corresponding to a 180° phase shift is therefore
called the ambiguity velocity; a larger velocity
magnitude will wrap around, reversing its sign.
Decreasing the code length increases the ambiguity
velocity at the cost of accuracy. Broader bandwidth increases accuracy and reduces the minimum available code length, but at the cost of
profiling range, which is critical for the LADCP.
The usual compromise is a medium-bandwidth
setting, for which the maximum ambiguity velocity for the BB-150 is about 3.3 m s
91 (along the
beam axis), and lower ambiguity velocities have
been used on occasion. Although the maximum
ambiguity velocity is large compared with average
LADCP lowering rates and relative currents, it can
be exceeded instantaneously as the package is
lifted and dropped with the pitch and roll of the
ship. Ambiguity errors can occur in any or all of
the four beams on a given ping. Some but not all
of the errors can be found using consistency checks
and glitch detection algorithms; editing is much
more effective when applied to single pings rather
than to multi-ping averages. After editing out the
detectable errors, bias may still result from the
undetected errors and from the selective removal of
samples with the largest velocities. Bias is largest in
the vertical velocity component but may be significant in the horizontal component also. We are not
aware of any quantitative estimates of the ambiguity error biases remaining in processed WHP
LADCP data sets.
Accuracy of LADCP data
This brings up an important question: how do we
evaluate LADCP performance in practice? And,
how good or bad is it? There have been only a few
comparisons between LADCP profiles and independent velocity profile measurements. Fischer and Visbeck (1993) showed the result of comparison with
Pegasus (an acoustically tracked free-fall probe)
profiles: rms differences of about 5 cm s
91 in each
component, up to a factor of two larger than the
rms difference between Pegasus up and down casts.
Hacker et al. (1996) made a similar comparison,
but compared only the depth-averaged velocity estimates from the two methods. (Poor Pegasus profile
quality, particularly at high vertical wavenumbers,
discouraged comparison of the relative velocity profiles.) Rms differences of the depth-averages were
about 1.5 cm s
91 on a cruise in 1992, and under
1 cm s
91 on a 1993 cruise using a better LADCP.
Given that direct comparisons between LADCP
and other profiling methods are rare, and clouded
by uncertainties in the alternative methods and by
spatial and temporal differences in sampling, we
are led to rely on other consistency checks. The
most general one is the comparison between up
and down casts. As noted above, this comparison
sometimes shows obvious problems. A second useful comparison is between the top of the LADCP
profile and simultaneous shipboard ADCP data.
This comparison is made separately for LADCP up
and down casts; temporal differences are often substantial, as verified by on-station shipboard ADCP
time series. Similarly, Send (1994) has shown that
Pegasus up–down differences are roughly consistent with a Garrett–Munk type internal wave spectrum. A third type of comparison is between the
bottom of the LADCP profile and the near-bottom
velocity calculated by tracking the bottom in addition to the water. Cunningham et al. (1997) have
shown cases where this method, together with the
shipboard ADCP comparison, reveal a disturbing
lack of consistency; the cause of the error is not yet
clear.
As this discussion of error sources should suggest, there is no good easy answer to the question,
‘What is the error in an LADCP profile?’ A reasonable but vague answer would be, ‘A few cm s
91
,
except when backscattering is very low, or something else goes wrong.’ A better answer would
point out that accuracy tends to be highest for the
depth average, but lowest for the lowest non-zero
vertical wavenumbers; that relative velocity profile
errors are larger for deep profiles than for shallow
ones, but the reverse may be true for the depthaveraged velocity; etc. More precisely quantifying
the errors in existing LADCP profiles, and finding
ways of reducing errors in future profiles, is an
ongoing project.
3.1.3.4 Applications: how are SADCP and
LADCP measurements changing our view of
ocean currents and physics?
The scientific role of ADCP observations is a function of their strengths and weaknesses relative to
other types of observations. The main strengths are:
1 measurement of absolute current profiles, in
contrast with geostrophic profiles, which are
always relative to an unknown reference;
2 high horizontal and vertical resolution;
3.1 Shipboard Observations during WOCE
117
King, Firing and Joyce
called the ambiguity velocity; a larger velocity
magnitude will wrap around, reversing its sign.
Decreasing the code length increases the ambiguity
velocity at the cost of accuracy. Broader bandwidth increases accuracy and reduces the minimum available code length, but at the cost of
profiling range, which is critical for the LADCP.
The usual compromise is a medium-bandwidth
setting, for which the maximum ambiguity velocity for the BB-150 is about 3.3 m s
91 (along the
beam axis), and lower ambiguity velocities have
been used on occasion. Although the maximum
ambiguity velocity is large compared with average
LADCP lowering rates and relative currents, it can
be exceeded instantaneously as the package is
lifted and dropped with the pitch and roll of the
ship. Ambiguity errors can occur in any or all of
the four beams on a given ping. Some but not all
of the errors can be found using consistency checks
and glitch detection algorithms; editing is much
more effective when applied to single pings rather
than to multi-ping averages. After editing out the
detectable errors, bias may still result from the
undetected errors and from the selective removal of
samples with the largest velocities. Bias is largest in
the vertical velocity component but may be significant in the horizontal component also. We are not
aware of any quantitative estimates of the ambiguity error biases remaining in processed WHP
LADCP data sets.
Accuracy of LADCP data
This brings up an important question: how do we
evaluate LADCP performance in practice? And,
how good or bad is it? There have been only a few
comparisons between LADCP profiles and independent velocity profile measurements. Fischer and Visbeck (1993) showed the result of comparison with
Pegasus (an acoustically tracked free-fall probe)
profiles: rms differences of about 5 cm s
91 in each
component, up to a factor of two larger than the
rms difference between Pegasus up and down casts.
Hacker et al. (1996) made a similar comparison,
but compared only the depth-averaged velocity estimates from the two methods. (Poor Pegasus profile
quality, particularly at high vertical wavenumbers,
discouraged comparison of the relative velocity profiles.) Rms differences of the depth-averages were
about 1.5 cm s
91 on a cruise in 1992, and under
1 cm s
91 on a 1993 cruise using a better LADCP.
Given that direct comparisons between LADCP
and other profiling methods are rare, and clouded
by uncertainties in the alternative methods and by
spatial and temporal differences in sampling, we
are led to rely on other consistency checks. The
most general one is the comparison between up
and down casts. As noted above, this comparison
sometimes shows obvious problems. A second useful comparison is between the top of the LADCP
profile and simultaneous shipboard ADCP data.
This comparison is made separately for LADCP up
and down casts; temporal differences are often substantial, as verified by on-station shipboard ADCP
time series. Similarly, Send (1994) has shown that
Pegasus up–down differences are roughly consistent with a Garrett–Munk type internal wave spectrum. A third type of comparison is between the
bottom of the LADCP profile and the near-bottom
velocity calculated by tracking the bottom in addition to the water. Cunningham et al. (1997) have
shown cases where this method, together with the
shipboard ADCP comparison, reveal a disturbing
lack of consistency; the cause of the error is not yet
clear.
As this discussion of error sources should suggest, there is no good easy answer to the question,
‘What is the error in an LADCP profile?’ A reasonable but vague answer would be, ‘A few cm s
91
,
except when backscattering is very low, or something else goes wrong.’ A better answer would
point out that accuracy tends to be highest for the
depth average, but lowest for the lowest non-zero
vertical wavenumbers; that relative velocity profile
errors are larger for deep profiles than for shallow
ones, but the reverse may be true for the depthaveraged velocity; etc. More precisely quantifying
the errors in existing LADCP profiles, and finding
ways of reducing errors in future profiles, is an
ongoing project.
3.1.3.4 Applications: how are SADCP and
LADCP measurements changing our view of
ocean currents and physics?
The scientific role of ADCP observations is a function of their strengths and weaknesses relative to
other types of observations. The main strengths are:
1 measurement of absolute current profiles, in
contrast with geostrophic profiles, which are
always relative to an unknown reference;
2 high horizontal and vertical resolution;
3.1 Shipboard Observations during WOCE
117
King, Firing and Joyce