ml_toar_station_classification

TOAR-classifier v2: A data-driven classification tool for global air quality stations

This notebook implements various machine learning approaches to obtain an objective station classification for global air quality monitoring stations as described in [1]. It has been developed in support of the international Tropospheric Ozone Assessment Report initiative, phase 2 (TOAR-II) [2]. TOAR has implemented a terabyte-scale database for global air quality data [3] with multiannual time series from over 23,000 stations. The objective of the station classification performed in this notebook is to create objective labels for the measurement sites as "urban", "suburban", or "rural" based on various features that provide hints of the characteristics of a station location. To this end, the machine learning models implemented here make use of the extensive metadata in the TOAR database, in particular the "global metadata" that is derived from various Earth Observation satellite data products (for details, see [1])

Files

• data/: Contains all datasets used in this work, including machine learning model predictions for station categories
• figures/: contains all the figures.
• TOAR-classifier_v2.ipynb: Jupyter notebook containing the implementation.
• requirements.txt: Lists all required Python packages.

Run the Code

Note: This has been tested on Ubuntu 24.04.

1. Install Python 3 if not already installed (most Linux systems have Python pre-installed).
2. Install Jupyter Notebook:
   • pip install notebook (for Jupyter Notebook) or
   • pip install jupyterlab (for JupyterLab).
3. clone the project by running the following command
   • git clone https://gitlab.jsc.fz-juelich.de/esde/toar-public/ml_toar_station_classification.git
4. Change directory to ml_toar_station_classification
   • cd ml_toar_station_classification
5. Creat virtual environment
   • python -m venv TOAR-classifier_v2 # feel free to change the virtual environment as convenient
6. Activate the created venv
   • python -m ipykernel install --user --name=TOAR-classifier_v2 --display-name "Python (TOAR-classifier_v2)"
7. Install required package
   • open jupyter notebook, jupyter-notebook and select kernel TOAR-classifier_v2
   • Install all the required packages for the project by uncommenting the first cell in the notebook and running the cell
8. Run the code cell by cell.

[1] Our paper https://egusphere.copernicus.org/preprints/2025/egusphere-2025-1399/

[2] a reference to https://igacproject.org/activities/TOAR/TOAR-II

[3] a reference to https://toar-data.fz-juelich.de

Citation

If you use this please cite

@article{Mache2025TOARClassifier, author = {Ramiyou Karim Mache and Sabine Schröder and Michael Langguth and Ankit Patnala and Martin G. Schultz}, title = {TOAR-classifier v2: A data-driven classification tool for global air quality stations}, year = {2025}, note = {Correspondence: Ramiyou Karim Mache (k.mache@fz-juelich.de)}, url = {} }