FireGuard

WildfireFireGuard v3.0 (SentinelCore v1.0 · F-GRADE 3.0 · F-CAST V2)

FireGuard is the original ForestGuard wildfire stack wrapped as a TerraSentinel engine: SentinelCore v1.0 composes F-GRADE 3.0 episode scoring, F-CAST V2 pre-ignition forecasting and the wind-forward attention cone, with evidence graph, calibration and counterfactuals intact.

Run FireGuard

Live mode uses keyless open providers; demo mode is deterministic and offline. Missing data lowers confidence — it never fakes a value.

LatitudeLongitudeData

Score formula

F-GRADE = 100 × P_true_fire × (0.30·episode_strength + 0.30·spread_potential + 0.25·exposure_impact + 0.15·context_modifier) − uncertainty_penalty (mode-specific FG-CIE formulas for forecast/candidate modes)

30%episode_strength

30%spread_potential

25%exposure_impact

15%context_modifier

Required inputs

NASA FIRMS active-fire detections (VIIRS/MODIS)
Open-Meteo fire-weather forecast (wind, RH, VPD, soil moisture)

Optional inputs

land cover (ESA WorldCover) + regional context packs
community feedback tallies

Missing optional inputs drop their formula term (weights renormalize) and reduce confidence — the engine never crashes or invents data.

Decision modes

no_fire_weather_risk

No elevated fire-weather risk

Quiet fire weather and no thermal evidence nearby.

pre_ignition_risk

Pre-ignition fire-weather risk

F-CAST fire-weather conditions are elevated before any detection.

candidate_hotspot

Candidate hotspot (unverified)

Thin satellite thermal evidence — not yet a corroborated episode.

active_fire_episode

Active fire episode

Corroborated satellite evidence of an active fire, scored by F-GRADE.

spread_attention

Spread attention (downwind)

Wind-forward attention cone indicates downwind exposure.

insufficient_data

Insufficient data

Not enough evidence or forecast data to assess.

Limitations

Satellite detections are not ground-confirmed; industrial heat and agricultural burning can false-positive.
F-CAST forecasts fire-weather risk and attention zones — it does not guarantee a fire will occur.

Model card

FireGuard is a self-built TerraSentinel engine: external providers supply live environmental variables; the score, mode, confidence, explanation and actions above are computed by this project’s own documented model. Provider risk indexes are used only as benchmark references and never as the output. API: GET /api/engines/fireguard/run?lat=&lon=

TerraSentinel is a decision-support tool and does not replace official emergency services, government warnings, or professional environmental management decisions.