pyturb

A microstructure processing toolbox for Rockland Scientific microstructure instruments.

Installation

Install using pip.

Usage

Pyturb is primarily a CLI but can also be integrated into scripts or notebooks. The processing should be run in the following order:

1. Convert p files to netCDF. Optionally, merge converted p files.
2. Calculate turbulence estimates on a per-profile basis.
3. Bin estimates onto a regular grid.

The three steps summarized above are applied in the CLI using subcommands p2nc, merge, eps, and bin.

p2nc - convert P files

Convert Rockland binary P-files to NetCDF format:

pyturb p2nc ./path/to/raw_data/*.p -o ./converted/

Note that unlike the ODAS toolbox, this conversion does not apply a velocity scaling to the microstructure shear or temperature gradient. The units of these variables are different to their ODAS counterparts. The scaling is applied later.

The merge utility enables merging of netcdf files (e.g. pyturb merge -o ./merged/merged.nc ./converted/*.nc). This may be useful in the case where files prematurely end before instrument recovery.

eps - calculate the dissipation rate

Estimate turbulent kinetic energy dissipation rate (epsilon) from converted NetCDF files:

pyturb eps ./converted/*.nc -o ./eps_output/

The eps command automatically detects multiple profiles within each input file. Output files are named {input_stem}_p{NNN}.nc for each profile found.

A selection of the option used:

• --diss-len: Dissipation window length in seconds (default: 4.0)
• --fft-len: FFT segment length in seconds (default: 1.0)
• --min-speed: Minimum speed threshold in m/s (default: 0.2)
• --direction: Profile direction to process: down, up, or both (default: down)
• --peaks-height: Minimum peak height for profile detection in dbar (default: 25.0). Relies on profinder
• --aux: Auxiliary NetCDF file with glider flight data (lat, lon, T, S)

Example processing just up casts:

pyturb eps -o ./eps_output/ --direction up ./converted/*.nc

bin - bin average the data

Bin epsilon estimates by depth and concatenate into a single file:

pyturb bin -o binned_profiles.nc --bin-width 2.0 --dmax 500 ./eps_output/*.nc

Options:

• --bin-width: Depth bin width in meters (default: 2.0)
• --dmin/--dmax: Depth range for binning (default: 0-1000 m)
• --pressure: Bin by pressure instead of depth

Processing Methods

Preprocessing Pipeline

Before computing epsilon, profiles undergo:

1. Low-pass filtering of speed (or dP/dt-derived speed) to remove high-frequency noise
2. Shear signals are scaled by 1/U² and temperature gradients by 1/U to convert to physical units
3. Iterative removal of outliers from shear and temperature gradient signals

Shear Spectrum Processing

The dissipation rate is estimated by fitting shear spectra to the Nasmyth spectrum:

1. Shear probe signals are converted to wavenumber spectra using Welch's method with overlapping FFT windows
2. A single-pole transfer function correction is applied to account for the spatial averaging of the shear probe
3. Epsilon is estimated by fitting the observed spectrum to the theoretical Nasmyth spectrum in the inertial subrange
4. Unresolved high-wavenumber variance is accounted for using the integrated Nasmyth spectrum

Python API

from pyturb.processing import batch_compute_epsilon, bin_profiles
from pyturb.pfile import batch_convert_to_netcdf

# Convert P-files
batch_convert_to_netcdf('./path/to/raw_data/*.p', output_dir='./converted/')

# Compute epsilon
batch_compute_epsilon('./converted/*.nc', output_dir='./eps/', diss_len_sec=4.0)

# Bin profiles
bin_profiles('./eps/*.nc', output_file='binned.nc', bin_width=2.0)