What the interval represents

A reported concentration is not an exact number without limits. Uncertainty combines contributions from the procedure and may be expressed as combined or expanded uncertainty. Where expanded uncertainty is reported, the coverage factor and corresponding interpretation must be stated. Adding a percentage without explaining its origin or included operations is insufficient.

Where it comes from

Air volume, collection, storage, analyte recovery, calibration and analytical response may contribute. Each element’s relevance varies by method. Laboratory-provided uncertainty may cover analysis only: whether it also incorporates sampling must be checked before assigning it to the entire measurement.

It is not between-day variability

Differences in production, tasks or people produce real variability in occupational exposure. That variability does not disappear with a more precise instrument. Selecting days and exposure groups belongs to assessment strategy; metrological uncertainty describes another part of the problem and does not replace representativeness.

Using it near a limit

Comparison must apply the relevant procedure and decision rule, documented before seeking a favourable result. It is not appropriate always to subtract uncertainty from concentration to declare compliance. Where a result is close to a criterion, better information, wider assessment or stronger controls may be needed according to the situation.

An example distinguishing concepts

In a hypothetical example, a laboratory improves analytical precision, but samples are still collected only on low-production days. Results may have lower analytical uncertainty while remaining unrepresentative of normal work. The needed improvement includes reviewing the survey, not merely choosing equipment with higher resolution.

What to request in the report

The scope of uncertainty, units, coverage factor, method and limitations should be identified. Retaining these conditions enables trend interpretation and supplier comparison without confusing figures calculated differently. A known bias or incident is investigated and managed according to the method; broad uncertainty does not automatically make a sample valid.