and S decreased as well, but quite slowly in relation to v.

6700 BA.¾E AND DEWAr: Gtxy STREAM BIMODALrrY AND VARIABU./q•

Instrument ,

E1T

E1B

E2T

E2B

E3T

Table 2. First-order Statistics for the 40 HRLP Time Series

Number of Standard Variable Observations Minimum Maximum Mean Deviation

T 818 6.75 17.53 12.90 2.64 S 818 35.00 36.38 35.71 0.45 u 818 -14.61 19.13 0.00 5.06 v 818 -45.37 133.60 42.50 34.55

T 808 6.9 8 15.67 10.36 2.21 S 808 35.12 36.28 35.53 0.40 u 808 -12.98 12.34 -0.01 4.02 v 808 -38.45 64.41 18.77 19.09

T 822 7.57 17.09 12.89 2.27 S 822 34.84 36.18 35.50 0.44 u 822 -22.95 16.36 -0.02 5.33 v 822 -23.72 145.25 59.28 31.80

T 844 7.08 15.13 10.76 1.90 S 844 32.96 35.84 34.97 0.56 u 844 - 12.17 29.51 -0.03 7.01 v 844 -22.53 92.24 43.80 19.53

T 807 7.57 16.64 12.72 2.36 S 807 34.73 36.17 35.45 0.46 u 807 -12.39 16.03 0.48 3.49 v 807 -34.63 144.26 57.89 37.44

E3B

FIT

F1B

T 823 7.04 15.19 10.48 2.04 S 823 35.10 36.18 35.51 0.40 u 823 .... v 823 ....

T 817 S 817 u 817 v 817

T 821 S 821 u 821 v 821

7.49 18.35 11.51 2.24 35.06 36.56 35.53 0.42

-52.49 34.12 0.00 15.22 -36.51 105.36 13.58 27.18

7.30 16.74 9.95 1.86 35.13 36.45 35.43 0.38

F2T

F2B

T 180 8.75 19.70 12.14 2.46 S 180 35.01 38.06 35.59 0.58 u 180 .... v 180 ....

T 839 6.47 21.22 10.89 2.14 S 839 21.94 49.02 35.51 1.75 u 839 -52.89 35.91 0.01 13.52 v 839 -43.24 79.50 15.00 26.64

F3T

F3B

T 131 11.74 13.42 12.25 0.50 S 131 35.06 37.83 35.60 0.68 u 131 .... v 131 ....

T 807 7.48 18.20 10.64 2.04 S 807 35.04 36.57 35.40 0.42 u 806 -62.17 66.14 0.00 17.55 v 806 -38.43 78.22 12.46 25.61

GTH* T 817 .... S 817 .... u 817 -51.15 44.07 0.00 13.69 v 817 -36.40 138.12 36.51 34.81

T 819 8.30 19.15 12.47 2.53 S 819 34.69 36.60 35.63 0.47 u 819 .... v 819 ....

*Direction as measured by GB was combined with speed as measured by GT to compute GTH, a hybrid velocity time series.

BANE AND DEWAR: GULF STREAM B[MODAI./TY AND VARIABIt.rrY 6701

150

tO0

o -50

Days 50 tO0 150 200 250

,

•4ø"•'PI E2T E2T is not significantly different from zero throughout the resolved frequency bands (Figure 6). There is evidence of downstream propagation at about 60 kin/day in the phase estimates between E2T and E3T (Figure 7), a rate consistent with eaxlier observations. This speed is not greatly different from the mean speed of the Gulf Stream observed at E2T, which is 59 cm/s (51 km/day). The only variables measured at the top E instruments which were appaxently uncorrelated

i were the cross-stream speeds at E2T and E3T. This may be -too , , • m , , , ,• an effect of topography, as E3 is closer to the crest of the

7a .... , .... , .... , .... , .... Charleston bump than is E2. We do stress, however, that 25

A.•. ^,._ • .... L ......... .,,,,,x ..... •,•_• the cross-stream speeds are relatively low. -ea v ...... - ......... , ..... '•'' "'"•" ' t The coherence between the top and bottom variability at -?5 • • , • , , the E2 mooring was significant down to periods of about 3 25 .... , .... , .... , ...... ' ' ß ' days. The phase between the two instruments, which were

2O

10

5 , I I I I I I

38 .... • .... • .... • .... • ....

37

• :)6

34 i i i I i i i i

SEP OCT NOV DEC JAN FEB MAR APR 1981 1982

separated 60 m vertically, was very small and negative, indicating that the top was leading the bottom. A detailed comparison of the data shows that this holds true on an event-by-event basis; thus it is unlikely that this region is baroclinically unstable. This result agrees with Dewar and Bane [1985], who found only marginal evidence for the eddy release of mean potential energy in the array E data.

The vertical structure at E1 is somewhat different from

that at E2. The E1T-E1B u time series pair is unambiguously coherent for only the lowest frequencies (coherence spectrum not shown). The 4- to 16-day band is marginally coherent at best, and the K '2 values at higher frequencies, while sometimes well above the 95% level, exhibit an uncomfortably large amount of scatter. For downstream flow at El, the top-to-bottom coherences are significant down to

Fig. 4. The 40 HRLP time series of downstream velocity periods of 3 days, with the possible exception of the 6.4- component v, cross-stream velocity component u, temperature T, and day band. We note, however, that /(2 is less than 0.5 in the salinity S measured at E2T (mean instrument depth of 229 m). The 10.7- to 8-day band, indicating that not more than half of transition from the weakly deflected state to the strongly deflected the energy in this band is common to the 210-m and 270-m state occurred near October 11 at this site, and is most noticeable in the temperature and salinity time series because of rapid currents. It is possible that the low E1T-E1B coherences are increases in those variables then. The transition back to the weakly due to effects of local topography. Recall that E1 was deflected state occurred during late January, when fluctuations of all moored on the 300-m isobath, with the E1B instrument variables shifted to more energetic higher frequencies than were only 30 m from the bottom. The low coherences in the u observed during the strongly deflected state. time series pair may be caused by a bottom-tra• or

Consequently, the water at E2T remained relatively warm (T > 15øC) and saline (S > 35.5 pP0 during the period of strong deflection, suggesting a change in the structure of the stream then. As was mentioned earlier, the dominant fluctuations in v, T, and S during this time had periods of about 14 to 20 days.

During late January the velocity fluctuations at array E began to shift back toward higher frequencies. The temperature was again low beginning in early February, followed by a warming trend which lasted until at least the end of the data record. The temperature variability also shifted toward more energetic higher frequencies during this time. By mid-February the stream had returned into its weakly deflected state, and the variability at array E was once again due primarily to meander/frontal eddy events.

Visual comparisons among the time series from the various E instruments are quite good. We have computed the squared coherence (K 2) and phase between several instnunent pairs, typical examples of which are shown in Figures 6 and

E2T

Period

10 • _•50 2010 5 2 1 I I I I I -

=-"'

' %t %

: U

10 ø ........ f , ....... 10-' 10-' 10 ø

Frequency (cltcZ•s/sec)

7. The uncertainty of the phase estimates is roughly 17 ø (10 ø) at a coherence of 0.6 (0.8). Notice that K '2 is Fig. 5. Record-length spectrum for the u and v 40 HRLP time series

at E2T. The v spectrum decays with roughly a -2 power law, and significant and quite high in essentially all variables down spectral values from the u time series are about 2 orders of to periods of about 2.5 days. The phase between E1T and magnitude less than those for v.

6702 BAN•E AND DEWAR: GULF STREAM BIMODALITY AND VARIABILITY

(a) PERIOD (DAYS)

O0 50 20 10 5 2 I 100 50 20 10 5 2

i i i i • i i i i

i i i i

_

ß

95•.

[ 10 -z

I I I ! IIIIJ I t ! I •l

10 -t

i i t ! trill I I I ! ltt•

i I I I

O0 ,50 20 10 5 2 l

, , • , , 180

9O

'------"" -v/"" 0

-90

, , ,,,,,,t , , ,,,,, -180

- 0.8

0.6 0.4

L 95% 95%

- f 0.2 • , , , , ,,,,,• , , , *,,,, 0

0 ø 10 -z 10 -t 10 ø 10 -a 10 -t LO ø FREQUENCY (CYCLES/DAY)

Fig. 6. Sample coherence and phase spectra from the E array. Shown here are the spectra from the E1T-E2T pair of instruments for (a) cross-stream velocity component u, (b) downstream velocity component v, and (c) temperature T.

bottom boundary layer phenomenon, the effects of which Array F are not reaching to E1T. In contrast, E2B is 130 m away Array F was positioned downstream of the Charleston from the bottom over the 400-m isobath, and it may not be bump in order to measure the subsurface variations measuring similar bottom phenomena. It is interesting that associated with the deflection process. The location of this the 10.7- to 4-day band is more coherent between E1T and array was selected so that the Gulf Stream would move E2T than between E1T and E1B. laterally away from the moorings during a strong seaward

(,,) (b) (o) PERIOD (DAYS)

100 50 20 lO 5 2 i i i

,

-

_ ! ! ! I !ill! ! i I ! lilt

i i i i i 95•.

f i ! I i IlllJ ! f I I 10 -z

100 50 20 1o 5 2 i i i i

i I i I 1111J I I I ! It ,

i ! i I i

95•. 95•.

, 10 -t 10 ø 10 -z 10 -t 10 ø 10 -z

FREQUENCY (CYCLES/DAY)

O0 50 20 10 5 2 1 1 i i i i

! i i ! 1111J i t ! t lilt

, •

10-'

180

90 U'• .<

0

0

-90 r•

-180

0.8 z

0.6

0.4

' ' 1

0.2 D

I I I ! Ill 0

lO ø

Fig. 7. Same as Figure 6, but for the E2T-E3T instrument pair.

BANE ANI) DEWAR: GULF STREAM BIMODAL1TY AND VARIABILITY 6703

150

tO0

o -5O

-tOO

• 40 HLP I F3B mean, the water was generally warmer than the Days September-October period, and the variability was

50 too 150 200 250 dominated by fluctuations with periods near 16 days. The ß , ""• ' . i .... i .... i .... i ....

ß transition back to the weakly deflected state is suggested best by the F3B temperature data. During January the water

2O

15

I , I I I I I , I I

SEP OCT NOV DEC JAN FEB MAR APR 1981 1982

37

g'• 36

35

34

became quite cold, and the peak-to-peak temperature variations decreased considerably. The large perturbations in v, T, and S observed during late January and the first few days of February were due to the final fluctuation of the stream while in its strongly deflected state.

The u and v spectra from F3B (Figure 9) are consistent with these transitions and their associated changes in the fluctuation character of the stream. For example, the v specmnu is red, with roughly a -1.5 to -2 power law, and alludes to the presence of energetic lower frequencies. The u specmnu is indistinguishable from a white specmnu out to the 5-day period. The 32-day band is significantly more energetic in the downstream flow than in the cross-stream flow; however, the energy in the remaining bands is roughly equally partitioned between u and v.

Examples of the record-long coherence and phase estimates for cross-stream and downstream velocities and

temperatures measured within array F are shown in Figure 10. The coherences in temperature between F2B and F3B are representative of the intra-array coherences at F. The coherence squared is nearly unity at a lag of 1 day over the 32- to 8-day band, and the phase is flat and nearly zero. This is consistent with a downstream propagation speed of 25 km/day. Visual comparisons of individual warm events in the two records during the strongly deflected period also

Fig. 8. The 40 HRLP t/me series of downstream velocity show a lag of about 1 day for the best comparison. The component v, cross-stream velocity component u, temperature T, and significance of the coherences becomes less certain beyond salinity S measured at F3B (mean instrument depth of 272 m). The the 8-day band, and the phase there shows no general trend. transition from the weakly to the strongly deflected state occurred This suggests that the higher-frequency fluctuations at array in late October at this site. While in the strongly deflected state, the signatures of several DAMEX biweekly meanders were recorded F have a coherence length scale shorter than 20 km. Dewar here, most notably near November 6, November 21, December 11, and Bane [1985] have commented on the effect of the and January 31. The transition back to the weakly deflected state Charleston bump on the character of the eddy ellipses at F. occurred during early February, when fluctuations shifted to higher The present calculations suggest that the higher frequencies frequencies than were observed during the strongly deflected state.

are sensitive to the bump topography.

deflection, whereas the stream would flow through the array when it was traveling more directly alongshore.

The u, v, T, and S data from F3B show several interesting features (Figures 2 and 8). First, the mean velocity is directed onshore at an angle of about 20 ø with respect to the local isobaths. Note that the F1 and F2 mean velocities

shown in Figure 2 are directed offshore at a comparable angle. Second, the cross-stream velocity component fluctuations have magnitudes comparable to the downstream fluctuations, in contrast to the observations at array E. Third, the transitions from the weakly deflected state to the strongly deflected state (late October at array F) and back again (late January) are quite evident in the time series. And last, several events were observed at this site during the strongly deflected period that gave large Eulerian u, v, T, and S fluctuation signatures.

Prior to the late October transition to the strongly deflected state, the variability at F3B was dominated by fluctuations with periods of about 4 to 8 days. Then, near

F3B

Period

zoto z • I i I I I - 95•, I

-

•-V -- U

,

- 1 10 ø ....... ,, ..... ,, 10 -• 10-' 10 ø

Prequency

the end of October a large amplitude event began at array F heralding a change in all variables. For the next 3 •'2 Fig. 9. Record-length spectrum for the u and v 40 HRLP time series at F3B. The u and v spectra have comparable magnitudes over most months, with the stream in its strongly deflected state, the of the frequency range, the exception being the lowest resolved downstream flow at F fluctuated slowly about a near-zero bands.

6704 BANE AND DEWAR: GULF STREAM BIMODALITY ANT) VARIABILITY

(a)

100 50 20 !0 5 2

i i I I lilt! i i i t lit

(b) PERIOD (DAYS)

100 50 20 10 5 2 1

t i i i i ' i

! I t t ,Ittl I I I I It

oo 50 20 10 5 2 ! --i i i i i

_

.

I I I I Illll I I I I lilt

180

90 U•

0

-90 •

-180

95•. 9,5. /• 9,5

f ........ ,,'Y ] i i i Ill I I I I trill I I II! I I I ! IIIll ! II 0

i0 -I 10 -t lO ø 10 -z i0 -s 100 10 -z 10 -I 10 ø FREQUENCY (CYCLES/DAY)

Fig. 10. Sample coherence and phase spectra from the F array. Shown here are the spectra from the F2B-F3B pair of instruments for (a) cross-stream velocity component u, (b) downstream velocity component v, and (c) temperature T. Cl'he F2B time series was lagged 1 day prior to spectral computations.)

During the period of strong deflection, at least four events velocity structure is in map perspective with the direction of were observed at F3B which were characterized by large propagation to the right and the coastline toward the top of fluctuations in all variables (Figure 8). They occurred on the figure. The anticyclonic flow within the warm meander about November 6, November 21, December 11, and crest is indicated in this diagram, with an offshore flow in January 31. It is possible that two other similar events were the leading portion of the crest and an onshore flow in the recorded on October 23 and December 26, although their trailing portion. This meander is folded enough that the signatures are not as clear as those of the other four. The flow in the trailing portion of the crest has an "upstream" highest temperatures (--14--18øC) and salinities (-35.9-36.6 component (to the left in this figure), indicative of the ppt) observed at F3B occurred during these events, as did strength of the anticyclonic circulation. The large some of the highest peak-to-peak velocity fluctuations amplitudes of the DAMEX biweekly meanders that result in (-75-100 cm/s). Periods between events averaged about 16 this circulation likely result from stronger amplification days, and coherence and lag estimates indicate that they processes occurring downstream of the Charleston bump progressed to the northeast at about 25 km/day. These during a strongly deflected condition than during a weakly events were the DAMEX biweekly meanders progressing deflected condition. past array F.

We were fortunate to obtain a clear satellite image of the SST field which shows the surface signature of one of the Mooring G biweekly meanders departing the array F area on November This single mooring was located 180 km downstream of 7 (Figure 11). Another folded back meander crest may be array F and 360 km downstream of array E on the 400-m seen off Onslow Bay. It is the fluctuation which passed F3B isobath. It was placed off Onslow Bay near the site of the about October 23, and it is probably a biweekly meander earlier Gulf Stream Meanders Experiment; thus data from also, although its signature at F3B was not as clear as the mooring G may be viewed in the context of a better others. The folded back form of this type of meander, along historical data base than those data from either array E or with its large amplitude, differentiates it from the smaller array F. meander/frontal eddy features that occur in a weakly deflected Because of instrument malfunctions we obtained reliable Gulf Stream. Eulerian velocity and temperature signatures measurements of speed at only GT (210 m), and of direction, from these meanders indicate a strong anticyclonic T, and S only at GB (270 m). Although significant vertical circulation throughout the meander crest. An example is shears exist in the Gulf Stream, it has been observed in this shown in Figure 12, which shows the velocity vectors and area that the flow directions usually change little with temperatures from the biweekly meander that passed F3B depth. In particular, Brooks and Bane [1981, 1983] and during late November (compare Figure 8). The velocity Bane et al. [1981] have demonstrated the vertical coherence sticks are plotted with the along-isobath direction to the of the velocity fluctuations at this location. Using this fight, and the time axis has been reversed. This means that information, and seeing favorable comparisons between GT because of the downstream propagation of the meander, the and GB directions where possible (the GT direction sensor

BANE AND DEWAR: GULF STREAM BEVIODALI• AND VARIAB• 6705

Fig. 11. AVHRR image from November 8, 1981, showing a DAMEX biweekly meander just departing the array F region. The fluctuation signatures in velocity, temperature, and salinity from this meander may be seen spanning most of the first part of November in the F3B time series in Figure 8. The folded back meander crest off Onslow Bay is the fluctuation which passed F3B about October 23, and is probably a biweekly meander also. (Image courtesy of R. Legeckis, NOAA.)

worked intermittently), we constructed a "hybrid" data set also confirmed for short periods during which the GT using the GT speed and the GB direction. The resulting time sensors provided reliable data.) series is expected to be a reasonable estimate of the 210-m Differences between the fluctuations that occurred at G flow because of the directional coherence, but not of the during the weakly and strongly deflected states are readily 270-m flow because of the vertical shear. These u and v time apparent in the GTH data. In particular, the early and late series are labeled GTH (G top hybrid) and are shown, along segments of the time series are dominated by with the GB T and S time series in Figure 13. Because of the meander/frontal eddy events with periods of about 4 to 8 known vertical coherence in temperature and salinity days, similar to both the E2T and F3B records (Figures 4 fluctuations at this site [Brooks and Bane, 1981], the 270-m and 8). Typical of these motions at this location are high level T and S time series may be considered representative coherences between v, T, and S; u and v being almost in of the 210-m level fluctuations in those quantities. (This is quadrature, with u leading v; and periods in the 4- to 8-day

6706 BANE AND DEWAR: GULF STREAM BIMODALITY A.N•) VARIABE.ITY

75-

5O

25

-25

-50

Meander Propagation

18

Ch 14

10

3'0 ' 25 2•0 November 1981

F3-B

and the DAMEX biweekly meanders during the strongly deflected state. In this subsection we will discuss the

connections between various events at E, F and G. It will be shown that the coherences between various mooring pairs depend upon the frequency band of interest. The record-long coherence spectra between all inter-array instrument pairs show there is little of significance in any of the resolved frequency bands, and the corresponding phase spectra do not indicate any systematic trends. Thus when spectra are averaged over the length of the experiment, it appears that there is little connection between fluctuations at the various

arrays. This is due in part to the changes in the deflection state of the stream during DAMEX. During periods when some processes were coherent between two arrays, other motions (which may have been coherent at other times) are

temporarily not. We have therefore examined the coherence and phase information for the approximate times of weak and strong Gulf Stream deflection.

Spectra were calculated for 64-day-long periods during the Hg. 12. Time series from the F3B current meter during late early and late portions of the time series in order to November showing the velocity and temperature signatures left by investigate the weakly deflected state. Only slight the passage of the warm crest of a biweekly meander. The vector differences existed between the sets of spectra from these sticks are oriented so that the northward, along-isobath direction is to the right, and time increases on the horizontal axis toward the left. Because of the downstream propagation of the meander, the velocity structure is in map perspective with the direction of propagation to the right and the coastline toward the top of the figure. The anticyclonic flow within the meander crest is evident, with an offshore flow in the leading portion of the crest and an onshore flow in the trailing portion. This meander was folded -•, 100 enough that the flow in the crest's trailing portion had an upstream

component, indicative of the strength of the anticyclonic • 50 circulation. • 0 ;> -50

band [Bane et al., 1981; Brooks and Bane, 1981, 1983]. All -too of these properties may be seen in the GTH records. By

refering to Figures 3 and 13, one may see that a meander crest was just passing mooring G on March 2, and its signature with high v, T, and S values was recorded in the time series then. • -75

During the central period, the time of strong deflection, the variability is due to the passage of the DAMEX •o biweekly meanders past G. Each of the biweekly meanders that left an identifiable signature at F3B may also be •.. discerned in the GTH time series, as will be discussed below.

1o

One important difference between the biweekly meander signatures at F and G is in the ratio of u variance to v variance. While this ratio was about 1 in the F3B time 38 series, it is much less than 1 for the GTH time series. This 37 suggests that a structural change occurred in the meanders as they progressed from F to G, and they became more • a6 elongated in the downstream direction. This evolution may ca be driven by energy fluxes from the meanders to the mean current, since the flow along this stretch the continental 34 margin is known to support such energy conversions [Webster, 1961b, 1965; Oort, 1964; Hood and Bane, 1983].

,o ,,., I GTH/GB Days

0 50 t00 is0 200 250

i i i I i i I i

ß ß ß ß I .... i .... i .... I ....

I I I I I I , I I

SEP OCT NOV DEC JAN FEB MAR APR 1981 1982

The temperature and salinitiy variances have comparable Fig. 13. The 40 HRLP time series of downstream velocity magnitudes between the two locations, although the component v, cross-stream velocity component u, temperature T, and durations of high T and S dtaing the passage of a given salinity S measured at the G mooring. (The velocity time series are meander are usually longer at G than at F, consistent with constructed from the speed time series at GT and the direction time the structural changes just mentioned. series at GB to give a hybrid time series GTH, a procedure necessitated by instrument malfunctions (see text). GTH represents

velocity at 210-m mean instrument depth.) The T and S time series Inter-Array Coherences are from GB (270-m mean depth). The transition from the weakly to

the strongly deflected state occurred near November 10 at this site. It is clear from the foregoing discussion that there is The signatures of the DAMEX biweekly meanders were recorded here

evidence for downstream propagation of beth also. The transition back to the weakly deflected state occurred meander/frontal eddy events during the weakly deflected state during mid-February.

BAN•E AN•D DEWAR: GULF STREAM BLVIODALWY A•X,I) VARIABII.rI"f 6707

two periods. During the first 64 days the squared coherence Cape Hatteras. We use observations from studies other than between E2 and G is modest (4).5) in the 2- to 5-day band DAMEX to show the existence of the strongly deflected at a lag of 9 days, corresponding to downstream state during other time periods, and we consider mechanisms propagation at 40 km/day. This, together with an energetic for initiating the transition from a weakly deflected to a peak in that same band in the E2T spectra (Figure 5) suggest strongly deflected state. that the 2- to 5-day period meanders which passed array E Other Observations probably made it to mooring G; however, only 50% of the

It has been known for some time that the Gulf Stream energy in this band observed at G is statistically related to

exhibits interesting behavior in the Charleston bump region that at E2. The spectra also showed that the energetic 8-day variability at G is incoherent with the 8-day variability at [Singer et al., 1983] (see Brooks [1986] for historical E2. The situation is differera for the F2-G coherences in that background), but the bimodal nature of its path and the the low-frequency bands including the energetic 8-day band effect that this bimodality has on Gulf Stream variability is are significantly coherent between the two locations at a lag only now becoming apparent. Significant attention was first of 4 days. Some coherence values are even quite striking; focused on the deflection process during the mid-1970's for example, K 2 for the F2-G u time series pair at a period of when satellite imagery repeatedly showed its occurrence 8 days is 0.9. At higher frequencies the coherences decline [Pashinski and Maul, 1973; Brooks and Bane, 1978; to marginal values. The v coherences between E and F at a Pietrafesa et al., 1978]. Legeckis [1979] was the first to lag of 4 days are significant in the 16-day band, marginally attempt to categorize the different states of the Gulf Stream's significant in the 8-day band, and not significant in the 2- path in the South Atlantic Bight using satellite imagery. He to 5-day band. No significant coherences between cross- suggested a classification scheme with five types of patterns stream flows at E and F were found. in the stream as follows: type I, deflection; type II,

The 64-day interval from late in the mooring period deflection with cyclonic rotation; type m, deflection showed coherences which are similar to those just discussed. moving downstream; type IV, stable waves; and type V, One difference is that F2 and G are coherent in all bands out unstable waves. From his presentation it is clear that the to 2 days, rather than only in the 8-day band. Connections first three types in this classification are examples of the between E2 and G in the 2- to 5-day band are established, strongly deflected Gulf Stream, the first type with no and the coherences between E2 and F2 in that band are again meanders and little surface temperature manifestation of the weak. cyclonic flow in the Charleston gyre, the second again with

A 64-day-long period was also selected from the middle of no meanders but with a warm "streamer" of water flowing the experiment to study the strongly deflected state. The southwestward along the shoreward flank of the Charleston variability during this interval was dominated by the gyre, and the third a case with a meander pattern similar to DAMEX biweekly meanders, and as was mentioned above, the DAMEX biweekly meanders moving downstream. His there is good evidence that these fluctuations propagated last two types are examples of trains of meander/frontal from F to G. There is weaker evidence that they may have eddy features propagating away from the Charleston bump propagated from E to F. The spectra from F and G during a period of weak Gulf Stream deflection. Type V

waves have frontal eddies with circulations strong enough to temperature time series pairs show significant coherences at a 6-day lag, and the phases are essentially zero. Visual form warm filaments along their shoreward edges, while the coherence between the time series (Figures 8 and 13) at a type IV waves do not have accompanying warm filaments. lag of 6 days is striking. The events match well in number Legeckis [1979] points out that type IV and type V waves and magnitude, and the lag suggests a 30-km/day appeared most often in the imagery which he analyzed. propagation speed, consistent with the 25-km/day speed Bane [1983] has presented a kinematical argument which measured at array F. The statistics suggesting downstream suggests that the sharp tuming of the stream at the propagation from the velocity time series are not as strong Charleston bump during the weakly deflected state is due to as those from the temperature data. Coherence in the amplification of meander/fontal eddies. The main body downstream flow is marginal in most bands, although in the of the Gulf Stream does not proceed very far seaward in this 16-day band it is significant. The scatter in /(2 is case, consistent with the "600-m isobath crossing" criterion

discussed above, but the stream's shoreward SST front does uncomfortably high and the phase tends to vary considerably with small changes in lag. Visual coherence become oriented east-west along the leading edge of a between events is good when coincidence of fast meander, thereby giving the impression of a seaward downstream flow is considered. Recall that the comparisons deflection of the stream in a satellite image (see our Figure between F and G time series discussed above suggested 3). No discussion was given in that paper on the strongly meander decay as these events progressed from F to G. Such deflected state; however, several views of the subsurface evolution in the flow field would tend to lower the temperature field in the Charleston-to-Hatteras region were coherences as computed here. presented. It is obvious from those and more recent sections

Evidence for the propagation of the DAMEX biweekly that there usually exists a dome or ridge of cold water to the northeast of the Charleston bump, between the shelf break meanders from E to F comes from the temperature and the main body of the Gulf Stream. Bane [1983] called coherences. The /(2 value at a 6-day lag between E2T and this the "cold dome," and Singer et al. [1983] presented F3B is significant. Again, velocity coherences are spotty.

Visual comparisons between temperature time series (Figures statistics for isotherm doming in this area using historical 4 and 8), while suggestive, are not as striking as those data. The smacture of this feature varies with the deflection between F and G. state of the stream. It is very much like a ridge oriented

roughly along the isobaths during the weakly deflected BIMODAL1TY IN THE PATH OF THE GULF STREAM state, while during times of strong deflection the cold water

In this section we discuss the notion of bimodality in the forms a stronger dome, which is apparently the central Gulf Strearn's path in the region between Charleston and portion of the spun-upCharleston gyre.

6708 BANE AND DEwAm GL,T.F STREAM BLMODAL.rD'

Using 5 years of satellite data, Olson et al. [1983] location: downstream velocity from 25 to 130 cm/s; showed that the position of the Gulf Stream's shoreward SST temperature from 8 ø to 16øC, and salinity from 35.0 to 36.1 front in the deflection region undergoes a seasonal ppt (Figure 4). Changes in the F3B and GTH time series oscillation. Their 60-day low-passed time series of frontal (Figures 8 and 13) associated with this transition in position along a transect oriented roughly perpendicular to deflection state began near October 21 (F3B)and November shore just downstream of the Charleston bump shows that 10 (GTH). These times of occurrence imply that the the stream was farther offshore in the fall and winter transition progressed in the northeastward direction at about seasons of all years but one from 1976 through 1980, 10 to 20 km/day, a property consistent with Leds [1984] consistent with the stream's being in a strongly deflected interpretation of a long-period traveling wave (see previous state for significant durations during those seasons. The section). histogram of the stream's SST frontal location along the There exist several candidate mechanisms for causing this transect indicates a bimodal distribution, presumably transition. Two which will be discussed briefly here are (1) because of the existence of both deflection states during the Gulf Stream path and transport changes at the Charleston 5-year period of their study. bump and (2) interactions between the stream and cold core

Observations from other studies suggest the existence of a Gulf Stream rings which populate the open ocean waters strongly deflected Gulf Stream during winter months. During seaward of the deflection region. A third possibility is the 1979 Gulf Stream Meanders Experiment the stream discussed by Lee et al. (1988), who have suggested that the became strongly deflected during mid-February. Bane et al. transition they observed during GALE was a result of [1981], Brooks and Bane [1981], and Hood and Bane [1983] coupling between a growing offshore meander of the stream have presented various aspects of that period of strong and an enlarging cyclonic frontal eddy just downstream from deflection, which lasted for about 1 month and ended as a the bump. strong meander progressed from the deflection region past The possibility of changes in the path or transport of the their mooring B (which was in the same location as the Gulf Stream as it approaches the Charleston bump is an DAMEX G mooring). That meander had characteristics intriguing one, since a similar explanation seems plausible similar to those of the DAMEX biweekly meanders and was for the changes from one state of deflection to the other in likely one of that type. Lee et al. ["Response of South the bimodal Kuroshio meander off Japan. Chao [1984] has Carolina shelf waters to wind and Gulf Stream forcing during shown with the use of a barotropic model that the winter of 1986", submitted to Journal of Geophysical bimodality of the Kuroshio's path is an example of multiple Research, 1988, hereinafter referred to as Lee et al. (1988)] equilibrium states. For small volume transports (<30 Sv), have used current meter data and satellite imagery from the only the small meander state exists, while larger transports 1986 Genesis of Atlantic Lows Experiment (GALE) to show (between 30 and 60 Sv) allow the possibility of either that the Gulf Stream underwent a transition from an meander state. The transition to the large meander state "onshore mode" to an "offshore, deflected mode" during occurs as a consequence of ocean spin-down, during which January, following the passage of large amplitude frontal the decreasing transport of the Kuroshio interacts with the eddy/meander events. Kyushu wedge, a bottom topographic perturbation off

Lee [1984] has analyzed current meter and temperature data southern Japan. Large transport changes in the Gulf Stream collected during the Blake Plateau Study, which was within the Florida Straits have been documented to have conducted during 1983 and 1984 in the same general region periods from a few days to at least a year [Lee et al., 1985]; as was DAMEX. His monthly mean temperature and current however, no comparable measurements have been made near maps for the 200-m level indicate that the Gulf Stream was the Charleston bump. Similar transport changes are in a strongly deflected state for the Febmary-April period of conceivable near the bump, though, and as such might be 1983, then moved shoreward for at least the next 3 months. responsible for the deflection state changes through Using additional statistical information, he suggests that interactions with the bottom topography. The decreasing the offshore position of the stream during that time period transport during the fall as part of the annual transport cycle was due to downstream propagating waves with periods near would suggest that time of year for a transition to the 40 days and phase speeds of about 23 km/day. The Blake strongly deflected state if Chao's [1984] Kuroshio Plateau Study data also suggest that periods of strong mechanism were applicable. Other, more rapid transport deflection may exist during seasons other than fall and changes due to local wind effects or meanders passing the winter, but these suggestions are not as convincing as the bump may be important as well. Further measurements would fall-winter observations from other studies. Difficulties in be necessary to test the transport hypothesis. obtaining good estimates of Gulf Stream SST frontal Another candidate mechanism is the interaction between locations during the summer months contribute to the fewer the Gulf Stream and cold core Gulf Stream tings. observations of summertime Gulf Stream deflection, and Observations have shown that there can be significant with these fewer summer observations it is not possible to distortions in the stream's path caused by ring-stream confirm that the strongly deflected state occurs interactions [Richardson, 1983], so it seems possible that preferentially during the fall and winter. the cold core rings which drift as far south as the deflection

region might contact the stream's seaward boundary, thereby The Transition to the Strongly changing its path into a strongly deflected configuration. Deflected State This is not an appealing mechanism for the following

DAMEX has shown that the transition from the weakly reasons. First, the location of the deflection is very limited, deflected state to the strongly deflected state can be quite occurring usually within 20 to 30 minutes of latitude of rapid. The changes at E2T which began on October 11 took 32øN [Brooks and Bane, 1978; Pietrafesa et al., 1978]. If only 2 days and resulted in the following increases at that rings were responsible for the deflection, then a more

BANE AND DEWAR: GLn_F STRSAM BE•IODALITY A•\•D ¾AKIABILrI%' 6709

uniform alongshore distribution of deflection position would offshore meander of the stream and an enlarging cyclonic be expected. Second, there should be no seasonal preference frontal eddy downstream of the bump may induce such a for ring-stream interactions, yet the evidence to date transition. It remains unclear as to which of these three is suggests that the strongly deflected state occurs the responsible mechanism, or if there is more than one preferentially during the fall and winter (although additional cause. summertime data are needed to verify or refute this). During The bimodal character of the stream's path affects the low- DAMEX, however, there were indications in the satellite frequency variability of the current between the Charleston imagery of a cold core ring transiting southward along the bump and Cape Hatteras in the following manner. Typical seaward edge of the stream on a trajectory such that it may meander/frontal eddy patterns dominate the mesoscale have interacted with the stream near the bump and changed fluctuation field during times of weak deflection, while its path. This ring was first clearly visible in a November larger-amplitude meanders with very sharp curvature and 15 image, and it was located just to the southeast of the strong anticyclonic circulations in their crests (fluctuations stream off Charleston. The December 3 image in Plate 1 which we refer to herein as the DAMEX biweekly meanders) shows this ring centered at about 30øN, 76.5øW on that exist during times of strong deflection. Furthermore, the date, approximately 200 km south of the deflection region. DAMEX biweekly meanders have generally slower phase

The DAMEX results have shown that the transition from a propagation speeds than do the typical meander/frontal eddy weakly deflected to a strongly deflected configuration may events (about 20-25 km/day compared with 35-60 km/day) occur rapidly. There is no unambiguous answer at this point as well as longer periods (about 16 days compared with 1 as to the cause of this transition. week).

SUMMARY Acknowledgments. This work was sponsored by the Office of Naval Research through contracts N00014-77-C-0354 and N00014-

Gulf Stream mean flow and variability have been studied 87-K-0233 to the University of North Carolina (LINC)and N00014- in the Gulf Stream Deflection and Meander Energetics 87-G-0209 to the Florida State University. Special thanks go to R.

Ault, K. Osgood, and J. Schultz for their assistance with several of Experiment (DAMEX). This study was conducted between the data processing tasks. We are grateful to P. Blankinship for his September 1981 and April 1982 in the area of the preparation and deployment of the DAMEX instruments and J. Charleston bump, off the coast of the southeastern United Woods for his efforts on the recovery cruise. The assistance given States. Using data from an array of current meters and by the crews of the research vessels Researcher (NOAA) and Cape satellite imagery, it has been shown that the seaward Hatteras (UNC/Duke University) is greatly appreciated. Satellite imagery was kindly supplied by O. Brown and R. Evans (University deflection of the Gulf Stream by the bump has a bimodal of Miami)and R. Legeckis (NOAA). character and that the Gulf Stream meander field downstream

of the bump differs between the two states of deflection. The two deflection conditions have been termed the weakly deflected state and the strongly deflected state in this paper. A weakly deflected Gulf Stream flows past the Charleston bump and close enough to the edge of the continental shelf that its shoreward SST front remains inshore of the 600-m

REFERENCF•

Bane, J. M., Jr., Initial observations of the subsurface structure and short-term variability of the seaward deflection of the Gulf Stream off Charleston, South Carolina, J. Geophys. Res., 88(C8), 4673-4684, 1983.

isobath between Charleston and Cape Hatteras. A strongly Bane, J. M., Jr., and D. A. Brooks, Gulf Stream meanders along the deflected Gulf Stream tums sharply eastward at the bump and continental margin from the Florida Straits to Cape Hatteras,

Geophys. Res. Lett., 6, 280-282, 1979. flows seaward far enough to have its shoreward SST front Bane, J. M., Jr., and W. K. Dewar, The Deflection and Meander offshore of the 600-m isobath for several tens of kilometers Energetics Experiment, Current meter and bottom pressure gauge along the continental slope downstream of the bump. The data report, Tech. Rep. CMS-83-2, Univ. of N. C., Chapel Hill, transitions between deflection states may occur very rapidly, 1983. and once in a strongly deflected state the Gulf Stream may Bane, J. M., Jr., D. A. Brooks, and K. R. Lorenson, Synoptic observations of the three-dimensional structure and propagation remain so for up to several months. Two transitions were of Gulf Stream meanders along the Carolina continental margin, observed during DAMEX, one from a weakly to a strongly J. Geophys. Res., 86(C7), 6411-6425, 1981. deflected state and one in the other direction. The transition Brooks, D. A., Doing the Charleston Bump, Eos Trans. AGU, 67(1),

from weakly to strongly deflected state took only a few days 4-5, 1986. Brooks, D. A., and J. M. Bane, Jr., Gulf Stream deflection by a at a given site, while the signatures of the other transition bottom feature off Charleston, SC, Science, 201, 1225-1226, were not as definite in the data records. The strongly 1978. deflected state occurred during the fall and winter in DAMEX Brooks, D. A., and J. M. Bane, Jr., Gulf Stream fluctuations and (October through January), and satellite observations from meanders over the Onslow Bay upper continental slope, J. Phys. earlier years suggest that it is this time of year in which the Oceanogr., 11, 247-256, 1981. Brooks, D. A., and J. M. Bane, Jr., Gulf Stream meanders off North strong deflection state preferentially occurs. More recent Carolina during winter and summer, 1979, J. Geophys. Res., current measurements in this region show that a strongly 88(C8), 4633-4650, 1983. deflected state may occur occasionally during other times of Brooks, D. A., J. M. Bane, Jr., R. L. Cohen, and P. Blankinship, the year also. It is not known how the transition from a The Gulf Stream Meanders Experiment: Current meter and

atmospheric data report for the August to November 1979 weakly deflected to a strongly deflected state may be mooring period, Rep. 81-3-T, Tex. A & M Univ., College induced; however, two candidate mechanisms discussed here Station, 1981. are as follows: (1) transport or path changes in the Gulf Chao, S. Y., Bimodality of the Kuroshio, J. Phys. Oceanogr., 14, Stream itself, which induce changes in the topographic 92-103, 1984.

Dewar, W. K., and J. M. Bane, Jr., The subsurface energetics of the interactions at the Charleston bump, and (2) interactions Gulf Stream near the Charleston bump, J. Phys. Oceanogr., 15, between the stream and cold core Gulf Stream rings. Lee et 1771-1789, 1985. al. (1988) have suggested that coupling between a growing Hood, C. A., and J. M. Bane, Jr., Subsurface energetics of the Gulf

6710 BANE AND DEWAR: GULF STREAM BLMODALITY AND VARIABnaTY

Stream cyclonic frontal zone off Onslow Bay, North Carolina, J. Pashinski, D., and G. A. Maul, Use of ocean temperature while Geophys. Res., 88(C8), 4651-4662, 1983. coasting between the Straits of Florida and Cape Hatteras, Mar.

Lee, T. N., Blake Plateau Current Measurements Study, Rep. 2, Gen. Weather Log, 17, 1-3, 1973. Oceanics Inc., Miami, Fla., 1984. Pietrafesa, L. J., and G. S. Janowitz, On the dynamics of the Gulf

Lee, T. N., and L. P. Atldnson, Low-frequency current and Stream front in the Carolina Capes, in Stratified Flows: The temperature variability from Gulf Stream frontal eddies and Second International Symposium on Strat•ed Flows, pp. atmospheric forcing along the southeast U.S. outer continental 184-197, Tapir, New York, 1980. shelf, J. Geophys. Res., 88(C8), 4541-4567, 1983. Pietrafesa, L. J., L. P. Atldnson, and J. O. Blanton, Evidence for

Lee, T. N., L. P. Atkinson, and R. Legeckis, Observations of a Gulf deflection of the Gulf Stream by the Charleston Rise, Gulf Stream frontal eddy on the Georgia continental shelf, April 1977, Stream, 4(9), 3-7, 1978. Deep Sea Res., Part A, 28(4), 347-378, 1981. Richardson, P. L., Gulf Stream rings, in Eddies in Marine Science,

Lee, T. N., F. Schott, and R. Zantopp, Florida Current: Low- edited by A. R. Robinson, pp. 19-45, Springer-Verlag, New frequency variability as observed with moored current meters York, 1983. during April 1982 to June 1983, Science, 27(4), 803-820, 1985. Singer, J. J., L. P. Atkinson, J. O. Blanton, and J. A. Yoder, Cape

Legeckis, R. V., The influence of bottom topography on the path Romain and the Charleston bump: Historical and recent of the Gulf Stream at latitude 31 N from NOAA's satellite imagery hydrographic observations, J. Geophys. Res., 88(C8),

Le(ab•tract}, Eo• Trfl•..AGU, J7 260 1976• 4685-4697, 1983. gec•as, t•. v., •atemte om•r•au•is or the influence of bottom yon Arx, W. S., D. F. Bumpus, and W. S. Richardson, On the fine topography on the seaward deflection of the Gulf Stream off structure of the Gulf Stream front, Deep Sea Res., 3, 46-65, Charleston, South Carolina, J. Phys. Oceanogr., 9, 483-497, 1955. 1979. Vukovich, F. M., and B. W. Cdssman, Further studies of a cold

Li, L., M. Wirebush, D. R. Watts, A. J. Brincko, and T. N. Lee, eddy on the eastern side of the Gulf Stream using satellite data Gulf Stream and wind-induced current variability on the Georgia and ship data, J. Phys. Oceanogr., 8, 838-843, 1978. continemal shelf, winter 1974, J. Geophys. Res., 90(C12), Webster, F., A description of Gulf Stream meanders off Ohslow Bay, 3199-3210, 1985. Deep Sea Res., 9, 130-143, 1961a.

Maul, G. A., P. W. deWitt, A. Yanaway, and S. R. Baig, Webster, F., The effect of meanders on the kinetic energy balance Geostationary satellite observations of Gulf Stream meanders: of the Gulf Stream, Tellus, 13, 392-401, 1961b. Infrared measurements and time series analysis, J. Geophys. Res., Webster, F., Measurements of eddy fluxes of momentum in the 83, 6123-6135, 1978. surface layer of the Gulf Stream, Tellus, 17, 239-245, 1965.

Olson, D. B., O. B. Brown, and S. R. Emmerson, Gulf Stream frontal statistics from Florida Straits to Cape Hatteras derived from satellite and historical data, J. Geophys. Res., 88(C8), J. M. Bane, Marine Sciences Program, University of North 4569-4577, 1983. Carolina, Chapel Hill, NC 27599.

Oort, A. H., Computations of eddy heat and density transports W. K. Dewar, Department of Oceanography, Florida State across the Gulf Stream, Tellus, 16, 55-63, 1964. University, Tallahassee, FL 32306.

Osgood, IC E., J. M. Bane, Jr., and W. IC Dewar, Vertical velocities and dynamical balances in Gulf Stream meanders, J. (Received January 18, 1988; Geophys. Res., 92(C12), 13,029-13,040, 1987. accepted February 10, 1988.)

BAA'E Ah•D DEWAR: GULF STREAM BLMODAI./FY AND VAKIABILITY 6963