620 



Fishery Bulletin 97(3), 1999 



A^^^. and n^^ were calculated by pooling together the 

 data from all good tows. For mature female crab, a 

 simplified model consisting of only the P. term was 

 fitted to data because the smallest mature females 

 were too large to fit through the belly meshes of the 

 trawl net. In both cases, the models were fitted to 

 the data by using the S+ function MS (Venables and 

 Ripley, 1994). 



The 95% confidence intervals for E^^ ^^, were esti- 

 piated by using a bootstrap analysis (Efron and 

 Tibshirani, 1993) which considered between-haul 

 variability but ignored within-haul binomial variabil- 

 ity which is relatively small. Each bootstrap estimate 

 of E^ ^j was calculated by randomly sampling indi- 

 vidual hauls with replacement from the original data, 

 then by fitting the model to the data from the chosen 

 hauls pooled together. After replicating the bootstrap 

 process 240 times, values oiE^^ ^^. at each increment of 

 carapace width were then ranked. The upper and lower 

 confidence intervals were chosen as the values of fi^^ ^^. 

 ranked 7* and 234* at each width increment. 



Two tests of model form were conducted. First, to 

 determine if the P^ term was a significant contribu- 

 tion to the model, the goodness of fit of the model, 

 including both the P, and P^ terms (five parameters), 

 was compared to that of a model including only the 

 P. term (three parameters) by using a likelihood ra- 

 tio test (pages 153-155 in Hilborn and Mangel, 1997). 

 Second, to determine if the models for mixed sexes 

 differed between crab species, the summed likeli- 

 hoods of the model fit to each species were compared 

 with the likelihood from a model fitted to the com- 

 bined data for both species by using a likelihood ra- 

 tio test. For both tests, significance of the likelihood 

 ratio was evaluated by using a chi square statistic 

 with degrees of freedom equal to the difference in 

 parameters between the models being tested (Hilborn 

 and Mangel, 1997). 



In addition to the tests on model form, two tests 

 were also conducted for differences in net efficiency 

 between biological subgroups of Tanner crab. First, 

 a test was conducted to determine whether net effi- 

 ciency differs between mature females and mixed 

 sexes restricted to the same range of carapace width 

 as mature females. Because of the restricted range 

 of carapace widths, we assumed that net efficiency 

 could be modeled as a simple logistic function of cara- 

 pace width and sex. This model was fitted by using 

 the S-i- function GLM (Venables and Ripley, 1994). 

 Second, a test was conducted to determine whether 

 net efficiency for mature male Tanner crab differed 

 from that for immature male Tanner crab of equal 

 size. Male crab were categorized as either immature 

 or mature on the basis of height of their chela in re- 

 lation to their carapace width by using the computer 



technique of Somerton (1980). Both categories were 

 then restricted to a common range of carapace width 

 that was spanned by the smallest mature and the 

 largest immature individuals. Net efficiency was then 

 modeled as a function of maturity and carapace width 

 by using logistic regression. Significance of the sex 

 term in the first test and the maturity term in the 

 second test were assessed with analysis of deviance 

 (page 186 in Venables and Ripley, 1994). When a term 

 was significant, the resulting model was evaluated to 

 predict net efficiency for each biological class at the 

 midpoint of the carapace width interval examined. 



Results 



Effect of the auxiliary net on trawl performance 



The attachment of the auxiliary net to the trawl net 

 resulted in a decrease in wing spread, presumably 

 from increased drag on the bottom. During the test- 

 ing phase when the codends of both the trawl and 

 auxiliary nets remained open, average wing spread 

 for hauls with standard 55-m bridles was 14.3 m 

 (n=3), considerably less than the mean width of the 

 standard net (mean= 17.0, n =29) used during the pre- 

 ceding survey hauls with the same length of towing 

 cable. To help compensate for the additional drag of 

 the auxiliary net, we shortened the bridles to 28 m. 

 This increased average wing spread to 15.4 m (n=4). 

 Wlien the codends were closed, however, the auxil- 

 iary net rapidly filled with epibenthic fauna, espe- 

 cially brittlestars, resulting in a progressive narrow- 

 ing of the wing spread by as much as 3 m over the 

 duration of a haul. After seven attempts at trawling 

 in the initial study area, we were forced to locate a 

 new study area with a lower abundance of brittle- 

 stars. The progressive narrowing was nearly elimi- 

 nated by the change of sampling site, but the mean 

 wing spread during the remainder of the experimen- 

 tal hauls (mean=15.9, ^(=24) was still significantly 

 less (^test, ^=5.18, P<0.001) than the standard net. 

 Although the wing spread was less during the ex- 

 periment than during the AFSC survey, the depar- 

 ture from the standard width may not have been 

 great enough to affect footrope contact with the bot- 

 tom. Footrope contact at the center of the trawl was 

 evaluated by using a video camera for three tows of 

 the standard trawl (mean width=17.5 m), three tows 

 of the experimental trawl with standard bridles 

 (mean width=14.3 m), and three tows of the experi- 

 mental trawl with short bridles (mean width=15.3 

 m ). In all cases, the footrope in the bosom of the trawl 

 did not contact the bottom, as evidenced by the lack 

 of mud clouds behind the footrope, but instead 



