Delhi's air looks -8 points vs last year on the meter, but that is the weather's work. Remove it, and the city is +22 points worse than last year. The improvement belongs to the weather, not to policy. Today's reading itself is weather-flattered by about 25 points.

Even after removing the weather's help, Delhi's fine-particle level (PM2.5) is 44 micrograms per cubic metre (µg/m³), which is 3.0× the WHO safe limit of 15. The worst weather-corrected hotspot is Wazirpur at 50 µg/m³, which is 3.3× the WHO limit.

After weather correction, 30 of 30 comparable stations are worse than last year; the typical station is 24 AQI points higher. This is a city-wide rise, not a few bad spots.

On days it has never seen, the weather-correction model explains about 37% of the variation, so we treat its outputs as estimates, not exact points. The weather-corrected year-on-year change for fine particles is measured to within roughly ±2 percentage points of sampling error, before model uncertainty. Gas data carries documented unit-inconsistency (peer-reviewed international research, 2025); we lead on the reliable PM2.5/PM10. Percentages compare the same stations in both years; single-day observed spikes are never used as the headline.

Weather-corrected claims allowed (model skill adequate) · PM10: weather-corrected claims allowed (model skill adequate) · NO2: direction reliable (same stations + units both years); absolute levels carry the units caveat · NOx: not part of CPCB's AQI, and the weather model is weak here, so we report observed facts only · SO2: the weather model is too weak to separate weather from policy, so we report observed facts only · CO: the weather model is too weak to separate weather from policy, so we report observed facts only · Ozone: rising ozone chemically signals falling traffic NO2 (standing rule: excluded so it can never be spun as 'transport improved')

PM2.5: 37 of 42 stations above the WHO daily limit as observed; even after weather correction, 42 of 42 remain above it. PM10: 37 of 44 stations above the WHO daily limit as observed; even after weather correction, 44 of 44 remain above it. (Weather-model matched panel.)