bottom, and (2) in deep water, for each major ship
acceleration such as stopping on station, by comparing the acceleration measured by GPS with the
acceleration relative to a water layer (assumed to
be spatially uniform) measured by the ADCP
(Joyce, 1989; Pollard and Read, 1989). Individual
estimates by each of these methods is noisy, particularly when SA is in effect, so many estimates must
be averaged to get a usable calibration factor. It is
possible to see slow (week-to-week) changes in
compass error, but no more than a very crude
correction for the time-variable heading error can
be made.
An opportunity to measure and therefore compensate for this error was again provided by GPS.
Signals are encoded on a carrier wave of approximately 20 cm wavelength. By examining the relative phase of the carrier at a pair of antennas, it is
possible to determine their relative position with
an accuracy of a few millimetres after resolving the
integer wavelength ambiguities.
The first attempt to measure heading using this
technique on a WHP cruise was in the Pacific in
May 1991 on RV Thomas Washington. Developed
by Ramon Cabrera and Eric Firing, the system
used two commercial GPS receivers with antennas
mounted 1–2 m apart. Gyrocompass heading helped
to resolve the ambiguity inherent in the phase measurements. This system was used on several ships
and other WHP cruises until it was supplanted by
the commercially available Ashtech GPS 3DF, a
four-antenna GPS attitude sensor (King and Cooper,
1993; Griffiths, 1994). The 3DF could track six
satellites on each antenna. Changing constellations, multipath interference, and occasional
shading of an antenna by superstructure meant
coverage was far from continuous, and varied
between ships depending on the detailed geometry
of the installation. Poor choice of setup parameters
in the receiver (too strict or too slack settings for
internal quality checks) could also have disastrous
effects. A data return of 70% was typical, leaving
significant gaps, sometimes several hours, without
a heading determination. The successor to the 3DF
was the ADU2. This had improved algorithms and
the ability to track up to 12 satellites simultaneously. With the arrival of the ADU2, coverage
became routinely near 100%.
The 3DF and ADU2 revolutionized shipboard
ADCP data quality by solving the single biggest
instrumental problem: the uncertainty in heading.
With a well-functioning GPS attitude sensor,
the total heading uncertainty can be reduced to
0.1° or less, bringing the corresponding velocity
uncertainty below 1 cm s
91
. 3DF installations on
research vessels have shown that gyrocompass
errors are typically <1–2° in low and middle latitudes, and transient errors can reach 5–10° peakto-peak in high latitudes. Without correction from
a GPS attitude sensor, the high-latitude errors are
doubly crippling – they are very large, and the
ocean currents are often weak (e.g. Heywood
et al., 1999). Errors include damped oscillations at
the Schuler period, 84 minutes, excited by accelerations of the ship; an offset proportional to the
northward velocity component of the ship, for
which the built-in compensation is not always
effective; and other short- and long-term drifts for
which we know of no model.
Other sources of error in SADCP velocities
Now that the problem of heading error has been
solved, and position fixes accurate to 10 m or better are widely available, other sources of error in
SADCP-measured currents gain attention and significance. Alderson and Cunningham (1999) show
that the hitherto ignored errors caused by pitch
and roll bias (most systems do not use any pitch or
roll estimates in vector-averaging the single-ping
velocity profiles) are significant in applications
such as the calculation of transport through Drake
Passage. They also note that errors due to correlations among pitch, roll and heave (Kosro, 1985)
may now be detectable. Firing (unpublished data),
using pitch and roll measured by a 3DF and heave
from the single-ping vertical velocity component
measured by the ADCP, has estimated such errors
as occasionally reaching 5 cm s
91 during rough
weather on a North Atlantic WHP cruise of the
RV Knorr. Another rough-weather error, a much
larger bias in the direction of motion of the ship,
was identified more than a decade ago, but
remains a problem on most ships. There has been
a variety of speculation about the explanation, but
none that the present authors find convincing. A
major source of small-scale shear error is the presence of strong scattering layers. As the sound from
a forward-looking beam, for example, encounters
a scattering layer (or the bottom), the part of the
beam with the shortest range and earliest return is
closer to vertical than the nominal beam angle; the
angle of encounter increases with time to a larger
SECTION 3 NEW WAYS OF OBSERVING THE OCEAN
114
acceleration such as stopping on station, by comparing the acceleration measured by GPS with the
acceleration relative to a water layer (assumed to
be spatially uniform) measured by the ADCP
(Joyce, 1989; Pollard and Read, 1989). Individual
estimates by each of these methods is noisy, particularly when SA is in effect, so many estimates must
be averaged to get a usable calibration factor. It is
possible to see slow (week-to-week) changes in
compass error, but no more than a very crude
correction for the time-variable heading error can
be made.
An opportunity to measure and therefore compensate for this error was again provided by GPS.
Signals are encoded on a carrier wave of approximately 20 cm wavelength. By examining the relative phase of the carrier at a pair of antennas, it is
possible to determine their relative position with
an accuracy of a few millimetres after resolving the
integer wavelength ambiguities.
The first attempt to measure heading using this
technique on a WHP cruise was in the Pacific in
May 1991 on RV Thomas Washington. Developed
by Ramon Cabrera and Eric Firing, the system
used two commercial GPS receivers with antennas
mounted 1–2 m apart. Gyrocompass heading helped
to resolve the ambiguity inherent in the phase measurements. This system was used on several ships
and other WHP cruises until it was supplanted by
the commercially available Ashtech GPS 3DF, a
four-antenna GPS attitude sensor (King and Cooper,
1993; Griffiths, 1994). The 3DF could track six
satellites on each antenna. Changing constellations, multipath interference, and occasional
shading of an antenna by superstructure meant
coverage was far from continuous, and varied
between ships depending on the detailed geometry
of the installation. Poor choice of setup parameters
in the receiver (too strict or too slack settings for
internal quality checks) could also have disastrous
effects. A data return of 70% was typical, leaving
significant gaps, sometimes several hours, without
a heading determination. The successor to the 3DF
was the ADU2. This had improved algorithms and
the ability to track up to 12 satellites simultaneously. With the arrival of the ADU2, coverage
became routinely near 100%.
The 3DF and ADU2 revolutionized shipboard
ADCP data quality by solving the single biggest
instrumental problem: the uncertainty in heading.
With a well-functioning GPS attitude sensor,
the total heading uncertainty can be reduced to
0.1° or less, bringing the corresponding velocity
uncertainty below 1 cm s
91
. 3DF installations on
research vessels have shown that gyrocompass
errors are typically <1–2° in low and middle latitudes, and transient errors can reach 5–10° peakto-peak in high latitudes. Without correction from
a GPS attitude sensor, the high-latitude errors are
doubly crippling – they are very large, and the
ocean currents are often weak (e.g. Heywood
et al., 1999). Errors include damped oscillations at
the Schuler period, 84 minutes, excited by accelerations of the ship; an offset proportional to the
northward velocity component of the ship, for
which the built-in compensation is not always
effective; and other short- and long-term drifts for
which we know of no model.
Other sources of error in SADCP velocities
Now that the problem of heading error has been
solved, and position fixes accurate to 10 m or better are widely available, other sources of error in
SADCP-measured currents gain attention and significance. Alderson and Cunningham (1999) show
that the hitherto ignored errors caused by pitch
and roll bias (most systems do not use any pitch or
roll estimates in vector-averaging the single-ping
velocity profiles) are significant in applications
such as the calculation of transport through Drake
Passage. They also note that errors due to correlations among pitch, roll and heave (Kosro, 1985)
may now be detectable. Firing (unpublished data),
using pitch and roll measured by a 3DF and heave
from the single-ping vertical velocity component
measured by the ADCP, has estimated such errors
as occasionally reaching 5 cm s
91 during rough
weather on a North Atlantic WHP cruise of the
RV Knorr. Another rough-weather error, a much
larger bias in the direction of motion of the ship,
was identified more than a decade ago, but
remains a problem on most ships. There has been
a variety of speculation about the explanation, but
none that the present authors find convincing. A
major source of small-scale shear error is the presence of strong scattering layers. As the sound from
a forward-looking beam, for example, encounters
a scattering layer (or the bottom), the part of the
beam with the shortest range and earliest return is
closer to vertical than the nominal beam angle; the
angle of encounter increases with time to a larger
SECTION 3 NEW WAYS OF OBSERVING THE OCEAN
114