SEDIMENT CONTAMINANTS AND BIOLOGICAL EFFECTS IN SOUTHERN CALIFORNIA -USE OF A MULTIVARIATE STATISTICAL APPROACH TO ASSESS BIOLOGICAL IMPACT

This study attempts to predict biological toxicity and benthic communi ty impact in sediments collected from two southern California sites. C ontaminant concentrations and grain size were evaluated as predictors using a two-step multivariate approach. The first step used principal component analysis (PCA) to describe contamination type and magnitude present at each site. Four dominant PC vectors, explaining 88% of the total variance, each corresponded to a unique physical and/or chemical signature. The four PC vectors, in decreasing order of importance, we re: (1) high molecular weight polynuclear aromatic hydrocarbons (PAH), most likely from combusted or weathered petroleum; (2) low molecular weight alkylated PAH, primarily from weathered fuel product; (3) low m olecular weight nonalkylated PAH, indicating a fresh petroleum-related origin; and (4) fine-grained sediments and metals. The second step us ed stepwise regression analysis to predict individual biological effec ts (dependent) variables using the four PC vectors as independent vari ables. Results showed that sediment grain size alone was the best pred ictor of amphipod mortality. Contaminant vectors showed discrete depos itional areas independent of grain size. Neither contaminant concentra tions nor PCA vectors were good predictors of biological effects, most likely due to the low concentrations in sediments.