Zeppelin et al.: Sizes of walleye pollock and Atka mackerel consumed by Eumetopias jubatus 



511 



In symmetrical fishes such as walleye pollock and Atka 

 mackerel the left and right otoliths are mirror images 

 of each other (Harkonen, 1986). We compared the left 

 and right-sided measurements for all seven structures 

 using a subsample of the structures used to develop the 

 regression equations. There was no significant difference 

 for either walleye pollock (paired Ntest, P<0.05, /? = 13 for 

 HYPO, 15 for QUAD, and 14 for all other structures) or 

 Atka mackerel (paired t-test, P<0.05, ra=14 for OTOS 

 and 17 for all other structures). 



Fish specimens used for regressions were collected 

 from the Gulf of Alaska and Bering Sea. Standard 

 length (SLi was converted to fork length for walleye 

 pollock (when fork length was not available for a small 

 number of otoliths included in the regressions) by using 

 the following equation: FL = 0.40+1. 07(SL) (Wilson 2 ). 

 We chose to use FL over SL for the regressions because 

 all fish were in good condition, thus allowing for ac- 

 curate measurements. Additionally, FL is the standard 

 used for commercial fishery and survey data by the 

 National Marine Fisheries Service for direct compari- 

 sons. A partial analysis of these data was previously 

 reported in Orchard (2001). We expanded the data set 

 reported in Orchard (2001) to reflect the size range of 

 bones found in Steller sea lion scats and included only 

 fish specimens collected within our study area. 



Linear regression models were fitted for most cranial 

 structures by using the following equation: 



Y= a+ PX, 



where Y = the fork length of the fish; 



X = the measurement of the cranial structure; 

 and 

 a and P are constants that define the regression 

 formula. 



However, some cranial structures provided a better fit 

 with the following quadratic regression equation: 



Y = a + PX + pX 2 . 



The strength of the relationship of the regression models 

 was assessed by using a coefficient of determination 

 (r 2 ). 



Erosion is a potential source of bias when estimating 

 prey body size from digested otoliths (Prime and Ham- 

 mond, 1987; Dellinger and Trillmich, 1988; Harvey, 

 1989). We used condition-specific digestion correction 

 factors (DCFs) developed by Tollit et al. (2004b, this 

 issue) to correct for the high degree of variation in the 

 erosion of cranial structures. DCFs were obtained from 

 feeding experiments on captive juvenile Steller sea lions 

 by using a subsample of fish collected for the regres- 

 sion analysis (Tollit et al., 2004b, this issue). Selected 

 cranial structures from three size groups of pollock 



2 Wilson, M. 2003. Persona] commun. Alaska Fisheries Sci- 

 ence Center, Natl. Mar. Fish. Serv., NOAA. Seattle, WA. 



(28.5-45.0 cm FL) and one size group of Atka mackerel 

 (30-36 cm FL) were used to develop the DCFs. 



Estimation of size of walleye pollock and Atka mackerel 

 consumed by Steller sea lions in the Bering Sea and 

 Gulf of Alaska 



Steller sea lion scats were collected from 1998 to 2000 

 along most of the U.S. range of the Alaskan western stock. 

 Scats were collected from rookery (breeding) and haul- 

 out (nonbreeding) sites in summer (June-September) 

 and haul-out sites in winter (February-March). We 

 assumed that scats collected on summer rookery sites 

 primarily represent the diet of adult females because 

 adult males present on rookeries usually fast during 

 this time. Juveniles of both sexes come ashore on rook- 

 eries during summer and undoubtedly are represented 

 in the data, but to a lesser degree than adult females. 

 Scats from juvenile Steller sea lions are more likely to 

 be sampled on haul-out sites during summer, where 

 juveniles make up the greatest proportion of individuals. 

 Scats collected on summer haul-out sites or any winter 

 site presumably represent a greater cross-section of 

 ages and sexes than collections from rookeries during 

 summer. 



Scats were rinsed through nested sieves of 4.8-, 1.4-, 

 0.7-, and 0.5-mm mesh. Bones and otoliths were iden- 

 tified to the lowest possible taxon by using reference 

 collection specimens. All recovered otoliths and selected 

 bones identified as either walleye pollock or Atka mack- 

 erel were given a condition grade based on the degree of 

 erosion (Tollit et al., 2004b, this issue). In general, cra- 

 nial structures considered in "good" condition had little 

 or no erosion, "fair" were moderately eroded (generally 

 up to about 20%), and "poor" were heavily digested 

 (Tollit et al., 2004b, this issue). All structures that were 

 given a condition grade of "good" or "fair" were identi- 

 fied as being from the left or right side and measured 

 to the nearest 0.01 mm with digital calipers. Cranial 

 structures graded as "poor" were not measured and ex- 

 cluded from further analyses because of high observed 

 intraspecific variation (Tollit et al., 1997; Tollit et al., 

 2004b, this issue). 



Fork-length estimates with and without DCFs applied 

 were calculated for each cranial structure and for all 

 structures combined. Otoliths were treated separate- 

 ly because most diet studies currently rely on otolith 

 length to estimate fish fork length. Ninety-five percent 

 confidence intervals around all mean size estimates 

 were calculated by using parametric bootstrapping pro- 

 cedures (Manly, 1997) in which error associated with 

 the regression equation and resampling error resulting 

 from variability within correction factors, and variabil- 

 ity in scats were taken into account. Full details of the 

 bootstrapping procedure are presented in Tollit et al. 

 (2004b, this issue). 



The same fish may be represented by multiple cranial 

 structures within a scat; therefore, in order to avoid 

 pseudoreplication. we selected a minimum number of 

 individuals (MNI; Ringrose, 1993) for each scat sample. 



