744 



Fishery Bulletin 102(4) 



Table 1 



Regression coefficients of wing spread (widthl and footrope distance from the bottom (footropel as a function of scope and gear 

 type based on tows from the capture efficiency experiment and tows using standard survey gear. Also provided are the results of 

 two-tailed Mests testing for the difference in intercept between gear types. 



Slope 



Intercept 



Experimental gear (n = 43) 



Standard gear (n = 33l 



P intercept 



Width (ml 

 Footrope (cm) 



0.0038 

 0.0087 



15.3206 

 -1.2582 



16.0713 

 -0.4241 



<0.0000 

 0.0023 



5 body orientation at the point of contact with the 

 footrope — footrope contact occurred along the body 

 axis (1) or from the side (2); 



6 crab size — small (0), medium (1), or large (2), where 

 size is expressed as an approximation based on 

 visual comparison of carapace length to the dimen- 

 sions of trawl parts, such as mesh or chain links. 

 Corresponding length intervals were approximately 

 <90 mm, 90-135 mm, and >135 mm. 



As a general rule, crabs could be seen in the videos 1-2 

 seconds prior to contact with the footrope. Assignment 

 of codes was typically straightforward. However, in some 

 instances, several reviews of the encounter were neces- 

 sary in order to determine a crab's position or orientation 

 in relation to the footrope. 



The probability of capture was estimated by using 

 stepwise generalized linear modeling (GLM; Venables 

 and Ripley, 1994) to fit a logistic model describing the 

 probability of capture as a function of crab size, where 

 body height, body orientation, average footrope distance 

 from the bottom during the tow, the use of artificial 

 light, and all possible first order interactions were con- 

 sidered as additional potential terms. The model fitting 

 procedure (with data from crabs for which all variables 

 were observed) entailed a stepwise backward model se- 

 lection process. The process began with fitting the model 

 to all interaction terms less one, and then calculating 

 and comparing the resulting AIC values. The interaction 

 producing the largest decrease in AIC was subsequently 

 eliminated. Next, the procedure was repeated with the 

 remaining terms until no interaction term could be 

 eliminated without increasing the AIC. Then, the above 

 process was repeated for the main effects. For the main 

 effects having an interaction term, both the main effects 

 and the interaction term were eliminated together as a 

 unit. The final model chosen contained those terms that 

 produced the minimum AIC value. 



Results 



Effect of the auxiliary net on trawl geometry 



Regressions of wing spread and footrope height on 

 scope, gear type, and their interaction were compared 



to determine how closely the two gear types fished. The 

 interaction term was not significant for wing spread 

 (P=0.08) nor for footrope height (P=0.82), indicating that 

 the slopes did not differ between gear types. However. 

 tests of the intercepts were significant for both wing 

 spread and footrope height and indicated that trawl 

 geometry differed between survey and experimental 

 trawls (Table 1). Predicted standard survey wing spreads 

 for the minimum (137 m), median (229 m), and maxi- 

 mum (320 m) scopes used were 16.6, 16.9. and 17.3 m 

 — approximately 0.8 m more than the experimental 

 gear at the same scopes. Predicted footrope distances 

 off the bottom were 0.8, 1.6, and 2.4 cm, at the above 

 three scope values — approximately 0.8 cm greater than 

 the experimental gear. Although we detected statistical 

 differences in the trawl geometry between the two gear 

 types, the actual difference in physical measurements 

 was small and presumably had only a nominal effect on 

 the results of the capture efficiency experiment. 



Our assumption that the auxiliary net caught all 

 escaping crabs was reinforced by two observations: 1) 

 the data from the bottom contact sensor on the chain 

 footrope indicated consistent contact with the sea floor; 

 and 2) the auxiliary net consistently had large catches 

 of benthic organisms other than crab, such as starfish 

 and shells, and produced enough drag on the system to 

 reduce wing spread. The effectiveness of the auxiliary 

 footrope at capturing escaping crab is in part due to 

 its weight and small diameter that enable it to sweep 

 beneath the crabs and in part due to the suspension of 

 benthic organisms initiated by the turbulence created 

 by the passing of the first footrope. 



Length-based capture probability 



Capture probability was estimated from length mea- 

 surements (n=3233) collected from 43 successful 

 experimental tows (21 north, 22 south) made within 

 11 standard EBS survey station blocks (Fig. 2). Male 

 samples (n = 1667) ranged in size from 23 to 184 mm 

 (Fig. 3). Female samples (/? = 1566) ranged in size from 

 51 to 162 mm. 



The two-parameter model (model 1) of capture prob- 

 ability was selected over the three-parameter model 

 (model 2) because it had a lower AIC value for both 

 male and female RKC (Table 2). For the comparison of 



