Latest National Agriculture Imagery Program releases distributed as visualization COGs
RGB JPEG Cloud-Optimized GeoTIFFs of USDA NAIP quarter quads, converted from the naip-analytic MRF files with the settings, file names and key layout of the AWS naip-visualization bucket. Every 2024 and 2025 state-year, which that bucket does not hold, and the latest flight of the four states last flown in 2023, with reduced pictures of each state and of the country for drawing a map.
One file per quarter quad per state-year, from flights of 2021 to 2025. Each file name holds the date it was flown; a 2023 name then holds a second date, the day the file was made. The quad index gives the flight date of every file.
<state>/<year>/<resolution>/rgb/<block>/<name>.tif, as in the AWS naip-visualization bucket, for example ut/2024/60cm/rgb/40111/m_4011127_ne_12_060_20240816.tif.overviews/<state>-<year>-utm<zone>.tif (one flight year of one state in one UTM zone, 9.6 m) and overviews/national-latest.tif (the latest flight of each state, EPSG:3857, about 306 m).stac/naip.parquet lists every quarter quad and stac/naip-overviews.parquet every overview file, as STAC GeoParquet: address, size, SHA-256, footprint, CRS and flight date. A reader finds the files for a place by querying the index on its bbox columns.Each index is a single GeoParquet file in the STAC GeoParquet layout: one row per STAC Item, with
the file's address in the image asset. There are no Item JSON files and no partitions. The
STAC Collection is stored in the file's own metadata under the stac-geoparquet key, and the
GeoParquet metadata under geo.
stac/naip.parquet has 245,554 rows, one per quarter quad file, covering 213,436 places; a place
flown in more than one listed year has a row for each. Its columns:
id: state and file name, for example ut_m_4011127_ne_12_060_20240816.geometry: the footprint polygon, WKB in longitude and latitude; bbox: its extent as
xmin, ymin, xmax, ymax.datetime: the flight date, at 00:00 UTC.gsd (0.3 or 0.6 m), naip:state, naip:year.proj:code, proj:epsg, proj:bbox, proj:shape, proj:transform: the file's UTM grid.assets.image: href (the file's address), file:size, file:checksum (SHA-256 as a
multihash), type and roles.naip:run: the conversion run that made the file.stac/naip-overviews.parquet has 115 rows, one per overview file, with the same columns except
naip:run, plus start_datetime and end_datetime (the flights drawn into the file) and
naip:quads (how many quarter quads it was made from). The national file's row has no state
or year.
Rows of the quad index are ordered along a Hilbert curve through the centre of each footprint,
newest flight first at the same place, with states and years mixed. They are written in 30
row groups of about 8,192 rows, with statistics and a page index, so a query for a small area
reads a few row groups instead of the whole 44 MB file. For the newest picture of a place, keep
the row with the latest datetime.
The quarter quads around Alta, Utah, newest first:
Use GeoParquet for analysis. Field dictionaries describe columns and units.
Column definitions in the catalog include names, storage types and descriptions. Use the field dictionaries and producer documentation above for units, coded values and limitations.
Query the remote GeoParquet directly:
USDA Farm Production and Conservation Business Center, Geospatial Enterprise Operations (NAIP). Source files from the NAIP on AWS open data buckets maintained by Esri.
NAIP imagery is produced by the USDA Farm Production and Conservation Business Center and is in the public domain. USDA asks for credit; these files are a format conversion, not a USDA product.
Independently maintained by nthh.
Conversion scripts and validation ยท Portolan map