1. Introduction
We present some results about the North Atlantic Deep Water (NADW) deep western boundary current (DWBC) inferred from moored current meter observations made in the mid–Brazil Basin (Fig. 1). The data were obtained as part of the World Ocean Circulation Experiment (WOCE) Deep Basin Experiment (Hogg et al. 1996), which was conducted in the Brazil Basin and the deep passages connecting this basin to other parts of the Atlantic Ocean. Part of the focus of the Deep Basin Experiment was on the abyssal circulation, including DWBCs, in the Brazil Basin.
The results, independent of the goals of the Deep Basin Experiment, are of interest because the DWBC measured is a direct part of the deep circulation of the great ocean conveyor scheme of Broecker (1991). The transport, as well as the flow direction, of the DWBC presented here should be of interest to those concerned with Broecker’s conveyor belt flow scheme and its role in climate.
In September–October 1993 one of us (GW) set an array of ten current meter moorings with a total of 34 current meters to study the NADW and Antarctic Bottom Water (AABW) DWBCs in the South Atlantic off of Brazil (Fig. 1). The mooring recovery cruise was in March 1995, and those data relating to the DWBC of NADW (hereafter DWBC) are discussed here.
Properties of the deep water masses found in the Brazil Basin are given in DeMadron and Weatherly (1994, hereafter DW). For those not familiar with this basin, NADW is distinguished from the water types that sandwich it, AAIW above about σ2 = 36.70 (depth ∼ 1200 m) and AABW below at about σ4 = 45.87 (depth ∼ 3600 m), as being relatively saline and oxygenated.
The location of the current meter mooring array was essentially along the western portion of the South Atlantic Ventilation Experiment (SAVE) 2 hydrographic section discussed in DW. Figure 2 shows the geostrophic velocity for the SAVE2 section (from DW) with the location of the moorings and current meters superimposed. [Figure 2 assumes, as is often done near the western boundary of the South Atlantic, that the interface between NADW and AABW is a level of no motion and that, if the water is too shallow for AABW, the interface between NADW and the overlying Antarctic Intermediate Water (AAIW) is a level of no motion. DeMadron and Weatherly noted that, if for the latter the bottom is taken as the reference level, essentially the same conclusions arise.] Figure 2 indicates that the DWBC is about 200 km wide and should be delineated from data from moorings 1 through 4. It also indicates recirculation (equatorward flow) of NADW farther offshore out to distances of ∼450 km, which should be revealed by moorings 4–6. DeMadron and Weatherly noted that the recirculating flow is associated with a deep bowllike feature in the density field [previously commented on by McCartney (1993)], which is seen in some sections from the Brazil Basin. DeMadron and Weatherly estimated the DWBC transport for the SAVE2 section at 18°S to be 16 Sv (Sv ≡ 1 × 106 m3 s−1) and found the transport of the recirculating, equatorward flow to be 5 Sv. McCartney (1993), looking at a section farther to the north at 11°S, estimated the DWBC poleward transport to be more than twice as large, 41.8 Sv, and the recirculating, equatorward transport to also be large, 22.2 Sv.
The outline of the remainder of this paper is as follows. After discussing the data, the mean and time-dependent volume transport of the DWBC are estimated. (The sensitivity of the estimates to data voids and data dropouts is examined in the appendix.) There is a summary and discussion section in which some of our conclusions are tempered with recent float observations. Finally, there are some concluding remarks.
2. Data
In the western central Brazil Basin where we placed the current meters, NADW is found approximately between 1200 and 3600 m depths (DW). On the moorings for studying the DWBC, we placed current meters at two levels in the NADW, at 1800-m depth in upper NADW (UNADW), which is approximately between 1200 and 1900 m depths (DW), and at 2800-m depth in lower NADW (LNADW), which is approximately between 2600 and 3600 m depths (DW). These depths were selected because they coincided with maxima seen in a freon section obtained during SAVE2 (W. Smethie Jr. 1993, personal communication). We also placed one current meter in the overlying AAIW at 900-m depth and one in the underlying AABW at 100 m above the bottom. In this study we locate the current meters sometimes by their depths or heights above the bottom and at other times by the water type they were set in.
The record-mean averaged currents are shown in Fig. 3. The current meters in NADW indicate that the DWBC flow is directed approximately to the southeast and not, as expected, approximately to the south (the mooring line was oriented 8.5° counterclockwise from east).
Some information about the current meter records is found in Table 1. [More information can be found in the data report by Harkema and Weatherly (1996), a copy of which is on the web site http://okean.ocean.fsu.edu and which has been submitted to the WOCE Current Meter Archive Center, CMDAC, College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR 97331, USA for general access.] It is apparent from Table 1 that not all the records are of 540-day duration, the approximate length of time the moored array was deployed. For example, all the AAIW records are of 1-yr duration due to a wiring problem. The AAIW records shown in Fig. 3, as well as some others of comparable length, indicate the well-known result that averaged currents estimated from records of duration ≲1 yr are often different from those expected from longer duration records. However, as will be noted later, there is a strong tendency for uniformity of the flow direction in the vertical. For example, the AAIW and AABW records on mooring 4 are both about the same duration and both have roughly the same flow direction (to the east).
The location of the current meters in the vertical plane is again indicated in Fig. 4. In this figure the measured velocity (along 140°T and for period I; see later) is contoured and the bounding isopycnals for NADW are indicated. The DWBC is ∼210 km wide (Fig. 4 indicates a 250-km width, but this was reduced by the sine factor given and explained in the following section) and is embedded in a flow extending downward into AABW to the bottom and upward into the overlying AAIW at least to 900-m depth. From Fig. 4 and Table 1 it is evident that no data were obtained from the AAIW meters on moorings 3 and 7 and no (useful) data from the LNADW meter on mooring 5. That there were no data from the AAIW level (900-m depth) on mooring 1 is due to the AAIW current meters being a relatively late addition to the moorings and that mooring 1 was not robust enough (cf. Harkema and Weatherly 1996) to extend it upward to 900-m depth to add an AAIW current meter.
3. Transport estimates
The intent is to estimate the poleward volume transport of the DWBC rather than its meridional transport as well as to illustrate the velocity structure of DWBC in a plane more normal to the flow direction. We present the velocity components oriented toward 140°T and the associated volume transport factored by sin[(140° − 90°) + 8.5°]. (Recall that 140° is the mean flow direction measured at the UNADW current meter on mooring 3 and is taken as a proxy of the mean DWBC flow direction; 8.5° is the orientation of the mooring section measured counterclockwise from east.) The transports so computed are equivalent to the normal transport across the mooring section.
Before discussing the transport estimates, we point out that our best guess at what the DWBC looks like is found in Fig. 4. While the results were being written, we became aware of some recent data from floats set to drift in AAIW in the Brazil Basin (Boebel et al. 1999). These results suggested that we should have chosen another way of filling in the missing data value at the AAIW level at mooring 3. Sensitivity studies, presented in an appendix, indicate that the effect on the transport of the DWBC is not major; however, the visual impression is rather different. Figure 3 is the velocity section of the averaged flow over the period September 1993–May 1994 when we had maximum data return (this is referred to later as period I) with the missing value at the AAIW level on mooring 3 determined by horizontal interpolation. This gives results more in agreement with the float data than the scheme used in vertical extrapolation to infer the missing value. The velocity core in Fig. 3 is in UNADW, consistent with what Lee et al. (1990, 1996) found at 26.5°N, while the earlier method we used indicated maximum velocities above the NADW in AAIW (see Fig. A1 for an example).
In the following we present transport estimates made when the missing value at the AAIW level on mooring 3 was determined by vertical extrapolation rather than the horizontal interpolation used for Fig. 4. As the average transports so estimated are only about 6% larger than those if we had consistently used the method used for Fig. 4 (see appendix) and the number of transport estimates made before becoming aware of the float results was substantial (e.g., over 500 were made for Fig. 9), we thought it unwarranted to redo all the calculations.
The cumulative poleward volume transport of the DWBC formed from the record-mean velocities shown in Fig. 4 is 36 Sv (Fig. 5). (The transports that we present were estimated using MATLAB. Tests that we ran indicated essentially the same conclusions obtained from hand-contoured velocity plots integrated using a planimeter.) The cumulative transport shown in Fig. 5 is estimated from averages all of which are not of the same duration because of data dropouts (Table 1). Even if the shorter-duration average currents (shorter than the 544-day duration deployment) are reasonable proxies for average flows (Fig. 3 suggests this may be a tenuous assumption), there are indications of a seasonal variability in the currents. If this is the case, the average transport may be somewhat under- or overestimated if the averaging interval is not an integer multiple of a year, as shown by Hall et al. (1997). That there may be some seasonal variability is suggested in some of the current meter time series, for example the stick plots of the daily averaged flows shown in Figs. 6 and 7 (as well as others in Harkema and Weatherly 1996). These figures indicate a tendency for stronger poleward flow in December–March and a tendency for weaker poleward (or stronger equatorward) flow in September–November.
Figure 8 is a daily time series estimate of the DWBC transport. It was formed from the daily averaged Gaussian-filtered currents after further smoothing by four passes with a Hanning filter (weights 1/4, 1/2, 1/4), equivalent to smoothing with an 8-day low-pass filter. Much short-term (period weeks) variability, sometimes as large as ∼100 Sv is evident. A seasonal variability is also suggested with a tendency for stronger poleward transports in February–March and weaker poleward transports in September–October. The seasonal variability is clearly evident in the 6-month running mean (the dashed curve in Fig. 8) where its amplitude is ∼10 Sv. The 12-month running average (the dot–dashed curve in Fig. 8) indicates a poleward mean transport of about 39 Sv.
Similar figures to Fig. 8 but computed for the following water classes, AAIW, UNADW, middle NADW (MNADW), LNADW, and AABW, are shown in Fig. 9. The bounding density surfaces for MNADW were taken as, again following DW, σ2 = 36.95 and σ2 = 37.02. The AAIW transport is only determined from its measured depth of 900 m to its lower bound. The 6-month running mean (dashed) curves in Fig. 9 and the amplitudes of the seasonal variability inferred from them and listed in Table 2 indicate that the seasonal variability is rather uniformly distributed throughout the NADW, rather than concentrated in one or two water types, and that it extends approximately barotropically upward into the AAIW and downward into the AABW. As for the NADW transport in Fig. 8 the 12-month running means for each class shown in Fig. 9 are rather straight. If the seasonal variability were essentially barotropic, then (assuming an ideal dataset) the date the 6- and 12-month running averages curves intersected would be the same for each water type. The approximate date that the 6- and 12-month running mean curves intersect, listed in Table 2, suggests that the variability of each water class may slightly lag that of the one immediately above, with the total lag of the deepest (AABW) seasonal signal relative to that of the shallowest (AAIW) being about 2 months. However, this lagging is more pronounced for the shallower levels where the 12-month running mean curves, while straight, are slightly sloped upward with increasing time. Whether this sloping is real or an artifice resulting from data dropouts, which are more pronounced in the shallowest level (see following section), is unknown.
4. Discussion and summary
a. The mean DWBC
The 12-month running average of the time series volume transport (Fig. 8) is rather steady at about 39 Sv. We think this is a better estimate of the mean poleward transport than the 36 Sv inferred from the record-mean currents (Fig. 5). This is because the duration of the experiment was not an integer multiple of a year (the seasonal variability is appreciable) and some records are shorter than others. Sensitivity estimates (presented in an appendix) indicate that this value is uncertain to about ±10% due to alternative ways that we could have filled in data blanks and chosen bounding depths for NADW, and due to data dropouts. However, the statistical uncertainties of the velocities given in Table 1, due to the 18-month duration of the experiment, introduce a much larger uncertainty of about 40% (16.1 Sv, Table 2). Thus our 39 Sv of poleward transport of NADW by the DWBC is uncertain to about 50% or about ±20 Sv.
The 39 Sv of mean poleward transport of the DWBC, albeit with large uncertainty, is greater than we would expect. (Here the DWBC transport is that between the σ2 = 36.70 and σ4 = 45.87 isopycnals seaward of and to about 300 km off the shelf break.) While it is approximately equal to the equatorward 35 Sv estimated from long-duration current meter records just north of the equator [if we add the 13 Sv of UNADW and MNADW inferred at 0°–3°N, 44°W by Fischer and Schott (1997) to the 22 Sv of LNADW inferred at 8°N, 52°W by Johns et al. (1993)], we would not expect all of this to be available for the DWBC in the Brazil Basin. First, there is evidence both with floats (Richardson and Schmitz 1993) and in the hydrography (e.g., Arhan et al. 1998) that the DWBC splits at the equator with part traveling eastward along the equator and the rest continuing southeastward along the Equatorial Channel into the South Atlantic. Second, whatever fraction makes it across the equator into the Equatorial Channel, not all of it may round the northeastern tip of Brazil at 5°S, 34°W (Cape Saõ Roque) to continue as a deep western boundary current in the Brazil Basin. Part of the NADW entering the Brazil Basin from the Equatorial Channel may continue to flow predominately zonally into the Brazil Basin (DW). It would appear that the DWBC has a transport in the mid–Brazil Basin (where our measurements were made) greater than it has farther to the north at its entry point at Cape Saõ Roque.
We are not the first to measure a NADW transport at a midbasin site much greater than expected. At 26.5°N in the Atlantic, Lee et al. (1990, 1996), using five years of moored current meter data, reported a mean DWBC equatorward transport of about 40 Sv, about three times larger than the estimated 13 Sv of NADW thought to be transiting the North Atlantic (Dickson and Brown 1994). This has resulted in speculation about a deep recirculation farther offshore in the Hatteras Abyssal Plain. A comparable deep offshore recirculation pattern, not captured in our relatively narrow array, appears in order for the Brazil Basin. However, recent float data at depths 2000 m and 4000 m (Hogg and Owens 1999) do not reveal such a recirculation.
The DWBC width, estimated from Fig. 4, is about 210 km (the ∼250 km width seen in Figs. 4 and 5 scaled by sin58.5°; see sec. 3). This is about twice the ∼100 km DWBC width measured by current meter arrays just north of the equator (Fischer and Schott 1997) and at 26.5°N (Lee et al. 1990). Another difference in our DWBC velocity section (Fig. 4) and those in the North Atlantic of Fischer and Scott (1997) and Lee et al. (1990, 1996) is that theirs show the DWBC to be hard against the continental slope while ours show it to be slightly seaward of the slope with a region of near zero, or counter, flow at the continental slope. This is perhaps not surprising because in the Northern Hemisphere both the Coriolis force and the beta effect (the meridional variation in the horizontal Coriolis force) act to push or steer the DWBC to the west [cf. Nof and Dewar (1993) for a discussion of the latter, beta effect], while in the Southern Hemisphere only the latter acts to do so.
The DWBC flowed approximately toward the southeast rather than toward the south. We expected it would flow roughly parallel to the local topography (to the south). The studies of DW and Zangenberg and Siedler (1998) indicate that the Vitoria–Trindade Seamount Chain, 200 km to the south (Figs. 1 and 5), has a pronounced effect on the path and transport of the DWBC. While neither DW nor Zangenberg and Siedler (1998) show the seamount chain influence extending as far north as our mooring line in their schematics of the NADW circulation, given that the current width and its distance from the Vitoria–Trindade Seamount Chain are comparable it would seem likely that this seamount chain does deflect the DWBC seaward.
It is evident from Figs. 3, 4, 8, and A1 that the AABW beneath the DWBC flows poleward. Figure 8 and Table 2 indicate that the poleward volume flux is about 4 Sv. This is comparable in magnitude to, but slightly less than, previous estimates of the transport of the DWBC of AABW at the same approximate latitude [DW estimated 7 Sv at ∼18°S and Zangenberg and Siedler (1998) estimated about 6 Sv at 19°S]. (Here the DWBC of AABW is taken to be the equatorward flow of water denser than σ4 = 45.87 along the lower continental slope.) The latter current is found to the east of the DWBC and its western edge can be seen in Figs. 4 and A1 at distance 400 km. That there may be about 4 Sv of AABW flowing poleward in the Brazil Basin is not surprising since of the 7 Sv of the AABW thought to enter the Brazil Basin from the Vema and Hunter Channels to the south (Hogg et al. 1996), only about 2 Sv exits into the western North Atlantic through the Equatorial Channel (Hall et al. 1997) and only about 1 Sv exits into the eastern Atlantic through the Romanche and Chain Fracture Zones (Mercier and Speer 1998). Thus not all the AABW can flow directly northward in the Brazil Basin to exit out the Equatorial Channel and the Romanche and Chain Fracture Zones. Some must recirculate before being upwelled into the overlying NADW as previously inferred by Speer and Zenk (1993) and DW. What is new here is that some of the recirculation occurs west of the DWBC of AABW.
While Fig. 9 shows that there is a net poleward flow of AAIW above the DWBC of about 8 Sv, Fig. 4 (and Fig. A1) indicate an equatorward flowing current of AAIW just seaward of the shelf break. We made no velocity measurements in this current; however, this feature appeared consistently independent of how we inferred velocities in this region. Strong equatorward AAIW flow shown just seaward of the shelf break is also indicated in the SAVE2 geostrophic sections (Fig. 2), the geostrophic section in Smethie and Weatherly (1998), in the moored current meter records at 20°S of Müller et al. (1998), and in float data of Boebel et al. (1999, their Intermediate Water Boundary Current).
b. The time-dependent flow
The largest fluctuations in the transport time series in Fig. 8 (which have amplitudes comparable to the mean) are those with timescales around a month. We think these fluctuations are due to topographic and/or Rossby waves. As a test we examined auto- and cross-spectra of the north–south velocity records obtained in the UNADW (not shown). All but one of the former showed a pronounced peak in the period range 40–75 day (the exception being the westernmost, shallowest record), and the latter all showed that these motions were coherent (at the 80% to 95% confidence level with one exception) and all had lags consistent with a westward phase propagation of 3–5 cm s−1. We think it likely that these motions are responsible for the horizontal banded structure evident above 3000 m in the SAVE2 geostrophic velocity section (Fig. 2).
We do not think the horizontal scale of the monthly timescale fluctuations were well resolved by our mooring array; the horizontal spacing of the moorings (which ranged from 53 km to 118 km) appears too coarse for the 100–200-km-wide banded structures seen in Fig. 2. We question whether their amplitudes were indeed as large as indicated in Fig. 8.
While the monthly timescale fluctuations in Fig. 8 may be aliased due to the horizontal spacing of the moorings, we think that the velocity section Fig. 4 is not. Averaging over periods longer than the timescale of these fluctuations appears to effectively filter them out. We think this is borne out by the smoothness of the 6- and 12-month running mean curves in Fig. 8.
Figure 8, particularly the 6-month running mean, indicates that the DWBC transport varies seasonally, with it being larger in magnitude around February–March and smaller around September, and that the amplitude of the variability ∼25% (∼10 Sv) of the mean (∼39 Sv). This is similar to what Fischer and Schott (1997) found for LNADW and MNADW just north of the equator in the Equatorial Channel (1°–4°N, 44°W). They found a seasonal variability in the currents with a phase such that maximum flow along the Channel toward the equator occurred in March, with the combined annual and semiannual variability accounting for about 50% of the variance. Hall et al. (1997) reported that the LNADW flow farther to the east in the Equatorial Channel around 36°W was out of phase with, and lagging by about a month, the AABW transport (which they were monitoring) that was maximum (toward the west) in September–October and minimum in February–March. While they did not report the transport of the LNADW, they did report that the amplitude of the seasonal variability of the AABW transport was about 25% of the mean. Thus our NADW seasonal variability is about out of phase with that of the AABW and in phase with that of the LNADW reported by Hall et al. (1997). Mercier and Speer (1998) reported a seasonal variability in LNADW and upper AABW flow through the Romanche and Chain Fracture Zones at the northeastern corner of the Brazil Basin. They found the seasonal contribution to be larger in March–April, enhancing the eastward flow out of the Brazil Basin, and to be in the opposite sense in September–October (K. G. Speer 1998, personal communication). They found the combined effect of the seasonal and biseasonal variability to be about 25% of the mean in amplitude with the biseasonal signal being larger.
Our ∼540 day duration study may be marginally long for detecting a seasonal signature. However, as Fischer and Schott (1997), Hall et al. (1997), and Mercier and Speer (1998) all detected seasonal variations in channels connecting the Brazil Basin to other basins, we think our inferred seasonal modulation is probably real. We think it unlikely that our NADW variability being in phase to within approximately one month to the NADW variability reported in these studies is a coincidence. [We assume here that the seasonal variability reported in Mercier and Speer (1998) in their LNADW/upper AABW is due to the contribution of the former.] Like these studies the amplitude of our seasonal variability was also between about one-fourth and one-half of the mean.
The seasonal variability of transport was nearly uniformly distributed in the vertical in the DWBC and appears due to a barotropic variability not limited to the DWBC (Fig. 9 and Table 2). Lee et al. (1996) also found a seasonal variability of the DWBC transport at 26.5°N, that was part of a barotropic variability. Their variability was found to be due to seasonal changes of the wind stress curl in the subtropical North Atlantic west of the Mid-Ocean Ridge. Their transport was largest (toward the south) when the wind stress curl was less (in the boreal summer), and weakest in the boreal winter when the wind stress curl was largest. Physically, the wind-driven barotropic western boundary current (always oppositely directed to the DWBC) countered the DWBC less when the wind stress curl was less. The amplitude of their variability, like ours, was about 10 Sv.
We suspect that the wind stress is responsible for the seasonal barotropic variability. There is a seasonal variability of the wind stress curl in the vicinity of our moorings (Fig. 10). The wind stress curl is smallest in the austral summer (around December) and largest in the austral winter (around June). Assuming the variability over the period October 1996–July 1997 shown in Fig. 10 is typical, a comparison of Figs. 8 and 10 indicates that the current transport variability lags that of the wind stress by two to three months.
However, based on the arguments presented by Lee et al. (1996) and that the wind stress curl in the Brazil Basin in the vicinity of our moorings is predominately positive [Fig. 10 and Hellerman and Rosenstein (1983, their Fig. 6)], we expected that the times of maximum and minimum transport seen in the 12-month running mean curve in Fig. 8 would be reversed. The times of maximum and minimum transport in Fig. 8 appear to be due to the seasonal changes in the region of the Brazil Basin where the wind stress curl is negative, that is, equatorward of about 15°S (Hellerman and Rosenstein 1983, their Figs. 6 and 9). Perhaps the Vitoria–Trindade Seamount Chain at about 22°S (Figs. 1 and 3) effectively blocks the northward flowing barotropic seasonally driven western boundary current in the region of positive wind stress curl (about 15°S to about 50°S; Hellerman and Rosenstein 1983) from extending northward of it, and hence permits the southward flowing seasonally driven western boundary current in the region of negative wind to extend farther poleward through our current meter array.
The cross-over of the 6- and 12-month running mean curves in Fig. 9 occurred at later times as the depth increased, suggesting a possible vertical phase shift within the seasonal signal with the total lag between the AABW and AAIW being about two months. However, in light of the relative shortness of the AAIW records and that most of this phase shift occurs in AAIW and UNADW (Table 2), this phase shift is suspect. Thierry (1998) reported a vertical phase shift in the seasonal variability over a comparable depth range above the Romanche and Chain Fracture Zones. However, this phase shift is considerably greater, about six months, and is attributed to second-mode baroclinic equatorial Rossby waves emanating from the African coast. We doubt that similar, lower mode waves can propagate as far south as our latitude. A 6-month (semiannual) variability is indicated in the spectrum (not shown) of the transport time series (Fig. 8). Fischer and Schott (1997) and Mercier and Speer (1998) also reported semiannual variability.
5. Some concluding remarks
Our view of the DWBC inferred from the current meter data differs significantly in one way from what we expected from McCartney (1993) and DW. We do not see a region of equatorward flow just seaward of the DWBC (associated with the deep bowl shape region in the isopycnals mentioned in the introduction) that they inferred. We think the discrepancy is due to our data not supporting their choice of a level of no motion at the NADW–AABW interface. Figure 2 from DW indicates that in the vicinity of mooring 4 the NADW flow should have been equatorward and at about 3600-m depth, as they assumed, there is no flow. However, Fig. 4 does not give such a picture. The current meter data indicate that mooring 4, on average, is located in the seaward edge of the DWBC and that the flow extends downward to the bottom. If we use in Fig. 2 a near-bottom reference velocity in the vicinity of mooring 4 consistent with the current meter results presented in Fig. 4 (i.e., the near-bottom flow near mooring 4 is several cm s−1 toward the south), then the equatorward flow there is replaced by a poleward flow. We do not think, based on the current meter data, that there is a region of recirculating, poleward flow of NADW just seaward of the DWBC.
If we are indeed correct that there is no recirculation or poleward flow region of NADW just seaward of the DWBC and that the interface of NADW and AABW beneath the DWBC is not a level of no motion, then the magnitude of previous estimates of the transport of this DWBC in the Brazil Basin made from hydrographic sections (e.g., McCartney 1993; DW; Zangenberg and Siedler 1998) may be too small.
A second difference with DW, and with some other instantaneous but direct velocity measurements in the Brazil Basin (Rhein et al. 1995), is that a second poleward flow resembling a weaker offshore branch of the DWBC (e.g., at distance about 500 km in Fig. 2) does not appear in the current meter mean pictures. There are, however, in the approximately 540 individual velocity sections used to create Fig. 8 many instances of a double-core structure to the DWBC. (When there are two cores, they are, in general, not separated by a strong equatorward flow region.) Further examination of the individual velocity sections used to create Fig. 8 indicates that the changes in the transport shown in Fig. 8 are due to variations in the width, thickness, and/or strength of the DWBC rather than to on/offshore meandering of the DWBC.
The large 39 Sv mean poleward DWBC transport and the apparent lack of the expected recirculating region just seaward of the DWBC appear to require further comment. Concerning the former, we suspect the excess transport over the maximum possible throughflow value of about 13 Sv is due to wave-induced abyssal flow similar to that proposed by Spall (1994). We noted earlier indications of mesoscale wave motions in the hydrography (Fig. 2) and of westward propagating mesoscale wave motions in auto- and cross-spectra of the UNADW records (not shown). Concerning the latter, the velocity section shown in Fig. 4 is reproduced in Fig. 11 with selected oxygen contours from the SAVE2 section (taken from DW94, Fig. 7) superimposed. Figure 11 indicates that the high oxygen core apparent in the western portion of this section (DW, Fig. 7) approximately coincides with the DWBC velocity core. No similar core is apparent farther seaward in the SAVE2 oxygen section (DW); this suggests that, if there is any poleward, compensating flow of NADW to the east in the Brazil Basin, it consists of NADW that has flowed sufficiently far and/or has sufficiently mixed with surrounding lower oxygen content waters not to have a distinguishing high-oxygen core.
In closing we note that just prior to submitting this manuscript we received a copy of a manuscript describing float data from 2500-m and 4000-m depths in the Brazil Basin by Hogg and Owen (1999). Their data in the vicinity of our moorings appears consistent with the results presented here; namely, they show that just to the north of the Vitoria–Trindade Seamount Chain the DWBC has a distinct eastward flow component. Their data also indicate that the DWBC tends to round the Vitoria–Trindade Seamount Chain and then flow westward rather than continuing eastward toward the Mid-Ocean Ridge. Our data also suggest that the DWBC might have difficulty continuing eastward to cross the Mid-Ocean Ridge because it is embedded in a flow that extends downward to depths appreciably greater than the ∼3500 m sill depth of the Mid-Ocean Ridge. However, Müller et al. (1998) found it difficult to identify the NADW in moored current records obtained just south of the Vitoria–Trindade Seamount Chain at 20°S. We suspect, in partial agreement with both Hogg and Owen (1999) and Müller et al. (1998), that some of the flow that we measured rounds the seamount chain to return as a DWBC off the coast of South America and that some continues eastward into the Brazil Basin. Perhaps some of the latter flow contributes to the eastward flowing Namib Col Current, which nearly spans the South Atlantic just south of 20°S from about 30°W to about 10°E (Speer et al. 1995).
Acknowledgments
We wish to express our thanks to the staff of the Current Meter Facility at The Florida State University and to I. Soltanovsky and O. Laurikov of the Shirshov Insitute of Oceanology ot the Russian Academy of Sciences for their excellent efforts in obtaining and processing the data, to the captains and crews of the research vessels Maurice Ewing and Oceanus for their excellent assistance in the mooring cruises, and to P. Jahromi for superb assistance in all the phases of this project. We also thank L. Ferreira da Silva, Y. Ikeda, X. Durieu de Madron, E. A. Kelley Jr., and F. Sandoval for their assistance in making the deployment cruise happen. This investigation was supported by the National Science Foundation under Grants OCE92-06117 and OCE97-30120. This was written when one of us (GW) was a visitor at the Laboratoire de Physique des Oceans at IFREMER/Brest. Travel Grant 98P4700 from the French Education Ministry for Research and Technology and the gracious hospitality of M. Arhan and Y. Desaubies are gratefully acknowledged. We profited from conversations with M. Arhan, D. Nof, M. Ollitrault, R. Pickard, S. Wacogne, and comments of the reviewers. D. Legler is thanked for providing NSCAT data and F. Sandoval and J. Moss for their help with some of the figures.
REFERENCES
Arhan, M., H. Mercier, B. Bourlès, and Y. Gouriou, 1998: Two hydrographic sections across the Atlantic at 7°30′N and 4°30′S. Deep-Sea Res., I,45, 829–872.
Boebel, O., R. E. Davis, M. Ollitrault, R. G. Peterson, P. L. Richardson, C. Schmid, and W. Zenk, 1999: First direct observations of the western South Atlantic circulation at intermediate depth. Geophys. Res. Lett.26, 3329–3332.
Broecker, W. S., 1991: The great ocean conveyer. Oceanography,4, 78–89.
DeMadron, X. D., and G. Weatherly, 1994: Circulation, transport and bottom boundary layers in the Brazil Basin. J. Mar. Res.,52, 583–638.
Dickson, R. R., and J. Brown, 1994: The production of North Atlantic Deep Water: Sources, rates, and pathways. J. Geophys. Res.,99, 12 319–12 341.
Fischer, J., and F. A. Schott, 1997: Seasonal transport variability of the Deep Western Boundary Current in the equatorial Atlantic. J. Geophys. Res.,102, 27 751–27 769.
Hall, M. M., M. McCartney, and J. A. Whitehead, 1997: Antarctic Bottom Water flux in the equatorial Western Atlantic. J. Phys. Oceanogr.,27, 1903–1926.
Harkema, R., and G. L. Weatherly, 1996: A compilation of moored current meter data from the western boundary of the Brazil Basin for the Deep-Basin Experiment. Dept. of Oceanography, Florida State University Tech. Rep. CMF-96-01, 72 pp. [Available from WOCE Current Meter Archive Center, CMDAC, College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR 97331.].
Hellerman, S., and M. Rosenstein, 1983: Normal monthly wind stress over the world ocean with error estimates. J. Phys. Oceanogr.,13, 1093–1104.
Hogg, N. G., and W. B. Owens, 1999: Direct measurements of the deep circulation within the Brazil Basin. Deep-Sea Res.,46, 437–453.
——, ——, G. Siedler, and W. Zenk, 1996: Circulation in the deep Brazil Basin. The South Atlantic: Present and Past Circulation, G. Wefer, W. H. Berger, G. Siedler, and D. J. Webb, Eds., Springer-Verlag, 355–361.
Johns, W. E., D. M. Fratantoni, and R. J. Zantopp, 1993: Deep western boundary current variability off northeastern Brazil. Deep-Sea Res.,40, 293–310.
Lee, T. N., W. Johns, F. Schott, and R. J. Zantopp, 1990: Western boundary current structure and variability east of Abaco, Bahamas, at 26.5°N. J. Phys. Oceanogr.,20, 446–466.
——, ——, R. J. Zantopp, and E. R. Fillenbaum, 1996: Moored observations of western boundary current variability and thermohaline circulation at 26.5°N in the subtropical North Atlantic. J. Phys. Oceanogr.,26, 962–983.
McCartney, M. S., 1993: Crossing of the equator by the deep western boundary current in the western Atlantic Ocean. J. Phys. Oceanogr.,23, 1953–1974.
Mercier, H., and K. G. Speer, 1998: Transport of bottom water in the Romanche Fracture Zone and the Chain Fracture Zone. J. Phys. Oceanogr.,28, 779–790.
Müller, T. J., Y. Ikeda, N. Zangenberg, and L. V. Nonato, 1998: Direct measurements of western boundary currents off Brazil between 20°S and 28°S. J. Geophys. Res.,103, 5429–5437.
Nof, D., and W. Dewar, 1993: The drift of mid-ocean jets. J. Phys. Oceanogr.,23, 2313–2355.
Rhein, M., L. Stramma, and U. Send, 1995: The Atlantic Deep Western Boundary Current: Water masses and transports near the equator. J. Geophys. Res.,100, 2441–2457.
Richardson, P. L., and W. J. Schmitz, 1993: Deep cross-equatorial flow in the Atlantic measured with SOFAR floats. J. Geophys. Res.,98, 8371–8387.
Smethie, W. M., Jr., and G. L. Weatherly, 1998: Scientists report preliminary results from the U.S. DBE Cruise. U.S. WOCE Notes,8, 18–22.
Spall, M. A., 1994: Wave-induced abyssal recirculations. J. Mar. Res.,52, 1051–1080.
Speer, K. G., and W. Zenk, 1993: The flow of Antarctic Bottom Water into the Brazil Basin. J. Phys. Oceangr.,23, 2667–2682.
——, G. Siedler, and L. Talley, 1995: The Namib Col Current. Deep-Sea Res.,42, 1933–1950.
Tennekes, H., and J. L. Lumley, 1972: A First Course in Turbulence. The MIT Press, 300 pp.
Thierry, V., 1998: Le signal saisonnier dans l’Atlantique equatorial profond et son enfluence sur la distribution des masses d’eau. Projet de thèse, Laboratoire de Physique des Océans, IFREMER/Centre de Brest, 49 pp. [Available from Laboratoire de Physique des Océans, IFREMER/Centre de Brest, B.P. 70, 29280 Plouzane, France.].
Zangenberg, N., and G. Siedler, 1998: Path of the North Atlantic Deep Water in the Brazil Basin. J. Geophys. Res.,103, 5419–5437.
APPENDIX
Sensitivity Estimates
Initial data voids
For the velocity sections used to compute the transports shown in Figs. 8 and 9, values for the AAIW level value for mooring 3 were inferred by vertically extrapolating the UNADW and LNADW values below. As mentioned in the text we now think this value should have been inferred by horizontal interpolation. Here we examine the sensitivity of the inferred transports to different choices in estimating values for data voids. Also, for Figs. 8 and 9 the AAIW value at mooring 1 was inferred by horizontally extrapolating the AAIW values at moorings 2 and 3 (the latter value in turn was an extrapolated value).
If an AAIW value for mooring 3 had instead been estimated by horizontal interpolation of the AAIW values on moorings 2 and 4, the net NADW transport indicated in Fig. 5 would diminish by 2 Sv (∼6%). If an AAIW value for mooring 1 had instead been inferred by vertically extrapolating the measured UNADW flow on the same mooring using the geostrophic shear inferred from the hydrographic data shown in Smethie and Weatherly (1998), the net NADW transport indicated in Fig. 5 would increase by 1 Sv (∼3%).
We did not estimate what would happen if vertical extrapolation using hydrographically inferred geostrophic shear had also been used to infer a value for the AAIW position on mooring 3. As explained in section 4, we think the geostrophic shear in the upper 3000 m inferred from hydrographic sections is appreciably aliased by Rossby and/or topographic waves along our mooring array except perhaps at mooring 1. The geostrophic shear in the vicinity of mooring 3 seen at the 9001–1700 m depth range in Smethie and Weatherly (1998) is different in sign and magnitude from that seen in the same region in the SAVE2 data (Fig. 2). We think the difference is due to the aforementioned Rossby and/or topographic waves. The three geostrophic sections for the Brazil Basin in DW and the one shown in Smethie and Weatherly (1998) show similar geostrophic shears just seaward of the shelf break in the UNADW–AAIW depth range (i.e., locations similar to mooring 1) but show quite different geostrophic shears about 125 km seaward of the shelf break in the same depth range (i.e., locations similar to mooring 3).
Subsequent data dropouts
The transport estimates in Fig. 8 are relatively more accurate at the beginning of the record until May 1994, referred to as period I. During this period all the current meters indicated in Fig. 4 worked. The data dropouts occurred after this period; they did not occur randomly but clustered around two times. During 20–21 May 1994, two current meters failed: the LNADW one on mooring 3 and the AABW one on mooring 4 (Table 1). The remaining failures occurred about a year into the experiment. Specifically, after 342–343 days all the AAIW current meters stopped recording data and after 346 days the LNADW current meter on mooring 2 failed (Table 1). The period 22 May–4 September 1994 is called II, and after that to the end of the record in March 1995 is called period III.
To get an idea about how the transport estimates may have been degraded in periods II and III due to data dropouts, the following was done. The velocity section and associated transport for period I, using the currents averaged over this period were calculated (Figs. A1a and A2a). The procedure was then repeated for period I except that the data dropouts for period II were omitted from the calculations (Figs. A1b and A2b). The procedure was finally repeated for period I except the dropouts for period III were omitted (Figs. A1c and A2c). Comparing the velocity sections in Fig. A1 indicates that the visual impression of the DWBC with or without the dropouts is about the same. Comparing the transports in Fig. A2 suggests that in periods II and III the transports were overestimated by approximately 1% and 9%, respectively.
Other
The depths of the bounding isopycnals for NADW used in our transport estimates are those at the time a CTD survey was made across our moored array (Smethie and Weatherly 1998). The depths of these isopycnals are shown in Fig. 4, and the upper bounding one, σ2 = 36.7, is often deeper (by ∼50 m) than the typical 1200-m depth quoted in DW, and the lower bounding one, σ4 = 45.8, is often shallower than (by several hundred meters) the typical 3600-m depth given by DW. Smethie and Weatherly (1998) noted that the DWBC transport estimated from the hydrographic data was relatively small at about 10 Sv. Thus our transports may be underestimated. To estimate how much, the transport for Fig. 5 was recalculated taking the upper and lower bounds to be 1200 m and 3600 m, respectively. The transport increased by about 5%, or 2 Sv.
Chart showing the location of the moored current meter array (dots) at latitude −18° just north of the Vitoria–Trindade Seamount Chain (VTSMt) in the Brazil Basin. ECh denotes the Equatorial Channel, HCh the Hunter Channel, RFZ the Romanche Fracture Zone, CFZ the Chain Fracture Zone, VCh the Vema Channel, RGR the Rio Grande Rise, and LSP the Lower Santos Plato. The crest of the Mid-Ocean Ridge is at about longitude −11°. Depth contours are 200 m, 3000 m, 4000 m, and 5000 m.
Citation: Journal of Physical Oceanography 30, 5; 10.1175/1520-0485(2000)030<0971:EMOTNA>2.0.CO;2
The South Atlantic Ventilation Experiment (SAVE) 2 geostrophic velocity section (from De Madron and Weatherly 1994) with the numbered current meter moorings superimposed (moorings 8 and 10 were not recovered due to release problems.) The solid dots and squares denote current meters. Positive contours denote poleward (southward) flow and negative contours denote equatorward flow. Units are centimeters per second, and the level of no motion is taken at the NADW–AABW interface, which is approximately at depth 3600 m. These results suggest a NADW DWBC extending approximately from mooring 1 to almost mooring 4, with a recirculating poleward flow just seaward of it, and a weaker, secondary core of the DWBC in the vicinity of mooring 7.
Citation: Journal of Physical Oceanography 30, 5; 10.1175/1520-0485(2000)030<0971:EMOTNA>2.0.CO;2
Expanded view of the region showing the current meter moorings (red crosses) with the record mean currents superimposed. Upper left: records from depth 900 m in Antarctic Intermediate Water (AAIW); upper right: records from depth 1800 m in upper North Atlantic Deep Water (UNADW); lower left: records from depth 2800 m in lower North Atlantic Deep Water (LNADW); and lower right: records from 100 m above the bottom (mab) in Antarctic Bottom Water (AABW). The current meter moorings are numbered 1, . . . , 7, 9 starting from the left, but not so identified in the figure.
Citation: Journal of Physical Oceanography 30, 5; 10.1175/1520-0485(2000)030<0971:EMOTNA>2.0.CO;2
Contours of the current component directed toward 140°T (in cm s−1) averaged over period I (see text and appendix) when the data recovery rate was highest. This figure is considered to be more visually representative of the velocity structure than that formed from the record-mean records as the averaging period is the same for all current meters and the data return was the highest. Locations of the current meters are indicated by dots, and the mooring numbers are shown at the top. The dashed curves are the bounding isopycnals for NADW (σ2 = 36.7 upper and σ4 = 45.87 lower) used in this study. Distance is measured from the shelf break.
Citation: Journal of Physical Oceanography 30, 5; 10.1175/1520-0485(2000)030<0971:EMOTNA>2.0.CO;2
Cumulative volume transport estimated from the record mean velocity component along 140°. The transport has been factored by sin(140° − 90° + 8.5°) (see text) and integrated between the depth range of the two isopycnals shown in Fig. 4. The smoothness of the curve results from finely griding (180 × 180) the MATLAB returned velocity section to estimate the transport.
Citation: Journal of Physical Oceanography 30, 5; 10.1175/1520-0485(2000)030<0971:EMOTNA>2.0.CO;2
Stick plots of the current meter records from mooring 2. Values are daily averaged ones after four subsequent passes with a Hanning filter (weights 1/4, 1/2, 1/4), equivalent to smoothing with an 8-day low-pass filter. Seasonal variability is hinted with stronger poleward flows in the austral spring–summer (Nov–Feb) and weaker poleward flows in the austral winter, early spring (Jul–Sep).
Citation: Journal of Physical Oceanography 30, 5; 10.1175/1520-0485(2000)030<0971:EMOTNA>2.0.CO;2
As in Fig. 6 except for mooring 4.
Citation: Journal of Physical Oceanography 30, 5; 10.1175/1520-0485(2000)030<0971:EMOTNA>2.0.CO;2
Transport time series formed from daily transport estimates. For each day a figure analogous to Fig. 4 was made from the daily averaged currents for that day, after subsequent four passes with a Hanning filter, and from that figure the transport was estimated as for Fig. 5. The transport plotted is that at distance 322 km (see Fig. 5), or equivalently from the shelf break to mooring 5. The dashed curve is the running 6-month mean, which indicates a seasonal variability. The dot–dashed curve is the 12-month running mean, which indicates a mean poleward transport of about 39 Sv. The interval to the left of the first vertical dashed line is period I referred to in the text and the appendix, while that between the two vertical dashed lines is period II. The interval to the right of the second vertical dashed line is period III.
Citation: Journal of Physical Oceanography 30, 5; 10.1175/1520-0485(2000)030<0971:EMOTNA>2.0.CO;2
As in Fig. 8 except (a) for Antarctic Intermediate Water between σ2 = 36.7 and depth 900 m, (b) for Upper North Atlantic Deep Water (36.7 < σ2 ⩽ 36.95), (c) for Middle North Atlantic Deep Water (36.95 < σ2 ⩽ 37.02), (d) for Lower North Atlantic Deep Water [(σ2 = 37.02) ⩽ σ ⩽ (σ4 = 45.87)], and (e) for Antarctic Bottom Water (σ4 > 45.87). The mean transport for each water type, estimated from its respective 12-month running mean (dot–dashed) curve, is listed in Table 2. Like Fig. 8 each 6-month running mean (dashed) curve indicates a seasonal variability.
Citation: Journal of Physical Oceanography 30, 5; 10.1175/1520-0485(2000)030<0971:EMOTNA>2.0.CO;2
The velocity section in Fig. 4 redrawn (in cm s−1) with oxygen values [in ml L−1 (darker lines)] from the SAVE2 section adapted from DW. The high oxygen core approximately coincides with the high velocity core. (The secondary oxygen core, the closed 5.82 ml L−1 contour, identifies LNADW. This secondary core suggests that at the time of the SAVE2 section the lower, LNADW, portion of the DWBC was slightly wider than its upper, UNADW, portion.)
Citation: Journal of Physical Oceanography 30, 5; 10.1175/1520-0485(2000)030<0971:EMOTNA>2.0.CO;2
Wind stress curl at 18.25°S averaged across the Brazil Basin for the period October 1996–July 1997. A seasonal variability (which leads that of the DWBC transport shown in Fig. 8 by about 2–3 months) is evident. Values were derived from NSCAT.
Citation: Journal of Physical Oceanography 30, 5; 10.1175/1520-0485(2000)030<0971:EMOTNA>2.0.CO;2
Fig. A1. (a) Velocity component along 140°T averaged over period I (see text and the appendix). (b) As in (a) except the data dropouts for period II (LNADW on mooring 3 and AABW on mooring 4) are excluded. (c) As in (b) except the data dropouts for period III (AAIW on moorings 2, 4, 5, 6, and LNADW mooring 2) are excluded. Note that Figs. 4 and A1 display the same data; however, for the former the missing value at mooring 3 at depth 900 m is estimated by horizontal interpolation, while that in the later it is estimated by vertical extrapolation.
Citation: Journal of Physical Oceanography 30, 5; 10.1175/1520-0485(2000)030<0971:EMOTNA>2.0.CO;2
Fig. A2. (a)–(c) The cumulative volume transport for Figs. A1a, A1b, and A1c, respectively.
Citation: Journal of Physical Oceanography 30, 5; 10.1175/1520-0485(2000)030<0971:EMOTNA>2.0.CO;2
Information about the record-mean east–west (u) and north–south velocity (υ) components. The uncertainties were estimated from the relation [
Mean transport of the different water types estimated from the 12-month running mean curves in Fig. 9 and the amplitude of their seasonal variability estimated from the 6-month running mean curves in Fig. 9. The uncertainty of the velocity values listed in Table 1 were used to estimate the transport uncertainties. The last column is the approximate date the 12 and 6 month curves intersect for each water type; this date should be the same if the seasonal variability were truly barotropic and the data set had no drop outs.
