622 



Fishery Bulletin 92(3). 1994 



spatial distribution patterns (Fager and McGowan, 

 1963). Species were included in a group based solely 

 on their presence or absence in catches. Fixed groups 

 were defined as the greatest number of members 

 having affinities with one another based on a 40% 

 affinity threshold. 



However, because fishermen are concerned prima- 

 rily with abundance in terms of biomass, the find- 

 ings of RGA were supplemented by cluster analysis 

 (CA), a method that incorporates sample catch 

 weights into the grouping process. CA calculates re- 

 semblance measures using the Bray-Curtis dissimi- 

 larity coefficient (Bray and Curtis, 1957) and then 

 clusters agglomeratively, using a flexible sort fusion 

 strategy with an assigned clustering coefficient value 

 (beta) equal to -0.25 (Lance and Williams, 1967). The 

 flexible strategy was selected over other clustering 

 methods because of its tendency to reduce the chain- 

 ing effect often seen in dendrograms. Dendrograms 

 display the similarities among species and groups in 

 hierarchical form, permitting greater flexibility in 

 the interpretation of associations than the RGA tech- 

 nique which produces set groups (Clifford and 

 Stephenson, 1975). Relative groupings can be dis- 

 tinguished at varying levels of dissimilarity where 

 the number indicates greatest resemblance. 



Prior to classifying assemblages, steps were taken 

 to reduce "noise" in the data. First, groundfish spe- 

 cies other than scorpaenids were eliminated, since 

 the objective of the study was to identify rockfish 

 assemblages without the masking effect caused by 

 the presence of other species. Next, the least frequent 

 rockfishes, defined as those taken at fewer than three 

 stations in each of the groundfish surveys, were omit- 

 ted. Knowledge of the path width of our survey gear 

 facilitated the standardization of catch data as catch 

 per unit of effort (CPUE), i.e. kilograms per hectare 

 towed ( kg/ha )\ or roughly equivalent to towing our 

 gear for 0.8 km. Finally, CPUE's were log transformed 

 (log ]Q (x+l)) to reduce the influence of high CPUE's 

 (Boesch 4 ). 



Results 



Over the six surveys, a total of 1,874 successful hauls 

 were made in the Columbia and U.S. -Vancouver 

 INPFC areas (Fig. 1). Rockfish were present in 79% 



:1 1 kg/ha=0.1 t/kra 2 . 



' Boesch, D F. 1977. Application of numerical classification in 

 ecological investigations of water pollution. Virginia Inst. Ma- 

 rine Science, Spec. Sci. Rep. 77, EPA-600/3-77-033, 114 p. 

 Environ. Res. Lab., Off. Res. Dev., U.S. Environ. Protection 

 Agency, Corvallis, OR 97330 



( 1,476) of the tows. Sampling effort was greatest in 

 1986 and considerably lower during 1977, 1989, and 

 1992. Most of the effort was applied between the 

 depths of 101 and 200 m (Fig. 2). Sampling at greater 

 depths was proportionally higher in 1977 than dur- 

 ing the other surveys. 



Catch composition and species diversity 



Thirty three rockfish species (shortspine thornyhead 

 Sebastolobus alascanus and 32 species of Scbastes) 

 were identified. Among these, 20 were commonly 

 caught in 1,468 hauls and included in the assem- 

 blage analysis (Table 1). 



Catches of rockfish varied widely in size and com- 

 position. Many of the catches were small: 25% had 

 CPUE's under 1.1 kg/ha and 50% had CPUE's under 

 4.8 kg/ha. In contrast, 8% were greater than 100 kg/ 

 ha while only 1% were greater than 500 kg/ha. Maxi- 

 mum rockfish CPUE's reached 4,126, 564, 1,253, 759, 

 2,303, and 828 kg/ha during the six respective sur- 

 veys. The average CPUE for each survey was 50, 16, 

 34, 21, 34, and 27 kg/ha, respectively. On average, 

 abundance levels increased in deeper water, peak- 

 ing in the 151-250 m depth interval (Fig. 3). 



Species diversity in survey catches depicts the 

 multispecies nature of the rockfish community vul- 

 nerable to the bottom trawl. Eighty three percent of 

 rockfish samples contained more than one species. 

 Of these, approximately 50% contained 2-5 species, 

 29% contained 6-10 species, and 4% contained 10- 

 16 species. Of the single-species catches, 78% were 

 under 1 kg/ha. In contrast, the two largest single- 

 species catches (canary rockfish) exceeded 100 kg/ 

 ha. Eighty five percent of the single-species samples 

 were either shortspine thornyhead (7%), canary 

 ( 26%' ), darkblotched (21%), yellowtail ( 18% ), or green- 

 striped, Sebastes elongatus (14%), rockfish. Silver- 

 gray, S. brevispinis, rosethorn, S. helvomaculatus, 

 redbanded, S. babcocki, and yellowmouth, S. reedi, 

 rockfish were never caught alone. 



Species diversity, like abundance, increased with 

 depth (Fig. 4). Hauls made at the shallowest sam- 

 pling sites (-55 m) demonstrated little variety. As 

 sampling depth increased, nearshore species, such 

 as black, S. melanops, and quillback, S. maliger, rock- 

 fish, were replaced by offshore rockfishes, including 

 juveniles of many species that inhabit even deeper 

 waters as adults. Over the middle-shelf, within the 

 55-150 m depth interval, up to 13 rockfish species 

 were taken in a single tow. About 22% of the hauls 

 made at these depths contained five or more species 

 (Table 2). Species diversity peaked along the outer- 

 shelf, where the centers of abundance for several 

 species overlapped. In waters 151-250 m deep, catches 



