forced to predict during periods of non-occurrence; the harmonic terms 

 described the curves otherwise. Among the trawl models, low R^ values 

 were associated with descriptions of species that were present 

 throughout the year, and had no seasonal pattern of occurrence. The 

 relationship between cooling-water-flow rate and impingement is well 

 establish in the literature (Murarka 1977; NUSCo 1981). Thus, including 

 flow as a multiplicative variable forced the impingement models to 

 predict low counts during plant shutdowns, but interacted with the 

 harmonic terms to allow for seasonality during normal operations. While 

 storms also influence daily impingement counts (NUSCo 1980) , these 

 affects were minimized by taking a weekly average of daily counts. Some 

 models included harmonic terms that described 5- or 6-yr cycles. 

 Frequently, these models described the 1983 pattern of occurrence well, 

 but some actual data were outside the 95% confidence interval. Also, 

 our data base was too short (six years) for us to observe a repeatable 

 five or six year pattern. Thus these harmonic terms described the 

 existing data reasonably well, but the period will probably change as 

 more data are added to the series. 



The results of the forecasting procedure produced bounds (95% 

 confidence limits) for what the 1983 abundances would be if they 

 followed historical patterns. Most of the 1983 catches did not deviate 

 greatly from what was predicted based on six years of data and therefore 

 had not been affected by power plant operations. Of those species whose 

 abundances fell outside the confidence interval, only tomcod fell below 

 the lower limit; apparently the 1983 tomcod peak abundance was delayed 

 somewhat. The 1983 data for anchovies, windowpane and cunner were above 

 the upper limit forecasted for these species. These discrepencies could 

 be the result of insufficient data (e.g. the length of the available 

 series was shorter than some natuarally occurring fluctuations) or 

 unusal climatic events (warm winter) , but could not be attributed to 

 power plant operations. 



The models resulting from the harmonic regression and 

 autoregression techniques used here for the first time provided the 

 necessary framework for future applications such as impact assessment 

 under 3 unit operation. As we gain experience using these techniques 

 and as more data become available, we will be able to provide even 



47 



