To Hit et al.: Sizes of walleye pollock consumed by Eumetopias jubatus 



525 



mated with the minimum number of individuals (MNI) 

 technique (Ringrose, 1993; Browne et al., 2002). This 

 technique is used to select structures within each scat 

 that preclude pseudoreplication or double counting of 

 fish. Within each scat, the structure with the great- 

 est MNI was selected, and right-sided structures were 

 selected over left-sided structures if both sides were 

 found in equal number because right-sided structures 

 are used in regression formulae. If two structures had 

 the same MNI estimate, then selection was made on 

 the structure with the larger regression determination 

 coefficient, r 2 (OTO-W>OTO-L>QUAD>DENT>HYPO> 

 INTER>ANGU>PHAR ). 



Geographical and temporal variation in sizes of 

 prey consumed 



All elements from the seven cranial structures in good or 

 fair condition were used to compare size of pollock con- 

 sumed by Steller sea lions in Southeast Alaska between 

 regions (inside haul-outs versus outside rookeries), across 

 years and across rookeries (with rookery data collected 

 in June and July), and across months (with data col- 

 lected from inside haul-outs). Biologically meaningful 

 differences in FL of pollock were assessed by grouping 

 corrected lengths into stage-class categories (juvenile 

 or 1-year-old fish FL<20 cm; adolescent 20<FL<34 cm; 

 subadult 34<FL<45 cm; and adult FL>45 cm) (Smith, 

 1981; Walline, 1983; Dorn et al., 2001). Adults were 

 considered to be mature fish >5 years old and targeted 

 by fisheries (Smith, 1981). Subadults were likely 3 or 

 4 years old, of which only a proportion had matured or 

 were targeted by the fishery. To avoid the possibility of 

 pseudoreplication in our chi-squared comparisons, we 

 used only the presence or absence of structures of each 

 stage class in a scat because individual fish eaten by a 

 sea lion may have come from an age-specific school and 

 were therefore not independent (Hunt et al., 1996). Pres- 

 ence-absence data was chosen over MNI data because 

 the former greatly reduces potential concerns regarding 

 size-dependent recovery of cranial structures (Tollit et 

 al., 1997). With the exception of our regional comparison, 

 data from juvenile and adolescent stage-classes were 

 pooled because of the low sample sizes of juvenile fish. A 

 Fisher's exact test was used as an alternative test to chi- 

 square comparisons when counts for a stage-class group- 

 ing were <5 (S-PLUS 2000, Mathsoft Inc., Seattle, WA). 



Overlap of prey size with size of fish caught by fisheries 



To assess the impact of using the new methods described 

 and to compare the size of pollock consumed by sea lions 

 with the size of pollock typically caught by fisheries, we 

 obtained randomly subsampled size-frequency landing 

 data from the Canadian commercial pollock fishery in 

 Dixon Entrance (1993-1999) (Saunders 4 ). This area is 



4 Saunders, M. 2002. Unpubl. data. Fisheries and Oceans 

 Canada, 3190 Hammond Bay Road, Nanaimo, B.C., Canada, 

 V9T 6N7. 



115-135 km SE of the Forrester Island rookery on the 

 southern border of Southeast Alaska (Fig. 1). 



Results 



Sizes of pollock consumed 



The traditional method of estimating prey size from 

 otoliths alone was not satisfactory because most otoliths 

 were in poor condition (86%, ra=247) or were broken 

 lengthwise (>89%) (or were both broken and in poor 

 condition). Cranial bones, on the other hand, occurred 

 in higher numbers than otoliths and were therefore more 

 useful for estimating prey size (Table 2). 



Sixty-one percent of scats (1215 of 1987) collected 

 from Southeast Alaska (1994-99) contained pollock 

 remains, with an average MNI of 1.57 ±1.66 individual 

 pollock per scat (range: 1-37 individuals). Many scats 

 contained hard parts that were not useful for estimat- 

 ing prey size (e.g., gill rakers), leaving 531 scats (26%) 

 with measurable selected structures. Of these, 303 scats 

 contained 1746 elements in good (n = 225), fair (n = 684). 

 and poor condition (« = 837). 



Applying digestion correction factors had a consider- 

 able effect on the estimated length and mass of fish 

 consumed, and on the proportion that were deemed 

 to be adults (Fig. 2). The estimated lengths of pollock 

 calculated from all structures graded in good or fair 

 condition (without accounting for digestion) was 34.4 

 ±9.7 cm (ra=909, modal range: 32-40) (Table 2, Fig. 2). 

 Lengths increased by 23% on average when appropri- 

 ate DCFs were applied to each structure to account for 

 the observed degree of digestion (mean FL = 42.4 ±11.6 

 cm, modal range: 44-52, 95% CI = 41. 0-43.9) (paired 

 t-test, i 908 =67.1, P<0.001). A DCF of 1.0 (no correction 

 required to account for digestion) was applied to 62 ele- 

 ments in good condition, resulting in a mean fork length 

 of 39.6 ±11.9 cm estimated from those bones. 



The size-frequency distribution of pollock consumed by 

 sea lions also varied significantly following the applica- 

 tion of DCFs (Kolmogorov-Smirnov, KS = 176.2, P<0.001) 

 and led to an increase in the proportion of fish thought 

 to have been adult (>45 cm FL) from 16% to 44%. This 

 result in turn reduced the proportion of fish thought 

 to have been subadults (29%), adolescents (25%), and 

 juveniles (<2%, <20 cm FL) (Fig. 2). The size range of 

 pollock eaten ranged widely regardless of whether DCFs 

 were applied (10-78 cm) or not (10-64 cm). When we 

 calculated fork lengths using only elements selected 

 according to MNI criteria, the means increased by just 

 0.5 cm for corrected and by just 0.3 cm for uncorrected 

 lengths, with near identical standard deviations and 

 distributions (Fig. 2) (Kolmogorov-Smirnov, uncorrected 

 KS = 0.33, P=0.89, corrected KS = 0.032, P=0.91). 



The use of all otoliths regardless of digestion state 

 resulted in a mean fork length that was only about 

 half of that derived by using all structures corrected 

 for digestion (Table 2). Excluding otoliths in poor con- 

 dition significantly reduced sample size (Table 2) but 



