What we don't know
— and intend to find out.

Yes — within limits. The first large field deployment of continuous ground smoke sensors (20 prescribed fires and wildfires, Canada 2023–2024) detected every fire over 100 hectares at median distances of ~1.5 km. Detection of fires ≤1 ha dropped to 30%.

Source: Fire (MDPI) 9(4):141, Jan 2026

DHS's own Inspector General found its contracted sensors (N5, 2020–2024) "did not consistently detect fires or provide early warning" — one sensor <100 ft from a fire alerted 30 minutes after dispatch, and wind direction caused misses. DHS ended the program. The same sensor class, used well in the Canadian study above, still delivered unique detections. The honest read: capability exists, but placement, wind, and verification are unsolved.

Source: DHS OIG-26-05, Apr 2026

No comparable open-hardware result exists yet. Controlled-burn demos with a BME688 + particulate sensor (BioBot project) detected smoke through obstruction at ~7 m; the DHS/NRCan networks detected large fires at 0.5–4.8 km. Where cheap open nodes fall in that range is exactly what we intend to measure.

Source: BioBot Project

Promising but undemonstrated. The biggest identified false-positive source is far-field regional smoke drifting hundreds of kilometers — vendor event algorithms currently fail during regional smoke episodes. Multi-node corroboration plus wind data is our planned approach; no published number exists yet.

Source: Fire (MDPI) 9(4):141, Jan 2026

Unknown, and we treat it as an experiment rather than a guarantee. Meshtastic excels at off-grid community messaging, but public-safety-grade delivery guarantees through canyons and forest canopy have not been demonstrated. We'll publish packet-delivery data from every field test.

Source: Meshtastic

Untested by us so far. Battery autonomy targets 2–3 days without sun; snow load, sustained inversion fog, and sensor aging are the known risks. Winter-survival trials are planned.

We don't claim it today. Satellites now deliver Ultra Real-Time fire detections in under 60 seconds for much of the US and Canada; camera networks like Pano AI already detect fires in Utah. If our instruments can't demonstrate earlier, unique, actionable detections in at least some repeatable circumstances, early detection stays a research track — never a promise.

Source: NASA FIRMS