Gudmundson et al Diet of Callorhinus ursinus 



447 



dividing the MNI of each prey species or group (for all 

 samples) by the total MNI of all prey for samples that 

 contained identifiable fish otoliths or cephalopod beaks 

 (or both). 



To determine if the occurrence of prey varied by is- 

 land or sample type (or both), the presence or absence 

 of the primary prey (?!=number of scats with prey re- 

 mains) was modeled as a binomial random variable by 

 using generalized linear models with island, sample 

 type, and the interaction of island and sample type 

 as explanatory variables (S-PLUS 2000, Insightful 

 Corp., Seattle, WA). The interaction was considered 

 significant if adding the interaction significantly re- 

 duced the deviance in the model. If the interaction was 

 significant, islands within each sample type or sample 

 types within each island were compared, depending 

 on whether sample type or island reduced more of the 

 overall deviance. If the interaction was not significant, 

 it was removed from the model, and island and sample 

 type effects were tested. Even though data were insuffi- 

 cient to test for differences in prey occurrence between 

 years, year was included in the model as an additive 

 variable. 



Comparisons of prey age and size 



Recovered walleye pollock otoliths were assigned a con- 

 dition grade ("good," "fair," or "poor") based on distinc- 

 tive features such as sulcus definition, shape, chipping, 

 breaks, and wear (Sinclair, 1988). Pollock otoliths of 

 "good" or "fair" condition were measured lengthwise 

 parallel to the sulcus to the nearest 0.1 mm using hand- 

 held digital calipers. A correction factor was applied to 

 "fair" otoliths to account for loss of otolith length as a 

 result of digestion (Sinclair, 1988; Antonelis et al., 1997). 

 Fork lengths of prey were estimated by using regression 

 formulae of otolith length against body length (Frost and 

 Lowry, 1981) and age class was estimated from fork- 

 length-age relationships (Sinclair et al., 1994). Pollock 

 otoliths of "poor" condition were enumerated, but not 

 measured, and were not included in prey size compari- 

 sons because of their high degree of erosion. 



Pollock otoliths recovered from scat samples processed 

 in the early 1990s were measured, but only age class 

 estimations were recorded in the database. Therefore, 

 to test how the size of walleye pollock otoliths varied 

 by sample type and island, otoliths of "good" and "fair" 

 grades were combined into two age categories; juvenile 

 (0-2 age) and adult (3-5-i- age) and each sample (scat 

 or spew) was categorized as containing juvenile, adult, 

 or mixed (juvenile and adult) pollock. Multidimensional 

 contingency tables with island, sample type, and age 

 category as variables were used to test interactive ef- 

 fects among variables. A saturated model including 

 all variables and interactions was compared with re- 

 stricted models by using chi-square goodness-of-fit test 

 (S-PLUS 2000, Insightful Corp., Seattle, WA). Samples 

 were pooled if sample type or island effects were condi- 

 tionally independent, and two-dimensional contingency 



tables were then used to test variables that were not 

 independent. 



The size range of Gb-Bm consumed by northern fur 

 seals was estimated by measuring the rostral length 

 of lower beaks recovered from scat and spews. Rostral 

 length was measured to the nearest 0.1 mm with an 

 optical micrometer. Because cephalopod beaks are more 

 resistant to digestion and to subsequent loss of length 

 than are otoliths (Sinclair et al., 1996; Tollit et al., 

 1997), Gb-Bm lower beaks were not assigned condition 

 grades prior to being measured. However, lower beaks 

 showing excessive wear, such as a broken rostral tip, 

 were not measured. 



To evaluate Gb-Bm prey size differences between 

 sample type and islands, we developed (using combined 

 samples of both species) regression equations for low- 

 er beak rostral length (LRL) against dorsal mantle 

 length (DML) and for DML against weight. Squid speci- 

 mens were opportunistically collected from commercial 

 pollock trawl fishery bycatch, research driftnets, and 

 NOAA research vessel mid-water trawl operations be- 

 tween 1979 and 2000. Sampling areas were broad rang- 

 ing throughout the North Pacific, at numerous localities 

 in the eastern Bering Sea, Gulf of Alaska, and subarctic 

 Pacific Ocean south of the western Aleutian Islands 

 (WalkerM. 



The regression of LRL against DML for Gb-Bm was 

 developed by using 757 lower beaks (n = 482 Gb, 275 

 Bm) with a dorsal mantle length range of 21-386 mm 

 and the DML-weight regression was developed by using 

 1676 lower beaks (/? = 1048 Gb, 628 Bm) with a dorsal 

 mantle length range of 17-386 mm. The regression in- 

 cluded size ranges of Gb-Bm found in northern fur seal 

 scats and spews examined in the present study. Linear 

 models were used to develop regressions of LRL (mm) 

 against DML (mm) and DML against weight (grams): 



DML = 39.37(Li?L)- 0.50. 

 Log(jmg/!n = 2.87(Log( DML)) -4.16. 



A high degree of correlation (LRL to DML P<0.001; 

 /•'- = 0.98; SE = 0.51); DML to weight P<0.001; ^2 = 0.99; 

 SE = 0.01) was found for both regression equations 

 (Walker^). Gb-Bm DML data were log transformed and 

 differences in DML between sample types and island 

 were determined by comparing means with a two-sample 

 f-test. Although the DML-weight regression is not used 

 in the analysis, it is included here for future use by other 

 researchers conducting bioenergetics studies. 



Prey age or size estimations were limited to wall- 

 eye pollock and Gb-Bm because sufficient numbers of 

 otoliths or beaks of other primary fish and cephalopod 

 prey species were not recovered from scat and spew 

 samples, or because regressions for the species were 

 unavailable. 



Walker, W. A. 2004. NMFS database. Alaska Fisheries 

 Science Center, National Marine Fisheries Service, NOAA, 

 7600 Sand Point Way N.E., Seattle, WA 98115. 



