Overture Divisions

Divisions are the administrative areas from Overture Maps: countries, regions, counties, cities, neighborhoods. In ATGeo they do two jobs. They're places you can look up, and they're what relations.within points at when a record needs to say "what city is this in".

Dataset overturemaps.org
Collection division
Tileset slug overture-division
Location a community.lexicon.location.bbox only
Licences ODbL 1.0, CC BY 4.0
Attribution Overture attribution

The origin is urn:atgeo:overturemaps.org:division:{id}, with Overture's GERS id (and the same caveat as places: Overture replaced all of its ids in release 2025-06-25.0).

Location

A division has a bounding box, not a point and not a polygon: north, south, east, west, as decimal strings. A box whose west is greater than its east crosses the antimeridian. The box is not the shape; a bounding box over a crescent-shaped city includes plenty of territory that isn't the city. Don't use it for point-in-polygon tests.

Attributes

Key Notes
subtype country, dependency, region, county, localadmin, locality, macrohood, neighborhood, microhood (borough is mapped but Overture has none right now)
country ISO 3166-1 alpha-2 code
region ISO 3166-2 code, e.g. US-PA
wikidata Wikidata item id, when Overture has one
population Always present. 0 means no population was recorded, not that nobody lives there

country, region and wikidata are left out when the source doesn't have them. Names in other languages arrive as variants.

Importance

Importance is 0 to 100, and ranks against Overture places as well as other divisions. It comes from how many people live in a division or how big it is, and nothing else; subtype and administrative level don't enter:

base = round(100 · max(min(ln(1 + population) / 20, 1), min(ln(1 + area) / 16, 1)))

with area the bounding box's area in km², standing in where Overture has no population. The base is scaled by 0.8, so a division tops out at 80 on size alone. A division with a wikidata item whose Wikipedia page is well known gets a lift into the remaining 20 points, which is what puts Paris above a large, empty district. The constants are the reference producer's choices, not protocol; the API page has the full formula.

Containment levels

relations.within on any record lists divisions broadest first: the order is the nesting, so you can take the last entry as the most specific and the first as the broadest. The wire carries each division's subtype as type and no numeric level. The gazetteer sorts by a level it derives from the subtype:

level subtype
10 country
15 dependency
25 region
35 county
45 localadmin
50 locality
55 borough
60 macrohood
65 neighborhood
70 microhood

The numbers aren't emitted; they're just the sort order, spaced out so there's room to insert. Overture's own admin_level isn't used, since it's mostly empty and ambiguous within a subtype. Each collection's descriptor lists the types its records actually use as containmentTypes; read that rather than hardcoding this table. A different gazetteer could use a different vocabulary.

The overview band

When a request covers too much ground for ordinary tiles, the server answers with a coarse sample instead (see the API). For divisions that sample holds the most important ones. Countries, dependencies and regions score above every other subtype, so they're the ones it keeps.

Example

Center City, a macrohood in Philadelphia, from a live tile (trimmed):

{
  "value": {
    "$type": "org.atgeo.place",
    "origin": "urn:atgeo:overturemaps.org:division:407ce93b-dadf-49eb-b25b-6942e077c089",
    "name": "Center City",
    "locations": [
      {
        "$type": "community.lexicon.location.bbox",
        "north": "39.960094",
        "south": "39.939913",
        "east": "-75.133937",
        "west": "-75.187627"
      }
    ],
    "variants": [
      { "name": "Downtown Philadelphia", "type": "alternate", "language": "en" }
    ],
    "attributes": {
      "subtype": "macrohood",
      "country": "US",
      "region": "US-PA",
      "wikidata": "Q2739353",
      "population": 173284
    },
    "relations": {
      "within": [
        { "uri": "urn:atgeo:overturemaps.org:division:f39eb4af-5206-481b-b19e-bd784ded3f05", "name": "United States", "type": "country" },
        { "uri": "urn:atgeo:overturemaps.org:division:9b885f43-e59d-4249-b114-4d8853d66bfb", "name": "Pennsylvania", "type": "region" },
        { "uri": "urn:atgeo:overturemaps.org:division:1a04f6e5-dd95-4e67-b25d-777814f7c93f", "name": "Philadelphia County", "type": "county" },
        { "uri": "urn:atgeo:overturemaps.org:division:10ddec2b-285d-477a-a39b-bb3209735f9b", "name": "Philadelphia", "type": "locality" }
      ]
    }
  },
  "importance": 54
}

Divisions are polygons, so one division is referenced from every tile it overlaps. Deduplicate by origin when you merge tiles (the client SDKs do this for you).