49 



MDS analysis, BIOENV was used to overlay specific environmental data over the MDS 

 results to qualitatively evaluate other environmental factors contributing to the differences 

 in the datasets (Clark and Warwick 1994). 



Clustering independently determines the similarities between sample data points 

 based on multiple variables, and then groups them accordingly. The cluster program 

 standardized the data, so the relative abundance of each species was used for analysis. The 

 abundance data were transformed by the fourth square root to minimize the dominance of 

 the very abundant species, so the rare species also contributed to determining the similarity 

 between samples. For each analysis, the Bray-Curtis similarity index was calculated and 

 used to create a similarity matrix. Hierarchical agglomerate clustering with group-average 

 linking was performed on the matrix for each dataset. The results were displayed in a 

 dendrogram showing station groupings on the basis of Bray-Curtis similarity in the 

 mineralogy composition or the micro fossil community structure. 



Non-metric MDS provides an ordination, or map, of samples showing the inter- 

 relationships between samples on a continuous scale. The MDS method compares the 

 extent to which the data groups determined by clustering are similar. MDS ordination was 

 performed on the similarity matrixes of the mineralogy data and the fourth square root 

 transformed microfossil data as in clustering analysis. The plots were constructed by an 

 iterative procedure, however, which successively refined the positions of the samples to 

 reflect the similarity relations between them. 



The BIOENV module of PRIMER was used to overlay various environmental 

 datasets, including grain size data and microfossil densities (number of individuals per 

 gram of picked material). These data groupings were overlaid on the MDS ordination 

 plots. The additional variables were represented as symbols of differing size, determined 

 by a simple linear function of the selected variable, and superimposed on the 2-dimensional 

 MDS ordination of the microfossil data. 



3.10.3 Analysis of Similarities 



The samples collected in the cores following the postcap survey were classified as 

 cap material (CDM), dredged material (pseudo-UDM), and ambient material, based on 

 both the visual appearance (Section 3.7.2) and microfossil analysis (Section 3.9.4). The 

 statistical strength of the differences between these pre-determined groups were evaluated 

 using the ANOSIM (analysis of similarities) randomization test, applied to test for the 

 statistical significance of differences displayed in the microfossil assemblages of the CDM 

 and pseudo-UDM samples. ANOSIM is based on a non-parametric permutation procedure 

 applied to the rank similarity matrix, described previously, which underlies the ordination 

 of samples (Clarke and Warwick 1994). The procedure is analogous to standard 

 parametric analysis of variance (ANOVA). 



The Portland Disposal Site Capping Demonstration Project, 1995-1997 



