What a Satellite Fire Detection Is, and What It Is Not

Direct answer: a satellite fire detection is a single image pixel that measured more heat than its surroundings. It is a thermal anomaly, not a confirmed wildfire. One large fire routinely produces dozens or hundreds of detections across several satellite passes, and some detections are not wildfires at all — agricultural burning, gas flares and hot industrial sites register the same way. A detection count is therefore never a fire count.

What the satellite is actually measuring

The instruments behind most global fire data — NASA’s VIIRS and MODIS, feeding the FIRMS system — do not see flames. They measure infrared radiation and look for a pixel markedly hotter than the pixels around it.

Each detection is therefore a small square of ground with a timestamp and a coordinate. VIIRS resolves roughly 375 metres per pixel; MODIS roughly 1 kilometre. The coordinate reported is the centre of that square, not the position of a flame. A detection plotted on a map at a neighbour’s house means something hot was somewhere inside a square that contains their house, yours, and a good deal else.

Why a count of detections is not a count of fires

This is the single most misread number in wildfire data, and the mechanics make it obvious once seen.

• One fire covers many pixels. A fire front several kilometres long lights up a line of them at once.
• Each satellite pass counts again. The same fire burning for three days is detected on every overpass that sees it, and those are separate rows in the data.
• Multiple satellites see the same ground. Different instruments on different platforms each contribute detections for one fire.

So a headline figure such as 991 detections describes how much hot ground was observed across a period and an area. It could be one enormous fire complex observed repeatedly, or many small agricultural burns, and the number alone does not distinguish them. It measures thermal activity; it is not an inventory of incidents.

Not everything hot is a wildfire

The sensors are honest about what they do: they find heat. Several things that are not wildfires are reliably hot.

• Agricultural burning. Across much of South Asia, Africa and South America, seasonal crop-residue burning produces enormous and highly regular detection clusters. They are intentional, managed, and not emergencies.
• Gas flares. Oil and gas infrastructure burns off gas continuously. Flares appear night after night in exactly the same place, which is what makes them recognisable.
• Industrial heat. Steelworks, refineries and similar sites can cross the threshold.
• Volcanic activity. Lava is hot, and it is detected as such.

Products that categorise detections exist, and every detection carries a confidence value — low, nominal or high — reflecting how strongly it stood out from its background. A low-confidence detection is a weak signal that may be a small fire, a partly clouded pixel, or an artefact.

The harder problem: what the satellite cannot see

Detections are evidence of fire. The absence of detections is not evidence of no fire, and treating it as such is the dangerous version of this misunderstanding.

• Cloud blocks the view completely. Thermal sensors do not see through thick cloud, so a fire under weather is simply absent from the data.
• Dense smoke degrades detection — which is a problem exactly when a fire is at its most intense.
• Satellites revisit; they do not watch. A polar-orbiting satellite passes a given point a small number of times per day. A fire that starts, runs and is contained between two passes may never be detected at all.
• Small or cool fires fall below the threshold. A fire under canopy, or a smouldering ground fire, can be invisible from orbit while being very real on the ground.

This is a specific case of a general rule worth internalising about every monitoring system: nothing showing does not mean nothing happening.

Why this data is still valuable

None of the above makes satellite detection weak. It makes it a particular kind of strong.

It is global, consistent and fast, and it depends on nobody on the ground reporting anything. For large fires in remote or under-monitored regions it is frequently the first indication that exists anywhere, hours ahead of any official announcement. It is also the only practical way to see a continental-scale fire season as one picture.

What it is not is a confirmed incident list. Ground observation, aircraft, camera networks and agency reporting do that job, and how a fire behaves once it starts is a question satellites can only partly answer.

How to read a fire map properly

• Read a cluster as one probable event, not as a count of separate fires.
• Check the time window. Detections accumulate, so a 24-hour map and a 7-day map of the same fire look wildly different.
• Treat a lone low-confidence point with caution, particularly near known industry or during a burning season.
• Never read empty ground as safe ground, especially under cloud.
• For anywhere you actually care about, check the responsible fire agency. They hold information satellites do not.

How this site fits

Fire activity shown here comes from NASA FIRMS, and where a group of detections is presented as a single event, the detection count is stated as what it is — a count of detections. We do not confirm fires and are not an emergency service. For evacuation information, containment status or road closures, the responsible fire agency is the only authority.

Sensor resolutions and confidence categories described here reflect the VIIRS and MODIS products as documented by NASA at the time of writing. Product specifications are revised periodically.

Frequently asked questions

Does one satellite fire detection mean one fire?

No. A detection is a single image pixel that measured more heat than its surroundings. One large fire produces many detections across its front and again on every satellite pass, so counts of detections are always higher than counts of fires.

What does a number like 991 detections actually mean?

That 991 hot pixels were observed across the area and period in question. It could be one large fire complex seen repeatedly, or many small agricultural burns. The figure measures thermal activity, not the number of incidents.

Are all detections wildfires?

No. Agricultural crop-residue burning, gas flares at oil and gas facilities, hot industrial sites and volcanic activity all register as thermal anomalies. Flares are recognisable because they appear in the same place night after night.

How accurate is the location of a detection?

It is the centre of a pixel, roughly 375 metres across for VIIRS and about 1 kilometre for MODIS. A point plotted on a map means something hot was somewhere within that square, not at that exact spot.

Can a fire be missed by satellites?

Yes, routinely. Thick cloud blocks thermal sensors entirely, dense smoke degrades detection, small or smouldering fires fall below the threshold, and polar-orbiting satellites pass a given point only a few times a day. A fire that starts and is contained between passes may never appear.

What does the confidence value mean?

How strongly the pixel stood out from its background, reported as low, nominal or high. A low-confidence detection may be a small fire, a partly clouded pixel or an artefact, and is worth treating cautiously on its own.

If there are no detections near me, is there no fire?

Not necessarily. Absence of detections is not evidence of absence of fire, especially under cloud or between satellite passes. For anywhere that matters to you, the responsible fire agency holds information satellites cannot provide.

Sources

• NASA FIRMS - Fire Information for Resource Management System
• NASA Earthdata - VIIRS 375 m active fire product
• NASA - MODIS active fire and thermal anomalies product
• NASA EONET - Earth Observatory Natural Event Tracker
• Copernicus EMS - European fire monitoring and rapid mapping
• GDACS - global wildfire alerting

Found an error? See our corrections policy.