FISHERY BULLETIN: VOL. 80, NO. 3 



variation apparent between clusters obtained 

 from the same bank in 3 adjacent years (Hawaii 

 in 1976, 1977, and 1978) was as great as the 

 variation in clustering found between different 

 banks and through longer periods of time. While 

 there were a few suggestions of differences in the 

 species composition of groups among the four 

 banks, these were relatively minor and were 

 ignored. Only one fairly consistent pattern of 

 grouping was repeatedly exhibited across banks 

 and through time, and this was confirmed by a 

 single clustering of all the data pooled together. 

 This pattern shows that the bottom fish fishery is 

 loosely composed of three species groups which 

 are apparently segregated on the basis of the 

 depth range of member species (Table 3) (for 

 depth distributions see Gosline and Brock 1960; 

 Brock and Chamberlain 1968; Strasburg et al. 

 1968). These groups represent species assem- 

 blages which are for the most part independent 

 of time and/or geographic location. 



The delimitation of these three species groups 

 is somewhat arbitrary and should not be viewed 

 as the only way in which an intermediate level of 

 species aggregation of the catch could be 

 achieved. Nevertheless this grouping structure 

 is reasonable and its use enhances the biological 

 realism of the multispecies model by identifying 

 and classifying those species which seem to share 

 the greatest correlation in fishing mortalities. In 

 addition the grouping structure allows an assess- 

 ment of the effect aggregation has on the fit of 

 data to a Schaefer stock-production analysis. A 

 brief discussion of each of these groups is ap- 

 propriate. 



The appearance of ulua, ta'ape, and a'awa in 

 the shallowest group (Group I) is consistent with 

 the observation that these three species are 

 frequently harvested with other types of fishing 

 gear. Members of this group are often seen by 

 scuba divers who venture below 30 m, although 

 the vertical distribution of these species in the 

 deep-sea handline fishery is centered around the 

 60 m terrace which circles much of the Hawaiian 

 Islands (Brock and Chamberlain 1968). Because 

 the name ulua refers to several different carangid 

 species, one of which (P. dentex) is most often 

 taken with members of Group II in deeper water, 

 it is evident that some inaccuracies in the classi- 

 fication exist. This particular defect is not so 

 much a result of the clustering process as it is a 

 result of faulty data. Kahala, on the other hand, 

 range widely (Gosline and Brock 1960) and are 

 known from throughout the depth ranges of both 



Table 3.— Bottom fish species groupings defined by cluster 



analysis. 



Group 



Species 



Approximate depth 

 range (m) 



I Ulua, uku, ta'ape, a'awa 



II Opakapaka, hapu'upu'u, kahala, 



gindai, lehi, nohu 

 III Onaga, ehu, kalekale 



30-140 



80-240 

 200-350 



Groups I and II and occur even shallower. Its 

 position in Group II may simply reflect the 

 relatively greater fishing pressure exerted in the 

 100-200 m depth range where other members of 

 Group II, such as the opakapaka and hapu'upu'u, 

 are centered. The deepest group (Group III) is 

 particularly well defined and is composed of 

 three lutjanid species, two of which are deep- 

 water eteline red snappers. 



Fishing Effort 



An attempt was made to evaluate the two mea- 

 sures of fishing effort, catch-records and fisher- 

 man-days, on the basis of their correlation with 

 catch per unit of effort (CPUE). The Schaefer 

 model predicts that plots of CPUE against effort 

 should demonstrate a linear relationship with a 

 negative slope if the production of the stock is 

 described by the logistic growth curve (Ricker 

 1975). Such a prediction generates a one-tailed 

 test of the hypothesis that p>0 against the alter- 

 native hypothesis that p<0 where p is the popula- 

 tion correlation coefficient between CPUE and/. 

 Even though a negative correlation between 

 CPUE and effort is expected in a situation where 

 catch and effort are completely unrelated 

 random variables, the degree of spurious 

 correlation due to this effect will be small if the 

 main cause of variation in CPUE is varying stock 

 abundance (Gulland 1974). 



Correlations were computed between these 

 two variables, using both measures of fishing 

 effort for each species group-bank combination 

 (3X4 = 12). Additional correlations were 

 computed for the total aggregated catch from 

 each of the four banks (1 X 4 = 4), resulting in 16 

 comparisons of the two measures of effort (Table 

 4). Comparisons which might be based on 

 treating species as independent stocks are in- 

 appropriate here because the two measures of 

 effort become equal in this limiting case. One 

 means of evaluating the effectiveness of these 

 two measures is to compare the signs of the cor- 

 relation coefficients (r) and the magnitudes of 

 the coefficients of determination (> -2 ) for each. It 



440 



