O'BRIEN AND MAYO: CPUE OF YELLOWTAIL FLOUNDER 



highly significant differences in only 50, 35, and 

 IdVc of the years for Georges Bank, Southern New 

 England, and Cape Cod grounds, respectively 

 (Table 4), while the tonnage class-depth interac- 

 tion was highly significant in no more than 20% 

 of the years on each of the three grounds. 



Interaction between tonnage class and statis- 

 tical area throughout the region was highly sig- 

 nificant in all cases. Further analyses yielded 

 highly significant differences in CPUE among the 

 three stocks and highly significant tonnage class- 

 stock interactions which suggests that relative 

 fishing power among vessel classes is not consis- 

 tent from stock to stock, implying a need for com- 

 puting a separate set of fishing power coefficients 

 for each stock. Within each stock, differences in 

 CPUE among statistical areas and tonnage 

 classes were also highly significant in most cases, 

 although the tonnage class-area interaction was 

 not. 



Standardized CPUE 



Annual fishing power coefficients, obtained by 

 retransforming linear model parameter esti- 

 mates for each tonnage class, are presented in 

 Table 5 by stock. Cursory examination of the coef- 

 ficients reveals distinct trends throughout the 20- 

 yr series. On Georges Bank and Southern New 

 England grounds, fishing power coefficients for 

 the smaller vessels (TC 23-25) relative to the 

 standard declined over time, whereas coefficients 

 for the larger vessels increased over time. On 

 Cape Cod grounds the coefficients increased for 

 the smaller vessels (TC 23 and 24) although 

 trends were less pronounced. These trends are 

 illustrated graphically by plotting annual devia- 

 tions from the 20-yr average in Figures 7-9. A 

 Durbin-Watson test for first order autocorrela- 

 tion of the annual deviations (Neter and Wasser- 

 man 1974) was significant for most tonnage 

 classes within each of the stocks, suggesting the 

 presence of a substantial tonnage class-time in- 

 teraction. 



The three-way linear model, modified to in- 

 clude interaction terms, also revealed highly sig- 

 nificant tonnage class-year as well as tonnage 

 class-season and year-season interactions within 

 each of the three stocks (Table 6). When interac- 

 tions are significant, they can be examined in 

 detail or absorbed in the error term when testing 

 for main effects. Since tonnage class effects have 

 already been examined on an annual basis, the 

 interaction terms were excluded from the three- 



way model used to obtain parameter estimates for 

 tonnage class, season, and year. The model is 

 specified as follows: 



f/ = a + X [Pi, Xy + ^2j X2j + Paj Xsj] + 



where ^y, P2/, P3; = model parameter estimates 



in logarithmic units for 

 category j for tonnage 

 class, season, and year, 



Xij , X2J, X2,j 



dummy variables for ton- 

 nage class, season, and 

 year ( = 1 when category 7 

 occurs; = otherwise). 



ANOVA results obtained without interaction 

 are presented in Table 6 for each of the three 

 stocks. For Georges Bank and Southern New 

 England stocks, year accounts for the greatest 

 reduction in error sums of squares; on Cape Cod 

 grounds tonnage class accounts for the greatest 

 overall reduction. 



Coefficients for tonnage class, year, and season, 

 derived from model parameter estimates for the 

 combined 1964-83 period are presented in 

 Table 7. Tonnage class coefficients for Georges 

 Bank and Southern New England are relatively 

 homogeneous, as compared with those obtained 

 for Cape Cod grounds, owing to the narrower 

 range of vessel tonnage classes which have con- 

 sistently exploited these fisheries. Seasonal coef- 

 ficients exhibit the same pattern on Georges 

 Bank and Southern New England with the 

 highest catch rates occurring during the third 

 quarter; on Cape Cod grounds the highest catch 

 rates occur during the second quarter. Trends in 

 annual coefficients are similar on all three 

 grounds. Standardized CPUE indices based on 

 the annual coefficients are illustrated in Figure 

 10, and traditional indices based on the methods 

 of Lux (1964), as given by Clark et al. (1984), are 

 also presented for comparative purposes. 



Although each series indicates similar trends, 

 CPUE indices obtained from the linear model for 

 Georges Bank and Southern New England have 

 remained slightly higher than the traditional in- 

 dices since the early 1970's. Prior to this, the tra- 

 ditional CPUE indices were greater than the re- 

 vised indices. On Cape Cod grounds, differences 

 between the two series are considerably greater, 

 particularly in the early years. 



101 



