550 



Fishery Bulletin 90(3). 1992 



niques similar to our submersible methods, estimated 

 abundances of juvenile and adult rockfishes along 

 transects on an inshore reef supporting an extensive 

 kelp forest. Their estimates of juvenile abundances are 

 much higher than ours: more than 46,000 fish/ha com- 

 pared with our maximum of 15,039 fish/ha (station 3). 

 Comparing abundances of adults at the same stations, 

 they estimated 3133-5046 fish/ha vs. 558-1724 fish/ha 

 at our station 3. The maximum estimated number of 

 adult fishes at any of our stations was 9635 fish/ha (sta- 

 tion 4). However, our estimates at the shallow bank top 

 stations (1 and 3) are low because they did not include 

 most of the schooling fishes, which were above the 

 submersible. We have no accurate estimates of the 

 abundances of these fishes, primarily yellowtail and 

 widow rockfish. They occurred in schools of thousands. 



Availability of comparative data 



There are few data sets comparable with ours that were 

 obtained by other methods. Rough bottoms are un- 

 trawlable, reducing usable gear to longlines or set nets. 

 Even where these are used, if substrate varies over the 

 length of the longline or net, they "integrate" the fishes 

 over those different bottom types, preventing associa- 

 tion of species with specific substrate types. Using otter 

 trawls with foot rope rollers, Barss et al. (1982) studied 

 fish assemblages associated with "rough" (rocky bot- 

 tom fishable with nets using rollers) and "smooth" bot- 

 tom on the west (offshore) side of Heceta Bank. The 

 areas with most relief were unfishable (including our 

 stations 1 and 3). They found distinct differences in 

 catches between the two types of areas, but admitted 

 that their results were biased by the type of gear they 

 were forced to use in order to trawl on rough bottom. 

 Recently, Matthews and Richards (In press) used 

 gillnets to compare fish assemblages on trawlable and 

 untrawlable bottoms west of Vancouver Island. Their 

 goal was to determine whether, as commercial fisher- 

 men believe, untrawlable bottom west of Vancouver 

 Island provides refuges for commercially exploited 

 fishes (primarily Pacific Ocean perch) caught nearby. 

 They concluded there were no reservoir populations. 

 However, given the mesh size of their bottom-set gill- 

 nets, they were unlikely to sample either juveniles or 

 semipelagic species such as yellowtail and widow 

 rockfish. Thus, given our current and previous obser- 

 vations of juvenile and yellowtail rockfish associated 

 with shallow, high-relief rocky ridges (Pearcy et al. 

 1989), we suggest that these unfished areas could still 

 provide refuges for fishes in either of those categories. 



Habitat shifts 



Ontogenetic habitat shifts, such as those desribed here 

 for pygmy rockfish, are common among rockfish 

 species. Westrheim (1970), Carlson and Haight (1976), 

 and Straty (1987) found that juvenile Pacific Ocean 

 perch Sebastes alutus were usually shallower than 

 adults. Carr (1983) described the growth-related migra- 

 tion of juvenile S. atrovirens, S. camatm, S. chryso- 

 melas, and S. caurinus to the bottom in a central 

 California kelp forest. However, Hallacher (1977), 

 studying adults and juveniles of 5. rnystinus and S. ser- 

 ranoides in Monterey Bay at depths of '^'25m or less, 

 found that abundances of both increased with depth! 

 maxima occurring at the greatest depths sampled. This 

 difference could be related to degree of association with 

 the bottom as adults. The species Carr (1983) studied 

 were benthic as adults, whereas the latter two species 

 occur in the water column. 



Conclusions 



The results presented here show the utility of using a 

 submersible rather than bottom set nets, traps, or long- 

 lines to study fish-substrate associations in deep water 

 areas where substrate is heterogeneous. Other meth- 

 ods, such as Remote Operated Vehicles, rely on video 

 and still camera images, which are not as adequate for 

 accurate identification as the human eye. Moreover, 

 other types of gear do not allow detailed characteriza- 

 tion of the substrate sampled, but rather integrate the 

 catch from a variety of habitats. We believe that the 

 methods presented here, in addition to describing basic 

 fish-habitat associations, allowed us to control the ef- 

 fects of sampling across a range of different habitats, 

 and increased our ability to detect meaningful spatial 

 variation in fish abundance. 



Acknowledgments 



We thank the crews of the RV William A. McGaw and 

 RS Delta for help in obtaining the data; their coopera- 

 tion and expertise are appreciated and gratefully 

 acknowledged. Comments by L. E. Hallacher and 

 L. Richards greatly improved the manuscript. The 

 research was supported by the Minerals Management 

 Service, Department of the Interior, under MMS 

 Agreement No. 14-12-0001-30445, and the National 

 Undersea Research Program of NOAA. 



