GET-Pak

Géosciences Environnement Toulouse – Processing and Analysis Workbench

GET-Pak is an open-source Python toolbox for reproducible inland-water quality research. It converts atmospherically corrected Sentinel-2 MSI reflectance products (especially GRS NetCDF outputs) into Level-2B water-quality maps such as suspended particulate matter, turbidity and chlorophyll-a. It also extracts statistics over user-defined regions of interest (ROI) in shapefile format, builds time series, and prepares satellite–in situ matchups for validation.

GET-Pak is designed for both interactive Jupyter workflows and automated batch processing (CLI), making it suitable for single-scene exploration, regional monitoring, and large image collections on local machines or HPC systems.

Workflow

Overview of the GET-Pak processing workflow, from input reflectance products and masking to Level-2B water-quality maps, ROI statistics, and report generation.

L2B Maps

Screenshot of a Turbidity map generated with the automated workflow.

Warning

To reduce file size, GET-Pak stores output GeoTIFF values as scaled integers. Divide the stored pixel values by the corresponding scale factor to recover the physical values:

• HySPM (mg/L) and turbidity (NTU): 100
• All other products: 10,000

Automated Report

Example of an automated spreadsheet report generated from the test dataset available on Zenodo.

Note

The automated report values are automatically divided by their respective scale factors.

Citation

A recommended citation will be added following publication of the GET-Pak software article. Auxiliary data for reproducibility purposes is available at Zenodo under the link: https://doi.org/10.5281/zenodo.20933323 containing the ROI file liangzi_lake_epsg32650.shp, 3 GRS images and 3 pre-computed water masks that can be used for a test-run to generate an example report.

Related research

The current release uses methods described in the following publications:

Harmel, T., Chami, M., Tormos, T., Reynaud, N., Danis, P.-A., 2018. Sunglint correction of the Multi-Spectral Instrument (MSI)-SENTINEL-2 imagery over inland and sea waters from SWIR bands. Remote Sensing of Environment 204, 308–321. https://doi.org/10.1016/j.rse.2017.10.022

Tavares, M.H., Guimarães, D., Roussillon, J., Baute, V., Cucherousset, J., Boulêtreau, S., Martinez, J.-M., 2025. A Framework to Retrieve Water Quality Parameters in Small, Optically Diverse Freshwater Ecosystems Using Sentinel-2 MSI Imagery. Remote Sensing 17, 2729. https://doi.org/10.3390/rs17152729

Cordeiro, M.C.R., Martinez, J.-M., Peña-Luque, S., 2021. Automatic water detection from multidimensional hierarchical clustering for Sentinel-2 images and a comparison with Level 2A processors. Remote Sensing of Environment 253, 112209. https://doi.org/10.1016/j.rse.2020.112209

TL;DR Install GET-Pak

Important

GET-Pak depends on GDAL. Using a Miniconda environment is therefore strongly recommended unless you are already comfortable managing native geospatial dependencies.

Create and activate a Conda environment with Python 3.10:

conda create --name gpk310 python=3.10
conda activate gpk310

Install GDAL before installing GET-Pak to avoid dependency-resolution errors:

conda install gdal

Clone or download the GET-Pak repository, enter its root directory, and install the required Python dependencies:

pip install -r requirements.txt

Some system-dependent packages are intentionally not included in requirements.txt. Install them separately:

conda install h5py libgdal-netcdf

Finally, install GET-Pak in editable mode:

pip install -e .

Done! You should now be able to verify the installation with a quick version check:

getpak --version

or:

getpak -v

If all went well, you should see something like:

(gpk310) user@MACHINE:~$ getpak -v

            _..._
          .'     '.      _
         /    .-""-\   _/ \ 
       .-|   /:.   |  |   | 
       |  \  |:.   /.-'-./ 
       | .-'-;:__.'    =/  ,ad8888ba,  88888888888 888888888888                          88
       .'=  *=|CNES _.='  d8"'    `"8b 88               88                               88
      /   _.  |    ;     d8'           88               88                               88
     ;-.-'|    \   |     88            88aaaaa          88        8b,dPPYba,  ,adPPYYba, 88   ,d8
    /   | \    _\  _\    88      88888 88"""""          88 aaaaaa 88P'    "8a ""     `Y8 88 ,a8"
    \__/'._;.  ==' ==\   Y8,        88 88               88 """""" 88       d8 ,adPPPPP88 8888[
    /|\  /|\ \    \   |   Y8a.    .a88 88               88        88b,   ,a8" 88,    ,88 88`"Yba,
   / | \/ | \/    /   /    `"Y88888P"  88888888888      88        88`YbbdP"'  `"8bbdP"Y8 88   `Y8a
  /  | || |  /-._/-._/                                            88
             \   `\  \                                            88
              `-._/._/
                        
GET-Pak version: 0.1.4

Tip

You can test-run GET-Pak with the files provided here (and remember to fix the paths inside your settings.ini before you run).

Running GET-Pak

Important

Review and adapt settings.ini before launching the workflow.

After installation, run the complete settings-driven workflow:

getpak run

The above command loads the settings.ini file included in the installation directory by default.
Or you can also point to a customized settings.ini file:

getpak run -c /path/to/your/customized_settings.ini

Run only the L2B processing step:

getpak l2b

Run only the report/time-series extraction step:

getpak report

You can also run the complete settings-driven workflow from the repository root:

python main.py run

Because the main.py is only a wrapper the automation module can also be launched directly, but that won't change performance:

python -m getpak.automation

If you liked GET-Pak or it supports your research, giving the repository a ⭐ can help others discover the project.