earth map with SAFDE location

Sub-Antarctic Flux and Dynamics Experiment (SAFDE):
Overview with Some Descriptive Results

Douglas S. Luther, Univ. of Hawaii at Manoa, Honolulu, HI
D. Randolph Watts, Univ. of Rhode Island, Kingston, RI
Alan D. Chave, Woods Hole Oceanographic Institution, Woods Hole, MA
James G. Richman, Oregon State Univ., Corvallis, OR
Stephen R. Rintoul, CSIRO Marine Laboratories, Hobart, Australia
John A. Church, CSIRO Marine Laboratories, Hobart, Australia
Jean H. Filloux, Scripps Institution of Oceanography, La Jolla, CA



The broad intent of SAFDE was to obtain long duration, spatially coherent measurements of current and temperature in order to provide information on property fluxes and dynamic balances, and their variability, relative to the location of the Sub-Antarctic Front (SAF) and with respect to time. The suite of instruments deployed incorporates both well established ideas of oceanographic experimentation, as well as novel approaches. The former is represented by a seven element mooring array, where each sub-surface mooring supported 1-5 current meters and up to 2 additional temperature sensors from 300 m to as deep as 3200 m. Straddling this local dynamics array, and extending approximately 450 km along a NNE-SSW line, is an array of 13 horizontal electrometers (HEMs) and 18 inverted echo sounders (IESs). These seafloor instruments are well suited for observing currents at mid- to high latitudes where large vertical scales dominate.

The HEMs alone provide direct measurements of the vertically averaged horizontal currents. The IESs and HEMs combined provide, at least, both horizontal components of a vertical average (barotropic) and a gravest baroclinic (empirical) mode description of the horizontal current field. The less conventional HEM/IES data is emphasized in this presentation. Summaries of the structure and variability of the absolute (barotropic and baroclinic) currents associated with the Sub-Antarctic Front (SAF) southwest of Tasmania (along WOCE SR3) are presented, using data from a large array of instruments deployed from 3/'95 to 3/'97. Several unusual aspects of the transport variability and the mean circulation are noted.

Purpose of SAFDE

Given logistical and cost barriers, observations in the Southern Ocean will always be sparse in time and space. Advances in our knowledge of the Antarctic Circumpolar Current (ACC) dynamics must come therefore through a synergy of models and observations. What this means for observations is that they must focus on measuring quantities, such as non-geostrophic terms of the momentum equations and terms of the vorticity equation, that can provide better discrimination of differing model assumptions and parameterizations.

SAFDE was designed with this principle in mind, to achieve observations of the ACC south of Tasmania that would permit direct evaluation of the momentum, energy and vorticity budgets. That these budgets differ in different sectors of the Southern Ocean is as obvious as the differing bottom topographies that exert such a great influence on the ACC. The assumption of zonal invariance of the dynamics of the ACC, occasionally necessary for the development of basic dynamical concepts, must certainly be rejected in detail (see, for instance, the analysis of FRAM output by Wells and de Cuevas, 1995). One way that limited area, high resolution, observation programs such as SAFDE can be employed in the study of the ACC's dynamics over larger regions is through comparison with, and validation of, conclusions based on satellite-derived sea surface heights; and this certainly is one of our goals. But perhaps the greatest value will be achieved when the high resolution observations are used to provide the kind of information that is most useful for discriminating the validity of differing model dynamics. Models validated with such observations would then be expected to more accurately simulate the flow and variability in other locations of the Southern Ocean.

SAFDE observations alone are insufficient for satisfactory model validation, but prudent experiments like SAFDE, as well as thoughtful analyses of surface drifter velocities and altimeter-derived velocity components at cross-over points, will take us a long way toward that goal. A good example of a comparison of higher moment statistics derived both from a numerical model of the Southern Ocean and from various altimeter, drifter and current meter datasets can be found in Wilkin and Morrow (1994).

SAFDE has achieved multi-year observations of currents and temperatures in both a small current meter mooring array with a diameter of ~70 km, and along a NNE-SSW section perpendicular to the expected mean axis of the ACC at the Sub-Antarctic Front (SAF). The experiment lasted two years, from March, 1995, to March, 1997. The measured variables are coherent horizontally and vertically in broad, sub-inertial frequency bands. The rarity of such horizontally coherent observations, required for estimating momentum budgets, etc., is revealed by noting that when SAFDE was proposed in 1991 only three locations in the ACC had been sampled with moored current meters for as long as one year, and only one location for as long as two years, with the horizontally coherent records all being no more than about a year long.


Figure 1 displays the SAFDE array (as recovered) overlaid on Smith and Sandwell (1997) topography. Triangles are IESs; stars are HEMs; large (small) circles are full (partial) moorings; the diamond indicates coincident HEM and partial mooring; and, the square is a pressure recorder. All locations shown yielded full two-year records, except the southernmost HEM which had 543 days and the pressure gauge which had a one-year record.

The current associated with the SAF, containing the majority of the ACC transport at this longitude, flows from the west at about 49.5S on the north side of the Southeast Indian Ridge. After the current crosses the 140E fracture zone into deeper water its mean position tends toward the east-southeast (e.g., Gille, 1994) though it executes large meanders as predicted by McCartney (1976). The SAFDE array was oriented along a NNE-SSW line overlapping the WOCE SR3 repeat hydrography track that was designed to be perpendicular to the approximate mean axis of the SAF. Figure 2 shows the locations of the HEMs (the row of stars in Fig. 1 ) relative to the position (50S - 52S) of the SAF, as observed in March, 1993, by S. Rintoul along the WOCE SR3 line.

At the center of the array, 9 sub-surface, nearly full depth moorings were deployed as a local dynamics array. Of these, 4 complete moorings and 3 partial moorings were recovered. Three of the complete moorings (indicated by circles that are overlaid by IES triangles in Fig. 1) were situated along the NNE-SSW SR3 track, and the last complete mooring was to the east. Figure 3 displays the zonal current from the four current meter depths (nominally, 300 m, 600 m, 1000 m and 2000 m) on the easternmost mooring (easternmost circle in Fig. 1 ). The high vertical coherence characteristic of currents at other locations along the ACC (e.g., Sciremammano et al., 1980; Bryden and Heath, 1985) is clearly evident here as well.

The three full moorings along SR3 are at the same locations as three moorings deployed by S. Rintoul from March, 1993, through January, 1995. This earlier experiment had a fourth mooring off to the west. Therefore, the combined experiments have yielded compact arrays of four tall moorings at nearly identical locations for a duration of almost four years, with measurements of currents and temperatures at 4 depths (at least).

The three partial moorings recovered in March, 1997, yielded 2000 m current and temperature records at the locations of the diamond and two small circles in Fig. 1 , two to the west of the line of 3 full moorings and one to the east. Therefore, at the 2000 m level, there are 7 two-year records of currents and temperatures.

The most novel aspect of SAFDE was the deployment of a suite of HEMs to obtain time series of the vertically averaged, horizontal water velocity, and a suite of IESs to obtain time series of the temperature and dynamic height structures. Seventeen HEMs were deployed with fifteen recovered, yielding 13 complete records at the locations shown in Fig. 1 Eighteen IESs were deployed in SAFDE. All were recovered with complete records.

IES Acoustic Travel Times and HEM Electric Fields: Interpretation and Processing

The IESs measure acoustic travel times (ATT) from the seafloor to the surface and back. Watts and Rossby (1977) showed that variations in the ATT are dominated by the largest vertical scales (lowest baroclinic mode) of temperature perturbations. Where there is high vertical coherence of the temperature perturbations, irrespective of modal content, each ATT can define almost uniquely a particular profile of temperature. If, in addition, dynamic height variability is dominated by temperature fluctuations, each ATT can also define almost uniquely a particular profile of dynamic height. To achieve the conversion from ATT to dynamic height structure there must be high vertical coherence of velocity and temperature fluctuations, a condition known to be met in the ACC as noted above, and there must be sufficient CTD data to fully map the range of measured travel times onto particular profiles of temperature or dynamic height, a condition met south of Tasmania due to the many repeat hydrographic sections conducted by S. Rintoul and collaborators over the past six years. See the poster at this meeting titled "Combining IES and Hydrography to Observe The Time-Varying Baroclinic Structure of the Sub-Antarctic Front" by Watts, Sun and Rintoul, for a discussion of the methodology of creating "gravest empirical mode" (GEM) representations of the vertical structure of temperature, specific volume anomaly, etc., as a function of ATT.

The HEMs measure the horizontal electric fields (HEFs) which are theoretically related to the conductivity weighted, vertically averaged, horizontal water velocity (Sanford, 1971; Chave and Luther, 1990), as follows:

electric field eqns.

where E is the electric field; the subscripts denote the direction - east (x), south (-y) or up (z) - in which that component is positive; u(z,t) and v(z,t) are zonal and meridional water velocity components; sigma is seawater electrical conductivity; F is the known geomagnetic field; and, C (always less than or equal to 1) is a scale factor that depends on sub-seafloor conductivity. C is estimated by intercomparisons, but is probably 0.95 +- 0.05 almost everywhere in the deep oceans. The brackets, [ ] , denote vertical average, while the caret, ^ , denotes deviation from a vertical average. The covariance integrals on the RHS of (1a&b) will be called the "conductivity correction."

The validity of the relations in (1a&b) has been demonstrated empirically (e.g., Luther et al., 1991) for periods greater than a few days. At shorter periods the measured HEFs are dominated by fields generated in the ionosphere that penetrate the ocean and crust.

The conductivity correction in (1a&b) means that the HEF is not a pure measure of vertically averaged horizontal water velocity. Where both the vertical shear of the horizontal currents is large and the conductivity varies strongly with depth, as in the Gulf Stream for instance, the conductivity correction can be up to 25% of the total. At mid- to high latitudes, conductivity varies weakly with depth, resulting in smaller corrections (Chave and Luther, 1990). In the SAF, the conductivity correction is small due to relatively weak stratification and shear. Even so, where the IES data are available they have been used to calculate the conductivity correction, which is then subtracted from the HEF measurement to yield vertically-averaged water velocity components [u(z,t)] and [v(z,t)]. Dynamic height profiles obtained from the IESs, via the GEM approach mentioned previously, were used to estimate top-to-bottom shears. GEMs relating seawater conductivity to IES ATTs were produced to generate conductivity profiles as a function of time. The shears and conductivity profiles at the HEM sites were then combined to yield the conductivity corrections per (1a&b). Using data from site H7, Figure 4 displays an example of the conductivity correction, along with the un-corrected, HEF derived, vertically averaged meridional current, [v(z,t)]*, and the resultant corrected estimate of [v(z,t)].

For sites H2-4, 16 & 17 (refer to Fig. 6 for site numbers), where no IES data is available, regression relations between the conductivity corrections and the appropriate [u(z,t)]* and [v(z,t)]* were estimated from sites H7 and H14, so that the conductivity corrections could be predicted from the observed electric fields. Given the small size of the conductivity correction throughout the array (e.g., Fig. 4 ), this method of obtaining the conductivity correction results in only a very small error in the final vertically averaged currents at H2-4, 16 & 17.

The HEF data required substantial, unexpected processing to arrive at estimates of [u(z,t)]* and [v(z,t)]* prior to application of the conductivity correction. Most of the compasses failed, and the HEFs appeared to be distorted by lateral heterogeneities in the conductivity structure of the crust. But a major advantage of using HEFs to study the ocean is that at periods shorter than a day the fields are dominated by signals generated in the ionosphere that can be used to inter-calibrate an array of HEFs. The ionospheric signals are highly coherent horizontally (typically, coherence amplitudes are greater than 0.95 at periods of 1-4 hours over the 32 km separation typical of SAFDE) and have large horizontal scales so that there is little phase change over hundreds of kilometers. The high frequency fields, specifically the fields in the 30 minute to 4 hour band, were used to align the measured HEFs with respect to each other and to calibrate the HEF amplitudes in a relative sense. Then, at site H9, the IES-derived shears, referenced by the current meter observations at 2000m there, were combined with IES-derived conductivity profiles and integrated to produce synthetic [u(z,t)]* and [v(z,t)]* time series per (1a&b). These time series provided the final absolute orientation and amplitude calibration needed for the suite of HEF-derived currents. As it turned out, comparison of the HEF-derived [ ]* currents with the synthetic [ ]* currents at H9 resulted in no change to the previously-determined orientation of the HEF-derived currents, and no change to the amplitude of the [v(z,t)]* current. Only [u(z,t)]* (and thus all the HEF-derived [u(z,t)]* currents along the HEM array) had to be adjusted upward in amplitude by 15%.

Finally, at site H12, where the HEF derived, absolute velocity exists for only 200 days (and so is not employed in the present analysis), a special procedure was used to derive a time series of absolute currents for the full SAFDE time period. Since the HEF derived [u(z,t)] and [v(z,t)] contain information on both baroclinic and barotropic motions, the baroclinic contributions to these currents at sites H11 & H13 were subtracted, by using the IES derived shears relative to the bottom. The resultant "barotropic" (Fofonoff, 1962, definition) currents were interpolated to site H12, and the interpolants combined with the IES derived shears for that site.

Vertically Averaged Currents - Variability

The bottom referenced, IES derived shears at sites H7-14 (refer to Fig. 6 for site numbers) have been vertically averaged for comparison with the HEF derived [u(z,t)] and [v(z,t)]. The IES average currents are plotted as red lines in Figure 5a and Figure 5b , while the HEF derived currents are plotted as blue lines. The time series have been truncated to a common time period of 701 days, except that H17 has only 543 days of data. Means (in m/s) of the plotted data are provided to the right of each series, and have standard errors no greater than 0.004 m/s based on integral time scales of 10-25 days.

The high coherences and similar amplitudes of the two independent estimates of [u] and [v] in Fig. 5 are striking, even more so when it is noted that the IES currents contain only baroclinic geostrophic information, while the HEF currents also contain barotropic and boundary layer current signals. At the southern end of the array, especially sites H13 & H14, the differences between the two estimates are larger, possibly as a result of a weakening stratification, hence weakening baroclinic currents.

Variability associated with the SAF is strongest in [v] (reaching a range of 0.6 m/s at H10), and strongest at sites H7-13, in the center of the array, as hoped. This variability, associated in most instances with meandering and eddying activity, is at times fairly weak (e.g., days 150 to 300, during austral winter, 1995). An annual cycle is not visually evident in any of the time series, and is only barely discernible in high resolution power spectra of these data.

Vertically Averaged Currents - Means

The means listed in Fig. 5 are displayed graphically in Figure 6 on a map of the SAFDE HEM array overlaid on topography. The IES average currents are displayed as red vectors and the HEM average currents are white vectors. Several aspects of the means are worth noting. Both HEM and IES currents indicate a northeastward mean flow at the expected latitude of the principal axis of the SAF around 50.5S. This orientation of the front was also found in 2 years of current measurements from mooring deployments in '93-'95 at locations just west of H9 & H10 (H. Phillips and S. Rintoul, personal communication), but had not been noticed previously in coarser-resolution hydrographic (e.g., Reid, 1997) or altimetry (e.g., Gille, 1994) datasets, which show the front oriented east-west or ESE-WNW. However, numerical models of the Southern Ocean often exhibit semi-permanent standing oscillations of the ACC. For example, the Semtner-Chervin GCM exhibits mean surface geostrophic flow to the northeast at the location of the SAFDE array (e.g., Gille, 1997, Fig. 3).

What no dataset or model has previously revealed, as far as we know, is the secondary maximum of current to the southeast around 52S. Based on mean temperature sections derived from the IESs, this feature appears to be a branch of the SAF, being too warm to be considered a part of the Polar Front (PF), although the PF can wander this far north. Even farther south, the HEM currents at H16-H17 exhibit strong mean southward flow nearly perpendicular to the principal axis of the Southeast Indian Ridge at these longitudes. The time series in Fig. 5b indicate that this mean southward flow is not the result of a few events but appears to be a stable feature of the flow at all times.

The result of the two semi-permanent features described above is that along the SAFDE HEM array there is a clear divergence of the current from 51S to 52S. The divergence is stronger in the HEF derived total current, than it is in the IES derived baroclinic current. In addition, there is a convergence of the current between H4 & H6.

Vertically Averaged Currents - Horizontal Integration

The zonal currents in Fig. 5a from sites H7-H14 have been integrated meridionally to obtain a pseudo-zonal transport through a 200 km meridional section ("pseudo" is used to describe this result due to the NNE-SSW tilt of the array). Similarly, the meridional currents of Fig. 5b have been integrated zonally to obtain a pseudo-meridional transport. Both pseudo-transports have been plotted in Figure 7 for both IES derived and HEF derived currents shown in Fig. 5. The purpose of this exercise is to further compare the results from the two observation techniques. The high coherence between the two estimators of transport is obvious, with the pesudo-meridional transport estimates being practically identical in amplitude as well. The pseudo-zonal transport estimates differ in amplitude, with the HEF estimates being more energetic.

A surprising feature of the plots in Figure 7 is that the means from the two estimators differ little. Non-baroclinic flows, that affect only the HEF-derived currents, actually subtract from the total baroclinic psuedo-zonal transport, rather than add to it.

The time series of the HEF derived, pseudo-zonal transport passing through the entire HEM array from H2 to H16 is shown in Figure 8 (top) . Despite the integration over a latitudinal range of 425 km, the variability in the pseudo-transport is large, ranging from -65 Sv to 190 Sv (and, these extremes are reached within 3 months of each other). Projecting the HEF derived currents onto an array-oriented coordinate system reduces this extreme variability, but the cross-array transport ( Fig. 8 (bottom) ) still ranges over 200 Sv from maximum to minimum (positive transport is to the ESE).

Orienting the transports into an array coordinate system permits comparison with cross-track transports calculated from WOCE SR3 hydrographic data, since the SAFDE array was aligned with SR3. The red asterisks in Fig. 8 (bottom) are the cross-track baroclinic transports (referenced to the bottom) calculated from 7 hydrographic cruises along SR3 for the same distance spanned by H2 to H16. The hydrographic data was collected in 10/'91, 3/'93, 1/'94, 1/'95, 4/'95, 7/'95, and 9/'96, and yielded transport estimates of 112, 101, 79, 85, 90, 91 & 74 Sv, respectively (S. Rintoul and J. Richman, personal communications). The last three transects occurred during the SAFDE deployment, so their transport values appear at the appropriate time in Fig. 8 (bottom) . The other 4 estimates are arrayed at the beginning of the time axis in Fig. 8 (bottom) in the order that they were obtained (though with a greatly fore-shortened time axis). The 4/'95, 7/'95, and 9/'96 SR3 cross-track estimates agree well with the HEF derived transports, further evidence of the minimal impact of non-baroclinic currents on the transport through this relatively narrow span that just straddles the SAF. But the mean transport over the 701 days of SAFDE is 66.2 Sv +- 4.7 Sv s.e., while the mean over the 7 SR3 cruises is 90.3 Sv, with no estimate smaller than 73 Sv. We will leave it to the reader to draw his/her own conclusions about interannual variability, sampling bias, and so on.


This very preliminary analysis of a portion of the SAFDE dataset has provided an intriguing view of the vertically averaged circulation and its variability in the vicinity of the Sub-Antarctic Front southwest of Tasmania, which already foresees achievement of one of SAFDE's goals, that is to place in context the variability observed in the repeat hydrographic sections along SR3. In this regard, it's worth noting that volume transports across the full Tasmania to Antarctica SR3 section have less variability than might be supposed from the discussion above. Apparently, the cross-track transport variability displayed in Fig. 8 (bottom) is canceled somewhat by transports outside the SAFDE array. Considering the latitudinal variations of water properties (especially temperature and salinity) along SR3, the implication is that, while the transport variations may be canceled in the total integral from Tasmania to Antarctica, variations in property transports will probably not be canceled likewise. These issues and many others will be explored eagerly with the SAFDE dataset, in combination with other WOCE-era measurements and models of the waters south of Tasmania.


Bryden, H.L. and R.A. Heath. 1985. Energetic eddies at the northern edge of the Antarctic Circumpolar Current in the Southwest Pacific. Prog. Oceanogr., 14, 65-87.

Chave, A.D. and D.S. Luther. 1990. Low-frequency, motionally induced electromagnetic fields in the ocean, 1, Theory. J. Geophys. Res., 95, 7185-7200

Fofonoff, N. P., 1962. Dynamics of Ocean Currents. In: The Sea, Vol. 1, M. N. Hill, Ed., Interscience Pub., N.Y.

Gille, S.T.. 1994. Mean sea surface height of the Antarctic Circumpolar Current from Geosat data: Method and application. J. Geophys. Res., 99, 18,255-18,273.

Gille, S. T., 1997. The Southern Ocean momentum balance: Evidence for topographic effects from numerical model output and altimeter data. J. Phys. Oceanogr., 27, 2219-2232.

Luther, D.S., J.H. Filloux and A.D. Chave. 1991. Low-frequency, motionally induced electromagnetic fields in the ocean, 2, Electric field and Eulerian current comparison from BEMPEX. J. Geophys. Res., 96, 12,797-12,814.

McCartney, M.S.. 1976. The interaction of zonal currents with topography with applications to the Southern Ocean. Deep-Sea Res., 23, 413-427.

Reid, J., 1997. On the total geostrophic circulation of the PAcific Ocean: flow patterns, tracers, and transports. Prog. Oceanogr., 39, 263-352.

Sanford, T.B. 1971. Motionally-induced electric and magnetic fields in the sea. J. Geophys. Res., 76, 3476-3492.

Sciremammano, F. Jr., R.D. Pillsbury, W.D. Nowlin, Jr., and T. Whitworth, III. 1980. Spatial scales of temperature and flow in Drake Passage. J. Geophys. Res., 85, 4015-4028.

Smith, W. H. F. and D. T. Sandwell. 1997. Global seafloor topography from satellite altimetry and ship depth soundings. Science, 277, 1956-1962.

Watts, D.R. and H.T. Rossby. 1977. Measuring dynamic heights with inverted echo sounders: Results from MODE. J. Phys. Oceanogr., 7, 345-358.

Wells, N.C. and B.A. de Cuevas, 1995: Depth-integrated vorticity budget of the Southern Ocean from a General Circulation Model, J. Phys. Oceanogr., 25, 2569-2582.

Wilkin, J.L. and R.A. Morrow. 1994. Eddy kinetic energy and momentum flux in the Southern Ocean: Comparison of a global eddy-resolving model with altimeter, drifter, and current-meter data. J. Geophys. Res., 99, 7903-7916.

WOCE. 1997. WOCE Data Guide 1997. WOCE Report No. 150/97, March 1997, WOCE International Project Office, Southampton, UK.

Return to SAFDE Abstracts & Posters Page