Skip to content

Latest commit

 

History

History
178 lines (156 loc) · 8.97 KB

Procedure.md

File metadata and controls

178 lines (156 loc) · 8.97 KB

Procedure

CTD/sampling stations that constitute a section are designated by a station.list file. Repeat the process in 1. to prepare station.list for each occupation. We use P06 section as an example.

1. Prepare station.list

1-1. Raw data in Matlab format and station list

  1. Download and unzip all CTD files necessary to form one complete section. Both whp_netcdf and exchange are acceptable. For example, three zip files p06e_nc_ctd.zip, p06c_nc_ctd.zip, and p06w_nc_ctd.zip are necessary for the P06 1992 section. Unzip the archive in one directory (say, work/P06/1992/). For whp_netcdf format, use read_ctd_nc.m in the next step. For exchange format, use read_ctd_exchange.m.
  2. Start Matlab and run read_ctd_nc or read_ctd_exchange depending on the downloaded format. In this example, output goes to P06/1992/p06_1992.mat.
>> read_ctd_exchange('work/P06/1992/', 'work/P06/p06_1992');
>> load 'work/P06/1992/p06_1992.mat'

1-2. Correspondence with existing products

  1. (optional. When JOA data is available for this occupation;) Use JOA application to convert The "Best" Vertical Section Data into the CSV format using File → Export as Spread Sheet with comma separation. On Mac/Linux, you might need to convert newline characters of the CSV file (e.g. nkf -Lu).
>> flagJ = findJstations(stations, 'work/P06/JOA/P06_1992_bottle.csv');

Recent JOA data are provided in CSV -- no conversion is needed. For these CSV data, apply

>> flagJ = findJOAstations(stations, 'work/A25/A25_OVIDE_2004_bot_clean_edited_hy1.csv');

to obtain the flag for JOA file.

  1. (optional. When Purkey's product is available for this occupation;) Find stations in Purkey's product.
>> load 'work/P06/ctd_all_gridded_P06.mat' % Purkey's product
>> flagP = findPstations(stations, D_ctd(1));
  1. (optional. When WOCE Atlas header is available for this occupation;) Find stations in WOCE Atlas. For example, the station file for the WOCE occupation of P06 is available here.
>> flagA = findAstations(stations, 'work/P06/Atlas/P06_sta_bdep.txt');

At this stage, it is likely a few (or more!) warnings pop out which have to be manually checked. See P06/README.md, for example.

1-3. Edit station list

  1. Output initial version.
>> station_list(stations, flagJ, flagP, flagA, 'P06/1992/p06_1992.list');

If any flag is missing, use zeros(1,length(stations)) as a dummy.

The station list's prepared above are the seed of the clean data. Those stations listed in the station list will be incorporated in the output. Those stations commented out will not.

  1. Use your favourite editor to edit the station list (p06_1992.list). Comment out (by # or %) those stations to be excluded. Output from station_check.m might be useful where station distance is calculated and a simple station map is shown.

2. Clean reported data

2-1. Salinity offset and user defined interpolator

Edit configuration.m. For each section, the user needs to define the salinity batch offset. It is a function of k. Here k is the number shown in the first column of the station list and not the original station number.

It is also necessary to define Matlab functions used in the vertical and horizontal interpolations. Each function must have the following argument list;

interpolated = vertical_interpolator(measured_pressure_p1(:), ...
                                     measured_variable_on_p1(:), ...
                                     interpolating_pressure_grid(:))

and

[interpolated, bottom_depth] = horizontal_interpolator(measured_latlon(:,:), ...
                                                       measured_variable(:,:), ...
                                                       bottom_depth_uninterpolated(:), ...
                                                       pressure_grid(:), ...
                                                       latlon_grid(:))

As default options, vinterp.m and hinterp.m are provided. The former uses Hanning smoothing and the latter shape-preserving piecewise cubic interpolation following Purkey and Johnson (2010). In configuration.m, the functions are listed as;

vinterp_handle = @vinterp;
hinterp_handle = @hinterp;

using Matlab's function handle. Feel free to change.

In order to avoid unrealistic interpolated field, the field between two stations more than MAX_SEPARATION degrees part is not interpolated. This paramter is also defined in configuration.m. The default is 2 degrees.

2-2 reported_data

It is necessary to use the right configuration.m file. Here % is Shell prompt and >> is Matlab prompt.

% cp P06/configuration_1992.m configuration.m
>> D_reported(1) = reported_data('P06/1992/p06_1992.list', 'P06/1992/p06_1992.mat');
% cp P06/configuration_2003.m configuration.m
>> D_reported(2) = reported_data('P06/2003/p06_2003.list', 'P06/2003/p06_2003.mat');
% cp P06/configuration_2010.m configuration.m
>> D_reported(3) = reported_data('P06/2010/p06_2010.list', 'P06/2010/p06_2010.mat');
% cp P06/configuration_2017.m configuration.m
>> D_reported(4) = reported_data('P06/2017/p06_2017.list', 'P06/2017/p06_2017.mat');
>> save 'output/reported/I05/i05.mat' D_reported

2-3 Output in WHP Exchange format

>> reported_WHPX(D_reported(1), 'output/reported/P06/work/P06_1992');
% zip P06_1992_ct1 P06_1992_*.csv
>> reported_WHPX(D_reported(2), 'output/reported/P06/work/P06_2003');
% zip P06_2003_ct1 P06_2003_*.csv
>> reported_WHPX(D_reported(3), 'output/reported/P06/work/P06_2010');
% zip P06_2010_ct1 P06_2010_*.csv
>> reported_WHPX(D_reported(4), 'output/reported/P06/work/P06_2017');
% zip P06_2017_ct1 P06_2017_*.csv

3. Clean data in uniform grid

Matlab

>> pr_grid = [0:10:6500];
>> ll_grid = [153.4:(1/10):288.6];
% cp P06/configuration_1992.m configuration.m
>> D_pr(1) = grid_data_pressure(D_reported(1), ll_grid, pr_grid);
% cp P06/configuration_2003.m configuration.m
>> D_pr(2) = grid_data_pressure(D_reported(2), ll_grid, pr_grid);
% cp P06/configuration_2010.m configuration.m
>> D_pr(3) = grid_data_pressure(D_reported(3), ll_grid, pr_grid);
% cp P06/configuration_2017.m configuration.m
>> D_pr(4) = grid_data_pressure(D_reported(4), ll_grid, pr_grid);
>> save 'output/gridded/P06_gridded.mat' D_pr ll_grid pr_grid

ASCII

The output can be visualized using e.g. The Generic Mapping Tools.

>> gridded_xyz(D_pr(1), 'output/gridded/P06/P06_1992.xyz', ll_grid, pr_grid);
>> gridded_xyz(D_pr(2), 'output/gridded/P06/P06_2003.xyz', ll_grid, pr_grid);
>> gridded_xyz(D_pr(3), 'output/gridded/P06/P06_2010.xyz', ll_grid, pr_grid);
>> gridded_xyz(D_pr(4), 'output/gridded/P06/P06_2017.xyz', ll_grid, pr_grid);

Binary

The data are in 4-byte, big-endian, IEEE754 floats. The data can be visualized using e.g. Grid Analysis and Display System (GrADS) A tentative control file for GrADS is also output, but note that GrADS is designed for three dimensinal data set (XYZ) and not for sectional data (XZ or YZ). The control file output from gridded_bin.m overwrites y as depth to avoid unnecesary re-ordering of the binary data set.

>> gridded_bin(D_pr, 'output/gridded/P06/p06.bin', ll_grid, pr_grid);

NetCDF

To output the data to NetCDF format, use gridded_nc with the following arguments:

gridded_nc(line_id,inpath,repopath,outfname)

-- line_id (eg, 'P06') as character array -- inpath full path to location of 'gridded' folder under which will sit individual line folders with the gridded mat files -- repopath full path to the 'GO-SHIP-Easy-Ocean' repository folder that contains all the code to build the product for each line -- outfname (optional) full path to output the netcdf gridded data to. Default is the same as inpath.

>> gridded_nc('P06','output/','./');

4. Notes

4-1. When bottom depth data are missing in the CTD file

This often happens. If depth data are available in the SUM file, follow these steps. First, extract the depth data with awk. For example

# for I05, 2002
tail -146 i05_74AB20020301su.txt | awk '{print $3, $4, $16}' > i05_2002.depth

or

# for I05, 2009
cat i05_33rr20090320su.txt | awk '/ROS/ && /BO/ {print $3, $4, $16}' > i05_2009.depth

This depth file can be fed as the 3rd argument of reported_data.m.

4-2. When bottom depth data are missing and the SUM file is unavailable.

If depth is one of -999, 0, 4 (used in 74AB20020301), NaN, and no SUM is available, reported_data.m assumes the CTD casts were terminated at exactly 10 dbar above the bottom.