Eddy-Detection-Velocity

Try to detect eddies based on the velocity field created with DIVAnd using the py-eddy-tracker.

Installation

That might be the trickiest part: installing the tool without conflict in the package versions.

Install Python

We worked with python3.11.15, other versions may work, and other will fail.

Download

The code is download from https://www.python.org/downloads/release/python-31115/ and extracted:

tar xvf Python-3.11.15.tar.xz

Compilation

cd Python-3.11.15
./configure
make
sudo make install

Set virtual environment

It is often recommanded to create a virtual environment in Python to ensure compatilibity between the installed modules.
We worked with the https://virtualenvwrapper.readthedocs.io/en/latest/, which should be installed with:

pip install virtualenvwrapper

then we can create a new environment (here called Eddy) by specifying the Python version we just installed:

mkvirtualenv --python=/usr/local/bin/python3.11 Eddy

Install module

We followed the official documentation: https://py-eddy-tracker.readthedocs.io/en/stable/installation.html.

First we install some requirements:

pip install "matplotlib<3.8" opencv-python pint polygon3 pyyaml requests scipy "zarr<3.0" netCDF4 numpy numba

then for the installation of pyEddyTracker there were some errors, so we ended up doing this:

pip install cython
pip install --no-build-isolation "netCDF4<1.6"  
pip install --no-build-isolation git+https://github.com/AntSimi/py-eddy-tracker.git

which solved the problem.

Create a Jupyter kernel

If you want to use Jupyter notebooks with the same environment (Eddy):

pip install ipykernel
python -m ipykernel install --user --name Eddy --display-name "Eddy"

so when you open Jupyter, you can now select Eddy as the kernel.

Running a detection

A full (and working) example is available in the notebook detect_eddy_medsea.