Taggart et al.: Estimating abundance of Cancer magister 



491 



crab was recovered during the same sampling event. 

 Thus, the probability of resampling crabs by returning 

 them to the water was very low. 



Sampling by divers 



Divers using scuba equipment censused crabs on 15 

 to 20, 2x100 m belt transects within each study site. 

 Approximately one day of sampling was required at 

 each study site. The dive transects were conducted per- 

 pendicular to the shoreline and they extended from the 

 shallow subtidal (0 m, mean lower low water) to 18 m 

 depth or to the end of the 100 m transect, whichever 

 came first. Divers did not go below 18 m depth in an 

 effort to reduce nitrogen accumulation in divers' blood 

 and to reduce the surface intervals required between 

 transects. From 1992 to 1997, transect locations were 

 randomly selected in the same areas as the crab-pot 

 sampling. The random locations selected in 1997 were 

 resampled during the following years of the study. 



Divers counted all Dungeness crabs located within 

 1 m of each side of the transect. An effort was made to 

 locate buried crabs by swimming close to the bottom 

 and looking for irregularities in the bottom or protrud- 

 ing crab eyestalks. Each crab was examined and the 

 following were recorded: legal males >165 mm carapace 

 width), sublegal males (<165 mm carapace width), ovig- 

 enous females, and nonovigerous females. 



Data analysis 



For each year, we calculated the average pot CPUE for 

 each site by reproductive class (males, nonovigerous 

 females, and ovigerous females). The number of pots 

 sometimes deviated from 50 when a pot was lost or when 

 the degradable cotton string securing the pot lid broke 

 (range: 44-50 pots). The number of crabs counted on 

 dive transects was averaged for each reproductive class 

 by site for each year. All dive transects were conducted 

 perpendicular to shore; thus the transects crossed the 

 shallow habitat where the shallow string of pots was 

 set and terminated at 18 m which was the center of 

 the depth we targeted for the deep pot set. Because the 

 deep pot set was at or slightly beyond the deep end of 

 the transect, we may have sampled more crabs from 

 deepwater habitats than from the shallower transects. 

 However, we did not think this was a significant bias 

 because we sampled crabs from a relatively large area. 

 We, therefore, pooled the pots from both depth strata 

 for analysis. 



We tested for differences between April and Septem- 

 ber for the pot CPUE data and the dive density data 

 with paired f-tests. CPUE and density data were not 

 normally distributed; therefore we transfo rmed the 

 data with a square-root transformation [Y=JiY + 3/8)] 

 for statistical analyses (Zar, 1996). These analyses were 

 conducted for males, nonovigerous females, and oviger- 

 ous females. Because seasonal increases in water tem- 

 perature could drive differences in CPUE between April 

 and September, we calculated mean water temperatures 



by averaging the water temperatures at the 5 m and 

 15 m depths at each site and year. This analysis was 

 limited to years and sites where we collected samples 

 in both April and September (1992-97, from five sites: 

 North Beardslee Islands, South Beardslee Islands, Berg 

 Bay, Bartlett Cove, and Gustavus Flats). We assessed 

 how CPUE was influenced by two independent vari- 

 ables, water temperature and season, with stepwise 

 regression. Because CPUE declined from 1992 to 1997 

 (Taggart et al., in press), we controlled for year so that 

 it would not confound our analysis. 



In order to assess sampling bias between pots and dive 

 transects, the percentages of females (females/all crabs), 

 nonovigerous females (nonovigerous females/all crabs), 

 and ovigerous females (ovigerous females/all crabs) were 

 calculated for each site and sampling time. We also com- 

 pared the percentage of the male population that was 

 legal size (legal-size male crabs/all male crabs) from the 

 pots and from the dives. The percentage estimates from 

 the pot data were compared to estimates from the dive 

 transects with a paired sign test (Zar, 1996). If percent- 

 age estimates for pot data were unbiased when compared 

 to estimates from dive data, the pot percentage esti- 

 mates would have an equal chance of being higher or 

 lower than the percentage estimates for the dive data. 

 Because small sample sizes exaggerate percentage com- 

 parisons, we excluded samples where the total number 

 of crabs collected was less than 25 crabs/site. 



The power of pots and dive transects to detect trends 

 in populations was compared with Monitor, a power 

 analysis program (Gibbs and Melvin, 1997; Gibbs, 

 1998). For our analyses, we varied the number of tran- 

 sects and pots, compared males and nonovigerous fe- 

 males, and varied the duration of the study. For all 

 analyses the following input parameters of the model 

 were held constant: "survey occasions" = annual, "type" 

 = linear, "significance level" = 0.05, "number of tails" = 

 2, "constant added" = 1, "trend variation" = 0, "round- 

 ing" = decimal, "trend coverage" = complete, and "rep- 

 lications" = 10,000. 



To estimate power, the model requires "count" and 

 "variance" for each plot across years for at least three 

 years. Pot and transect data collected from 1992 to 

 1998 from five sites (North Beardslee Islands, South 

 Beardslee Islands, Berg Bay, Bartlett Cove, and Gusta- 

 vus Flats) were used for these analyses. The data were 

 limited to September to avoid seasonal bias. The aver- 

 age across years was calculated for each transect and 

 each pot. These averages were input into the model's 

 variable called "plot count." For each pot and transect 

 a linear regression was calculated among years (CPUE 

 vs. year for pots; density vs. year for dive transects) 

 and the residual mean square was the "plot variance" 

 variable (Thomas and Krebs, 1997). 



To estimate the effect of sample size on power we set 

 the "number [surveys] conducted" to four and limited 

 the analysis to males. We varied the number of "plots" 

 (pots and transects). For pots, we randomly selected 

 subsamples of the 250 pots and ran simulations from 

 25 pots to 250 pots in 25-pot increments. The number 



