velocity of the package over the ground. When the
velocity is small, the bottom-contaminated velocity
estimates are similar to the surrounding uncontaminated estimates, and the interference is not visually
evident in the calculated velocity profile. When the
package velocity is larger than 10 cm s
91 or so, the
interference shows up as a velocity spike and/or
offset in the processed LADCP profile, if no special
editing is done. Editing out the contaminated depth
range leaves a gap in the composite shear profile
and therefore an uncertain offset between the parts
of the velocity profile on either side of the gap.
The problem can be avoided entirely by using a
staggered ping sequence, so that the interference
appears in two non-overlapping depth ranges on
alternate pings. Editing out the interference then
leaves no gap in the composite shear profile. The
use of two ADCPs, one looking up and the other
down is described on a web site by Visbeck (http:
//www.ldeo.columbia.edu/~visbeck/ladcp). This also
solves the problem: the upward-looking profiler’s
data remains uncontaminated.
The accuracy of the depth-averaged velocity
depends almost entirely on the accuracy of the
position fixes at the start and end of the cast, and
on the accuracy of the time-integrated velocity of
the water relative to the package. Fix accuracy is
not a major concern, given that SA now has been
turned off, and that P/Y code fixes and DGPS
(real-time or post-processed) were widely available
during the WHP. Even when only C/A code fixes
with SA were available, the resulting uncertainty in
velocity is less than 1.6 cm s
91 for 95% of all casts
lasting 2.5 h or longer. Of greater concern is the
velocity integral, for which there are two types of
error: that of the velocity measurement itself, and
that due to gaps in the sampling. Long gaps, as
opposed to occasional ping dropouts and the normal interval between pings, are caused primarily by
interference from sound reflecting off the ocean
bottom instead of the water. Although such gaps
can last several minutes, this interval is short compared with the entire profile, and can be filled by
interpolating a low-pass filtered time series of the
water velocity relative to the package. Even if the
interpolated velocity is in error by 10 cm s
91 on
average, and the gap is 5 min, the contribution to
the depth-averaged velocity error will be only
0.3 cm s
91 for a 2.5-h profile. Therefore the most
worrying type of error is that which contributes a
bias to the velocity measurement. Of the possible
causes, we will mention two here: compass error
and ambiguity error.
Velocity errors arising from LADCP magnetic
compass errors
Self-contained ADCPs use magnetic flux-gate compasses. These can fail to measure the true orientation of the package, either because of inherent
instrument error, or because the presence of other
magnetic materials on the package (e.g. mounting
hardware or pressure cases) can generate their
own local magnetic fields or distort the earth’s
magnetic field. Compass errors can vary with tilt
as well as heading.
Compass error affects both the relative velocity
profile and the depth-averaged velocity calculation. The LADCP is much less sensitive to compass
error than a shipboard ADCP because the magnitude of the velocity error is proportional to the
speed of the water relative to the instrument,
which is usually smaller for the LADCP by a
factor of 10 or more. For example, if the horizontal speed of the water relative to the LADCP
is 20 cm s
91
, a 5° compass error will cause a
1.7 cm s
91 velocity error perpendicular to the mean
velocity. Compass accuracy varies with geographic
position and instrument tilt, becoming increasingly
problematic near the magnetic poles. Systematic
errors exceeding 5–10° sometimes can be detected
by comparing simultaneous LADCP and SADCP
profiles in their common depth range; errors up to
60° were identified on one high-latitude cruise,
but the cause was not found (Hummon and Firing:
Post-cruise compass calibrations for North Atlantic
WOCE LADCP, unpublished document); errors
were much smaller on other cruises in the same
region. With respect to compass error, the LADCP
at the end of WOCE is at the same stage the
SADCP was prior to WOCE: the error can be
large, and it is very difficult to measure. Unfortunately, no technical solution analogous to a GPS
heading sensor is presently known.
Ambiguity error
Much of the WHP LADCP data set was obtained
using 150 kHz broadband sonars (BB-150) from
RD Instruments. Unlike the narrow-bandwidth
sonars with which most of the WHP shipboard
ADCP data were collected, the BB-150 transmits
pulses modulated with repeated pseudorandom
codes to estimate Doppler phase shift in the range
<180° (Pinkel and Smith, 1992). The velocity
SECTION 3 NEW WAYS OF OBSERVING THE OCEAN
116
velocity is small, the bottom-contaminated velocity
estimates are similar to the surrounding uncontaminated estimates, and the interference is not visually
evident in the calculated velocity profile. When the
package velocity is larger than 10 cm s
91 or so, the
interference shows up as a velocity spike and/or
offset in the processed LADCP profile, if no special
editing is done. Editing out the contaminated depth
range leaves a gap in the composite shear profile
and therefore an uncertain offset between the parts
of the velocity profile on either side of the gap.
The problem can be avoided entirely by using a
staggered ping sequence, so that the interference
appears in two non-overlapping depth ranges on
alternate pings. Editing out the interference then
leaves no gap in the composite shear profile. The
use of two ADCPs, one looking up and the other
down is described on a web site by Visbeck (http:
//www.ldeo.columbia.edu/~visbeck/ladcp). This also
solves the problem: the upward-looking profiler’s
data remains uncontaminated.
The accuracy of the depth-averaged velocity
depends almost entirely on the accuracy of the
position fixes at the start and end of the cast, and
on the accuracy of the time-integrated velocity of
the water relative to the package. Fix accuracy is
not a major concern, given that SA now has been
turned off, and that P/Y code fixes and DGPS
(real-time or post-processed) were widely available
during the WHP. Even when only C/A code fixes
with SA were available, the resulting uncertainty in
velocity is less than 1.6 cm s
91 for 95% of all casts
lasting 2.5 h or longer. Of greater concern is the
velocity integral, for which there are two types of
error: that of the velocity measurement itself, and
that due to gaps in the sampling. Long gaps, as
opposed to occasional ping dropouts and the normal interval between pings, are caused primarily by
interference from sound reflecting off the ocean
bottom instead of the water. Although such gaps
can last several minutes, this interval is short compared with the entire profile, and can be filled by
interpolating a low-pass filtered time series of the
water velocity relative to the package. Even if the
interpolated velocity is in error by 10 cm s
91 on
average, and the gap is 5 min, the contribution to
the depth-averaged velocity error will be only
0.3 cm s
91 for a 2.5-h profile. Therefore the most
worrying type of error is that which contributes a
bias to the velocity measurement. Of the possible
causes, we will mention two here: compass error
and ambiguity error.
Velocity errors arising from LADCP magnetic
compass errors
Self-contained ADCPs use magnetic flux-gate compasses. These can fail to measure the true orientation of the package, either because of inherent
instrument error, or because the presence of other
magnetic materials on the package (e.g. mounting
hardware or pressure cases) can generate their
own local magnetic fields or distort the earth’s
magnetic field. Compass errors can vary with tilt
as well as heading.
Compass error affects both the relative velocity
profile and the depth-averaged velocity calculation. The LADCP is much less sensitive to compass
error than a shipboard ADCP because the magnitude of the velocity error is proportional to the
speed of the water relative to the instrument,
which is usually smaller for the LADCP by a
factor of 10 or more. For example, if the horizontal speed of the water relative to the LADCP
is 20 cm s
91
, a 5° compass error will cause a
1.7 cm s
91 velocity error perpendicular to the mean
velocity. Compass accuracy varies with geographic
position and instrument tilt, becoming increasingly
problematic near the magnetic poles. Systematic
errors exceeding 5–10° sometimes can be detected
by comparing simultaneous LADCP and SADCP
profiles in their common depth range; errors up to
60° were identified on one high-latitude cruise,
but the cause was not found (Hummon and Firing:
Post-cruise compass calibrations for North Atlantic
WOCE LADCP, unpublished document); errors
were much smaller on other cruises in the same
region. With respect to compass error, the LADCP
at the end of WOCE is at the same stage the
SADCP was prior to WOCE: the error can be
large, and it is very difficult to measure. Unfortunately, no technical solution analogous to a GPS
heading sensor is presently known.
Ambiguity error
Much of the WHP LADCP data set was obtained
using 150 kHz broadband sonars (BB-150) from
RD Instruments. Unlike the narrow-bandwidth
sonars with which most of the WHP shipboard
ADCP data were collected, the BB-150 transmits
pulses modulated with repeated pseudorandom
codes to estimate Doppler phase shift in the range
<180° (Pinkel and Smith, 1992). The velocity
SECTION 3 NEW WAYS OF OBSERVING THE OCEAN
116