LADCP experiment
Since this experiment
relies heavily on direct velocity measurements, we think it necessary
to detail both these issues explicitly and outline how we plan to deal with
them. Few experiments have been designed for LADCP sampling, so this program
provides a unique opportunity to optimise data collection techniques
specifically for direct velocity profiling. The LADCP is ideally
suited to work in fast, narrow currents over steep topography.
INSTRUMENT/PROCESSING
The LADCP provides a full-depth profile of ocean current from a
self-contained ADCP mounted on the CTD rosette. During processing,
overlapping profiles of vertical shear of horizontal velocity are
averaged and gridded, to form a full-depth shear profile. The shear
profile is integrated vertically to obtain the baroclinic velocity and
the resulting unknown integration constant is the depth-averaged or
barotropic velocity. This barotropic component is then computed as the sum
of the time-averaged, measured velocity and the ship drift (minus a
small correction, less than 1 cm/sec, to account for a nonconstant fall
rate) (Fischer and Visbeck, 1993; Firing, 1998). Errors in the baroclinic profile
accumulate as 1/sqrt(N) where N is the number of samples
(Firing and Gordon, 1990). This
error translates to the lowest baroclinic mode and, for a cast of 2500
m depth, it is about 2.4 cm/sec (Beal and Bryden, 1999). The
barotropic component is inherently more accurate, because the errors
result from navigational inaccuracies alone. These are quite small
with P-code GPS, about 1 cm/sec (2 to 4 cm/sec without). Comparisons with
Pegasus suggest that the LADCP can measure the depth-averaged velocity
to within 1 \cms (Hacker et al., 1996). The rms difference between
Pegasus and LADCP absolute profiles are within the expected oceanic
variability, 3-5 cm/sec (Send, 1994), due primarily to high frequency
internal waves.
Because our measurements will be taken over steep topography
where bottom velocities are high, we will use the new bottom track and
inverse solutions method of Visbeck (2000) as a post-processing
step in the production of final LADCP data. This method has the
potential to further improve absolute velocities by constraining the shear
profile, using the least squares method, to bottom tracked velocities,
ships navigation and shipboard ADCP data.
In previous experiments the interference layer, which results from the
previous ping reflecting off the bottom, has caused a large data
gap in the LADCP profile, causing an uncertain velocity offset
(several cm/sec) between the parts of the profile on either side of the
gap. For this experiment bottom velocities will be greatly improved by
using Chereskin's upgraded instrument which will ping asynchronously,
thereby avoiding complete data loss in the interference layer. A
second problem with data loss arises at the bottom of a CTD/LADCP
cast, when the package is held 10 m above the sea bed for bottle
sampling. At this distance the instrument is 'blind' since the blank
after transmit is order 20 m, and a time gap in the data stream will
result in an uncertainty in the absolute velocity. We propose to
finish a cast at an optimum 80m above the sea bed, avoiding data loss
altogether while maintaining a good return from the water column for
profiling down to the sea bed.
AGEOSTROPHIC MOTION
Ageostrophic motions captured in LADCP measurements may represent
large errors when trying to discern a mean circulation, and ordinarily
remain unaccounted for. In this instance, we have estimated the
magnitude of such motions using results from the ACE current
time series at 32°.
These energies translate to ageostrophic motions of amplitude
up to 4 cm/sec close to the surface, assuming a pure sinusoidal wave,
and of 3 cm/sec or less at the depths of the AUC. Since average
velocities are over 10 cm/sec in the Undercurrent, and
instantaneous velocities over 30 cm/sec, these oscillations
will certainly not overcome the signal. In fact, they are of the order
of the measurement accuracy. Low frequency modifications of the
Agulhas Current, caused by meandering of the jet, are far more
energetic than these inertial and tidal oscillations. They
correspond to changes in velocity of 35 cm/sec (less than 5 cm/sec in
the AUC), assuming sinusoidal variability. This variability will
affect hydrographic and direct velocity measurements alike. The
barotropic component of the tides will be estimated and
removed from LADCP velocities using the Oregon State University
global model of ocean tides (Egbert, 1994). The model
output has been found to match closely with the barotropic component
of velocity from LADCP (Firing, 1998) in regions of large tidal
amplitude.
SHIPBOARD ADCP COMPLEMENT
Shipboard ADCP data can complement LADCP measurements in a number of
ways, helping to reduce errors. First, it affords continuous
resolution in the horizontal, where LADCP is limited by station
spacing. Second, velocities at the surface of the ocean down to depths
of 50 m are not measured with the lowering technique, due to the
instrument's blank after transmit and this gap can be (partially) filled
with ADCP data (Beal and Bryden, 1999). Third, comparisons of
ADCP and LADCP over their common depth provides a good quality check
for the lowered data and allows for quantification of some
errors. Furthermore, ADCP velocities are essential for estimates of Ekman
layer fluxes (Chereskin et al., 1997), which remain unresolved by
either LADCP or geostrophy. Because ocean current is a small residual
between the rotated ADCP measurement and the ship's velocity, the
calculation is especially sensitive to errors in heading and
transducer alignment. Heading bias errors, formerly the most limiting
error in SADCP absolute velocity estimates, have by and large been
eliminated by the use of GPS attitude measurements (King and Cooper, 1993;
Griffiths,1994;Chereskin and Harris, 1997).
The predicted error in ship-relative currents is
less than 1 cm/sec for a 10-min average (Chereskin and
Harding, 1993). Use of P-code GPS during WOCE resulted
in an absolute velocity accuracy of 2 cm/sec for 10-min averages
(Chereskin and Harris, 1997).
DIRECT VERSUS GEOSTROPHIC VELOCITIES
In regions with steep topography and narrow currents, direct measurements of
velocity have an advantage over geostrophic velocities (Beal et al., 2000).
The geostrophic method can significantly under-estimate
the peak velocities and total transport of a slope current. This is
due to the missing horizontal density gradient below the deepest common level
of each station pair. This data gap is referred to as a 'bottom
triangle' and for bottom-trapped currents, such as the AUC, it
represents a significant portion of the flow. Another disadvantage of
the geostrophic method in this case, is that it requires integration
along pressure horizons. This tends to smear out and diminish the
signature of a bottom-trapped current, which has largest velocities
being observed successively deeper on each station. In the case of the
Somali Deep Western Boundary Current, Beal et al. (2000) found
that these two sampling effects meant that only 0.5 Sv of a directly
estimated 5 Sv transport was captured by geostrophy. Thus, while the
strength of the geostrophic method lies in capturing the net current
over a region between station pairs, the LADCP can better resolve
individual features of width not much greater than station-spacing,
especially over topography.