First, velocities were passed through a MATLAB built-in one dimensional smoothing filter (filter.m) with a window size of ten. Second, the difference was taken between the raw and smoothed time series and the standard deviation of that difference was calculated. Third, points with a difference greater than three times the standard deviation were identified, and replaced with the smoothed signal value at the same time. The first, second, and third, steps were then repeated three times, with the previously de-spiked signal replacing the input each time, which produced final, quality controlled, de-spiked and smoothed de-spiked time series of velocities.De-spiked and smoothed de-spiked records of 3D velocity data, time, and depth information were re-saved in MATLAB readable ‘.mat’ files. De-spiked and smoothed de-spiked velocity data were treated with an additional quality control step.
A measure of signal quality, called Correlation, was recorded by the Vectrino profiler along with the raw data. Correlation is the measure of the data correlation as returned by each of the three acoustic beams of the profiler. Time points with correlation less than 70% are replaced in the record with non-number (NaN) values.Quality controlled records of de-spiked and smoothed de-spiked velocity were saved in MATLAB readable ‘.mat’ files.