When the GEM-ETTA method is applied to the BEST mooring data, timeseries of T(z) and S(z) result (Figs. 20 and 21). Qualitatively, these timeseries exhibit the kind of variability that is expected in this region, with vertically coherent mesoscale features containing thermohaline anomalies propagating through the array at frequent intervals.

The BEST project data present a number of challenges. Two of the moorings, SN 56 and SN 59, lacked bottom pressure recorders, leaving SSH uncorrected for up to 4 cm of sub-tidal barotropic sea level variability at those locations. Furthermore, using the bottom pressure records that are available, up to 4 cm of additional error aliased to a ~60-day periodicity was found in the CSR model tide, which was used in the altimetric correction. At three of the five moorings, an in situ tide record can be generated, and this addition is planned as a future refinement.

One of the BEST moorings, SN 58, was located over 75 km away from the nearest TOPEX/POSEIDON groundtrack. Both the altimetry and the mooring data for this location were lowpass filtered to help reduce the presumably uncorrelated high frequency noise. However, the filtering will alter the SSH-tau parameter value pairs which are the index into GEM-ETTA thermohaline estimates. The effect of the spatial separation and filtering on the thermohaline estimates has not yet been quantified but will be investigated using one of other the moorings and a record of SSH from ~75 km away along the nearest TOPEX groundtrack.

A comparison of GEM-ETTA estimates and five reasonably simultaneous hydrocasts collected during the BEST-2 cruise was made. The small number of validation stations did not allow us to generate accurate statistics, but qualitatively speaking the in situ profiles and their GEM-ETTA estimates show good agreement, with the exception of the estimated salinity at SN 62 (on the continental slope). Because the GEM-ETTA fields were calculated using stations that contained data to 1500 dbar, in fact, no profiles from the continental slope or shelf were included. This result indicates that a separate GEM-ETTA regression should be created for use with with moorings on the continental slope or shelf, particularly in light of the strong Benguela upwelling, a process which can significantly alter the water mass distribution over the shelf and slope.

In addition, high rms error in the near-surface estimates (Fig. 6b) indicates the need for a seasonal decomposition to be added to the GEM-ETTA fields in the upper 150 m. Note that satellite sea surface temperatures (SSTs) provide an additional source of data which could also be used to enhance estimates of near-surface temperature and salinity values. These additional steps (seasonal decomposition, use of SST) will be incorporated in the coming year.

Lastly, while the stations selected for validation were obviously co-located with the BEST mooring deployment sites, they were not necessarily occupied at the satellite overflight times. The repeat cycle of TOPEX is 10 days, meaning that up to a 5-day lag could occur between the in situ measurement and the GEM-ETTA estimate. In fact, the minimum delay was 3 days and the maximum, 4.2 days.

The upcoming field component of ASTTEX will measure tau and SSH at mesoscale (~80 km) resolution using an array of twelve PIES moorings deployed across the Agulhas eddy corridor into the South Atlantic. The application of GEM-ETTA to these data is expected to result in detailed thermohaline flux timeseries for the Indian-South Atlantic exchange. Because these moorings will be located along a single TOPEX/Jason-1 groundtrack, the horizontal scale of anomalies propagating through the array will be able to be estimated using the relatively high resolution (~7 km) altimeter data. This will enhance the information able to be derived from the ASTTEX mooring measurements, in particular allowing us to define accurate length scales for the mesoscale thermohaline anomalies detected and leading to more accurate thermohaline and mass flux estimates.

In addition, a validation mooring collecting in situ temperature and salinity at 6 depths (200, 400, 600, 800 and 1400 m) will provide 72 validation profiles exactly coincident with the altimeter overflight. This is a great deal more than we were able to provide from the BEST data, and will allow a quantitative analysis of the errors in the GEM-ETTA estimates. Furthermore, the profile data will be collected hourly, which means that the temporal decorrelation of the thermohaline structure will be able to be estimated, giving us additional insight into the constraints necessary for the rigorous application of the GEM-ETTA method.

The Indian Ocean provides heat and salt to the southeastern South Atlantic. This inter-ocean connection affects the thermohaline balance of South Atlantic surface, central, and intermediate waters, helping to precondition the water column for the formation of North Atlantic Deep Water (NADW) and so to strengthen the Atlantic meridional overturning circulation (MOC). Thermohaline anomalies of Indian Ocean origin appear in the Cape Basin hydrographic record as well as in integrated measures of water column properties such as acoustic travel time (tau) and sea surface height (SSH). With the Gravest Empirical Mode-Enhanced Thermohaline Transport Analysis (GEM-ETTA) method, timeseries of these properties can be used to monitor the thermohaline variability in the water column and the associated fluxes on an ongoing basis. Monitoring requires independent estimates of tau and SSH, which can be obtained using an inverted echo sounder and bottom pressure mooring (PIES) in combination with a satellite altimeter.

The Agulhas-South Atlantic Thermohaline Transport Experiment (ASTTEX) will investigate fluxes of heat, salt and mass from the Agulhas Retroflection into the South Atlantic subtropical gyre using the GEM-ETTA method. This experiment is designed to provide a quantitative, multi-year Eulerian measurement of the strengths and characteristic scales of Agulhas-South Atlantic (ASA) fluxes. These have been shown by the principal investigators and others to contain a strong mesoscale component associated with features such as rings shed from the Agulhas Retroflection. ASTTEX represents the first attempt to measure the Eulerian flow field with a spatial resolution sufficient to resolve this mesoscale component of the thermohaline fluxes.

In this part of the work, a large number of historic hydrographic observations in the region were synthesized and projected onto tau and SSH. Measurements of tau and SSH made during the BEST project were then used to estimate temperature and salinity profiles in the water column. Simulated results from the application of GEM-ETTA to a 14-station test suite of hydrographic data acquired as part of the Benguela Sources and Transports (BEST) project (Fig. 6) look promising, as do the preliminary results from a comparison of dtau/SSH-estimated profiles derived from BEST mooring data with 5 in situ profiles (Fig. 19).

This retrospective analysis, while not able to produce information as detailed or accurate as that expected from the upcoming experiment, will provide a reference and large-scale, long-term comparison for the ASTTEX flux measurements. The comparison of results from the application of GEM-ETTA to each of these two experiments (BEST: 1992-1993, ASTTEX: 2002-2004) will result in a first estimate of decadal-scale variability in the Atlantic-Indian Ocean component of the MOC.