156 



Fishery Bulletin 101(1) 



reported by Sinclair and Zeppelin (2002) suggest 

 that the dominant prey in more recently collected 

 samples from the Gulf of Alaska and Aleutian Is- 

 lands in summer (breeding females on rookeries) 

 and winter (juvenile and adult males and females 

 on nonbreeding haul-outs) were similar to those 

 reported by Merrick et al. (1997). Sinclair and Zep- 

 pelin (2002) found walleye pollock more frequently 

 than any other prey species in Steller sea lion scats 

 from the Gulf of Alaska and eastern Aleutian Is- 

 lands during the summer and winter, whereas Atka 

 mackerel was the second most frequently occurring 

 prey species in the eastern Aleutian Islands, and the 

 most frequently occurring prey species in the central 

 and western Aleutian Islands during summer and 

 winter. However, Sinclair and Zeppelin (2002) did 

 find significant seasonal changes in the frequency 

 of certain prey species in the diets of Steller sea 

 lions at specific sites in Alaska. For example. Pacific 

 cod occurred more frequently in scats collected on 

 haul-outs during the winter than in scats collected 

 on rookeries during the summer in all regions (Gulf 

 of Alaska through western Aleutian Islands). Pacific 

 salmon occurred more frequently in the summer 

 than in the winter in the Gulf of Alaska and the 

 eastern Aleutian Islands, while this seasonal dif- 

 ference was reversed in the central and western 

 Aleutian Islands. A seasonal change in the propor- 

 tion of the diet comprising high-energy prey species 

 like salmon can have a substantial effect on the total 

 and per capita amount of food biomass required by 

 Steller sea lions (e.g. our results for southeast Alas- 

 ka). Nevertheless, given the similarities between the 

 two summer and winter data sets (from Sinclair and 

 Zeppelin, 2002, and Merrick et al, 1997), and given 

 the level of uncertainty that we incorporated in our 

 diet compositions, we feel that the diet compositions 

 that we assumed for Steller sea lions in the western 

 Aleutian Islands through Gulf of Alaska regions in 

 1998 were reasonable approximations. 



The scat data we used to estimate the diet of Steller sea 

 lions in southeast Alaska had better temporal and demo- 

 graphic coverage. Trites and Calkins'* reported data from 

 scats collected in every month except September Although 

 the scat data from the summer months were again from 

 breeding females on rookeries, the scat data from the rest 

 of the year were from nonbreeding animals on haul-outs. 

 Animals using these haul-outs included adult and juvenile 

 males and females. Thus, those scats were more represen- 

 tative of the average diet of the population than scats 

 collected on rookeries during the breeding season. There 

 is evidence that the diet of mature females on rookeries 

 differs from the diet of nonbreeding animals on haul-outs 

 during the summer (Trites and Calkins'), but it is difficult 

 to translate this difference into sex- or age-specific dietary 

 differences. 



In addition to sampling limitations, there are at least 

 two other potential biases associated with using scat 

 data to assess diet composition. The first potential bias 

 is the possibility that some of the consumed prey species 



15 

 10 



5 





 15 



10 



5 





 30 



20 



10 





 90 



60 



30 



60 



40 



20 





 30 

 20 

 10 





 30 



20 



10 



1234567 1234567 



Region of Alaska 



Figure 5 



Annual prey biomass consumption for Steller sea lions in 1998 in 

 the seven study areas of Alaska. Values for southeast Alaska were 

 estimated by using diet information for all four seasons. Error 

 bars represent ±1 SD (obtained by using Monte Carlo simula- 

 tions-1000 runsl. 



were not represented in the scat samples (Bowen, 2000). 

 Although the identification of prey hard parts other than 

 otoliths in scats increases the probability of detection of 

 prey species, cartilaginous fish or fish with small or fragile 

 bony structures may be completely digested and not evi- 

 dent in scat (Olesiuk et al., 1990). For example, in captive 

 Steller sea lion feeding trials, the average number of hard 

 parts recovered in scat was 31.2 per pollock, but only 7.9 

 per herring (Cottrell and Trites, 2002). Thus, there was 

 a greater chance of an individual herring being missed 

 compared to an individual pollock. However, small fish are 

 likely consuined in larger numbers, which would increase 

 the likelihood of detecting their presence in scat. 



The second potential source of dietary bias arises from 

 using the "split-sample frequency of occurrence" technique 

 to estimate the percentage of biomass that different prey 

 represent in the diet. This technique assumes that the 

 prey identified in a scat sample represent all the prey con- 

 sumed in a meal, and that all prey species of a meal arc 

 consumed in equal biomasses (Olesiuk et al., 1990). This 



