Replica Parcel-Level Land Use Data

Replica also offers a certainty table that provides county-level metadata on the precision of our parcel-level land use data. Replica defines 3 levels of confidence at the parcel-level: 

1. labeled: parcels that are “labeled” were assigned their land use based on available parcel-level features in ground truth data, such as their Point of Interest NAICS industry code, USPS delivery codes, zoning and use code text, and so on. This is our highest level of confidence.
2. sampled: parcels that are “sampled” are missing parcel-level ground truth information, but are located in tracts in which aggregate land use data is known. Their land use is sampled according to aggregate land use distributions.
3. unknown: parcels that are “unknown” are missing parcel-level and aggregate land use information alike, so their land use and confidence level is unknown. In the absence of any ground truth, Replica does not attempt to impute a use.

‍

See the Data in Action

Click here to see a sample visualization of the data in Foursquare Studio.

What is the source(s) of the land use data?

Replica's land use data is modeled from a composite of data sources, including UrbanFootprint, Regrid, and other public and private sources:

• Parcel-level land use data
• Aggregate land use data
• Places-of-Interest (POIs) data (e.g. restaurants, stadiums, theaters)
• Building footprint data
• Census housing unit data
• Purpose-specific auxiliary datasets (e.g. schools, airports)

‍

What is the methodology used to model parcel-level land use data? 

Parcel-level land use data was collected for about 150M of all parcels in the country. These parcels were then enriched with data on POIs, e.g. restaurants, retail stores, and museums. Parcels with unknown land use were modeled based on aggregate land use data and other attributes (e.g. POIs and parcel size). Lastly, parcels were matched to buildings to estimate building area and mixed-use buildings.

‍

How did Replica decide on the land use classifications listed above? 

The land use of parcels is classified using direct input data from proprietary sources where available, and otherwise modeled using tract-level characteristics. Parcel-level data can come in a variety of forms, including spatially joined points of interest, zoning descriptions, and postal delivery codes. Tract-level characteristics consist of aggregate land use proportions that are further weighted using the relative, block-level concentration of housing units in Census data. We defined categories that were as consistent as possible based on our various input data providers’ schemas and standards such as the American Planning Association’s LBCS.

‍

What year is the land use data based on?

We currently offer Fall 2021 parcel-level land use data and plan to refresh the data annually.

‍