22 
Fishery Bulletin 111(1) 
predation. Additional research to verify and refine our 
estimates of diet composition, and to begin quantifying 
rockfish population dynamics and the influence of pin- 
niped predation through incorporation of information 
on harbor seal consumption rates (Howard, 2009; How- 
ard et ah, 2013) is warranted. 
Although rockfish species appear to constitute a 
more foundational prey resource for harbor seals than 
was recognized previously, harbor seal diets do not ap- 
pear to be homogeneous, a finding that is consistent 
with the results of observational studies of preda- 
tory behavior (Suryan and Harvey, 1998; Tollit et al., 
1998; London, 2006; Wright et al., 2007; Hardee, 2008; 
Thomas et al., 2011; Peterson et al., 2012). Substan- 
tial spatial heterogeneity in diet composition was de- 
tected among seals from the 4 sampling locations. For 
example, the mean diet of seals sampled near Padilla 
Bay was dominated by Shiner Perch, a common spe- 
cies in bays and estuaries throughout the west coast 
of North America (Hart, 1973). Seals sampled from the 
other locations, which are characterized by deeper and 
more open waters and greater rocky relief, tended to 
rely more on species of rockfish and salmon and Pa- 
cific Herring. Spatial patterns of habitat suitability un- 
doubtedly underlie the relative abundance of prey in 
local areas — a dynamic that is subsequently reflected 
in seal diets. Heterogeneity among sexes also was ob- 
served; a more diverse diet and greater use of rockfish 
species and Spiny Dogfish were observed for male seals 
than for females. Sex-based heterogeneity in diet was 
not expected, given the slight sexual dimorphism in 
harbor seals, but it may reflect a number of factors, in- 
cluding intersexual competition for food resources, for- 
aging behavior, predatory efficiency, and differences in 
reproductive investment. For example, reproductively 
active females tend to make shorter foraging trips dur- 
ing early lactation (Boness et al., 1994) — behavior that 
may reduce their access to some prey classes. 
Although the sampling location and sex covariates 
explained primary patterns among estimates of seal 
diet composition, substantial unexplained heterogene- 
ity was observed in the estimates. In particular. Black 
and Yellowtail Rockfish were among the most impor- 
tant prey species for a number of individual seals, es- 
pecially males, but they were absent from the diets of 
other seals. Whether differences between individual 
seals could be explained by unmeasured covariates or 
are attributable to individual preference or specializa- 
tion is unknown. In either case, this heterogeneity with 
respect to rockfish predation is an intriguing aspect of 
the results of this study. 
Our estimates of mean diet composition are not 
thought to provide an accurate assessment of harbor 
seal diets on an annual basis. Most seals were sam- 
pled in the spring (Table 1), and no seals were sampled 
from late May through late October. One would expect 
season to be an important covariate that could explain 
differences in diets, especially given the large changes 
in the relative abundance of prey during the spring 
spawning migration of Pacific Herring and the summer 
availability of migrating adult salmon species (Stasko 
et ah, 1976; Willson and Womble, 2006; Therriault et 
al., 2009; Thomas et al., 2011). We surmise that such 
temporal heterogeneity exists, but that evidence of 
these seasonally available prey species in harbor seal 
blubber was diminished by late October. The lack of 
summer seal samples may partially explain the differ- 
ence between these results and assessments of harbor 
seal diet based on scats, in which salmon species and 
Pacific Herring are prevalent (Luxa, 2008; Lance et al., 
2012). A complete assessment of seasonal variation in 
harbor seal diets would require a somewhat expanded 
investigation, in which the distribution of sampling ef- 
fort would be designed to investigate potential changes 
in diet expected on the basis of seasonally predictable 
shifts in the availability of prey species. The expected 
deposition and turnover rates of fatty acid compounds 
in adipose tissue (Nordstrom et al., 2008) also would 
contribute importantly to an optimized sample design. 
On the basis of the results of this investigation, an ex- 
panded effort to more fully explore spatial, temporal, 
and demographic patterns in harbor seal diets likely 
would be successful. 
Two estimates of mean diet composition, one unad- 
justed and one adjusted for differential fat mass of prey, 
were provided for all seals combined (Fig. 2). However, 
no adjustment for differential fat mass was made for 
the estimates stratified by location and sex. The large 
differences in fat composition among the prey classes 
(Table 2) and, to a lesser extent, the lack of total mass 
data for mature Chinook, Sockeye, and Pink Salmon, 
all of which have high fat content, somewhat reduce 
our confidence in the fat-adjusted estimates. The es- 
timates unadjusted for differential fat mass are infor- 
mative ecologically, providing information on the likely 
sources of adipose tissue ingested by harbor seals. Fat- 
adjusted estimates may be of greater interest from the 
perspective of prey population demographics because 
rescaling the estimates with mean fat per prey con- 
verts the units to the relative numbers (proportions) of 
prey animals consumed. Given an estimate of the num- 
ber of fish consumed per unit of time, the fat-adjusted 
estimates would facilitate the investigation of preda- 
tion rates by prey class. 
Although QFASA is a powerful method for investi- 
gation of predator diets, it is important to recognize 
potential problems with its use. With respect to marine 
mammals, logistical constraints and permit require- 
ments may limit sample sizes and preclude comprehen- 
sive investigations of free-ranging populations. From a 
statistical perspective, it is important to acknowledge 
that estimates of diet composition are conditioned on 
the calibration coefficients, the suitability of which in 
any particular application cannot be verified. In the in- 
stance of this investigation, the calibration coefficients 
were estimated during a controlled feeding study of 
captive harbor seals (Nordstrom et al., 2008), the spe- 
cies of interest. Even so, the degree to which the coef- 
