Adaad · Volume 1, Issue 4 · Wednesday 18 November 2026 · Climate, energy and resources
Pollution/Volume 1, Issue 4 · Plotline

The coast is clearer: Karachi emits more, but Lahore breathes worse

Karachi emits more fine particles per person in EDGAR’s annual estimates. Yet from 1 November 2025 to 16 October 2026, Lahore’s sensor readings fell in the unhealthy-or-worse range on 209 days, against Karachi’s 85. The comparison raises a question: what turns emissions into the air a city breathes?

Hiba Sameen · analysis
Adaad data desk · charts and files
Wednesday 18 November 2026 · Volume 1, Issue 4
4 min read · Download the data (CSV)

In November 2024, Lahore's air reached record levels of pollution and the Punjab government closed schools across the province and set up a "smog war room". Smog is usually discussed as Lahore's own problem, the product of the city's traffic, factories and fires. Comparing Lahore with Karachi shows why that framing is incomplete.

Karachi is Pakistan's largest city, with 20.6mn people in its built-up area against Lahore's 14.1mn. We set two kinds of data side by side for the two cities. The first is the European Commission's EDGAR database, which estimates how much of each pollutant is released inside each city in a year. The second is 350 days of readings from Hawanama, the station network run by the Pakistan Air Quality Initiative (PAQI), which reported from 78 locations in Lahore and 27 in Karachi over the period.

Figure 1 shows both, month by month, for each city. The grey row shows the fine particles the city releases, per person per day. The coloured row shows the air on a typical day that month, measured as fine particles (PM2.5) in micrograms per cubic metre of air. Choosing "All rows" adds the gases each city releases, the fires detected within 200km of it and how well its air is ventilated.

Figure 1 · What Karachi and Lahore release and breathe, month by month · interactive
Karachi releases more fine particles per person than Lahore in EDGAR's estimates, but Lahore's air was unhealthy on 209 of 350 days and Karachi's on 85

Emissions: 2024 annual estimates (sulphur dioxide: 2022), split across the months by Adaad using EDGAR's monthly profiles for Pakistan. Air, and the fires and ventilation under All rows: 1 November 2025 to 16 October 2026 (350 days).

Show the numbers
Emissions are EDGAR's annual estimates for 2024 (2022 for sulphur dioxide). Adaad split each source's annual total across the months with EDGAR's monthly profile for that source in Pakistan, then divided by the days in the month and by GHSL population, so the monthly tiles are an allocation, not a measurement, and the annual figures in the Year column do not depend on it. Fine particles count only those emitted as particles. Sources follow EDGAR Cities' sectors: buildings is small-scale fuel burning in homes, shops and offices; transport is road, rail, domestic shipping and domestic aviation; "Other" is agriculture and waste. Sulphur dioxide is not in EDGAR Cities, so it comes from EDGAR's 2022 maps (about 10 km squares) cut to each city's urban centre. "All rows" adds nitrogen oxides, sulphur dioxide, fires and ventilation. "By source", or clicking a row there, splits each release row by source; the sources add up to the row above, and the percentages are each source's share of the year. Fires are satellite detections of burning (NASA VIIRS, nominal and high confidence) within 200 km of the city centre, on either side of the border, averaged over the days of the month the record covers. They are a proxy for burning, not a count of fires or a measure of smoke, and they miss fires lit between the satellite's early-afternoon and night passes. The record lacks seven days of November 2025. Ventilation is ERA5's estimate of the depth of the layer of air that pollution mixes into, multiplied by the wind speed 10 metres above the ground: a rough measure of how fast the air over a city is replaced, in thousand square metres a second. Air is 1 November 2025 to 16 October 2026 (350 days). For each day, the median of the daily PM2.5 readings from the Hawanama stations in the city (at least three needed); the typical day is the median day of the month. Colours apply the US EPA's 24-hour PM2.5 bands to that network median (µg/m³): good up to 9.0, moderate to 35.4, unhealthy for sensitive groups to 55.4, unhealthy to 125.4, very unhealthy to 225.4, hazardous above. It is not an official city AQI. Hover or tap a month for days in each band. Source: EDGAR Cities and EDGAR v8.1, European Commission JRC (CC BY 4.0); GHSL Urban Centre Database 2025, JRC; NASA VIIRS active fires via Global Forest Watch; ERA5 (Copernicus) via Open-Meteo; Pakistan Air Quality Initiative / Hawanama, station readings of 17 October 2026; US EPA PM2.5 AQI breakpoints (2024); Adaad calculation.

The release figures are EDGAR's estimates for 2024 (2022 for sulphur dioxide), the latest years available. We split each annual total across the months using EDGAR's monthly profiles for Pakistan, so the monthly tiles are an allocation, not a measurement. The air readings are measured, at stations whose hardware Hawanama does not identify and whose readings it publishes unvalidated. PAQI's own published series for 2024 also puts Lahore well above Karachi, at 105 µg/m³ against 46.

On releases, Karachi leads. In EDGAR's 2024 estimates, each resident accounts for 2.9kg of fine particles a year, against 2.3kg in Lahore, or 27 per cent more. In total, Karachi released about 59,000 tonnes and Lahore about 32,000. Karachi also releases more nitrogen oxides and sulphur dioxide per person, two gases that form fine particles in the air.

The breakdown by source shows where Karachi's extra comes from. Fuel burnt in buildings, which EDGAR defines as small-scale combustion in homes, shops and offices, is the largest source in both cities and almost identical per person (1.6kg in Karachi, 1.7kg in Lahore). The gap is industrial. Karachi's industry releases twice as much per person as Lahore's, and its power plants about ten times as much. Transport, which covers road, rail and domestic shipping and aviation, accounts for about 1 per cent of fine particles released in each city. It is larger in nitrogen oxides, at 27 per cent of Lahore's and 13 per cent of Karachi's, and larger per person in Lahore (0.68kg a year against 0.42kg).

On air, the order reverses. We apply the US Environmental Protection Agency's 24-hour bands to the median reading across each city's stations. By that measure, Lahore was in the unhealthy range or worse on 209 of 350 days, and hazardous on five. Karachi was unhealthy or worse on 85 days and never hazardous. Lahore's typical day was worse than Karachi's in every month. In November it held 181 µg/m³ against Karachi's 81. In April, Lahore's cleanest month, it held 37 against 27.

The rankings therefore point in opposite directions. That mismatch is not a test of whether local emissions matter. What a city breathes also depends on how fast its air is replaced and on what arrives from outside.

Karachi is on the coast and Lahore sits inland on the Punjab plain. In ERA5, the European reconstruction of past weather, the wind over Karachi averaged 4.7 metres a second, against 2.1 over Lahore, where the daily average fell below 2 on 166 of the 350 days. In December the layer of air that pollution mixes into was about 155 metres deep over Lahore, against 346 over Karachi. Wind multiplied by that depth gives a rough measure of ventilation, and Karachi's averaged about three times Lahore's.

Satellites recorded an average of 449 fire detections a day within 200km of Lahore in May 2026, after the wheat harvest, and no more than 25 around Karachi in any month. November 2025, after the rice harvest, averaged 130 around Lahore, far below the 1,075 typical of November over 2013 to 2024. Part of that fall reflects timing, because most large farm fires in Indian Punjab are now lit after 3pm, once the satellite has passed. Lahore's air was still about twice as polluted as Karachi's in August and September, when almost nothing was burning.

The release figures also leave out road dust and the particles that form later in the air from gases such as nitrogen oxides, sulphur dioxide and ammonia, and they cover different years from the readings. This data cannot say how much each of these differences contributes.

For policy, closing schools protects children for a day. The comparison strengthens the case for measuring what Lahore's particles are made of and where they come from, so local and regional pollution controls can be targeted more effectively.

Method note

Cities. Each city is its urban centre in the Global Human Settlement Layer (GHSL) Urban Centre Database, the continuous built-up area of the city rather than its municipal or district boundary. Karachi's urban centre may leave out parts of Port Qasim, an industrial and power-plant zone.

Emissions. Fine particles and nitrogen oxides for 2024 come from EDGAR Cities, which publishes emissions by sector for each urban centre. Fine particles here are those emitted directly as particles. Sector names follow EDGAR Cities: buildings is small-scale non-industrial stationary combustion; transport is road, rail, domestic shipping and domestic aviation; other is agriculture and waste. Sulphur dioxide is not in EDGAR Cities, so it is taken from EDGAR v8.1's 2022 maps, which divide the world into squares of about 10km, and summed over the squares each city covers.

Monthly allocation. EDGAR publishes no city-level monthly figures. We spread each sector's annual total across the months with EDGAR's monthly profile for that sector and pollutant in Pakistan, averaged over 2018 to 2022, then divided by the days in the month. The profile places 27 per cent of a year's emissions from buildings in January, which suits a heating season and may overstate winter emissions in Karachi. Because the two cities have different mixes of sources, the monthly comparison between them depends on this choice. If Karachi's buildings emissions are instead spread evenly across the year, keeping its annual total, its January figure falls from 17.7g to 8.1g per person a day, against Lahore's 16.2g. This is a sensitivity test, not a preferred estimate. The annual figures do not depend on the allocation.

Population. Per-person figures divide by each urban centre's population from GHSL, interpolated to 2024 (2022 for sulphur dioxide).

Air quality. Readings come from Hawanama's public data service, saved on 17 October 2026. Hawanama combines PAQI's low-cost sensors with Punjab EPA and Urban Unit reference monitors, but it does not identify the hardware at each station, none of the stations used here matches a Punjab EPA station name, and all readings are published unvalidated. Each daily value is the service's mean of a station's hourly readings over a UTC day (5am to 5am Pakistan time). The service does not report how many hours each day contains, so we cannot confirm complete 24-hour means. We drop 17 October, of which at most two hours had passed when the data was saved, leaving 1 November 2025 to 16 October 2026. Devices at the same location are averaged, and a day counts when at least three locations report. The city's value for the day is the median across locations, and the typical day of a month is the median of those daily values.

Weather. Wind speed 10 metres above the ground and boundary layer height, which we call the mixing depth, are hourly ERA5 values from the Open-Meteo historical archive for the grid points nearest central Lahore (31.5°N, 74.25°E) and central Karachi (25.0°N, 67.0°E), averaged over the same UTC days as the air readings. ERA5 is a model reconstruction on a grid of about 25km, not a measurement in the city, and Karachi's grid square straddles the coast. Ventilation is each day's mean mixing depth multiplied by its mean wind speed, a simplified form of the ventilation index used in air quality forecasting, which uses the wind through the whole mixing layer. It averaged 2,810 square metres a second over Karachi and 969 over Lahore. It was below 500 on 102 days in Lahore and four in Karachi. In each city the air was worst on the least ventilated days. The median reading on the least ventilated quarter of days was 161 µg/m³ in Lahore and 70 in Karachi, against 47 and 25 on the most ventilated quarter. That is an association across the seasons, not an estimate of the effect of weather. Rain does not favour Karachi: ERA5 puts 43mm over Karachi in the period and 338mm over Lahore.

Fires. NASA VIIRS 375m active fire detections (nominal and high confidence) from the Global Forest Watch copy of the near-real-time record, counted within 200km of each city centre on either side of the border, by UTC day. No cropland mask is applied, and one fire can be detected more than once. The satellite passes at about 1.30pm and 1.30am, so fires lit and burnt out between passes are not recorded. The copy is missing 18 of the 350 days, seven of them in November 2025 (2 to 7 and 16 November). Monthly values average over the days observed, which assumes the missing days were like the rest of the month. The 2013 to 2024 average for each month uses months with at least 20 observed days. October 2026 covers the first 16 days only, before burning usually peaks. Within 100km the daily averages are 198 in May and 53 in November around Lahore, and at most 8 around Karachi. Within 300km they are 578 and 216, and at most 52. About 13 per cent of detections around Lahore and 38 per cent around Karachi since 2012 fall in places that burn in at least 24 separate months, which are more likely to be industrial heat sources than fields. Leaving those out gives 396 and 112 around Lahore and at most 19 around Karachi. Across October and November, detections within 200km of Lahore averaged 1,164 a day in 2021 and 159 in 2025. The late-burning finding is from the Stubble Burning Status Report 2025 by the International Forum for Environment, Sustainability and Technology, as reported by The Tribune. It found that over 90 per cent of large farm fires in Indian Punjab in 2024 and 2025 occurred after 3pm, against 3 per cent in 2021, and that the area burnt had also fallen. Detections are a proxy for burning. They do not measure smoke, or how much of it reaches either city.

Stations. Over the period, 78 devices at 73 locations reported in Lahore and 27 devices at 24 locations in Karachi, but not all at once. On a typical day 52 locations reported in Lahore (between 24 and 62) and 12 in Karachi (between 4 and 18). Of these locations, 70 in Lahore and 22 in Karachi lie inside the GHSL urban centres used for emissions.

Checks. We recounted the unhealthy-or-worse days under five alternatives: dropping station days that appear to hold a single hourly reading; keeping only stations that report on at least half of all days, or at least 80 per cent; keeping only stations inside the urban centre; and adding every Hawanama station inside the urban centre. Lahore's count ranges from 204 to 210 days and Karachi's from 81 to 101. The gap between the cities holds in every case.

PAQI series. PAQI's published daily series gives annual means of 105 µg/m³ in Lahore and 46 in Karachi for 2024. PAQI runs both that series and Hawanama, and we have not established which instruments each uses or how far they overlap, so it is a consistent ranking from the same organisation, not an independent check.

Files. The numbers as drawn in Figure 1, the daily city series behind the air rows, the monthly fire and weather figures and the robustness counts are downloadable below.

Sources

Al Jazeera, "Pakistan's Punjab shuts schools, sets up 'smog war room' over air pollution", 6 November 2024.

Copernicus Climate Change Service, ERA5 hourly data on single levels, accessed through the Open-Meteo historical weather API.

European Commission Joint Research Centre, EDGAR Cities (sector definitions, Table 1).

European Commission Joint Research Centre, EDGAR v8.1 air pollutants.

Global Forest Watch, VIIRS active fire alerts, from NASA's Fire Information for Resource Management System.

Mari Rivero and others, Stats in the City: the GHSL Urban Centre Database 2025, Publications Office of the EU.

Omar, A. and Pakistan Air Quality Initiative, Air quality monitoring data for Pakistan (2016–2024), Zenodo.

Pakistan Air Quality Initiative, Hawanama.

The Tribune, "Satellites miss majority of stubble fires, Delhi air pollution underestimated: Report", December 2025.

US Environmental Protection Agency, AQI breakpoints.

Data and methods downloads