trawl models only NR had an R"^ exceeding 0.50. The forecast errors for 

 all the trawl models were high, and the actual 1983 data for trawls fell 

 within the 95% confidence interval for only the NB model. 



Catches of windowpane in both trawls and impingement were at a 6-yr 

 high in 1983. This peak could be accounted for in several ways. First, 

 the windowpane population may have a cycle of abundance longer than 6 

 years (the length of MNPS database). Secondly, unusually warm winter 

 water temperatures in 1983 (Fig. 22) may have made it unnecessary for 



21- 



Figure 22. Predicted ( ) and actual weekly means (+) of daily intake 



water temperatures. The equation for the predicted values is 

 T=11.13-6.47*cos(time * 365 .25/12)-6.46*sin(time * 365.25/12) 

 with r2 = 0.97. 



the windowpane to migrate offshore to avoid cold inshore temperatures as 

 usual (Austin 1973) . Model errors for trawl-caught windowpane were 

 high, although impingement models seemed to accurately describe the 

 pattern of windowpane fluctuations. Actual 1983 data exceeded the upper 

 95% confidence interval for both trawls and impingement models. More 

 years of data may be needed to accurately describe catch fluctuations 

 for this species. 



38 



