Insights

What 25,585 recordings in buildings and 6,947 journeys on public transport reveal about the air we share indoors. A first look at the open dataset, as of 4 November 2026.

The dataset

Volunteers measure places open to the public with their own CO2 monitors, using the free IndoorCO2Map app. Each recording is linked to a place on OpenStreetMap, which gives every place a fixed ID and a type such as café or doctor’s practice. The app reads the monitor once a minute for at least five minutes, and the recording then appears on the map and in the open dataset.

25,585

recordings in buildings

12,009

places

57

countries

6,947

journeys on 1,337 transit lines

Open data export of 4 November 2026, recordings since April 2024. Around 95% of the data so far was collected with Aranet4 monitors.

Read this first: the limits of the data

  • A first, exploratory look. For many categories the number of recordings is still low and not representative, so the results should be interpreted with caution. Some breakdowns rest on only a few places or journeys, such as 95 pubs or 17 Eurostar journeys, and mainly show what becomes possible as the dataset grows.
  • Volunteers choose where to measure. That can bias the results, probably less for supermarkets, chemists and doctors’ practices and more for restaurants.
  • Mostly Germany and Austria. About 70% of all recordings come from these two countries, so results for all countries mainly reflect them.
  • Snapshots, not safety ratings. A recording shows the air at one moment, in one part of a building. The same place can be very different at another time, during an event or in another room.

Places

All countries, but most data from Germany and Austria

About three in ten pubs and bars average above 1,400 ppm, not a single DIY store does

Each bar shows how the measured places of one type are spread across the three CO2 levels of the GO IAQS Score by GO AQS (Global Open Air Quality Standards): good up to 800 ppm, moderate from 801 to 1,400 ppm and unhealthy above 1,400 ppm. Each place counts once, with the average of its recordings.

  • Good: up to 800 ppm
  • Moderate: 801 to 1,400 ppm
  • Unhealthy: above 1,400 ppm
Pubs (95)29%38%33%
Bars (118)43%28%29%
Places of worship (101)48%35%18%
Sports centres (80)49%34%18%
Theatres (178)39%45%16%
Doctors’ practices (268)39%49%13%
Cinemas (173)54%36%
Restaurants (842)60%31%
Supermarkets (2,686)37%55%
Cafés (488)59%33%
Hospitals (147)74%22%
Museums (322)70%26%
Pharmacies (197)72%25%
Clothes shops (446)78%20%
Fast food (449)82%16%
DIY stores (196)94%
All place types (12,009)59%34%
All countries, open data export of 4 November 2026. Place types from OpenStreetMap; numbers in brackets are measured places. Each place counts once, with the average of its recordings. Levels as defined by the GO IAQS Score.

13%

of doctors’ practices average above 1,400 ppm

Only 39% stay at or below 800 ppm. In hospitals, the share above 1,400 ppm is 4%.

Only 29%

of pubs stay at or below 800 ppm

Pubs do worst overall. Yet every type of place in the chart also has places with good air, which shows that good air is achievable everywhere.

The bad tail by place type

Typical visits differ little between place types, the worst ones a lot

Averages hide the bad outliers. Each box shows the middle half of the visits to one place type, with the median as a line; the whiskers run from the 10th to the 90th percentile. Every place counts the same, however often it was measured, as if you visited each place once. The medians only range from about 570 ppm in DIY stores to 1,040 ppm in pubs, but the 90th percentiles from 720 to 2,720 ppm.

  • Middle half of visits
  • Median
  • 10th to 90th percentile
  • Numbers: 90th percentile in ppm
  • Moderate from 801 ppm
  • Unhealthy above 1,400 ppm
Pubs2,720
Bars2,479
Places of worship1,631
Theatres1,580
Schools1,571
Doctors’ practices1,504
Cinemas1,401
Supermarkets1,387
Restaurants1,372
Cafés1,350
Hotels1,284
Hospitals1,176
Bakeries1,156
Drugstores1,110
Museums1,102
Clothes shops980
Shopping centres956
Fast food920
DIY stores723
Location-weighted percentiles of visit averages: every place counts the same in total, whatever its number of recordings. Dashed lines: the GO IAQS limits at 800 and 1,400 ppm. Open data export of 4 November 2026, all countries, place types with at least 60 measured places, sorted by the 90th percentile.

Pubs and bars: one visit in ten is above 2,700 ppm in pubs and nearly 2,500 ppm in bars, one in twenty above 3,900 ppm in pubs and 3,500 ppm in bars.

Doctors’ practices: one visit in ten is above 1,500 ppm, one in twenty about 1,800 ppm or more.

Schools are probably biased: most school measurements are likely taken by teachers who control the windows and are more aware of air quality than the average teacher.

Supermarkets in Germany

What the data shows · Germany

More than a third of Lidl stores and over a quarter of Aldi Süd stores average above 1,400 ppm

The chains sell similar goods in similar shops and still end up far apart. Kaufland is the only chain where most stores stay at or below 800 ppm; at Lidl and Aldi Süd, fewer than one store in ten does.

  • Good: up to 800 ppm
  • Moderate: 801 to 1,400 ppm
  • Unhealthy: above 1,400 ppm
Lidl (186)55%36%
Aldi Süd (137)62%28%
Aldi Nord (105)18%67%15%
Netto (139)21%71%
Norma (53)43%49%
Kaufland (69)54%39%
Rewe (303)29%64%
Penny (80)26%68%
Edeka (250)42%54%
All supermarkets (1,703)31%58%11%
Germany, open data export of 4 November 2026. National chains with at least 50 measured stores; numbers in brackets are stores. Each store counts once, with the average of its recordings (5,753 supermarket recordings in total). Levels as defined by the GO IAQS Score.

Public transport

Public transport · All countries

Trains have the worst air in public transport, trams the best

  • Good: up to 800 ppm
  • Moderate: 801 to 1,400 ppm
  • Unhealthy: above 1,400 ppm
Regional trains (1,332)34%53%13%
Long-distance trains (482)19%71%
Bus (1,157)58%34%
S-Bahn (533)42%50%
Metro (1,847)52%44%
Tram and light rail (1,526)62%37%
All journeys (6,877)47%45%
Transit data export of 4 November 2026: 6,947 journeys on 1,337 lines, all countries. A journey’s value is its average over the recording; numbers in brackets are journeys. Each line counts the same, however often it was measured. Levels as defined by the GO IAQS Score.

62%

of tram and light rail journeys stay at or below 800 ppm

Buses follow with 58% and metros with 52%.

13%

of regional train journeys average above 1,400 ppm

More than on any other mode. On long-distance trains, only 19% of journeys stay at or below 800 ppm, and their median of 1,013 ppm is the highest of all.

Long-distance trains: half of all ICE journeys average between 906 and 1,204 ppm, and only 0.7% average 2,000 ppm or more. Intercity trains have a longer tail: one journey in ten averages above 1,850 ppm. One EuroCity journey from Munich to Zurich averaged 8,441 ppm.

Public transport · All countries

Buses and trams get worse in winter, long-distance trains barely change

The same bus and tram lines average about 135 to 180 ppm more from October to March, and close to four in five of these lines are worse in winter. Long-distance trains stay between 1,030 and 1,060 ppm in both halves of the year, on the same 41 lines.

  • April to September
  • October to March
Long-distance trains (41)1,0561,033
Regional trains (85)915972
S-Bahn (32)822875
Metro (56)768804
Bus (80)677854
Tram and light rail (81)647782
All countries, transit data export of 4 November 2026. Each line’s average CO2 in both halves of the year, then the median across lines measured in both; numbers in brackets are lines.

Activity by region

Where the data comes from

Vienna, Leipzig and Berlin lead

277 regions have at least one recording and 6 have more than 1,000, with California as the only one outside Europe. The top 10 regions hold 48% of all building recordings.

Vienna2,732
Leipzig1,611
Berlin1,487
Upper Bavaria (Munich)1,183
Düsseldorf region1,042
California1,040
Cologne region991
Schleswig-Holstein854
Dresden region829
Zuid-Holland557
NUTS 2 regions in Europe, states and provinces elsewhere. Building recordings only. Open data export of 4 November 2026.

Check it yourself

All data behind these charts is open. Download it, explore it online or look at the code.

Help us collect more data

The more places are measured, the clearer the picture gets: more countries, more place types and enough recordings to compare cities, chains and seasons. All you need is a portable CO2 monitor and the free app.

Contribute