244 



Fishery Bulletin 105(2) 



Hexagrammid — 

 Cephalopod 



Gadid 



zy 



Rookeries 



Forage 

 Salmon 



3 



Rockfish 

 Flatfish 

 Other  



:d- 



Hexagrammid 



Cephalopod 

 Gadid 



J- 



Rockfish 



Forage 

 Salmon 



IF 



Haulouts 



0.35 



40 



45 



50 



0.55 



Distance between clusters 



Figure 8 



Hierarchal clustering trees of prey groups found in Steller 

 sea lion (Eumetopias jubatus) scats from three Southeast 

 Alaskan rookeries (Forrester, Hazy, and White Sisters) in 

 summer (Jun-Aug) and from haulouts the rest of the year 

 (Sep-May). Distances between pairs of prey groups were 

 defined as (l=partial correlation coefficient)l2 and cluster- 

 ing was done using "hclust" software (S-Plus 4.0) (MathSoft 

 Inc., Cambridge, MA). 



the consistency of the annual summer diet at Forrester 

 Island (which was sampled in all years. Fig. 2) and on 

 the relative stability of ocean conditions during the 

 1990s (which is believed to determine the relative abun- 

 dances of suites of prey available to sea lions and other 

 species; Benson and Trites, 2002; King, 2005; Trites et 

 al., 2007). Thus we feel that the seasonal description 

 of diet accurately reflects what Steller sea lions were 

 eating in Southeast Alaska during the 1990s. 



A bias could have been introduced in our analysis if 

 the sizes of collected scats differed significantly between 

 haulouts and rookeries, and if the number of prey spe- 

 cies recovered was correlated with the size of scats. 

 In checking this potential source of error, we found 

 numbers of prey types per scat were positively related to 

 size for scats <250 mL but were independent of scat size 

 beyond this volume (Fig. 7). In general, there was little 

 difference in the sizes of scats and numbers of species 

 recovered from rookeries and haulouts, and thus there 

 was no apparent effect of size of scat on our results. 



By inferring the dietary importance of prey species 

 from their frequencies of occurrence in scats one im- 

 plicitly assumes that the probability of finding prey in 

 scats is proportional to the number or mass of that prey 

 consumed, and that this proportionality does not vary 













<<} 



/ 



/ 





2- 



-40% 



Population change (1990-1994) 



Figure 9 



Diet diversity versus rate of population change 

 of Steller sea lions iEumetopias jubatus) between 

 1990 and 1994 (total number of adults and juveniles 

 counted on rookeries during the summer; Strick et 

 al., 1997) by region of Alaska. The fitted line is a 

 least squares linear regression. Diet diversity was 

 calculated using split-sample frequency of occurrence 

 and a Shannon-Wiener index. The locations and all 

 the data used to calculate diet diversity (except for 

 Southeast Alaska) were taken from Merrick et al. 

 (1997). Data from Southeast Alaska reflect the mean 

 summer rookery diet of Steller sea lions and rockfish 

 were grouped with the "other" prey type. 



among prey species. However, controlled feeding studies 

 with captive Steller sea lions, and other pinnipeds, have 

 shown that the types and proportions of prey hard parts 

 that pass through the digestive tract vary, depending 

 on the species of prey and the size of prey (Cottrell and 

 Trites, 2002; Tollit et al., 2003; Tollit et al., 2004b). 

 Pollock bones, for example, tend to be more robust than 

 the bones of other species and have a higher likelihood 

 of being recovered than the more fragile bones of other 

 species, such as sand lance. However, smaller schooling 

 species, such as sand lance, are likely to be consumed 

 in higher numbers than pollock, and the greater con- 

 sumption of these smaller species would increase the 

 likelihood of some of the smaller bones passing through 

 the digestive tract. Thus, the probability of detecting 

 different prey species in scats can vary. Recording the 

 presence of all identifiable hard parts as we did (i.e., 

 not relying only on otoliths) significantly reduces the 

 likelihood of any species passing undetected (Cottrell 

 and Trites, 2002). 



Captive feeding experiments indicate that the aver- 

 age scat probably contains the remains of prey con- 



