In all, 32 variables were used during initial mul- 

 tiple regression analyses. These included 2 sedi- 

 mentary parameters, 2 seasonal/reproductive com- 

 ponents and 7 climatic variables, each of which 

 had 4 values representing daily and quarterly high 

 and low extremes. 



Model Selection Procedure 



The quarterly mean values were first detrended 

 using a polynomial regression equation to create 

 the residuals needed for regression analysis. If no 

 significant long-term trend was evident, residuals 

 were created by subtracting the quarterly mean 

 from the 1 1 year mean. A step-wise multiple 

 regression was then used on these residuals to 

 select explanatory abiotic variables and combina- 

 tions of variables that were significantly different 

 from zero, at a probability level of p < 0.05. 1 his 

 probability was deemed sufficient to guard against 

 fitting more parameters than can be reliably esti- 

 mated, given the sample size. The model that 

 minimized the mean square error and maximized 

 the R was selected as best model describing 

 observed variability in community abundance, 

 numbers of species and abundances of selected 

 species. 



Analyses of covariance were then conducted to 

 test for annual differences in abundance and spe- 

 cies numbers using significant explanatory vari- 

 ables as covariates. Results of these analyses and 

 pair-wise t-tests on adjusted means (least square 

 means) were used to identify significant (p < 0.05) 

 interannual differences in the abundance and spe- 

 cies number. In this report, pair-wise differences 

 among years involving only 1986 and 1987 will 

 be emphasized. 



Biological Index Value 



The Biological Index Value (BIV) of McCloskey 

 (1970), an index of dominance, was calculated for 

 the 1 most abundant taxa at each statios ■ collected 

 from 1986-1987. To provide comparisons with 

 the pre-operational data, a BIV was also calculated 

 for 1980-85. To calculate the BIV, the top ten 

 numerically abundant species in each sampling 



year are ranked from highest to lowest and ranlcs 

 summed for each year. The BIV is sum of the 

 ranks across all years for each taxon expressed as 

 a percentage of a theoretical maximum sum that 

 occurs if a species ranked first in all sampling 

 years. For example, the BIV would be equal to 

 100% and the theoretical maximum equal to 60 

 when a species ranks first in abundance in each 

 of six years and a total of 10 species are collected. 



Species Diversity 



Species diversity at each station was calculated 

 using the Shannon information index: 



1^' = Z^l"g2 5 {Pielou 1977) 



where n \ = number of individuals of the i 

 species, N = total number of individuals for all 

 species and S = number of species. An evenness 

 component of diversity was calculated as: 



//' 



{Piehu 1977) 



where II max ~'og 2 ■'' ^^^ represents the theo- 

 retical maximum diversity when all species are 

 equally abundant. Evenness ranges from zero to 

 one and increases as the numbers of individuals 

 among species become more evenly distributed. 

 Diversity calculations excluded oligochaetes and 

 rhynchocoels (groups that sometimes accounted 

 for over 80% of the total organisms collected) 

 because they were not identified to species. Sim- 

 ilarly, other organisms that could not be identified 

 to species, either because they were juveniles or 

 in poor physical condition, were excluded from 

 this analysis. 



Numerical Classification and Cluster 

 Analyses 



(Cluster analyses, based on annual abundances 

 of organisms, were performed using the Bray- 

 Curtis similarity coefficient. This coefficient is cal- 



Benthic Infauna 



63 



