Wilson; Effects of year and region on abundance and size of Theragra chalcogramma 



827 







50 



100 



I "ISO 



I 200 



t 250 



ra 



CD 



50 



en 



100 



150 ^ 

 200 



250 



75 150 225 300 75 150 225 300 



Bottom depth (m) 



Figure 2 



The vertical distribution of age-0 pollock density in the western Gulf of 

 Alaska, August-October, 1985-88. Symbol size indicates the mortality- 

 adjusted number of fisli/m'. The diagonal line indicates bottom depth. 



effect on age-0 density. Annual differences were always 

 limited to the Kodiak region where fish densities were high 

 during 1985 and 1988 in relation to 1986 and 1987 (Table 

 2). Within other regions, annual differences were not sig- 

 nificant. As to within-year patterns, regional differences 

 were mostly limited to 1987 when fish density increased 

 strongly from east to west ( Fig. 3, Table 2 ). The distribution 

 in each other year was more even. For brevity, only results 

 from the first statistical procedure are tabulated. 



In the first procedure, significant year-region interaction 

 was indicated by each two-way AN OVA of transformed fish 

 densities (Table 3). Fishers LSD test indicated annual dif- 

 ferences only in the Kodiali region (Table 4), where age-0 

 density was relatively high during 1985 and 1988 (Table 

 2). The LSD test also indicated regional differences, but 

 only during 1987 or 1988, depending upon transformation 

 (Table 4). When log-transformed, the regional effect was 

 associated with low age-0 densities in the Kodiak region 

 during 1987. When rank-transformed, it was associated 

 with high densities in the same region but during 1988. 

 The LInimak region was excluded from these tests. 



In the second procedure, only two of the seven Kruskal- 

 Wallis tests rejected the null hypothesis of equal fish densi- 

 ties. The first indicated an annual effect within the Kodiak 

 region (Ki-uskal-Wallis test statistic=25.31, P«0.0001). 

 The second indicated a regional effect within 1987, which 

 included the Unimak region (Kruskal-Wallis test statis- 

 tic=32.24,P«0.0001). 



In the third procedure, only one chi-square test out- 

 come was significant (chi-square=15. 19, P=0.0005 ). During 

 1987, the frequency of occurrence around Kodiak Island 

 was low (65%) in relation to occurrences in the Unimak 

 region (98^7^) (Table 2). 



In the fourth procedure, the two-way AN OVA of nonzero 

 densities revealed a significant year-region interaction 

 (df=3,241, F=4.462, P=0.005). Fisher's LSD test indicated 

 that all differences occurred within the Kodiak region 

 where nonzero density was relatively high during 1985 

 and 1988 (Table 2). 



Length 



Overall, 35,530 age-0 pollock were measured. They were 

 collected from 335 trawl hauls and ranged from 3 to 13 

 cm FL. Significant effects of year and region were detected 

 after accounting for a day-of-year effect (Table 5). 



Daily increases in mean length per haul averaged 0.090 

 cm (Fig. 4), which is within the range of published growth 

 rates for these fish (0.6-1.2 mm/d) (Bailey et al., 1996b). 

 Two samples had unusually low mean length per haul (in 

 1986 and 1987) and were from bays; these represented few 

 fish and so had little effect on the fitted lines. 



Annual differences in length were due to a comparatively 

 small mean size during 1986, a medium size during 1985, 

 and large sizes during 1987 and 1988 (Fig. 4, Table 2). A Bon- 

 ferroni multiple comparison test indicated that 1985 and 



