Winship and Trites Prey consumption of Eumetopias /ubatus 



151 



(percentage of biomass that each prey category repre- 

 sented in the diet). These distributions had medians from 

 Table 2 and upper and lower limits equal to ±45% of me- 

 dians >10%, or ±98% of medians <10%. These percentages 

 were then standardized so that all prey categories were 

 summed to 100% for a given diet. The ranges of the as- 

 sumed errors in diet composition were determined by us- 

 ing estimates of the minimum and maximum split-sample 

 frequencies of occurrence of prey categories (Olesiuk et al., 

 1990; Olesiuk, 1993; see "Discussion" section). 



The energy density offish is a function of their chemical 

 composition, especially their lipid content (Stansby, 1976; 

 Hartman and Brandt, 1995). Thus, the energy density 

 of fish can vary with age (older fish tend to store more 

 lipid; Brett, 1983; Harris et al., 1986; Paul et al., 1998a), 

 season (lipid content can vary with foraging conditions; 

 Paul et al, 1993; Paul et al., 1998a; Robards et al., 1999), 

 reproductive status (lipid content of spawning fish can be 

 different from nonspawning fish; Dygert, 1990; Smith et 

 al., 1990; Hendry and Berg, 1999), and geographic loca- 

 tion (feeding conditions can vary with location; Paul and 

 Willette, 1997; Lawson et al., 1998; Paul et al., 1998b). The 

 quantity and resolution of data on the energy density of 

 prey of Steller sea lions varied depending on the prey spe- 

 cies (Appendix I). When detailed season-specific energy- 

 density data were available for prey species, we generally 

 incorporated seasonal changes in energy density. Unfortu- 

 nately, no detailed geographic-specific energy density data 

 were available for any prey species; therefore we assumed 

 that the energy density of prey did not vary among regions 

 of Alaska. We used relatively wide ranges of possible en- 

 ergy-density values for all prey in order to incorporate the 

 uncertainty in how energy density varies with season and 

 geographic location. 



Many data were available on the energy density of for- 

 age fish (Appendix I). The energy densities of forage fish 

 species are relatively high, but vary seasonally in relation 

 to spawning periods, the over-winter fast, and spring and 

 autumn phytoplankton blooms (Anthony et al., 2000). For 

 example, eulachon had a very high energy density (7.5- 

 11.1 kJ/g wet mass) and its energy density was slightly 

 higher in the summer than in late winter (Payne et al., 

 1999). The energy density of capelin was lower, ranging 

 from 3.5 to 7.0 kJ/g wet mass. In the Gulf of Alaska, the 

 energy density of capelin was high in June (start of spawn- 

 ing) after the spring phytoplankton bloom and decreased 

 through the summer with advancing reproductive stage 

 (Anthony et al., 2000). The energy density of capelin in- 

 creases again in the fall and early winter as the fish feed 

 on the autumn phytoplankton bloom (Lawson et al.. 1998; 

 Payne et al., 1999; Anthony et al., 2000). 



Pacific herring and Pacific sandlance were the two main 

 forage fish species consumed by Steller sea lions in Alaska 

 in the 1990s (Merrick et al.. 1997, Trites and Calkins^). 

 The energy density of Pacific herring increased with age, 

 and the energy density of adults (age >0) ranged from 

 4.4-11.7 kJ/g wet mass (Appendix I). In the Gulf of Alaska, 

 Pacific herring were highest in energy content in the au- 

 tumn and lowest in energy content in the spring (after the 

 overwinter fast; Paul et al., 1998a), but the exact timing of 

 these seasonal changes varied depending on the region of 

 Alaska (Perez, 1994). Pacific sandlance (age >0) ranged in 

 energy density from about 3.2 to 6.1 kJ/g wet mass. Pacific 

 sandlance from the Gulf of Alaska was highest in energy 

 content in June (after the spring bloom), and its energy 

 content decreased through autumn (spawn mid-autumn) 

 and remained low throughout the winter fasting period 

 (Robards et al., 1999; Anthony et al., 2000). We assumed 



