436 



Fishery Bulletin 100(3) 



minimum number of prey («,) consumed. For some hard 

 parts, it may not be possible to identify the species, but 

 the hard part can be classified to a group of species, such 

 as family. In these cases, the species can be grouped into 

 a single unit for estimation (e.g. all hexagrammids) or the 

 unidentified prey can be partitioned to species based on 

 the sample of species-specific hard parts. For example, 

 most salmonid bones are currently indistinguishable to 

 species; therefore salmonids in scats represented by bones 

 (u) can be apportioned into species from proportions ()^) 

 obsei-ved from otoliths (o, ). If we denote f to be the set of 

 all identified salmonids represented by otoliths, the num- 

 ber of salmonid prey in species ; can be estimated as 



where 



-o, +u 7, for / e f . 



y, = =i^ for (■ 6 T 



(4) 



(5) 



Pinniped diet can be quite variable in response to 

 prey availability (Pierce et al., 1991; Tolht et al., 1997a; 

 Browne et al., 2002; Beach et al.'; Brown et al.^). Because 

 scat collected at a single date may reflect what was lo- 

 cally available at that time, collecting scats at different 

 times throughout a season or year will provide a better 

 representation of diet. From scats collected across several 

 occasions, what is the best way to estimate average diet 

 composition? If the amount of scat collected and resulting 

 reconstructed biomass for an occasion is proportional to 

 the amount of prey consumed, the data should be pooled 

 for a single ratio estimate. However, in many cases the 

 amount of scat collected will reflect a multitude of factors, 

 such as tide height, storms, human disturbance, and the 

 length of time that the seals were at the haulout prior to 

 collection. Also, in many circumstances a fixed number of 

 scats are collected rather than some fixed proportion of 

 the scats available at the haulout. Therefore, we suggest 

 that an average of the proportions (ratios) is appropriate. 

 If there are T occasions, the average proportion is 



Prey mass can be determined from morphometric re- 

 lationships between hard part dimensions (e.g. typically 

 otolith length) and mass. Adjustments should be made to 

 account for partial digestion of the hard part. From regres- 

 sions between otolith length (corrected for degradation) 

 and fish length and between fish length and mass, an esti- 

 mate of the average mass of prey represented in scat can 

 be constructed (Harvey et al., 2000; Browne et al., 2002). 

 In some scats, a prey species may be represented only by 

 hard parts (e.g. gillrakers) that do not have a quantifiable 

 relationship to mass. We have to assume that the prey 

 with unknown mass are represented by the average mass 

 determined from the measurable hard parts (e.g. unbro- 

 ken otoliths) of that species. 



Thus, the biomass cannot strictly be measured but must 

 be estimated by using estimated average mass (w^) and 

 estimated number of individual prey (r\): 



X'^" 



(6) 



b, = n, -^ 



where o* = the number of hard parts (typically unbroken 

 otoliths); and 

 ii',, = the estimated mass derived from the regres- 

 sions relating otolith length to fish mass. 



To estimate number of prey consumed (P, ) rather than bio- 

 mass, estimated biomass is divided by average mass: 



w, w, 



(7) 



where p, 



the number of prey species / consumed 

 per unit of biomass consumed. 



(8) 



Often there are seasonal shifts in diet resulting from prey 

 availability (Olesiuk et al., 1990; Tollit and Thompson, 

 1996; Browne et al., 2002). In such cases, the analysis 

 should be stratified by season. If data are collected over 

 several years, again a simple average of the seasonal 

 proportions is appropriate. In the appendix, we provide 

 variance estimators for diet composition and consumption 

 estimates for data collected over several years stratified 

 by season. 



Data collection and analysis 



For the data used in this paper, Browne et al. (2002) have 

 described the scat collection and analytical methods and 

 have provided descriptions of the food habits from these 

 data. As in Browne et al. ( 2002 ), we stratified our data col- 

 lection and analysis into three seasons, spring ( 1 March- 

 14 May), summer (15 May-15 July) and fall (16 July-15 

 October), based on the timing of chinook salmon runs at 

 the Bonneville Dam, offset by two weeks to account for the 

 travel of salmonids from the lower Columbia River to the 

 Dam (at river km 235). We describe here additional meth- 



' Beach, R. J., A. C. Geiger, S. J. Jefferies, S. D. Treacy, and B. L. 

 Troutman. 1985. Marine mammals and their interactions 

 with fisheries of the Columbia River and adjacent waters, 

 1980-1982. NWAFC Processed Rep. 85-03, 316 p. Northwest 

 and Alaska Fisheries Sci. Cent., Nat. Mar. Fish. Serv., NOAA, 

 7600 Sand Point Way NE, Seattle, WA 98115. 



2 Brown, R. F, S. D. Riemer, and S. J. Jeffries. 1995. Food of 

 pinnipeds collected during the Columbia River Area Commer- 

 cial Salmon Gillnet Observation Program, 1991-1994. Wildlife 

 Diversity Program Tech. Rep. 95-6-01, 16 p. Oregon Depart- 

 ment of Fish and Wildlife, 2501 SW 1st Ave., PO Box 59, Port- 

 land, OR 92707. 



