RBRproc

Jan 23, 2018 update:

RBRproc was written before RSKtools, the RBR Ltd official Matlab software toolbox, had the ability to post-process RBR CTD data. RSKtools v2.0.0, released in July 2017, contains a number of functions for CTD data post-processing, and is now in my opinion a more useful toolbox than RBRproc. Futhermore, as of v2.0.0, RSKtools stores profile data in a different manner than older versions, and RBRproc has not been updated to handle the changes. Visit this link to download RSKtools.

RBRproc is still a useful way to handle CTD data in general, and so long as the CTD structure is formatted in a particular way, then RBRproc will work well (I still use it for Seabird CTD data and autonomous float data).

-- Mark / Jan 23 2018

old header:

RBRproc is a collection of Matlab routines designed to process RBR profiler data (e.g., RBR Concerto). The approach here largely resembles the processing chain used by Seabird profilers.

At the moment the toolbox is designed to use RBR's RSKtools Matlab toolbox to read raw 'rsk' sqlite files. The output structure from RSKtools is then converted into a more convenient structure. This structure is then used as input to the various RBRproc processing routines.

For the older RBR loggers that output hexadecimal files, first use Ruskin to export the file into any one of the Ruskin outputs (mat, txt, Excel, etc.). Then read that file into Matlab. Following this, insert the data into the blank structure created by the function blankRBRstruct.m. Matlab functions to read RBR hex or txt files do not exist as far as I know (volunteers?), but reading Excel files and mat files in Matlab is trivial.

Requirements

RSKtools

Official releases:

RBR Global link

Bleeding edge development version:

Bitbucket

Gibbs SeaWater Matlab tool box

TEOS-10 link

Example usage

In this example we will read an rsk file and extract the downcasts using functions in RSKtools, and then perform some basic processing with RBRproc.

% read logger data into Matlab using RSKtools
rsk = RSKopen('/full/path/to/rskfile');

% parse the file into profiles, keeping only the down casts
profiles = RSKreadprofiles(rsk,[],'down');

% recast things into a 1 x Ncast structure with easy-to-access fields.
% the field names come from rsk.channels.longName
profiles = flattenRSK(profiles);

% plot a time series of pressure to see the profiles
plot(cat(1,profiles.mtime),cat(1,profiles.Pressure)-10.1325,'.-')
title(profiles(1).fileName,'interpreter','none')

% find the maximum pressure of each profile
arrayfun(@(x) max(x.Pressure),profiles)

% extract the 4th profile for this example
profile = profiles(4);

% add coordinates to structure
profile.Latitude = 52;
profile.Longitude = -129;

% subtract a user-chosen atmospheric pressure from total pressure
patm = 9.9; % say a mid-latitude low pressure system was passing by 
profile = rmPatmRBR(profile,patm);

% despike the fluorometer and turbidity profiles.  Use an 11 point
% median filter to find the spikes, and replace them with NaN
profile = despikeRBR(profile,{'Chlorophyll','Turbidity'},'median',11,'NaN');

% low pass filter (filtfilt) T/C with running 3 pt triangular window
profile = filterRBR(profile,{'Temperature','Conductivity'},3);

% lag conductivity by 0.33 seconds (2 scans at 6 Hz) to reduce
% salinity spiking 
profile = alignRBR(profile,'Conductivity',-2/6);

% Identify scans when the descent rate was below 50 cm/s data, and
% replace with `NaN`.
profile = loopRBR(profile,'NaN',0.5);

% now calculate practical salinity
if isfield(profile,'Salinity'),
  profile = rmfield(profile,'Salinity'); % remove RBR's calculation
end

profile.PracticalSalinity = gsw_SP_from_C(profile.Conductivity,....
                                          profile.Temperature,...
                                          profile.Pressure);

% interactive function to choose start and end points of profile
% (this could use some work from a GUI expert)
profile = trimRBR(profile);

%% bin average all variables by pressure into 1 dbar bins
profile = binRBR(profile,'pressure',1);

Laundry list

1. Implement better input handling.

2. Modify despikeRBR.m to operate on blocks of data instead of full profile.

3. Make trimRBR not suck.

4. Depth vector has the incorrect length when it is calculated by loopRBR, but then binned by pressure.