low latitudes (Firing et al., 1998), where geostrophic
calculations from synoptic sections are poor.
Perhaps less expected has been their utility in high
latitudes, where they have shown narrow currents
and eddies extending to the bottom – features that
would be misrepresented by typical geostrophic
sections. King et al. (1996) show the Antarctic
Circumpolar Current at Drake Passage resolved
at 3 km resolution along track. An eddy with a
strong barotropic component was encountered at
60°S on that section: the top-to-bottom geostrophic
shear from CTD data was no more than 30 cm s
91
(King and Alderson, 1994), while the ADCP
revealed surface speeds up to 60 cm s
91 .
SADCP and LADCP measurements can be
blended with geostrophic shear calculated from
CTD profiles for improved estimates of geostrophic
currents and transports. The simplest method is
to use SADCP velocities as the reference for the
geostrophic shear profiles. The ageostrophic velocity component is usually the largest source of error.
Saunders and King (1995a,b) reduced the error by
smoothing along track, then estimated a noise
threshold below which the ADCP-referenced
bottom velocities were considered negligible and
set to zero. The ageostrophic noise can also be
reduced by averaging in the vertical over the thickest layer with consistently good measurements, but
excluding the mixed layer (Bacon, 1994). Wijffels
et al. (1998) found the problem of ageostrophic
‘noise’ in the ADCP velocities to be an insurmountable problem for referencing. Instead, they
used deep portions of LADCP profiles to reference
geostrophic profiles in the Kuroshio and its
recirculation region, and the full LADCP profiles
inshore of the Kuroshio. Compared with the
SADCP, LADCP profiles have the advantage of
sampling the full water column, most of which we
expect to have lower levels of ageostrophic noise,
but the disadvantage of providing only point measurements in the horizontal rather than a genuine
average between stations. Close station spacing is
therefore crucial.
A more complex method of blending ADCP and
geostrophic current estimates is via an inverse
model. Joyce et al. (1986) set up a system with
conservation in 13 layers, and investigated the
impact of using ADCP currents as additional constraints in the box inverse, compared with using
hydrographic data only. Bingham and Talley
(1991) compared two methods of using the
SADCP velocities in the inverse, as additional constraints or as initial values, and found the latter
preferable. These studies are based on pre-WOCE
data sets; inversions using the much higher-quality
WOCE ADCP data are in progress, and it is too
early to say how much (or how little) the velocity
measurements will contribute to the end result.
Some hints are provided by an intriguing study in
which inversions were performed on the output of
a numerical model, FRAM. First, McIntosh and
Rintoul (1997) showed that an initial estimate of
the section-averaged reference velocity differing
greatly from the correct value could bias the inversion under some conditions; in such a case, if the
section-averaged ADCP velocities are sufficiently
accurate, they would improve the solution. Second, the small-scale structure of the actual reference layer velocity was not strongly constrained by
the geostrophic shear and water property distributions, so that good flux estimates could be
produced with highly smoothed reference-layer
velocity estimates. This implies that when the section-averaged ADCP-derived reference velocity is
not significantly different from the default of zero,
it may have little effect on the net fluxes from the
inversion; but the inversion without any direct
velocity input will misrepresent the horizontal structure of the velocity field at the reference
level.
SADCP data from 44 cruises combined with
CTD data from 29 of these have been used to elucidate the current and potential vorticity structure
of subsurface currents near the equator in the
Pacific. Averaging in density coordinates in the
vertical, and in stream coordinates meridionally
centred on the eastward-flowing North and South
Subsurface Countercurrents, Rowe et al. (2000)
showed that the mean zonal velocity at the core of
each current is sharply peaked as a function of latitude; the corresponding relative vorticity reversal
contributes to a potential vorticity front at the
core no wider than 40 km. The high horizontal
resolution of the ADCP measurements was critical
in resolving this structure.
SADCP data have been used in the Atlantic
(Chereskin and Roemmich, 1991), the Pacific
(Wijffels et al., 1994), and the Indian Ocean
(Chereskin et al., 1997) to measure the surface
Ekman layer transport across a low-latitude zonal
section, and its distribution in the vertical. The
cross-track Ekman velocity was estimated as the
3.1 Shipboard Observations during WOCE
119
King, Firing and Joyce
calculations from synoptic sections are poor.
Perhaps less expected has been their utility in high
latitudes, where they have shown narrow currents
and eddies extending to the bottom – features that
would be misrepresented by typical geostrophic
sections. King et al. (1996) show the Antarctic
Circumpolar Current at Drake Passage resolved
at 3 km resolution along track. An eddy with a
strong barotropic component was encountered at
60°S on that section: the top-to-bottom geostrophic
shear from CTD data was no more than 30 cm s
91
(King and Alderson, 1994), while the ADCP
revealed surface speeds up to 60 cm s
91 .
SADCP and LADCP measurements can be
blended with geostrophic shear calculated from
CTD profiles for improved estimates of geostrophic
currents and transports. The simplest method is
to use SADCP velocities as the reference for the
geostrophic shear profiles. The ageostrophic velocity component is usually the largest source of error.
Saunders and King (1995a,b) reduced the error by
smoothing along track, then estimated a noise
threshold below which the ADCP-referenced
bottom velocities were considered negligible and
set to zero. The ageostrophic noise can also be
reduced by averaging in the vertical over the thickest layer with consistently good measurements, but
excluding the mixed layer (Bacon, 1994). Wijffels
et al. (1998) found the problem of ageostrophic
‘noise’ in the ADCP velocities to be an insurmountable problem for referencing. Instead, they
used deep portions of LADCP profiles to reference
geostrophic profiles in the Kuroshio and its
recirculation region, and the full LADCP profiles
inshore of the Kuroshio. Compared with the
SADCP, LADCP profiles have the advantage of
sampling the full water column, most of which we
expect to have lower levels of ageostrophic noise,
but the disadvantage of providing only point measurements in the horizontal rather than a genuine
average between stations. Close station spacing is
therefore crucial.
A more complex method of blending ADCP and
geostrophic current estimates is via an inverse
model. Joyce et al. (1986) set up a system with
conservation in 13 layers, and investigated the
impact of using ADCP currents as additional constraints in the box inverse, compared with using
hydrographic data only. Bingham and Talley
(1991) compared two methods of using the
SADCP velocities in the inverse, as additional constraints or as initial values, and found the latter
preferable. These studies are based on pre-WOCE
data sets; inversions using the much higher-quality
WOCE ADCP data are in progress, and it is too
early to say how much (or how little) the velocity
measurements will contribute to the end result.
Some hints are provided by an intriguing study in
which inversions were performed on the output of
a numerical model, FRAM. First, McIntosh and
Rintoul (1997) showed that an initial estimate of
the section-averaged reference velocity differing
greatly from the correct value could bias the inversion under some conditions; in such a case, if the
section-averaged ADCP velocities are sufficiently
accurate, they would improve the solution. Second, the small-scale structure of the actual reference layer velocity was not strongly constrained by
the geostrophic shear and water property distributions, so that good flux estimates could be
produced with highly smoothed reference-layer
velocity estimates. This implies that when the section-averaged ADCP-derived reference velocity is
not significantly different from the default of zero,
it may have little effect on the net fluxes from the
inversion; but the inversion without any direct
velocity input will misrepresent the horizontal structure of the velocity field at the reference
level.
SADCP data from 44 cruises combined with
CTD data from 29 of these have been used to elucidate the current and potential vorticity structure
of subsurface currents near the equator in the
Pacific. Averaging in density coordinates in the
vertical, and in stream coordinates meridionally
centred on the eastward-flowing North and South
Subsurface Countercurrents, Rowe et al. (2000)
showed that the mean zonal velocity at the core of
each current is sharply peaked as a function of latitude; the corresponding relative vorticity reversal
contributes to a potential vorticity front at the
core no wider than 40 km. The high horizontal
resolution of the ADCP measurements was critical
in resolving this structure.
SADCP data have been used in the Atlantic
(Chereskin and Roemmich, 1991), the Pacific
(Wijffels et al., 1994), and the Indian Ocean
(Chereskin et al., 1997) to measure the surface
Ekman layer transport across a low-latitude zonal
section, and its distribution in the vertical. The
cross-track Ekman velocity was estimated as the
3.1 Shipboard Observations during WOCE
119
King, Firing and Joyce