euroargodev / terrain-following Goto Github PK

This is a readme file for the terrain-following interpolation algorithm 
documented in Yamazaki+(2020) https://doi.org/10.1029/2019JC015406.

The codes are originally designed to be applied for Argo data in *_prof.nc format fetched from Argo Global Data Assembly Center.

interpolation_v2.py
    ...the interpolation algorithm with some plotting function. This is slightly modified from Yamazaki+(2020) for clarity of the scheme and customizability.

Test trajectory data (testData.csv) is included in this folder, which can be directly loaded to interpolation_v2.py.
The result of interpolation for test data is shown in two .png files.

Following codes would help pre-processing of the trajectory data:

nc2csv.py
    ...extract trajectory data and position quality flag (qpos) from netcdf file in the Argo profile format (*_prof.nc).

process.py
    ...change id for data seeming sequential as trajectory by id-cast but problematic in lon-lat coords.	

For a trial, you can interpolate testData.csv with the GEBCO 1arcmin topography (GRIDONE_1D.nc),
which you may find at: https://drive.google.com/drive/folders/14RizKyM064OsjEo5J0u0RtzdveD_mU-j?usp=sharing
The result of interpolation is presented in two .png files, in which interpolated positions (green) seem to follow the isobaths.
(Yellow points are linearly interpolated positions, while not all position-lacking data have linearly interpolated positions by default.)

This scheme is slightly modified from Yamazaki+(2020), such that the "backward revision" for interpolated positions obtained by forwarding interpolation (Fig.A, panels c-d) is replaced with the "weighted average" for the forwarding and backwarding interpolations, with a weighting function defined from inverted distances from the first and last points of the interpolation section. This modification greatly reduces "asymmetry" in interpolated positions, which can result in sharp artificial curves near the end of the interpolation section to connect to the positioned point. By customizing the weighting function and tuning parameters (the length of search range and the searching resolution), you may optimize the scheme for your purpose.

Correspondence: Kaihe Yamazaki ([email protected])
written in 2019 September, modified in 2022 July.