Most station and site summaries, and all analyses, used only species that were 

 considered part of the benthic infauna. Transient, water column, or terrestrial species were 

 excluded. 



Data analyses included multivariate cluster analyses and univariate Analysis of 

 Variance (ANOVA) techniques. The multivariate technique employed to examine community 

 differences consisted of classificatory procedures (Clifford and Stephenson, 1975) based on 

 multiple attributes (e.g., species composition). The primary assumption underlying this 

 approach was that optimal areas (habitats) for a particular species within an environment 

 were inhabited by greater abundances of that particular species. Areas with similar species 

 composition (in terms of both types and abundance) were assumed to provide similar 

 physical/chemical microenvironments. Conversely, areas that supported modified or 

 different assemblages of species were assumed to provide altered or different 

 microenvironments. 



Two classifications were performed in which entities were grouped by specific common 

 attributes. The sampling stations (entities) were grouped by similarities in species 

 composition (attributes). This is termed the "normal" analysis by Clifford and Stephenson 

 (1975). The "inverse" analysis grouped the species (entities) with respect to their 

 distribution among stations (attributes). Both analyses used only identifiable species that 

 were considered to be important in the benthic infauna. Taxa not used in the analyses 

 included species not part of the benthic infaunal community and rare taxa. A number of rare 

 taxa were found in this study. They carried little classification information (Boesch, 1977), 

 but they could mask much of the information carried by the more common species. To ensure 

 that they did not do so, taxa that did not occur in at least two replicates and at a minimum 

 of two stations were excluded from the analyses. 



The classification analyses involved three procedures. The first was a calculation of an 

 inter-entity similarity (distance) matrix using the Bray-Curtis index (Clifford and 

 Stephenson, 1975). In the normal analysis, abundance data were square root-transformed and 

 standardized by species mean. Next, the step-across procedure (Bradfield and Kenkel, 1987; 

 Williamson, 1978) was applied prior to application of the flexible sorting strategy (Lance 

 and Williams, 1967; Clifford and Stephenson, 1975). The third procedure was sorting, by 

 which the entities were clustered into a hierarchical dendrogram. Dendrograms from both 

 the normal and inverse analyses were combined into a two-way coincidence table (Clifford 

 and Stephenson, 1975). The values of relative abundance of each species were replaced by 

 symbols (Smith, 1976) in the body of the two-way table as an aid to presenting the patterns 

 of species distributions. 



ANOVA techniques were applied to dominant species, taxonomic groups, and community 

 parameters to support statistically (e.g., with probability levels and confidence limits) the 

 community differences that were found. Station and site differences were assessed using the 

 Student-Newman-Keuls (S-N-K) test (Sokal and Rohlf, 1969). 



Differences between sites were tested by contrasts on site means. The site means were 

 calculated by averaging the means of the station within each site. Since six contrasts were 

 performed, the a error rate for each contrast was adjusted, using Bonferroni's equation, to 

 (1=0.05/6=0.008. 



The ANOVA and S-N-K tests were applied to: 1) community parameters (including 

 total abundance), number of species, diversity (H), evenness (J), and dominance (I-J); 2) 

 abundance and biomass in five major taxonomic groups, and total biomass; and 3) abundance 

 data for the five numerically dominant species at each station and site. Tests of abundance 

 used log (x+1) transformed data. 



18 



