Delhi’s official air quality index dropped from 366 to 309 between November 2 and 3, but a detailed analysis of monitoring data raises questions about the reliability of pollution readings during one of the city’s most toxic weeks of the year.
Missing data, suspicious measurement patterns, and algorithmic loopholes in how the city’s average AQI is calculated appear to have combined to produce readings that may not accurately reflect ground conditions, an HT analysis of Central Pollution Control Board data shows.
To understand the issue, it is essential to unpack the mechanism to calculate AQI, which contains built-in flexibility designed to ensure readings are available even when some stations malfunction. But this flexibility can be exploited to generate more favourable pollution assessments, particularly when data goes missing during the most polluted hours of the day or when the dominant pollutant at a station switches from one that generates a higher index to one that shows lower readings.
The analysis of 168 hours of data from October 28 to November 4 found that missing station data was not random. With more data missing during polluted hours than clean ones, the net effect would be to make Delhi’s air quality appear better than actual conditions.
What are the loopholes in AQI calculation?
The AQI calculation works through a multi-step process. The 24-hour average concentration of six pollutants and 8-hour average for carbon monoxide and ozone is converted into a sub-index at each of Delhi’s 39 air quality monitoring stations. The highest sub-index among the eight pollutants is declared the AQI for that station. The average AQI across the 39 stations becomes Delhi’s official reading.
However, the system allows three significant relaxations:
One: An AQI need not be calculated across all 39 stations for the city average.
Two: Only 16 of 24 hours of data is sufficient for calculating the sub-index for a pollutant.
Three: A 24-hour sub-index is not needed for all eight pollutants. If the sub-index is available for three pollutants, and one of them is either PM2.5 or PM10, the station’s AQI can be calculated.
Have the loopholes been in play in AQI calculation in the past week?
Yes. All 39 stations were used only on November 1 in the CPCB bulletin for the week ending November 3, which gives the 24-hour average up to 4pm. On other days, only 37 or 38 stations were included.
But, more crucially, the prominent pollutant — the one whose sub-index determines a station’s AQI — also changed during the week. PM2.5, the expected prominent pollutant at this time of year when pollution peaks, was prominent at only 32-36 stations. PM10 was the prominent pollutant at all other stations, except on November 3, when nitrogen dioxide was the most prominent pollutant at the Lodhi Road station run by the Indian Institute of Tropical Meteorology.
Chart 1
The prominent pollutant changing at a station would not necessarily indicate a problem if data was collected for all 24 hours for at least PM2.5 and PM10, both of which typically generate higher sub-indices than other pollutants around this time. However, even among stations where PM2.5 data was available for the required 16 hours, between 8% (October 28 and 29) and 19% (November 3) had less than 24 hours of data.
Chart 2
Does use of the loopholes affect AQI data?
At a theoretical level, missing data can either increase or decrease the 24-hour average depending on whether the missing hours are relatively clean or polluted. An examination of what happened in reality for PM2.5 data over the 168 hours shows the hours when the fewest stations were recording data were the four hours from noon to 3 p.m. — typically the cleanest hours of the day. This pattern would tend to push the PM2.5 average up.
The next set of hours with the most missing stations were the five hours from 7 a.m. to 11 a.m. and the hour starting at 2 am — relatively polluted periods. This would push the average down.
Whether this represents a long-term trend or a recent pattern is difficult to verify. The CPCB dashboard that provides sub-indices for all pollutants takes hours to extract even a week’s worth of data, making historical analysis impractical.
Chart 3
The trends between November 2 and November 3 show multiple anomalies
The city’s average AQI dropped from 366 to 309 between November 2 and 3. While meteorological factors contributed to this decline — the AQI decreased at all but three stations that generated readings on both days — data anomalies also appear to have played a role.
Station-level AQI improved most dramatically at three locations: the Lodhi Road station run by IITM, the Shri Aurobindo Marg station run by the Delhi Pollution Control Committee, and the ITO station run by CPCB. The average AQI at these stations nearly halved (from 319 to 164, 294 to 157, and 280 to 155 respectively) even as the city’s average decreased by only 16%.
The prominent pollutant changed at all three stations: from PM2.5 to PM10 at ITO and Aurobindo Marg, and from PM2.5 to nitrogen dioxide at Lodhi Road-IITM.
While this shift appears plausible at Aurobindo Marg, where data for both PM2.5 and PM10 was missing during the same two hours, the patterns at the other two stations raise concerns. At Lodhi Road-IITM, the PM2.5 index trended above the PM10 index for most of the day but was much lower in three hours. Moreover, at Lodhi Road, the IITM station showed significantly lower PM2.5 readings than the India Meteorological Department station at the same location, suggesting measurement discrepancies between monitoring systems.
At ITO, both PM2.5 and PM10 sub-indices were below 50 between 4-5 a.m. when data transmission stopped. When the station resumed transmitting at noon, both indices had jumped above 350 — a trajectory that appears difficult to explain through normal atmospheric conditions.
Chart 4
What does this mean?
With one to two stations missing PM2.5 data — the most significant pollutant during this season — at any given hour, Delhi’s AQI data may not fully capture pollution levels across the city. When this occurs at heavily polluted locations like ITO, it can lower the city average without reflecting actual conditions on the ground.
Stations also appear to generate sudden spikes and drops in readings, which could result from calibration issues or external factors affecting measurements. Environmental groups have reported observing water being sprinkled near some monitoring stations in recent days, though officials have not confirmed whether this occurred or how it might affect readings.
These patterns suggest that day-to-day trends in the city’s average AQI — which is meant to communicate health risk to residents — may not provide a reliable picture of actual pollution levels during critical periods.