Pragmatic estimation of uncertainty in normalized concentrations of contaminants in sediments.

No abstracts are to be cited without prior reference to the author.

International assessments of environmental contaminant concentrations commonly seek to combine data over time and between laboratories. It has long been recognized that significant differences can exist in the quality of the underlying analytical data. For example, performance might be better in more experienced laboratories and might have improved with time as instrumentation has developed. Concerns about poor quality data and the possible misleading impacts that such data might have on assessments have led to various approaches being taken to minimize the influence of poor quality data. Contaminant data held on the ICES Environmental Database must be accompanied by supporting analytical quality control (AQC) information, such as analyses of certified reference materials or results from laboratory performance studies. Some assessments have used these data to develop thresholds for minimum analytical quality and excluded data that failed to pass the thresholds. These procedures were effective, but were subsequently criticized for eliminating data that originators considered to be of good quality even though they were not supported by good AQC information. More recently, methods have been developed that incorporate data of different analytical quality through the fitting of weighted smoothers. Poor quality data are not eliminated, but have lower weighting in the analysis. These methods require an estimate of the analytical uncertainty of each data point. This paper describes the derivation of analytical uncertainties of normalized contaminant concentrations in sediment, using information from the OSPAR Co-ordinated Environmental Monitoring Programme (CEMP) held on the ICES Database in an assessment of temporal trends. The paper also describes how the uncertainties were used in the trend assessment and developments in the methodology that will be considered in the 2005 assessment.