100 
Fishery Bulletin 117(1-2) 
Genetic stock identification 
Estimates of the stock of origin for each fish were 
made by using genetic stock identification (GSI). To 
perform GSI, we used the algorithms employed by the 
software ONCOR (Kalinowski et al., 2007). The ON- 
COR program uses the partial-Bayesian algorithms 
of Rannala and Mountain (1997) with an expectation 
maximization algorithm to assign individuals to their 
most likely population of origin by using a baseline of 
genetic information collected from samples of known 
origin. A critical assumption of this method is that all 
possible source populations are in the baseline. This 
assumption is unlikely to be met in most studies of this 
type, including ours. However, assignment likelihoods 
are correlated with the hierarchical genetic structure 
typical of salmonids such that, if the source population 
is not in the baseline but a genetically similar popula¬ 
tion is, the unknown individual will likely be assigned 
to the correct hierarchical genetic group for the actual 
source population. The groups to which we assigned 
our samples were determined by examining the ac¬ 
curacy of different configurations with leave-one-out 
cross-validation analyses. 
Once the genetic stock groups were established, we 
employed a 2-tiered approach to determine the genetic 
stock origin of each ocean-caught steelhead. First, we 
used the program ONCOR to assign individuals to 1 
of 7 genetic stock groups from a baseline of 131 SNP 
loci from 148 populations sampled in an area from the 
Central Valley of California to Puget Sound, Wash¬ 
ington. This first tier of GSI assignments provided us 
with the resolution adequate for determining whether 
samples originated from the Columbia River Basin or 
from another region. The accuracy of this baseline was 
estimated by using the leave-one-out method as imple¬ 
mented in ONCOR. This leave-one-out method removes 
an individual fish from the baseline and treats it like 
an individual with an unknown origin, while assigning 
it to a group by using the remaining individuals in the 
baseline. This procedure is repeated for all individuals 
in the baseline, and a running total of how many of 
the assignments were correct is used to estimate the 
accuracy of the baseline. 
Second, samples that were assigned to the Columbia 
River genetic stock group during the first analysis were 
then analyzed with ONCOR by using a baseline of 180 
SNP loci from 116 populations of the Columbia River 
(Hess et al. 1 ), to determine to which of the 4 genetic 
stock groups of the Columbia River each of those fish 
belonged. The accuracy of this baseline was also tested 
by using the leave-one-out method. 
Some of the assignments of fish sampled from the 
Columbia River were made by using the technique 
known as parentage-based tagging, with the software 
SNPPIT, vers. 1.0 (Anderson 3 ). Parentage-based tag¬ 
3 Anderson, E. C. 2010. Computational algorithms and 
user-friendly software for parentage-based tagging of Pa¬ 
cific salmonids. Final report submitted to the Pacific Salmon 
ging is accomplished by genotyping as many hatchery 
adult broodstock within a region as possible and then 
implementing parentage assignments to identify the 
offspring of those adults when they are sampled in a 
mixture of unknown individuals (Steele et al., 2013). 
Although we were not necessarily interested in know¬ 
ing the hatchery of origin of our samples, this method 
can increase the accuracy of the GSI assignments be¬ 
cause offspring are assigned directly to their parents 
rather than to a geographic region on the basis of 
their genotype frequencies. At the time of our analy¬ 
ses, parentage-based tagging data were available only 
for adults that spawned in 2009-2012 at hatcheries in 
the Snake River (data available from website, accessed 
January 2016). 
Stock distributions and catch 
Using the genetic stock assignments for each fish, we 
examined stock-specific distributions in our study area. 
To obtain a measure of steelhead relative abundance, 
we calculated values of stock-specific catch per unit of 
effort (CPUE) simply as the number of fish caught di¬ 
vided by the number of kilometers trawled. We then 
compared CPUE among years. In addition, we calculat¬ 
ed the average CPUE for each stock for each year and 
then averaged those values across years to estimate 
an overall stock-specific CPUE, giving equal weight to 
all years. 
Steelhead that originate from the Columbia River 
represent several distinct genetic stock groups and ori¬ 
gins (hatchery and wild), but they all enter the ocean 
at close to the same location, the mouth of the Co¬ 
lumbia River. To examine differences in early marine 
migration among these groups, latitude and longitude 
values were averaged for all fish originating from the 
same genetic stock group. Latitude and longitude val¬ 
ues had been recorded for each trawl haul during sam¬ 
pling. These annual averages of latitude and longitude 
for each genetic stock group then were averaged over 
all years, to equally weight each year, and standard 
deviations were calculated. Differences in latitude also 
were converted to distance in kilometers (by using a 
conversion calculator available from the NOAA Na¬ 
tional Hurricane Center at website, accessed Novem¬ 
ber 2018). 
Results 
Genetic stock groups 
The genetic stock groups we chose for genetic assign¬ 
ments were those that provided the highest degree of 
geographic resolution, while still retaining high ac¬ 
curacy and close alignment with the distinct popula¬ 
tion segments identified in assessments completed as 
Commission’s Chinook Technical Committee (US Section), 43 
p. [Available from website, accessed July 2018.] 
