smartcitiesmission.com
Title card: Delhi's Diwali AQI Data Went Blank at the Worst Hour, Air Quality Sensors, smartcitiesmission.com
Internet of Things

Delhi’s Diwali AQI Data Went Blank at the Worst Hour

Smart City Bot

On Diwali night, PM2.5 readings at most Delhi monitoring stations went blank near 1,000 µg/m³, the official AQI came in lower than the air, and the new pollution command centre raised no city-wide alert because its alarms look for one station spiking, not all of them

This post argues that Delhi’s Diwali data gap shows how a city sensor network should behave on its worst night: a reading at the instrument’s ceiling, a station that falls silent and a whole city turning bad together each need their own rule, written into the command centre’s alarm logic and the operator’s contract.

At 1 am on 9 November, the peak of Diwali night, 36 of Delhi’s 59 monitoring stations had no PM2.5 reading in the Central Pollution Control Board’s records.[3] PM2.5 is the fine particulate smoke from firecrackers. When the Delhi Pollution Control Committee (DPCC) released its own station data on Monday evening, the missing hours showed concentrations of up to 3,744 micrograms per cubic metre at Mundka, more than 60 times the national 24-hour standard of 60.[3] The highest figure anywhere in CPCB’s records for the night was exactly 1,000.[2]

The smoke peaked at 1 am, and so did the number of stations that stopped reporting

The Centre for Research on Energy and Clean Air (CREA) counted one station with missing data at 7 pm on Diwali, seven at 10 pm, 29 by midnight and 36 at 1 am.[4] Across 39 long-standing stations, Hindustan Times found 42% of station-hours between 10 pm and 6 am blank, against about 7% on ordinary nights this season.[2]

The stations were not dead. At 1 am, 38 of those 39 stations sent a nitrogen dioxide reading, while only 13 sent one for PM2.5, and the particulate readings mostly stopped as they approached 1,000 µg/m³. Stations run by all three operators, CPCB, DPCC and the Indian Institute of Tropical Meteorology (IITM), went blank.[2]

This mattered for the number most people saw. The official AQI is a running 24-hour average, and CPCB’s method needs only 16 hours of data for a pollutant’s sub-index, so a station with blank hours still contributes a figure built from its cleaner hours.[8] Delhi’s AQI in the 4 pm bulletin the day after Diwali was 344. By HT’s count, that broke a run since 2015 in which the day after Diwali was always worse than the day before.[2] Adding the readings DPCC later released, HT estimates the bulletin would have read about 360, and even that is a floor, because blanks at CPCB and IITM stations remain.[2][3]

The high readings existed, but they never reached the official record

Every one of the 125 station-hours missing from CPCB’s records at DPCC stations shows a value above 1,000 on DPCC’s own website, and where both sources have a reading, the figures match.[3] So the instruments measured the smoke. Somewhere between the station and the central record, values above 1,000 dropped out. HT said the pattern points to a cut-off at about 1,000 µg/m³, though why it exists, or whether it is new, is unclear.[2]

Officials have offered partial explanations. A DPCC official told the Times of India there was a “technical glitch” in the display data, that no station was switched off and that all data was available in the log.[5] The Indian Express reported that officials put the outages down to a mix of software clipping on central servers and physical stress on the sensors, with sudden spikes triggering filter-tape changes or inlet purging that cause hourly dropouts.[6] CPCB had not responded to questions.[5]

Clipping means anything above a set limit is cut off. We do not know which analysers Delhi uses, but Met One’s BAM 1020, a widely used beta attenuation monitor for particulate matter, lists a measurement range of up to 10,000 µg/m³, with selectable output ranges that include 0 to 1 mg/m³, which is 1,000 µg/m³.[12] On instruments like that, a 1,000 ceiling can be a configured setting rather than a physical limit, which is why it should be written down in advance.

Then there is the question of trust in the numbers that did arrive. DPCC’s peak hourly PM2.5 reading for Jahangirpuri was shown as 719 µg/m³, then revised to 4,985, then revised again to 776, all on the same Monday.[5] Envirocatalysts founder Sunil Dahiya pointed out that a near-zero reading of 0.1 from Sirifort was counted in the city average.[5] A feed that changes three times in a day without a stated reason will be read as manipulation, and the opposition duly called it “data fraud”.[4]

An alarm built for one bad station stays quiet when all 59 turn bad together

Delhi has a new tool for nights like this. The DPCC’s Integrated Command and Control Centre (ICCC), opened on 1 October at the DMRC IT Park in Shastri Park, combines air quality stations, AI cameras watching more than 700 construction sites, GPS tracking of road sweepers and field teams, and a complaints desk.[9] At the regional level, the Commission for Air Quality Management’s PAVAN-I3C platform pulls in data from 111 stations across Delhi-NCR and escalates alerts to the agency concerned when norms are breached[10] (see our weekly roundup).

On Diwali night the centre had 18 staff on duty against the usual 11, and it handled 14 public complaints and 74 dust alerts in the 24 hours after Diwali, about 51 of them from construction sites.[1] The site-level alerts did their job. At one house construction site in Panchsheel Park, PM10 touched 1,000 against a limit of 300; a data engineer called the site, asked for sprinklers and covers, and the level fell below 300 within half an hour.[1] That loop rests on DPCC’s rule that construction sites of 500 square metres or more must install PM2.5 and PM10 sensors and share a live camera feed.[11]

The city-wide picture was different. An official told The Indian Express that the system generates an alert only when pollution rises sharply at a particular station, and that no such alerts came on Diwali night because levels were “more or less similar across stations”, nearly all close to severe.[1] An alarm designed to find the odd one out had nothing to find when everyone was bad, and it cannot notice half its stations going silent, because a silent station does not rise sharply either. As our guide to command and control centres stresses, a control room is only as good as the rules behind its screens.

Four rules would have turned Diwali’s blank hours into a working alarm

None of this needs new hardware. The first rule is to treat a reading at the ceiling as “at least”, not as missing. CREA analyst Manoj Kumar argued that such hours should be flagged as possibly beyond the instrument’s range and counted as 1,000 µg/m³, a lower bound, instead of being dropped from averages.[6] The official AQI is capped at 500 anyway, and HT notes that any hour above about 420 µg/m³ already maxes it out, so keeping these hours in costs little and gains honesty.[3]

The second is to make silence itself an alarm. A station that stops sending one pollutant while still sending others, after a climbing last value, is a high-priority event. CREA’s count had 23 stations silent by 11 pm,[4] so a simple tally of reporting stations on the video wall, with a threshold, would have flagged trouble well before midnight.

The third is to pair spike alerts with absolute ones. Comparing stations finds a local source like a burning dump; a partner rule should fire when most stations cross a fixed level together, because that is when health advisories matter. GRAP Stage III had been in force since 6 November, and Stage IV is meant for an AQI above 450.[7]

The fourth belongs in the operations contract. Data capture should be measured hour by hour and reported separately for high-pollution episodes, since a good monthly average can hide a failure on the one night that counts, and every revision to published data should carry a timestamp and a reason. We argued in our piece on Ahmedabad’s digital twin that upkeep is the real test of city technology, and an air network is no different.

Building owners should ask the same questions of their own sensors. Many buildings use PM sensors to decide when to bring in fresh air. Take a hypothetical office tower whose outdoor sensor reports nothing above 1,000: if the building management system reads that blank as “keep the normal fresh-air schedule”, it pulls smoke indoors on the night it should be closing dampers. Ask your BMS integrator for the sensor’s range, what the controller does with an out-of-range or missing value, and whether that raises an alarm.

More bad nights are coming, with stubble smoke and falling temperatures ahead.[1] The sensors are in place; the alarms and published numbers now have to hold up on the nights when the air is worst.

Planning a command and control centre? Talk to Trisim Global Solutions.

Trisim Global Solutions control room solutions

Trisim’s view: a blank reading on a bad night should set off an alarm, and any command centre, city or building, should have that rule in writing before the next Diwali.

Sources

  1. The Indian Express, 14 complaints, 74 alerts: Inside Delhi’s pollution control war room after Diwali, 10 November 2026
  2. Hindustan Times, Delhi pollution monitors went dark as AQI surged on Diwali, 10 November 2026
  3. Hindustan Times, Delhi’s missing Diwali night data released; PM2.5 at over 60 times national standard, 10 November 2026
  4. The New Indian Express, Delhi AQI data under scrutiny as monitoring stations go offline amid peak Diwali pollution, 9 November 2026
  5. The Times of India, Missing readings, conflicting peaks: As pollution peaks in Delhi, 35 stations go blank for hours, 10 November 2026
  6. The Indian Express, As Delhi’s air turned toxic on Diwali, 45 monitoring stations had missing readings, 9 November 2026
  7. The Indian Express, Delhi’s AQI is 370, but some platforms show 1,000: Why do readings differ?, 9 November 2026
  8. Central Pollution Control Board, How is AQI calculated? (PDF), undated, accessed 10 November 2026
  9. The Indian Express, Delhi government launches control centre to monitor pollution sources in real-time, AI analysis, 2 October 2026
  10. Press Information Bureau, Union Environment Minister visits CAQM’s ‘PAVAN-I3C’ Integrated Command and Control Centre to review Air Pollution Mitigation Operations, 30 October 2026
  11. The Economic Times (PTI), DPCC rolls out Dust Portal 2.0 to tighten monitoring of construction-related pollution, 8 July 2026
  12. Met One Instruments, BAM 1020 Beta Attenuation Mass Monitor datasheet (PDF), March 2026

Leave a Reply