Adams et al.: Population estimates of Pacific coast groundfishes 



453 



avoided it. Pacific whiting had the strongest response 

 (69%), appearing to be attracted to the ROV. Sable- 

 fish had the next strongest response (62%), but their 

 numbers were low, averaging 3.75 individuals over all 

 depths. Their behavior was variable: some animals 

 swam away, some swam toward the ROV, and some 

 ignored the ROV. Catsharks showed approximately 

 equal numbers responding and not responding, and 

 those that did respond consistently avoided the ROV. 

 Finally, the red octopus, Octopus rubescens, was 

 the most abundant animal counted from the ROV. It 

 was observed from the ROV only at the 200-m depth 

 at numbers of 1,610 per hectare (SE=516.4) and was 

 not captured in the trawls. The ROV red octopus 

 abundance estimates were higher than any fish es- 

 timate from the trawls. 



Discussion 



Much of the difference between the ROV and trawl 

 estimates was due to the different mechanical na- 

 ture of the sampling gear and to the body shape and 

 behavior of the fishes. The ROV intensively sampled 

 a narrow area directly in front and extending up a 

 short height off the bottom. The trawl sampled a 

 much wider area, including a larger area off the bot- 

 tom, but with what must have been a great deal of 

 escapement. The result was that ROV abundance es- 

 timates were higher and had lower coefficients of 

 variation for fishes either strongly associated with 

 the bottom or with a body shape and size, or both, 

 that would pass through the mesh ( such as the red 

 octopus). Conversely, animals with highly patchy dis- 

 tributions and off-bottom behavior (catsharks, rock- 

 fish at the 200-m depth) had higher abundance esti- 

 mates from the larger volume trawl. Although both 

 the ROV and trawl estimates were adjusted to the 

 same surface area, there was no adjustment made 

 to reflect the 7.7 times greater off-bottom height 

 sampled by the trawl (Fig. 2). Probably neither 

 method does a particularly good job of estimating off- 

 bottom fishes that are patchily distributed. Visual 

 estimates offish abundance either from submersibles 

 (Uzmann et al., 1977) or from divers (Kulbicki and 

 Wantiez, 1990) are reported to be higher than those 

 from otter trawl estimates for multispecies assess- 

 ments, but the reverse was true in this study for rock- 

 fishes (see also Krieger, 1993). 



The smaller sample sizes required to detect a 50% 

 reduction of the ROV abundance estimates would 

 improve the ability to detect trends in abundance 

 such that a smaller increment of change would be 

 detectable. The degree on improvement would vary 

 with species. Reductions of the order of 1.3 to 19 times 



in required sample size are sufficient to increase the 

 ability to detect true changes in population size. 

 Much of the decrease in required sample size (and 

 increase in statistical power) comes from the much 

 higher ROV abundance estimates. If samples are 

 drawn from two populations with similar levels of 

 variation, a 50% decrease in abundance of the larger 

 estimate is much larger and therefore easier to de- 

 tect. Unfortunately, a great deal of variation due to 

 patchiness in fish distribution remains. 



For an environmental assessment, the ROV esti- 

 mates provide a better overall picture of the commu- 

 nity than do the trawl estimates. While the similar- 

 ity in the presence and absence of species in the' two 

 methods was surprisingly high, the ROV abundance 

 estimates generally tended to be higher and had 

 lower coefficients of variation. Species that were usu- 

 ally in direct contact with the bottom had much 

 higher ROV abundance estimates than did those cap- 

 tured in the trawl. Small, cylindrically shaped fishes 

 (hagfish, poachers, and eelpouts) had particularly 

 large differences between ROV and trawl abundance 

 estimates, probably owing to escapement under the 

 footrope or through the trawl webbing. The most 

 obvious example of escapement or avoidance was that 

 provided by the red octopus, which was more abun- 

 dant in the ROV sampling than any fish, but which 

 was not captured by the trawl. 



For the commercially important species at these 

 depths (Dover sole, thornyheads, and sablefish), the 

 results were mixed. For Dover sole and thornyheads, 

 ROV abundance estimates were higher and coeffi- 

 cients of variation were lower. For sablefish, 

 abudance estimates were higher for the trawl sam- 

 pling at 200-m and 600-m depths. Although rockfish 

 at the 200-m depth are not commercially important, 

 the trawl estimates for this taxon were higher, and this 

 is likely to be true of commercially important rockfishes. 



Stock assessment integrates patterns of removals 

 and information on year-to-year variation in recruit- 

 ment from catch-at-age data with mean levels of 

 abundance and longer term trends from survey data 

 (Deriso et al., 1985; Methot, 1990). Using simulations, 

 Kimura (1989) showed that, if survey data are bi- 

 ased, the results can be a precise, but biased, infer- 

 ence regarding the population. The higher ROV abun- 

 dance estimates mean that trawl estimates are bi- 

 ased too low and that there is actually a larger dif- 

 ference between years of low population size and 

 threshold levels of overfishing set by these assess- 

 ments. The risk of overfishing is actually lower than 

 that which was assumed. A more dangerous situa- 

 tion arises when bottom trawl estimates are larger 

 than the ROV estimates, as with rockfish (also see 

 Krieger, 1993). Here the difference between years of 



