biweekly trawl and weekly entrainment and impingement data were 

 determined using spectral analysis (PROC SPECTRA, SAS 1982). Because 

 the seine, offshore fish larvae and all the fish egg data were collected 

 at unequal intervals, spectral analyses were not done on these data. 

 Fourth, stepwise analyses were used to find the best set of variables 

 (predictors) for each mathematical model. 



Variables considered as potentially good predictors were time, 

 water temperature, deviations from normal water temperatures, barometric 

 pressure, species abundance at other stations, zooplankton abundance' 

 season and flow (water volume entrained by the cooling system) . Of 

 these, time, season, and flow were the most useful. The latter was only 

 considered for the impingement models. The variable "season" was a 

 dummy variable that was set equal to 1 for those months corresponding to 

 when at least 98% of the total annual abundance occurred for each 

 species; it was set to during other times (see Table A). The variable 

 "time" always appeared in the argument of the sine and cosine functions. 

 The actual argument of these trigonometric functions is the time (in 

 days) expressed as radians scaled for the period of the cycle being 

 described by each harmonic component (see Bliss (1958) and Lorda (1983) 

 for more details) . The following periods (multiples or even fractions 

 of a basic period of one year) were initially considered for possible 

 harmonic components: 



6 year cycle 6 months 



5 year 4 months 



4 year 3 months 



3 year 2 months 

 2 year 

 1 year 



Dummy predictor variables that represented interactions were created by 

 multiplying the values of the interacting variables. In the impingement 

 models, flow appeared as a multiplier of the other predictors and forced 

 a low impingement prediction whenever flow was low. Three general 

 classes of deterministic models were found most descriptive: 



