Every year Diwali night does something no report or press release can. It makes Delhi’s air pollution visible, measurable and impossible to ignore, all in a few hours.This year was no different. On the night of November 8–9, PM2.5 in Delhi climbed from about 160 micrograms per cubic metre (µg/m³) at 8 pm to an official city average of 592 µg/m³ in the hour after midnight. The real peak was higher. As levels rose, dozens of stations dropped out of the Central Pollution Control Board (CPCB)’s public feed. Delhi Pollution Control Committee (DPCC) data for the same monitors show that they had not failed. They were still recording, and every missing hour read above 1,000 µg/m³. Readings that high appear to be thrown out of the national feed rather than published. Twenty-one stations recorded at least 900 µg/m³ before they went quiet.Sulphur dioxide nearly quadrupled, fine particles came to make up about 90% of PM10, and about three-quarters of the extra pollution arrived after the 10 pm deadline set by the Supreme Court.So Diwali is a pollution problem, and it deserves a firm response. But stopping the story there misses the most important number of the night.The number before the first crackerAt 4 pm on Diwali day, before a single cracker went off, Delhi’s PM2.5 was already about 136 µg/m³. That is more than twice the national 24-hour standard of 60 µg/m³ and about nine times the World Health Organisation’s 24-hour guideline.That afternoon figure is the city’s baseload. It comes from vehicles and road dust, power plants and factories across the region, construction sites, burning garbage and landfills, wood and dung burnt for cooking and heating, and the secondary particles that form from all of these in the air. On 7 November, the real-time source apportionment run by the Indian Institute of Tropical Meteorology put secondary aerosols at about a third of Delhi’s PM2.5, biomass burning at about a quarter, waste burning at 13% and vehicles at 10%. None of these sources stop after Diwali. None of them started on Diwali.Firecrackers and stubble fires are episodes. The baseload runs 24 hours a day, 365 days a year. The episodes highlight the problem; the baseload is the problem.Same emissions, different weatherSo why is the air clean enough to be talked about in August and lethal in November? Mostly the weather.From 2018 to 2025, Delhi’s PM2.5 averaged 34–38 µg/m³ in July and August. In November it averaged 219 µg/m³, in December 199 and in January 180. Delhi’s traffic, industry and power demand do not change six-fold between August and November. What changes is the atmosphere’s capacity to carry pollution away.During the monsoon, rain washes particles out of the air, winds are steady and the mixing layer (the depth of air into which pollution can spread) is high. In winter the opposite happens. Cold, calm nights produce temperature inversions that put a lid on the city. On Diwali night this year, the mixing layer was only 55–80 metres deep from 11 pm to 7 am, winds at street level were near calm and humidity rose above 80%. Our model suggests the weather on its own would have added only about 50–60 µg/m³ to an ordinary night. Its real effect was to trap whatever was emitted close to the ground.Also read: Northern India’s Air Crisis Is a Class and Public Health EmergencyThe lesson goes both ways. The monsoon is not proof that Delhi’s air is under control: even in August the average is close to the national annual standard and several times the WHO guideline. Winter is not proof that something unusual is being emitted. Winter shows what Delhi emits every day once the atmosphere stops diluting it. Between 2018 and 2025, Delhi met even its own lenient daily standard on only 100–176 days a year, and the annual average stayed between 94 and 115 µg/m³.Stubble: Real, but not the whole storyStubble burning follows the same pattern. It matters, and on certain days in late October and early November, when north-westerly winds carry smoke from Punjab and Haryana, it can add tens of micrograms to Delhi’s air. Our trajectory-based analysis puts its contribution at roughly 5–20% of Delhi’s October–November PM2.5 over 2018–2025, and only about 1.5–5% of the annual average.The telling result is what happened when fires fell. Fire exposure along Delhi’s air paths dropped by about 90% between 2019–21 and 2024–25, partly because burning appears to have moved to later in the day, when satellites pass overhead less often. Yet Delhi’s October–November PM2.5 stayed at 165–170 µg/m³. If stubble were the main cause, the season would have cleaned up. It did not.Use the momentum, at the sourceNone of this argues for going easy on crackers or crop fires. Each winter, these episodes are the only time air pollution leads the national news, hearings and political speeches. That attention should be turned into lasting action at the source, not more emergency measures that end with the season.For firecrackers, that means rules that can be enforced and checked: licensed manufacturing whose chemistry is independently tested, traceable sales, real penalties for bursting outside permitted hours, and a steady move towards community displays. It also means honest data. A monitoring system that drops its highest readings during the worst night of the year understates what people breathed.For stubble, the fixes are well known but still underfunded. Farmers need assured markets for straw from biomass power, cement kilns, compressed biogas and pellets. They need machinery that is available on time, and procurement and price support that make it worthwhile to diversify away from paddy in the groundwater-stressed north-west. Monitoring must account for burning that happens after satellite overpasses, or we will keep congratulating ourselves on fire counts that measure observation times rather than fires.You cannot fix what you cannot measureDiwali also showed how fragile our evidence is. To check whether the Supreme Court’s green-cracker order made any difference, we needed hour-by-hour readings from every station for the whole night. That is exactly when the public record broke down. As concentrations passed 1,000 µg/m³, readings vanished from the CPCB feed. By 1 am only 23 of 59 Delhi stations were still reporting there, and the gaps were never filled in afterwards (at least till the dat awas checked on 9th November 2026, 4 PM). On the national data alone, the city’s peak looks like 592 µg/m³. Once the readings DPCC kept for the same monitors are added back, even when capped at 1,000, the peak is about 730 µg/m³. The difference is large enough to change the answer to the question the court asked.Also read: Firecrackers Contributed More Than 70% of Major Air Pollutant PM2.5 On Diwali in DelhiThere were other problems. Several analysers stopped climbing at a fixed ceiling, so their true peaks are unknown. One station came back reading 0.1 µg/m³ in the middle of the night. The national station list carries 79 names for 60 actual sites in Delhi, because stations get renamed. And 19 new stations, added in less-polluted areas this year, pull the city average down, so a simple comparison with last year’s average shows an improvement that is partly an artefact of where the monitors are. Without fixing these issues site by site, any claim that a policy worked, or failed, is a guess.This matters far beyond one night. Every intervention, whether a cracker ban, a stubble scheme, a power plant standard or an electric bus fleet, has to be judged on whether it reduced pollution once the weather is accounted for. That needs:complete, unedited data, including readings above 1,000 µg/m³, flagged where necessary but never silently dropped;stable, documented station networks, with a public record of every new site, renaming, relocation and calibration;validated data published quickly, not months later, together with the raw readings so that independent analysts can check the corrections;open weather and emissions data, so that the effects of meteorology and of emissions can be separated, as this piece has tried to do;monitoring beyond the big cities, in the towns and rural areas of the wider airshed, where much of the pollution starts and where there are almost no monitors.Good data does more than grade past policies. It shows where the next ones should go. Source apportionment, emission inventories and satellite records only steer policy as well as the ground measurements they are checked against. If the measurements are weakest during the worst episodes, we lose both the evidence of what is not working and the signal of what needs to be done next.The other 363 daysThe bigger task is the baseload, and it cannot be done within Delhi’s borders or only in winter.Power: Coal-fired plants across the NCR and the wider Indo-Gangetic Plain must meet the 2015 sulphur dioxide and nitrogen oxide standards. Instead, flue-gas desulphurisation deadlines have been pushed back repeatedly, and in 2025 the requirement was dropped for most plants away from large cities. Sulphur dioxide from these plants turns into fine sulphate particles that travel hundreds of kilometres.Industry: Clean fuels such as piped gas and electricity across all NCR industrial clusters, continuous emission monitoring with public data, and enforcement that reaches the thousands of small, unregistered units, not just the large ones.Waste: An end to open burning and to fires at legacy landfills, with segregation and processing of waste at source, every day of the year, not just under emergency rules.Transport: Rapid electrification of buses, two- and three-wheelers and delivery fleets, cleaner freight and bypass routes for trucks, and public transport good enough to win people out of cars.Construction and dust: Dust controls enforced year-round, paved and vacuum-swept roads, and green cover on open land.Homes: Affordable LPG refills and clean heating, so that poor households and night workers do not have to burn wood, dung and waste to cook and stay warm.All of this has to be planned for the airshed, not the city. Delhi’s air is shaped by emissions from Haryana, Uttar Pradesh, Rajasthan, Punjab and beyond, and the same airshed covers the entire Indo-Gangetic Plain. The National Clean Air Programme should move from city-level PM10 targets to binding, region-wide cuts in emissions measured in tonnes, with PM2.5 as the benchmark and funding tied to actual reductions. The Graded Response Action Plan should become a last line of defence, not the main policy.Diwali night will keep coming back every year, and so will the smoke from the fields. Each time, they will show us the same thing: a city, and a country, whose everyday air is already unsafe before the festivals begin. We should be grateful for the highlighter. Then we should deal with what it is pointing at.Sunil Dahiya is the founder and lead analyst of EnviroCatalysts. Views are personal. Figures for 2026 are based on provisional CPCB and DPCC data.