282 



Fishery Bulletin 105(2) 



Table 1 



Selectivity model selection and estimated parameters of the best fitting model. The value of the Akaike Information Criterion 

 (AIC) is shown for each of four possible models described in Munro and Somerton (2001): 2-parameter logistic, 3-parameter 

 logistic, 4-parameter logistic, and cubic spline models. The estimated parameters of the best fitting model, that is, the one with 

 the lowest value of the AIC, are shown. Parameter notation (a Ay) is the same as in Equations 4-6 in the text. Although, for 

 arrowtooth flounder [Atheresthes stomias) the best fitting model is the cubic spline, the parameters of the best fitting parametric 

 model are also included for use at lengths >20 cm. 



AIC 



Species 



2-parameter 

 logistic 



3-parameter 

 logistic 



4-parameter 

 logistic 



Cubic 

 spline 



Estimated 

 parameters 



)' 



Arrowtooth Rounder (Atheresthes stomias) 6568.8 6502.0 6504.0 



Flathead sole (//;ppog/ossoirfes e/ossodo;; ) 1804.3 1794.7 1796.7 



Rex sole (Glyptocephalus zachirus) 1642.8 1594.0 1596.0 



Dover sole (Micros/omuspaci^cus) 1245.9 1247.9 1249.9 



6382.7 -2.65 0.156 0.966 



1863.4 -3.66 0.208 0.861 



1634.9 -7.45 0.376 0.873 



1256.6 2.27 -0.055 



Estimating whole-gear efficiency 



Estimates of trawl efficiency, by 1 cm length categories, 

 were obtained by substituting into Equation 3 the esti- 

 mates of mean W,, and W, for the standard bridle length 

 from the herding experiment, /e„ from the net efficiency 

 experiment, h from the herding experiment, and W^,, 

 from the bridle measurement experiment. Variance of £, 

 which was derived from Equation 2 by using the delta 

 method (Seber, 1973) and by assuming no covariance 

 between any of the parameters, was calculated as 



V(E) = 





^^1 V(h) + 



w. 



V(k„)+ 



vW,/, 



V(WJ + 



w. 



V(\^„.) + 



(10) 



*..(W,.+/!W„„) 



wj 



V(W, 



The variance variables V(W„), and V(W^y) were estimated 

 as the variance of these dimensions during the herding 

 experiment. Variance variables V(^„), V(/i), and V(W^„) 

 were calculated as described earlier. 



Results 



Net efficiency experiment 



All four species of flatfish were present in each of the 

 34 tows successfully completed. Total number and size 

 range of measured fish were the following: 9512; 5-84 

 cm TL (arrowtooth flounder); 1701; 6-45 cm TL (flathead 

 sole); 2142; 10-61 cm TL (rex sole); and 949; 30-57 cm 

 TL (Dover sole). 



Estimates of /c„ 



The best fitting model of /;■„ as a function of length dif- 

 fered among the four flatfish species. For arrowtooth 

 flounder, the best fitting model was the cubic spline 

 (Table 1, Fig. 2), primarily because of its flexibility to fit 

 the selectivity of small (<20 cm TL) fish. At larger sizes 

 (>20 cm TL), however, the best fitting parametric model 

 was the 3-parameter logistic model. The cubic spline 

 model fitted almost equally well (Fig. 2); therefore, we 

 have also included the parameter estimates of this model 

 in Table 1. For flathead sole and rex sole, the best fitting 

 selectivity model was the 3-parameter logistic with a 

 maximum capture probability substantially below that 

 for arrowtooth flounder (Table 1; Fig. 2), indicating that 

 the escapement beneath the footrope for these species is 

 substantial even at the largest sizes. For Dover sole, the 

 best fitting model was the 2-parameter logistic model, 

 with the parameters chosen such that the predicted 

 capture probability decreased monotonically, and nearly 

 linearly, over the observed length range of fish. 



Herding experiment 



Seventeen geographic blocks, each containing three 

 hauls, were successfully completed. For arrowtooth 

 flounder, 11,510 fish were measured and all 17 blocks had 

 nonzero catches at all bridle lengths. For the remain- 

 ing species, the statistics are as follows: flathead sole 

 (6632 measurements, 17 blocks), rex sole (620 measure- 

 ments, 13 blocks), and Dover sole (392 measurements, 

 12 blocks). 



The width of the area contacted by the bridles (Vr„„) 

 increased dramatically with increasing bridle length 

 (Table 2), primarily because of the increase between 

 the short and standard bridle length. At the shortest 

 bridle length, the estimated value of W^,,, indicated 

 that the lower bridle was typically lifted off the bottom 

 along its entire length. Other aspects of trawl geometry 

 changed with increasing bridle length. Wingspread 



