At Forty Samples a Month, Two Coliform Positives Stop Triggering an Assessment

A distribution system logs two total coliform-positive samples in a single month. The laboratory work is not in dispute and every required repeat sample was collected. Whether those two positives start a Level 1 assessment turns on a number that has nothing to do with the water: how many routine samples the system took that month. At 30 samples the two positives are a trigger. At 40 samples they are not.

That is not a rounding artifact. It is the shape of the paragraph. The Level 1 trigger in the Revised Total Coliform Rule is written as two alternative tests, one that counts positives and one that measures a percentage, and the paragraph picks between them using the sample count itself. The seam between the two tests falls at 40 samples per month, and the arithmetic on either side of that seam moves in opposite directions.

One Paragraph, Three Branches

The triggers sit in eCFR, 40 CFR 141.859, Coliform treatment technique triggers and assessment requirements. Paragraph (a) opens by tying the whole apparatus to the two trigger lists: "Systems must conduct assessments in accordance with paragraph (b) of this section after exceeding treatment technique triggers in paragraphs (a)(1) and (a)(2) of this section."

Paragraph (a)(1), the Level 1 list, has three subparagraphs, and each of the three has to be counted before any conclusion about a given month. The first is the percentage test: "For systems taking 40 or more samples per month, the system exceeds 5.0% total coliform-positive samples for the month." The second is the counting test: "For systems taking fewer than 40 samples per month, the system has two or more total coliform-positive samples in the same month." The third is procedural rather than numeric: "The system fails to take every required repeat sample after any single total coliform-positive sample."

Read the first two together and the seam is explicit. Subparagraph (a)(1)(i) governs a month in which 40 or more samples were taken. Subparagraph (a)(1)(ii) governs a month in which fewer than 40 were taken. The two are keyed to samples taken, not to population, not to system size, and not to the number of samples the system was required to take. That distinction does real work later.

Where the Sample Count Comes From

The sample count is not a choice. For systems above a thousand people it comes from a table in eCFR, 40 CFR 141.857, Routine monitoring requirements for public water systems serving more than 1,000 people. The section applies where "The provisions of this section apply to public water systems serving more than 1,000 persons," and the frequency paragraph says that "The monitoring frequency for total coliforms is based on the population served by the system, as follows:".

Before reading a value out of that table, read what its second column measures. The caption is "Total Coliform Monitoring Frequency for Public Water Systems Serving More Than 1,000 People" and the two column headings are "Population served" and "Minimum number of samples per month". The second column is a floor on collection effort. It is not a count of positives and it is not a percentage. The trigger paragraph then reads a different quantity, the samples actually taken, and the two quantities coincide only when a system collects exactly its minimum.

Four rows of that table bracket the seam. Population 25,001 to 33,000 carries a minimum of 30 samples per month. Population 33,001 to 41,000 carries 40. Population 41,001 to 50,000 carries 50. Population 50,001 to 59,000 carries 60. So a utility serving 33,000 people that collects exactly its minimum takes 30 samples and lands in the counting branch, while one serving 33,001 takes 40 and lands in the percentage branch. One additional resident moves the system across the seam.

The Arithmetic at the Seam

Take the two positives through both branches. At 30 samples, subparagraph (a)(1)(ii) applies, the test is two or more, and two positives reach it. At 40 samples, subparagraph (a)(1)(i) applies and the test is whether the month exceeds 5.0 percent. Two positives out of 40 samples is 5.0 percent exactly. Exactly 5.0 percent does not exceed 5.0 percent, so the month passes. The smallest count that clears the percentage test at 40 samples is 3, which is 7.5 percent.

That produces a gradient that bends backward at the seam. In the counting branch the implied positive rate falls as the sample count rises, because the threshold stays at two positives: 2 of 30 is 6.7 percent, and 2 of 39 is 5.1 percent. Cross into the percentage branch at 40 samples and the required rate jumps to 7.5 percent. It then falls again as the denominator grows, because the smallest qualifying integer count is the floor of five percent of the sample total plus one: 3 of 50 is 6.0 percent, and 4 of 60 is 6.7 percent.

The row at 40 samples per month is therefore the least sensitive row in this range. A system serving 39,000 people needs a worse month, in percentage terms, than either neighbor on the table, the 30-sample row below it or the 50-sample row above. Smaller rows bite harder again: two positives out of 25 samples is 8.0 percent. A reader who has followed the ninetieth-percentile arithmetic in how a water system can pass its lead test while individual taps fail will recognize the pattern: the rule sets a rate, the sampling plan sets a denominator, and the denominator decides which rates are even reachable.

The seam also has a path that runs the other way. Because (a)(1)(ii) is keyed to samples taken, a system whose minimum is 40 but that collects only 38 in a month is, on the face of the paragraph, a system taking fewer than 40 samples that month, which is the two-positive branch. Collecting fewer samples than required is separately addressed. In eCFR, 40 CFR 141.860, Violations, paragraph (c)(1) reads: "Failure to take every required routine or additional routine sample in a compliance period is a monitoring violation." So the shortfall carries its own consequence and moves the month into the stricter branch at the same time.

The Chain

A trigger is a starting gun, not a finding. Section 141.860 separates the two with care. Paragraph (b)(1) reads: "A treatment technique violation occurs when a system exceeds a treatment technique trigger specified in § 141.859(a) and then fails to conduct the required assessment or corrective actions within the timeframe specified in § 141.859(b) and (c)." The violation is the second clause, not the first. Exceeding the trigger is the condition that opens the window; missing the window is what closes it as a violation.

The window is stated in days. For Level 1, "The system must submit the completed Level 1 assessment form to the State within 30 days after the system learns that it has exceeded a trigger." For Level 2 the sentence is nearly identical: "The system must submit a completed Level 2 assessment form to the State within 30 days after the system learns that it has exceeded a trigger." The same 30 days governs both levels, even though a Level 2 assessment is the heavier instrument. The clock runs from the day the system learns it has exceeded a trigger, not from the month the samples were collected, so laboratory turnaround sits upstream of the deadline rather than inside it.

Escalation to Level 2 has its own three branches in paragraph (a)(2). The first is an E. coli failure: "An E. coli MCL violation, as specified in § 141.860(a)." The second is repetition, and it carries a condition that changes the reading completely: "A second Level 1 trigger as defined in paragraph (a)(1) of this section, within a rolling 12-month period, unless the State has determined a likely reason that the samples that caused the first Level 1 treatment technique trigger were total coliform-positive and has established that the system has corrected the problem." The third covers the smallest systems: "For systems with approved annual monitoring, a Level 1 trigger in two consecutive years."

Two features of (a)(2)(ii) matter for anyone reading a compliance record. The window is a rolling 12-month period, not a calendar year, so a trigger in November and another the following September are inside it while a calendar-year table would show one in each row. And the escalation is conditional. If the State both found a likely reason for the first trigger and established that the system corrected the problem, the second trigger does not escalate. The same two events, the same twelve months, and the record reads as one Level 1 or as a Level 2 depending on whether a State determination was filed in between.

The E. coli route reaches a maximum contaminant level violation directly. Paragraph (a) of 141.860 states: "A system is in violation of the MCL for E. coli when any of the conditions identified in paragraphs (a)(1) through (a)(4) of this section occur." The four conditions are an E. coli-positive repeat after a total coliform-positive routine sample, a total coliform-positive repeat after an E. coli-positive routine sample, failure to take all required repeats after an E. coli-positive routine sample, and failure to test for E. coli when a repeat sample is total coliform-positive. Two of the four turn on a laboratory result; the other two are about work that was not done.

What the Trigger Replaced

The reason a month with two positives now opens an assessment rather than producing a violation is that the 2013 revision moved that month out of the violation column. The final rule, EPA, National Primary Drinking Water Regulations: Revisions to the Total Coliform Rule, 78 FR 10270, states in its summary that "Under the RTCR there is no longer a monthly maximum contaminant level (MCL) violation for multiple total coliform detections." The next sentence says what took its place: "Instead, the revisions require systems that have an indication of coliform contamination in the distribution system to assess the problem and take corrective action that may reduce cases of illnesses and deaths due to potential fecal contamination and waterborne pathogen exposure."

The agency's standing rule page, EPA, Revised Total Coliform Rule And Total Coliform Rule, puts the same change first in its own list: "The biggest change is no acute or monthly MCL violation for total coliform positive samples only." The page describes the replacement as "Setting a total coliform treatment technique (TT) requirement," and states that systems "must comply with the RTCR starting April 1, 2016, or an earlier state effective date." It also records that "Minor corrections to the final RTCR became effective on April 28, 2014."

The trigger routes a month into an investigation rather than into an enforcement count. The agency page states the obligation plainly: "PWSs are required to conduct a Level 1 or Level 2 assessment if conditions indicate they might be vulnerable to contamination," and "Any sanitary defect identified during a Level 1 or Level 2 assessment must be corrected by the PWS." That is why the seam at 40 samples deserves attention. It is not deciding whether a utility is penalized. It is deciding whether anybody goes and looks.

Where This Doesn't Apply

The 40-sample seam is reachable only by systems whose monitoring plan puts them near it. Subpart Y places smaller systems under separate monitoring sections, 40 CFR 141.854 through 141.856, which cover non-community and community systems serving a thousand or fewer people on ground water only, and subpart H systems serving a thousand or fewer. The highest frequency those three sections set is one sample a month, so for them the Level 1 test is the two-positive count and the percentage sentence does no work.

The counting branch is also where the smallest systems meet the annual-monitoring escalation in (a)(2)(iii) rather than the rolling 12-month one, so the same phrase, a second Level 1 trigger, is measured over months for one system and over years for another.

Seasonal systems carry a path of their own that has nothing to do with sample counts. Paragraph (b)(2) of 141.860 reads: "A treatment technique violation occurs when a seasonal system fails to complete a State-approved start-up procedure prior to serving water to the public." A seasonal system can reach a treatment technique violation without any positive sample at all.

Finally, the quoted sentences are the federal floor. Several of the determinations in them belong to the State, including the close-out determination in (a)(2)(ii) and the approval of annual monitoring in (a)(2)(iii). A primacy agency that has adopted more stringent trigger language will not produce the arithmetic above. Readers tracking how differently two jurisdictions can treat the same distribution-system event will find a companion case in how boil water advisories follow pressure rather than laboratory results.

What Would Invalidate This

The 2013 preamble's own reasoning for splitting the Level 1 trigger at 40 samples per month is unverified here. That discussion sits in section III.E of the preamble, and the rendered text of the final rule stops inside section III.D, so the phrases that would carry the reasoning, 40 or more samples and the 5.0 percent threshold, appear nowhere in what was read. If the preamble supplies a statistical rationale, such as a detection-power argument that makes 7.5 percent at 40 samples a deliberate choice, the backward bend described above is a design feature rather than a seam artifact.

One gap remains. 40 CFR 141.858, Repeat monitoring and E. coli requirements, was not read, so every statement above about repeat samples rests on the sentences quoted from 141.859(a)(1)(iii) and 141.860(a). Two others have closed. The 141.857 table runs from 2 samples a month at population 1,001 to 480 at 3,960,001 or more, and no row at or above 40 samples requires a higher positive rate than the 7.5 percent at 40. And paragraph (d) of 141.860, Reporting violations, carries three subparagraphs, none of which anything above relies on.

The practical check for an operator or an analyst is short. Find the month. Count the samples actually taken, not the minimum. Pick the branch from that count. Then read the result against the right test: two or more positives below 40 samples, more than 5.0 percent at 40 or above. If the count sits at exactly five percent of the sample total, the month is on the safe side of the word exceeds, and the next question is whether a required sample was missed, because that single omission moves the month into the other branch and carries a monitoring violation of its own.