Key takeaway
A value accepted for a reporting purpose can have a different evidential meaning from a direct measurement of operating behavior.
Ask what the hourly value represents
An emissions export may contain a number for every operating hour. That completeness can be useful for reconstructing a report while concealing that some values were calculated when a monitor did not provide valid data. Before presenting the archive as measured plant behavior, retain how each value was obtained and which reporting version it belongs to. Filling all cells is not the same as recovering all measurements.
EPA’s 2021 Monitoring Insights explains that substitute data for missing or invalid Part 75 observations can become deliberately conservative as missingness increases. Its method-of-determination codes identify measured or calculated values. This historical primary explanation was read 11 October 2026. It supports a provenance distinction, not a current rule determination or an instruction to change a facility’s submitted values.
Compare two interpretations of the same file
This hypothetical example uses five equal operating hours of one concentration parameter in ppm. Three values are direct measurements: 80, 100 and 120. Two are illustrative substitute values, 300 and 400. They are invented to expose the interpretation problem, not outputs from an approved EPA algorithm. The archive retains one row per hour and labels the basis explicitly.
The all-row average is 1,000 / 5 = 200 ppm; the direct-measurement subset average is 300 / 3 = 100 ppm. Neither establishes the actual five-hour physical average. The measured subset may be selected by which hours the monitor worked, and the two larger substitutes were not direct observations. The useful conclusion is narrower: the populated series is 100% complete, while direct measurements cover 60% of these operating hours.
| Hour | Value, ppm | Provenance | Use in this review |
|---|---|---|---|
| H1 | 80 | Direct measurement | Measured subset |
| H2 | 100 | Direct measurement | Measured subset |
| H3 | 120 | Direct measurement | Measured subset |
| H4 | 300 | Illustrative substitute | Retain reporting value; flag |
| H5 | 400 | Illustrative substitute | Retain reporting value; flag |
Separate a backup measurement from an estimate
EPA’s Plain English Guide distinguishes quality-assured backup-monitor or reference-method information from substitute values when primary quality-assured data are unavailable. These are different evidence paths. An archive label saying primary monitor unavailable does not, by itself, establish that the replacement number was estimated. Recover the actual method and validity evidence rather than deriving it from a generic outage flag.
Create a parameter-specific dictionary that preserves original method code, source monitor or calculation reference, quality state, operating-hour basis, unit and any later correction. Keep the historical code dictionary and algorithm version when interpreting old records. The guide contains historical program references and is used here only for that evidence distinction. An operator must verify the currently applicable rules, approvals and reporting instructions for a real compliance question.
Choose the evaluation use before filtering
A recipient testing retrieval of reporting explanations might need measured and substitute hours together. A recipient modeling physical operating behavior may need separate strata, missingness information and a justified selection rule. Neither use authorizes silently deleting high values or replacing a reporting substitute with a convenient interpolation. State the question and preserve a trace from any evaluation view to the source reporting history.
Review what happens around monitor failures, maintenance and quality-test boundaries. Are operating hours still identifiable? Do several parameters have different validity states during the same hour? Did a later correction replace the original report? A single row-wide clean flag may erase those differences. The review should show both the raw provenance state and the proposed task-specific inclusion decision, with an owner responsible for unresolved method codes.
Describe coverage with its basis
For the fictional file, the completed decision permits a description of five populated concentration hours, including three direct measurements and two illustrative substitutes. It does not permit a claim that average actual emissions doubled or that two hours exceeded a real limit. Concentration alone is also not mass emissions; flow and the applicable calculation would need their own review. No physical total is inferred from the table.
The internal deliverable is a coverage register by parameter and provenance, alongside the versioned reporting series. Inventory records those categories; Readiness assigns missing code dictionaries or quality-state evidence. Keep facility identifiers, operating details and compliance records in their authorized environment. A separately reviewed external sample may have a narrower purpose than the original report, and a complete data file supplies no licensing demand or price evidence.