256 
Fishery Bulletin 116(3-4) 
et al. 2 ). Since 1990, over 150 RSTs have been deployed 
in the Columbia River basin. Many RSTs operating 
for more than 20 years (Pacific States Marine Fish¬ 
eries Commission, Columbia Basin PIT Tag Informa¬ 
tion System, database available from website, accessed 
July 2015). Abundance estimates obtained from RST 
data are used to assess population productivity from 
the freshwater habitat and as the foundation for other 
life-cycle performance metrics (e.g., Venditti et al. 1 ; 
Copeland et al., 2014). This demographic information 
is central for monitoring salmonid populations and is 
often used to inform conservation actions (Venditti et 
al. 1 ; Copeland et al. 2 ). 
Rotary screw traps are passively operating traps 
constructed of a partially submerged cone mounted to 
2 pontoons (Johnson et al., 2007). The pontoons enable 
a RST to float while the cone funnels fish into a hold¬ 
ing box located at the stern of the RST. Once fish are 
in the holding box, a helical twist within the cone pre¬ 
vents individuals from swimming upstream and out of 
the trap. When a RST is in operation, a trap tender 
removes captured fish from the holding box. Targeted 
individuals are anesthetized, inspected for tags, and 
length and weight data are recorded (Johnson et al., 
2007). An allotted number of individuals are marked 
and released upstream of the RST (for a single trap 
design) or downstream (for a double-trap design) to be 
recaptured during proceeding days. 
Rotary screw traps in Idaho typically operate from 
early March until freezing temperatures and frazil ice 
make them inoperable in November. The exact dates of 
RST installation and removal are dependent on yearly 
environmental conditions and sampling duration can 
vary by 1 or 2 weeks among years. Unexpected ice, 
high water, and RST mechanical failures can reduce 
trap efficiency or halt RST operations for several days 
to several weeks within a year. During these outages, 
it is assumed fish continue to pass the RST because 
this migratory behavior was observed during previous 
years. Most stream-type Chinook salmon in Idaho fol¬ 
low a recurring bimodal migratory pattern where the 
majority of a cohort begins emigration out of headwa¬ 
ter rearing habitat in the fall and the remainder of the 
cohort emigrates in the spring (Bjornn, 1971, 1978). 
Parameterization of time-stratified mark-recapture models 
Data collection at RSTs consists of the daily number 
of unmarked and marked fish captured, and the num¬ 
ber of marked fish released the day prior. Schwarz 
and Bonner 3 found that weekly stratification of RST 
2 Copeland, T., R. V. Roberts, B. N. Oldemeyer, and K. A. Ap- 
person. 2013. Idaho steelhead monitoring and evaluation 
studies: annual progress report, January 1, 2012-December 
31, 2012. Idaho Dep. Fish Game Rep. 13-07, 47 p. [Avail¬ 
able from website.] 
3 Schwarz, C. 1. and S. J. Bonner. 2012. An application of a Bayes¬ 
ian stratified-Petersen model to estimate the number of outgoing fish 
on the Cheakamus River, British Columbia. Simon Fraser Univ. Rep. 
2012-02-22 . [Available from website.] 
mark-recapture data provided a sufficient balance be¬ 
tween maintaining run characteristics while avoiding 
unnecessary data sparsity issues under the assumption 
that daily capture probabilities were similar within the 
week. Therefore, we opted to stratify year (j=l,...,t) by 
ordinal week (i=l,...,s). If capture probabilities were 
subject to high variability within a weekly stratum, 
stratum size could be decreased. For our model that 
used multiple years of data, the weekly stratification 
of the number of unmarked fish captured in the ith 
stratum in the y'th year was denoted as u y, the number 
of marked fish released in the jth stratum in the jth 
year as /? y, and the number of recaptured fish captured 
in the ith stratum in the jth year as my. For the 3 
models that used data from 1 year, symbol designation 
remained the same but the subscript denoting year, j, 
was removed. 
The likelihood function of time-stratified mark-re- 
capture models implemented in the Bayesian frame¬ 
work consisted of 2 primary components: the prob¬ 
ability an individual was captured at the RST and the 
estimated number of unmarked individuals passing the 
RST. The numbers of individuals recaptured in a stra¬ 
tum, n?ij, were assumed to be binomially distributed by 
the number of marked individuals released upstream 
of the RST within the stratum, n ip and the probability 
that an individual passing the RST was captured, p^: 
my ~ BinomiaKtiij, p i; j). (2) 
Previous studies have modeled m X) as a multinomial 
distribution incorporating an additional parameter 
describing the process of an individual’s probability of 
being available for recapture during a later stratum 
(Mantyniemi and Romakkaniemi, 2002; Bonner and 
Schwarz, 2011). We chose to exclude this parameter to 
simplify our models because >96% of juvenile Chinook 
salmon recaptured at RSTs in our study did so within 
the proceeding day of release. The likelihood of the 
model is complete when the number of unmarked indi¬ 
viduals captured within a stratum, u ip is incorporated 
by using the binomial distribution: 
Uij ~ BinomiaKU py), (3) 
where [/,, = the estimated number of unmarked fish 
passing the RST during the stratum. 
The assumptions of the time-stratified Lincoln-Peters- 
en model are (Otis et al., 1978) as follows: 
1 Individuals do not emigrate or die between marking 
and recapture; 
2 Marks or tags are not shed; 
3 Marks or tags are detected if present at recapture; 
4 Marked and unmarked individuals within a stratum 
have the same probability of capture; 
5 Individual movements within a stratum are inde¬ 
pendent; and 
6 Individuals passing or being released below the RST 
are emigrating downstream and remain below the 
RST. 
