Nationwide assessment of per- and polyfluoroalkyl substances in US public drinking water systems: detection patterns, screening exceedances and predictors of contamination from UCMR 5 monitoring data
Tanveer, H.U.; Ali, M.; Fatima, S.; Tanveer, A.; Azam, A.; Tanveer, H.; Aslam, R.F.; Hussain, A.; Ali, A.M.
In preparation for ACS ES&T Water
What the paper does
Between 2023 and 2025 the US Environmental Protection Agency (EPA) made every large public water system, and a sample of small ones, test for 29 per- and polyfluoroalkyl substances (PFAS) under the fifth Unregulated Contaminant Monitoring Rule (UCMR 5). The analysis works from 1,863,306 analytical results across 10,297 systems in the data as released so far. UCMR 5 is still being added to, so the specific data release and as-of date will be stated with the manuscript. Several organizations have already put the data on a map. We asked what it actually says about how many systems have a problem, how big the problem is, and what predicts it.
At least one PFAS turned up in 34.4 percent of systems. About 16.7 percent exceeded a screening threshold for one of the new legal limits, and those systems serve somewhere between 64 and 80 million people. When we averaged the way the compliance rule will, about 6.8 percent of systems still exceed. The strongest predictors were geographic: which EPA region a system is in and how close it sits to a Superfund site. The apparent links to demographics faded once we accounted for state-level clustering. Only about one in ten of the exceeding systems reported any PFAS-specific treatment. Compared with the previous national survey a decade earlier, detection at the same analytical threshold has fallen slightly, consistent with the manufacturing phase-outs.
From detection to exceedance
How many systems, depending on what you mean
Why it matters
Between 64 and 80 million people are drinking from systems that screen above the new limits, and nine in ten of those systems have not yet installed treatment. The paper also tells regulators where to look first. It warns that headline exceedance counts depend heavily on how you handle the results below the detection limit, the same problem I worked on with disinfection byproducts at George Mason. The analysis is in Python and the code will be released with the paper.