Laake et al : Pinniped diet composition 



443 



a: 

 m 



6 

 u. 



Median weight = 173g 



•  



10 



 1000 



10000 



Adult ctiinook 



Equivalent estimates 



Weight (g) on log 10 scale 



Figure 3 



Comparison of spring consumption estimates for harbor seals on the Columbia River based on split-sample 

 frequency of occurrence (SSFO) and biomass reconstruction (BR) in relation to prey weight. Similar estimates 

 were obtained for prey near the median weight. SSFO increases estimates of smaller prey items (e.g. smelt) and 

 decreases estimates of larger prey items (e.g. adult chinook salmon) in relation to BR. 



We have assumed an equal probability of recovering 

 identifiable skeletal remains from all prey sizes and that 

 masses predicted from otolith measurements represent 

 prey identified from other structures. There are several 

 situations when recovered otoliths may not correctly rep- 

 resent the size of prey consumed. Small otoliths from 

 small individuals of a species may be more likely to be 

 completely digested. In that case, biomass would be over- 

 estimated because the larger otoliths would be recovered 

 from larger fish. Also, otoliths may have different passage 

 rates due to changes in their structure as fish age result- 

 ing in a size bias. Consumption estimates could thus be 

 erroneously high or low because prey would be calculated 

 from a single size group rather than the range of prey 

 consumed. Unequal digestion may also create errors in 

 estimated mass from the otoliths that are recovered. Our 

 estimated mass may have been biased because we applied 

 an average degradation factor to adjust otolith length. A 

 better alternative would be to grade otolith condition and 

 apply condition-specific degradation factors (Tollit et al., 

 1997b). 



Counting the number of individual prey items is not 

 possible with nonunique bones. Instead, we estimated a 

 minimum number of individuals (MNI) contained in the 



scat {7!,^=MNI). MNI from all skeletal elements is a mini- 

 mum estimate because the presence of many nonunique 

 structures are assigned to a single prey item when they 

 could represent several different prey. This error would 

 not bias diet composition if it did not vary over species and 

 size. However, differential passage of unique structures 

 and identifiability among species result in the greater 

 probability of detecting some prey (Browne et al., 2002). 

 To minimize interspecific biases, we could use an MNI of 

 1 for all bones, regardless of the enumeration of unique 

 structures for some species. While this would reduce some 

 problems with species differences, it would exacerbate 

 differences associated with prey size because large prey 

 would be accurately reflected by an MNI of 1 and small 

 prey that are eaten in greater quantities would be se- 

 verely underestimated. 



Bowen (2000) proposed estimating the number of prey 

 consumed by correcting the otolith count with rates of oto- 

 lith recovery from feeding trials. These correction factors 

 vary widely between seals and studies and are influenced 

 by a variety of factors, including size of individual prey, 

 meal size, and activity of harbor seals (Harvey, 1989; Har- 

 vey and Antonelis, 1994; Cottrell et al., 1996; Tollit et al., 

 1997b). Browne et al. (2002) examined the ratio of otolith- 



