474 



Fishery Bulletin 103(3) 



calculated and compared for each of the 24 species- 

 year combinations. One of the three total abundance 

 estimates was most precise for each of the species-year 

 combinations. For each species-year combination, the 

 habitat stratum sample size (used in the estimate post- 

 stratified by habitat), the HFD stratum sample size, 

 and the LFD stratum sample size (both used in the 

 estimate poststratified by habitat and fish density) were 

 plotted in relation to the total abundance estimator that 

 was most precise in order to investigate the influence of 

 sample size on the relative precision of the three total 

 abundance estimators. 



Indices of abundance Three indices were constructed 

 for each species in each year to determine interannual 

 variations in relative abundance (mean CPUE): an all- 

 site index, a habitat index, and a HFD index. For each 

 species and year, the all-site index was the mean CPUE 

 from all sites sampled. The habitat index was the mean 

 CPUE from all sites sampled within the species' habitat 

 area. The HFD index was the mean CPUE from all sites 

 sampled within the species' HFD area. 



CPUE values were not normally distributed and 

 therefore the Kruskal-Wallis nonparametric analysis 

 of variance test was used to test the three indices for 

 each species' differences in mean CPUE among years. 

 For species that showed significant differences (o=0.05), 

 a Tukey HSD (honestly significant difference) multiple 

 comparison test for unequal sample sizes was conducted 

 to determine which years differed (a=0.05). The Tukey 

 multiple comparison test was used because it is robust 

 with respect to departures from population normality 

 and homogeneity of variance (Keselman, 1976). The 

 results for the three indices for each species were com- 

 pared to see how the differences in estimating abun- 

 dance with the three indices affected conclusions of 

 significant differences in abundance between years. 



Numerous sources of bias can affect estimators of 

 abundance from survey data. The poststratification 

 estimator and other design-based estimators may be 

 biased when applied to data that were not collected 

 under a probability sampling design, as done in the 

 present study. For a qualitative estimate of possible 

 design bias in the estimates, the annual proportion of 

 sample sites in each stratum (habitat, nonhabitat, HFD, 

 and LFD strata) were compared with the proportion of 

 area (km 2 ) in that stratum. First, we compared the size 

 of the habitat area, in relation to the size of the total 

 survey area, with the number of samples taken in the 

 habitat area, in relation to the number taken in the 

 total survey area. 



number of samples taken in the HFD area, in relation 

 to the number taken in the total habitat area. 



Size of the HFD area 



Size of the habitat area 



Number of samples taken 

 in the habitat area 

 Size of the total survey area Number of samples taken 



in the total survey area 



Second, we compared the size of the HFD area, in 

 relation to the size of the total habitat areas, with the 



Number of samples taken 

 in the HFD area 

 Size of the habitat area Number of samples taken 



in the habitat area 



Recognizing that the distribution of individuals var- 

 ied within and across strata, two measures were used 

 to better understand the distribution of each species in 

 each year. The proportion of zero catches (e.g., a "zero 

 catch" for rock sole indicates a tow in which no rock sole 

 were caught) and the mean CPUE of nonzero catches 

 were calculated for each species in each year over four 

 areas: the total survey area, the habitat area, the HFD 

 area, and the LFD area. 



Results 



Fish CPUE statistics were calculated for a total of 244 

 quantitative tows over the six sampling years (Fig. 2) 

 in habitats ranging from 1 to 111 m depth and from 0% 

 to 99% sand. Based on compiled data from all six years, 

 the habitat area for rock sole was defined by 1-84 m 

 depth and 2-99% sand; for yellowfin sole, by 2-43 m 

 depth and 24-99% sand; for Pacific halibut, by 2-27 m 

 depth and 2-99% sand; and for flathead sole, by 12-87 m 

 depth and 8-97%. sand (Fig. 3). The HFD area, defined 

 by depth and percent sand, was determined for each of 

 the four species in each of the six years (Table 1, Fig. 3). 

 Although the range of depth and the range of percent 

 sand were determined independently in each year, they 

 remained quite constant for each species over the six 

 sampling years. 



The size of habitat area in relation to total area 

 ranged across species from 0.62 to 0.92 and, for each 

 species, the proportion of habitat sites to total sites 

 varied among years (Table 2). The proportion of sample 

 sites in habitat to sample sites in the total survey area 

 ranged from 0.88 to 1.00 for rock sole, 0.60 to 0.87 

 for yellowfin sole, 0.52 to 0.93 for Pacific halibut, and 

 0.29 to 0.67 for flathead sole. The relative number of 

 samples taken in each species' habitat area exceeded 

 the relative size of their habitat area (i.e., a positive 

 disproportion of samples in habitat), except for rock 

 sole in 1991 and 1994, yellowfin sole in 1993 and 1994, 

 Pacific halibut in 1993 and 1994, and all years for 

 flathead sole. On average, rock sole had a 5% positive 

 disproportion of samples in its habitat area, yellowfin 

 sole and Pacific halibut had an 11% positive dispropor- 

 tion of samples in their habitat area, and flathead sole 

 had a 15% negative disproportion of samples in its 

 habitat area. 



The size of the HFD area in relation to habitat area, 

 and the number of sites sampled in the HFD area in 

 relation to the number sampled in the entire habitat 

 area, varied over the six sampling years for each of the 

 four species (Table 2). On average over the six years, 



