Methodology & limitations
Model version wildfire-baseline-0.2.0. Coverage: Spain, France, Italy and Greece (mainland plus major islands).
What this predicts
This tool estimates environmental susceptibility to wildfire: how favourable current and forecast conditions are to ignition and rapid fire growth. It also shows real satellite fire detections.
What it does not predict
It does not predict exact ignition points, fire spread, or evacuation needs. It is not an emergency service. Always follow local authorities.
How the wildfire score works
The 0–100 score is a transparent weighted blend of normalised components: fire weather (35%), fuel dryness (20%), wind spread (15%) and drought (10%). Terrain and historical-fire context (10% each) are not yet ingested, so the score is renormalised over the components we can compute, and confidence is reduced accordingly. Every score exposes its top drivers with the underlying numbers.
Fire Weather Index
We compute the Canadian Forest Fire Weather Index (FWI) System from noon temperature, humidity, wind and 24-hour rainfall. The moisture codes (FFMC, DMC, DC) are spun up over the preceding ~90 days so they reflect antecedent dryness. Danger classes follow the EFFIS thresholds.
Risk categories
Low (0–19), Moderate (20–39), High (40–59), Severe (60–79), Extreme (80–100). Categories are always shown as text, never colour alone.
Data sources
- Weather forecast & recent history: Open-Meteo (ECMWF/GFS blend), ~9 km resolution.
- Active fire detections: NASA FIRMS (VIIRS 375 m).
- Place names: BigDataCloud reverse geocoding.
- Globe imagery: Esri World Imagery; terrain via Cesium (optional).
Spatial resolution & uncertainty
Weather model resolution is roughly 9 km. Scores describe conditions near a point, not a precise address. Confidence falls with forecast lead time and with any missing inputs. Stale or missing data is labelled, never hidden.
Known limitations
- No terrain (slope/aspect) or historical burn context yet.
- Temperature anomalies vs climatology are not yet included.
- Country-wide risk overlays are a planned addition.
- The model is not yet scientifically back-tested.
This system provides environmental risk intelligence from public weather and satellite data. It is not an emergency warning service.