How city area is measured

How CompareCitySizes calculates area

Every outline on our map is either a boundary our team has stored or one returned by OpenStreetMap's Nominatim search service. When a city is added, our server first checks for a boundary it already holds, whether curated by our team or saved from an earlier lookup. Only if it finds none does it search Nominatim for the city name and country.

What happens next depends on what Nominatim returns. If the result has a polygon of its own, we use that polygon as it is; internally this source is called nominatim_city. If the result is only a point, we look up the administrative areas named in its address, in the order municipality, city, town, village and county, and use the first one that returns a matching polygon. That outline belongs to the containing area rather than the city itself, so it is stored as nominatim_admin and the app labels it "Approximate boundary". If neither of the server's searches returns a polygon, the server returns no boundary. The third source, admin_override, is a boundary curated by our team, and it is never marked approximate.

Unless our team has stored an area with a curated boundary, the area is calculated from the polygon itself. We use the spherical polygon formula described by Chamberlain and Duquette (JPL, 2007), on a sphere with the Earth's mean radius of 6,371,008.8 metres: the area of each outer ring, minus any holes, summed over every part of the shape. A calculated number is therefore a measurement of the outline, not an official figure, and it is only as accurate as that outline.

In October 2026 we corrected an error in the formula used to calculate areas from outlines, which under-reported areas at mid-latitudes: for the commune of Paris it gave about a fifth of the published 105.4 km²^5,6.

Why the same city has different areas

Across the eight countries we have researched, the differences come from four recurring causes.

The first is definition. A city name can label a council, a statistical urban area or a wider region, and each has its own boundary. Our United Kingdom guide shows the City of London, a single local authority district, inside Greater London. Our Australia guide finds three Sydneys: the City of Sydney council area, 26.7 km²^7,8; the Sydney Significant Urban Area, 4,197 km²^9,10; and Greater Sydney, the statistical region the Australian Bureau of Statistics uses for the capital city, 12,369 km²^11,12. Japan (Tokyo Metropolis and its 23 special wards) and France (the commune of Paris and the Métropole du Grand Paris) follow the same pattern.

The second is rural land. In some countries the legal city is a whole territory, rural land included. Our Brazil guide explains that a município covers all the rural land inside its borders: Altamira, in Pará, covers 159,533 km²^13,14, about 105 times the São Paulo município by our calculation. In China, Chongqing Municipality, 82,400 km²^15,16, is a province-level territory that takes in counties as well as districts: 26 districts and 12 counties (2024 yearbook).

The third is gaps in boundary data. An official figure can exist with no outline to match it, or a search can return an outline that does not match the figure. Our India guide shows both: we found no OpenStreetMap relation confirmed as the boundary of Mumbai's municipal corporation, and the outline a search for Delhi returned is smaller than the whole territory.

The fourth is measurement choice: what counts as land, and which edition a figure comes from. The US Census Gazetteer lists land and water separately and gives no total; San Francisco has 120.9 km²^17,18 of land and 479.7 km²^19,20 of water. The Office for National Statistics publishes a total extent for Greater London that includes the tidal Thames, 1,595 km²^3,4, and a land area of 1,572 km²^21,22. Japan's Geospatial Information Authority (GSI) takes the coastline at high water and counts rivers and lakes as land. Figures also change between editions: IBGE publishes a new edition of every Brazilian município's area each year, and the Census Gazetteer is reissued every year.

Problems getting accurate boundaries

A search can return a point instead of a boundary. On 17 October 2026, Nominatim answered "Mumbai, India" with a single point. Repeating the app's fallback search, which looks for an enclosing administrative area, returned only the same point, so Nominatim gave no polygon for Mumbai at all.

A boundary can be missing. We found no OpenStreetMap relation confirmed as the boundary of either municipal corporation in our India guide, and no OpenStreetMap boundary for Australia's Greater Capital City Statistical Areas, or for Chongqing's urban planning area. In Japan, each of Tokyo's 23 special wards is its own relation, and our Japan guide found no single boundary for the wards combined, so we do not draw them as one shape.

A boundary can be the wrong shape. The OpenStreetMap polygon we fetched for Xicheng District in Beijing measured about three times the official 50.7 km²^23,24, and the one for the Mumbai Metropolitan Region about two-thirds of MMRDA's 6,328 km²^25,26. We found no source that explains either gap, so our guides do not draw either.

A boundary can take in the sea. The relation a search for Tokyo returned on 17 October 2026 included wide areas of open sea around the outlying islands; measured by us from OpenStreetMap on that date, it is about 19 times GSI's 2,200 km²^27,28. The Hamamatsu relation extends offshore too, and measured about a third larger than GSI's figure.

An outline can include water that the official figure leaves out. In Jacksonville, San Francisco and New York City, the city limits run out across water, so by our measurement each OpenStreetMap polygon is within one percent of the Census land and water figures added together, and well above the land figure alone. San Francisco's outline reaches as far as the Farallon Islands.

A boundary may not be settled. GSI marks the area of any Japanese municipality whose boundary is not fully determined (境界未定) as a reference value (参考値), and says the area is then an approximate figure given for convenience. Chiyoda ward, the 23-ward total and the Tokyo Metropolis total all carry the mark.

Finally, some figures could not be published: we left out a statistical yearbook table we could not archive, figures for which we found no second source from a different publisher, and a figure for Paris without its two woods that has no primary source.

Country guides

• China

  Chongqing Municipality 82,400 km²^15,16 vs its urban planning area 5,473 km²^29,30

• Japan

  Tokyo Metropolis 2,200 km²^27,28 vs the 23 special wards 627.5 km²^31,32 (GSI, both 2026, both reference values, 参考値)

• United Kingdom

  City of London 3.1 km²^1,2 vs Greater London 1,595 km²^3,4 (ONS total extent, both 2024)

• United States

  San Francisco: 120.9 km²^17,18 of land vs 479.7 km²^19,20 of water (Census Gazetteer)

• France

  Commune of Paris 105.4 km²^5,6 vs Métropole du Grand Paris 814 km²^33,34

• India

  Greater Mumbai's corporation: 483.1 km²^35,36 in its Year Book 2025 vs 437.7 km²^37,38 in its 2011 city profile

• Australia

  City of Sydney council area 26.7 km²^7,8 vs Greater Sydney 12,369 km²^11,12 (ABS)

• Brazil

  São Paulo município 1,523 km²^39,40 (IBGE, 2025) vs the urbanized area of the São Paulo urban concentration (37 municípios) 2,134 km²^41,42 (IBGE, 2019)