Air quality

Europe publishes hourly measurements from thousands of monitors, a satellite maps nitrogen dioxide over the whole continent every day, and a model fills in everything between. Almost none of it can be read at face value. Marylebone Road’s 2023 NO2 is 41.6 in µg/m³ and 21.8 in ppb, over the EU limit in one unit and comfortably under it in the other; the UK air quality index rates North Kensington’s year as worse than the kerb 4 km away, because the index is mostly about ozone.

The spatial traps are sharper still. Monitoring networks are designed rather than sampled, so half of Berlin’s stations describe their own street and interpolating them paints discs around every monitor. A satellite pixel covers 19 km² and tracks the background station (r = 0.68) far better than the kerb (0.30). A 10 km model agrees with the stations suspiciously well because it assimilates them. And exposure figures change with the reporting scale: the same field puts 52% of Berliners above a threshold by grid cell and 34% by district.

The concepts below settle units, indices, networks, meteorology, satellite columns and exposure; the how-to guides download and analyse real station data from London and Berlin, map Sentinel-5P, evaluate CAMS, fit a land-use regression and weight by population; and the fixes are the five ways these numbers go wrong without looking wrong.

Learning paths run through this topic too — see the end of this page.