492 



Fishery Bulletin 102(3) 



of dive transects for which data were collected for mul- 

 tiple years was 75. For simulations with a sample size 

 less than 75, we randomly subsampled the data in the 

 same manner as we did with pots. For simulations with 

 sample sizes greater than 75, we amplified the samples 

 with simple bootstrapping to obtain samples from 100 to 

 250 transects in 25-transect increments (Wonnacott and 

 Wonnacott, 1990). For each sample size, we modeled 

 three annual rates of change (0.02, 0.03, and 0.05). 



To evaluate how study duration affects power, we lim- 

 ited the analysis to males, varied study duration ("num- 

 ber [surveys] conducted") from two years to 12 years in 

 two-year increments, and compared three annual rates 

 of change (0.02, 0.03, and 0.05) for both pots and tran- 

 sects. To hold effort constant between the two sampling 

 techniques, we set the pot and transect sample size to 

 the number we could accomplish in a five-day research 

 cruise (250 pots and 75 transects). 



To explore the relationship between annual trend in 

 population and power, we held effort constant (250 pots 

 and 75 transects) and varied the annual trend (from 

 -0.10 to +0.10 in 0.01 increments) for both males and 

 nonovigerous females. It was not possible to conduct a 

 power analysis for ovigerous females because a large 

 proportion of the pots and transects had no ovigerous 

 female crabs. 



Results 



The pot CPUE estimates for males, nonovigerous females, 

 and ovigerous females was significantly different in 

 April than in the following September (Fig. 2, A, C, and 

 E). Male and nonovigerous female CPUE was higher in 

 September (Fig. 2, A and C) and ovigerous female CPUE 

 was lower in September (Fig. 2E). In contrast, April den- 

 sity estimates from dive transects were not significantly 

 different from the following September density estimates 

 for males (Fig. 2B). Dive density estimates for nonovig- 

 erous females were higher in September than in April 

 (Fig. 2D); density estimates for ovigerous females were 

 lower in September than in April (Fig. 2F). 



When we tested the influence of temperature and 

 season on male CPUE with stepwise regression, season 

 was selected first; temperature was not selected because 

 it did not have a significant additional effect (Table 2). 

 Because no significant difference was found between 

 the April and September density estimates from dive 

 transects (Fig. 2B), we did not conduct a stepwise re- 

 gression for the dive data. 



Percentage estimates of females from sampling with 

 pots were lower than percentage estimates from dive 

 transects for a significant number of samples for both 

 April and September (Fig. 3Ai; therefore pots were bi- 

 ased against sampling females. When females were 

 split by reproductive status, no bias was detected for 

 sampling nonovigerous females with pots (Fig. 3B). 

 In contrast, the percentage estimates for ovigerous fe- 

 males remained biased and the magnitude of the bias 

 increased (Fig. 3C). To test potential sampling bias 



co 2.5 • 



3.25 



2.75  



2.25 



O 1.5 • 



1.75 



2.75 



2.25 



P=0.40 



P=0.04 



0.75 



1.75 

 3.75 



2.75 \ 



2.75 



P=0.04 



April 



September 



April 



September 



Figure 2 



Within-year paired comparisons by site of catch in 

 pots (left column) and density on dive transects (right 

 column) for: (A and B) male Dungeness crabs {Cancer 

 magister); (C and D) nonovigerous female crabs; and 

 (E and F) ovigerous female crabs. Catch and density 

 data were transformed with a square-root transforma- 

 tion. P-values indicate results from paired /-tests and 

 significant results show differences between April and 

 September. Lines on the graphs are parallel if measure- 

 ments at sites were consistently higher or lower in April 

 and September. 



related to crab size, we compared the proportion of 

 the male population that was legal size sampled with 

 pots and dives (Fig. 4). There was no significant bias 

 when pots and transects were compared with a sign 

 test (April. P>0.999; September, P=0.06). 



CPUE estimates from pots had a higher power than 

 density estimates from dive transects for the same 

 sample size (Fig. 5). Because more time is required 

 to conduct a dive transect than to set and pull a crab 

 pot, the power of transects compared to pots was even 

 lower when effort was incorporated into the analysis 

 (Fig 6). The power can be increased for both pots and 



