81 



from artificial mixing of a sharp chemical 

 boundary. 



Statistical analyses were conducted to 

 document the presence or absence of a 

 sharp chemical boundary between sample 

 intervals (Appendix D). Initially a 

 principal components analysis (PCA) was 

 conducted to reduce the large number of 

 chemical analytes to a smaller, more 

 workable number of uncorrected variables 

 (PCA axes). Each PCA axis represents a 

 set of analytes which covary. Scores from 

 the PCA axes were then examined for the 

 presence of a sharp chemical boundary. 

 Regression analysis was conducted 

 between the PCA axes and a set of dummy 

 variables, each of which simulated a sharp 

 boundary at a different depth interval. A 

 single sharp chemical boundary was 

 considered present when the R 2 value for a 

 PCA axis and a dummy variable was close 

 to 1 . In addition, PCA axis scores for 

 each depth interval were plotted to confirm 

 the regression results. 



Three different principal components 

 analyses were conducted in this study. 

 The first was run on all of the samples and 

 included only those analytes sampled in all 

 cores (three metals and TRPH). The 

 second was run on nine of the cores which 

 were sampled for three metals, TRPH, and 

 PAHs. The third was conducted on only 

 three cores and included all of the 

 analytes. From these analyses, sharp 

 boundaries were found for some cores 

 using PCA axes generated from metal and 

 TRPH data. The results showed that 

 metals and TRPH are the best boundary 

 indicators for the cores examined. The 



frequency distribution of the PCA R 2 

 values resulted in a bimodal distribution of 

 values less than 0.5 and greater than 0.9. 

 This division is convenient for using the 

 PCA analyses to describe the relative 

 presence (>0.9) or absence (<0.5) of a 

 boundary between sample intervals. R 2 

 values for several of these PCA axes and 

 their dummy variables are presented in the 

 following discussion. 



Data from STNH-N, which had the 

 most visually obvious cap and mound 

 distinction, also resulted in well-correlated 

 statistical boundaries. Metal data from 

 STNH-N showed a dual concentration 

 pattern between relatively low and higher 

 values (Figure 4-5); the higher values were 

 within the Stamford ranges, although there 

 was quite a bit of overlap between 

 Stamford and New Haven metal 

 concentrations. R 2 values for the 

 cap/mound boundary documented in visual 

 core descriptions were 0.984 at CTR, 

 0.989 at 60E, and 0.915 at 40N. 



The only evidence for chemical 

 gradients at STNH-N was at 40W, where 

 the visual and chemical interface between 

 cap and mound was unclear. The increase 

 in contaminant values from 40 to 100 cm 

 in the 40W core was coincident with the 

 described variability in texture. This 

 coincidence was noted commonly in this 

 coring investigation and is discussed more 

 fully below. The greatest R 2 value 

 occurred at 80 cm (0.879), which indicated 

 that mound material was indeed found 

 below that point. Organic chemical data, 

 in general, agreed with the metal data at 

 STNH-N. An exception was a peak of 



Sediment Capping of Subaqueous Dredged Material Disposal Mounds 



