Harding et at: Regional and seasonal patterns of epipelagic fish assemblages from the central California Current 
271 
□ summer 
□ fa" 
LU 
CO 
+1 
12- 
'cx> 
s 
9- 
> 
-C 
Q. 
O 
6- 
O 
-C 
o 
3- 
c 
CO 
CD 
0- 
LU 
CO 
+1 
26.0 -I 
E 
25.5- 
CJ) 
-id 
25.0 
( f ) 
c 
Q) 
"O 
c 
24.5 
Re* 
Se ns 
Re ' 
Se' 
LU 
c n 
Figure 4 
Environmental variables measured immediately before or after trawling at stations off 
northern and central California, averaged by region and season across years and shown 
here untransformed. NC=north coast, GF=Gulf of the Farallones. Significance levels for 
two-way analysis of variance (ANOVA) tests were determined separately for each variable 
and are represented in each graph: Re=region, Se = season, ns= not significant at P>0.05, 
*P<0.05, **P<0.01, ***P<0.001. All RexSe interaction terms were not significant. 
fied by the SIMPER routine were typical of more than 
one assemblage, which is not surprising for communities 
consisting of mobile pelagic species. 
The arrangement of stations in PCA plots, based 
solely on environmental measures, was broadly similar 
to MDS ordinations based on fish abundance and diver- 
sity. Because the resemblance matrices underlying the 
corresponding PCA and MDS plots were independently 
derived, the match is unlikely to be purely coinciden- 
tal. It is reasonable to assume that one or more of the 
oceanographic variables are potentially causal fac- 
tors affecting fish community structure in this region. 
Matrix-matching tests, where community patterns were 
compared to oceanographic features, helped narrow the 
field down to three or four important variables, namely 
DEP, DIS, and any one or two of the upwelling-related 
variables TMP, SAL, and DEN, which together proved 
to be fairly good indicators of variation in the commu- 
