Brodeur et al.: Summer distribution of early life stages of Theragra chalcogramma 



605 



Because we were uncertain whether any larger 

 T. chalcogramma were present at the time of sam- 

 pling, we conducted a set of gear comparisons at four 

 different stations to examine whether the Methot 

 trawl was adequately sampling the largest age-0 in- 

 dividuals available. Paired tows (2 pairs during the 

 day and 2 at night) were made with the Methot trawl 

 and an anchovy trawl. The anchovy trawl has a vari- 

 able mouth opening depending on depth and was 

 estimated to range from 110 to 135 m 2 for the four 

 tows on the basis of the relationships given by Wil- 

 son et al. (in press). The trawl body contained vari- 

 able-size mesh grading from 15.2 cm (stretched) in 

 the forward section to 3.8 cm in the codend, which 

 contained a 3-mm liner. At each gear comparison sta- 

 tion, oblique tows with each gear type were done in 

 random order down to the same depth. Standard- 

 ized catches of T. chalcogramma in each gear were 

 compared by length categories by using a nested 

 ANOVA, with haul as the nesting factor. 



The trawl samples were preserved in 5% formalin 

 buffered with marble chips. The samples were later 

 sorted in the laboratory and all the fish were re- 

 moved. Fish were identified by using Hart (1973), 

 Eschmeyer et al. (1983), and Matarese et al. (1989). 

 Standard length (SL) of all fish was measured to the 

 nearest millimeter with a stage micrometer or mea- 

 suring board. Larval and juvenile stages were sepa- 

 rated by using a combination of information on length 

 of transformation and osteology provided by 

 Matarese et al. (1989) and Busby (unpubl. data). 

 Larval and juvenile T. chalcogramma and Pacific cod, 

 Gadus macrocephalus, were separated at 25.0 mm 

 SL by using the criteria of Dunn and Matarese 

 (1987). 



Raw numbers of larvae and juveniles from each 

 taxon collected were converted to number per unit 

 area or density per volume filtered. Consistent with 

 the results of previous studies on larval and juvenile 

 T. chalcogramma (Hinckley et al., 1991; Shima and 

 Bailey, 1994), we did not find significant density dif- 

 ferences overall between day and night sampling 

 (Mann- Whitney Test, P=0.09); thus, we did not make 

 corrections in our abundance estimates for time of 

 day of sampling. Total abundance of each taxon in 

 the study area was calculated by multiplying the 

 weighted mean catch per 10 m 2 for each station by 

 the polygonal area represented by that station (see 

 Richardson [1981] and Kendall and Picquelle [1990] 

 for details). Abundances were calculated separately 

 for both larval and juvenile G. macrocephalus and 

 T. chalcogramma and then summed for the total 

 abundance for each taxon. 



Classification of the catches was done with analy- 

 ses by using both occurrence and abundance data. 



Species associations were identified on the basis of 

 co-occurrence of taxa in catches by using Recurrent 

 Group Analysis (Fager, 1957). This analysis places 

 taxa that co-occur into groups based on an affinity 

 level set at 0.4, as previously used for other assem- 

 blage analyses in this region (Kendall and Dunn, 

 1985; Doyle et al., 1995). Only taxa that occurred in 

 more than 15% of the collections were included in 

 this analysis. A second type of hierarchical classifi- 

 cation was performed on the abundance data to see 

 whether a different technique produced different 

 results in identifying assemblages. Two-way Indica- 

 tor Species Analysis, a polythetic, divisive technique 

 (Gauch, 1982), was used in conjunction with the pro- 

 gram TWINSPAN (Hill, 1979). This analysis starts 

 with all the entities (taxa or stations) belonging to 

 one group and then ordinates them by reciprocal 

 averaging. Each group is progressively divided until 

 it contains no more than the predetermined min- 

 imum number of members, as opposed to an agglo- 

 merative technique, such as cluster analysis, which 

 starts with individual entities and progressively com- 

 bines them. Station groupings were also formed by 

 using TWINSPAN, and these were described in re- 

 lation to the species matrix on the basis of whether a 

 particular taxon had a high or low affinity with that 

 station grouping. 



To interpret the ecological significance of these sta- 

 tion groupings, we examined environmental and sta- 

 tion variables such as water depth and temperatures 

 from different depth intervals available from expend- 

 able bathythermograph data taken at each station. 

 We also calculated station position variables, such 

 as distance from nearest land (including islands) and 

 alongshore distance from a line perpendicular to the 

 coast just northeast of our first transect of stations 

 (Fig. 1). Differences among the median values for all 

 variables by the different TWINSPAN groupings 

 were tested by using a Kruskal-Wallis test. 



We compared the abundance estimates from the 

 1990 AFSC Gulf of Alaska groundfish trawl survey 

 with those determined from the Methot trawl sur- 

 vey and another ichthyoplankton survey conducted 

 a few months earlier than our study. The trawl sur- 

 vey took place from 1 June to 9 September 1990, cov- 

 ered a broader area of the Gulf of Alaska (132- 

 170°W), and sampled depths ranging from 20 to 530 

 m (see Stark and Clausen [1995] for additional sam- 

 pling details). For the purposes of this analysis, only 

 abundances from the western Gulf of Alaska strata 

 (506 stations) were summarized. Abundances for 

 each stratum were estimated by dividing the bio- 

 mass of each species caught by its mean weight (given 

 in Stark and Clausen [1995]), and then these were 

 summed across all depths and strata. 



