Published: 22 September 2026 | Author: Team JanVayu | Reading time: 6 min
Delhi's AQI went from 88 on Friday to 169 on Monday, and the explanation offered was that the monsoon is withdrawing, the winds have dropped, and dust and exhaust are staying up longer. All of that is true. None of it is new, and the four-day number is the least informative part of the story.
Here is what our own record says about that week, and what it would take for a week like it to read differently.
Start with what the two numbers mean
India's Central Pollution Control Board calls 51 to 100 Satisfactory and 101 to 200 Moderate. So the move being reported is from Satisfactory to Moderate. It is real, and it is a long way from the categories Delhi spends its winter in.
It is worth knowing how unusual the starting point was. We hold CPCB's own daily bulletins for Delhi from 2015 to now, 4,133 days of them. Across all of those days the city was rated Good on 15 of them, which is 0.36 per cent. It was Satisfactory on 694, about one day in six. Friday's 88 was not a normal Delhi day that then deteriorated. It was one of the better days Delhi gets, and it did not last.
One caution before you compare that with a number on a phone app. CPCB and the US EPA convert the same PM2.5 concentration into different AQI numbers with different labels, and WAQI (aqicn.org), for one, publishes the US EPA scale. The AQI explainer sets the two tables side by side. When the label and the reading disagree, read the µg/m³.
The seasonal turn has happened in all 43 years we hold
We publish a reconstruction of daily PM2.5 for 783 Indian districts going back to 1980, from Wang and colleagues (2024) in Earth System Science Data. It lets us ask a question that a four-day comparison cannot: how often does this actually happen?
Take each year's monsoon months (June to September) and compare them with the post-monsoon (October and November), for every year from 1980 to 2022.
| City | Monsoon | Post-monsoon | Median rise | Years the rise did not happen |
|---|---|---|---|---|
| Delhi | 38.4 | 111.8 | +70.7 µg/m³ (2.62×) | 0 of 43 |
| Lucknow | 31.5 | 103.8 | +69.0 µg/m³ (2.77×) | 0 of 43 |
| Patna | 29.9 | 80.3 | +49.5 µg/m³ (2.60×) | 0 of 43 |
| Kolkata | 22.1 | 62.8 | +41.2 µg/m³ (2.84×) | 0 of 43 |
| Mumbai | 18.9 | 43.5 | +24.8 µg/m³ (2.31×) | 0 of 43 |
| Chennai | 20.0 | 23.8 | +4.6 µg/m³ (1.24×) | 0 of 43 |
Medians, µg/m³, LongPMInd district means. The last column counts years in which the post-monsoon figure came in at or below the monsoon figure.
Across forty-three years and six cities, the rise arrives every single time. Last Friday did not start it. It is the year turning.
Notice also how much the size of it varies. Delhi's post-monsoon runs at about two and a half times its monsoon; Chennai's rises by a quarter. The reason is likely geographic rather than a matter of governance, though these two ratios alone do not prove it. Our airshed panel shows something related: annual PM2.5 is mostly set by region, with 89 per cent of district variance sitting between states.
The peak has grown by decade, and its timing has not
The season is old. Its scale is not. Delhi's average October and November reading, by decade:
| Decade | Post-monsoon PM2.5 |
|---|---|
| 1980s | 59.8 µg/m³ |
| 1990s | 68.3 µg/m³ |
| 2000s | 138.7 µg/m³ |
| 2010s | 134.7 µg/m³ |
| 2020 to 2022 | 142.4 µg/m³ |
The reconstruction shows Delhi's post-monsoon level roughly doubling at 2000, but the dataset's authors attribute that jump partly to missing satellite inputs before 2000 (Wang et al. 2024, ESSD), so compare 2000 to 2022 with itself rather than with earlier decades. India's own annual limit is 40. A reader looking for the story in this week's numbers is looking two decades too late.
Where the wind explanation holds, and where it stops
"Weak winds are reducing dispersion" is a claim we can test rather than repeat, because we run meteorological normalisation on 44 Indian cities: a model learns how each city's daily PM2.5 responds to wind, temperature and humidity, then every day is re-read under randomly resampled weather so that no day is flattered by a lucky westerly. The method is Grange and colleagues (2018); we described how it works in Was It Policy, or Was It the Wind?.
For Delhi, over 41 stations and 91,911 station-days from 2018 to 2024, that model accounts for 89.6 per cent of the variance in daily readings knowing only the weather, the date, the day of the week and which station reported. It is told nothing about traffic, construction or stubble. Season, station and weather together explain that share, so a four-day move of this size is well within what weather and the calendar produce. That is our inference: it does not tell you whether emissions changed between Friday and Monday.
Stretch the same data to years and it stops carrying the story. Delhi's raw trend across those seven years is −1.75 µg/m³ per year. With the weather removed it is −1.78. The two agree almost exactly, which means the weather is not hiding an improvement and is not inventing one either. Delhi's normalised 2024 figure is 97.8 µg/m³. At that rate it would reach 40 around 2056, if the trend stayed linear for three more decades, which nothing guarantees.
Thirty-three of the 44 cities are falling once weather is removed. Delhi is one of them. The direction is right, and the speed is far too slow.
What would move it
We built a calculator for this, because the arithmetic is worth doing once rather than arguing about.
Delhi's annual mean is 89.0 µg/m³. Reaching 40 means removing 49.0, or 55.1 per cent of the mass. ARAI and TERI put transport at 28 per cent of Delhi's winter PM2.5 (17 per cent in summer; dispersion model, base year 2016). Applying that winter share to an annual mean, take every vehicle off the road and the city lands at 64.1, still 24.1 above the limit. Remove the four named sectors (89 per cent of the study's winter modelled concentration) and it lands at 9.8. What is left is the study's "Others" category, which is itself local (diesel generators, refuse burning and similar). The study does not isolate pollution arriving from outside Delhi as a separate slice.
That last figure is the reason a four-day AQI reading makes a poor scorecard and an airshed makes a better unit of government than a city.
The useful thing to watch
Skip this week against last week. Watch your district's annual record since 1980, which is on the site for 783 districts, and the annual mean with the weather taken out of it.
If you find an error in any of the figures above, write to us at contribute@janvayu.in. Every number from our own analysis comes from a file you can download; the four-day AQI figures are News18's and the source shares are ARAI and TERI's.
Sources: CPCB daily AQI bulletins 2015-2026 (2015-2025 parsed by UrbanEmissions.Info); LongPMInd, Wang et al. 2024, ESSD 16, 3565-3577, CC BY 4.0; India Air Quality Database, XKDR Forum, CC BY 4.0; ARAI & TERI 2018 source apportionment for Delhi NCR. The four-day AQI figures are as reported by News18.