Cloud-optimized versions of Global terrestrial ecoregions from the World Wildlife Fund (WWF), based on the landmark Olson et al. (2001) classification. The dataset delineates 847 ecoregions grouped into 14 biomes and 9 biogeographic realms. Each ecoregion represents a distinct community of plants and animals sharing evolutionary history.
1,616 polygons · EPSG:4326 · H3 native resolution 8 (parent 0) · license: public domain (US federal work, 17 U.S.C. § 105 — see https://edg.epa.gov/EPA_Data_License.htm)
US EPA Ecoregions of the United States, Level III, from the Omernik & Griffith framework — the US-native ecoregion standard, nested inside the North American (CEC) Level I/II tiers. 1,616 polygons. Covers the conterminous United States (1,250 polygons, from EPA's us_eco_l3) and Alaska (366 polygons, from EPA's ak_eco_l3), merged into a single layer of 1,616 polygons and distinguished by the region column. Hawaii, Puerto Rico and the other territories are NOT covered: the Omernik framework extends only to CONUS and Alaska. For those areas use the global WWF terrestrial ecoregions collection in this bucket.
Every polygon carries the FULL hierarchy as attributes (NA_L1CODE/NAME, NA_L2CODE/NAME, US_L3CODE/NAME), so Levels I and II require no separate dataset — group this layer by the level you want. Distinct classes present: 12 Level I, 25 Level II, 105 Level III.
Delivered as GeoParquet for analysis, PMTiles for web mapping, and H3 hexagonal parquet at native resolution 8 (parent resolution 0) for spatial joins and aggregation. The Aleutian Islands place 13 polygons east of the antimeridian (longitudes 172.46°E to 179.78°E), so the collection bbox is given in the STAC antimeridian form with xmin > xmax, and the hex partition set includes the dateline resolution-0 cell 576707042908045311. No single polygon straddles the antimeridian, so no geometry spans the globe.
Reprojected from the source Albers equal-area projections to EPSG:4326 with GDAL during preprocessing; geometry is otherwise unmodified. Attribute spelling and letter-case are verbatim from EPA — note that Level I/II names are ALL-CAPS while Level III/IV names are Title Case.
Merge note: the two source products use disjoint Level III numbering (CONUS 1–85, Alaska 101–120), so US_L3CODE and US_L3NAME are unique across the merged layer and safe to group by without also grouping by region.
Group by an identifier column rather than a name column. The identifier columns are internally consistent, but a few of EPA's own names carry more than one spelling for the same unit, so grouping by the name splits that unit in two. Measured in this release:
Provenance. EPA publishes these files without a version label and overwrites them in place, and the previous download host has already stopped serving them, so the copy staged in this bucket is the durable record of what was ingested, retrieved 2026-08-27:
Counting: one row is one polygon, not one ecoregion. The 1,616 polygons make up 105 Level III ecoregions, 59 of which are split into several separate polygons.
STAC: https://s3-west.nrp-nautilus.io/public-ecoregion/epa-ecoregions-l3/stac-collection.json
The source-layer is epa-ecoregions-l3. It is the last path segment of the dataset name, NOT the
EPA shapefile layer name (us_eco_l3 / ak_eco_l3). Using anything else renders a blank map — this is
the single most common mistake.
Note the Level I / Level II names are ALL-CAPS in the source while Level III / Level IV
names are Title Case. Match the values exactly as enumerated in the STAC values arrays.
h8 is the catalog's universal join key, so this layer joins straight to any other
res-8-carrying hex asset.
⚠️ Never SUM(Shape_Area) or SUM(Shape_Leng) on the hex asset. They are per-polygon
totals repeated on every cell the polygon covers. Dedup first
(SELECT DISTINCT _cng_fid, Shape_Area …), or derive area from the H3 footprint as above.
Source: EPA Office of Research and Development, Ecoregions of North America
(https://www.epa.gov/eco-research/ecoregions), staged to https://s3-west.nrp-nautilus.io/public-ecoregion/raw/.
Reprojected from the source Albers equal-area projections to EPSG:4326 with GDAL, then
converted, tiled and hexed with cng-datasets.
Build recipe: catalog/ecoregion/k8s/epa-ecoregions*/ in
boettiger-lab/data-workflows (issue #633).
Cite the framework as Omernik, J.M. (1987) Ecoregions of the conterminous United States, Annals of the Association of American Geographers 77(1):118-125, and Omernik, J.M. & Griffith, G.E. (2014) Ecoregions of the conterminous United States: evolution of a hierarchical spatial framework, Environmental Management 54(6):1249-1266.