About the data
What you are looking at, where it comes from, and what it can and cannot tell you.
What is this?
Firefinder shows weekly wildfire ignition risk along power-grid corridors for Portugal and Spain, with France coming next. The country is divided into hexagons of roughly 5 km², and every week each hexagon gets a probability that a fire starts there, computed from satellite imagery, weather, terrain and land cover.
That risk is then rolled up onto every power line corridor in the national grid and ranked, because vegetation near power lines is a leading cause of both wildfires and large-scale outages. Knowing which corridors to worry about this week is more useful than knowing which regions are generally fire-prone.
Everything is built from free public data, and the whole project is open source.
The data sources
- Sentinel-2 (ESA, via AWS Open Data)
- Raw satellite imagery at 10 to 20 m resolution. We do our own cloud masking and build monthly composites of vegetation greenness (NDVI) and moisture (NDMI). The current month is rebuilt daily as new scenes arrive.
- ERA5 reanalysis via Open-Meteo
- Daily temperature, humidity, wind and rain, with NASA POWER as a fallback source. New days are appended daily.
- EFFIS (Copernicus Emergency Management)
- Official European burnt-area perimeters since 2016. These are both the labels the model learns from and the historical fires layer in the app. Refreshed daily.
- NASA FIRMS
- Live satellite fire detections from the VIIRS instruments on the Suomi NPP and NOAA-20 satellites, with roughly 3 hours of latency and a 375 m pixel, refreshed every 15 minutes in the app. To be clear about what a detection is: a satellite seeing heat. The position is approximate to about 375 m, small or brief fires can be missed entirely, and false positives happen (industrial heat, gas flares).
- OpenStreetMap
- The power grid itself: transmission and distribution lines with voltage and operator, split into corridors of up to 5 km and named by the nearest town.
- Copernicus DEM and ESA WorldCover
- Elevation and slope from the Copernicus digital elevation model, and land-cover fractions (tree, shrub, grass, crop, built-up) from ESA WorldCover as proxies for the fuel available to a fire.
The forecasting model
The forecasts come from gradient-boosted decision trees (XGBoost) trained on nearly a decade of fire history. For each hexagon and week, the model weighs 21 factors: how dry and stressed the vegetation looks from space, the week's fire weather, drought over the last 30 and 90 days, how steep the terrain is, what kinds of fuel cover the ground, distance to power lines, and the time of the fire season.
Some honest numbers. Ignition in any given cell in any given week is rare: about 1 in 700 to 1 in 2,600 depending on the country. The model ranks risk well (ROC-AUC of 0.85 for Portugal and 0.86 for Spain, measured on held-out years it never saw during training), but absolute probabilities at base rates this low are hard. That is why the app talks in tiers (Severe, High, and so on) rather than pretending to a precision the data cannot support.
The "why" bars you see when you click a corridor or cell are SHAP attributions: the model's own accounting of which factor pushed the score up or down for that place and week.
Limitations, stated plainly: the weather used is observed, not forecast. Small fires are missing from the training labels. And corridor risk means fire risk near the line, not a claim that the line causes fires; no public European dataset attributes ignitions to grid equipment, so that distinction is beyond what open data can support.
Corridor scoring
Cells within 500 m of a power line contribute to that corridor's score, weighted as 0.65 times the worst cell plus 0.35 times the average, so a corridor through one severe spot matters more than one through many mild ones. Corridors are ranked nationally each week.
How fresh is what you see
| Layer | Updated |
|---|---|
| Risk scores (cells and corridors) | Daily at 05:00 UTC, via a public GitHub Actions run |
| Live fire detections | Every 15 minutes while the app is open |
| Historical fire perimeters | Daily |
| Grid geometry and terrain | Static |
Open source
Everything (the data pipeline, the model, and this app) lives at github.com/davidcoallier/firefinder.eu, with instructions to run the whole thing yourself. All the data sources are free and require no API keys. Issues and pull requests are welcome.
Contains modified Copernicus Sentinel data. EFFIS © European Union. Fire detections courtesy of NASA FIRMS. Map data © OpenStreetMap contributors. Satellite basemap © Esri, Maxar, Earthstar Geographics.