The system on board the Malcolm Baldrige is a 150 kHz RD Instruments unit with a hull-mounted transducer.   Standard data acquisition and processing procedures are described in Wilson and Routt (1992), along with data from this cruise not included in this paper.  Single-ping profiles of water velocity relative to the ship are collected using eight meter vertical bins, with the center of the first bin at 14 meters. These profiles are corrected for sound velocity at the transducer, rotated into north and east components using the ship's gyrocompass heading, and averaged into one minute ensembles (approximately 45 pings) for storage.  Assuming an unbiased ping-to-ping standard deviation of 15 cm/s for this configuration (RD Instruments manufacturer's specifications) one minute ensembles would have estimated errors of less than 3 cm/s. The fifteen minute averages used for analysis would have errors of less than 1 cm/s.
	The one-minute profiles are corrected for transducer-gyrocompass misalignment using a method similar to that described by Pollard and Read (1990).  Depth-averaged (10 - 150 m) absolute velocities measured on station (ship moving slowly) are compared with values measured just before or after the station (ship moving at cruising speed).   Measurements from these comparisons spread over the duration of the cruise give calibration coefficients of  the misalignment angle scaling factor .  While there are numerous nearly random errors that could contribute to these calculated  misalignment values (short period gyrocompass oscillations, or actual differences in currents between station arrival and departure), a longer-lasting bias of the magnitude of the uncertainty could lead to large (3-7 cm/s)  systematic errors in velocity, which must be considered in determining the error bounds of transport calculations.
 	Errors are also introduced into estimates of absolute water velocity by uncertainties in
ship velocity. Global Positioning System (GPS) data (time, latitude, longitude, speed/course over
ground) are logged at 5 second intervals, and all Transit satellite fixes are recorded. In August -  September 1989, GPS data were available only about 60% of the time due to incomplete 
satellite coverage (Figure 2b). When available, GPS position data are filtered and differentiated to estimate mean ship velocity over each one-minute ADCP ensemble. GPS speed and heading data are also filtered and averaged over the ensemble interval; these data are used as a quality check to edit ship speeds derived from position data.  Based on an estimated GPS position uncertainty of 25 m (these GPS data were not subjected to Selective Availability degradation) fifteen minute averages of ship speed used for analysis are accurate to 3 - 4 cm/s; these estimates should be unbiased.
	When GPS data are not available, ADCP velocity ensemble averages are referenced using Transit fixes. ADCP shear measurements are used to calculate the mean velocity at a deep (roughly 100 to 150 meters) reference level between two Transit positions. Transit fixes may be in error by as much as 100 meters.  Fixes spaced 60 minutes or more apart (typical time between satellite fixes in the study area) therefore can  result in errors as large as 5 cm/s. Furthermore, resolution at less than  the period between satellite fixes is lost, since a single reference level velocity is calculated for the entire region between Transit fixes.   After referencing, the one-minute averages of absolute velocity are further averaged into 15-minute means, which are the basis for all of the velocity data presentations.