Methodology & sources

Two tiers of measurement

Every PFAS value is one of two tiers, encoded in the field name so the two never blur together. All concentrations are published in parts per trillion (ppt).

The federal limits we compare against

Every safety-critical number enters through an owner-signed sources file with a primary-source citation — never from memory. These are the current, legally enforceable EPA maximum contaminant levels (MCLs), set by the April 2024 PFAS National Primary Drinking Water Regulation (40 CFR 141 Subpart Z).

• PFBS has no individual MCL. It is regulated only as a Hazard-Index component, with a health-based water concentration of 2000 ppt used as its denominator in the mixture calculation.
• DoD interim-action screening level: 12 ppt for PFOA and PFOS — three times the EPA MCL — the level at which the Department of Defense prioritizes off-base interim action. This is a DoD cleanup trigger, not a drinking-water MCL.

The five severity bands

Each water system, and each installation on the map, is colored by a single severity band — the worst PFAS result among the systems serving the surrounding area. The five bands, worst to least severe:

How we associate a water system with an installation

We join the community water systems whose service area overlaps or lies within 3 miles of the installation boundary. The association is strictly geographic proximity — it is never a statement that the installation is the source of anything measured in a nearby system. For each installation we test three geographic relationships in order and stop at the first that matches:

1. Overlap — service area overlaps the installation boundary.
2. Within 3 miles — service area within 3 miles of the installation boundary.
3. County / ZIP fallback — same county as the installation (county/ZIP fallback; no measured service-area geometry). This is only used when a base has no service-area boundary at all ( community water systems serving the same county as the installation (county/ZIP fallback, used only when no service-area boundary is available)); those rows carry no measured distance.

Separately, federal water systems located on the installation itself (on-base systems) are attached as their own rows and never merged into a community-system row — each is flagged as an “on-base federal water system”.

Why the buffer is 3 miles

The 3-mile buffer is an owner-signed method parameter, chosen after re-running the geographic join at 1, 3, and 5 miles so the number could be signed against evidence rather than picked arbitrarily. The counts below cover the polygon-buffer join only (county/ZIP fallback and on-base federal attachment are radius-independent).

Widening from 1 to 3 miles adds 688 system rows and 25 more installations with at least one joined system. Widening again from 3 to 5 miles adds 815 rows but only 10 more installations — the extra distance mostly pulls in systems that are no longer plausibly local. Three miles captures the adjacent-community systems while keeping the join defensibly tight.

What a UCMR5 result can and cannot tell you

• The reporting-limit floor

  UCMR5 laboratories report only down to a minimum reporting level (MRL) — for PFOA and PFOS that floor is 4 ppt. UCMR5 cannot see below the MRL, so a “non-detect” means “not detected above the reporting level,” not zero. We label such a result exactly that: “non-detect (sampled under UCMR5; result below the minimum reporting level)”.

• Not-tested is distinct from non-detect

  A system that was never part of UCMR5 PFAS monitoring is “not tested for PFAS under UCMR5” — a separate, honest state, not a clean result. We never present an untested system as a non-detect: its UCMR5 fields stay null and every derived flag stays off.

• Monitoring data, not a violation

  UCMR5 is federal monitoring data, not a violation determination. A UCMR5 value above a limit is never reported here as a violation. The only violation counts in this dataset come from EPA ECHO Safe Drinking Water Act (SDWA) public water system records.

Current law versus what is only proposed

The April 2024 MCLs above are current law. In 2026 the EPA published two proposals affecting this rule. Proposals are not final rules — the April 2024 limits remain in force unless and until any change is finalized. Both comment periods close 2026-07-20.

• 2026 EPA proposed rescission of the PFHxS/PFNA/HFPO-DA/PFBS + Hazard-Index limits — EPA proposed to rescind the regulatory determinations for the PFHxS, PFNA, HFPO-DA, and PFBS limits together with the Hazard Index (docket EPA-HQ-OW-2025-0654, Federal Register document 2026-10085; comment period closes 2026-07-20).
• 2026 EPA proposal to extend the PFOA/PFOS MCL compliance deadline — EPA proposed to adjust the compliance timing for the PFOA and PFOS MCLs, not the limits themselves (docket EPA-HQ-OW-2025-1742, Federal Register document 2026-10086; comment period closes 2026-07-20).

Coverage, in numbers

These are the honest coverage figures from the committed coverage report, current as of 2026-07-11.

Limitations we are explicit about

• County / ZIP served is imperfect. When a base has no service-area boundary we fall back to systems serving the same county, using ECHO's counties- and ZIPs-served fields. County geography is coarse, so a fallback join can pull in a system across a large county or miss one just over a county line. About 41.1% of joined systems use this fallback.
• Polygon coverage has gaps. Service-area boundaries do not exist for every system, and some installations have only a point or a manual centroid (closed or realigned bases absent from the MIRTA boundary set), which we approximate with a 1-mile disc. A missing boundary can leave a genuinely nearby system unjoined.
• The roster coverage gap. Of 723 installations, only 352 are resolved to a boundary or centroid; the remaining 371 are listed without geometry. As a result, about 376 installations show zero joined systems. They are kept in the dataset and shown honestly, never dropped.
• UCMR5 does not cover every system. Only about 47.4% of joined systems (1,309 of 2,762) have UCMR5 PFAS data; the rest are marked not tested — which, again, is not the same as clean.

Cite this data

This dataset is published under a Creative Commons Attribution 4.0 (CC BY 4.0) license — free to republish, chart, or build on with attribution to TapWaterData Editorial. Suggested citation:

TapWaterData Editorial. "Military PFAS Dataset: U.S. Department of Defense PFAS-assessed installations joined to nearby community water systems and federal PFAS drinking-water monitoring (UCMR5)." Data updated 2026-07-11. Licensed under CC BY 4.0. https://tapwaterdata.com/military-pfas