A PFAS result is two numbers before it is one. The first is the concentration the laboratory reports. The second is the reporting limit, the level below which the method could not assign a value with the confidence the laboratory is prepared to state. A non-detect is therefore a boundary, not a measurement. The record is only as sound as the way that boundary survives the trip from the laboratory report into a spreadsheet, a board memo, or a public summary.
A recent item from Circle of Blue, written by Alissa de Chassey and dated October 23, 2026, takes up the consequence. Its summary describes a new study suggesting PFAS contamination in drinking water may be underestimated, and it notes that when concentrations "fall below laboratory reporting limits," the result tells the reader less than a measured value would. The summary does not describe the study's sampling design or its statistical method, so this article makes no claim about how large any gap might be. Readers who need the methods should go to the original article before drawing conclusions. The point for this sector is narrower: a censored result carries less information than its neighbors, and any total built from it inherits that loss.
Zero, the Limit, or a Model: The Substitution Is a Decision
Totals across a set of samples need a value for each non-detect. Substituting zero produces the lowest possible sum and can make a site look cleaner than the laboratory could show. Substituting the reporting limit produces the highest plausible sum and can make the same site look worse. Using half the limit, or a distribution fitted to the detected values, sits between the two and needs its own justification. None of these is a neutral default. Each changes the total, and a changed total can change a decision about treatment, sampling frequency, or disclosure.
The reporting limit itself is not fixed. It depends on instrument sensitivity, sample volume, matrix, extraction, and any dilution the laboratory applied to that sample. Two samples collected in the same week can carry different limits because they ran in different batches. An apparent drop from a detected value to a non-detect may therefore reflect a higher limit in one batch rather than a cleaner tap. Method detection limits and reporting limits are also not the same thing, and a file that mixes them invites error.
What Has to Travel With the Number
A usable file keeps, for each result, the reported value, the qualifier that marks it as a detect, an estimate, or a non-detect, the reporting limit applied to that specific sample, the method and laboratory, the sample date, and the substitution rule used in any total. Losing one of these fields is common. Export templates often drop qualifier columns, and a column of bare numbers reads as though every value were measured.
The second break occurs in summary documents. Annual reports and public tables sometimes pool detects and non-detects under one heading, or quote a maximum without saying whether it was a measured value. Where a source does not state its reporting limit, the accurate line is that the limit is unstated, not that the result is zero.
A File Standard for the Next Review
Keep the laboratory report as the source document and treat spreadsheet values as copies. Store the reporting limit per sample rather than per program. Write the substitution rule into the methods line before any total is calculated, and do not change it after seeing the result. The standard should also name who may change a qualifier after review and record the reason for the change. When quoting an outside study, state what the source says and what it does not say, and leave out any figure the source does not give.
Uncertainty does not disappear with a better form. It can be stated plainly. A record that says a result is below a stated limit, measured by a named method on a given date, gives the next reader enough to judge what the number can bear. A record that says only "not detected" leaves that reader to reconstruct a decision the original author already made, often without the information needed to check it.