Weinberg et al.: Capture probablity of a survey trawl for Paralithodes camtschaticus 



743 



Towing sites were selected according to catch rates 

 and carapace lengths obtained from the recently com- 

 pleted 2002 EBS survey (Stevens 2 ). Tows were made in 

 pairs, one in a northerly direction, the other in a south- 

 erly direction and were offset to the east or west by a 

 minimum of 0.1 nmi; the initial direction was chosen 

 randomly in order to mitigate any bias that the current 

 flow might have on footrope contact with the bottom 

 (Weinberg, 2003). Increased effort was given to sites 

 producing favorable numbers and crab lengths by add- 

 ing additional towing pairs. For each tow the total catch 

 of all species from each net was first weighed before all 

 RKC were removed from the catch, weighed, coded by 

 sex, and measured to the nearest millimeter. 



Data analysis 



Trawl geometry Trawl geometry for standard survey 

 nets and experimental nets was measured to confirm 

 that the two gear types fished similarly. Average wing 

 spreads and footrope heights off-bottom for experimental 

 tows were compared to those from 33 standard survey 

 gear tows taken at the same or nearby sampling loca- 

 tions. Because the depth of sampling varied, wing spread 

 and footrope height were linearly regressed on scope, 

 a factor variable indicating gear type (i.e., survey or 

 experimental), and their interaction. Two-tailed r-tests 

 were used to test for the difference in the slopes and the 

 intercepts between gear types. Significance of the inter- 

 action term indicated that slopes differed between gear 

 types. For nonsignificant interaction, significance of the 

 intercepts indicated that wing spread or footrope height 

 differed between gear types by a constant amount. 



Capture probability Capture probability for the experi- 

 mental gear was estimated from catch data of the trawl 

 and the auxiliary net as a function of carapace length 

 (L) for both male and female crab. Based on the assump- 

 tion that the auxiliary net allows no escapement, the 

 probability of capture at the footrope was modeled as a 

 logistic function (Munro and Somerton, 2002) by using 

 SPLUS software (version 6.1, Insightful Corporation, 

 Seattle, WA). Two models were considered: the first, 

 a two-parameter model which reaches an asymptotic 

 maximum of 1 (unity): 



PlLh 



l +e -<a+pLr 



(1) 



and the second, a three-parameter model which reaches 

 an asymptotic maximum less than 1: 



PiD- 



1 + e 



-la+/3L) 



(2) 



2 Stevens, B. G., R. A. Macintosh, J. A. Haaga, C. E. Armistead, 

 and R. S. Otto. 2002. Report to industry on the 2002 

 eastern Bering Sea crab survey. AFSC Proc. Rep. 2002-5, 

 59 p. Alaska Fish. Sci. Cent., Natl. Mar. Fish. Serv., NOAA, 

 Kodiak Fishery Research Center, 301 Research Court, Kodiak 

 AK 99615. 



Because crab capture at the footrope is a binomial pro- 

 cess (i.e., crabs are either captured or they escape), the 

 models were fitted to the capture and length data by 

 using maximum likelihood (Millar, 1992; Munro and 

 Somerton, 2001) and the data were pooled across tows. 

 For each sex, both models were fitted to the data, and the 

 best of the competing models was selected according to 

 the lowest obtained value of the Akaike information cri- 

 terion (AIC; Burnham and Anderson, 1998), defined as 



AIC = -2(log likelihood ) + 2( number of parameters). 



After choosing a model for each sex, we examined whether 

 the capture probability curves differed between sexes by 

 fitting a model to the data for both sexes combined, and 

 then comparing the value of AIC for this model to the 

 sum of the AIC values for the models fitted to each sex, 

 again using the minimum AIC value to objectively select 

 the better of the two models. 



Bootstrapped confidence intervals were constructed for 

 the mean capture probability for each 1-mm length cat- 

 egory, between the smallest and the largest individuals 

 (Efron and Tibshirani, 1993) by resampling the catch- 

 at-size data from individual hauls 1000 times. Empirical 

 95% confidence intervals were then determined as the 

 range between the 25th highest and the 25th lowest of 

 the bootstrap capture probability estimates. 



Video data analyses 



To understand the factors associated with crab escape- 

 ment under the footrope, a video camera was mounted 

 on the trawl to observe RKC interaction with the center 

 of the trawl footrope. These in situ video observations 

 included tows made in 2000 on the standard trawl and in 

 2002 on the experimental trawl. Artificial light was pro- 

 vided for all of the 2000 tows, and for some of the 2002 

 trial tows made before the capture efficiency experiment 

 began. All the 2002 experimental tows were made under 

 natural light conditions. RKC encounters observed on 

 the videotapes were counted from the time the footrope 

 settled to the bottom until the time the footrope was 

 lifted off-bottom at the end of the tow. The probability of 

 capture was predicted as a function of several explana- 

 tory variables. Variables observed and codes (in paren- 

 theses) for each individual included the following: 



1 the capture event — the crab escaped beneath the 

 footrope (0) or was captured (1); 



2 use of artificial light — the tow was made with (0) 

 or without artificial light (1); 



3 estimated mean footrope height above the sea floor 

 over the course of the entire tow was based on the 

 bottom contact sensor and was expressed in centi- 

 meters (0-5); 



4 body height — the crab was observed to be crouching 

 with its legs either tucked beneath the carapace or 

 stretched out so that the carapace was very close 

 to the bottom (0) or standing upright on its dactyls 

 (1); 



