EPA Community Water System Service Area Boundaries (v3)

Polygon service-area boundaries for community public water systems (CWS) in the United States — version 3 (March 2026) of the U.S. EPA Office of Research and Development's Service Area Boundary (SAB) dataset, reprocessed into cloud-native formats.

• Coverage: 44,656 community water systems (~99% of U.S. population served by CWS), all 50 states + DC + tribal/territorial. The H3 hex asset represents 44,128 of these (528 very-small features did not cover a single H3 res-10 cell).
• Provenance: ~60% authoritative (state primacy agencies, utilities); ~40% EPA-modeled from building footprints, parcels, population, OSM. The Model_Method column is empty for authoritative boundaries.
• Geometry: MultiPolygon, reprojected from NAD83 (EPSG:4269) to WGS84 (EPSG:4326).
• License: U.S. federal data — public domain (no restrictions).
• Source: https://www.epa.gov/ground-water-and-drinking-water/community-water-system-service-area-boundaries
• Citation: EPA Office of Research and Development. Community Water System Service Area Boundaries v3, March 2026. https://github.com/USEPA/ORD_SAB_Model

What's in this collection

The full v3 release also contains a T_NTNC layer (transient + non-transient non-community water systems — schools, campgrounds, gas stations, etc., ~77k features). Only the CWS layer is published here; T_NTNC may be added later.

The raw EPA zip (with documentation PDF and census crosswalk tables) is preserved at raw/PWS_Boundaries_Latest.zip.

Key columns

See the STAC collection for the full data dictionary including values for coded columns and per-feature aggregation warnings on the hex asset.

Usage

DuckDB (GeoParquet)

1INSTALL spatial; LOAD spatial;
2INSTALL httpfs; LOAD httpfs;
3
4-- All Texas community water systems serving >100k people
5SELECT PWSID, PWS_Name, Population_Served_Count, Service_Area_Type
6FROM read_parquet('https://s3-west.nrp-nautilus.io/public-epa-water/epa-sab-v3/cws.parquet')
7WHERE Primacy_Agency = 'TX' AND Pop_Cat_5 = '>100,000'
8ORDER BY Population_Served_Count DESC;
1INSTALL spatial; LOAD spatial;
2INSTALL httpfs; LOAD httpfs;
3
4-- All Texas community water systems serving >100k people
5SELECT PWSID, PWS_Name, Population_Served_Count, Service_Area_Type
6FROM read_parquet('https://s3-west.nrp-nautilus.io/public-epa-water/epa-sab-v3/cws.parquet')
7WHERE Primacy_Agency = 'TX' AND Pop_Cat_5 = '>100,000'
8ORDER BY Population_Served_Count DESC;

DuckDB (H3 hex — spatial joins)

H3 hex parquet is hive-partitioned by h0. Always include an h0 filter when querying a region — otherwise you scan all 122 partitions.

1-- Population served by any CWS overlapping a HydroBasins HUC8 (example)
2WITH region AS (
3  SELECT DISTINCT h8 FROM read_parquet(
4    'https://s3-west.nrp-nautilus.io/public-hydrobasins/.../hex/h0=*/data_0.parquet'
5  ) WHERE HYBAS_ID = 7080023860
6),
7systems AS (
8  SELECT DISTINCT cws.PWSID, cws.PWS_Name, cws.Population_Served_Count
9  FROM read_parquet(
10    'https://s3-west.nrp-nautilus.io/public-epa-water/epa-sab-v3/cws/hex/h0=*/data_0.parquet'
11  ) cws
12  JOIN region USING (h8)
13)
14SELECT SUM(Population_Served_Count) AS total_pop, COUNT(*) AS n_systems FROM systems;
1-- Population served by any CWS overlapping a HydroBasins HUC8 (example)
2WITH region AS (
3  SELECT DISTINCT h8 FROM read_parquet(
4    'https://s3-west.nrp-nautilus.io/public-hydrobasins/.../hex/h0=*/data_0.parquet'
5  ) WHERE HYBAS_ID = 7080023860
6),
7systems AS (
8  SELECT DISTINCT cws.PWSID, cws.PWS_Name, cws.Population_Served_Count
9  FROM read_parquet(
10    'https://s3-west.nrp-nautilus.io/public-epa-water/epa-sab-v3/cws/hex/h0=*/data_0.parquet'
11  ) cws
12  JOIN region USING (h8)
13)
14SELECT SUM(Population_Served_Count) AS total_pop, COUNT(*) AS n_systems FROM systems;

The SELECT DISTINCT cws.PWSID, ... step is essential — Population_Served_Count is repeated on every hex row a service area covers. See the hex asset's table:columns description for the full duplication caveat.

MapLibre GL JS

1map.addSource('epa-sab-cws', {
2  type: 'vector',
3  url: 'pmtiles://https://s3-west.nrp-nautilus.io/public-epa-water/epa-sab-v3/cws.pmtiles'
4});
5
6map.addLayer({
7  id: 'epa-sab-cws-fill',
8  type: 'fill',
9  source: 'epa-sab-cws',
10  'source-layer': 'cws',   // <-- last segment of --dataset, NOT the GDB layer name
11  paint: {
12    'fill-color': '#1f78b4',
13    'fill-opacity': 0.35,
14    'fill-outline-color': '#08306b'
15  }
16});
1map.addSource('epa-sab-cws', {
2  type: 'vector',
3  url: 'pmtiles://https://s3-west.nrp-nautilus.io/public-epa-water/epa-sab-v3/cws.pmtiles'
4});
5
6map.addLayer({
7  id: 'epa-sab-cws-fill',
8  type: 'fill',
9  source: 'epa-sab-cws',
10  'source-layer': 'cws',   // <-- last segment of --dataset, NOT the GDB layer name
11  paint: {
12    'fill-color': '#1f78b4',
13    'fill-opacity': 0.35,
14    'fill-outline-color': '#08306b'
15  }
16});

Processing notes

• Reprojected from NAD83 (EPSG:4269) to WGS84 (EPSG:4326) at conversion time.
• Shape_Area and Shape_Length are carried through from the source dataset in the original CRS units; they are not recomputed after reprojection. For accurate area, compute from the WGS84 geometry or use the H3 hex count.
• H3 native resolution 10 (≈15,000 m² cells), with parent rollups at 9, 8, and 0.
• One row per (CWS feature, H3 cell) pair in the hex asset.
• Built with cng-datasets on NRP Nautilus.