Once a fire ignites, the decisive question is not where it is but where it will be. Ground-based forecasters rely on weather-station networks that are sparse in remote terrain and fuel maps that may be years out of date. Without continuous satellite-derived inputs—live fire perimeters, canopy moisture, surface wind divergence—spread models run on stale assumptions and produce evacuation windows that are too narrow, too late, or simply wrong.
A sovereign constellation combines mid-wave infrared (MWIR) thermal sensors for sub-hourly perimeter updates with hyperspectral passes for fuel moisture estimation, feeding a numerical fire-spread engine such as FARSITE or Phoenix RapidFire. Atmospheric wind fields pulled from the same satellite network close the loop between the fire model and the local mesoscale boundary layer that drives spotting and flanking behaviour. The architecture can ingest commercial weather-satellite wind products as supplementary streams without depending on them as the primary source.
The operational outcome is a probabilistic spread cone, refreshed every 30–60 minutes, delivered directly to incident commanders and civil protection authorities. Evacuation orders shift from reactive to anticipatory: communities are moved before the fire arrives rather than after it crests a ridge. Nations that have suffered catastrophic fire seasons—Australia in 2019–20, Greece in 2021, Canada in 2023—all discovered that the data pipeline, not the firefighting resources, was the binding constraint. Sovereign control over that pipeline is the fix.
Frequently asked
Why can't we just use free NASA FIRMS or Copernicus EMS data?
NASA FIRMS and Copernicus EMS are invaluable global baselines, but both are governed by third-party tasking priorities and data-access policies that a sovereign nation cannot control during an emergency. A nation running its own constellation sets its own revisit schedule, prioritises its own territory, and can downlink directly to its own emergency management centres without queuing behind other users or depending on a foreign government's uptime guarantee.
What orbit is best for fire spread forecasting?
LEO (450–550 km SSO) gives the spatial resolution and thermal sensitivity needed for perimeter mapping; a constellation of 6–12 microsatellites achieves sub-30-minute revisit over a target region. Geostationary thermal IR (e.g. EUMETSAT Meteosat SEVIRI at 3 km) complements with high temporal cadence but cannot resolve individual fire fronts below roughly 1 km width. The sovereign default is a national LEO constellation with geostationary data as a backstop, not the reverse.
How does fire spread forecasting differ from active fire detection?
Active fire detection (§6.2.1) identifies where fire is burning right now from satellite thermal anomalies. Fire spread forecasting takes that detected perimeter as an initial condition and runs a physical or machine-learning spread model — incorporating wind fields, fuel type, slope, and atmospheric moisture — to project where the fire will be in 1, 6, 12, and 72 hours. The forecast product is what drives evacuation orders and resource pre-positioning.
What sensors do sovereign fire-spread satellites typically carry?
The minimum useful payload is a mid-wave infrared (MWIR, ~3.5–4.0 µm) and long-wave infrared (LWIR, ~10–12 µm) dual-band imager capable of detecting fire radiative power above roughly 5 MW at nadir. Microsatellite platforms such as those demonstrated by ESA's Φ-sat-2 and ICEYE's thermal IR pathfinders show this is achievable below 100 kg. Optional additions include shortwave IR (~2.2 µm) for smouldering detection and a visible/NIR channel for post-front burned area delineation.
Can artificial intelligence replace physics-based spread models?
Not yet as a sole approach. ML models trained on historical fire perimeters (e.g. using USGS Landsat and MODIS burn records) can outperform physics models under 'normal' conditions, but they extrapolate poorly to extreme fire behaviour — precisely the cases that matter most operationally. The current best practice, used by the US National Interagency Fire Center and Canada's NRCan, couples a physics-based core (Rothermel, FlamMap, Phoenix RapidFire) with ML bias correction. A sovereign system should plan for the same hybrid architecture.
How many satellites does a national fire spread forecasting constellation actually need?
For a mid-sized country (roughly Australia to Spain in area), modelling shows 6 microsatellites in SSO give median revisit of 25–35 minutes over any point; 12 satellites reduces this below 15 minutes. Smaller island or city-state nations may achieve adequate coverage with 3 satellites plus data-sharing agreements. The ITU filing and frequency coordination required for each spacecraft slot means the constellation should be planned and registered at once even if launched in tranches.
What happens to fire spread forecasting capability if a commercial data provider cuts access?
A government that has not built sovereign capacity must either accept data blackout — potentially for the most politically sensitive fire events — or pay emergency spot-market prices. During Australia's 2019–2020 'Black Summer', Planet and Maxar tasked assets extensively on commercial terms; countries without their own observation rights experienced multi-hour data gaps over active fronts. Sovereignty of the sensor layer is the only structural fix.
How are fire spread forecasts disseminated to ground crews and the public?
Operational products are typically served as OGC WMS/WFS layers into incident management platforms (e.g. ESRI ArcGIS Emergency Management, Palantir Gotham), as GeoTIFF push-files to state emergency services, and as simplified risk polygons to public alert systems such as Australia's Emergency Alert or Canada's National Alert Aggregation and Dissemination System. A sovereign space programme should own the full chain from sensor downlink through model output to dissemination API, so no single third-party link can break it.