Flow chart for shear probes

The processing of shear-probe data can be divided into the following five major steps and these steps apply to data collected with any platform or vehicle. There are many sub-steps to these major steps. The major steps are;

Conversion to physical units.

• Determine the speed of profiling of the shear-probe through the water.
• Determine the temperature of the water.
• Convert the shear probe data samples into physical units
• Convert all other signals per the recommendations of the manufacturer of the sensor or instruments that produce these signals.

"Section" selection.

Before you can process your shear-probe data to derive the rate of dissipation you must select the section of data that you wish to process. For a vertically profiling instrument, this is traditionnally referred to as a "profile". We adopt the term "section" as this is platform independent and will include time series for dissipation estimates along horizontal or slanted trajectories as well as from moored shear probes. You must make sure that the selection is meaningful and sensible. For example, the shear probe must be profiling through the water with a speed, direction, and orientation that is fairly stationary. The selection of data can be partially automated by requiring that the kinematics of your instrument achieve certain minimum criteria. The steps to section selection are as follows:

• Choose the minimum speed of profiling.
• Choose the direction of the vertical velocity of the profiler.
• Choose the minimum depth.
• Choose the maximum pitch and roll of the profiler.
• Choose the minimum duration over which the minimum speed through maximum pitch and roll must be satisfied.

Choosing the processing parameters.

• Choose the length of data (in meters) used for each dissipation estimate – diss-length.
• Choose the lowest wavenumber of spectral estimation – fft-length.
• Translate diss-length and fft-length into duration (time).
• Round these up to nearest power-of-two number of samples.
• Choose a high-pass filter cut-off frequency to be consistent with (b).
• Choose de-spiking parameters.
• Choose vibration-coherent noise removal.

Compute the dissipation rate estimates.

The following items break down the derivation of the turbulent dissipation rate of kinetic energy ([math]\displaystyle{ \varepsilon }[/math]). Explanations for each step can be found after.

1. Extract the section defined in step 2 ("Section" selection).
2. High-pass filter the shear-probe and (optionally) the vibration data.
3. Identify each diss-length segment in the profile.
4. De-spike the shear-probe data, and track the fraction of data affected by de-spiking within each diss-length segment. This will become a quality-control metric.
5. Calculate the frequency spectra and cross-spectra of shear and vibrations for each diss-length segment.
6. Extract the original and the vibration-coherent clean shear-probe frequency spectra with the Goodman algorithm.
7. Correct shear and vibration frequency spectra for the high-pass filter.
8. Correct the cleaned frequency spectra for the bias induced by the Goodman algorithm.
9. Convert the frequency spectra into wavenumber spectra using the mean speed for each diss-length segment. That is, make the wavenumber [math]\displaystyle{ \begin{equation}k=f/U\end{equation} }[/math] and the wavenumber kinetic energy spectrum [math]\displaystyle{ \begin{equation}E(k)=UE(f)\end{equation} }[/math] .
10. Correct the spectra of shear for the wavenumber response of the shear probe.
11. Apply an iterative spectral integration algorithm to estimate the variance of shear.
12. Calculate the turbulent dissipation rate by multiplying the shear variance by [math]\displaystyle{ \begin{equation} \frac{15}{2}\nu\end{equation} }[/math] where [math]\displaystyle{ \nu }[/math] is the temperature-dependent kinematic viscosity.
13. Determine the figure of merit (FM) for each shear-probe spectrum using the method described here.
14. Calculate the expected variance of each dissipation estimate using the method described here.

Apply quality-control metrics.

Spikes in epsilon estimates arise from a number of causes such as collisions of sensor tips with suspended particles (e.g. detritus, plankton, jelly fish, seaweed), electronic noise due to other sensors, or mechanical platform vibrations. This section describes quality control measures and its coding.

In a first step, epsilon estimates are flagged based on quality control metric and disagreement between dissipation estimates from redundant sensors.

1. Quality-control metrics (see also Processing Steps section V) that accompanied dissipation estimates are used to flag individual estimates. In particular, quality control thresholds for

   • figure of merit (FM)
   • fraction of shear-probe data altered by the de-spiking routine
   • number of iterations of the de-spiking routine required to clean the data
   • (more to be discussed)

2. Agreement between dissipation estimates from redundant sensors (i.e. two or more shear probes) does not exist.

Flagged data will receive quality control coding 4.

In a second step of quality control, a review of ensembles that have been flagged is performed. Individual shear spectra, associated tilt and acceleration data and microstructure thermistor data/spectra are visually examined for consistency. Previously flagged data that appears to be good data will receive quality control coding 2.

Details for QC coding can be found here.

---

Return to Shear Probe Welcome Page