Accuracy of shipboard doppler current profiling during CODE

During the Coastal Ocean Dynamics Experiment (CODE, 1981-1982), a shipboard system was used to collect approximately 1800 hours of upper ocean current profiles from the research vessel Wecoma. The data included approximately 100 acoustic log measurements per minute of the Doppler shift in each of 32 range cells along four acoustic beams, gyroscopic measurements of ship pitch, roll and heading, and LORAN-C measurements of the ship´s position. The means by which this data set was used to estimate currents and the accuracy of those estimates are the subject of this talk (for a more detailed discussion, see (1)). After screening for gross errors, the acoustic Doppler shift was converted to relative velocity of the ocean as seen from the ship´s frame of reference. At fixed range, the spectrum of high frequency variability in relative velocity was dominated by a peak 0(20-35 cm/s) at the surface-wave encounter frequency rising above a white noise background 0(10 cm/s). Block averaging over 100 profiles reduced this variability to 0(2 cm/s) in each horizontal component. Prior to averaging, however, data had to be transformed from the ship´s time-varying reference frame into geographic coordinates. For this data set, pitch and roll corrections caused only small changes in average profiles of horizontal velocity, and were neglected in subsequent processing. More significant were potential errors from uncertainties in gyrocompass derived heading. Sequential LORAN-C positions were used to compute, and remove from each averaged profile, the velocity of the ship with respect to the Earth, yielding one estimated profile of ocean currents per minute. Within each profile, errors in LORAN-C measurements contribute vertically constant noise 0(80 cm/s) which had to be removed through additional filtering. Spectral analysis of the current estimates showed that it was reasonable to model LORAN-C errors as serially uncorrelated, in which case the optimal filtering algorithm in a constant current field was derived, and shown to be equivalent to the least-squares fit of the ship´s drift to a straight line. True variability in the currents, both spatial and temporal, limits the time over which such a linear fit is appropriate. The measured spectrum was used to determine the ave- raging interval of 30 minutes, with expected noise of 0(2-3 cm/s) in the alongshore currents and 0(4-5 cm/s) in the cross-shore currents. Finally, changes in the estimated currents during the ship´s starts, stops and turns were used to determine errors in calibration and alignment of the acoustic system and correct them in post-processing. Currents measured with the shipboard system were compared with those from the large CODE array of moored VMCMs. For ship/mooring separations of 1 km or less, correlation between moored and shipboard measurements was 0.95-0.97 for the energetic alongshore component of the current, 0.76-0.82 for the weaker cross-shore component, at depths of 20-90 m; differences between measurements had standard deviations of 4-5 cm/s, about what was expected from the independent error analysis. These differences increased rapidly as a function of ship/mooring separation, indicating some geophysical variability contributed to the observed differences. Small, but statistically significant, correlation (0.2-0.3) was also found between the ship´s speed and the difference magnitude.