Data sources

Every line in a sitedia diagram comes from public data. Nothing is drawn by hand and nothing is invented. This page names each dataset, links to it, and says how complete it actually is — including where it is thin.

Two things are worth knowing before the tables. First, most of the panels come from national datasets (Overture, OpenStreetMap, USGS, NAIP). Second, Building age, Building use and Plot size depend on the city's own tax-lot data — which is why the US edition covers twelve cities for now: New York City, the City of Los Angeles, San Francisco, Boston, Washington, DC, Chicago, Miami, Dallas, Houston, Denver, Seattle and Minneapolis.

Nationwide layers

These render for any US address.

Overture is itself a compilation — its building footprints draw on OpenStreetMap, Microsoft and Esri open data. Where a footprint originates in OpenStreetMap it carries the ODbL license, so if you redistribute a derived database the share-alike term applies. Drawings and images made from it do not trigger that; see Terms.

Building heights

Overture's heights are OpenStreetMap tags or, where there is none, Microsoft's estimate from aerial imagery — an average over the roof that runs low on tall buildings. In seven cities we use the height the city or county measured instead, building by building, and where a city publishes none we measure it ourselves from public LiDAR:

We take the city's height only when both footprints are the same building (they overlap by at least half), the city's height is 2.5 m or more, and — where Overture puts a building above 30 m — the city's height is at least half of Overture's. A much lower city height there usually means the tower went up after the survey, so Overture's height stays. The same rules apply to our LiDAR heights. Everywhere else, heights are Overture's.

City data

The US edition covers New York City (the five boroughs), the City of Los Angeles, San Francisco, Boston, Washington, DC, Chicago, Miami, Dallas, Houston, Denver, Seattle and Minneapolis for now. Building age and building use come from each city's tax-lot data, joined to footprints by lot. The Details tables add parks, streets, zoning and transit names from the sources below.

Zoning panel. Each city's zoning districts (the «Zoning» rows below; in New York, MapPLUTO's zoning district of each tax lot) are grouped into eight classes so one legend reads everywhere — residential, mixed use, commercial, industrial, public and institutional, open space, planned development and other. The grouping is ours; the district code in the panel details is the official designation. Houston has no zoning ordinance, so the panel is not drawn there.

Special districts panel. Each city's own maps of special plans (special purpose and special use districts, specific plans, special area plans, planned developments, institutional master plans), local historic districts, overlay districts and redevelopment or tax increment areas (the «Special districts» rows below), redrawn in four colors. Expired, rescinded and inactive areas are left out, and so are areas that only restrict a kind of business (liquor sales, check cashing, signs). Overlay districts larger than 2 km² (0.8 sq mi) are drawn as outlines only, so they do not tint the whole panel; the details window still lists them.

Public facilities panel. Schools, civic and cultural buildings, markets, and parking lots and rail land as mapped in OpenStreetMap (ODbL), from Geofabrik's state extracts — the sites themselves, not designated land (the US has no designation of planned public facilities). Parking lots under 1,600 sq ft (150 m²) are left out.

New York City

Los Angeles

San Francisco

Boston

Washington, DC

Every DC dataset below comes from Open Data DC, Government of the District of Columbia, under CC BY 4.0.

Chicago

Parcels and building records come from Cook County Open Data (Cook County Assessor's Office); zoning, parks and street classes from the City of Chicago Data Portal.

Miami

Parcels, condominium units, streets, parks and transit come from Miami-Dade County GIS, provided as is; zoning from the City of Miami Open Data under CC BY 4.0.

Dallas

Parcels and appraisal records come from the Dallas Central Appraisal District (DCAD), provided as is; zoning, parks and street classes from City of Dallas GIS. The City's data is for informational purposes only — it is not a survey and not suitable for legal, engineering or surveying purposes (Texas Government Code § 2051.102).

Houston

Parcels and years built come from the Harris Central Appraisal District through the Texas Geographic Information Office's StratMap 2026 Land Parcels (CC0); land use, parks and major thoroughfares from City of Houston GIS, which is in the public domain and provided for general reference only; bus stops from METRO's transit data. Houston has no zoning ordinance, so there is no zoning column.

Route and arrival data provided by permission of METRO. METRO is the Metropolitan Transit Authority of Harris County, Texas; it does not endorse sitedia.

Denver

Parcels, original years built, land use, zoning, parks and street classes come from the City of Denver Open Data Catalog, licensed under CC BY 3.0. The City provides the data as is, without warranty, and not for engineering purposes; the City does not endorse sitedia. Bus stops and rail stations come from the Regional Transportation District's GTFS schedule data, used under RTD's license agreement; sitedia is not endorsed by, sponsored by or affiliated with RTD.

Seattle

Parcels, years built and present use come from the King County Department of Assessments, through the City of Seattle's parcel layer and King County GIS Open Data; zoning, parks and street classes come from the City of Seattle's open data. Both are provided as is, without warranty; neither King County nor the City endorses sitedia. Station and stop names come from OpenStreetMap.

Minneapolis

Building use, years built, floor area, stories and units come from the City of Minneapolis Assessing Department's 2026 parcel data, and zoning, parks and street classes from the City's open data — all dedicated to the public domain (CC0). Parcel outlines come from Hennepin County GIS, provided as is. Station and stop names come from OpenStreetMap.

How complete is it, really

Coverage is not a promise, it is a measurement. The table below is the share of buildings that actually carry a year or a use inside a ½-mile circle in Midtown Manhattan, Downtown Los Angeles, Union Square in San Francisco, Back Bay in Boston, Downtown Washington, DC, the Loop in Chicago, Brickell in Miami, Downtown Dallas, Downtown Houston, Downtown Denver, Downtown Seattle and Downtown Minneapolis, counted the same way from our own database on 2026-10-03 (Chicago, Miami, Dallas, Houston and Denver 2026-10-04, Seattle and Minneapolis 2026-10-06).

* In Chicago, rooftop structures and footprints under 20 m² are left out of the count — about a third of the features in the Loop (with them: 802 buildings, 59.1% and 88.9%). Most of the remaining gaps in the Loop are exempt buildings on park land, such as the Art Institute and Millennium Park, which have no assessor record. Lincoln Park, Wicker Park and Hyde Park come out at 97.0%, 98.0% and 89.6% for year built. In Miami, Little Havana, Coconut Grove and Wynwood come out at 96.8%, 94.1% and 89.2%; before condominium units were read, Brickell was at 64.8%. In Dallas, Uptown, Deep Ellum and Lower Greenville come out at 96.4%, 94.9% and 95.8%; across all 334,646 buildings inside the city limits, 94.8% carry a year and 99.5% a use. In Houston, Midtown, Montrose and the Heights come out at 95.8%, 98.4% and 98.7% for year built; across all 636,602 buildings inside the city limits, 94.6% carry a year and 89.4% a use. In Denver, LoHi, Capitol Hill and Cherry Creek come out at 97.1%, 97.6% and 97.0% for year built, and Union Station at 89.4%; across all 266,331 buildings inside the city limits, 98.5% carry a year and 99.7% a use. In Seattle, Capitol Hill, Fremont and Ballard come out at 95.7%, 96.3% and 94.4% for year built; Pioneer Square (72.5%) and South Lake Union (77.0%) are lower for the same reason as Downtown — structures on street right-of-way, public buildings, piers and port terminals have no assessor year. Across all 300,350 buildings inside the city limits, 97.1% carry a year and 99.0% a use. In Minneapolis, Uptown, Northeast and Dinkytown come out at 99.0%, 99.9% and 95.1% for year built, and the North Loop and Mill District at 91.2% and 82.9%; across all 177,282 buildings inside the city limits, 99.1% carry a year and 99.8% a use.

Buildings without a match are drawn in light gray and labeled Unknown. They are not guessed at and not hidden.

Known limits

In Chicago, Miami and Dallas the trees panel was checked against canopy maps, not tree inventories. Chicago: within 2.8 points on average of the Chicago Region Trees Initiative's census-tract canopy across twelve tracts. Miami: 15.8% against 15.2% for Miami-Dade County's 2023 canopy across eight areas, 96% of it within 3 m of our detection; around the Brickell and Downtown towers our detection runs 2–5 points high. Dallas: 1.8 points high on average and 3.9 points in absolute terms against the City's hexagon canopy across 24 random hexagons; 1.9 points low in six Downtown and Uptown hexagons.

Houston. Bus stops come from METRO's own stop list, so the transit panel shows every stop METRO publishes (OpenStreetMap had only about 80 around Midtown and Montrose); stations come from OpenStreetMap. The trees panel was checked against the City's street tree inventory: 91.0% of 5,416 trees in eight neighborhoods fall on detected canopy. In the parts of Houston in Fort Bend and Montgomery counties (about 3,300 buildings, 0.5% of the city) the state's parcel files carry no year built, so those buildings show their use but not their age. Land use is the appraisal district's 2022 code as the City published it.

Minneapolis. The City's tree inventory dates from 2012 and its canopy by neighborhood carries no survey year, so the trees panel was checked against the Metropolitan Council's Growing Shade 2021 canopy for the City's eleven communities (Sentinel-2 imagery at 10 m, calibrated to the University of Minnesota's 1 m land cover). Covering each community with ½-mile windows, the panel comes out 2.7 points high on average and 3.0 points off in absolute terms; the densest communities read highest — Powderhorn 7.3, Downtown (Central) 6.4 and Phillips 6.3 points high.

Seattle. The trees panel was checked against the City's 2021 tree canopy by census tract. With the settings used elsewhere our detection ran about 12 points high — lawns, playfields and golf courses read as canopy — so Seattle uses its own lawn filter. Of the 57 tracts that fit inside a ½-mile window, 29 were used to set it; on the other 28 the panel comes out 1.1 points high on average and 2.7 points off in absolute terms, and 96.3% of the City's street trees (6 in. trunk or larger, ±3 m) fall on detected canopy. Downtown and South Lake Union still read about 5 points high. Year built is the original year of the newest building on each parcel; the City's other parcel layers carry an effective (remodel-adjusted) year, which is not used.

Denver. The trees panel was checked against the City's 2020 tree canopy by neighborhood. With the settings used elsewhere our detection ran about 6 points high, mostly from small lawns in residential neighborhoods, so Denver uses a stricter lawn filter: across 13 neighborhoods not used to set it, the panel comes out 0.8 points high on average and 2.5 points off in absolute terms (Congress Park the furthest, 6.3 points low), and 92.5% of 9,322 trees in the City's tree inventory fall on detected canopy. On sites with a radius over about 0.75 mi (1.2 km) the aerial image is coarser and the canopy reads 2–3 points lower. Bus stops and stations follow RTD's current schedule, which changes in January, May and August; the panel is refreshed with each new schedule.

Dallas beyond Dallas County. The Dallas Central Appraisal District covers Dallas County only. The Collin County part of the city takes its years from the Collin Central Appraisal District's 2026 data, the Denton and Kaufman County parts from the state's StratMap 2026 parcels. Rockwall County publishes no year built, so the few dozen buildings on the Rockwall side of Lake Ray Hubbard show their use but not their age.

In Washington, DC the trees panel is the District's measured canopy, not our detection. Checked against the District's 2020 land cover, our aerial-photo detection painted the long shadows of downtown buildings as trees (Downtown 25% against 10%). DC publishes its 2020 Urban Tree Canopy at 1 m, so the panel draws that directly; it does not show trees planted or lost since 2020. Beyond the District line (Maryland, Virginia) the panel falls back to the same NAIP detection as the other cities.

Terrain is relative, not absolute. The elevation tiles are good for reading slope and landform, but they are not a survey. In dense downtowns they sometimes carry bridges, elevated roadways or structures as if they were ground. We remove artifacts narrower than about 260 ft, which leaves real hills untouched, but wider ones survive — a broad elevated highway can still read a few feet higher than the street beside it. Never quote a spot elevation from these diagrams.

Aerial imagery is not live. NAIP is flown on a multi-year cycle by state, so a recently completed building may be missing from the photograph while appearing in the Buildings panel.

Plot size uses the same lot polygons as building age and land use — MapPLUTO in New York, the County Assessor's parcels in Los Angeles — so the lot grain and the building data always agree.

Zoning and city-planning facilities are still Korea-only. There is no national equivalent in the United States — zoning is written by each municipality in its own categories, and getting a legally consequential boundary subtly wrong is worse than not drawing it. Both cities' zoning does appear in the parcel Details table — New York's from MapPLUTO (ZoneDist1), Los Angeles' from City Planning — but not as a panel. For Los Angeles we do not list a maximum FAR: it depends on the height district, D and Q conditions and community plan footnotes, so check ZIMAS.

Corrections

If a diagram disagrees with what you know about a site, the fault is usually upstream and worth reporting to the publisher above — but tell us too, using the feedback button on the main page, and we will note it or drop the layer.

← Back to sitedia · About · Terms · Privacy