Winship and Trltes: Prey consumption of Eumetopias /ubatus 



155 



24 



™ 22- 



20 - 



18 



16 



14 - 



6 



Western Central Central Central 



Aleutians Aleutians 3 Aleutians 2 Aleutians 1 



Eastern 



Aleutians 



Gulf of 



Alaska 



Southeast 

 Alaska 



12 3 4 5 6 7 



Region 



Figure 4 



Estimated average daily food biomass requirements per individual Steller sea 

 lion (all ages combined) from seven regions of Alaska based on summer diet 

 information (filled circles). For southeast Alaska, the estimated per capita 

 food requirement based on diet information for all four seasons is also shown 

 (unfilled circle). Error bars represent ±1 SD (obtained by using Monte Carlo 

 simulations-1000 runs). 



mid biomass of any region (41,000 |±8070] t). CVs of indi- 

 vidual prey category consumption ranged from 20 to 399c. 



Discussion 



Uncertainty in model predictions 



An important aspect of our model is that it produces distri- 

 butions of predicted food requirements rather than point 

 estimates (Winship et al.. 2002). This allowed us not only 

 to estimate mean predicted food requirements but also to 

 estimate the potential error in these mean predictions by 

 using either SD or C V ( SD as a percentage of the mean ). We 

 found that the CVs of mean predicted food requirements 

 at the population level (both total biomass and individual 

 prey categories) were approximately 20-40%. The ranges 

 of food requirements predicted by the model were of course 

 much wider than ±1 CV. For instance, 5% of the predicted 

 values lie beyond ±1.96 CV if the normal distribution is 

 used to approximate the distribution of model predictions. 

 The minimum and maximum food requirement estimates 

 predicted by the model were generally <40'7f of the mean 

 and >160% of the mean, respectively, if the CV of the mean 

 predicted food requirement was 30%. Thus, the ranges of 



predictions produced by our model reflect considerable 

 uncertainty in the food requirements of Steller sea lions 

 because of the assumed errors that we attributed to cer- 

 tain bioenergetic parameters (e.g. metabolic rate at sea), 

 population parameters (e.g. age- and sex-specific survival 

 rates), and diet parameters (e.g. diet composition). Future 

 research on key parameters in our model will help to refine 

 the predictions of this model and improve the accuracy of 

 estimates of the food requirements of this species. 



Biases in diet composition 



The diet compositions that we used were estimated from 

 the hard parts of prey found in scats collected on haul-outs 

 and rookeries and were limited by incomplete sampling 

 coverage by time of the year and by sex- and age-class. Diet 

 data from the western Aleutian Islands through the Gulf 

 of Alaska came mainly from mature females on breeding 

 rookeries during late June, early July, and early August 

 in the early 1990s (Merrick et al., 1997). Thus, these data 

 reflect a specific segment of the population during a short 

 period of the year. 



We applied the diets reported by Merrick et al. (1997) 

 to all age- and sex-classes of Steller sea lions in 1998 and 

 assumed that those diets did not change seasonally. Data 



