Using smoothers for comprehensive assessments of contaminant time series in marine biota

Abstract

It is shown how locally weighted regression smoothers can be used in comprehensive assessments of contaminant time series in marine biota. The smoothers are used to make formal statistical tests of linear and non-linear temporal trends, to predict future contaminant levels, and to compare current and predicted levels with environmental reference values. The smoothers also provide graphical summaries that are easy to interpret, and simultaneously characterize several features of a contaminant time series. The methodology is demonstrated using time series of γHCH and dieldrin levels in fish liver. Finally, it is shown that smoothers have reasonable power for detecting a range of temporal trends, when compared with other methods of trend detection.