A headline that says PFAS contamination in drinking water may be underestimated invites a yes or no answer. For a utility, a dealer, or a lab, the more useful question is how the total was assembled. In a Circle of Blue report on PFAS undercounting, written by Alissa de Chassey and dated October 23, 2026, the reporting starts at the edge of the measurement itself: "When PFAS concentrations in drinking water fall below laboratory reporting limits" the number that results is no longer a direct reading. The piece notes that it originally appeared on Inside Climate News.
A reporting limit is the lowest concentration a laboratory will report as quantified under its method, calibration, and quality controls. Below that floor, the lab may report a non-detect, a flagged estimate, or a value outside its formal limits. Each output looks different in a data file, and each has to be handled before anyone adds the results together.
Three Ways to Write a Below-Limit Result
Analysts use several conventions when they substitute a value for a non-detect before summing. The three most common are zero, half the reporting limit, and the full reporting limit. Zero gives a floor, since it assumes nothing is present below the limit. The full limit gives a ceiling, since it assumes every non-detect sits right at the threshold. Half the limit falls between the two and is a modeling choice, not a measurement. Some analysts drop non-detects altogether, which biases the total in a direction that depends on how many samples fell below the limit.
When few samples sit near the floor, the choice matters little. When many do, the spread between floor and ceiling can grow larger than the difference the study is trying to detect. That is the arithmetic behind any claim of underestimation. A total built on zeros will read lower than one built on limits, so the claim is only as strong as the convention behind it. Timing adds another layer. Laboratories revise methods over time, and a limit that was reasonable in one year may not be the limit used the next, so a multi-year total can mix floors without showing it.
What to Check Before Quoting the Total
A reader who wants to use the study in a board packet, a dealer's sales file, or a lab's method notes should check the full report for four things. First, which reporting limit applied to each compound, and whether it changed across the laboratories or years included. Second, how non-detects entered the sum. Third, whether the sampling frame covered systems that test rarely or only at required intervals. Fourth, whether the estimate of underestimation came from comparing like totals or from comparing one convention against a measured result.
The excerpt available here does not show which convention the study used or how it treated differing laboratory limits. Treat any total or proportion drawn from it as unverified until the methods section has been read against those four questions. The headline may describe a real gap. The summary alone does not establish how large that gap is.
Keep the Study Apart From Your Own File
A study total is not a utility's compliance record. Your own monthly and quarterly sums carry your laboratory's reporting limits, your data flags, and your handling rule. Lining them up against a national or regional figure requires that both use the same convention. If they do not, the difference you see may be an artifact of bookkeeping rather than a change in the water. Where the utility's annual report and the study use different periods, the comparison is weaker still, and the mismatch should be written down before anyone draws a trend.
A household that wants to see its own results should start with the utility's report and, where that report does not answer the question, a private test drawn under a documented procedure. Questions about coverage for home systems and appliances come after that record exists, because the test result is what allows any equipment claim to be checked.
Read the headline as a statement about method until the method is shown. The undercount may be real. The summary available here does not, on its own, settle its size, and a careful reader keeps that uncertainty in the file rather than resolving it in the headline.