Wilson et at: Regional variation in the feeding cycle of juvenile Theragra chalcogramma 
319 
Data analysis 
Stomach content data were used to estimate food con- 
sumption and diet composition. The amount of food 
consumed was assessed by using SCW normalized to 
somatic body weight (whole-body wet weight less stom- 
ach content weight) and expressed as percent somatic 
body weight (%BW). Diet composition was quantified by 
using prey counts or weights summed across fish and 
then expressing them as a percentage of total prey count 
or weight (%No., %W, respectively). We also computed 
the frequency of occurrence of a particular prey type 
as the percentage of all stomachs containing that par- 
ticular prey type (%FO). The cumulative number of prey 
types detected was examined in relation to the number 
of fish examined, as in Ferry and Cailliet (1996). We did 
not incorporate estimates of fish catch in our computa- 
tions (e.g., cluster sampling estimator [Buckel et al., 
1999]) because of uncertainty about how to standardize 
sampling effort among the different gears, sampling 
objectives, and methods used during the various cruises. 
We conducted a modeling exercise to integrate our 
observations on feeding with previous findings on body 
condition (Buchheister et al., 2006) and prey energy 
density (Mazur et al., 2007) to explore the implications 
for juvenile walleye pollock growth. The exercise con- 
sisted of first estimating daily ration (DR), and then 
inputting it and other empirical data into a bioenerget- 
ics model to output fish growth rate. Daily ration was 
estimated empirically with a simple evacuation rate 
model (Elliott and Persson, 1978): 
C = 24 E S, (1) 
where C - daily food consumption (%BW/d); 
24 = the number of hours in a day; 
E = the instantaneous rate of evacuation 
(%BW/h); and 
S = average SCW over the course of the day. 
We used £=0.28 %BW/h (Merati and Brodeur, 1996). S 
was computed as the average of mean SCW for each 3-h 
time bin (e.g., 0-3, >3-6, >6-9 h) (Elliott and Persson, 
1978). All SCWs were arcsine transformed so that the 
errors approximated a normal distribution. 
Growth rate was estimated with an age-0 walleye 
pollock bioenergetics model (Ciannelli et al., 1998). 
Model inputs were fish body weight (g), diet (%W), DR 
(%BW), predator energy density (J/g, Buchheister et 
al., 2006), taxon-specific prey energy densities (J/g, 
Mazur et al., 2007), and water temperature (°C). Fresh 
fish body weight was estimated from mean whole wet 
weight of thawed fish by using fresh-frozen weight 
relationships (Buchheister and Wilson, 2005). Preda- 
tor energy density was estimated by using fish body 
weight-energy density relationships (A. Buchheister, 
unpubl. data, but see Buchheister et al., 2006). Water 
temperature at a depth of 40 m was measured with 
a calibrated SBE-19 or SBE-39 temperature profiler 
(Sea-Bird Electronics, Bellevue, WA), or a microbathy- 
thermograph (Richard Brancker Research, Ltd., On- 
tario, Canada) and averaged across sites where fish 
were collected. Model output growth rates were based 
on estimates of excess daily consumption of prey in 
grams of food per gram of body weight per day (g/g/d) 
after budgeting for egestion, excretion, respiration, and 
specific dynamic activity. We used the length-weight 
relationships from Buchheister et al. (2006) to convert 
growth rate units to mm SL/d for direct comparison 
with otolith-based growth rates. 
Results 
Most of the 1732 juvenile walleye pollock examined were 
from the Semidi and Shumagin regions during LSumOO 
(Table 1) when samples were collected with midwater 
trawl nets. The use of different survey gear (i.e., midwa- 
ter or bottom trawl nets) among research cruises did not 
appear to influence fish size or stomach content weight. 
When compared by season and region, neither mean SL 
(ANOVA, P=0.101), nor mean stomach content weight 
(ANOVA, P=0.102), differed by survey gear. The bulk 
of the stomach contents (61%) was composed of juvenile 
and adult euphausiids, which also did not significantly 
vary by survey gear (ANOVA, P=0.426). 
Stomach content weight 
Overall, SCW averaged 0.72% BW, but season-to-season 
fluctuations were evident within each region. In all 
regions, mean SCW decreased from LSumOO to WinOl 
and then increased into SumOl before continuing to 
decline (Fig. 2). The Semidi region was associated with 
low mean SCW (0.14-0.82% BW) compared to all other 
regions except Kodiak during LSumOl (0.10% BW), 
which was represented by only seven fish. The Kodiak 
region had the highest mean SCW during LSumOO 
(1.40% BW) and during WinOl (0.56% BW). Later, the 
Shumagin region had the highest mean SCW during 
SumOl (0.76% BW) and LSumOl (0.49% BW), but we 
acknowledge low confidence in the Shumagin-Kodiak 
comparison because of the low number of fish available 
from the Kodiak region. Thus, the region with the high- 
est mean SCW appeared to have shifted from Kodiak to 
Shumagin after WinOl. 
Diet 
The cumulative number of identifiable prey types encoun- 
tered in juvenile walleye pollock reached an asymptote 
at 16 after the contents of about 60 stomachs were 
examined (Fig. 3). Overall, 19 categories of items were 
represented in juvenile walleye pollock stomachs, but we 
did not consider two of these categories to be part of the 
diet (hard items [e.g., sand] and parasites), and one cat- 
egory consisted of unidentifiable items (Table 2). As the 
number of stomachs examined decreased below 60, the 
cumulative number of prey types dropped sharply, caus- 
ing a negative bias on diet breadth. The smallest sample 
