Where PFAS data is — and isn't — public

Three different things are called “PFAS data”

Most confusion about PFAS data availability comes from treating three unrelated objects as one. They differ in what they prove, and they differ enormously in how public they are.

Monitoring measurements are concentrations someone measured in water, groundwater, sediment, soil or tissue at a known location. They are evidence about the environment. Site-investigation records are the reports produced when a specific site is characterised — including the depth-resolved soil profiles that say how much PFAS is in the ground and how far down. Potential-source inventories are neither: they are lists of facilities whose industry suggests PFAS may have been used, which tell you where to look and nothing about whether anything was found.

What is public: the monitoring layer

Ambient monitoring is the healthy part of the picture. SWATGenX consolidates federal and state monitoring programs into a single inventory spanning almost every state in the country and six media — water, groundwater, sediment, tissue, soil and porewater. It is being actively expanded, so the station and observation totals move week to week; the live map and the API report the current holding rather than a figure quoted here that would be out of date by the time you read it.

One property of that inventory is routinely misread, and it is stable enough to state plainly even while the totals move: most PFAS observations are non-detects. Across every snapshot we have taken, the clear majority of records report that the analyte was not found above the reporting limit. That is not a defect in the dataset. A non-detect is a real measurement, and often the more useful one, because it bounds where PFAS is not — which is exactly what a transport model needs in order to be falsifiable. But an observation count quoted on its own invites a reader to hear that many detections, and that is not what it means.

What isn’t: the site-investigation record

This is the gap, and it is the one that matters for modeling. A soil concentration without its depth cannot become a source term — the whole question at a legacy site is how much material is still held up in unsaturated soil and how long it will take to arrive. Depth is the variable the model needs, and depth is the variable that most public summaries drop.

When those reports were pursued systematically, the result was stark. Across 619 military and AFFF installations for which at least one targeted agency repository query was made, 58 — about 9% — had a publicly retrievable depth-resolved soil report. The reports mostly exist. They are scanned PDFs in records rooms, unindexed document portals, and public-records responses; nothing links them, few are machine-readable, and depth frequently appears only inside an image of a table. The barrier is retrievability, not secrecy — which is a more tractable problem, and the reason assembling the data was worth doing at all.

The result of that effort is open: 30,972 depth-resolved records from 5,812 borings across 41 states, assembled from 141 separate agency sources and covering 72 analytes, published with a DOI so that it can be cited and checked. It is browsable as the PFAS soil inventory.

The screening layer, and what it does not mean

The platform also carries 196,710 potential-source facilities in four categories — industrial and manufacturing, oil, gas and mining, wastewater and biosolids, and AFFF and defense. This layer marks where PFAS may have been used or released. It is not a map of confirmed contamination, and a facility appearing on it is not an allegation about that facility. It is a screening aid whose purpose is to tell you where a monitoring gap is worth worrying about — useful precisely because the monitoring network was never designed around PFAS sources.

How to get at it

All three layers are open to read. The map shows monitoring stations and potential sources together, with a preview limit on how many features an anonymous session pulls at once so the public endpoints stay responsive; a free account lifts it. The soil inventory is separately downloadable in full from its DOI deposit with no account at all, and the same holdings are queryable programmatically through the API and the MCP server. What you do with the data afterwards — turning a measured soil profile into a source term and running it forward — is what the site-scale modeling workflow is for.