Catalogue

Each layer on the globe is one or more curated collections. A card says what one record is and is not, who publishes it and on what terms, what it covers, how current it is and where that was read, what is left out and why, and the call that serves it. Counts and dates are as read on the day given, not a live status.

The same entries are JSON at /api/catalog. 25 of 84 layers are catalogued so far; the rest are listed under Signal sources in the globe's About panel. Place documents are indexed at /place.

Land and heritage

National Wetlands Inventory polygons

nwi-wetlands · layer wetlands · show on the globe

What it holds

USFWS's map of wetland and deepwater habitat, one polygon per mapped unit, each with its Cowardin code decoded locally into its system, subsystem, class and water regime names and the year of the imagery it was drawn from (NWI Data_Source).

A record is

A habitat polygon mapped from aerial imagery of a given year. It is not a jurisdictional (Clean Water Act) delineation and does not say the area is wet today.

Sources

National Wetlands Inventory, U.S. Fish and Wildlife Service. Licence and credit: public domain

Coverage

United States and territories; polygons below 10 km camera height, 500 per request (NWI's maxRecordCount); imagery years vary by quad, often decades old.

Freshness

Each polygon carries the year of the imagery it was mapped from; the service does not publish an edit date. Read from Data_Source MapServer layer 3 (Image Year); Wetlands layer 0 ?f=json has no editingInfo, 2026-10-13.

Left out, and why

jurisdictional status: NWI is a habitat inventory, not a delineation; nothing here says what a regulator would decide
guessed names for unknown codes: a code the table does not answer leaves the names null

Caveats

A habitat inventory mapped from imagery that can be decades old, not a jurisdictional (Clean Water Act) delineation.

Sensitivity

Habitat polygons only; no ownership or personal data.

Searchable fields

none: the layer answers by place, not by name

Embedded as record text

WETLAND_TYPE, SYSTEM_NAME, CLASS_NAME, WATER_REGIME_NAME

API

/api/land?op=wetlands

Related collections

National Hydrography Dataset streams and water bodies (picture) 0.62; National Wetlands Inventory raster (zoomed out) 0.60; Where the water goes: NLDI traces on the NHDPlus V2 network 0.59Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

National Wetlands Inventory raster (zoomed out)

nwi-wetlands-raster · layer wetlands · show on the globe

What it holds

NWI's national generalised raster, about 100 m pixels in eight wetland types, drawn above 10 km where the polygons are not loaded. The legend is the raster's own attribute table.

A record is

A pixel class at about 100 m: a picture of where NWI maps wetland types at that scale.

Sources

National Wetlands Inventory, U.S. Fish and Wildlife Service. Licence and credit: public domain

Coverage

United States and territories; above 10 km camera height.

Freshness

Read from the service named Wetlands_Raster, which carries no version: USFWS can change it in place, and the weekly endpoint sweep checks only that it answers. It replaced the retired Wetlands_Raster_v5 (Service not found) with the same eight classes and colours in its rasterAttributeTable (2026-10-18). Read from service name Wetlands_Raster; rasterAttributeTable, 2026-10-18.

Left out, and why

small wetlands: lost at 100 m pixels
a click answer on the raster: raster identify is deferred; below 10 km the polygons answer

Caveats

Above 10 km the map is NWI's national ~100 m generalised raster, eight classes. Small wetlands can be missing at that scale.

Sensitivity

A picture of habitat classes; no personal data.

Searchable fields

none: the layer answers by place, not by name

Embedded as record text

nothing: no record text

API

/api/terrain?op=products

Related collections

National Wetlands Inventory polygons 0.60; National Hydrography Dataset streams and water bodies (picture) 0.54; National Heritage Areas 0.48Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

National Register of Historic Places, listed properties

nrhp-listed · layer historic · show on the globe

What it holds

Properties listed on the National Register as NPS publishes them in its public, non-restricted spatial data: points for every listing it maps, and boundaries for districts and sites, each with its NRIS reference number, resource type, NHL flag, listing date and National Archives nomination link.

A record is

A listing on the National Register: a property NPS has listed, at the location NPS publishes. Absence is not absence on the ground.

Sources

National Register of Historic Places spatial data (public, non-restricted listings; nrhp_locations MapServer), National Park Service, Cultural Resources GIS. Licence and credit: US federal (public domain); general reference only, not Section 106 compliance; NPS publishes only non-sensitive, non-restricted listings

Coverage

United States; points below 25 km, boundaries below 10 km, 2,000 per layer and box. 73,928 listed points (layer 0), restricted addresses excluded; 22,855 listed boundaries on layer 1 (counted 2026-10-12).

Freshness

The service publishes no edit date; NPS describes it as its current public, non-restricted spatial data. Read from nrhp_locations MapServer ?f=json (no editingInfo; serviceDescription), 2026-10-13.

Left out, and why

removed listings (648) and listings with no status (60): the layer shows what is listed, as STATUS='Listed'
boundaries whose Address reads "Address Restricted" (3 listed): NPS marks their location restricted; filtered in the query, Address never requested
archaeological sites whose locations are withheld under NHPA s.304 / ARPA: NPS does not publish them in this service, and eye does not seek them elsewhere (no SHPO files, DINAA, tDAR or mirrors)
street addresses: never requested

Caveats

Only National Register properties that NPS publishes as non-restricted. Archaeological sites withheld under NHPA s.304 / ARPA are not in this source and are not shown. Absence on the map does not mean no resource is present. Not a substitute for Section 106 consultation or a SHPO file search.
Property names can include a person's name, as NPS lists them, and are not searchable.

Sensitivity

Agency-published, non-restricted listings only; restricted-address records are filtered out; property names (RESNAME) can be a person's and are never searched or embedded.

Searchable fields

NRIS_Refnum

Embedded as record text

ResType, Is_NHL, CertDate, City, County, State

API

/api/land?op=historic

Related collections

National Historic Landmarks (the 'Landmarks only' chip on Historic places) 0.82; National Heritage Areas 0.75; World Heritage Sites with an NPS boundary 0.69Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Flood hazard zones (FEMA National Flood Hazard Layer)

fema-nfhl-zones · layer flood · show on the globe

What it holds

FEMA's regulatory flood hazard zones (NFHL layer 28) in a box below 5 km: zone and subtype with FEMA's own meaning, whether the ground is in the special flood hazard area, the static base flood elevation where FEMA published one, depth and its units and datum, and the FIRM study it comes from (DFIRM_ID, study type). Under the zones, FEMA's own rendering of the same layer as a picture.

A record is

A flood hazard zone on FEMA's regulatory map (FIRM): the 1% annual-chance floodplain and others. Not a forecast, and the absence of a zone is not a finding of no flood risk.

Sources

National Flood Hazard Layer, Federal Emergency Management Agency. Licence and credit: public domain

Coverage

United States, where FEMA has digital flood maps; below 5 km in a ~9 km box; each zone's effective FIRM study (by DFIRM_ID).

Freshness

FEMA's map service publishes no edit date; each zone carries the FIRM study it comes from (DFIRM_ID). Not counted nationally: counts over the full extent are slow (a 0.02 degree San Antonio box answered 47 zones). Read from NFHL MapServer/28 ?f=json (no editingInfo, empty copyright and description), 2026-10-15.

Left out, and why

effective dates per panel: not requested; the zone's FIRM study id is shown instead
preliminary and pending maps: not in layer 28
a base flood elevation of -9999: FEMA's 'none' sentinel; dropped, never shown as a number

Caveats

A regulatory flood map (FIRM), not a forecast: it says nothing about whether anything is flooding now.
The absence of a zone is not a finding of no flood risk: FEMA has not mapped everywhere, and zone D means no FEMA analysis.

Sensitivity

Regulatory map polygons; no ownership or personal data.

Searchable fields

FLD_ZONE, FLD_AR_ID

Embedded as record text

FLD_ZONE, ZONE_SUBTY, SFHA_TF, STUDY_TYP

API

/api/land?op=flood

Related collections

Disaster declarations by county (FEMA OpenFEMA Disaster Declarations Summaries v2) 0.56; FEMA National Risk Index by county 0.53; Landslide susceptibility (USGS slope-relief threshold models, 2024) 0.50Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Protected areas and public lands (USGS PAD-US 4.1)

padus-units · layer publiclands · show on the globe

What it holds

USGS's Protected Areas Database (PAD-US 4.1, Manager Name service) in the four categories the route asks for, fee lands, easements, designations and proclamation boundaries: each unit's owner and owner type, local owner, manager and manager type, designation, GAP status, IUCN category, public access, acres and year established, and for easements the holder and holder type, decoded with the service's own domains.

A record is

A protected-area unit as PAD-US publishes it. Public access is PAD-US's code, not permission to enter; land inside a proclamation boundary is not all public; an easement's owner keeps the land.

Sources

Protected Areas Database of the United States (PAD-US) 4.1, U.S. Geological Survey Gap Analysis Project. Licence and credit: public domain

Coverage

United States; below 250 km (the 600 largest units when zoomed out past 60 km); PAD-US 4.1 (2025). 647,844 units in the Fee, Easement, Designation and Proclamation categories (counted 2026-10-15).

Freshness

USGS last edited the service's data on 2025-06-04; the service cites PAD-US 4.1 (doi:10.5066/P96WBCHS). Read from PADUS4_1ManagerName FeatureServer/0 ?f=json editingInfo.dataLastEditDate and copyrightText, 2026-10-15.

Left out, and why

the Marine and Other categories: not asked by the route
Loc_Own, Own_Name and EsmtHldr as search keys or record text: an owner or easement holder can be a private person; shown as PAD-US publishes them, never searched or embedded

Caveats

Public access is PAD-US's code for each unit; it is not permission to enter.
The land inside a proclamation boundary is not all public, and an easement's owner keeps the land (the holder has the rights the easement grants).

Sensitivity

Ownership is shown as PAD-US publishes it (a federal inventory); what stays out is the search from an owner or easement holder to its holdings. A privately owned unit (Own_Type PVT) is often named for a family, so it is labelled by PAD-US id and its published name is shown in the record but never searched or embedded. A unit on federal or state land (Own_Type FED or STAT) that is not an easement has its Unit_Nm embedded as the record text's name segment (B0, 2026-10-18); local, district, joint and NGO units are not, since many are acquisitions named for a former owner.

Searchable fields

Unit_Nm, Loc_Nm

Embedded as record text

Category, Own_Type, Mang_Type, Des_Tp, State_Nm, Pub_Access, GAP_Sts, IUCN_Cat

API

/api/land?op=publiclands

Related collections

National Park System units, authorised boundaries 0.75; Trust land: BIA American Indian and Alaska Native Land Area Representation 0.60; National Register of Historic Places, listed properties 0.55Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

National Historic Landmarks (the 'Landmarks only' chip on Historic places)

nhl-listed · layer historic · show on the globe

What it holds

The National Register listings NPS flags as National Historic Landmarks (Is_NHL = 'X'), from the same nrhp_locations service and query the Historic places layer already makes, with the NHL clause ANDed on: points for every landmark NPS maps, and boundaries below 10 km.

A record is

A National Register listing that is also a National Historic Landmark, at the location NPS publishes. The same exclusions as every listing: listed only, no restricted address; absence is not absence on the ground.

Sources

Coverage

United States; points below 25 km, boundaries below 10 km. 1,196 listed NHL points (layer 0, Is_NHL='X' AND STATUS='Listed', none with a restricted address); 1,699 listed NHL boundaries on layer 1 after the restricted-address exclusion (1,700 before) (counted 2026-10-16).

Freshness

As nrhp-listed: the service publishes no edit date. Read from nrhp_locations MapServer ?f=json (no editingInfo), 2026-10-16.

Left out, and why

the one listed NHL boundary whose Address reads restricted (layer 1): the NHL clause is ANDed onto the restricted-address exclusion and never replaces it
every listing that is not an NHL, while the chip is on: landmarks only; switching the chip off brings them back

Caveats

Only what NPS publishes as non-restricted, as for every National Register listing; withheld archaeological sites are not in this source.

Sensitivity

Agency-published, non-restricted listings only; restricted-address records are filtered out; property names (RESNAME) can be a person's and are never searched or embedded.

Searchable fields

NRIS_Refnum

Embedded as record text

ResType, Is_NHL, CertDate, City, County, State

API

/api/land?op=historic

Related collections

National Register of Historic Places, listed properties 0.82; National Heritage Areas 0.72; World Heritage Sites with an NPS boundary 0.66Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

National Park System units, authorised boundaries

nps-units · layer heritage · show on the globe

What it holds

Every National Park System unit as NPS's Land Resources Division publishes it: the authorised boundary (layer 2) below 250 km, and the unit's centroid (layer 0) from farther away, with the unit code, name, type (national park, monument, historic site, battlefield, ...), state, NPS region and the date the boundary was last edited.

A record is

A park unit's designation and its authorised outline. Not all land inside is federally owned: ownership and access are Public lands (PAD-US).

Sources

NPS Land Resources Division Boundary and Tract Data Service: nps_boundary (layer 2) and nps_boundary_centroids (layer 0), National Park Service, Land Resources Division. Licence and credit: US federal (public domain); credit: National Park Service, Land Resources Division

Coverage

United States and territories; boundaries in a box of at most 2 degrees below 250 km, centroids nationally. 437 unit boundaries (layer 2); 437 centroids (layer 0) (counted 2026-10-16).

Freshness

NPS last edited the boundary layer on 2026-10-16. Read from FeatureServer layer 2 editingInfo.lastEditDate; DATE_EDIT per unit, 2026-10-16.

Left out, and why

tract-level land status (layer 1): ownership is PAD-US's (Public lands); this layer is designation
the Creator, Editor and metadata fields: bookkeeping, and an editor field can hold a person's name
fetching nps.gov: the unit page is an outbound link only; nps.gov reset TLS from the build box

Caveats

Outlines are NPS's authorised unit boundaries. Not all land inside is federally owned; for ownership and access, see Public lands.

Sensitivity

Agency boundaries of park units; no ownership or personal data.

Searchable fields

UNIT_CODE, UNIT_NAME

Embedded as record text

UNIT_TYPE, STATE, REGION

API

/api/land?op=heritage

Related collections

Protected areas and public lands (USGS PAD-US 4.1) 0.75; National Heritage Areas 0.71; National Register of Historic Places, listed properties 0.68Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

National Heritage Areas

nps-nha · layer heritage · show on the globe

What it holds

The 62 National Heritage Areas Congress has designated, as NPS's Heritage Partnerships Program maps them, each with NPS's paragraph about the area and a link to the area's own site; one national answer, simplified to 0.01 degrees.

A record is

A partnership region designated by Congress: not federal land and not a park unit. NPS's description is shown as NPS serves it.

Sources

National Heritage Area Boundaries (WASO_NHA_Map2026 FeatureServer), National Park Service, Heritage Partnerships Program; Cultural Resources GIS. Licence and credit: US federal (public domain); copyrightText "Cultural Resources GIS, National Park Service"

Coverage

United States; one national answer. 62 heritage area polygons (WASO_NHA_Map2026 layer 0) (counted 2026-10-16).

Freshness

The service says its geometry was last updated April 2025 and its metadata December 2025; last edited 2026-01-07. Read from FeatureServer layer 0 description and editingInfo.lastEditDate, 2026-10-16.

Left out, and why

embedding the descriptions: they are free text, which the record-text rules keep out; a reviewed exception for these 62 agency paragraphs is Cam's call (D-HE2)
EDIT_BY, ORIGINATOR and the survey bookkeeping fields: never requested; an editor field can hold a person's name
the older WASO_BND_NHA layer: 12 features, superseded by the 2026 map

Caveats

A heritage area is a partnership region, not federal land.

Sensitivity

Agency-published partnership regions; descriptions are NPS text.

Searchable fields

UNIT

Embedded as record text

nothing: no record text

API

/api/land?op=heritage

Related collections

National Register of Historic Places, listed properties 0.75; National Historic Landmarks (the 'Landmarks only' chip on Historic places) 0.72; National Park System units, authorised boundaries 0.71Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Battlefields of the Civil War, Revolutionary War and War of 1812 (ABPP survey areas)

abpp-battlefields · layer heritage · show on the globe

What it holds

The American Battlefield Protection Program's survey boundaries for the principal battlefields of three wars: study areas outlined and core areas filled, coloured by war, with ABPP's site code, battle name, state and acres; below 1,500 km, in a box of at most 8 degrees.

A record is

A survey area: where ABPP's survey placed a battle's tactical context (study area) or where most of the fighting occurred inside it (core area). Not a property line and not a National Register listing.

Sources

ABPP Battlefield Boundaries: Civil War, Revolutionary War and War of 1812 study and core areas (ABPP_Battlefield_Boundaries FeatureServer layers 0, 1, 5, 6, 8, 9), American Battlefield Protection Program, National Park Service; Cultural Resources GIS. Licence and credit: US federal; item licenseInfo (read 2026-10-16): NPS "shall not be held liable for improper or incorrect use"; "These data and related graphics are not legal documents"; credit: American Battlefield Protection Program, National Park Service, Department of the Interior (ABPP-NPS)

Coverage

United States, where ABPP surveyed. 342 Civil War study areas (layer 0); core areas 332; Revolutionary War 159 and 146; War of 1812 80 and 70 (counted 2026-10-16).

Freshness

ABPP last edited the service on 2025-09-09; the Civil War survey update was last revised in 2014. Read from FeatureServer editingInfo.lastEditDate; item description, 2026-10-16.

Left out, and why

Potential National Register boundaries and the Civil War resurvey layers (2-4, 7, 10): a third tier reads as 'listed'; the Register itself is Historic places
owner or person fields: none exist in the source

Caveats

Battlefield boundaries are survey boundaries, not property lines: most of the land inside is private and not open to visitors. Core areas are drawn as ABPP delineates them, inside the study area.
NPS: the draft Update to the Report on the Nation's Civil War Battlefields "defined 'study area' as the maximum delineation of a battlefield, providing the tactical context of the battle and including all places related or contributing to the battle event. The Battlefield Boundary is where troops deployed and maneuvered before, during, and after the engagement."

Sensitivity

Survey areas only; no owner or person fields exist in the source.

Searchable fields

SITE

Embedded as record text

WAR, STATE, BOUNDARY

API

/api/land?op=heritage

Related collections

National Park System units, authorised boundaries 0.58; Trust land: BIA American Indian and Alaska Native Land Area Representation 0.52; National Historic Landmarks (the 'Landmarks only' chip on Historic places) 0.47Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

World Heritage Sites with an NPS boundary

nps-whs · layer heritage · show on the globe

What it holds

The World Heritage Sites NPS publishes boundaries for (35 rows), one site per inscription: the nine Frank Lloyd Wright buildings of inscription 1496rev are one site with their components listed. The UNESCO page and the site's own page are linked as the service gives them.

A record is

An inscription on UNESCO's World Heritage List, at the boundary NPS publishes. A row with no name reads 'unnamed in source', never filled in; a row with no geometry is not drawn.

Sources

World Heritage Sites with NPS-published boundaries (World_Heritage_Sites FeatureServer), National Park Service (Office of International Affairs). Licence and credit: US federal (public domain); site pages and UNESCO links as the service gives them; credit: National Park Service

Coverage

United States; one national answer. 35 rows in World_Heritage_Sites layer 0 (27 inscriptions after grouping; one row has no name and no geometry, so 26 are drawn) (counted 2026-10-16).

Freshness

NPS last edited the service on 2024-12-10. Read from FeatureServer layer 0 editingInfo.lastEditDate, 2026-10-16.

Left out, and why

SiteOwner: organisations, but owners all the same: never requested, shown or searched
inscriptions without an NPS-published boundary: not in this service; absence here is not absence from the List

Caveats

World Heritage Sites with an NPS-published boundary, one site per inscription (a serial inscription's components are drawn together); the UNESCO page is linked as the service gives it.

Sensitivity

Inscribed sites; SiteOwner values are organisations and are not shown or searched.

Searchable fields

WHS_Ref

Embedded as record text

nothing: no record text

API

/api/land?op=heritage

Related collections

National Register of Historic Places, listed properties 0.69; National Historic Landmarks (the 'Landmarks only' chip on Historic places) 0.66; National Heritage Areas 0.64Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Trust land: BIA American Indian and Alaska Native Land Area Representation

bia-lar · layer tribal · show on the globe

What it holds

BIA's Land Area Representation: the outer extent of land held in trust or restricted fee for federally recognized tribes and individual Indians, compiled from BIA land title records, as one national answer of 335 land areas (reservations, rancherias, pueblos and communities), each with its BIA region, classification code as BIA codes it, and GIS acres. Parts under 0.7 km2 are dropped, except each land area's largest part, which is drawn even when smaller so no land area disappears.

A record is

A land area as BIA names it, held in trust, restricted fee or mixed ownership. For illustrative and reference purposes only: not legal, survey or jurisdictional. Not a Census tribal statistical area (those are on the Field layer).

Sources

American Indian and Alaska Native Land Area Representation (BIA_AIAN_National_LAR FeatureServer layer 0), Bureau of Indian Affairs, Office of Trust Services, Division of Land Titles and Records, Branch of Geospatial Support. Licence and credit: US federal work (public domain), BIA attribution requested; BIA: for illustrative and reference purposes only, not for legal, survey, engineering or navigation purposes. The agency's published contact is in the layer's copyright text and is not reproduced in records.

Coverage

11 of BIA's 12 regions; no Eastern Oklahoma region rows and no land in Oklahoma; Alaska has one land area (Annette Island); not every tribe's land is in the layer, so a blank area does not mean no tribal land; current as BIA publishes it; no edit date is published. 335 land areas in BIA_AIAN_National_LAR layer 0 (returnCountOnly); CLASSIFICATION '1' 333, '3' 2; 4,279 polygon parts at offset 0.01, 3,009 under 0.7 km2 (counted 2026-10-17).

Freshness

No edit date published; BIA's copyright text says the dataset is continuously being updated. Recounted 2026-10-17. Read from FeatureServer layer 0 ?f=json (no editingInfo); copyrightText, 2026-10-17.

Left out, and why

Census AIANNH tribal areas: already on fabric (kind=tribal, 'Tribal land'), a different thing: a Census tabulation geography
polygon parts under 0.7 km2: allotment-sized parts held for individual Indians, dropped so they are not drawn beside parcels (Cam, 2026-10-17); each land area's largest part is kept and drawn even when it is under 0.7 km2 (82 such parts on the 2026-10-17 count), so no land area disappears; both counts are in the route's caveat
allotment-level or parcel data, owners, names or addresses from any source: none fetched, none drawn, nothing joined to the parcels layer
sacred sites or restricted cultural data: not in the source and never fetched from any other
BIA's detail level (offset 0.001): held for Cam; only the national 0.01 pull ships
the agency phone and email in the layer's copyright text: kept in provenance as the agency's published contact; never in a record, searched or embedded

Caveats

The Bureau of Indian Affairs' Land Area Representation: the outer extent of land held in trust or restricted fee for federally recognized tribes and individual Indians, compiled from BIA land title records. BIA prepares it for illustrative and reference purposes only, not for legal, survey or jurisdictional use. Land inside an outline can be in other ownership, and land outside can be trust land. BIA's layer has no Eastern Oklahoma region rows and no land in Oklahoma, and does not include every tribe's land, so a blank area does not mean there is no tribal land there. Alaska Native corporation lands are not trust land and are not shown; Census tribal areas are on the Field layer.
This GIS Dataset is prepared strictly for illustrative and reference purposes only and should not be used, and is not intended for legal, survey, engineering or navigation purposes.

Sensitivity

Exterior extents of tribal trust and restricted-fee land as BIA publishes them; allotment-sized parts under 0.7 km2 are dropped, except each land area's largest part, which is drawn even when smaller (82 on the 2026-10-17 count) so no land area disappears; no allotment, parcel, owner or sacred-site data. LARNAME is a land area's name (all 335 read: reservations, rancherias, pueblos, communities; none a person's) and is embedded on Cam's call (2026-10-17).

Searchable fields

LARID, LARNAME

Embedded as record text

LARNAME, REGION

API

/api/land?op=tribal

Related collections

National Park System units, authorised boundaries 0.65; Protected areas and public lands (USGS PAD-US 4.1) 0.60; National Register of Historic Places, listed properties 0.54Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Species

USFWS designated critical habitat, final and proposed

fws-crithab · layer crithab · show on the globe

What it holds

ESA critical habitat designations under the U.S. Fish and Wildlife Service, one polygon set per species and rule, final and proposed, drawn from FWS's own tile caches and answered at a click from its FeatureServer: species, listing status, Final or Proposed, Federal Register rule and publication date.

A record is

A regulatory designation under the Endangered Species Act. It is not species range or occurrence, and FWS says the GIS file is not the legal boundary (the Federal Register rule and 50 CFR 17 are).

Sources

Threatened and Endangered Species Critical Habitat (final and proposed; USFWS_Critical_Habitat FeatureServer and the usfws_critical_habitat_final / _proposed tile caches), U.S. Fish and Wildlife Service. Licence and credit: US federal (public domain); FWS: the GIS files are not the legal source for critical habitat boundaries (the Federal Register rule and 50 CFR 17 are); see fws.gov/disclaimer

Coverage

United States and territories, FWS-lead species. 803 final per-species designations (layer 0); 70 proposed (layer 2) (counted 2026-10-13).

Freshness

FWS last edited the final layer on 2026-08-31. Read from USFWS_Critical_Habitat FeatureServer layer 0 editingInfo.lastEditDate, 2026-10-13.

Left out, and why

occurrence points of any kind (GBIF, iNaturalist, NatureServe): precise locations of sensitive species are not shown; only agency regulatory polygons
species images: fws.gov/disclaimer: images can be copyrighted
unitname, effectdate, accuracy: FWS fills them with its placeholder "Please check current species specific shapefile"

Caveats

Many listed species have no designation, and a polygon does not mean the species is present. Proposed designations are always labelled as proposed. Marine and anadromous species are in the NOAA Fisheries card.

Sensitivity

Agency-published regulatory polygons only; no occurrence points.

Searchable fields

none: the layer answers by place, not by name

Embedded as record text

comname, sciname, listing_status, status, pubdate

API

/api/land?op=crithab

Related collections

NOAA Fisheries designated critical habitat (marine and anadromous species) 0.78; Listed species ranges (EPA OPP, USFWS and NOAA) 0.71; Superfund site boundaries (EPA) 0.50Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

NOAA Fisheries designated critical habitat (marine and anadromous species)

nmfs-crithab · layer crithab · show on the globe

What it holds

ESA critical habitat for species under NOAA Fisheries (whales, sea turtles, corals, salmon and steelhead, and others), drawn from NOAA's MapServer /export over FWS's caches and answered at a click: polygons containing the point (layer 226) and river and stream lines within about 200 m (layer 2), with listing status, Final or Proposed, unit, Federal Register rule, eCFR section and dates.

A record is

A regulatory designation under the Endangered Species Act, as NOAA maps it. Not range or occurrence; the eCFR text is the legal boundary.

Sources

All NMFS Critical Habitat (combined polygon layer 226 and line layer 2), NOAA National Marine Fisheries Service. Licence and credit: US federal (public domain); copyrightText 'National Marine Fisheries Service, Southeast Region' (layer 226) and 'National Marine Fisheries Service' (layer 2); the eCFR text, not the GIS, is the legal boundary

Coverage

US waters and the rivers and streams of anadromous species; rules published to 2025-07-15 (polygons); to 2017-08-17 (lines). 2,117 polygons in layer 226; 59,888 lines in layer 2 (counted 2026-10-13).

Freshness

The polygon layer is named 20230502 but holds rules to July 2025 (Indo-Pacific corals, Nassau grouper, proposed Rice's whale and six proposed green sea turtle DPSs); the line layer's latest PUBDATE is 8/17/2017. Read from layer names and PUBDATE maxima, 2026-10-13.

Left out, and why

NOAA's OPR folder: answers 499 Token Required
EPA's Critical_Habitat FeatureServer copy: it disagreed with NOAA at the Seattle point (no bocaccio)
occurrence and sighting data of any kind: only agency regulatory polygons and lines

Caveats

Marine and anadromous species come from NOAA Fisheries. NOAA's combined polygon layer is dated 2023-05-02 in its name but holds rules to July 2025; its line layer ends in 2017. The eCFR text is the legal boundary. A line row is one within about 200 m of the point, not one the point is on.

Sensitivity

Agency-published regulatory polygons and lines only; no occurrence points.

Searchable fields

none: the layer answers by place, not by name

Embedded as record text

COMNAME, SCIENAME, LISTSTATUS, CHSTATUS, PUBDATE, HABTYPE

API

/api/land?op=crithab

Related collections

USFWS designated critical habitat, final and proposed 0.78; Listed species ranges (EPA OPP, USFWS and NOAA) 0.76; Superfund site boundaries (EPA) 0.57Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Listed species ranges (EPA OPP, USFWS and NOAA)

opp-ranges · layer crithab · show on the globe

What it holds

ESA listed and proposed species ranges as USFWS and NOAA Fisheries delineate them, published by EPA's Office of Pesticide Programs for Section 7 work. A click with Critical habitat on lists the species whose range covers the point, grouped by status, with taxonomic group, lead agency, range date and ECOS page. The polygons are never drawn.

A record is

A range: where a species could occur. Not presence, not habitat, not a sighting.

Sources

Endangered Species Range Areas (Species_Ranges FeatureServer, diced layer 3), U.S. EPA Office of Pesticide Programs; data stewards USFWS Ecological Services and NOAA Fisheries. Licence and credit: US EPA Data Licensing Agreement for public use (the layer's licence line, kept as a credit)

Coverage

United States and its waters. 12,293 diced range polygons in layer 3 (counted 2026-10-13).

Freshness

EPA last edited the layer's data on 2026-10-11; each species carries its own range date (2014-2026 at the points checked). Read from Species_Ranges FeatureServer layer 3 editingInfo.dataLastEditDate; Range_Shapefile_Last_Updated, 2026-10-13.

Left out, and why

drawing the range polygons: the answer is a list, so it does not suggest where a species is
Office_Name and Address_City (FWS field offices): they add nothing to the question
occurrence points (GBIF, iNaturalist, NatureServe): precise locations of sensitive species are not shown

Caveats

A range is where a species could occur, as USFWS or NOAA delineate it; it is not presence. It includes proposed species. Range dates differ by species (2014-2026 at the points checked, 2026-10-13).

Sensitivity

Agency-published coarse ranges only, never drawn; no occurrence points.

Searchable fields

none: the layer answers by place, not by name

Embedded as record text

Common_Name, Scientific_Name, ESA_Listing_Status, Entity_Description, Lead_Agency, Taxonomic_Group

API

/api/land?op=crithab

Related collections

NOAA Fisheries designated critical habitat (marine and anadromous species) 0.76; USFWS designated critical habitat, final and proposed 0.71; Superfund site boundaries (EPA) 0.54Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Water

Impaired waters (EPA ATTAINS, 303(d))

attains-impaired · layer impaired · show on the globe

What it holds

Assessment units that states, territories and tribes report to EPA as impaired (isimpaired = 'Y'): Integrated Report category 4 and 5 stream reaches (lines) and lakes, estuaries and watersheds (areas), each with its category, 303(d) listing, TMDL, reporting cycle, overall status, EPA action status and How's My Waterway page.

A record is

A state's report that a whole assessment unit is impaired, for one reporting cycle. A regulatory listing, not a measured value at a point.

Sources

ATTAINS assessment units (ATTAINS_Assessment FeatureServer, lines and areas), U.S. Environmental Protection Agency, Office of Water, from state, territorial and tribal Integrated Reports. Licence and credit: US federal; EPA EGAC disclaimer (approved for use, no liability, not EPA's official position); copyrightText 'US Environment Protection Agency'

Coverage

United States, states, territories and tribes that report to ATTAINS; below 10 km, 2,000 per layer and box; reporting cycles 2022-2026 (they differ by state). 150,218 impaired line records (layer 1); 32,061 impaired area records (layer 2) (counted 2026-10-13).

Freshness

EPA last edited the line layer's data on 2026-09-02; 81,709 of the 150,218 impaired line records are a state's submittal EPA has not acted on yet. Read from ATTAINS_Assessment FeatureServer layer 1 editingInfo.dataLastEditDate; cyclestatus counts by statistics query, 2026-10-13.

Left out, and why

waters not reported impaired, including waters never assessed: the layer shows listings; absence is not a finding that a water is clean
the per-cause columns (mercury, PCBs, pathogens, nutrients, PFAS and about 35 more): deferred to a dossier follow-up
a zoomed-out picture above 10 km: EPA's gispub /export answered 504 on 3 of 5 calls when probed from this box (2026-10-13); recheck from Vercel

Caveats

Waters that states report to EPA as impaired, as they reported them. A category is a regulatory listing for a whole assessment unit, not a measured value at the point you click. Reporting cycles differ by state (2022-2026). About 54% of the impaired line records shown are a state's submittal that EPA has not acted on yet; the EPA action status says which. Waters not shown were not reported impaired: that includes waters never assessed, which is not the same as clean.

Sensitivity

Agency regulatory listings of water bodies; no personal data.

Searchable fields

assessmentunitidentifier

Embedded as record text

state, reportingcycle, ircategory, on303dlist, hastmdl, cyclestatus

API

/api/water?op=impaired

Related collections

Community water system service areas (EPA) 0.67; PFAS in public drinking water (EPA UCMR 5) 0.65; Toxics Release Inventory reported releases (EPA) 0.59Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

PFAS in public drinking water (EPA UCMR 5)

ucmr5-pfas · layer pfas · show on the globe

What it holds

UCMR 5 PFAS results per public water system, as EPA's PFAS Analytic Tools publish them: for each system serving a box and each of 29 PFAS, the highest single result at or above the minimum reporting level in ng/L, how many results reached it and how many results there were; or that the system was sampled with none at or above it, was not in UCMR 5, or is unknown this time. Systems are found by their service areas first (cws-service-areas), then by EPA's estimated location.

A record is

A water system's UCMR 5 results for one PFAS: samples taken where treated water enters the distribution system, 2023-2025. Not a tap, not current water quality, and not compared with any health limit.

Sources

PFAS Analytic Tools, UCMR results (PFAS_Analytic_Tools_Layers FeatureServer layer 1, UCMR 5 rows), U.S. EPA Office of Enforcement and Compliance Assurance (ECHO PFAS Analytic Tools); data from EPA's Office of Ground Water and Drinking Water, UCMR 5. Licence and credit: US federal public data; credit: U.S. EPA Office of Enforcement and Compliance Assurance

Coverage

United States and territories, public water systems in UCMR 5; below 25 km (service areas below 10 km); UCMR 5 sampling, 2023-2025. 1,978,735 UCMR 5 rows in PAT layer 1 (38,562 at or above the reporting level); 221,831 UCMR 3 rows are not used (counted 2026-10-15).

Freshness

EPA last edited the layer's data on 2026-10-13 (03:16 UTC). Read from PFAS_Analytic_Tools_Layers FeatureServer/1 ?f=json editingInfo.dataLastEditDate, 2026-10-15.

Left out, and why

Hazard Index rows: 53,408 rows with a null value and date; excluded by the request (Result_At_or_Above_UCMR_MRL IS NOT NULL)
Result_Above_HBSL and any health-limit comparison: the flag is never Yes on UCMR 5 rows, and no threshold has been read from a live EPA source
UCMR 3 rows (2013-2015): different analytes and reporting levels; deferred
facility and sample-point names and ids: never requested; a result is shown per system
an average: the value shown is the highest single result
a system placed at a guessed point: a system with no service area and no estimated location (3,011 UCMR 5 rows have no geolocation method) is listed as not drawn

Caveats

UCMR 5 results for public water systems, 2023-2025: samples taken where treated water enters the distribution system, not at any tap. A system is drawn on its service area as EPA maps it (state-sourced or modelled, as marked), or at an estimated location when EPA has no boundary for it. 'Detected' means at or above UCMR 5's minimum reporting level; the value shown is the highest single result in ng/L, not an average and not a comparison with any health limit. A system with no results here may not have been selected: of systems serving fewer than 3,300 people, EPA sampled a representative 800. Results are not a statement about current water quality.

Sensitivity

Per water system, never per tap or well; the non-community layer (which carries a parcel number) is not used. PWS_Name can be an apartment complex, a nursing home or a rental park: it is shown in the record and on the label only, never searched or embedded. The layer has no contact fields.

Searchable fields

F_PWS_ID

Embedded as record text

Contaminant, State, Result_At_or_Above_UCMR_MRL, UCMR_Geolocation_Method

API

/api/water?op=pfas

Related collections

Community water system service areas (EPA) 0.68; Impaired waters (EPA ATTAINS, 303(d)) 0.65; Superfund sites (EPA EMEF, SEMS) 0.58Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Community water system service areas (EPA)

cws-service-areas · layer pfas · show on the globe

What it holds

EPA's Community Water System service-area boundaries, the geography UCMR 5 results are drawn on below 10 km: each with its PWSID, population served and category, service-area type, whether the boundary comes from the system or state or is modelled by EPA, its original data provider, and the system's ECHO report.

A record is

The area a community water system serves, as its state or the system mapped it or as EPA modelled it. Not a pipe network, and not a statement that every address inside is served.

Sources

Community Water System Service Area Boundaries (Water_System_Boundaries FeatureServer), U.S. EPA; the original data provider of each polygon (a state agency, the system, or EPA's Office of Research and Development for modelled boundaries). Licence and credit: MIT License, as the ArcGIS item (80c6912ef14f46e480f5afd807767b4b) licenseInfo states it, read 2026-10-18; copyrightText 'US EPA'. The notice, verbatim: MIT License Copyright (c) 2025 US EPA Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Coverage

United States and territories, community water systems; below 10 km, 2,000 per box, simplified on the server. 44,615 community water system polygons: 25,375 sourced from the system or state, 19,240 modelled by EPA (counted 2026-10-15).

Freshness

EPA last edited the layer's data on 2026-05-29. Read from Water_System_Boundaries FeatureServer/0 ?f=json editingInfo.dataLastEditDate, 2026-10-15.

Left out, and why

transient and non-transient non-community systems' polygons: that layer carries a parcel number; deferred, and PAT's estimated location places those systems
Model_Method, Verification_Status and Confirmed: answered empty when probed; not requested

Caveats

Service areas are EPA's Community Water System boundaries: 25,375 are sourced from the system or state and 19,240 are modelled by EPA (2026-10-15); a modelled boundary is marked and drawn dashed.

Sensitivity

Utility service areas; no contact fields. PWS_Name is not searched or embedded.

Searchable fields

PWSID

Embedded as record text

Pop_Cat_5, Service_Area_Type, Symbology_Field, Primacy_Agency, Data_Provider_Type

API

/api/water?op=pfas

Related collections

PFAS in public drinking water (EPA UCMR 5) 0.68; Impaired waters (EPA ATTAINS, 303(d)) 0.67; Superfund site boundaries (EPA) 0.65Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

National Hydrography Dataset streams and water bodies (picture)

nhd-hydro-picture · layer hydro · show on the globe

What it holds

The USGS National Hydrography Dataset as The National Map draws it: streams, rivers, canals and water bodies as transparent line art in USGS's cached tiles, loaded by the browser from basemap.nationalmap.gov at every height. While it is on, the Surface water layer stops drawing Natural Earth's generalised rivers.

A record is

A picture of NHD's mapped surface water at the tile's scale (high-resolution NHD at large scales, NHDPlus at small ones, as the service describes). A drawn line is a mapped channel, not a statement that water flows in it today.

Sources

National Hydrography Dataset, cartographic tile cache (USGSHydroCached MapServer), U.S. Geological Survey, The National Map. Licence and credit: US federal (public domain); credit as the service's copyrightText: "USGS The National Map: National Hydrography Dataset. Data refreshed September, 2026." (read 2026-10-16)

Coverage

United States; tiles cached to level 23 (drawn to 16 here).

Freshness

The service says its data was refreshed in September 2026. Read from USGSHydroCached MapServer ?f=json copyrightText ("Data refreshed September, 2026.v1.1"), 2026-10-16.

Left out, and why

a click answer on the picture itself: pixels carry no attributes; a click asks NLDI and EPA for the NHDPlus V2 reach the point drains to (nhdplus-v21-reach)
NHDPlus HR as the click answer: deferred (D-H1): its point query took 51 s and 18.6 s on two of five calls on 2026-10-16 and named a canal at the Boise River

Caveats

Streams drawn from the USGS National Hydrography Dataset. Trace and flow use the medium-resolution NHDPlus Version 2 network, so a small stream on the map may drain to a larger NHDPlus reach. Mean annual flow is a model estimate, not a measurement; measured flow is on the gauges.

Sensitivity

A picture of streams and water bodies; no personal data.

Searchable fields

none: the layer answers by place, not by name

Embedded as record text

nothing: no record text

API

/api/terrain?op=products

Related collections

Where the water goes: NLDI traces on the NHDPlus V2 network 0.73; The NHDPlus V2.1 reach a point drains to 0.67; National Wetlands Inventory polygons 0.62Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Where the water goes: NLDI traces on the NHDPlus V2 network

nldi-trace · layer hydro · show on the globe

What it holds

Three navigations from the NHDPlus V2 reach a point drains to, through the USGS Network Linked Data Index: the main stem downstream as flowlines (default 100 km), the drainage basin as one polygon (NLDI's simplified basin, at most 2,000 vertices), and the USGS stream gauges NLDI indexes upstream (default 150 km). Offered in the Rivers and streams ground dossier and on every USGS gauge of the Surface water layer; a gauge the Surface water layer already draws is drawn once.

A record is

A path, a basin or a list of gauges computed by NLDI on the medium-resolution NHDPlus V2 network for one reach and one distance; the count of gauges depends on the distance asked. A river path, not the HUC-12 watersheds it passes through (fabric's "Watersheds downstream").

Sources

Network Linked Data Index (NLDI) on the NHDPlus Version 2 network: comid position, navigation (main stem downstream, upstream with tributaries) and basins, U.S. Geological Survey, Water Mission Area. Licence and credit: US federal (public domain); credit: USGS Network Linked Data Index

USGS Water Data API, U.S. Geological Survey. Licence and credit: public domain

Coverage

Conterminous United States (the NHDPlus V2 network); a point outside it has no reach.

Freshness

The NHDPlus V2 network is static between releases; NLDI's gauge index follows USGS's site list. Read from NLDI linked-data root (sources list: comid, nwissite, ...); EPA V2.1 fdate per flowline, 2026-10-16.

Left out, and why

a stored path or basin per place: computed per click from NLDI and held 30 days; not enumerated, so not counted
the basin as county shares: a derived number, not a published one (D-H3)
a second gauge client: upstream gauges are NLDI's site list; live readings stay with the Surface water layer's USGS gauges

Caveats

Streams drawn from the USGS National Hydrography Dataset. Trace and flow use the medium-resolution NHDPlus Version 2 network, so a small stream on the map may drain to a larger NHDPlus reach. Mean annual flow is a model estimate, not a measurement; measured flow is on the gauges.
NLDI's upstream gauge list can include inactive sites; the count depends on the distance asked.

Sensitivity

Network paths and public stream-gauge sites only; no personal data.

Searchable fields

none: the layer answers by place, not by name

Embedded as record text

nothing: no record text

API

/api/water?op=trace

Related collections

The NHDPlus V2.1 reach a point drains to 0.83; National Hydrography Dataset streams and water bodies (picture) 0.73; National Wetlands Inventory polygons 0.59Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

The NHDPlus V2.1 reach a point drains to

nhdplus-v21-reach · layer hydro · show on the globe

What it holds

For a clicked point, the NHDPlus V2 catchment containing it (NLDI comid/position) and that reach's attributes from EPA's NHDPlus V2.1 Network Flowline layer: GNIS name, NHD feature type, Strahler stream order, total drainage area, and two modelled mean annual flows (EROM gauge-adjusted QE_MA and runoff-only QA_MA, cubic feet per second).

A record is

A flowline of the medium-resolution NHDPlus V2 network: where the point drains to, which near a small stream can be a larger river. Flows are model output, not measurements; -9999 and missing values read 'not published'.

Sources

NHDPlus Version 2.1 Network Flowline attributes (NHDPlus MapServer layer 2, watersgeo.epa.gov), U.S. EPA Office of Water, with USGS. Licence and credit: US federal (public domain); credit: US EPA Office of Water, NHDPlus Version 2.1

Coverage

Conterminous United States and the NHDPlus V2 vector processing units. 2,724,612 flowlines in EPA's NHDPlus V2.1 Network Flowline layer (returnCountOnly) (counted 2026-10-16).

Freshness

Each flowline carries its NHD feature date (Boise River reach 23398863: 2001-03-08); the V2.1 attributes are static. Read from fdate field, NHDPlus MapServer layer 2, 2026-10-16.

Left out, and why

uppercase field names: the layer answers 400 to them; lowercase names are asked
a sentinel as a value: -9999 drainage areas, stream order 0 or -9 and null flows read 'not published', never 0
NHDPlus HR attributes: deferred (D-H1): slow and erratic on 2026-10-16

Caveats

Streams drawn from the USGS National Hydrography Dataset. Trace and flow use the medium-resolution NHDPlus Version 2 network, so a small stream on the map may drain to a larger NHDPlus reach. Mean annual flow is a model estimate, not a measurement; measured flow is on the gauges.
Both flows are NHDPlus V2.1 model output in cubic feet per second: the EROM estimate adjusted to gauges (QE_MA) and the runoff-only estimate (QA_MA). They can differ several-fold; neither is a measurement.

Sensitivity

Agency hydrography; waterbody names are place names.

Searchable fields

gnis_name, reachcode

Embedded as record text

ftype

API

/api/water?op=reach

Related collections

Where the water goes: NLDI traces on the NHDPlus V2 network 0.83; National Hydrography Dataset streams and water bodies (picture) 0.67; Community water system service areas (EPA) 0.55Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Contaminants

Superfund sites (EPA EMEF, SEMS)

emef-sems · layer contaminants · show on the globe

What it holds

Superfund sites as EPA's Envirofacts EMEF map service publishes them (layer 0, program system SEMS), each with its EPA program id, FRS registry id and EPA facility page.

A record is

A Superfund program listing at EPA's facility coordinates. A regulatory status, not a finding of contamination at the point.

Sources

Envirofacts EMEF facility points (efpoints MapServer: Superfund/SEMS, TRI, RCRAInfo, ACRES), U.S. Environmental Protection Agency (Envirofacts / EMEF). Licence and credit: US federal public data (EPA Envirofacts); service copyright text "EPA OMS"; service terms page not read (EPA convention, not a confirmed licence)

Coverage

United States; below 25 km, 1,000 per program and box. 1,374 points in EMEF efpoints layer 0 (SEMS) (counted 2026-10-13).

Freshness

The map service publishes no edit date; EPA links its metadata record from the layer description. Read from EMEF efpoints MapServer layer 0 ?f=json (no editingInfo), 2026-10-13.

Left out, and why

location_address, postal_code and every contact field: never requested
site boundaries and NPL status: not in this layer: boundaries are drawn from EPA's boundary layer (npl-boundaries) and status and dates are joined from the NPL list by EPA ID (npl-status)
sites deleted from the NPL and Superfund sites never listed: not in this layer: its 1,374 points are exactly the current NPL (1,337 final, 37 proposed), joined by EPA ID on 2026-10-15

Caveats

EPA's Superfund (SEMS) site layer is the current National Priorities List: on 2026-10-15 every one of its 1,374 points was on EPA's NPL list by EPA ID, 1,337 final and 37 proposed. The 466 sites deleted from the NPL, and Superfund sites never listed, are not in it.

Sensitivity

Federal regulatory records; no address or contact field is requested. The facility name (primary_name) can be a person's or a street address and is never searched or embedded.

Searchable fields

registry_id, pgm_sys_id

Embedded as record text

pgm_sys_acrnm

API

/api/permits?op=contaminants

Related collections

Superfund site boundaries (EPA) 0.74; Toxics Release Inventory reporters (EPA EMEF) 0.69; Brownfields properties (EPA EMEF, ACRES) 0.65Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Toxics Release Inventory reporters (EPA EMEF)

emef-tri · layer contaminants · show on the globe

What it holds

Facilities that report to the Toxics Release Inventory, as EMEF layer 1 publishes them, with the TRI facility id (TRIFID), FRS registry id and EPA facility page.

A record is

A TRI reporter: a facility that filed with TRI. The listing alone says nothing about how much was released.

Sources

Coverage

United States; below 25 km, 1,000 per program and box. 64,501 points in EMEF efpoints layer 1 (TRI) (counted 2026-10-13).

Freshness

The map service publishes no edit date. Read from EMEF efpoints MapServer layer 1 ?f=json (no editingInfo), 2026-10-13.

Left out, and why

location_address and every contact field: never requested

Caveats

A listing is a regulatory status, not a finding of contamination at the point.

Sensitivity

Federal regulatory records; no address or contact field is requested; the facility name is never searched or embedded.

Searchable fields

registry_id, pgm_sys_id

Embedded as record text

pgm_sys_acrnm

API

/api/permits?op=contaminants

Related collections

Toxics Release Inventory reported releases (EPA) 0.71; Superfund sites (EPA EMEF, SEMS) 0.69; RCRA hazardous-waste handlers (EPA EMEF, RCRAInfo) 0.60Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

RCRA hazardous-waste handlers (EPA EMEF, RCRAInfo)

emef-rcra · layer contaminants · show on the globe

What it holds

Handlers of hazardous waste registered in RCRAInfo, as EMEF layer 4 publishes them.

A record is

A RCRA handler registration: a regulatory status, not a finding of contamination.

Sources

Coverage

United States; below 25 km, 1,000 per box (RCRA is dense: 3,001 in a 0.4 by 0.25 degree Houston box). 1,552,273 points in EMEF efpoints layer 4 (RCRAInfo) (counted 2026-10-13).

Freshness

The map service publishes no edit date. Read from EMEF efpoints MapServer layer 4 ?f=json (no editingInfo), 2026-10-13.

Left out, and why

location_address and every contact field: never requested
EMEF water (layer 2) and air (layer 3) facilities: the Environmental permits layer already draws them from ECHO
handlers past the first 1,000 in a box: the cap keeps the answer bounded; the coverage note says when it was hit

Caveats

A listing is a regulatory status, not a finding of contamination at the point.

Sensitivity

Federal regulatory records; no address or contact field is requested; the facility name is never searched or embedded.

Searchable fields

registry_id, pgm_sys_id

Embedded as record text

pgm_sys_acrnm

API

/api/permits?op=contaminants

Related collections

Superfund sites (EPA EMEF, SEMS) 0.61; Toxics Release Inventory reporters (EPA EMEF) 0.60; Toxics Release Inventory reported releases (EPA) 0.49Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Brownfields properties (EPA EMEF, ACRES)

emef-acres · layer contaminants · show on the globe

What it holds

Brownfields properties in EPA's ACRES system, as EMEF layer 5 publishes them.

A record is

A Brownfields property record: a property in an EPA grant program, not a finding of contamination.

Sources

Coverage

United States; below 25 km, 1,000 per box. 44,445 points in EMEF efpoints layer 5 (ACRES) (counted 2026-10-13).

Freshness

The map service publishes no edit date. Read from EMEF efpoints MapServer layer 5 ?f=json (no editingInfo), 2026-10-13.

Left out, and why

location_address and every contact field: never requested; a Brownfields property's EPA name is often its street address, so it is never searched

Caveats

A listing is a regulatory status, not a finding of contamination at the point.

Sensitivity

Federal regulatory records; no address or contact field is requested; the property name can be an address and is never searched or embedded.

Searchable fields

registry_id, pgm_sys_id

Embedded as record text

pgm_sys_acrnm

API

/api/permits?op=contaminants

Related collections

Superfund sites (EPA EMEF, SEMS) 0.65; Superfund site boundaries (EPA) 0.61; Toxics Release Inventory reported releases (EPA) 0.56Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Toxics Release Inventory reported releases (EPA)

tri-releases · layer contaminants · show on the globe

What it holds

What a TRI facility reported releasing on site in a reporting year, by chemical and medium (stack and fugitive air, surface water, underground injection, landfills, surface impoundments, land treatment), in its reported units and estimation basis, with the chemicals it certified on Form A. Asked from a TRI listing's record; never drawn.

A record is

One Form R release line: a facility's own report of a quantity to one medium for one chemical and year. Not exposure, not concentration, not a measurement by EPA.

Sources

Toxics Release Inventory, Form R release quantities and Form A certifications (Envirofacts tri_reporting_form + tri_release_qty + tri_chem_info), U.S. Environmental Protection Agency, TRI Program. Licence and credit: US federal public data (Envirofacts); code meanings from Envirofacts' TRI metadata pages, read 2026-10-13

Coverage

TRI reporting facilities in the United States; reporting years 1987 onward; the newest complete year is two behind the calendar (2024 answered, 2025 returned no rows on 2026-10-13).

Freshness

Envirofacts serves the TRI tables as filed; a year's data completes about two years later. Read from Envirofacts dmapservice TRI tables (2025 empty, 2024 populated, 2026-10-13), 2026-10-13.

Left out, and why

the certifying official's name, title and signature, and every public and technical contact name, phone, extension and email: sent by Envirofacts on every row; dropped by allowlist before anything is cached or returned
TRI's own facility coordinates: pref_longitude is stored unsigned; the map point stays EMEF's
totals across units: pounds and grams (dioxins) are never added
range reports and 'NA' lines in totals: a range is not a number, and NA is not zero
off-site transfers and waste management quantities: this collection is on-site releases (tri_release_qty) only

Caveats

What the facility reported to TRI for that year, in its reported units and estimation basis. Reported releases are not exposure or concentration. 'Not applicable' and range reports are shown as such and never counted in a total; totals are per medium and unit, so pounds and grams are never added. A Form A certifies a chemical without reporting quantities. Small facilities and some chemicals are not required to report.

Sensitivity

Federal regulatory records kept by allowlist; no person's name, phone or email leaves the parser.

Searchable fields

none: the layer answers by place, not by name

Embedded as record text

cas_chem_name, environmental_medium, unit_of_measure, reporting_year

API

/api/permits?op=tri

Related collections

Toxics Release Inventory reporters (EPA EMEF) 0.71; Leaking underground storage tank releases by county (EPA UST Finder) 0.63; Impaired waters (EPA ATTAINS, 303(d)) 0.59Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Superfund site boundaries (EPA)

npl-boundaries · layer contaminants · show on the globe

What it holds

Superfund site boundary polygons as EPA regions map them, harvested into EPA_GEO's national layer: site areas, ground boundaries, operable-unit outlines and extents of contamination, each with its EPA ID, NPL status decoded from the layer's own domain, federal-facility flag, EPA program as published, city, county, state and EPA site page, and the site's NPL list row joined by EPA ID (npl-status). Drawn below 10 km, coloured by NPL status.

A record is

One mapped feature of a Superfund site: a regulatory area as an EPA region drew it. Not a contamination footprint. A site can have several overlapping features; they are not merged, and each says how many its site has.

Sources

National NPL Superfund Site Boundaries (Public), FAC_Superfund_Site_Boundaries_EPA_Public FeatureServer, U.S. EPA, Office of Land and Emergency Management (EPA_GEO). Licence and credit: US EPA Data Licensing Agreement for public use (layer copyrightText, read in full 2026-10-15): "The Agency is providing this geospatial information as a public service and does not vouch for the accuracy, completeness, or currency of data. Data provided by external parties is not independently verified by EPA. This data is made available to the public strictly for informational purposes. Data does not represent EPA's official position, viewpoint, or opinion, express or implied. This information is not intended for use in establishing liability or calculating Cost Recovery Statutes of Limitations and cannot be relied upon to create any rights, substantive or procedural, enforceable by any party in litigation with the United States or third parties. EPA reserves the right to change these data at any time without public notice."

Coverage

United States; below 10 km, 2,000 per box, simplified on the server to 0.0005 degrees (about 50 m). 2,113 polygons in the Superfund program filter, for 1,907 EPA IDs (2,114 in the layer) (counted 2026-10-15).

Freshness

EPA last edited the layer's data on 2026-10-15; the service describes it as harvested weekly from regional datasets. Read from FAC_Superfund_Site_Boundaries_EPA_Public FeatureServer/0 ?f=json editingInfo.dataLastEditDate (2026-10-15 16:23 UTC) and the service description, 2026-10-15.

Left out, and why

EPA staff contact name, email and phone; street address, address comment and ZIP: never requested
SITE_FEATURE_NAME and SITE_FEATURE_DESCRIPTION: free text, not reviewed
GIS_AREA and its units: the units mix acres, square miles, 'Miles' and null
one 'Michigan LUST' polygon: a state tank-program record in a Superfund layer; the program filter (EPA_PROGRAM null or Superfund*) leaves it out
a decoded EPA_PROGRAM: its live values (Superfund, Superfund Remedial, Superfund Removal, Superfund Site Assessment) do not match the field's own code list, so it is passed through as published

Caveats

Superfund site boundaries as EPA regions map them, with NPL status. A boundary is a regulatory area, not a contamination footprint. A site can have several overlapping mapped features (site area, ground boundary, extent of contamination). EPA does not vouch for the accuracy, completeness or currency of these boundaries, and they are not for liability or cost-recovery decisions. NPL status and dates are from EPA's NPL list. A site missing from that list is shown as such, not as 'not NPL'.

Sensitivity

Agency regulatory records. EPA staff contacts and the street address are never requested. SITE_NAME is shown in the record and on the label, never searched or embedded: a site can be named for a person or a business.

Searchable fields

EPA_ID

Embedded as record text

NPL_STATUS_CODE, FEDERAL_FACILITY_DETER_CODE, EPA_PROGRAM, SITE_FEATURE_TYPE, CITY_NAME, COUNTY, STATE_CODE

API

/api/permits?op=contaminants

Related collections

Superfund sites (EPA EMEF, SEMS) 0.74; National Priorities List status (EPA) 0.71; Community water system service areas (EPA) 0.65Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

National Priorities List status (EPA)

npl-status · layer contaminants · show on the globe

What it holds

EPA's NPL list: each site's status (final, proposed, deleted), Hazard Ranking System score, proposal, listing, construction-completion, intent-to-delete and deletion dates, partial deletion, and its Federal Register notices and site progress profile. Joined by EPA ID to every Superfund point and boundary in a box, never by the box itself (a site's point can sit outside a box its polygon enters). Read in the record; not drawn on its own.

A record is

A site's row on the NPL list. A site with no row is shown as not on EPA's NPL status list, which is not a finding that it is clean; a list that did not answer is shown as unknown.

Sources

Superfund National Priorities List (NPL) Sites with Status Information, U.S. Environmental Protection Agency. Licence and credit: US federal public data; item Security Classification: Public

Coverage

United States; listings from 1983. 1,840 sites: 1,337 final, 37 proposed, 466 deleted (counted 2026-10-15).

Freshness

EPA last edited the list's data on 2026-09-30. Read from Superfund NPL Sites with Status Information FeatureServer/0 ?f=json editingInfo.dataLastEditDate, 2026-10-15.

Left out, and why

the HTML of the notice and profile fields: only the anchor's text and an https epa.gov link are kept
City, County and the narrative fields: the boundary carries the place; the narratives are not reviewed

Caveats

Dates are as EPA's list gives them (MM/DD/YYYY, shown as ISO dates); a missing or unreadable date is left out, never guessed.

Sensitivity

Federal regulatory list; no personal data is requested. Site_Name is not searched or embedded.

Searchable fields

Site_EPA_ID

Embedded as record text

Status, Proposed_Date, Listing_Date, Construction_Completion_Date, Deletion_Date

API

/api/permits?op=contaminants

/api/catalog?op=place-export

Related collections

Superfund site boundaries (EPA) 0.71; National Register of Historic Places, listed properties 0.61; Superfund sites (EPA EMEF, SEMS) 0.55Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Formerly Used Defense Sites: properties (USACE FUDS)

fuds-properties · layer contaminants · show on the globe

What it holds

Properties the Department of Defense once used and USACE is responsible for investigating and cleaning up, as USACE's FY2024 FUDS inventory publishes them: every property as a point (layer 1) and, below 10 km, the property's boundary where USACE has drawn one (layer 4), each with its FUDS property number, status, eligibility, county, state, closest city, USACE district, data year, FUDS portal page and its cleanup projects (fuds-projects).

A record is

A property in the FUDS program. A program status, not a finding of contamination at the point; a property with no projects was found to need none or is not eligible.

Sources

Formerly Used Defense Sites (FUDS), FY24 property, project and munitions response site data (fuds FeatureServer, layers 1-4), U.S. Army Corps of Engineers. Licence and credit: US federal; item licenseInfo: no usage constraints (read 2026-10-15); USACE describes the data as preliminary, not comprehensive and not for legal use; credit: U.S. Army Corps of Engineers

Coverage

United States and its territories; points below 25 km, boundaries below 10 km, 1,000 per layer and box; FY2024 data. 10,123 property points (layer 1); 3,003 property boundaries (layer 4) (counted 2026-10-15).

Freshness

USACE updated the item on 2026-10-07; the layers carry FY2024 data (FISCALYEAR 2024 on every row checked) and USACE calls it preliminary. Read from item 3f8354667d5b4b1b8ad7a6e00c3cf3b1 (sharing/rest modified) and FeatureServer/1 and 4 ?f=json editingInfo.dataLastEditDate (2026-10-07), 2026-10-15.

Left out, and why

CURRENTOWNER: free text holding private individuals' names, P.O. boxes and phone numbers; never requested
FEATUREDESCRIPTION: unreviewed narratives; never requested
a property's point where its boundary is drawn: one property, one symbol (27 of 126 points in the NJ/NY test box)
a guessed fiscal year or status: FISCALYEAR is text on one layer and a number on the other and is read strictly; the two STATUS vocabularies map to one enum, and an unknown value is null

Caveats

USACE's Formerly Used Defense Sites inventory, FY2024 data: properties the Department of Defense once used and USACE is responsible for investigating and cleaning up. A property's listing is a program status, not a finding of contamination at the point. Only about 3,000 of 10,123 properties have a mapped boundary; the rest are points. A munitions response site marks where munitions may be present, not a confirmed hazard footprint. USACE describes the data as preliminary and not for legal use. Property names are as USACE lists them and can include a person's name; they are not searchable.

Sensitivity

Agency-published DoD cleanup records. The owner and narrative fields stay at the source. Property names (FEATURENAME, DODFORMERLYUSEDDEFENSESITENAME) can carry a private surname (13 names on layer 1 contain 'PROPERTY', e.g. 'BARTLETT PROPERTY (WURTSMITH AIR FORCE BASE)'); they are shown in the record and on the label, never searched or embedded.

Searchable fields

FUDSUNIQUEPROPERTYNUMBER

Embedded as record text

STATUS, ELIGIBILITY, CLOSESTCITY, COUNTY, STATE, USACEDISTRICT, FISCALYEAR

API

/api/permits?op=contaminants

/api/catalog?op=place-export

Related collections

Formerly Used Defense Sites: munitions response sites (USACE FUDS) 0.79; Formerly Used Defense Sites: cleanup projects (USACE FUDS) 0.74; Superfund sites (EPA EMEF, SEMS) 0.56Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Formerly Used Defense Sites: cleanup projects (USACE FUDS)

fuds-projects · layer contaminants · show on the globe

What it holds

The cleanup projects on each FUDS property (layer 2): project number, restoration category (military munitions, hazardous, toxic and radioactive waste, containerized, building demolition and debris removal, potentially responsible party), program (IRP or MMRP), status, remedy-in-place, response-complete and closeout dates, risk value, USACE's cost-to-complete estimate and any overlapping federal land. Listed in their property's record; not drawn.

A record is

One environmental restoration project on a FUDS property, as USACE tracks it. A date after today is USACE's planned date and is marked so.

Sources

Coverage

United States and its territories; the projects of the properties drawn in a box; FY2024 data. 5,584 projects (layer 2) (counted 2026-10-15).

Freshness

USACE updated the item on 2026-10-07; the layers carry FY2024 data (FISCALYEAR 2024 on every row checked) and USACE calls it preliminary. Read from item 3f8354667d5b4b1b8ad7a6e00c3cf3b1 (sharing/rest modified) and FeatureServer/2 ?f=json editingInfo.dataLastEditDate (2026-10-07), 2026-10-15.

Left out, and why

a cost to complete of 0 on a project that is not complete: not an estimate; shown as 'not given' (3 such projects nationally on 2026-10-15). A 0 on a completed project is kept
a project whose property is not drawn in the box: one site, one symbol: projects are listed under their property
the project's own point: not drawn; the property carries it

Caveats

Costs are USACE's estimates to complete, in dollars as published; dates after today are planned.

Sensitivity

Agency-published DoD cleanup records. ENVRESTORATIONPROJECTNAME and DODFORMERLYUSEDDEFENSESITENAME are shown in the record only, never searched or embedded.

Searchable fields

FUDSPROPERTYID, PROJECTNUMBER

Embedded as record text

ENVRESTORATIONCATEGORY, DERPPROGRAM, PROJECTSTATUS, RISKVALUEMRSPP, OVERLAPPINGFEDERALLANDAGENCY

API

/api/permits?op=contaminants

Related collections

Formerly Used Defense Sites: properties (USACE FUDS) 0.74; Formerly Used Defense Sites: munitions response sites (USACE FUDS) 0.72; Superfund sites (EPA EMEF, SEMS) 0.56Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Formerly Used Defense Sites: munitions response sites (USACE FUDS)

fuds-mrs · layer contaminants · show on the globe

What it holds

Munitions Response Site polygons on FUDS properties (layer 3), with property number, site id, project number, status, program and relative priority. Drawn below 10 km, dashed.

A record is

Where munitions may be present on a former defense property, as USACE maps it. Not a confirmed hazard footprint.

Sources

Coverage

United States and its territories; below 10 km, 1,000 per box; FY2024 data. 1,682 munitions response site polygons (layer 3); relative priority is null on 1,661 (counted 2026-10-15).

Freshness

USACE updated the item on 2026-10-07; the layers carry FY2024 data (FISCALYEAR 2024 on every row checked) and USACE calls it preliminary. Read from item 3f8354667d5b4b1b8ad7a6e00c3cf3b1 (sharing/rest modified) and FeatureServer/3 ?f=json editingInfo.dataLastEditDate (2026-10-07), 2026-10-15.

Left out, and why

FEATUREDESCRIPTION and OFFICIALSITENAME: never requested
a relative priority where USACE gives none: shown as 'not given', never ranked

Caveats

A munitions response site marks where munitions may be present, not a confirmed hazard footprint. USACE publishes these polygons; they are shown to help people avoid hazards, with that caveat.

Sensitivity

Agency-published DoD cleanup records; FEATURENAME is shown in the record and on the label only.

Searchable fields

PROJECTNUMBER, FUDSPROPERTYID

Embedded as record text

PROJECTSTATUS, DERPPROGRAM, RELATIVEPRIORITY

API

/api/permits?op=contaminants

Related collections

Formerly Used Defense Sites: properties (USACE FUDS) 0.79; Formerly Used Defense Sites: cleanup projects (USACE FUDS) 0.72; Superfund site boundaries (EPA) 0.56Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Leaking underground storage tank releases by county (EPA UST Finder)

ust-releases-county · layer ust · show on the globe

What it holds

EPA UST Finder's county counts of reported leaking underground storage tank (LUST) releases: total, open, and no further action, with the county's population as the layer gives it. Drawn as county outlines filled by open releases.

A record is

A county's count of reported tank leaks in state records as of 2018-2021. A release is a reported leak; 'no further action' is the state's closure, not proof that no contamination remains. A snapshot, not current status.

Sources

UST Finder (national composite of USTs, UST facilities and LUST releases), county layers (UST_Finder_Feature_Layer_2eb, layers 2 and 3), U.S. EPA Office of Underground Storage Tanks and Office of Research and Development; The Association of State and Territorial Solid Waste Management Officials. Licence and credit: US federal public data; credit as the service's copyrightText gives it: EPA's Office of Underground Storage Tanks, EPA's Office of Research and Development, The Association of State and Territorial Solid Waste Management Officials

Coverage

United States, by county; between 50 and 1,500 km, a 4 degree box; state records as of 2018-2021; tribal records to October 2025. 3,220 counties; their release counts sum to 530,062 of UST Finder's 530,798 release points (736 are in no county row) (counted 2026-10-15).

Freshness

The service describes itself as a national composite as of 2018-2021, tribal data to October 2025; EPA last edited this layer's data on 2026-02-13. Read from UST_Finder_Feature_Layer_2eb FeatureServer ?f=json serviceDescription, and layer 3 ?f=json editingInfo.dataLastEditDate, 2026-10-15.

Left out, and why

individual release and facility sites (UST Finder layers 0 and 1): many are private homes ('%RESIDENCE%' matches 4,615 releases) and no field marks a residential site reliably; county counts carry no location below the county
an 'unknown status' count by subtraction: 13,179 releases nationally have an unknown status, yet some counties sum exactly; the arithmetic is not trustworthy, so nothing is derived
a per-capita rate: not computed; the layer's population is shown as given

Caveats

EPA's national composite of state tank records as of 2018-2021 (tribal records to October 2025), counted by county: a snapshot, not current status. A release is a reported leak; 'no further action' means the state closed it, not that no contamination remains. Counts depend on how each state reported. Individual sites are not shown because many are private homes.
Open and no-further-action counts: the layer never publishes a 0 in either field (505 and 225 nulls), so a null there most likely means 0 for many counties; it is kept null, and missingness is not random. Open: all 120 Kentucky counties plus 385 counties in 34 other states, 359 of which have release count = no-further-action count (0 left open). No further action: all 78 of Puerto Rico's plus 147 elsewhere (Kentucky 119, Washington 21, 7 others). Kalawao, HI and 8 other counties publish 0 releases with null status counts. Evidence in meta.nullPattern.

Sensitivity

County aggregates only. No site, name or address is requested.

Searchable fields

County

Embedded as record text

County, State

API

/api/permits?op=ust

/api/catalog?op=place-export

Related collections

Underground storage tank facilities by county (EPA UST Finder) 0.77; Toxics Release Inventory reported releases (EPA) 0.63; Toxics Release Inventory reporters (EPA EMEF) 0.51Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Underground storage tank facilities by county (EPA UST Finder)

ust-facilities-county · layer ust · show on the globe

What it holds

EPA UST Finder's county counts of UST facilities and tanks (open, closed, temporarily out of service), joined to the release counties by county name and state; the county layers carry no FIPS code.

A record is

A county's count of tank facilities and tanks in state records as of 2018-2021. A snapshot, not current status.

Sources

Coverage

United States, by county; with the release counties; state records as of 2018-2021. 3,220 counties; their facility counts sum to 741,447 of UST Finder's 742,844 facility points (counted 2026-10-15).

Freshness

The service describes itself as a national composite as of 2018-2021, tribal data to October 2025; EPA last edited this layer's data on 2026-02-13. Read from UST_Finder_Feature_Layer_2eb FeatureServer ?f=json serviceDescription, and layer 2 ?f=json editingInfo.dataLastEditDate, 2026-10-15.

Left out, and why

a FIPS join: the layers have no FIPS code; the name join leaves a non-matching or non-unique county unjoined and says so (all 57 counties joined in the Gulf Coast test box, 2026-10-15)
facility names and addresses: never requested; county counts only

Caveats

EPA's national composite of state tank records as of 2018-2021 (tribal records to October 2025), counted by county: a snapshot, not current status. A release is a reported leak; 'no further action' means the state closed it, not that no contamination remains. Counts depend on how each state reported. Individual sites are not shown because many are private homes.
Tank counts: closed and temporarily-out-of-service tanks are null in 118 counties each (all 78 of Puerto Rico's, 38 California counties including Los Angeles, and Georgia's 13123 and 13125), open tanks in 80 (Puerto Rico and the same two Georgia counties). Kept null, never 0 (meta.nullCellNote).

Sensitivity

County aggregates only. The NAME field here is the county's name; it is not embedded under that field name (the deny list keeps any bare NAME out).

Searchable fields

none: the layer answers by place, not by name

Embedded as record text

State

API

/api/permits?op=ust

/api/catalog?op=place-export

Related collections

Leaking underground storage tank releases by county (EPA UST Finder) 0.77; Community water system service areas (EPA) 0.63; Superfund site boundaries (EPA) 0.58Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Hazards and risk

FEMA National Risk Index by county

nri-counties · layer riskindex · show on the globe

What it holds

FEMA's National Risk Index for every county (Connecticut in its planning regions): the composite risk score and rating, expected annual loss in dollars and its rating, social vulnerability and community resilience as published in the index, and a rating for each of 18 natural hazards (avalanche to winter weather). Joined by FIPS onto the county outlines the globe already draws (Census TIGERweb); NRI's own county geometry is not fetched.

A record is

A county's percentile among all US counties, in FEMA's rating words. 'Very High' is high compared with everywhere else, not likely; expected annual loss is a modelled long-run average, not a forecast. 'Not Applicable', 'Insufficient Data', 'No Rating' and 'Data Unavailable' are not a class and never read as low.

Sources

National Risk Index, December 2025 (1.20.0): National_Risk_Index_Counties and National_Risk_Index_Census_Tracts FeatureServers, Federal Emergency Management Agency. Licence and credit: US federal; FEMA Terms & Conditions (item licenseInfo, read 2026-10-16): for planning purposes only, not a substitute for localized risk assessment; required statement: "This product uses the Federal Emergency Management Agency's National Risk Index dataset API or downloadable datasets but is not endorsed by FEMA. The Federal Government or FEMA cannot vouch for the data or analyses derived from these data after the data have been retrieved from the Agency's website(s)." Credit: FEMA National Risk Index, December 2025

Coverage

United States and territories, by county (Connecticut planning regions 09110-09190); drawn below 2,500 km; NRI version December 2025. 3,232 county rows in National_Risk_Index_Counties (returnCountOnly) (counted 2026-10-16).

Freshness

NRI_VER 'December 2025' (1.20.0); the layer's data last edited 2025-12-16. Read from NRI_VER field; FeatureServer layer 0 editingInfo.dataLastEditDate; item description, 2026-10-16.

Left out, and why

NRI's own county polygons: the globe has one county outline source (TIGERweb, via /api/economy?op=areas); counties are joined by FIPS
the other 430 or so fields (per-hazard frequencies, exposures, loss ratios, percentiles): the dossier shows the composite, loss, social vulnerability, resilience and the 18 ratings; the rest are FEMA's inputs
CDC/ATSDR SVI: on its own layer (vulnerability), shown beside this score in the click answer, never merged with it: FEMA's social vulnerability comes from the Census Community Resilience Estimate, CDC's from ACS, and they can disagree

Caveats

A relative index: a county's score is its percentile among all US counties, and a tract's among all tracts. 'Very High' does not mean likely. It means high compared with everywhere else.
Expected annual loss is a modelled average over many years, in dollars of buildings, population equivalence and agriculture. It is not a forecast for this year.
County and tract scores are computed separately and can't be compared with each other.
Social vulnerability as published in the National Risk Index. FEMA: "Data for Social Vulnerability are provided by the U.S. Census Community Resilience Estimate"; it is not the CDC/ATSDR Social Vulnerability Index, and the two can disagree.
FEMA: "The National Risk Index data are meant for planning purposes only. This National Risk Index dataset was created for broad nationwide comparisons and is not a substitute for localized risk assessment analysis."

Sensitivity

Published county indices; no personal data.

Searchable fields

STCOFIPS, NRI_ID

Embedded as record text

COUNTY, STATEABBRV, RISK_RATNG, EAL_RATNG, SOVI_RATNG, RESL_RATNG, AVLN_RISKR, CFLD_RISKR, CWAV_RISKR, DRGT_RISKR, ERQK_RISKR, HAIL_RISKR, HWAV_RISKR, HRCN_RISKR, ISTM_RISKR, IFLD_RISKR, LNDS_RISKR, LTNG_RISKR, SWND_RISKR, TRND_RISKR, TSUN_RISKR, VLCN_RISKR, WFIR_RISKR, WNTW_RISKR

API

/api/hazards?op=nri

Related collections

CDC/ATSDR Social Vulnerability Index 2022 by county 0.74; Wildfire risk to communities, county ranks (USDA Forest Service, 2nd edition, 2024) 0.73; FEMA National Risk Index by census tract 0.65Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

FEMA National Risk Index by census tract

nri-tracts · layer riskindex · show on the globe

What it holds

The same National Risk Index fields for every census tract, drawn on NRI's own tract polygons below 60 km, paged by FEMA's 2,000-record limit.

A record is

A tract's percentile among all US tracts, in FEMA's rating words; computed separately from the county index and not comparable with it. A tract has no name: it is shown by its 11-digit FIPS with its county and state.

Sources

Coverage

United States and territories, by census tract; drawn below 60 km; NRI version December 2025. 85,154 tract rows in National_Risk_Index_Census_Tracts (returnCountOnly) (counted 2026-10-16).

Freshness

NRI_VER 'December 2025'; the layer's data last edited 2025-12-16. Read from NRI_VER field; FeatureServer layer 0 editingInfo.dataLastEditDate, 2026-10-16.

Left out, and why

tracts above 60 km camera height: 85,154 polygons; the county index is shown instead
a tract's score compared with a county's: FEMA computes them separately

Caveats

A relative index: a county's score is its percentile among all US counties, and a tract's among all tracts. 'Very High' does not mean likely. It means high compared with everywhere else.
Expected annual loss is a modelled average over many years, in dollars of buildings, population equivalence and agriculture. It is not a forecast for this year.
County and tract scores are computed separately and can't be compared with each other.
Social vulnerability as published in the National Risk Index. FEMA: "Data for Social Vulnerability are provided by the U.S. Census Community Resilience Estimate"; it is not the CDC/ATSDR Social Vulnerability Index, and the two can disagree.
FEMA: "The National Risk Index data are meant for planning purposes only. This National Risk Index dataset was created for broad nationwide comparisons and is not a substitute for localized risk assessment analysis."

Sensitivity

Published tract indices; no personal data.

Searchable fields

TRACTFIPS, STCOFIPS, NRI_ID

Embedded as record text

STATEABBRV, RISK_RATNG, EAL_RATNG, SOVI_RATNG, RESL_RATNG, AVLN_RISKR, CFLD_RISKR, CWAV_RISKR, DRGT_RISKR, ERQK_RISKR, HAIL_RISKR, HWAV_RISKR, HRCN_RISKR, ISTM_RISKR, IFLD_RISKR, LNDS_RISKR, LTNG_RISKR, SWND_RISKR, TRND_RISKR, TSUN_RISKR, VLCN_RISKR, WFIR_RISKR, WNTW_RISKR

API

/api/hazards?op=nri

Related collections

CDC/ATSDR Social Vulnerability Index 2022 by census tract 0.81; CDC PLACES community health estimates by census tract 0.73; FEMA National Risk Index by county 0.65Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Wildfire risk to communities, county ranks (USDA Forest Service, 2nd edition, 2024)

wrc-counties · layer wildfirerisk · show on the globe

What it holds

USDA Forest Service Wildfire Risk to Communities for every county: national and state percentile ranks for four measures named by the service's own aliases (RISK_NATIONAL_RANK, BP_NATIONAL_RANK for burn probability, WHP_NATIONAL_RANK for wildfire hazard potential, EA_HURISK_NATIONAL_RANK), the shares of buildings and housing units in direct and indirect exposure, and building and housing-unit counts, from one national pull joined by FIPS onto the county outlines.

A record is

A county's modelled percentile rank among all US counties (or the state's), as USFS publishes it. Relative, not a forecast, and not a local assessment.

Sources

Wildfire Risk to Communities, 2nd edition (2024): county and community summary data (county_boundaries_conus_ak_hi and place_boundaries_conus_ak_hi FeatureServers), USDA Forest Service, Fire Modeling Institute, Rocky Mountain Research Station; Pyrologix. Licence and credit: CC0 1.0 and US public domain (item licenseInfo, read 2026-10-17); credit: USDA Forest Service, Wildfire Risk to Communities

Coverage

The 50 states and DC by county (Connecticut in planning regions 09110-09190); no territories; drawn below 2,500 km; WRC 2nd edition (2024): LANDFIRE 2020 fuels, burn probability as of end-2020, intensity as of end-2022; 2023 TIGER boundaries. 3,144 county rows in WRC_County_Summary_Data (returnCountOnly); 1 (District of Columbia, 11001) with every rank blank (counted 2026-10-17).

Freshness

Layer data last edited 2025-06-05; county item modified 2026-01-16; boundaries made May 2025 from the 2023 TIGER/Line files. May 2025 is the boundaries' date, not the model's. Read from FeatureServer layer 0 editingInfo.dataLastEditDate; ArcGIS item; field `vintage`, 2026-10-17.

Left out, and why

risk-to-potential-structures, burn probability and hazard rasters as values: the services publish no scale for their pixel values; the Wildfire hazard layer draws WHP as a picture (D5-2)
WRC's own county polygons: the globe has one county outline source; counties are joined by FIPS
territories: not in WRC
a plain-English name for RISK_* or EA_HURISK_*: the service does not say which is wildfirerisk.org's 'Risk to Homes'; the aliases are shown

Caveats

USDA Forest Service Wildfire Risk to Communities (2nd edition, 2024): modelled ranks that compare each county or community with all others nationally or in its state. A high rank is relative, not a forecast, and does not replace a local assessment. Fuels as of 2020-2022; 2023 Census boundaries.
The measures are named by the service's own field aliases (RISK_NATIONAL_RANK, BP_NATIONAL_RANK, WHP_NATIONAL_RANK, EA_HURISK_NATIONAL_RANK). The service does not say which of RISK_* and EA_HURISK_* is wildfirerisk.org's 'Risk to Homes', so neither is given a plain-English name here.

Sensitivity

Published modelled ranks per county; no building-level or person data. CC0 1.0.

Searchable fields

GEOID

Embedded as record text

STUSPS

API

/api/hazards?op=wrc

Related collections

Wildfire risk to communities, community ranks (USDA Forest Service, 2nd edition, 2024) 0.75; FEMA National Risk Index by county 0.73; CDC/ATSDR Social Vulnerability Index 2022 by county 0.68Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Wildfire risk to communities, community ranks (USDA Forest Service, 2nd edition, 2024)

wrc-communities · layer wildfirerisk · show on the globe

What it holds

The same ranks for every community (Census place) on WRC's own place polygons, below 150 km in a box of at most 1 degree, at most 1,000 a box; communities USFS excludes are drawn unfilled with USFS's reason.

A record is

A Census place's modelled percentile rank among all US communities (or the state's), by its 7-digit GEOID. A community USFS excludes (no people, no buildings, or DC as a state) has no rank: not ranked, never low.

Sources

Coverage

The 50 states and DC by Census place; drawn below 150 km; WRC 2nd edition (2024); 2023 TIGER boundaries. 32,033 place rows in WRC_Community_Summary_Data (returnCountOnly); 484 with exclude = 1 (counted 2026-10-17).

Freshness

As wrc-counties: layer data last edited 2025-06-05. Read from FeatureServer layer 0 editingInfo.dataLastEditDate, 2026-10-17.

Left out, and why

communities above 150 km camera height: 32,033 polygons; the county ranks are shown instead
a community's rank compared with a county's: USFS ranks each level among its own
the Census place name in the record vectors: names are never embedded (the palette matches them letter by letter); the state, the RISK and BP national ranks as the layer's own bins (never the number), and USFS's exclude reason are

Caveats

USDA Forest Service Wildfire Risk to Communities (2nd edition, 2024): modelled ranks that compare each county or community with all others nationally or in its state. A high rank is relative, not a forecast, and does not replace a local assessment. Fuels as of 2020-2022; 2023 Census boundaries.

Sensitivity

Published modelled ranks per Census place; no building-level or person data. CC0 1.0.

Searchable fields

GEOID, NAMELSAD

Embedded as record text

STUSPS, RISK_NATIO, BP_NATIONA, exclude_re

API

/api/hazards?op=wrc

Related collections

Wildfire risk to communities, county ranks (USDA Forest Service, 2nd edition, 2024) 0.75; CDC/ATSDR Social Vulnerability Index 2022 by census tract 0.65; FEMA National Risk Index by county 0.63Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Disaster declarations by county (FEMA OpenFEMA Disaster Declarations Summaries v2)

fema-declarations · layer riskindex · show on the globe

What it holds

Every federal disaster declaration naming a county, as FEMA's OpenFEMA API publishes it: the disaster number, declaration type (major disaster DR, emergency EM, fire management FM), declaration and incident dates, incident type, title, the programs declared (IH, IA, PA, HM) and the designated area, counted per year by type for the county's hazard history; the list opens under a county click of the National Risk Index, Social vulnerability or Wildfire risk ranks layers.

A record is

A federal decision that a disaster or emergency affected the county enough for federal help: a record of declarations, not a measure of damage or of how often hazards happen. Tribal and statewide designations name no county and are not attached to one.

Sources

OpenFEMA Disaster Declarations Summaries v2, Federal Emergency Management Agency. Licence and credit: US federal work (public domain); FEMA's terms page answered 403 to a keyless request on 2026-10-17 and the dataset metadata carries no licence. Credit: FEMA OpenFEMA, Disaster Declarations Summaries v2. This product uses the FEMA OpenFEMA API, but is not endorsed by FEMA.

Coverage

The 50 states, DC and territories as FEMA codes them; Connecticut in its eight legacy counties, so its planning regions have none; Alaska as FEMA files its Regional Educational Attendance Areas under census-area codes; Oglala Lakota, SD and Kusilvak, AK are read under their current and their pre-2015 codes (113 and 270), where FEMA still files most of their rows, each disaster once; every legacy code FEMA still files under (lib/hazards/history.ts FEMA_LEGACY, checked against a full OpenFEMA sweep on 2026-10-19) is read for its successors: Valdez-Cordova 261 for Chugach and Copper River, Prince of Wales-Outer Ketchikan 201 for Prince of Wales-Hyder, Ketchikan Gateway and Wrangell, Wrangell-Petersburg 280 for Wrangell and Petersburg, Bedford city 515 for Bedford County, VA. A split code's row is placed by its own designated area (FEMA_SPLIT_AREAS, each mapping cited to Census geography): the whole former census area to every successor, a school district or REAA only to the successor it lies in, an area no cited source places to none (listed as missing); marked filedUnder, never divided; declarations since 1953. 70,567 rows in DisasterDeclarationsSummaries v2 ($inlinecount and the dataset's recordCount); 1,612 with county code 000 (tribal and statewide); 11 under Alaska's retired code 261 (Valdez-Cordova, read for Chugach and Copper River) (counted 2026-10-17).

Freshness

lastDataSetRefresh 2026-10-17T20:46:55Z (refreshed as declarations are made); lastRefresh 2025-09-25. Read from OpenFEMA v1 OpenFemaDataSets metadata for DisasterDeclarationsSummaries (lastDataSetRefresh, lastRefresh, recordCount), 2026-10-17.

Left out, and why

FEMA's Individual Assistance, Public Assistance and NFIP claims datasets: applicant-level or address-adjacent; a county's declarations answer the question without them
tribal and statewide designations (county code 000): they name no county, so they are counted in the answer's meta.notCounty and attached to none
Connecticut's planning regions: FEMA publishes the legacy counties; no crosswalk is invented
a zero for a year with no declaration: waits on Cam's zero rule (HISTORY_ZERO_FILL is off): the year is absent, not 0

Caveats

Disaster declarations are FEMA's record of federal declarations naming this county. A declaration is a federal decision, not a measure of damage. Drought is the U.S. Drought Monitor's weekly index (0-500) since 2000. Climate is NOAA's county series. Tornadoes are SPC's record since 1950, which undercounts weak tornadoes before the 1990s radar network, and tornadoes with no county recorded are left out. Landslide susceptibility is a USGS 2024 model of terrain, not a forecast. Connecticut is published by FEMA, NOAA and USGS in its old counties.
FEMA OpenFEMA, Disaster Declarations Summaries v2. This product uses the FEMA OpenFEMA API, but is not endorsed by FEMA.

Sensitivity

Federal declarations by county as FEMA publishes them; no applicant, household or assistance-recipient data. Tribal and statewide designations name no county and are not attached to one. The declaration title and designated area are shown verbatim on click and never embedded.

Searchable fields

disasterNumber

Embedded as record text

incidentType, declarationType

API

/api/hazards?op=history

Related collections

Wildfire risk to communities, county ranks (USDA Forest Service, 2nd edition, 2024) 0.63; FEMA National Risk Index by county 0.62; Storm events per county per year (NOAA NCEI Storm Events, county-coded rows, 1996 onward) 0.57Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Drought by county: U.S. Drought Monitor Drought Severity and Coverage Index (weekly)

usdm-county-dsci · layer riskindex · show on the globe

What it holds

The U.S. Drought Monitor's county statistics: the weekly Drought Severity and Coverage Index (0 to 500) since January 2000 and this week's share of the county's area in each category (none, D0 abnormally dry to D4 exceptional drought), from USDM's data services, in the county's hazard history.

A record is

A county's weekly drought index as USDM publishes it, a weighted sum of the area in each drought category: a published single number, not a forecast. The dossier also draws the highest week of each year, computed here and labelled so.

Sources

U.S. Drought Monitor, National Drought Mitigation Center, USDA, NOAA. Licence and credit: public with attribution

Coverage

The 50 states, DC and Puerto Rico by county; Connecticut in its planning regions (the only round-7 source that publishes them), so a legacy Connecticut code answers no weeks; weekly since 2000-01-04 (Tuesday map dates).

Freshness

Weekly: the map is released on Thursday for the Tuesday before. The newest week on 2026-10-17 was 2026-10-13. Read from GetDSCI and GetDroughtSeverityStatisticsByAreaPercent answers (mapDate), 2026-10-17.

Left out, and why

a collection count: USDM publishes none; 1,398 weeks per county is a series length, not a count of records
an annual statistic as a published value: the dossier's yearly bars are the highest week, computed and labelled; the tower's annual mean or max waits on Cam

Caveats

U.S. Drought Monitor, National Drought Mitigation Center at the University of Nebraska-Lincoln, USDA and NOAA.
Not a federal-only work: produced by the National Drought Mitigation Center at the University of Nebraska-Lincoln with USDA and NOAA; credited to all three. USDM's data-services page redirected a keyless request on 2026-10-17, so the credit is USDM's standard one.

Sensitivity

Published county drought index; no person data.

Searchable fields

fips

Embedded as record text

nothing: no record text

API

/api/hazards?op=history

Related collections

FEMA National Risk Index by county 0.63; County climate: NOAA Climate at a Glance annual temperature and precipitation 0.60; CDC/ATSDR Social Vulnerability Index 2022 by county 0.57Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

County climate: NOAA Climate at a Glance annual temperature and precipitation

noaa-cag-county · layer riskindex · show on the globe

What it holds

NOAA NCEI's Climate at a Glance county time series: the January-December average temperature (degrees F) and precipitation (inches) for each year since 1895, as published, in the county's hazard history, with the 1901-2000 mean computed from the same series as a labelled reference line.

A record is

A county's annual value from NOAA's gridded divisional record (nClimDiv), as published: a climate series, not a station reading and not a hazard.

Sources

Climate at a Glance: County Time Series (annual average temperature and precipitation), NOAA National Centers for Environmental Information. Licence and credit: US federal work (public domain); credit: NOAA National Centers for Environmental Information, Climate at a Glance: County Time Series

Coverage

The 50 states and DC by county, Connecticut only in its legacy counties (no planning regions), no Puerto Rico; annual since 1895 (Alaska 1925, Hawaii 1991).

Freshness

Monthly; a year's annual value appears after December. On 2026-10-17 the latest annual value was 2025. Read from data.json answers (the last key, e.g. 202512), 2026-10-17.

Left out, and why

Connecticut's planning regions and Puerto Rico: not published (CT-110 and PR-127 answered 404 on 2026-10-17)
a collection count: 131 annual values is a series length per county, not a count of records

Caveats

Climate is NOAA's county series. A 1901-2000 mean is computed only where all 100 years are published (not Alaska or Hawaii).

Sensitivity

Published county climate series; no person data.

Searchable fields

fips

Embedded as record text

nothing: no record text

API

/api/hazards?op=history

Related collections

Drought by county: U.S. Drought Monitor Drought Severity and Coverage Index (weekly) 0.60; Storm events per county per year (NOAA NCEI Storm Events, county-coded rows, 1996 onward) 0.59; FEMA National Risk Index by county 0.59Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Tornadoes per county per year (NOAA Storm Prediction Center, 1950-2025)

spc-tornado-county · layer riskindex · show on the globe

What it holds

SPC's record of tornadoes since 1950, baked into distinct tornadoes per county per year (lib/hazards/data/spc-tornado-county.json, scripts/bake-hazard-history.mjs) from SPC's all-tornadoes file: a tornado counts once in each county any of its rows names in f1 to f4, its state segments and its continuation rows past four counties included; the whole-track summary row of a multi-state tornado is skipped, because its state segments name the counties. In the county's hazard history.

A record is

How many tornadoes SPC records in the county in a year. Not a casualty record, not a damage measure, and not comparable across eras: weak tornadoes before the 1990s radar network are undercounted.

Sources

Storm Prediction Center tornado database, 1950-2025 all tornadoes (1950-2025_all_tornadoes.csv), NOAA National Weather Service, Storm Prediction Center. Licence and credit: US federal work (public domain); credit: NOAA/NWS Storm Prediction Center

Coverage

The 50 states, DC, Puerto Rico and the Virgin Islands as SPC codes them; 3,036 current counties have at least one; Connecticut's eight legacy counties and 16 codes that name no current county are listed, not placed; 1950-2025. 74,956 rows in 1950-2025_all_tornadoes.csv (Last-Modified 2026-04-24): 72,788 tornadoes in one state, 1,358 state segments of multi-state tornadoes, 140 continuation rows for tornadoes crossing more than four counties, and 670 whole-track summaries (skipped); 82,211 county-tornadoes placed in current counties; 11 tornado-state records name no county and are left out (counted 2026-10-17).

Freshness

Annual; SPC publishes the year's final file in spring. The file read was last modified 2026-04-24. Read from Last-Modified header of 1950-2025_all_tornadoes.csv (2026-10-17), 2026-10-17.

Left out, and why

injury, fatality and loss columns: not needed for a count; the counts are not a casualty record
NCEI Storm Events' Tornado rows: Storm Events files one row per county segment of a tornado; SPC's distinct tornadoes are the count here, and the ncei-storm-events-county card (round 8) leaves Tornado out
the whole-track summary rows of multi-state tornadoes (670): the tornado's state segments name its counties; counting both would count it twice
tornadoes with no county recorded (11 tornado-state records): counted in the bake's meta.noCounty, never placed
codes that name no current county: Connecticut's legacy counties and 16 others are listed by code in meta; only Census-documented renames (Dade to Miami-Dade, Shannon to Oglala Lakota) are mapped

Caveats

Tornadoes are SPC's record since 1950, which undercounts weak tornadoes before the 1990s radar network, and tornadoes with no county recorded are left out.

Sensitivity

Counts of tornadoes per county per year baked from SPC's public record; casualty columns are not carried.

Searchable fields

fips

Embedded as record text

nothing: no record text

API

/api/hazards?op=history

Related collections

Storm events per county per year (NOAA NCEI Storm Events, county-coded rows, 1996 onward) 0.75; Disaster declarations by county (FEMA OpenFEMA Disaster Declarations Summaries v2) 0.56; County climate: NOAA Climate at a Glance annual temperature and precipitation 0.54Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Landslide susceptibility (USGS slope-relief threshold models, 2024)

usgs-landslide-susc · layer landslidesusc · show on the globe

What it holds

USGS's 2024 picture of where terrain is steep and high enough for landslides, at 90 m for the 50 states and Puerto Rico, drawn from USGS's ArcGIS Online tile cache in its own colours (classes 1-5 as the tile service draws them), with each county's share of susceptible terrain and its count of inventoried landslides from the release's county table on a ground click and in the county's hazard history.

A record is

A modelled terrain surface: where slope and relief pass the model's threshold. Not a forecast, not a hazard rating and not a record of landslides. The county share is susceptible km2 over the county's land km2, as USGS publishes it.

Sources

Slope-Relief Threshold Landslide Susceptibility Models for the United States and Puerto Rico (2024 data release; county_analysis.csv and the US_Landslide_Susceptibility tile service), U.S. Geological Survey (Belair, Jones, Martinez, Mirus and Wood). Licence and credit: CC0 1.0 (ScienceBase rights, read 2026-10-17); credit: U.S. Geological Survey: Belair and others (2024), Slope-Relief Threshold Landslide Susceptibility Models for the United States and Puerto Rico, doi:10.5066/P13KAGU3; tiles from USGS's ArcGIS Online service

Coverage

The 50 states and Puerto Rico; county shares on 2010 counties, so Connecticut's planning regions and six counties whose boundaries changed since 2010 (Chugach, Copper River, Petersburg Borough, Hoonah-Angoon, Prince of Wales-Hyder, and Bedford County, VA) have none. 3,221 rows in county_analysis.csv (2010 counties); 3,207 joined to current FIPS (3,204 by exact name, 3 by Census renames); Hoonah-Angoon, Prince of Wales-Hyder and Bedford County, VA match by name but are left out, because Census lists their boundaries as changed in 2013; tiles cached at levels 1-12 (counted 2026-10-17).

Freshness

Data release published 2024-08-22; tiles last-modified 2024-09-06; static. Read from ScienceBase item 65ccea5bd34ef4b119cb3bac (dates, rights); tile Last-Modified headers, 2026-10-17.

Left out, and why

the 90 m GeoTIFFs as values: 588 MB and 565 MB zips; the picture is enough
landslides.csv (67 MB): the Landslides layer already draws USGS's inventory (version 3)
the class under the click: the tile service has no identify, and decoding tile colours into classes would be our own inversion
class names (low to high): neither the service, the item nor the FGDC record names the five classes or says which of the release's two models the tiles draw
Connecticut's planning regions and the six counties whose boundaries changed since 2010: the county table is on 2010 counties; no crosswalk is invented, and a 2010 row is never attached to a changed boundary

Caveats

USGS's 2024 slope-relief model of where terrain is steep and high enough for landslides, at 90 m, for the 50 states and Puerto Rico. Susceptible terrain is not a forecast, not a hazard rating and not a record of landslides. The tile service does not say which of the release's two models it draws. County shares are USGS's, on 2010 counties, so Connecticut's planning regions and six counties whose boundaries changed since 2010 (Chugach, Copper River, Petersburg Borough, Hoonah-Angoon and Prince of Wales-Hyder in Alaska, and Bedford County, Virginia) have none.

Sensitivity

A modelled terrain surface and county shares; no property, person or site data. CC0 1.0.

Searchable fields

none: the layer answers by place, not by name

Embedded as record text

nothing: no record text

API

/api/hazards?op=history

Related collections

FEMA National Risk Index by county 0.55; USGS historical topographic maps, 1879-2006 0.53; Wildfire risk to communities, county ranks (USDA Forest Service, 2nd edition, 2024) 0.50Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Storm events per county per year (NOAA NCEI Storm Events, county-coded rows, 1996 onward)

ncei-storm-events-county · layer riskindex · show on the globe

What it holds

NCEI's Storm Events Database, baked into event reports per county per year per event type (lib/hazards/data/storm-events/<state>.json, scripts/bake-storm-events.mjs) from the 31 yearly details files 1996-2026: hail, thunderstorm wind, flash floods and every other type NWS logs by county, with tornadoes left to SPC's record. In the county's hazard history, as three bar series (hail, thunderstorm wind, flash floods) and a line naming the other types.

A record is

How many reports of an event type NCEI files under the county in a year. A count of reports, not of storms: one thunderstorm makes many wind and hail reports, and reporting grew with population and spotter networks. Not a casualty or damage record.

Sources

Storm Events Database, details files 1996-2026 (StormEvents_details-ftp_v1.0_d<year>_c<compiled>.csv.gz), NOAA National Centers for Environmental Information. Licence and credit: US federal work (public domain); credit: NOAA National Centers for Environmental Information, Storm Events Database

Coverage

The 50 states except Alaska, DC and Puerto Rico, by county; 3,182 current counties have at least one report; Connecticut's eight legacy counties are kept under their own codes and its planning regions get none; Alaska gets none, because every county-coded Alaska row (596) is under NCEI's own areas; 1996 to 2026-06 (the newest compiled month of the c20260918 file). 1,035,369 county-coded, non-tornado rows in the 31 details files 1996-2026 (1,819,792 rows in all; 733,143 zone-coded and 43,765 county-coded Tornado rows are not attached); 1,032,287 attached to 3,190 county codes; 1,039 refused because the county code's name disagrees with CZ_NAME, 1,184 under codes that name no current county, 859 in American Samoa, Guam and the Virgin Islands (counted 2026-10-17).

Freshness

NCEI recompiles the files on its own schedule (c stamps 20260323 for most years, 20260519 for 2017, 20260625 for 2022, 20260728 for 2024, 20260819 for 2025, 20260918 for 2026); the bake is per round, not monthly. Read from the c stamps in NCEI's csvfiles directory listing; meta.json records each file's name and sha256, 2026-10-17.

Left out, and why

forecast-zone rows (CZ_TYPE Z, 733,143): zones cross county lines, and NWS publishes its zone-county table only for 2025 and 2026 (bp18mr25, bp16ap26); zone rows are tallied per year and type and never attached (D8-1)
1950-1995: NCEI's scope before 1996 is tornado, hail and thunderstorm wind only, so a 1995-to-1996 jump would be the database's, not the weather's
Tornado rows: SPC's tornado record (spc-tornado-county) owns tornadoes; Storm Events files a row per county segment, so two tornado counts would disagree
EPISODE_NARRATIVE and EVENT_NARRATIVE: free text; never read
INJURIES_*, DEATHS_* and DAMAGE_*: not needed for a count; the counts are not a casualty or loss record
the fatalities and locations files: not needed for a county count
rows whose county code and CZ_NAME disagree: Puerto Rico's multi-municipio areas (1996-2008), Oklahoma's 1996 and Georgia's 2025 rows filed under forecast-zone numbers: listed by year in meta.nameMismatch, never attached

Caveats

NOAA's Storm Events Database as NCEI compiles it, county-coded rows only, 1996 onward: hail, thunderstorm wind, flash floods and other events NWS logs by county. Winter, heat, drought and wind events NWS logs by forecast zone are not attached to counties, because zones cross county lines; a type's year is blank when 10 % or more of its rows that year nationally were zone-coded. Tornadoes are counted from SPC's record above. A count is of reports, which grew with population and spotter networks.
A type's year is drawn only where fewer than 10 % of its reports nationally were coded by forecast zone (meta.zoneRows / countyRows in the bake); otherwise it is blank, "zone-coded", never 0.

Sensitivity

Counts of county-coded event reports per year; no narrative, casualty, damage or source text is carried.

Searchable fields

fips

Embedded as record text

nothing: no record text

API

/api/hazards?op=history

Related collections

Tornadoes per county per year (NOAA Storm Prediction Center, 1950-2025) 0.75; County climate: NOAA Climate at a Glance annual temperature and precipitation 0.59; Disaster declarations by county (FEMA OpenFEMA Disaster Declarations Summaries v2) 0.57Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Health

CDC PLACES community health estimates by county

places-county · layer health · show on the globe

What it holds

CDC's PLACES model-based estimates for 40 measures (health outcomes, prevention, risk behaviours, health status, disability and health-related social needs) for every county, age-adjusted, one measure at a time from one whole-measure pull, joined by FIPS onto the shared county outlines and binned by national quintiles; a click lists every measure for the county by CDC category.

A record is

A model-based estimate of the share of adults with a condition or behaviour in a county, with its 95% confidence interval and data year (2023; 2022 for five measures). Not a count of diagnosed people and not for ranking small differences.

Sources

PLACES: Local Data for Better Health, 2025 release, county (swc5-untb) and census tract (cwsq-ngmh) data, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Division of Population Health. Licence and credit: Public Domain (data.cdc.gov dataset licence, read 2026-10-16); credit: Centers for Disease Control and Prevention, PLACES: Local Data for Better Health, 2025 release

Coverage

United States counties (Connecticut planning regions); drawn below 2,500 km; data year 2023 (2022 for colorectal screening, dental visits, mammography, sleep and tooth loss). 2,957 counties with an age-adjusted DIABETES value (2,958 rows with the national one; 229,298 county rows in all measures and value types) (counted 2026-10-16).

Freshness

The 2025 release; the county dataset's rows were last updated 2025-12-04. Read from data.cdc.gov swc5-untb rowsUpdatedAt and dataset name, 2026-10-16.

Left out, and why

a per-county or within_box query while panning: a cold per-location call took 13.8 s and a within_box 19.6 s on 2026-10-16; the county level uses one whole-measure pull held a week
state rows: PLACES has none (only US); the layer asks the viewer to descend to counties
ZCTA and place datasets: out of this round

Caveats

Model-based estimates from CDC's BRFSS survey, ACS and the Census population; they are not counts of diagnosed people and are not for ranking small differences. Tracts with very small populations are not published.

Sensitivity

Model-based estimates for counties and tracts; no individual records; small-population tracts are withheld by CDC.

Searchable fields

locationid, locationname

Embedded as record text

statedesc, category, measure, year, data_value_type

API

/api/civic?op=health

Related collections

CDC/ATSDR Social Vulnerability Index 2022 by county 0.77; CDC/ATSDR Social Vulnerability Index 2022 by census tract 0.61; FEMA National Risk Index by county 0.59Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

CDC PLACES community health estimates by census tract

places-tract · layer health · show on the globe

What it holds

The same 40 measures for every published census tract, crude prevalence as CDC publishes tracts, asked by countyfips for at most six counties in view below 100 km and drawn on TIGER tract outlines (fabric op=field), binned by quintiles of the tracts loaded.

A record is

A crude model-based estimate for a tract (by its 11-digit FIPS). Tracts with very small populations are not published, and a tract with no row is shown as not published, never 0.

Sources

Coverage

United States census tracts (Connecticut in planning-region tract codes); drawn below 100 km, six counties at a time; data year 2023 (2022 for five measures). 83,522 tracts in cwsq-ngmh (3,047,284 tract rows in all measures) (counted 2026-10-16).

Freshness

The 2025 release; CDC last updated the tract dataset 2025-12-12. Read from data.cdc.gov cwsq-ngmh (civic lane read; the release date per the county dataset), 2026-10-16.

Left out, and why

within_box tract queries: 19.6 s cold; tracts are asked by county
national tract quintiles: 3 million rows; the bins are quintiles of the tracts loaded, and the legend says so

Caveats

Model-based estimates from CDC's BRFSS survey, ACS and the Census population; they are not counts of diagnosed people and are not for ranking small differences. Tracts with very small populations are not published.

Sensitivity

Model-based estimates for counties and tracts; no individual records; small-population tracts are withheld by CDC.

Searchable fields

locationname

Embedded as record text

category, measure, year, data_value_type

API

/api/civic?op=health

Related collections

CDC/ATSDR Social Vulnerability Index 2022 by census tract 0.79; FEMA National Risk Index by census tract 0.73; LIHTC Qualified Census Tracts (HUD, titled 2026) 0.67Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

CDC/ATSDR Social Vulnerability Index 2022 by county

svi-counties · layer vulnerability · show on the globe

What it holds

CDC/ATSDR's Social Vulnerability Index 2022 for every county (Connecticut in its planning regions): the overall percentile, the four theme percentiles (socioeconomic status, household characteristics, racial and ethnic minority status, housing type and transportation) and the flag count, from one national pull joined by FIPS onto the county outlines the globe already draws; a click adds CDC's 16 ACS estimates with their margins and, in its own block, FEMA's NRI social vulnerability for the same county.

A record is

A county's percentile among all US counties, 0 least to 1 most vulnerable, from ACS 2018-2022 estimates. A relative measure, not a measure of risk by itself, and not FEMA's NRI social vulnerability (a different index from the Census Community Resilience Estimate).

Sources

CDC/ATSDR Social Vulnerability Index 2022, US counties and census tracts (CDC_ATSDR_Social_Vulnerability_Index_2022_USA FeatureServer layers 1 and 2), Centers for Disease Control and Prevention / Agency for Toxic Substances and Disease Registry, Geospatial Research, Analysis, and Services Program. Licence and credit: US federal (public domain); credit: Centers for Disease Control and Prevention/Agency for Toxic Substances and Disease Registry/Geospatial Research, Analysis, and Services Program. CDC/ATSDR Social Vulnerability Index 2022 Database US.

Coverage

United States (the states and DC), by county; Connecticut planning regions 09110-09190; drawn below 2,500 km; ACS 2018-2022 (SVI 2022). 3,144 county rows in SVI2022 US county (layer 1, returnCountOnly); 0 with RPL_THEMES < 0 (counted 2026-10-16).

Freshness

CDC last edited the county layer's data on 2024-12-18. The service description still says '2020 Social Vulnerability Index (SVI) USA' while every layer is named SVI2022; the layer names are trusted. Read from FeatureServer layer 1 editingInfo.dataLastEditDate; layer names; service description, 2026-10-16.

Left out, and why

SVI's own county polygons: the globe has one county outline source (TIGERweb, via /api/economy?op=areas); counties are joined by FIPS
the EP_ estimates on the globe's colour: click only: the map colours the published percentile ranks
Puerto Rico and the other territories: not in SVI's US database; they read 'not in this release'
a number made from both SVI and FEMA's NRI: they are shown side by side in their native forms, each titled with its publisher; nothing is converted, averaged or ranked across them

Caveats

Percentile rank within the United States (0 least, 1 most vulnerable) from ACS 2018-2022 estimates, by CDC/ATSDR. A relative measure, not a measure of risk by itself. Tracts with no population or unreliable estimates are not ranked. This is not FEMA's social vulnerability score in the National Risk Index, which comes from the Census Community Resilience Estimate; the two can disagree.
The service describes itself as the 2020 SVI; its layers are named SVI2022 and the values match the 2022 release (Travis County, TX overall 0.5702), so the layer names are trusted.

Sensitivity

Published county percentile ranks from ACS estimates; no person-level data.

Searchable fields

FIPS

Embedded as record text

COUNTY, STATE

API

/api/civic?op=svi

Related collections

CDC PLACES community health estimates by county 0.77; FEMA National Risk Index by county 0.74; CDC/ATSDR Social Vulnerability Index 2022 by census tract 0.70Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

CDC/ATSDR Social Vulnerability Index 2022 by census tract

svi-tracts · layer vulnerability · show on the globe

What it holds

The same rankings for every census tract, on SVI's own tract polygons below 60 km (the height FEMA's NRI tracts load at), in a box of at most 0.4 degrees, paged by CDC's 2,000-record limit.

A record is

A tract's percentile among all US tracts, by its 11-digit FIPS with its county and state. CDC does not rank tracts with no population or unreliable estimates (-999): shown as not ranked, never as least vulnerable.

Sources

Coverage

United States (the states and DC), by census tract; drawn below 60 km; ACS 2018-2022 (SVI 2022). 84,120 tract rows in SVI2022 US tract (layer 2, returnCountOnly); 778 tracts not ranked (RPL_THEMES < 0) (counted 2026-10-16).

Freshness

As svi-counties: SVI 2022, the layers named SVI2022 although the service description says 2020. Read from FeatureServer layer 2 name; layer 1 editingInfo, 2026-10-16.

Left out, and why

tracts above 60 km camera height: 84,120 polygons; the county ranking is shown instead
a tract's percentile compared with a county's: CDC ranks each level among its own

Caveats

Percentile rank within the United States (0 least, 1 most vulnerable) from ACS 2018-2022 estimates, by CDC/ATSDR. A relative measure, not a measure of risk by itself. Tracts with no population or unreliable estimates are not ranked. This is not FEMA's social vulnerability score in the National Risk Index, which comes from the Census Community Resilience Estimate; the two can disagree.

Sensitivity

Published tract percentile ranks; CDC does not rank tracts with no population or unreliable estimates.

Searchable fields

FIPS

Embedded as record text

STATE, COUNTY

API

/api/civic?op=svi

Related collections

FEMA National Risk Index by census tract 0.81; CDC PLACES community health estimates by census tract 0.79; CDC/ATSDR Social Vulnerability Index 2022 by county 0.70Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Hospitals (HIFLD, FEMA's copy)

hifld-hospitals · layer hospitals · show on the globe

What it holds

Every hospital in HIFLD's hospital layer as FEMA's Resilience Analysis and Planning Tool last copied it: open hospitals nationally from orbit with a short record (name, type, status, beds, trauma level, city, state, county FIPS), and below 300 km the full record: address, beds, trauma level, helipad, owner type, NAICS industry, the state licensing source with its date, and the date HIFLD checked the location against imagery.

A record is

A licensed hospital as its state licensing source published it on the date shown, which can be years old. Not a list of emergency departments, and not a measure of capacity today.

Sources

Hospitals (HIFLD), FEMA Resilience Analysis and Planning Tool copy: Hospitals_RAPT FeatureServer layer 6 (data last edited 2026-04-22), Homeland Infrastructure Foundation-Level Data (DHS Geospatial Management Office, Oak Ridge National Laboratory), hosted by FEMA (FEMA_HQ_Preparedness_NIC). Licence and credit: U.S. Government data (HIFLD Open, public); credit: Homeland Infrastructure Foundation-Level Data (DHS Geospatial Management Office, Oak Ridge National Laboratory), via FEMA RAPT

Coverage

United States and territories; national dots from orbit, the full record in a box of at most 6 degrees below 300 km; per-row SOURCEDATE 2012-06-05 to 2025-07-03. 8,514 hospitals in Hospitals_RAPT layer 6 (returnCountOnly); 8,140 STATUS='OPEN', 374 CLOSED; 481 with BEDS = -999 (439 of them open) (counted 2026-10-16).

Freshness

FEMA last edited the layer's data on 2026-04-22. HIFLD Open's own hospital service no longer exists, so this copy is not refreshed from it; each row carries the date its source published it. Read from FeatureServer layer 6 editingInfo.dataLastEditDate; SOURCEDATE min and max (outStatistics, 2026-10-16), 2026-10-16.

Left out, and why

TELEPHONE, TTL_STAFF, ALT_NAME and VAL_METHOD: never requested: a contact field, a staff count and bookkeeping
CMS Hospital General Information (star ratings): not part of this record: the cms-hospital-ratings card (round 8) joins CMS's rows to these hospitals by exact address and ZIP in the box tier's click; this card's embedded record text is unchanged by it
closed hospitals, by default: 374 rows are CLOSED; a 'Closed too' chip draws them grey

Caveats

HIFLD's hospital layer as FEMA last copied it (data edited April 2026). Each hospital's details are as its state licensing source published them on the date shown, which can be years old. HIFLD's public site has closed; this is the last public copy. Not a list of emergency departments.

Sensitivity

Institutions as their licensing agency published them; no staff, patient or contact data. A hospital is labelled by its published name, an institution's (as a public dam keeps its name), so the palette's letter match finds it by name; NAME as a field stays on the deny list, and the same published name is embedded once as the record text's name segment (B0, 2026-10-18), with its CITY. Whether institution names may be searched is flagged for review; the fallback labels by HIFLD id. ADDRESS, WEBSITE and SOURCE are shown on click and never searched or embedded; OWNER (a category: non-profit, proprietary, government) is shown on click only, because the deny list matches the field name.

Searchable fields

ID, COUNTYFIPS, CITY, STATE

Embedded as record text

TYPE, TRAUMA, HELIPAD, CITY, STATE

API

/api/infra?op=hospitals

Related collections

Hospital star ratings (CMS Care Compare), on the HIFLD hospital click 0.67; CDC PLACES community health estimates by county 0.53; Formerly Used Defense Sites: properties (USACE FUDS) 0.53Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Hospital star ratings (CMS Care Compare), on the HIFLD hospital click

cms-hospital-ratings · layer hospitals · show on the globe

What it holds

CMS's Hospital General Information (Care Compare): each hospital's overall star rating (1-5) or CMS's 'Not Available' with its footnote, with CMS's hospital type, ownership, emergency services and birthing-friendly designation, baked into a join file (lib/infra/data/cms-hospitals.json, scripts/bake-cms-hospitals.mjs) and shown in the HIFLD hospital's full record below 300 km.

A record is

CMS's overall rating as Care Compare published it, matched to a HIFLD hospital by street address and ZIP code (exactly, or once a PO box or drawer is set aside, ordinal street words are read as numbers and U S as US) because HIFLD carries no CMS certification number; a match whose HIFLD name or type marks a tenant of another hospital is refused. A rating summarizes measures a hospital reported; it is not a measure of emergency capacity, and 'not rated' is not 'no match'.

Sources

Hospital General Information (Care Compare, Provider Data Catalog dataset xubh-q36u), Centers for Medicare & Medicaid Services. Licence and credit: no licence stated (metastore license: null, read 2026-10-17); US federal work; credit: Centers for Medicare & Medicaid Services, Care Compare (Hospital General Information)

Coverage

The CMS hospitals that match one HIFLD hospital one to one by address and ZIP, in the 50 states, DC and the territories CMS lists. 5,419 rows in Hospital General Information (5,419 CCNs; datastore count): 4,264 attached to a HIFLD hospital one to one (4,256 open, 2,544 of them rated; 8 to HIFLD rows not open; 221 of the 4,264 by the second pass), 224 address keys shared by more than one hospital on either side refused, 9 one-to-one matches refused as a tenant and another facility, and 2,245 'Not Available' ratings nationally (counted 2026-10-17).

Freshness

CMS's metastore says the dataset was modified 2026-07-22 and states no update cadence; the bake records that date, and the click prints it. Read from metastore schemas/dataset/items/xubh-q36u modified and accrualPeriodicity (read 2026-10-17), 2026-10-17.

Left out, and why

telephone_number and countyparish: never requested: a contact field, and HIFLD's own record already places the hospital
facility_name: read in the bake only, to refuse a tenant match (tenantConflict); never stored or shown: HIFLD's own record labels the hospital
a match on house number and ZIP alone: it attaches more (Texas: 450 against 338) but 29 of Texas's 450 attach the wrong hospital (a long-term-care unit to the general hospital around it)
addresses shared by two or more hospitals on either side (224 keys): a tenant unit would show its host's rating; listed in the bake's meta.sharedAddress, never attached
the measure groups (mortality, safety, readmission, patient experience, timeliness counts): not needed for the rating line; a later round may add them
a county place tower: ratings cover 2,544 of 8,140 open hospitals and are skewed by type, so a county mean would mostly measure which hospitals CMS rates

Caveats

CMS's overall hospital star rating as Care Compare published it (dataset modified 2026-07-22), joined to HIFLD's hospital by street address and ZIP code (exactly, or once a PO box or drawer is set aside, ordinal street words are read as numbers and U S as US) because HIFLD carries no CMS certification number; about one CMS hospital in five could not be matched that way, hospitals sharing an address are never matched, and a match whose HIFLD name or type marks a tenant of another hospital is refused. CMS rates acute-care and critical-access hospitals that report enough measures; many are not rated. A rating summarizes reported measures; it is not a measure of a hospital's emergency capacity.

Sensitivity

Institutions' published quality ratings; no patient, staff or contact data. CMS's telephone column is never carried.

Searchable fields

facility_id

Embedded as record text

nothing: no record text

API

/api/infra?op=hospitals

Related collections

Hospitals (HIFLD, FEMA's copy) 0.67; CDC PLACES community health estimates by county 0.52; Wildfire risk to communities, community ranks (USDA Forest Service, 2nd edition, 2024) 0.43Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Education

Public schools (NCES EDGE, school year 2024-25)

nces-public-schools · layer schools · show on the globe

What it holds

Every public school in NCES's 2024-25 administrative data (the Common Core of Data directory and membership), as points below 25 km inside the civic box: name, district, grades, level, school type, charter and virtual status, locale, enrollment, free or reduced-price lunch, FTE teachers and the student-teacher ratio as NCES publishes them, with a link to NCES's own page for the school.

A record is

A public school as its state reported it to NCES for 2024-25; enrollment is the October count. A full virtual school's point is its administrative address, not where students are. A negative value is one of CCD's reporting codes, never 0.

Sources

Public School Administrative Data 2024-25 (EDGE_ADMINDATA_PUBLICSCH_2425 MapServer layer 1), U.S. Department of Education, National Center for Education Statistics, Education Demographic and Geographic Estimates (EDGE) program. Licence and credit: US federal (public domain); credit: U.S. Department of Education, National Center for Education Statistics, EDGE / Common Core of Data, school year 2024-25

Coverage

United States, as NCES lists public schools; points below 25 km, in a box of at most 0.2 degrees; school year 2024-25. 101,110 schools in EDGE_ADMINDATA_PUBLICSCH_2425 layer 1 (returnCountOnly); 1,865 report a negative TOTAL; 501 Inactive and 372 Future are not drawn (counted 2026-10-16).

Freshness

The service publishes no edit date (a MapServer); its name and layer say school year 2024-25. Read from EDGE_ADMINDATA_PUBLICSCH_2425 MapServer layer 1 ?f=json (no editingInfo); layer name, 2026-10-16.

Left out, and why

race and ethnicity, sex and grade-level enrollment cells (AM..WH, *ALM/*ALF, PK..AE, TOTMENROL, TOTFENROL): never requested: at a small school a cell can identify a child
PHONE and LSTREET2: never requested
private schools: a different NCES survey (EDGE private schools); deferred (D5-9)
school district outlines: fabric already draws school districts (kind=school)
Inactive and Future schools: not operating in 2024-25; counted in the layer note, not drawn

Caveats

NCES Common Core of Data for school year 2024-25, as states reported it. Enrollment is the October count. A virtual school's point is its office. Public schools only; private schools are not on this layer.

Sensitivity

Public institutions' directory data as NCES publishes it; school-level race and sex counts are deliberately not read. LEA_NAME is never embedded; SCH_NAME, the school's label, is embedded once as the record text's name segment with LCITY and NMCNTY (B0, 2026-10-18), and is searchable (a public school's name is an institution's, flagged for review; the fallback is NCESSCH only). LSTREET1 is shown on click and never searched or embedded. Free or reduced-price lunch is shown as NCES publishes it, with no added suppression threshold for very small schools (flagged for review).

Searchable fields

NCESSCH, SCH_NAME

Embedded as record text

SCHOOL_LEVEL, SCHOOL_TYPE_TEXT, CHARTER_TEXT, VIRTUAL, SY_STATUS_TEXT, ULOCALE, GSLO, GSHI, LCITY, NMCNTY, LSTATE

API

/api/civic?op=schools

Related collections

Wildfire risk to communities, county ranks (USDA Forest Service, 2nd edition, 2024) 0.54; National Register of Historic Places, listed properties 0.54; Wildfire risk to communities, community ranks (USDA Forest Service, 2nd edition, 2024) 0.53Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Imagery and maps

World Imagery Wayback releases

esri-wayback · layer wayback · show on the globe

What it holds

Every published release of Esri World Imagery since 2014-02-20, each its own tile set, picked by release date; the release index comes from Esri's waybackconfig.json.

A record is

A release: the basemap as Esri published it on that date. A release date is not a capture date.

Sources

World Imagery Wayback (every published release of Esri World Imagery since 2014-02-20), Esri. Licence and credit: Esri Master License Agreement / ArcGIS Terms of Use (item licenseInfo read 2026-10-12; the linked Terms of Use page not read); not for offline tile export; attribution: Esri, Vantor, Earthstar Geographics, and the GIS User Community

Coverage

Worldwide; releases 2014-02-20 to 2026-09-24. 197 dated releases in waybackconfig.json (counted 2026-10-12).

Freshness

The newest release in the index was 2026-09-24. Read from waybackconfig.json itemTitle (Wayback YYYY-MM-DD), 2026-10-12.

Left out, and why

tile proxying, caching or export: the item's licence: not intended to be used to export tiles for offline; tiles load from Esri in the browser
imagery before 2014 as a release: Wayback starts at 2014-02-20 (older captures inside a release are dated by the capture-date collection)

Caveats

Esri World Imagery as published on each release date, 2014-02-20 onward. A release date is not a capture date: pixels can be years older, and an area that did not change shows the same image across releases. This is not pre-2014 aerial history.

Sensitivity

Imagery as Esri publishes it; nothing is stored here.

Searchable fields

none: the layer answers by place, not by name

Embedded as record text

nothing: no record text

API

/api/terrain?op=wayback

Related collections

World Imagery Wayback capture dates 0.75; USDA NAIP aerial photographs by year, 2010 onward 0.51; USGS historical topographic maps, 1879-2006 0.50Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

World Imagery Wayback capture dates

wayback-capture · layer wayback · show on the globe

What it holds

Esri's per-release imagery metadata: for each Wayback release, polygons saying which capture each region of that release came from, in 14 resolution bands, with the capture date, resolution, accuracy, provider and Esri's collection name. A click on Wayback imagery reads it at the point for the release drawn and the zoom it is drawn at.

A record is

A capture: when and by whom the pixels in that release were taken, as Esri records it. A point can sit on more than one capture.

Sources

World Imagery Wayback imagery metadata (per-release capture date, resolution, accuracy and provider; World_Imagery_Metadata_<YYYY>_r<NN> MapServers), Esri. Licence and credit: as esri-wayback (Esri Master License Agreement, kept as a credit); credit as the metadata services publish it: Esri, DigitalGlobe, Earthstar Geograpics, CNES/Airbus DS, USDA FSA, USGS, Aerogrid, IGN, IGP, and the GIS User Community

Coverage

Worldwide, where Esri publishes metadata (some areas have none); captures back to 2002 in the 2014-02-20 release, e.g. 2010-10-26 at San Francisco. 197 releases whose index entry names a metadata service on metadata.maptiles.arcgis.com (counted 2026-10-13).

Freshness

A release's metadata is fixed when the release is published; new releases bring new services. Read from waybackconfig.json metadataLayerUrl, one MapServer per release, 2026-10-13.

Left out, and why

a single 'best' capture per point: rows tie on DrawOrder (two 2010 captures at 90 in the 2014 release at San Francisco), so every row is shown
a guessed date where Esri lists none: no metadata answers 'Esri lists no capture date here'
metadata URLs from the caller: the route reads them from Esri's index by release id

Caveats

Capture dates are Esri's own metadata for the release shown. A point can sit on more than one capture. Esri publishes no capture metadata for some areas.

Sensitivity

Imagery provenance only; no personal data.

Searchable fields

none: the layer answers by place, not by name

Embedded as record text

SRC_DATE, NICE_NAME

API

/api/terrain?op=waybackdate

Related collections

World Imagery Wayback releases 0.75; USDA NAIP aerial photographs by year, 2010 onward 0.52; USGS historical topographic maps, 1879-2006 0.48Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

USGS historical topographic maps, 1879-2006

usgs-htmc · layer topo · show on the globe

What it holds

Every scale and edition of the topographic map USGS published, scanned (the Historical Topographic Map Collection), drawn one edition year at a time from Esri's HTMC ImageServer; a click lists the sheets at the point with their USGS citation, and their GeoPDF and GeoTIFF from USGS TNM Access matched by scan id.

A record is

A sheet: one edition of one USGS quadrangle as published. The edition year (DateCurrent) is when USGS revised the map, not a survey date; parts of a sheet can be decades older.

Sources

Historical Topographic Map Collection (HTMC), via Esri's USGS_Historical_Topographic_Maps ImageServer and USGS TNM Access, U.S. Geological Survey; hosted by Esri. Licence and credit: USGS maps are public domain; Esri's hosting is under the Esri Master License Agreement (kept as a credit, not a blocker); credit: Historical Topographic Map Collection courtesy of the U.S. Geological Survey, Esri (the service's copyrightText)

Coverage

United States and territories; tiles draw from about level 11 (about 10 km or closer); editions 1879-2006 (DateCurrent minimum and maximum). 181,008 sheets in the ImageServer (Category=1) (counted 2026-10-13).

Freshness

A closed collection: USGS's historical series ends in 2006, and the service describes it as all scales and all editions published since the program began. Read from ImageServer ?f=json description; DateCurrent max statistic (2006), 2026-10-13.

Left out, and why

the HTMC's own GeoPDF links (Download_GeoPDF): their host, ims.er.usgs.gov, no longer resolves (2026-10-13); the GeoPDF and GeoTIFF come from TNM instead
a year the view centre has no sheet for, in the picker: a year with no sheet under a tile draws blank
tiles coarser than level 11: a sheet draws only finer than its MaxPS; a 20 km tile came back blank

Caveats

USGS topographic maps as published, 1879-2006, one edition at a time. A map shows what USGS surveyed or revised for that edition; the edition year is not a survey date, and parts of a sheet can be decades older. Sheets of different scales and editions can sit side by side. Sheet collars and stamps are part of the scan. Draws only when zoomed in.

Sensitivity

Published federal maps; no personal data. Sheet names are place names.

Searchable fields

none: the layer answers by place, not by name

Embedded as record text

Map_Name, Map_Scale, DateCurrent, Imprint_Year, Survey_Year

API

/api/terrain?op=topo

Related collections

Landslide susceptibility (USGS slope-relief threshold models, 2024) 0.53; World Imagery Wayback releases 0.50; CDC PLACES community health estimates by census tract 0.49Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

USDA NAIP aerial photographs by year, 2010 onward

naip-years · layer naip · show on the globe

What it holds

The National Agriculture Imagery Program's leaf-on aerial photographs (0.3 to 1 m), one year at a time, from Microsoft Planetary Computer: the years flown at the view centre from its STAC search, and each year's mosaic registered at run time (never a hard-coded id), loaded by the browser from Planetary Computer. A click gives the flight date, resolution, state and item id of the quarter-quad under the cursor.

A record is

A STAC item: one quarter-quad photograph of one NAIP year, with its acquisition date. A year's mosaic joins many flights; the date on click is the photograph under the cursor, not the whole view.

Sources

National Agriculture Imagery Program (NAIP), STAC collection naip and per-year mosaics on Microsoft Planetary Computer, USDA Farm Service Agency (producer and licensor); hosted by Microsoft Planetary Computer. Licence and credit: collection licence "proprietary", pointing to USDA FSA's imagery policy (not a blocker under Cam's licence rule; credit kept): USDA Farm Service Agency, National Agriculture Imagery Program (NAIP), via Microsoft Planetary Computer

Coverage

Conterminous United States; tiles from level 11 (about 20 km or closer); 2010 onward; 6 to 11 years per place (Des Moines 8, San Francisco 6, DC 11, Austin 6 in the imagery lane's probes; Des Moines re-probed 2026-10-16: 2011-2023).

Freshness

The newest item at Des Moines is 2023 (ia_m_4109328_nw_15_030_20230902_20240104); NAIP flies each state every two to three years. Read from STAC search at -93.62, 41.59 (datetime and naip:year of the newest item), 2026-10-16.

Left out, and why

years before 2010 and state historic aerials (Texas, North Carolina): no keyless national source before 2010; the state services are phase 2 (D-I1, D-I2)
tile proxying or caching here: tiles load from Planetary Computer in the browser; Vercel never carries the image bytes
a per-place count: the years differ by place; counted per point, not nationally

Caveats

Agricultural-season aerial photographs flown state by state, so the years available differ by place. Each year is a mosaic of flights; the date on click is the photo under the cursor. Not every state was flown every year.

Sensitivity

Public aerial imagery at 0.3-1 m; no personal data is read or stored.

Searchable fields

Embedded as record text

naip:year, naip:state, datetime

API

/api/terrain?op=naip

/api/terrain?op=naipmosaic

Related collections

World Imagery Wayback capture dates 0.52; World Imagery Wayback releases 0.51; USGS historical topographic maps, 1879-2006 0.43Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Infrastructure

Dams (USACE National Inventory of Dams)

nid-dams · layer dams · show on the globe

What it holds

Dams in USACE's National Inventory of Dams public service: NID id, hazard potential with its meaning, the regulator's condition assessment and its date, purposes, dam type, height, storage, surface and drainage area, year completed, river, county, state, owner type, emergency action plan, operational status and the record's update date.

A record is

A dam as NID lists it. Hazard potential is the damage a failure would cause downstream, not the dam's condition or how likely it is to fail.

Sources

National Inventory of Dams (public service), U.S. Army Corps of Engineers. Licence and credit: public domain

Coverage

United States. 92,628 dams in NID's public service (layer 0) (counted 2026-10-15).

Freshness

NID's public service publishes no edit date; each dam carries the date its record was updated (DATA_UPDATED). Read from National_Inventory_of_Dams_Public_Service FeatureServer/0 ?f=json (no editingInfo), 2026-10-15.

Left out, and why

owner and representative names (CONG_REPRESNTATIVE, OTHER_NAMES, FORMER_NAMES, DESIGNER_NAMES and the owner names): never requested
a condition where the regulator published none: left out, never guessed

Caveats

Hazard potential is the damage a failure would cause downstream (High: loss of life likely), not the dam's condition and not how likely it is to fail.
Condition assessments are the regulating agency's, with the date it assessed; many dams have none published.

Sensitivity

Public infrastructure records; no owner or representative name is requested. A privately owned dam can be named for its owner ('<Surname> Pond Dam'), so a dam keeps its published name as its label, and has NAME embedded as the record text's name segment (B0, 2026-10-18), only when its primary owner type is Federal, State, Local Government, Public Utility or Tribal Government and every entry of NID's OWNER_TYPES list is one of those too. Every other dam (Private anywhere in its owner types, a primary type contradicted by its list, or an owner type that is Not Listed, missing or any other value: a non-answer is unknown, not public) is labelled by NID id, and its published name is shown in the record but never searched or embedded.

Searchable fields

NIDID

Embedded as record text

PRIMARY_PURPOSE, PRIMARY_DAM_TYPE, HAZARD_POTENTIAL, CONDITION_ASSESSMENT, COUNTYSTATE, STATE, YEAR_COMPLETED

API

/api/infra?op=dams

Related collections

Disaster declarations by county (FEMA OpenFEMA Disaster Declarations Summaries v2) 0.54; Formerly Used Defense Sites: cleanup projects (USACE FUDS) 0.53; Impaired waters (EPA ATTAINS, 303(d)) 0.51Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Broadband availability by county (FCC Broadband Data Collection, December 2025, via Esri)

fcc-bdc-counties · layer broadband · show on the globe

What it holds

The FCC's Broadband Data Collection for December 2025 as Esri summarizes it, for every county: Broadband Serviceable Locations in total and by the FCC's served, underserved and unserved tiers, served by fiber, cable and licensed fixed wireless, the number of providers, and the unserved count 12 months earlier, from one national pull joined by FIPS onto the county outlines the globe already draws; a click adds every technology column.

A record is

A county's count of locations where providers report fixed broadband available, by speed tier and technology. Available is not subscribed, and a provider's claim is not a measured speed. A share on the map is a ratio of two published counts.

Sources

FCC Broadband Data Collection December 2025 (Latest), counties and H3 resolution-8 hexagons (FCC_Broadband_Data_Collection_December_2024_View FeatureServer layers 1 and 5), Federal Communications Commission (data); Esri Policy Maps (summary layer, owner esri_policymaps). Licence and credit: Esri Master License Agreement on the summary layer (item licenseInfo, read 2026-10-17); figures from the FCC's public Broadband Data Collection (accessInformation); credit: Federal Communications Commission, Broadband Data Collection (December 2025), as summarized by Esri

Coverage

United States and territories as BDC publishes them, by county; Connecticut only in its eight legacy counties, so its nine planning regions are not shown at county level; drawn below 2,500 km; BDC December 2025 filing. 3,234 county rows in FCC_Broadband_Data_Collection_December_2024_View layer 1 (returnCountOnly); 0 with TotalBSLs = 0 (counted 2026-10-17).

Freshness

County data last edited 2026-06-10; item modified 2026-06-29. The service is named FCC_Broadband_Data_Collection_December_2024_View; its item is titled 'FCC Broadband Data Collection December 2025 (Latest)' and its description says December 2025. The item title is trusted (Esri's own statement of the vintage; the county layer's data was last edited 2026-06-10). Read from FeatureServer layer 1 editingInfo.dataLastEditDate; ArcGIS item e1343efcefc344709057260ee57290a0 title, modified and licenseInfo, 2026-10-17.

Left out, and why

the provider-offering related tables (per-ISP rows): availability by tier and technology answers the question; per-provider offerings are not needed for it
BDC's own county polygons: the globe has one county outline source (TIGERweb, via /api/economy?op=areas); counties are joined by FIPS
census block and block group layers: BDC's H3 hexagons are drawn below 60 km instead (fcc-bdc-hex)
Connecticut's planning regions at county level: not published in that geography: BDC still uses the eight legacy counties, and no crosswalk is invented
the 6- and 12-month-previous columns as a series: they are Esri's columns on one vintage; a series waits for the FCC's own files (D6-1)

Caveats

Fixed broadband availability as providers reported it to the FCC for December 2025, counted at Broadband Serviceable Locations. Available is not subscribed, and a provider's claim is not a measured speed. Summarized by Esri from FCC data; Esri's layer is under the Esri Master License Agreement. Connecticut is published here in its old counties, so its planning regions are not shown.
The service is named FCC_Broadband_Data_Collection_December_2024_View; its item is titled 'FCC Broadband Data Collection December 2025 (Latest)' and its description says December 2025. The item title is trusted (Esri's own statement of the vintage; the county layer's data was last edited 2026-06-10).

Sensitivity

Location counts per county as the FCC aggregates them; no address-level or subscriber data. Esri's summary layer is under the Esri Master License Agreement; the figures are the FCC's, credited to both.

Searchable fields

GEOID

Embedded as record text

StateAbbr

API

/api/civic?op=broadband

Related collections

Broadband availability by H3 hexagon (FCC BDC, December 2025, via Esri) 0.59; Community water system service areas (EPA) 0.54; Underground storage tank facilities by county (EPA UST Finder) 0.51Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Broadband availability by H3 hexagon (FCC BDC, December 2025, via Esri)

fcc-bdc-hex · layer broadband · show on the globe

What it holds

The same counts for BDC's own H3 resolution-8 hexagons (about 0.7 km2 each), below 60 km in a box of at most 0.3 degrees, one page of 2,000.

A record is

An H3 hexagon's count of locations by tier and technology as the FCC publishes it, by its H3 id. No names, no addresses.

Sources

Coverage

United States and territories, by H3 resolution-8 hexagon; drawn below 60 km; BDC December 2025 filing. 3,835,270 hexagons in layer 5 (H3 Resolution 8, returnCountOnly); 0 with TotalBSLs = 0; 911,479 with exactly one location (critique re-probe) (counted 2026-10-17).

Freshness

Hexagon data last edited 2026-08-26; same item and vintage as fcc-bdc-counties. Read from FeatureServer layer 5 editingInfo.dataLastEditDate, 2026-10-17.

Left out, and why

hexagons above 60 km camera height: 3.8 million polygons; the county figures are shown instead
a suppression of small hexagons: Cam's call (2026-10-17): ship hexagon counts as published, including one-location hexagons; the FCC publishes availability per location on its own map

Caveats

Fixed broadband availability as providers reported it to the FCC for December 2025, counted at Broadband Serviceable Locations. Available is not subscribed, and a provider's claim is not a measured speed. Summarized by Esri from FCC data; Esri's layer is under the Esri Master License Agreement. Connecticut is published here in its old counties, so its planning regions are not shown.

Sensitivity

Location counts per H3 hexagon (about 0.7 km2); no address or subscriber data. A quarter of hexagons hold a single location, so a hexagon can describe one home's availability as the FCC publishes it; shipped as published on Cam's call (2026-10-17).

Searchable fields

GEOID

Embedded as record text

nothing: no record text

API

/api/civic?op=broadband

Related collections

Broadband availability by county (FCC Broadband Data Collection, December 2025, via Esri) 0.59; Wildfire risk to communities, community ranks (USDA Forest Service, 2nd edition, 2024) 0.50; CDC/ATSDR Social Vulnerability Index 2022 by census tract 0.44Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Housing

Housing Choice Vouchers by census tract (HUD)

hud-hcv-tracts · layer housing · show on the globe

What it holds

HUD's count of Housing Choice Voucher households in each 2020 census tract and HUD's share of the tract's renter-occupied units (2020 Census DHC table H4), through December 2025, on HUD's own tract polygons in a box around the view below 100 km, coloured by the share or the count on fixed breaks (0-2-5-10-20-40-100 % and 11-25-50-100-250-500 households, this build's choice).

A record is

Voucher households in a tract as HUD aggregates them. A null is HUD's suppression of tracts with 10 or fewer voucher holders (drawn grey, '10 or fewer, or none'), never 0 and never 'no vouchers'. A sum of published tracts would be a floor, so there is no county total.

Sources

HUD eGIS: Housing Choice Vouchers by Tract, Qualified Census Tracts 2026, Difficult Development Areas 2026 and LIHTC properties (FeatureServers on HUD's ArcGIS Online organisation), U.S. Department of Housing and Urban Development (with the U.S. Census Bureau). Licence and credit: US federal work; item licenseInfo is HUD's no-warranty disclaimer with no use restriction (read 2026-10-17); credit: U.S. Department of Housing and Urban Development, U.S. Census Bureau

Coverage

The 50 states, DC and Puerto Rico by 2020 census tract; drawn below 100 km in a box of at most 0.3 degrees; Connecticut is not published (all 881 of its tracts null); through December 2025 (last updated January 2026). 85,154 tracts in Housing_Choice_Vouchers_by_Tract layer 0 (returnCountOnly); 47,463 with HCV_PUBLIC null, of which 881 are every Connecticut tract; 0 zeros; published minimum 11 (counted 2026-10-17).

Freshness

Layer data last edited 2026-07-21; description: 'Date of Coverage: Up to 12/2025, Last Updated: 01/2026'. Read from FeatureServer layer 0 editingInfo.dataLastEditDate (1784657444163) and the layer description, 2026-10-17.

Left out, and why

EANAME (HUD's entitlement-area name): a place name not needed for the count; never fetched
LIHTC properties' contact block: the layer carries contact names and phone numbers (CONTACT, CO_TEL); never fetched. The properties themselves are the hud-lihtc card (round 8), from an allow-list
Public Housing Developments: near-personal shares at small developments; held
a county roll-up or place tower: 56 % of tracts are suppressed or unpublished, so a sum would be a floor dressed as a total
Connecticut as suppression: HUD has a count for none of its 881 tracts, which is not the 10-or-fewer rule; drawn 'not published'

Caveats

Housing Choice Voucher households per 2020 census tract as HUD aggregates them, through December 2025. HUD omits tracts with 10 or fewer voucher holders, so a grey tract may have a few vouchers or none. The share is HUD's: vouchers over renter-occupied units in the 2020 Census. Qualified Census Tracts are a tax-credit designation, not a measure of conditions. HUD titles this list 2026, but its descriptions also say 2025 and 2024.
Connecticut: HUD's voucher layer has all 881 Connecticut tracts (filed under the old county codes 09001-09015) and a count for none of them, so Connecticut is 'not published', never 'suppressed'. HUD's QCT 2026 list files Connecticut under the planning regions (09110-09190). Each layer draws its own tracts; nothing is joined to a county.

Sensitivity

Voucher households per census tract as HUD aggregates them, with HUD's suppression of tracts with 10 or fewer; no tenant, landlord or address data. Fields are an allow-list (GEOID, HCV_PUBLIC, HCV_PUBLIC_PCT), never *.

Searchable fields

GEOID

Embedded as record text

nothing: no record text

API

/api/civic?op=housing

Related collections

LIHTC Qualified Census Tracts (HUD, titled 2026) 0.75; CDC/ATSDR Social Vulnerability Index 2022 by census tract 0.64; CDC PLACES community health estimates by census tract 0.61Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

LIHTC Qualified Census Tracts (HUD, titled 2026)

hud-qct · layer housing · show on the globe

What it holds

HUD's list of Qualified Census Tracts for the Low-Income Housing Tax Credit, drawn as amber dashed outlines over the voucher tracts, from the QUALIFIED_CENSUS_TRACTS_2026 service; a tract click says whether it is on the list.

A record is

A tract HUD designates for the tax credit (at least 50 % of households under 60 % of area median gross income, or a poverty rate of at least 25 %): a designation, not a measure of conditions.

Sources

Coverage

The 50 states, DC and Puerto Rico by census tract; Connecticut in planning-region codes (09110-09190); the designation year HUD titles 2026. 14,496 tracts in QUALIFIED_CENSUS_TRACTS_2026 layer 0 (returnCountOnly); the plain QUALIFIED_CENSUS_TRACTS service holds 15,727 under a layer named _2025 (counted 2026-10-17).

Freshness

Layer data last edited 2025-09-22. HUD's QCT year is stated three ways: the service and its item description say 2026 (QUALIFIED_CENSUS_TRACTS_2026), the item snippet says 2025, and the layer description says 'for 2024' (Date Updated: 10/2023). huduser.gov answered a keyless request with a WAF challenge (2026-10-17), so it is labelled 'titled 2026 by HUD'. Read from FeatureServer layer 0 editingInfo.dataLastEditDate (1758570469600); the service name, the item snippet and description, and the layer description, 2026-10-17.

Left out, and why

the QCT 2025 service (15,727 tracts): superseded by the list HUD titles 2026
Difficult Development Areas 2026 (2,902): a second designation: the hud-dda card (round 8), drawn by the same layer
NAME ("Census Tract 6.08"): a tract label; the GEOID is shown

Caveats

HUD's QCT year is stated three ways: the service and its item description say 2026 (QUALIFIED_CENSUS_TRACTS_2026), the item snippet says 2025, and the layer description says 'for 2024' (Date Updated: 10/2023). huduser.gov answered a keyless request with a WAF challenge (2026-10-17), so it is labelled 'titled 2026 by HUD'.

Sensitivity

A HUD tax-credit designation per census tract; no property or person data.

Searchable fields

GEOID

Embedded as record text

nothing: no record text

API

/api/civic?op=housing

Related collections

Housing Choice Vouchers by census tract (HUD) 0.75; Difficult Development Areas (HUD, titled 2026) 0.75; CDC PLACES community health estimates by census tract 0.67Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Difficult Development Areas (HUD, titled 2026)

hud-dda · layer housing · show on the globe

What it holds

HUD's Difficult Development Areas for the Low-Income Housing Tax Credit, from the Difficult_Development_Areas_2026 service: metro ZIP code areas (ZCTAs) and non-metro counties, drawn as violet dashed outlines over the voucher tracts when the 'Tax-credit areas (DDA)' chip is on; a click says which kind it is and names it as HUD does.

A record is

An area HUD designates for the tax credit because construction, land and utility costs are high relative to area median income, so a project there can claim a larger basis: a designation, not a measure of conditions.

Sources

Coverage

The 50 states, DC, Puerto Rico and the territories HUD lists; metro areas by ZCTA, non-metro areas by county (Connecticut's by planning region, NCNTY09160 and others); drawn below 100 km in a box of at most 0.3 degrees; the designation year HUD titles 2026. 2,902 areas in Difficult_Development_Areas_2026 layer 0 (returnCountOnly): 2,615 metro ZCTAs (DDA_TYPE SA) and 287 non-metro counties (NM); the older Difficult_Development_Areas service holds 2,958 (counted 2026-10-17).

Freshness

Layer data last edited 2025-09-22. HUD's DDA year is stated three ways: the service says 2026 (Difficult_Development_Areas_2026), its data was last edited 2025-09-22, and the layer description says 'Date of Coverage: 2023-2024 Last Updated: 10-2023'. huduser.gov answered a keyless request with a WAF challenge (2026-10-17), so it is labelled 'titled 2026 by HUD', as the QCT list is. Read from FeatureServer layer 0 editingInfo.dataLastEditDate (2025-09-22) and the layer description (read 2026-10-17), 2026-10-17.

Left out, and why

the older Difficult_Development_Areas service (2,958): superseded by the list HUD titles 2026
a county place tower: metro DDAs are ZIP code areas, so a county flag would mean different things in metro and non-metro counties

Caveats

HUD's DDA year is stated three ways: the service says 2026 (Difficult_Development_Areas_2026), its data was last edited 2025-09-22, and the layer description says 'Date of Coverage: 2023-2024 Last Updated: 10-2023'. huduser.gov answered a keyless request with a WAF challenge (2026-10-17), so it is labelled 'titled 2026 by HUD', as the QCT list is.
Connecticut: HUD's non-metro DDAs there are planning regions (NCNTY09160, Northwest Hills); its LIHTC rows carry the old county codes (CURCNTY 003, 009, 013).

Sensitivity

Published designations of ZIP code areas and counties; no person data.

Searchable fields

ZCTA5, DDA_CODE

Embedded as record text

nothing: no record text

API

/api/civic?op=housing

Related collections

LIHTC Qualified Census Tracts (HUD, titled 2026) 0.75; Low-Income Housing Tax Credit properties (HUD) 0.73; Housing Choice Vouchers by census tract (HUD) 0.61Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.

Low-Income Housing Tax Credit properties (HUD)

hud-lihtc · layer housing · show on the globe

What it holds

HUD's national database of LIHTC projects, drawn as teal points sized by low-income units when the 'Tax-credit properties (LIHTC)' chip is on: projects of five or more units at the address the state allocating agency reported, with the project's name, address, units, years placed in service and allocated, and HUD's coded fields (nonprofit sponsor, credit type, construction type, target population, tax-exempt bond, DDA and QCT flags) shown as HUD's codes.

A record is

A tax-credit rental project as the state agency reported it to HUD, located at the building with the most units. Not a vacancy or rent listing, and not a statement that the project is still in its compliance period.

Sources

Coverage

The 50 states, DC, Puerto Rico and the territories HUD lists; drawn below 100 km in a box of at most 0.3 degrees; only projects HUD geocoded to a rooftop (R) or a ZIP+4 centroid (4) are drawn, as HUD's layer description asks; allocated through 2021, placed in service through 2022 (YR_ALLOC and YR_PIS maxima; 8888 and 9999 are HUD's codes, not years). 50,566 projects in the LIHTC layer 0 (returnCountOnly): 45,765 with five or more units, of which 40,239 are geocoded R or 4 (drawn) and the rest to a block group, tract or ZIP centroid or not at all (counted, not drawn); 4,211 with one to four units (2,221 with one) and 590 with no unit count are never asked for; 2,367 have no coordinates (counted 2026-10-17).

Freshness

Layer data last edited 2024-12-09; the newest YR_ALLOC is 2021 and the newest YR_PIS 2022. Read from FeatureServer layer 0 editingInfo.dataLastEditDate; YR_ALLOC and YR_PIS outStatistics (read 2026-10-17), 2026-10-17.

Left out, and why

CONTACT, COMPANY, CO_ADD, CO_CTY, CO_ST, CO_ZIP and CO_TEL: the owner company, its contact person, their address and telephone: never requested (an allow-list, and a test denies them)
the geocoder bookkeeping (STD_*, DPV*, C1P*, MSG*, APT_*): not needed; never requested
projects of fewer than five units (4,211) and projects with no unit count (590): a small project is a household's home, so its point and address would say where a private person lives; the rule is in the upstream where (N_UNITS >= 5), and no answer counts withheld projects (Cam's decision, pending)
projects HUD located only to a block group, tract or ZIP centroid: HUD's layer description says to map only R and 4; they are counted in the layer's note, not drawn at a centroid
meanings of HUD's codes: HUD's LIHTC data dictionary is on huduser.gov, which did not answer keyless (2026-10-17), and the layer has no coded-value domains
HUD's Public Housing Developments and Multifamily Assisted properties: contact fields and per-development resident race, income and disability shares; held (D8-6)
record vectors: records-eval is pinned on 18 cards and would need re-measuring (D8-2)

Caveats

HUD's Difficult Development Areas, titled 2026 by HUD (the layer's own text says 2023-2024; HUD's page could not be read keyless), are metro ZIP code areas and non-metro counties with high costs relative to income, where tax-credit projects can get a larger basis. HUD's LIHTC database lists projects allocated through 2021 and placed in service through 2022, at the address the state agency reported. Projects of fewer than five units (4,211 of 50,566 nationally) and projects with no unit count (590) are never drawn, because a small project is a household's home. Projects HUD located only to a ZIP, tract or block-group centroid are counted, not drawn, as HUD's layer description asks. HUD has no administrative authority over the program; the states administer it.
HUD's LIHTC data dictionary is on huduser.gov, which did not answer keyless (2026-10-17), and the layer has no coded-value domains: coded fields are shown as HUD's codes, never a guessed meaning, and 8888 or 9999 in a year field is HUD's code, not a year.
Connecticut: HUD's non-metro DDAs there are planning regions (NCNTY09160, Northwest Hills); its LIHTC rows carry the old county codes (CURCNTY 003, 009, 013).

Sensitivity

Tax-credit rental properties as state agencies reported them to HUD. Owner company, contact person, their address and telephone are never requested. A project of fewer than five units is a household's home and is not drawn; its row never reaches the server. No LIHTC project name or address is searchable: the point is labelled by PROJECT as HUD publishes it (HUD sometimes files the street address as the name, and the label is then that address), the label is never matched by the palette or voice, and PROJ_ADD is never a searchable key; search finds a project only by its HUD id (HUD_ID). The address (PROJ_ADD, city, state, ZIP) appears on click, by design: these are public records of rental buildings of five or more units. A project with no PROJECT is labelled "LIHTC project · HUD id <HUD_ID>". Neither the name nor the address is embedded.

Searchable fields

HUD_ID

Embedded as record text

nothing: no record text

API

/api/civic?op=housing

Related collections

Difficult Development Areas (HUD, titled 2026) 0.73; LIHTC Qualified Census Tracts (HUD, titled 2026) 0.62; Formerly Used Defense Sites: properties (USACE FUDS) 0.55Nearest by text embedding of the card descriptions: they read alike, which is not a claim that the data agree.