Satellite Detection of Active Fires

Satellites detect active agricultural fires by sensing the unusual heat released while vegetation or crop residue is burning. Active-fire detection identifies thermal anomalies in satellite observations; it does not directly measure all the land burned or the amount of smoke emitted.

How a burning field becomes a satellite detection

Satellite sensors measure radiation in several wavelength bands. A fire produces a particularly strong signal in the mid-wave infrared, allowing sufficiently intense burning to be detected even when it occupies only part of a satellite pixel.

• Algorithms compare the candidate pixel's thermal signal with background temperatures in surrounding pixels, using tests designed to distinguish fires from normally warm surfaces.
• Measurements in longer-wave thermal infrared bands help distinguish the spectral signature of a fire from general surface heating.
• Cloud screening and additional spectral tests reduce false detections caused by clouds, sunlight reflections and other non-fire features.
• Pixels passing the detection tests are reported as active-fire locations, usually with observation time and confidence information.

Sensors and identification of agricultural fires

Widely used instruments include MODIS, with a nominal active-fire pixel size of about 1 kilometre, and VIIRS, whose finer-resolution active-fire product uses 375-metre observations. Pixel size describes the observation footprint, not the dimensions of the flame.

• A detection is associated with agricultural land by comparing its location with cropland maps or other land-use information.
• Cropland location and the crop calendar can support an interpretation of residue burning, but a thermal anomaly alone cannot establish the fuel type or cause.
• Finer spatial resolution generally improves the detection and separation of small fires, although detection also depends on fire intensity and observing conditions.

What detections reveal, and what they miss

Active-fire counts are observations of burning at satellite viewing times, not a complete census of individual farm fires. Their interpretation must account for both missed fires and repeated observations.

• Brief fires may start and finish between satellite overpasses; weak or very small fires may not generate a detectable signal.
• Cloud cover, thick smoke and obstructing vegetation can prevent or reduce detection.
• One fire may generate several detected pixels or be observed on successive passes, so detection counts need not equal the number of separate incidents.
• Burned area requires separate mapping of the affected surface; emissions estimates additionally require information about fuel consumed and emission factors.