Measurement uncertainty in determining concentrations of living organisms

Ballast water management systems (BWMS) are shipboard devices designed to treat ballast water such that concentrations of living or viable organisms in discharged water are sufficiently low to meet state, national, or international limits. At land-based, purpose-built test facilities, BWMS undergo a rigorous verification process to confirm their efficacy. Currently, concentrations of living organisms in treated water are determined by manual microscope counts. We considered sources of uncertainty associated with these measurements and performed empirical estimates of variability in samples representative of the initial, uptake water (with correspondingly high concentrations of organisms) and in samples representative of treated water (with low concentrations of organisms, near the discharge limits for organisms ≥ 50 µm and organisms ≥ 10 and < 50 µm in minimum dimension). The sources of uncertainty included both systematic error—here, the loss of organisms during sampling—and random error—represented by variability in organisms counts among analysts and among replicate subsamples, as well as variability across measurements of sample volumes. Systematic error for organisms ≥ 50 µm was measured by the recovery rates of 150-µm diameter microbeads in field-scale trials. This error was substantial: recovery rates were 76 ± 11%; n = 18 replicate trials. The greatest source of random error was due to variation in counts of organisms among replicate subsamples, with average coefficients of variation (CV, %) ranging from 10% to 37% for organisms ≥ 50 µm. The variations among analysts’ counts of organisms and volumetric measurements, however, were negligible, approximately one order of magnitude smaller. Likewise, pipetting errors were unimportant.