Weitkamp et al.: Seasonal and interannuai variation in [uvenile salmonids in the lower Columbia River 
435 
Table 3 
Mean catch per unit of effort (CPUE, number/1000 m 2 ) and coefficients of variation (CV; %) estimated by cruise and averaged 
by year for the most commonly caught fish species in our study in open-water habitats of the lower Columbia River estuary. Also 
indicated is the rank of each species based on mean abundances averaged across years (l=most abundant). 
Species-and-age class 
Rank 
Year 
2007 
2008 
2009 
2010 
CPUE 
CV 
CPUE 
CV 
CPUE 
CV 
CPUE 
CV 
American shad 
5 
28.9 
165.1 
33.5 
153.3 
26.3 
94.5 
31.7 
110.4 
Chinook salmon 
subyearling 
7 
4.3 
98.1 
1.7 
128.1 
4.0 
127.3 
3.3 
113.7 
yearling 
9 
0.6 
45.5 
0.8 
102.1 
1.2 
106.0 
2.4 
132.1 
Coho salmon 
8 
1.1 
111.0 
1.1 
166.8 
1.6 
142.9 
1.3 
132.6 
Northern anchovy 
1 
1254.9 
157.4 
1.1 
240.4 
0.6 
125.2 
114.9 
165.8 
Pacific herring 
3 
295.2 
185.3 
1.8 
173.8 
39.9 
118.4 
0.8 
100.5 
Shiner perch 
6 
19.4 
183.1 
10.4 
137.8 
10.2 
118.4 
0.8 
127.1 
Steelhead 
10 
0.6 
103.5 
0.8 
165.6 
0.8 
130.2 
1.6 
108.3 
Surf smelt 
4 
14.7 
54.5 
67.2 
162.5 
30.6 
100.1 
10.0 
47.0 
Threespine stickleback 
2 
88.6 
138.0 
230.5 
127.1 
207.7 
115.2 
102.6 
94.4 
similarity among hauls made >5 h apart (mean=32.7%; 
KW H- 99.6, P<0.05). In addition, pairwise similarities 
among sets made during the same cruise within 1 h of 
each other relative to low tide but at different stations 
(mean=51.5%) were lower than similarities among sets 
made at the same station (64.0%; MW [7=6.9, P<0.05) 
and comparable with hauls made at the same station 
but 3-4 h apart. Taken together, these results suggest 
a highly dynamic estuarine fish assemblage at fine time 
scales with modest spatial variation. 
We also used MDS plots to graphically evaluate 
variation in the fish assemblage at 2 scales (haul and 
cruise). This evaluation indicated that at fine time 
scales (haul) there was little correspondence between 
the fish assemblage and any time scale (e.g., time af- 
ter low tide, cruise, biweek, year) or station (Fig. 3B). 
For example, hauls occurring within any particular 
2-week period failed to form obvious assemblage groups 
but instead more or less spanned the range of MDS 
space. By contrast, when considered by cruise, there 
was a fairly clear pattern with the larger time scale 
(Fig. 3A). Results from the ANOSIM analysis were 
consistent with patterns observed in the MDS plots, 
showing better defined groups (i.e., higher Global R 
values) by cruise than by haul. When each haul was 
considered independently, no single or pair of variables 
produced well-defined groups (i?<0.40), consistent with 
the lack of obvious groups in the MDS plot (Fig. 3B). 
In this analysis, the variable “biweek” produced the 
best groups (Global R = 0.22), and inclusion of year in 
a 2-factor analysis increased biweek group separation 
(Global i? = 0.35) although year itself did not produce 
well-defined groups (Global 77=0.00). When evaluated by 
cruise, results were similar to those results produced by 
haul: the single variable biweek provided the greatest 
group separation (Global 77=0.43), and biweek produced 
more distinct groups (Global 77 = 0.52) when combined 
with year. 
Environmental forcing of assemblage composition We 
explored the environmental variables that best fitted 
the fish assemblage data, either by haul or averaged 
by cruise. When examined by haul, the environmental 
model producing the best fit (r=0.40, P<0.05) to the 
species composition data consisted only of in situ 7-m 
salinity, and models with fits that were only slightly 
poorer (0.39<r<0.40) included 7-m salinity, 1-m salinity, 
1-m temperature, and coastal SST (which was correlated 
with 1-m temperature). These results suggest the fish 
assemblage is actively responding to environmental 
forcing at short temporal scales (hours) associated with 
tidal inundation and to a lesser extent with seasonal 
changes. The same analysis for fish assemblage aver- 
aged by cruise also indicated that the highest correlation 
occurred with a single variable, river water temperature 
(/• = 0.57, P<0.05), which is largely a seasonal signal. 
Other models with fairly high correlations (0.46<r<0.49) 
with the fish data included river temperature, coastal 
SST, PDO, and river flow. Given the strong correlations 
between river temperature and SST, and between river 
flow and the PDO, these results suggest the fish com- 
munity is influenced from both types of climate signals: 
seasonal and river flow. 
Juvenile salmon abundance and timing 
in the Columbia River estuary 
Despite high variability in overall catches, the abun- 
dance and timing of juvenile salmon in the estuary 
was surprisingly consistent among years (Fig. 4). In 
