492 
Fishery Bulletin 107(4) 
Karluk commercial fishery and Karluk weir. Harvest, 
escapement, and age composition were available from 
the National Archives in Anchorage, the Alaska Depart- 
ment of Fish and Game, the National Marine Fisheries 
Service, and the Fishery Research Institute at the Uni- 
versity of Washington. 
With this approach, residuals (cj) of the fitted Ricker 
curve represented survival in the form of the annual 
deviation from the expected numbers of recruits based 
on the numbers of parents (escapement). The linear 
form of the survival model is given as 
In [_R, +r / E t ~\ = a + (5E t + e, 
( 1 ) 
where E 
R 
a 
P 
£ 
number of sockeye salmon counted at the 
Karluk weir during the spawning migration 
in year t; 
number of sockeye salmon counted at Karluk 
weir and caught in the fishery in years r 
from brood year t; 
density-independent parameter; 
density-dependent parameter; and 
residuals, the survival index. 
The survival index was lagged to match each of the 
three years of data of marine residence (i.e., juvenile, 
immature, and maturing stages) for the age 2.2 fish 
in the brood and was related to annual scale growth 
measurements and C-0 indices. 
The residual method is unfortunately less reliable 
than direct measurements of marine survival and re- 
quires several assumptions. One assumption is that the 
survival variability of the brood represents the survival 
of age 2.2 fish, a dominant age class in the brood. 
Also, because the recruitment estimate is based on the 
number of salmon counted at the Karluk weir and the 
number of salmon captured in the commercial fishery 
near the mouth of the Karluk River during the early 
and late run, it was assumed that there was equal 
fishing effort, minimal catch of non-Karluk sockeye 
salmon, and limited size selectivity of the fishing gear 
between runs and among years. It was also assumed 
that the fitted curve accurately reflected the overall 
density-dependent response of a stock at a given envi- 
ronmental state. 
Analytical techniques 
Ocean regime trends and comparison of regimes To 
describe and visually assess the low-frequency fluctua- 
tion in growth, climate indices, and survival in relation 
to the three ocean regimes, we fitted a loess regression 
line to the annual values using SigmaPlot software 
(Systat Software, Inc., Chicago, IL). The loess method 
of smoothing is based on tricube weighting and polyno- 
mial regression (Cleveland and Devlin, 1988). The loess 
regression method uses a variant of the local regression 
algorithm to approximate a nonlinear surface. Assume 
a series x- for i=l, n. . The basic idea involves estimating 
a smoothed series y i 
y i = g(x i )+e i , (2) 
where points in the near neighbor of {y i , x-) have more 
weight than points further away. 
In contrast use of an ordinary least-squares regression 
method, all points are equally weighted. In the analysis 
it is assumed that 30% of data points are used to com- 
pute each smoothed value and a 1° polynomial is then 
estimated. To determine the differences among regimes, 
an analysis of variance (ANOVA) was used to compare 
an average of the annual means of growth among C-0 
regimes. When a difference occurred we used the Tukey 
and Dunn pairwise comparison test to determine the 
regimes that differed. 
Correlation analyses To describe the relationships 
among the scale growth, the C-0 indices, and survival, 
we used the Pearson product moment correlation method. 
The null hypotheses of no associations among indices 
(H 0 : p=0) was tested against the alternative hypoth- 
eses that growth, climate, and survival were positively 
(H a : p>0) or negatively (H a : p<0) related at a 5% level 
of significance (a=0.05). A Bonferonni correction factor 
was used to adjust the critical P-value for the number 
of correlations = 0.05/ the number of correlations. Cor- 
relations among growth variables within broods were 
used to determine the dependence of growth on growth 
in the previous year. 
Results 
Influence of C-O regimes on mean 
sockeye salmon growth 
Juvenile (Ml) growth and immature (M2) growth were 
lower during the cool regime and higher during the 
recent warm regime, whereas maturing growth (M3) 
was lower during the early warm regime and higher 
during the cool and recent regime (Fig. 2). Growth 
varied over time by 14% for FW, 10% for Ml, 15% for 
M2, and 46% for M3. FW declined 12% during the early 
warm regime, was relatively constant during the cool 
regime, increased 8% from 1970 to 1980, and decreased 
8% from 1980 to 1998. Ml decreased during the early 
warm regime, increased 7% from the later part of the 
early warm regime to mid-cool regime, increased 5% 
from 1970 to 1980, and decreased 2% from 1980 to 1998. 
M2 was low and constant during the early warm and 
cool regimes and increased 15% from the early 1970s to 
the mid 1980s. M3 increased 40% from the mid 1920s 
to the mid-cool regime period and decreased 18% in the 
mid-1980s. 
For all four growth variables, at least one mean 
or median value was significantly different (ANOVA: 
P<0.002) among periods (Table 1). Mean Ml and M2 
were significantly higher (Tukey test: P<0.05) dur- 
ing the 1977-2000 warm C-0 regime than during the 
1922-46 warm and 1947-76 cool regimes. Mean FW 
