U.S. Mussel Watch Data from 1986 to 1994:  Temporal Trend Detection at Large Spatial Scales

Abstract

National Mussel Watch Programs collect bivalves at many sites along their respective coastlines in order to assess chemical contamination of coastal waters. This paper aims at detecting temporal trends at large spatial scales using contaminant data from 1986 to 1994 of the U.S. National Status and Trends Mussel Watch Program. U.S. coasts are divided into 10 large areas. After a logarithmic transformation of the data, linear models are used and include a ‘sampling site' effect, a linear trend estimated for each area, weather covariates, and an ‘analysis' effect reflecting changes of analytical procedures or even laboratories. Except for the Californian coast, chlorinated organic compounds show rather uniform decreases, which can be attributed to bans put on these contaminants. A few increasing trends were detected for trace elements, which also showed few decreases. Additions of weather covariates and possible analysis effects were intended to model apparently random between-year variations in contaminant time−series. Although low in terms of percent age of variability explained, the addition of these effects might have avoided the detection of spurious trends for some contaminants.