Swechha SearchAct
A hillside scarred by fire in Uttarakhand

Fire

What do India's fire numbers count?

Learn

The widely quoted Indian fire figures are counts of satellite detections: pixels that registered hot enough to flag. One fire can produce several detections on one satellite pass and none on the next, and a detection can be a crop field, a brick kiln or a rubbish heap rather than a forest. It is a count of observations, not a count of fires.

What a detection is

A polar-orbiting instrument passes overhead a few times a day and records where the surface was radiating heat.

NASA's FIRMS service publishes active-fire detections from the VIIRS and MODIS instruments. VIIRS resolves at 375 metres. A detection means that pixel was hot at the moment of overpass — nothing more, and nothing about what was burning.

Detections are therefore not fires. A single large fire spans several pixels and produces several detections; a small fire that starts and ends between passes produces none; cloud cover hides fires entirely; and a fire burning at night is seen by a different pass than one burning at noon.

Nor are they forest fires. The instrument cannot tell forest from farmland. Across northern India in late autumn a large share of detections are crop-residue fires, and industrial heat sources register too. Calling the total "forest fires" attributes a land cover the data does not carry.

Detection
One hot pixel on one satellite pass. The unit of every published fire count.
FRP
Fire radiative power, in megawatts. A measure of how much energy the detection was radiating — closer to intensity than a count is.
Overpass
One satellite transit. A fire is only counted if it is burning, uncovered and hot enough when a pass happens.

There is no threshold to break

Every other reading on this site is judged against a published limit. This one cannot be, and says so rather than inventing a benchmark.

Legal limit on fire detections

No legal threshold.

No statute publishes one

Sensor resolution

375 m

VIIRS S-NPP, science quality

Comparable window sampled each year

21-30 March

This site's own sampling method

Highest year in the sampled window

2,021 — 1,78,058 detections

NASA FIRMS, VIIRS S-NPP

Lowest year in the sampled window

2,024 — 65,208 detections

NASA FIRMS, VIIRS S-NPP

The ratio between the highest and lowest years in that fixed window is about 2.7. A window chosen to chase the peak would produce a different and less comparable series, which is why this one is fixed and stated.

Fire detections

How to read a fire count

The comparability of the series matters more than the size of any one year.

Which sensor? MODIS at 1 km and VIIRS at 375 m detect different numbers of fires from the same landscape. A series that switches sensors mid-way has a step in it that is instrumental, not real.

Which window? This site samples a fixed ten-day window in late March every year. FIRMS caps requests at five days, and a full season across fourteen years is several hundred requests. A fixed window is a sample rather than a season total — and it is comparable across years precisely because it does not chase the peak.

Detections or area? Burnt-area products measure how much land burned. Detection counts measure how many hot pixels were seen. They answer different questions and only one of them scales with the size of the fire.

What this number cannot tell you

  • It cannot tell you how much land burned.
  • It cannot tell you what was burning — forest, stubble, waste or industry.
  • It cannot see fires under cloud, small fires between passes, or fires that never got hot enough for the threshold.
  • It cannot be compared with a count from a different sensor, a different confidence filter or a different time window.

Where this comes from

Every figure above is read out of one of these, at the address printed beside it.

Reuse freely — CC BY 4.0. The grant covers this page. Each source above keeps its own terms, which is why every one of them is named.

Next

The reading this explains, the explanations next to it, and the one place to stand in it.

Live reading Fire detections One fixed March window, one sensor, fourteen years — sampled the same way each time. Reading environmental data Measured or modelled? The first question to ask of any environmental number, and the one most often skipped. Forests Is India losing forest? India's own report says the forest grew. The satellites say it went. Both are right. Reading environmental data How to read environmental data Seven questions that catch most of the ways a true number becomes a false claim. Fire What does stubble burning actually add? The seasonal fires north-west of Delhi, counted honestly — and the sensor gap that decides the number. Stand in it NatureScapes Fire scars are legible on the ground years afterwards, which is the check a pixel count cannot give you.

Work with us