878 



Fishery Bulletin 101(4) 



Table 3 



The results of fitting generalized linear models on the probability of encountering crabs, and the catch rate of crabs (numbers per 

 pot) for nonzero catches. ANOVA results; the significance of adding each parameter to the general linear model is presented. For 

 the binomial model of the probability of encounter, a chi-squared significance test was used; for the Gaussian model of ln(CPUEl 

 an F test was used. Coefficients p,..£)j are given with standard error in parentheses. For those models including "area" the standard 

 case was for East South Georgia (EG), and the parameters for North South Georgia and Shag Rocks (NG, SR) are given. Analyses 

 were performed with Splus statistical software. Final models were constructed by using only the significant parameters (italicized 

 in this table), n = number of crabs in sample. Poly (1), Poly (2), and Poly (3) are the coefficients of each of the three orders in a 3rd 

 order polynomial. 



Probability of encountering 



Ln(CPUE) for nonzero 

 catches of P. formosa 



Probability of encountering 

 P. spinosissima 



Ln(CPUE) for nonzero 

 catches of P. spinosissima 



91 



<0.001 

 >0.05 

 >0.05 

 <0.05 



91 



1.7229(0.4301) 



Poly(l) 

 Poly(2): 

 Poly(3): 



4.6387 (5.3605) 

 -8.4647 (6.2257) 

 -1.3614 (4.4808) 



0.0447 (0.0221) 



82 



<0.001 



<0.01 



>0.05 



<0.01 



82 



3.217(0.4877) 

 -0.0055 (0.0008) 



NG: 1.1304(0.3241) 



SR: 1.2771(0.2875) 



0.0638(0.0191) 



of bad weather, and to eliminate these the maximum 

 soak time was Hmited to 39 hours. The mid-depth of the 

 set (average of the depth at the start and end of the set) 

 was used to indicate setting depth, and the analysis was 

 restricted to sets whose depth range was less than 200 m. 

 The data from 45 sets (19% of total) were omitted from 

 analyses because they did not meet these criteria for area, 

 depth, or soak time. 



For the binomial encounter models, the only significant 

 factor was depth (Table 3). F'or the Gaussian catch per pot 

 model, depth, and soak time were significant for both spe- 

 cies and area was significant for P. spinosissima. However, 

 the depth effects were opposite for the two species, so that 

 P. spinosissima decreased in abundance with depth and P. 



formosa increased in abundance with depth, at least up to 

 about 1000 m depth (Fig. 3). There were insufficient data 

 from the cruises in 2001 to establish the effect of depths 

 shallower than 300 m and deeper than 1100 m, although 

 the limited sampling at depths greater than 1200 m in year 

 2000 (Fig. 1) suggests that catch rates of P formosa would 

 continue to decline at these depths, as suggested by the 

 generalized linear models in Figure 3. 



The sex ratio of P. formosa was skewed towards males 

 in shallow water (<800 m) and females in deep water 

 (Fig. 4). The mean size of P. formosa of both sexes also 

 decreased significantly with an increase in depth (Fig. 

 5). Although catch rates in numbers were usually much 

 higher in deeper water, smaller crabs of no commercial 



