Laake et al ; Pinniped diet composition 



435 



based on analysis of scat colloctions. During 1995-97, 

 harbor seal scats were collected at the Dusdemona Sands 

 haulout from 1 March to 15 October. By combining diet 

 composition obtained from scat analysis and contempo- 

 raneous surveys of seal abundance, we estimated the 

 average consumption of salmonids and other prey in the 

 Columbia River by harbor seals during spring, summer, 

 and fall of 1995-97. 



In our study, we focussed on the method of estimating diet 

 composition from a sample of scats. We describe an estima- 

 tor for diet composition based on reconstruction of the prey 

 biomass represented in the scat and show how it is related 

 to an alternative estimator described by Olesiuk (1993). 

 Using the data collected on harbor seals in the Columbia 

 River, we demonstrate the sensitivity of the consumption 

 estimates to the method for estimating diet composition. 



Materials and methods 



parts may be affected by factors that influence digestion 

 and deposition (e.g. seal activity level). Also, it may not 

 always be possible to collect an entire scat or even reason- 

 ably define a scat as a discrete entity. Thus, even though 

 consumption by seals varies to some degree, the biomass 

 reconstructed from a scat may vary much more than the 

 variation in consumption. Thus, one could argue that each 

 scat should be treated as a "representative" variable-size 

 sample of a nearly constant amount of biomass consumed 

 during some feeding interval. With that conceptual sam- 

 pling model, the most appropriate estimator would be a 

 simple average of the proportions in each scat: 



I*^' 



X "'*"''* 



(2) 



Diet composition models 



If we could randomly select a sample of /( prey items con- 

 sumed by pinnipeds, a ratio estimator (Cochran, 1977) 

 would be appropriate to estimate the proportion of bio- 

 mass (./r,) represented by the ;'*' prey species from w pos- 

 sible prey species: 



where 6^^ = the biomass of species /; 



«,;(. = the number of species i consumed; and 



u\f, = the average mass of species / in the k*^ scat. 



Equation 2 is similar to the estimator of Olesiuk (1993), 

 which he called split-sample frequency of occurrence 

 (SSFO): 



Z^' S"'"' 



(1) 



where 6, = the total biomass of the n^ prey items that are 

 species /; 

 w^ = 6, /«, = the average mass for species i; and 



"=X"'- 



(3) 



where /^^ = an indicator variable which equals 1 if one (or 

 more) prey items of species ; is in scat k, and 

 otherwise. 



Prey hard-parts in a scat represent a filtered selection 

 of the prey species that were consumed by a single animal 

 over some unknown and variable amount of time. Captive 

 feeding studies have shown that a scat does not represent 

 a single meal or even a single fixed period of feeding time 

 (Harvey, 1989). Moreover, pinnipeds are unlikely to con- 

 sume the same amount of prey in each meal or in a speci- 

 fied amount of time. Therefore, the biomass represented 

 by the prey remains in a scat is unlikely to be constant 

 because consumption varies. From a collection of s scats in 

 which each scat represents a variable amount of biomass 

 that is proportional to consumption, the ratio estimator 

 (Eq. 1) is also appropriate, where 6, is the total biomass of 

 species ; from the s scats. We will refer to (Eq. 1 ) as biomass 

 reconstruction (BR), which is equivalent to the estimator 

 used by Harvey (1988) and Hammond and Rothery (1996). 



Alternatively, one could argue that the biomass recon- 

 structed from prey remains in a scat may vary for numer- 

 ous reasons other than consumption. Variation in scat 

 volume and production and the resulting amount of hard 



Equation 2 is equivalent to 3 when you make Olesiuk's 

 (1993) assumption that an equal amount of biomass of 

 each species in the scat was consumed. SSFO requires 

 only a determination of the presence or absence of the 

 prey in a scat, and thus, it is much easier to implement 

 than either Equations 1 or 2, which require an enumera- 

 tion of the individual prey in each scat and their mass. 



Enumeration of prey in a scat sample is straightforward 

 with unique structures such as otoliths. However, by using 

 nonunique hard parts (e.g. gillrakers, vertebrae) to reduce 

 selection bias resulting from unequal digestibility of oto- 

 liths, problems are introduced with enumeration. When 

 prey are exclusively represented in scat by nonunique 

 structures, it may be possible only to determine that a sin- 

 gle individual was consumed or at best a minimum num- 

 ber can be constructed by enumerating nonunique struc- 

 tures and dividing by the average number of structures 

 per fish (e.g. count of vertebrae divided by average number 

 of vertebrae per fish). By including nonunique structures, 

 enumeration of prey is replaced with an estimate of the 



