Satellites recorded 16,070 farm-fire incidents across northern India this kharif harvesting season. Experts said mapping burnt areas instead of just active fires could measure stubble burning more accurately. The Centre got reports of monitoring gaps and asked authorities to use ground patrols after 3 pm to track fires better.
Experts suggest that mapping burnt areas, rather than active fires alone, could help measure incidents of stubble burning more accurately. Past assessments, too, have underscored the need for both tools.
Satellites have recorded 16,070 farm-fire incidents across northern India so far this kharif (monsoon) harvesting season. The actual scale of stubble burning, however, could be higher.
At a high-level inter-ministerial review last month, the Centre acknowledged gaps in satellite-based monitoring and directed district authorities to intensify ground patrols after 3 pm, when satellite observations are mostly unavailable. Meanwhile, there is still no consensus on a standardised method to measure the total area affected by stubble burning.
We explain how farm fires are tracked, the gaps in India's satellite-based monitoring, and why mapping burnt areas is just as important to assess the scale of stubble burning.
How farm fires are tracked
During the post-monsoon period, the narrow 10-to-20-day turnaround window (generally from October to mid-November) between paddy harvesting and wheat sowing prompts farmers to burn crop residue, severely degrading air quality across the Indo-Gangetic Plain. In peak season, stubble burning can contribute as much as more than 30% of Delhi's PM2.5, data by the Indian Institute of Tropical Meteorology (IITM) shows.
PM2.5, which refers to particulate matter with a diameter of 2.5 micrometres or less, is a microscopic mix of solid particles and liquid droplets suspended in the air, including dust, soot, smoke, and chemical aerosols. Sources of PM2.5 include vehicle and industrial emissions, coal or oil burning, power plants, and natural phenomena such as wildfires and windblown dust.
To monitor farm fires, India relies on daily thermal-anomaly observations -- a measure of significant deviations in a particular area's average temperature -- captured by sensors aboard polar-orbiting satellites such as Aqua, Suomi-NPP, and Terra, operated by US agencies such as NASA and the National Oceanic and Atmospheric Administration.
These satellites detect fires active only during fixed satellite overpass windows primarily between 10.30 am and 1.30 pm. So, fires lit in the late afternoon or evening hours risk going undetected.
For instance, the Suomi-NPP satellite, equipped with the Visible Infrared Imaging Radiometer Suite (VIIRS) sensor, passes over the region at around 1.30 pm. Its sensor detects fires at a spatial resolution of 375 metres -- meaning the satellite takes a series of pictures of the Earth, wherein each picture covers a 375-metre by 375-metre patch of land.
Whereas the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor, aboard the Aqua satellite, has a coarser resolution of about 1 km (as in, the VIIRS sensor produces sharper images). Changes in Aqua's orbit since 2023 have shifted its overpass from 1.30 pm to approximately 4 pm now, allowing it to capture some later fires.
Importantly, the resultant data of fire counts is fed into estimates of emissions from crop-residue burning, which the IITM uses in air-quality assessments and forecasts for Delhi-NCR. Hence, their accuracy has implications not only for estimating the pollution burden but also over policy decisions -- this includes the imposition of a Graded Response Action Plan, a set of progressively stricter emergency measures taken to control air pollution when Air Quality Index levels worsen.
What satellites miss
A November 2025 study by ISRO's Space Applications Centre, Ahmedabad, documented a shift in peak farm-burning activity towards late afternoon and evening hours in Punjab and Haryana.
Hiren Jethva, senior research scientist at Morgan State University and the NASA Goddard Space Flight Center, told The Indian Express that this shift meant that data from conventional polar-orbiting satellites was inadequate to map actual burning.
"Since most agencies rely on Suomi-NPP/VIIRS active fire detection products, the estimates of fire statistics derived from VIIRS is inadequate in capturing the near-totality of fire incidences in Punjab," Jethva said.
Active fire detection products -- essentially datasets processed from the infrared imagery captured by sensors such as VIIRS and MODIS -- detect thermal anomalies and incidents of stubble burning.
A 2019 study by researchers mainly affiliated with Columbia University found that the MODIS active-fire product missed more than 90% of fires in northwest India. The undercount, it said, was partly owing to the small size and short duration of most agricultural fires, lasting no more than 30 minutes which could occur entirely between satellite overpasses. Partial burning, common in parts of Haryana, posed another challenge as it produces weaker thermal signals that satellites may not reliably detect.
Using South Korea's GEO-KOMPSAT-2A satellite, Jethva in a study published earlier this year in the journal Remote Sensing found a sharp spike in farm fires around and after 4 pm - a trend gradually building since 2022 and and turning into a new fire regime since the last two years. The study found that aerosol pollution over the Indo-Gangetic Plain rose between 2022 and 2025, even as official satellite fire counts fell.
Another challenge remains cloud cover. In its daily bulletins, the New Delhi-based Indian Agricultural Research Institute has routinely flagged that clouds over northern states can obstruct satellite observations and result in fewer fires being detected.
Fewer farm fires, more burnt area
At an October 13 inter-ministerial review co-chaired by Union Agriculture Minister Shivraj Singh Chouhan and Union Environment Minister Bhupender Yadav, the latter called for ground surveillance to supplement satellite observations.
The official statement asked District Collectors to undertake ground assessments and strengthen enforcement, particularly after 3 pm, when no satellite pass is available. Past assessments have also underscored the need for both tools, rather than fire counts alone.
An assessment by the non-profit International Forum for Environment, Sustainability and Technology (iFOREST) found that in Haryana, recorded farm-fire incidents fell from 1,406 in 2024 to 662 in 2025, while its satellite-derived estimate of burnt area (BA) increased from 7,117 sq km to 8,812 sq km.
In Punjab, recorded incidents declined from 10,909 to 5,114 over the same period, but the estimated BA fell by a considerably smaller proportion, from 23,262 sq km to 19,757 sq km.
What burnt-area mapping measures
While active-fire monitoring identifies thermal anomalies at the time of a satellite's observation, BA mapping detects changes to the land surface left behind after a fire.
Stubble burning alters the way the ground reflects sunlight, particularly in the near-infrared and shortwave-infrared wavelengths. These changes, known as burn signatures, can remain detectable even after smoke plumes have subsided.
Two commonly used indices are the Normalised Difference Vegetation Index, which measures vegetation greenness, and the Normalised Burn Ratio (NBR), which compares near-infrared and shortwave-infrared reflectance (simply, light reflected by a surface). A fall in NBR can indicate damage to vegetation and a burnt or exposed surface.
Researchers compare satellite images taken before and after a suspected fire to identify changes in these indices and estimate the affected area.
Professor Emilio Chuvieco, a specialist in satellite remote sensing and wildfire monitoring at the University of Alcalá in Spain, said BA mapping provides a more accurate measure of the total land affected by fires than active-fire counts.
"Active fires are easier to identify than burned areas... but active fires are only detected at the satellite overpass and therefore small and/or lower intensity fires may be missed. Burnt area signal is weaker but lasts longer, and therefore BA products are much more accurate than AF to estimate total area affected by fires," he said.
The European Space Agency's Sentinel-2 satellites, currently being used in BA mapping exercises for internal assessments in India, provide imagery at resolutions of up to 10 metres in some spectral bands and revisit an area every five days.
Yet identifying burnt agricultural land presents its own technical challenges. Chuvieco said crop-residue fires are generally smaller and less intense than wildfires because they involve less biomass. "Therefore, medium resolution sensors, such as MODIS or Sentinel-3 are much less sensitive to agricultural fires than wildfires," he said.
Also, a change in vegetation or reflectance does not necessarily mean a field has burnt. Harvesting itself alters the spectral signature (how a material or surface reflects, absorbs, or emits light) of a field as standing crops are removed, exposing soil and crop residue. These changes can resemble burn signatures, making it difficult to distinguish burnt patches from harvested fields.
The short interval between harvesting and the next sowing season adds to the difficulty. Burn scars can change as fields are cleared, ploughed or prepared for the next crop, while cloud cover and haze can reduce the availability of usable satellite images. Chuvieco said satellite-derived BA maps are regularly compared with higher-resolution imagery to assess their accuracy.
