Reducing and quantifying uncertainty for pollution estimates calculated by modelling replicated benthic count data

The main contribution in this work is an improvement of a method for quantifying the level of pollution using benthic count data. The suggested improvement utilises the information found in replicated samples in order to obtain more precise estimates of the pollution. Simultaneously, the uncertainties of the pollution estimates are quantified. This is obtained by making pseudo samples by randomly permuting the counts of the species observed within a group of replicates. The pseudo samples are representative for the variation at the sampling site since the permutations are both species specific and restricted to replicates within the sampling site. The pollution is quantified by a modelling approach. Sediment samples are collected in an area surrounding a contamination source, e.g. an oilrig. The non‐polluted samples are used to build a linear model of the natural variation in the survey area. The remaining samples are projected onto the calculated model. Then, the distance from the projection to the sample gives an estimate of the pollution in the current sample. The last part of the work is devoted to investigate the robustness of the suggested improvements to pollution quantification from benthic count data. This is done by means of perturbations, and it appears that the modifications give both reliable and robust results. Copyright © 2002 John Wiley & Sons, Ltd.