Interactive web explorer for Google's Groundsource dataset — a collection of 2.6 million flood events extracted from news articles across 150+ countries (2000–2026).
- Heatmap — geographic density visualization of flood events on a dark-themed world map
- Timeline — bar chart showing monthly event counts with interactive range selection
- Animation — step through months to see how flood reporting evolves over time
- Brush selection — drag handles on the timeline to filter the map to a specific date range
- Detail view — zoom in to see per-event geometries with popups for UUID, dates and area
- Dataset info overlay — in-app description of methodology, validation and references
The application can be explorer via https://tberends.github.io/google-groundsource-explorer/
The raw dataset (groundsource_2026.parquet, 636 MB) is not included in the repository. The site uses pre-processed JSON files in the data/ directory generated by the preprocessing scripts.
- Download the parquet file from Zenodo and place it in the repo root
- Install dependencies:
pip install -r requirements.txt - Generate timeline and heatmap data:
python scripts/preprocess.py - Generate per-tile geometry detail files:
python scripts/preprocess_tiles.py
Mayo, R., Zlydenko, O., et al. (2026). Groundsource: A Dataset of Flood Events from News. Google Research.