Environmental Data Pipeline & Dashboard

A Data Engineering & Scientific Computing team project at ESIEE Paris. We built a pipeline and dashboard to explore environmental radioactivity measurements alongside rainfall data in metropolitan France.

Source code & setup Video walkthrough

Data and implementation

The project combines RNM/ASNR soil and water sample measurements, Météo-France precipitation data, and a municipality gazetteer. The documented cleaned dataset contains 171,545 records across 395 municipalities, from January 2020 to January 2025.

Three CLI commands separate ingestion, transformation and presentation:

Download: Playwright-based collection and CSV ingestion into a raw-data directory.
Clean: municipality-name normalization, Lambert-93 to WGS84 coordinate conversion with pyproj, and nearest-neighbor spatial joins with scikit-learn’s BallTree.
Dashboard: a Plotly Dash application with maps, date filters, distributions and measurement-count views.

My contribution

I designed the CLI and repository structure, implemented the data ingestion and cleaning pipeline, and contributed to the dashboard layout and callbacks.

Interpretation limits

This is an exploratory data-engineering project. Radioactivity results must be compared within compatible units, radionuclides and sample media: Bq/kg and Bq/L are activity concentrations, not radiation doses. Sampling campaigns and the distance between weather and sample locations also affect interpretation. The dashboard does not establish that rainfall causes changes in radioactivity.

Dashboard time series showing the number of environmental measurements collected over time — Measurement coverage varies over time; the dashboard makes that sampling pattern visible.

Tools: Python, pandas, NumPy, Playwright, pyproj, scikit-learn, Plotly Dash.