Punjab’s 90% Drop in Farm Fires May Be a Trick of the Satellite’s Clock

Every winter for the past several years, Punjab’s government has pointed to one number: farm fires are down, sharply. Officials cite a decline of roughly 90 percent in stubble-burning incidents since 2021, evidence, they say, of years of awareness campaigns, fines and crop-residue-management machinery finally working. This year’s satellite counts, now being logged daily by the Punjab Remote Sensing Centre as the kharif harvest peaks, will likely extend that trend on paper.

An ISRO Space Applications Centre study, published in Current Science, complicates that story considerably. Comparing geostationary satellite observations, which scan the same region every 15 to 30 minutes, against the polar-orbiting satellites that generate the official fire count, researchers found that peak burning activity in Punjab and Haryana has shifted from roughly 1:30pm in 2020 to around 5pm by 2024. The polar-orbiting satellites that produce India’s official farm-fire statistics cross northwest India only once around midday, meaning a fire lit in the late afternoon or evening may never register in the number the government reports.

A falling fire count can mean fewer fires, or it can mean the same fires, better timed to dodge the count.

The Commission for Air Quality Management has responded by asking ISRO for an improved protocol this season that pairs the existing satellite-based count with geostationary imagery and, where possible, night-time detection, and by pushing Punjab and Haryana to ground-truth the satellite record against actual field inspections. A parallel test is already running on the smaller wheat-stubble season: 10,207 fire events were recorded in Punjab and 1,832 in Haryana during April-May 2025 under the existing protocol, numbers CAQM itself now flags as an uncertain floor rather than a precise count.

None of this proves farm fires haven’t declined. Crop-residue-management machinery deployment has genuinely scaled up, and procurement data shows farmers adopting some of the alternatives the government has subsidised. What the ISRO findings do complicate is the specific claim of a 90 percent decline, since a systematic shift in burning time toward the satellite’s blind spot would mechanically deflate the count regardless of what farmers are actually doing in their fields. Given that stubble burning’s contribution to Delhi-NCR’s wintertime particulate load can spike from around 15 percent of PM2.5 to nearly 44 percent in the worst weeks, according to CAQM’s own attribution work, the gap between the reported count and the true burn rate isn’t just an academic quibble. It feeds directly into how aggressively emergency pollution measures get triggered, and how much credit gets claimed for a problem that may not have shrunk as much as the headline number suggests.

The stakes go beyond bragging rights over a single statistic. Delhi’s new Winter Pollution Action Plan, and the Graded Response Action Plan it runs alongside, both calibrate their emergency measures partly against how much of the capital’s particulate load is attributed to upwind stubble burning versus local sources like vehicles and construction dust. If the true burn rate is meaningfully higher than the official count suggests, that attribution skews toward under-blaming agriculture and over-relying on Delhi-side curbs to fix a problem partly seeded two states away, a mismatch that no amount of construction bans or staggered office timings inside the capital can fully correct.

This is a live test right now, not a settled finding. The kharif stubble-burning season runs from mid-September to the end of November, meaning the improved protocol CAQM has asked for is being built and applied in real time even as this year’s fires are being counted, both ways, under the old method and whatever ground-truthing the states can manage alongside it.

For a newsroom’s purposes, the story is a reminder that satellite-derived environmental statistics, now central to enforcement and policy credit-claiming across many countries, are only as good as the assumptions built into how and when they observe. Any metric that the people being measured have an incentive to game will eventually get gamed around, and the fix usually arrives only after the headline decline has already been celebrated for several seasons.