Adams et al.: Population estimates of Pacific coast groundfishes 



451 



any abundance (greater than 1 fish/hectare in either 

 estimate). For the 200-m, 400-m, and 600-m depths 

 (Fig. 3), the mean ROV estimates were higher than 

 the trawl estimates for 13 of 20, 15 of 16, and 9 of 10 

 comparisons, respectively (sign test 200 m: P=0.180; 

 400 m: P=0.001; and 600 m: P=0.011). 



Especially large differences between the ROV and 

 trawl estimates existed for strongly bottom-associ- 

 ated groups: skates, flatfish (Pleuronectiformes), and 

 thornyheads, and particularly for small, cylindrically 

 shaped fish that remained on the bottom: hagfish 

 (Eptatretus spp.) poachers (Agonidae), and eelpouts. 

 No significant differences were found between ROV 

 and trawl estimates for fishes with more off-bottom 

 behavior ("roundfish"), such as Pacific whiting, 

 Merluccius productus, sablefish, and rockfish. Spiny 

 dogfish, Squalus acanthias, were captured at every 

 depth in low numbers in the trawls and were the 

 only species to occur consistently in the trawls that 

 were not seen with the ROV. 



Coefficients of variation were generally smaller for 

 the ROV estimates than for the trawl estimates. For 

 species that occurred in both the ROV and the trawls, 

 the coefficients were smaller for the ROV estimate 

 in 10 of 15 comparisons at the 200-m depth, in 10 of 

 13 comparisons at the 400-m depth, and in 8 of 10 

 comparisons at the 600-m depth ([Fig. 4] sign test; 

 200 m: P=0.170; 400 m: P=0.050; and 600 m: 

 P=0.050). ROV coefficients of variation were always 

 larger for catsharks only and were mixed for Pacific 

 whiting and some species of flatfish. At all depths, 

 ROV coefficients of variation were smaller for rock- 

 fish and thornyheads, species typically with very 

 large variances. Coefficients of variation for total fish 



number were much larger for the trawl estimates 

 than for the ROV estimates at all depths (200-m, 2.4 

 times; 400-m, 4.6 times; and 600-m depths, 3.0 times). 



The required sample sizes to detect a 50% reduc- 

 tion in the log-transformed means for a fixed level of 

 power ( 1-/3, 0.80) were smaller for the ROV than for 

 the trawl, except for catsharks (Table 2). When the 

 ROV sample sizes were smaller, the difference in 

 sample sizes ranged from 1.3 to 19 times. For 

 catsharks, a taxon with high variability, the ROV 

 sample size was nearly six times larger than that 

 for the trawls. 



Most fishes showed no response to the ROV (Table 

 3). Considering all fish taxa observed, 80% showed 

 no response to the ROV, 6% were attracted, and 14% 



