266 
Fishery Bulletin 109(3) 
Table 2 
Results of permutational multivariate analysis of variance (PERMANOVA) pairwise tests for differences in fish assemblages 
and environmental variables between regions of the north coast (NC) of California and the Gulf of the Farallones (GF) for both 
seasons, and between summer and fall for both regions. 
Fish assemblages Environmental variables 
Factor 
Level 
Pairs 
pseudo-t 
P 
pseudo-t 
P 
Season 
summer 
NC, GF 
2.05 
0.0001 
1.80 
0.0149 
fall 
NC, GF 
2.17 
0.0001 
3.26 
0.0001 
Region 
NC 
summer, fall 
2.69 
0.0001 
2.30 
0.0022 
GF 
summer, fall 
1.03 
0.3471 
2.04 
0.0073 
fishes heavily dominated the catch. One hundred and 
thirty-one hauls were taken in summer and fall cruises 
along the NC. The catch in that area was dominated 
numerically by jacksmelt ( Atherinopsis californien- 
sis, 39%) and Pacific herring ( Clupea pallasii, 39%), 
with smaller landings of northern anchovy ( Engraulis 
mordax, 6%), juvenile Chinook salmon (5%), and surf 
smelt ( Hypomesus pretiosus, 2%). The five most fre- 
quently captured species were juvenile Chinook salmon 
(60%), jacksmelt (48%), adult Chinook salmon (34%), 
medusafish ( Icichthys lockingtoni, 30%), and Pacific 
sardine ( Sardinops sagax, 21%). 
Eighty-seven hauls were taken in summer and fall 
cruises in the GF. Species composition was dominated 
numerically by northern anchovy (45%) and Pacific her- 
ring (44%), followed by jacksmelt (4%), Pacific sardine 
(4%), and whitebait smelt ( Allosmerus elongates, 2%). 
The five most frequently captured species in this region 
were juvenile Chinook salmon (55%), Pacific herring 
(44%), jacksmelt (33%), adult Chinook salmon (32%), 
and medusafish (31%). 
Although the interpretation of yearly patterns along 
the NC (but not the GF) was hampered somewhat by 
the increasing sample size over time in the north, the 
haul-averaged catch density indicated that 2004 and 
2005 stand out as unusual years for several of the 
common species (those with >15% frequency of occur- 
rence) (Appendix 1). In the NC, the average density of 
jacksmelt was much higher in both seasons in 2004 
and 2005 than at any other time; Chinook salmon 
(both juveniles and adults) and Pacific herring aver- 
age density was highest in summer of 2004 and 2005; 
northern anchovy and Pacific sardine density was high- 
est in fall of 2004 and 2005; and jack mackerel (Tra- 
churus symmetricus) were captured only in 2004 (both 
seasons) and during the summer of 2005. In the GF, 
the average density of jacksmelt was especially high 
in the fall of 2004 and during both seasons in 2005, 
sardine density was especially high in spring of 2004 
and 2005, anchovy density was above average in both 
seasons in 2005, and Pacific butterfish ( Peprilus simil- 
limus ) average density was highest in both seasons in 
2005. No common species were notably absent in these 
two years. 
Multivariate biotic patterns 
All interaction terms in the three-way PERMANOVA 
were highly nonsignificant (P>0.40); these terms were 
sequentially removed by pooling their components of 
variation and degrees of freedom with residuals to 
increase statistical power for the remaining terms in the 
final reduced model. Main effects in the reduced model 
were all significant (region: pseudo-^ 16 =6.14, P=0.0001; 
season: pseudo-F x lg =2.07, P=0.0334; year: pseudo- 
P 5 16 = 1.52, P=0.0293). The two-way PERMANOVA 
for regional and seasonal community differences was 
also highly significant for both main effects (region: 
pseudo-Fj 60 =7.60, P=0.0001; season: pseudo-Fj g0 =4.62, 
P=0.0001) and not significant for regionxseason interac- 
tion (pseudo-Fj go =1.30, P=0.196). Subsequent pairwise 
comparisons (Table 2) showed strong differences in 
community structure between the NC and GF regions 
for both seasons, and strong seasonal differences within 
the NC region. However, there was no apparent seasonal 
difference within the GF region, where summer and fall 
communities were not statistically distinguishable. 
Interannual community pattern The ordination of trawl 
catch averaged broadly by region, year, and season (Fig. 
2A) showed clear separation of samples by region, but 
other differences due to years and seasons are not well 
supported by this configuration. With the exception of 
one point (‘04 GF-summer), 2004 and 2005 occupied a 
separate quadrant of the data cloud, indicating that 
there may have been some commonality of structure 
shared by those two years alone. Samples from the 
remaining four years did not show an annual pattern 
at this level of resolution. Seasonal (summer vs. fall) 
samples also appeared to be randomly mixed in this 
configuration, especially for the GF region. Seasonal 
differences in community structure must be examined 
at a finer scale of resolution, and samples averaged more 
narrowly by station and season across all six years, for 
recognizable seasonal patterns in MDS plots to emerge. 
Regional community pattern Ordinations of samples 
averaged by station and season visually supported the 
result of the two-way PERMANOVA and subsequent 
