82 



Fishery Bulletin 92(1), 1994 



was used as a measure of presence of juvenile 

 salmon. 



Five null hypotheses were tested during fish 

 length analyses of the four species. The first four 

 hypotheses stated that size of a species did not dif- 

 fer for fish from 1) outside and inside waters, 2) 

 outside waters >37 km offshore and <37 km offshore, 

 3) northern and southern waters, and 4) July and 

 August of 1984. The alternate hypotheses stated 

 that fish were larger in 1) outside than inside wa- 

 ters, 2) outside waters >37 km offshore than outside 

 waters <37 km offshore, 3) northern than southern 

 waters, and 4) August than July of 1984. The fifth 

 hypothesis stated that length did not differ among 

 species caught within each period. 



A number of one-tailed, two-sample ^-tests were 

 conducted under null hypotheses 1-4. Only cells 

 that varied in one dimension were directly com- 

 pared. (For example, under the hypothesis that 

 mean sizes of fish from northern and southern wa- 

 ters did not differ, the mean lengths of pink salmon 

 in the inside waters of northern and southern South- 

 east Alaska in August 1983 could be compared be- 

 cause the difference between these two cells was in 

 only one dimension — north versus south. ) Each pos- 

 sible pairwise comparison under one of the hypoth- 

 eses was treated as a separate, single, and indepen- 

 dent test, and all comparisons were equally weighted. 

 No ^-tests could be conducted if one cell had only one 

 fish length. For the overall probability statement, the 

 following statistic was used (Winer, 1971): 



22>" 



where it, 



lnP. 



Under the hypothesis that the observed probabili- 

 ties were a random sample from a population of 

 probabilities having a mean of 0.50, the % 2 statistic 

 has a sampling distribution which is approximated 

 by the x 2 distribution having 2k degrees of freedom, 

 where k is the number of comparisons (Winer, 1971). 

 For size hypothesis 5 (no difference in mean fork 

 length among salmon species), ANOVA was applied 

 by pooling observations for each species from all 

 habitats and regions. In effect, the pooled species 

 length distribution is a weighted sum of the compo- 

 nent distributions represented by the individual 

 samples. Mean lengths of different species were 

 compared separately for each period. If the overall 

 F-test was significant, all possible species compari- 

 sons within a period were tested with two-tailed t- 

 tests. Experimentwise error was controlled at a = 

 0.05 by adjusting the critical value for each t-test 

 to a = 0.0085, by using the Dunn-Sidak method 

 (Sokal and Rohlf, 1981). 



Results 



Total catch 



Over 10,000 juvenile Pacific salmon were captured 

 in 252 seine sets during the three sampling periods 

 (Table 1). The catch consisted of 66% pink salmon, 

 11% chum salmon, 12%' sockeye salmon, 10% coho 

 salmon, and 1%> chinook salmon. Pink salmon were 

 the most abundant species (CPUE=27), with 6,804 

 caught. Chinook salmon were the least abundant 

 species (CPUE=0.23), with only 57 caught. 



Statistical distribution of catch 



Catch distribution of juvenile salmon was extremely 

 patchy. None were caught in 22% of the sets; more 

 than half were captured in 5% of the sets. Plotting 

 catch abundance against quantiles illustrated that 

 the underlying statistical distribution for each spe- 

 cies was clustered around zero (Fig. 2). Chinook 

 salmon had the lowest FO in catches ( 12%), followed 

 by sockeye salmon (32%), chum salmon (397/ ), pink 

 salmon (45%), and coho salmon (54%). Coho salmon 

 (median catch=l) was the only species with a me- 

 dian catch >0. 



Juvenile salmon had highly aggregated distribu- 

 tions. Morisita's Index of Aggregation (I & ) was sig- 

 nificantly (P<0.001) greater than 1, indicating all 

 species had aggregated distributions in each habi- 

 tat and for all habitats pooled (Table 2). 



Species associations 



Pink, chum, and sockeye salmon catches were 

 closely associated with each other. Catches of pink, 

 chum, and sockeye salmon were positively and sig- 

 nificantly (P<0.05) correlated (Table 3). In contrast, 

 coho salmon abundance was not correlated with that 

 of other salmon (Table 3). 



Abundance 



By habitat In Southeast Alaska and British Co- 

 lumbia combined, pink salmon were the most abun- 

 dant species in each habitat (Table 1). The total pink 

 salmon catch exceeded the catch of each of the other 

 species by six times or more. 



In Southeast Alaska, the CPUE of juvenile pink, 

 chum, coho, and chinook salmon was greater in in- 

 side waters than in outside waters (Fig. 3), whereas 

 sockeye salmon were more abundant in outside 

 waters than inside waters (Fig. 3). For each species, 

 the lowest CPUE and FO were in the outer coast 

 inlets; sockeye salmon were never captured in an 

 outer coast inlet (Fig. 3). The FO of pink, chum, and 

 sockeye salmon was higher in outside than inside 

 waters; the opposite was true for coho salmon (Fig. 3). 



