sample size if necessary) , of a particular taxon and dividing by the 

 total number of fish caught. 



Totals of fish impinged were determined by summing the counts on 

 days sampled and the estimated numbers on days not sampled. The 

 estimates for days not sampled were calculated by dividing the sum of 

 the fish collected in a month by the sum of the condenser cooling 

 water-flow on collection days to determine a monthly impingement rate 

 (n/m^) . This rate was then multiplied by the flow during non-sampling 

 days in each month to obtain estimates for those days. The actual and 

 estimated daily totals were summed to provide monthly estimates. Annual 

 totals were the sum of the estimated monthly totals. 



Since previous studies (NUSCo 1982a, 1983a) have shown that the 

 trawl, seine, plankton and impingement data came from highly skewed, 

 non-normal distributions, all these data were log-transformed: 



Y .=ln(catch*c + 1) for trawls and seines 

 sti 



Y .=ln(count*c + 1) for impingement 



Y .=ln(density + 1) for ichthyoplankton 



where 's' refers to a sampling station, 't' designates the sampling 

 period, and 'i' indicates a sample within the sampling period. The 

 constants, 'c', are multipliers (1, 10 or 100) to insure that 

 incrementing the catch or count by 1 added less than 0.01%. When 

 replicates were taken in a sampling period, the transformed data were 

 averaged so that a single value was available for each sampling period 

 and station combination. The sampling periods were weekly for 

 impingement and plankton samples, biweekly for trawl samples and monthly 

 for seine samples. 



The general approach used to build the mathematical models for the 

 observed fluctuations included the following steps. First, abundant 

 species were selected and the data for these species were zero-filled. 

 That is, a zero was recorded for every sampling period-station-replicate 

 combination in which no individuals of a selected species were found. 

 Second, the data on the selected species were limited to those stations 

 where the species was most abundant. Third, the data were 

 log-transformed as described above and the dominant periods for the 



