Everyone who works with neighborhoods eventually asks a mapping database for the outline, and OpenStreetMap's answer is the most misunderstood in the whole project: sometimes there is a polygon, sometimes there is only a point, and the difference is not an error. It is two modeling traditions living in one database. In the first, the neighborhood is an administrative unit, mapped as a boundary relation with boundary=administrative and a local admin_level, and it has a true, extractable outline. In the second, the neighborhood is a place, mapped as a node with place=neighbourhood, a labeled point asserting that the neighborhood exists and roughly where its heart is, with no boundary claimed at all, because in most of the world neighborhoods have names, reputations, and centers, but no legally surveyed edges. This guide teaches both species, the admin_level system that ranks boundaries, the extraction routes through Nominatim and Overpass that the boundary literature uses, and the honest failure mode that everyone who goes hunting for neighborhood polygons meets. The geocoder itself is owned by the Nominatim guide linked below; here it appears as a tool in the extraction story.
Home / GIS and Open Map Data for Beginners
OpenStreetMap Neighborhood Boundaries: Relations with Polygons, Nodes Without

Photo: Jakob Jin / Pexels
Ask Which Species First: The Boundaries Checklist
Check off each step as you finish it. Your progress stays on this page.
0 of 7 complete
0% done
The Administrative Species: boundary=administrative and admin_level
Administrative boundaries in OpenStreetMap are relations: a boundary object that assembles its outline from way segments, carries the unit's name and metadata, and is tagged boundary=administrative with an admin_level ranking it in its country's hierarchy, from 2 for national borders down through regions, counties, and municipalities to the local levels, 9, 10, and below in many schemes, where districts, quarters, and neighborhoods live when a country administers them at that grain. The ranking is national, not universal: the wiki's boundary documentation and country pages record each nation's scheme, and the United States Boundaries page, for example, places the neighborhood level at admin_level 10, with border_type carrying the local vocabulary of what such units are called. Where a city genuinely administers its neighborhoods, as Paris does its quartiers, the relation species exists and the polygon is real. The relation's structure is the price of admission for extractors: the outline must be assembled from the relation's member ways, which is why the extraction routes below exist as tools rather than as a download button.
The Place Species: place=neighbourhood, a Point with a Name
The other species is a node: a single point tagged place=neighbourhood, placed at the neighborhood's perceived center, carrying its name and nothing spatial beyond the point itself. Its sibling values rank the settlement hierarchy, suburb, quarter, neighbourhood, down to the finest grain the project's place vocabulary names. This is not a failed boundary; it is a different claim. The node says the neighborhood exists, is called this, and is centered about here, and for rendering labels, for geocoding, for the gazetteer function of the map, that claim is exactly what is needed. What it cannot do is answer an area question: no polygon means no inside or outside, no acreage, no which-neighborhood-is-this-address for points near an edge. The practical consequence dominates the rest of this guide: whether your neighborhood has an outline in OpenStreetMap depends on which species your city was mapped as, and discovering which is the first step of any boundary project, before tools are chosen or deadlines promised.

Extraction Route One: Nominatim's Polygon
Where the relation species exists, the friendliest extraction runs through Nominatim, the project's geocoder. A search or lookup for the place can return the boundary itself: the API's polygon output options (the polygon_geojson family of parameters) attach the administrative outline to the result as GeoJSON, ready for a GIS. The tutorial literature on boundary extraction is built on this route because it needs no query language: find the unit in Nominatim, confirm the result is the boundary relation and not a node, request the polygon, save the geometry. The confirmation step is the craft: a search for a neighborhood name may return the place node, the administrative relation, or a same-named feature of another kind, and only the relation carries the outline. Our Nominatim guide owns the geocoder's endpoints and usage policy in full; here the polygon route is one application of them, and its policy limits, a request a second, cached results, no bulk harvesting, apply to boundary work as to all work through that door.
Extraction Route Two: Overpass and the Relation Tools
The second route speaks to the database directly. Overpass API queries can fetch boundary relations by their tags, boundary=administrative with the admin_level and name you seek, across a city or a country at once, which is the route for assembling a whole set of neighborhood polygons rather than one. The community's extraction threads document both its power and its characteristic disappointment: a query over a city's neighborhoods returns the few that exist as relations, in Paris famously at admin_level 10, and silence for the rest, because the rest are nodes or unmapped, and no query can return a polygon the database does not contain. A third, smaller tool completes the kit: services such as polygons.openstreetmap.fr take a relation's ID and return its assembled outline as GeoJSON or other formats, sparing you the member-assembly work when you already know the relation you want. The three routes are one toolbox: Nominatim to discover and fetch singly, Overpass to harvest systematically, the polygon tools to convert a known relation. All three share the same precondition, that the relation exists.
The Honest Failure Mode, and What to Do With It
Every practitioner of neighborhood extraction meets the wall: most neighborhoods, in most cities, have no polygon in OpenStreetMap, because most neighborhoods have no official polygon anywhere, and the database declines to invent one. The mature responses are three. Use the node honestly: for labeling, centering, and proximity work, the place node is the correct data, and a map that shows the point and not a pretend outline is telling the truth. Draw your own with provenance: where an application needs an outline, derive one from a stated method, a municipal open-data boundary, a published definition, a buffer you disclose, and keep the derivation documented, because an undocumented invented boundary becomes somebody else's false fact at the next reuse. Contribute the real thing where it exists: where a city does administer its neighborhoods and an importable, license-compatible source exists, the community's import process is how relation species are born. What not to do is the fourth response, the common one: tracing an impressionistic outline from a neighborhood's reputation and tagging it as administrative. The database's two-species honesty is a feature; the node that claims only a center is worth more than the polygon that claims a false precision.
Reading a City's Answer
The working procedure, compressed: search the neighborhood in Nominatim and identify what the result is, relation or node. If a relation, extract by polygon output or by ID through the polygon tools, and record the admin_level and the country scheme it belongs to. If a node, stop and choose: the node itself for point work, or a documented derivation for area work, with the method written down. If nothing, the neighborhood is unmapped, and the honest output of your project is that sentence, plus optionally the mapping work to fix it. That procedure, unglamorous as it is, is the whole professional secret of neighborhood boundaries in OpenStreetMap: the database knows two kinds of neighborhoods, and the craft is asking which kind yours is before promising anyone an outline.
Reading a Country's Scheme Before You Trust a Level
Because admin_level is national, the number alone tells a stranger almost nothing, and the professional habit is to read the scheme before the level. The wiki's boundary documentation maintains the country-by-country tables: which level is the state or province, which the county, which the municipality, and where, if anywhere, the neighborhood grain sits, with border_type values recording the local names, ward, district, quartier, barrio, that the levels carry in their own languages. Two practical corollaries follow for extraction work. First, a cross-border harvest must translate per country: a query for admin_level 10 across a frontier assembles neighborhoods on one side and municipalities on the other, and the error is silent because both are polygons with names. Second, the place hierarchy is the steadier cross-border instrument: place=neighbourhood means roughly the same claim in every country, because it claims so little, a name and a center, while the administrative ladder means something precise and different in each. Knowing which instrument your question needs, the polygon with its national semantics or the point with its universal modesty, is the judgment this whole guide has been building toward.
Assembling a Boundary Set: A Worked Pattern
For the practitioner assembling many neighborhoods at once, the routes combine into a pattern worth writing down. Discover through Nominatim: search the city's neighborhoods and record which resolve to relations and which to nodes, because that census, not hope, sets the project's scope. Harvest the relations through Overpass in one query per country scheme, filtering on boundary and admin_level, and export the set as GeoJSON through the polygon tooling where assembly is needed. Reconcile names across the two species, since the node and the relation for one neighborhood may carry variant names, and a boundary set that cannot join to its own gazetteer is half-finished. Document the gaps in the deliverable itself: the neighborhoods present as points, listed as points, so the set's consumer inherits your knowledge rather than your silence. That pattern, census, harvest, reconcile, document, is the difference between a boundary dataset and a folder of polygons, and it is achievable in an afternoon for most cities once the two-species reality is accepted.
And a final caution that belongs to every use of either species: names repeat. Every country has its Springfield, every city its Riverside, and a boundary or node harvested by name alone, without a geographic sanity check against the place you meant, will eventually be the wrong neighborhood with total confidence. Verify the geometry sits where your neighborhood sits; the database rewards the check.
Related Guides
If this page answered part of the question, these GIS guides cover the rest of the same ground:
- OpenStreetMap Nominatim: The Geocoder and Its Rules: The extraction route's engine, its endpoints and usage policy owned there.
- OpenStreetMap Tags: The Grammar of the Database: The boundary and place vocabularies in their grammatical home.
- OpenStreetMap Download: The Generic Doors: The extract routes for whole datasets, beside this guide's surgical ones.
- OpenStreetMap Power Lines: The power Key Family: Another dataset whose absences must be read as unmapped, not unreal.
Frequently Asked Questions
Why does one neighborhood have a boundary and another only a point?
They are different species in the data. A neighborhood administered as a unit is mapped as a boundary=administrative relation with a local admin_level, and has a polygon. A neighborhood known by name and center is mapped as a place=neighbourhood node, a point with no outline claimed.
What admin_level is a neighborhood?
It depends on the country's scheme, which the wiki documents per nation. In the United States scheme, admin_level 10 is the neighborhood level; Paris's quartiers likewise sit at level 10 in the French scheme. There is no universal level, only national ones.
How do I extract a boundary polygon?
If the relation exists: through Nominatim with its polygon output options, through Overpass queries on boundary=administrative, or by feeding the relation ID to a tool such as polygons.openstreetmap.fr, which returns the assembled outline as GeoJSON.
Can I get boundaries for all of a city's neighborhoods at once?
Only for those mapped as relations, via Overpass. In most cities that is a minority, sometimes a small one, because most neighborhoods exist in the data as place nodes. No query can return polygons the database does not contain.
What should I do when I need an outline and only a node exists?
Use the node for point work, or derive an outline by a documented method, a municipal source or a disclosed buffer, and record the derivation. Do not trace an impressionistic boundary and tag it administrative; a false polygon is worse than an honest point.
Photo credits: Jakob Jin (Pexels); Huy Phan (Pexels). Photos are served from this site; follow the credit to the library page for the original.