Johansson et al.: Seasonal distribution and environmental associations of larval and juvenile Sebastes spp. 
269 
quenced by using BigDye Terminator cycle sequencing 
kits (Thermo Fisher Scientific) and internal primers 
CBInf2 and CBInr2 (Bauble et al., 2012). Sequencing 
products were cleaned with Sephadex (GE Healthcare 
Bio-Sciences, Pittsburgh, PA) and visualized on an Ap¬ 
plied Biosystems 3730 xl DNA Analyzer (Thermo Fisher 
Scientific). Sequence data were aligned and edited with 
Sequencher, vers. 4.7 (Gene Codes Corp., Ann Arbor, 
MI) software. All sequencing and analysis were per¬ 
formed by the authors. Reference and unknown se¬ 
quence data are available from ScholarsArchive@OSU 
(available at website). 
Haplotype sequences from trawl-collected larvae 
were compared with a reference data set of 374 in¬ 
dependent haplotypes from 67 species of morphologi¬ 
cally identified adult rockfish (see Taylor et al., 2004 
for species included in the reference data set). Identi¬ 
cal sequences were identified in Sequencher, and all 
unique sequences were compared with the reference 
data set by using a custom script in R, vers., 3.3.1 (R 
Core Team, 2016) and the program PAUP*, vers. 4bl0; 
Sinauer Associates, Sunderland, MA) with the optimal¬ 
ity criterion set to distance (number of bp differences 
divided by total length of sequence in bp). 
Nonparametric bootstrapping (1000 replicates) was 
used to cluster individual unknown sequences within 
the reference data set of 374 known adult sequences. 
Some species in the reference data set failed to form 
monophyletic clades: the widow rockfish ( S. entome- 
las) and blue rockfish ( S. mystinus) (EM complex) and 
pygmy rockfish (S. wilsoni), Puget Sound rockfish (S. 
emphaeus), harlequin rockfish (S. variegatus), and 
sharpehin rockfish (S. zacentrus) (WEVZ complex). 
Others, such as the splitnose rockfish (S. diploproa), 
have few differences from sister species, a feature 
that affects the robustness of bootstrap results. For 
samples that were identified as falling within the 
EM complex, we manually compared sequences in 
Sequencher with those of known widow rockfish and 
blue rockfish. Known widow and blue rockfish differ 
consistently at a single nucleotide in our database 
of known sequences (base position 627, where widow 
rockfish sequences contain a cytosine and blue rock¬ 
fish sequences contain an adenine), allowing us to 
identify all our juveniles as widow rockfish. For all 
other samples, we performed identifications with 2 
different levels of stringency and compared outcomes. 
If an individual clustered within a monophyletic sin¬ 
gle-species clade with a bootstrap value >70%, this 
result was considered a high-confidence identification 
of the individual. Bootstrap values between 50% and 
69% were considered low-confidence identifications. 
Other outcomes (i.e., nonmonophyletic affinities, poly- 
tomies) were coded as failed identifications. We also 
calculated F84 distance (Felsenstein and Churchill, 
1996) between unknown and reference haplotypes us¬ 
ing the R package ape, vers. 3.5 (Paradis et al., 2004) 
to confirm that unknowns fell within the range of ex¬ 
pected intraspecific diversity on the basis of reference 
data. 
Physical data 
Physical data were collated from local, regional, and 
basin-wide spatiotemporal scales for comparison with 
rockfish community composition to determine whether 
species of Sebastes were associated with a particular 
level of environmental variability. In situ measure¬ 
ments of temperature and salinity were collected at 
the surface (1 m) and at trawling depth (30 m) by 
using a cast of an SBE25pZ«s Sealogger CTD conduc¬ 
tivity, temperature, and pressure recorder (Sea-Bird 
Scientific, Bellevue, WA) at each sampling station. At 
the regional scale, monthly averaged upwelling indi¬ 
ces, based on estimates of offshore Ekman transport 
driven by geostrophic wind stress at 45°N and 145°W, 
were obtained from the Pacific Fisheries Environmen¬ 
tal Laboratory (website). At the basin scale, monthly 
indices of the Pacific Decadal Oscillation (PDO; Man¬ 
tua et al., 1997) and the North Pacific Gyre Oscillation 
(NPGO; Di Lorenzo et al., 2008) were included in the 
analysis because they are oceanic expressions of the 2 
dominant modes of North Pacific atmospheric variabili¬ 
ty—the Aleutian Low and the North Pacific Oscillation, 
which are linked to the different phases of the El Nino 
Southern Oscillation cycle (Di Lorenzo et al., 2010). 
The PDO captures variability in sea-surface tempera¬ 
ture (SST) and the NPGO captures the variability in 
salinity, nutrients, and chlorophyll-a. 
Community analyses 
To evaluate seasonal and annual variability in the spe¬ 
cies numerical composition of late-larval and juvenile 
rockfish sampled for this study, nonmetric multidimen¬ 
sional scaling (NMS) was used to ordinate rockfish 
samples (densities of fish identified in the laboratory 
by visual means and proportions of rockfish identified 
on the basis of genetics) in -dimensional space. NMS 
is an iterative process that searches for the best posi¬ 
tions of n entities on k dimensions (axes) that minimize 
the stress of the ^-dimensional configuration (McCune 
et al., 2002). The goal is to generate a plot of sample 
points (individual tows) such that the rank order of 
the dissimilarity values between all pairs of samples is 
preserved in the spatial arrangement of points in the 
final plot. A total of 17 rockfish species were visually 
identified in the laboratory from 3266 individuals col¬ 
lected at 93 sampling stations, but only 7 taxa that oc¬ 
curred in >10% of the sampling stations were included 
in the analysis (total number of individuals of the 7 
visually identified taxa: 3072). Densities were In (x+1) 
transformed and relativized by species maxima to ho¬ 
mogenize variance. For rockfishes identified by using 
genetics, 24 taxa (total number of individuals of the 24 
genetically identified taxa: 915) were used, excluding 
individuals classified to species complexes as WEVZ 
(1489 individuals) or samples for which identification 
to species failed (25 individuals). This analysis was 
conducted with PC-ORD, vers. 6; MjM Software, Gl- 
eneden Beach, OR; McCune and Mefford, 2011) with 
