Identifying Government-Owned Parcels

We start by running our  Who Owns America® (WHOA) national analysis . This analysis takes all available  Regrid  parcel data and, after extensive cleaning and normalizing of key fields, classifies each parcel in the nation into distinct ownership types. Our WHOA general ownership types include government, owner-occupied, in-state corporate, out-of-state corporate, in-state non-corporate, and out-of-state non-corporate entities.

For classifying government-owned lands, our WHOA analysis relies heavily on checking the owner name listed in the parcel dataset. We have compiled lists of federal, state, and local government terms to compare parcel owner names against, as well as match rules for each terms. Our model compares parcel owner names against our lists of government key terms to determine if they are in fact government-owned. To classify a parcel as government-owned, some terms must be an exact match (such as USA or U.S.A), and others can be a partial match (such as Commonwealth of [State Name] or [City] Redevelopment Authority). Our lists also attempt to take into account the many potential variations in spelling of a name -- for example, in our search on federal lands, we have found over 20 variations in the spelling of agency names like the Bureau of Land Reclamation or the US Army Corps of Engineers. Using a combination of word recognition algorithms, we aim to identify as many government key terms as possible through our search.

We collected a number of datasets prepared by federal agencies and partners to crosscheck our parcel data against. By running spatial intersects, we compared what we classified as government-owned against what datasets recognized as government owned. For this initial sweep, we used data prepared by the U.S. Geological Survey (USGS), such as the  Protected Areas Database of the United States  and various National Geospatial Data Asset (NGDA) data made available on Esri's Living Atlas, such as the  USA Federal Lands  layer.

For a second pass, we compiled and collected state agency data, such as the  Vermont Protected Lands Database (VLPD) , the  State Owned Lands dataset for North Carolina , and the  California State Property Inventory , to compare our parcel data against. As with the initial check against national, authoritative datasets, we compared our parcels against these state datasets to make sure we accounted for and included all government-owned lands in our parcel data.

We completed a third sweep of checks by compiling, where possible, local authoritative data on government-owned lands. Recognizing that many county and/or city governments maintain and share datasets on their properties, we searched and compiled local datasets from county and/or city agencies, such as  Davidson County, Tennessee's Open Data Portal  and the  City of Phoenix Open Data Portal .

These additional checks not only allow us to check our work, they also help us identify additional key government terms or additional variations in government names to consider. We then ran recursive searches back on our dataset to find additional parcels with owner names containing these additional government terms. These checks against federal, state and local data thus allow us to compile as complete a dataset as possible, of government-owned parcels across the country.

Using data pulled from the Living Atlas, we flag parcels that intersect with  Federal American Indian Reservations ,  State American Indian Reservations ,  Off-Reservation Trust Lands , and  American Indian Joint-Use Areas .

Using data from the  Protected Areas Database of the United States , as well as  USA Parks  and previously collected local sources, we flag parcels that are protected and/or conserved as green space. We also used data from the  Landmarks and Government Buildings  dataset, as well as other layers made available through the Living Atlas, to flag parcels that represent particular types of use, such as  medical response structures ,  law enforcement structures ,  educational structures  and  dams .

We also flag parcels that intersect with  flood hazard areas , using data derived from the Federal Emergency Management Agency's  National Flood Hazard Layer .

Starting with Regrid's  right of way  attribute, we flag parcels that appear to represent transportation and utility networks. To make sure we account for as many road networks as possible, we intersect our parcels against  transportation  networks, flagging those as Right of Ways if they are not already marked as such by Regrid.

We end with flagging parcels that intersect with  Qualified Opportunity Zones ,  Difficult Development Areas , and  Justice40 census tracts .

Using information provided through Regrid's parcel schema, we flagged parcels with no buildings and structures with a parcel area of at least 20,000 sqft. We then filtered to parcels with buildings, and calculated the building area percentage -- or the percent of parcel area covered by the building. Parcels where this percentage was less than 5%, where the existing building and/or structure was less than 1000 sqft, and where the parcel had an overall area of at least 20,000 sqft were flagged as buildable.

Using a local copy of EPA's Smart Locations Mapping data, we created a subset of census block groups that met our condition for transit-accessible. We then flagged any parcels that are within one of these census block groups as transit-accessible.

To define urban areas, we used The Census Bureau's  2020 USA Census Urban Areas  dataset, flagging parcels that fall within a USA census urban area.

We started by using the Federal Aviation Administration's  Runways layer , flagging parcels that intersect this layer. We then used the Federal Aviation Administration's Airport points layer as a way to manually check and flag parcels near these points, comparing the parcels against an aerial imagery basemap to flag parcels that appeared to be part of the airport.

As a final step to try to remove as many road networks and slivers from our final dataset, we also ran searches for parcels with terms like ROADS, TRANSIT, TRANSPORTATION, and HIGHWAY and spot-checked results, manually flagging noticeable road slivers and segments as Right of Ways in the parcel dataset.