DISCUSSION 



This report examined finfish data from the impingement, plankton, 

 trawl and seine monitoring programs at MNPS . Of the numerous species 

 recorded, 10 were discussed in detail. Winter flounder was discussed 

 extensively in the the Winter Flounder Population Studies section and 

 the remaining nine (American sand lance, anchovies, sticklebacks, 

 silversides, Atlantic tomcod, grubby, windowpane, tautog and cunner) 

 were discussed here. All taxa except anchovies were considered 

 permanent residents that make limited movements or have some behavioral 

 pattern that makes them unavailable to capture during certain times. 

 Anchovies were seasonal residents in the greater Millstone Bight, and 

 were sporadically present in monitoring collections. Young sand lance, 

 sticklebacks and silversides, use the shore zone in summer and other 

 nearby habitats in colder seasons. Cunner and tautog are both 

 year-round residents that become torpid in winter and are less 

 susceptible to collection. 



In this report, we tried a new approach for assessing impacts. It 

 involved building harmonic regression models that described the observed 

 patterns of finfish catch fluctuations. All models had deterministic 

 components. Besides time (used to describe periodicities through 

 harmonic terms) , season and flow were found to be useful predictor 

 variables that determined the basic shape of the predicted curves. Some 

 models also included stochastic autoregressive terms which modified the 

 basic shape of the curve to account for autocorrelated deviations. 

 Previous modeling efforts used only stochastic components (NUSCo 1983a) 

 and produced less reliable descriptions of fluctuation patterns in the 

 forecast year. 



The reliability (as measured by R^) of 41 out of the 56 models 

 produced from historical data was good (> 0.70). The models fitted to 

 plankton and impingement data provided more accurate descriptions of 

 historical conditions than those fitted to trawl data. The higher R^ 

 values were associated with models that described very seasonal patterns 

 of occurrence; e.g. the times of the year when eggs and larvae were 

 present near MNPS were very well defined. When season interacted with 

 the sine and cosine terms in a multiplicative sense, the models were 



46 



