You are here:
Statistical considerations in estimating organism concentrations in ballast water discharges
Citation:
FRAZIER, M. R., H. LEE, II, AND D. REUSSER. Statistical considerations in estimating organism concentrations in ballast water discharges. Presented at Sixth International Conference on Maraine Bioinvasions, Portland, OR, August 24 - 27, 2009.
Impact/Purpose:
Sampling probabilities may affect the practical use of different ballast water performance standards which establish the acceptable concentration of organisms in ballast discharges.
Description:
Sampling probabilities may affect the practical use of different ballast water performance standards which establish the acceptable concentration of organisms in ballast discharges. The International Maritime Organization (IMO) has initiated a ballast water standard of <10 viable organisms m-3 for organisms >50 µm. More stringent standards have been considered by others, such as the <0.1 organisms m-3 in U.S. Senate bills 363/1224 to the <0.001 organisms m-3 based on an analysis of natural invasion rates. To accurately estimate the concentrations of organisms at these low levels it is critical to consider issues of sampling probability. For example, given a performance standard of <0.1 organisms m-3 and a true concentration of 0.2 m-3 (two times the proposed standard), about 20 m3 of ballast water must be sampled to have a 90% probability of detecting this exceedence. Given a performance standard of <0.001 organisms m-3 and a true concentration of 0.002 m-3, nearly 2000 m3 of ballast water must be sampled. It is also possible to fail when the true density is below the performance standard (i.e., false positive). For example, given the IMO performance standard of <10 organisms m-3 and a 1 m3 sampling volume, a ship with a true density of 7 organisms m-3 will appear to exceed the performance standard about 10% of the time. For these calculations, organisms are assumed to be randomly distributed in the ballast water (i.e., a Poisson distribution). If organisms are sufficiently aggregated then estimating densities will require sampling larger quantities of water. In the example with a performance standard of 0.1 organisms m-3, by adding an aggregation component (dispersion parameter of 0.5) to the model, the quantity of water sampled must increase from 20 to about 30 m3 to maintain the same probability of detection. Determining the distribution of organisms in ballast water to assess the extent of aggregation will be an important step for developing protocols to accurately estimate the concentration of organisms