VIIRS: active fire at 375 metres

VIIRS measures brightness temperature in kelvin in two infrared channels: a short-wave one that responds to the hot flame, and a long-wave one for the surroundings. The detection arises from the contrast between the two. To this sensor a fire is not a visual pattern but a thermal outlier, which is why detection works at night, and why it fails under cloud.

For every detection FIRMS supplies the fire radiative power in megawatts. It is the best available measure of a fire's intensity and the right quantity for prioritising reports: it distinguishes a burning hillside from a burning garden pile, which a plain count of points does not.

375 metres is the value at nadir, that is directly beneath the satellite. Towards the edge of the scanned strip the pixels grow larger; FIRMS therefore carries the actual ground footprint with every detection, measured at 0.38 to 0.73 kilometres.

Unlike MODIS, the VIIRS aggregation scheme keeps the pixel size between 375 and 750 metres out to roughly 1,200 kilometres either side of the ground track; only beyond that does it become markedly coarser. An overpass that merely clips the working area therefore still delivers usable but coarser geometry, the samoEKO overpass forecast marks it as an edge pass rather than concealing it.

Three satellites carry VIIRS: Suomi-NPP, NOAA-20 and NOAA-21. Over South-East Europe their overpasses fall between 23:00 and 02:00 and between 10:00 and 13:00 UTC, six observation windows a day. In between, roughly from 15:00 to 01:00 local time, there is virtually no VIIRS observation. That is precisely the Mediterranean afternoon peak of fire ignition.

Detections become available about three hours after the overpass. The sub-60-second ultra real-time detection NASA offers for some regions exists only for the United States and Canada via four direct readout stations; it is not available for South-East Europe.

Schroeder and colleagues quantified the detection threshold in 2014: a flaming fire of roughly 2 square metres at night, or roughly 40 square metres by day, is detected with 50 per cent probability. For fires below 50 hectares the detection rate lies between 15 and 70 per cent.

Cloud cover blocks thermal detection completely a null result is therefore never an all-clear; it may simply mean cloud. Day and night are also not directly comparable in sensitivity: a 2025 study suggests that VIIRS systematically withholds low-confidence detections at night.

And a single fire produces many pixels, which three satellites with overlapping strips see repeatedly. Raw detection counts are not fire counts; every count must be preceded by spatio-temporal clustering on the order of a 750-metre radius and 12 to 24 hours.

samoEKO retrieves the detections of all three satellites through the FIRMS area interface as soon as an access key is configured, and keeps a keyless second path through the static Europe files a blocked access key must not blind the situation picture. Without a key the second path carries the load alone; it lags the retrieval by minutes.

In one week of the high season the working area yielded 869 detections from NOAA-20, 876 from Suomi-NPP and 850 from NOAA-21, against 176 from MODIS. VIIRS therefore delivers roughly five times as much; for the small karst and maquis fires of the Adriatic coast there is no alternative.

For permanent industrial heat sources, the Zenica steelworks, the Tuzla and Kakanj power stations, the Lukavac cement works a separate FIRMS query exists that marks such points. It requires the same access key as the area interface. If it does not run, these plants appear as ordinary detections; the current state of the mask is stated on the Methodology page and in the retrieval log of the operations centre.