The Explorer of the Seas ADCP program has posed some unique challenges compared to Research Vessel mounted systems. First, the average ship speed is 20 knots, twice that of most research vessels, and this exacerbates the error in ocean velocity caused by heading inaccuracies. Figure 1 shows an evaluation of the accuracy of the gyrocompass when compared to Ashtech heading from a cruise in February 2006. During periods of manoeuvring the gyrocompass heading is poor, with a standard deviation greater than 1 degree, but these errors decrease to 0.2 - 0.3 degrees during a steady steam, giving an estimated ocean velocity error of order 4 - 6 cm/s. In addition, the gyrocompass exhibits heading related biases which add another ~4 cm/s of uncertainty. Overall, errors are three or four times higher than would be expected from a research vessel system. In spring 2006 a heading correction device, the Ashtech antenna array, was installed on Explorer and as a result velocity errors are decreased and biases removed, except when the instrument is not logging correctly. In summary, heading inaccuracies lead to ocean velocity errors of order 5 cm/s for the OS38 and OS150 data sets between 2001 and May 2006, plus potential biases up to 5 cm/s (these biases change over time, dependent on position and heading of the ship). Since May 2006, heading corrections are applied to the data set, where available, which eliminate bias and reduce error velocities to 2-3 cm/s. We estimate errors are still higher than for research vessel mounted systems owing to bubble contamination, as explained below. Because the Ashtech system is not 100% reliable, consult the Data Quality Summary Tables to check whether heading corrections were available for a particular cruise or not.

Second, bubble contamination of the acoustic data is significant, owing to Explorer's flat-bottomed hull and high cruising speeds. Bubbles cause two significant problems: they block the acoustic signal, and they manifest as large measured velocity biases in the forward-ship direction. Figure 2 shows an example of the amount of bubble contamination for a winter cruise. It is high during periods of acceleration/decceleration and during rough sea states. Figures 3 and 4 show ocean velocity profiles for a one and a half day period during a winter cruise. In figure 3 we are using the long-term-averaged (LTA) data from the OS38. The paucity of data and high noise levels (shown as jitter) are due to bubbles. In figure 4 we have used the single-ping (ENX) data and implemented an editing algorithm for bubbles before averaging. About 50% of the data is recovered and noise is reduced by a factor of five. Hence, single-ping editing was used to produce all the Explorer data sets.