ArticlePDF Available

Multiple Challenges Confront an Inland Recreational Fishery in Decline

  • Central Coast Indigenous Resource Alliance


Catch and release regulations designed to protect fisheries may fail to halt population declines, particularly in situations where fishing effort is high and when multiple stressors threaten a population. We demonstrate this claim using Alberta’s Bow River, which supports a high-effort Rainbow Trout Oncorhynchus mykiss fishery where anglers voluntarily release > 99% of their catch. We examined the population trend of adult trout, which were tagged and recaptured using electrofishing surveys conducted intermittently during 2003-2013. We constructed Bayesian multi-session capture-recapture models in Stan to obtain abundance estimates for trout, and regressed trend during two periods to account for variation in sampling locations. General patterns from all models indicated the population declined throughout the study. Potential stressors to this system that may have contributed to the decline include Whirling Disease Myxobolus cerebralis, which was detected for the first time in 2016, notable floods, and release mortality. Because disease and floods are largely uncontrollable from a management perspective, we suggest that stringent tactics such as angler effort restrictions may be necessary to maintain similar fisheries.
Multiple challenges confront a high-effort inland recreational
fishery in decline
Christopher L. Cahill, Stephanie Mogensen, Kyle L. Wilson, Ariane Cantin, R. Nilo Sinnatamby,
Andrew J. Paul, Paul Christensen, Jessica R. Reilly, Linda Winkel, Anne Farineau, and John R. Post
Abstract: Catch-and-release regulations designed to protect fisheries may fail to halt population declines, particularly in
situations where fishing effort is high and when multiple stressors threaten a population. We demonstrate this claim using
Alberta’s Bow River, which supports a high-effort rainbow trout (Oncorhynchus mykiss) fishery where anglers voluntarily re-
lease >99% of their catch. We examined the population trend of adult trout, which were tagged and recaptured using electro-
fishing surveys conducted intermittently during 2003–2013. We constructed Bayesian multisession capture–recapture models in
Stan to obtain abundance estimates for trout and regressed trend during two periods to account for variation in sampling
locations. General patterns from all models indicated the population declined throughout the study. Potential stressors to this
system that may have contributed to the decline include whirling disease (Myxobolus cerebralis), which was detected for the first
time in 2016, notable floods, and release mortality. Because disease and floods are largely uncontrollable from a management
perspective, we suggest that stringent tactics such as angler effort restrictions may be necessary to maintain similar fisheries.
Résumé : La réglementation sur la pêche avec remise à l’eau visant à protéger les ressources halieutiques peut ne pas prévenir
le déclin de populations, particulièrement dans des situations où l’effort de pêche est grand et de multiples facteurs de stress
menacent une population. Nous faisons la démonstration de ce postulat avec l’exemple de la rivière Bow, en Alberta, qui
supporte une pêche à la truite arc-en-ciel (Oncorhynchus mykiss) à effort élevé dans laquelle les pêcheurs remettent volontairement
à l’eau plus de 99 % de leurs prises. Nous avons examiné la tendance démographique de truites adultes qui ont été étiquetées et
recapturées dans le cadre de relevés à la pêche électrique menés de manière intermittente de 2003 à 2013. Nous avons construit
des modèles bayésiens de capture-recapture à sessions multiples dans le logiciel Stan pour obtenir des estimations de
l’abondance pour les truites et obtenu une tendance par régression pour deux périodes pour tenir compte de la variation des
lieux d’échantillonnage. Les motifs généraux obtenus de tous les modèles indiquent que la population a baissé tout au long de
l’étude. Les facteurs de stress pour ce système qui pourraient avoir joué un rôle dans cette baisse comprennent le tournis des
truites (infection à Myxobolus cerebralis), qui a été détecté pour la première fois en 2016, des inondations notables et la mortalité
après remise à l’eau. Comme il est difficilement possible de contrôler, dans une perspective de gestion, les infections et les
inondations, nous suggérons que des mesures draconiennes, comme des restrictions de l’effort des pêcheurs à la ligne, pour-
raient être nécessaires pour assurer le maintien de ressources halieutiques semblables. [Traduit par la Rédaction]
Recreational fisheries can decline and collapse when exploita-
tion rate is high relative to the productivity of the targeted popu-
lation (Post et al. 2002;Sullivan 2003;Johnston et al. 2007), when
environmental conditions are unfavorable (Beamish et al. 1999),
or when multiple stressors surpass the compensatory abilities of a
given fish population (Hansen et al. 2015). However, declines doc-
umented as reductions in catch or abundance are generally mul-
tifaceted and subject to much debate (Hilborn and Walters 1992).
Failure to achieve scientific consensus on the cause(s) of a decline
stems from the fact that data collected by management agencies
are observational in nature (Hilborn 2016), extracted from ecosys-
tems with complex dynamics (Sugihara et al. 2012;Glaser et al.
2014), and often lack the experimental manipulations, controls,
and replication necessary for causal inference (Jones and Hansen
2014). Consequently, disentangling key factors responsible for a
decline, such as the effects of fishing versus abiotic or environ-
mental stressors, is often difficult (Walters 1986).
Several policy options are available to managers seeking to off-
set the effects of fishing-related mortality in inland recreational
fisheries. Traditionally, fishing mortality has been managed using
passive regulations such as bag or length limits (Lester et al. 2003),
seasonal or area closures (Welcomme 2008), and gear restrictions
(Rahel 2016). Such regulations either (i) manipulate the catch or
harvest rate of individual anglers by decreasing catchability or
vulnerable fish population size or (ii) indirectly control total an-
gling effort through voluntary choices by anglers. Thus, the abil-
ity of these passive regulations to offset issues related to high
angling effort remains tenuous (Cox et al. 2003;Pereira and
Hansen 2003). For example, minimum length limits may fail to
prevent recruitment overfishing when angler effort is unre-
sponsive to declines in fishery quality (Allen et al. 2013). Simi-
larly, catch-and-release regulations (i.e., zero harvest) may fail
Received 6 March 2018. Accepted 24 June 2018.
C.L. Cahill, S. Mogensen, K.L. Wilson, A. Cantin, R.N. Sinnatamby, A. Farineau, and J.R. Post. Department of Biological Sciences, University of
Calgary, 2500 University Drive NW, Calgary, AB T3A 4X5, Canada.
A.J. Paul. Fish and Wildlife Policy, Alberta Environment and Parks, Provincial Building, Cochrane, AB T4C 1A5, Canada.
P. Christensen, J.R. Reilly, and L. Winkel. Operations Division, Resource Management, Alberta Environment and Parks, Provincial Building,
Cochrane, AB T4C 1A5, Canada.
Corresponding author: Christopher L. Cahill (email:
Copyright remains with the author(s) or their institution(s). Permission for reuse (free in most cases) can be obtained from RightsLink.
Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)
Can. J. Fish. Aquat. Sci. 00: 1–12 (0000) Published at on 29 June 2018.
Can. J. Fish. Aquat. Sci. Downloaded from by UNIV CALGARY on 07/16/18
For personal use only.
to achieve management objectives or maintain fish populations
in situations where effort is high or when release mortality is
present (Post et al. 2003). Despite the potential for problems in
high-effort situations, catch-and-release regulations are generally
the most stringent regulations enacted by inland recreational
fisheries managers in North America (Hubert and Quist 2010).
Anthropogenic stressors such as fishing-related mortality re-
duce a population’s ability to offset stochastic natural distur-
bances or disease outbreaks (Holling 1973). For example, severe
floods may reduce fish density through direct mortality or down-
stream displacement (Warren et al. 2009). Pathogens may also
alter population dynamics through reductions in growth (Johnson
et al. 2004) or competitive ability (Godwin et al. 2015), changes in
behavior (Barber et al. 2000), or through direct mortality events
(Hatai and Hoshiai 1992). For instance, viral hemorrhagic septice-
mia virus resulted in mass mortality events in freshwater drum
(Aplodinotus grunniens)(Lumsden et al. 2007) and muskellunge (Esox
masquinongy)(Elsayed et al. 2006) within the Laurentian Great
Lakes. The importance of pathogen-induced mortality was also
illustrated throughout the intermountain-west region of the
United States, where whirling disease (Myxobolus cerebralis) re-
duced salmonid abundance by up to 90% in some watersheds
(Nehring and Walker 1996;Vincent 1996). While stochastic events
and disease outbreaks influence fish population dynamics, biolo-
gists have few direct tools for managing stressors such as these.
A central challenge facing recreational fisheries managers is
how to mitigate the combined effects of anthropogenic and sto-
chastic environmental threats, as freshwater systems appear par-
ticularly vulnerable to the effects of multiple stressors (Ormerod
et al. 2010). Unfortunately, traditional fisheries management ap-
proaches rarely incorporate the effects of multiple, potentially
changing stressors on population sustainability (Lynch et al. 2016;
Paukert et al. 2016). Thus, approaches that explicitly recognize
management trade-offs among key stressors may be useful. For
example, S.R. Carpenter et al. (2017) developed a multidimen-
sional sustainable safe operating space (i.e., “SOS”) that distin-
guished among stressors that were controllable (e.g., harvest) and
largely uncontrollable (e.g., environmental change or disease out-
breaks) by fisheries managers. These authors suggested that man-
agers may need to offset the effects of uncontrollable variables
with those they can influence to maintain a quality fishery within
a particular SOS (S.R. Carpenter et al. 2017).
The Lower Bow River (LBR) rainbow trout (Oncorhynchus mykiss)
fishery highlights the challenges of managing inland recreational
fisheries in the presence of multiple stressors. The LBR is reputed
as a world-class rainbow trout fishery (Post et al. 2006;Askey et al.
2007) and experiences high angling effort in part because of its
proximity to a large city (Calgary, Alberta; 1.2 million people;
Rhodes 2005;Statistics Canada 2016). Notable floods occurred in
2005 and 2013 (i.e., 1/10 and 1/100-year flood events, respectively;
Veiga et al. 2015). Furthermore, whirling disease was detected in
the LBR for the first time in 2016, although it is unknown when
the disease arrived (Canadian Food Inspection Agency 2017). Given
that the potential for these threats to impact rainbow trout in the
LBR appears high, our objectives were to determine (i) abundance
of rainbow trout ≥250 mm fork length (FL) and (ii) assess trends in
abundance during 2003–2013. Results from this study are then
used to demonstrate the potential importance of effort restric-
tions as an inland recreational fisheries management tool, partic-
ularly when angling effort is high and the remaining stressors
cannot easily be controlled.
Study area
The Bow River originates in the Rocky Mountains of southwest-
ern Alberta and is a major tributary of the South Saskatchewan
River (Fig. 1). Our study occurred in the LBR, which designates a
224 km section of the river beginning at the Bearspaw Dam and
flowing eastward through Calgary, Alberta, to the Bassano Dam
(Fig. 1). The LBR receives substantial nutrient inputs from waste-
water effluent as it flows through Calgary (Sosiak 2002), which
increases production of top trophic levels in the system (Askey
et al. 2007).
Rainbow trout were introduced into the LBR by stocking be-
tween 1933 and 1947 (Gilmour 1950) and are now naturalized
(Rhodes 2005). Rainbow trout are the most sought-after species in
the LBR fishery, and most fish are caught via fly-fishing during
August. This system supports the highest effort fishery in Alberta
(mean 161 angler hours per hectare annually; Council and Ripley
2006), has been functionally catch and release since at least 2006
(e.g., 0.005% of captured rainbow trout were kept by anglers;
Council and Ripley 2006), and is worth CAN$24.5 million per year
(Crowe-Swords 2016).
Field sampling
Alberta Environment and Parks biologists conducted seven
multiday capture–recapture surveys during 2003–2013 to assess
trout population trends in the LBR. A contiguous 4.32 km section
of the river was sampled in 2003, 2005, 2007, and 2008 (Fig. 1).
During 2011–2013, sampling occurred at four randomly selected
1 km sections (Fig. 1). In all years, fish were sampled using two
Smith-Root 5.0 or 7.5 Generator Powered Pulsators and jet-
powered boom electrofishers on four consecutive days during late
August to early September. Boats fished opposite banks in the
downstream direction, and fish were transferred to tubs until the
end of the section was reached. Rainbow trout ≥250 mm FL were
tagged behind the dorsal fin using individually numbered Floy
anchor tags, and fish were released in the center of the section
they were captured in.
Multisession capture–recapture models were constructed to ob-
tain derived abundance estimates of rainbow trout ≥250 mm FL,
which were regressed through time to estimate the population
trend during 2003–2008 and 2003–2013 to account for sampling
changes (described above). Multisession capture–recapture mod-
els generalize a simple closed capture–recapture model to multi-
ple sessions or years and are useful for combining data from
multiple years into a single analysis (Converse and Royle 2012;
Kéry and Royle 2015). Our models estimated capture probabilities
for each day within each year to account for imperfect detection
of fish in our surveys. Data were modeled in an integrated Bayes-
ian framework. The integrated approach propagated uncertainty
through the abundance analysis and ultimately into estimates of
population trend, rather than analyzing the outputs of models as
data (Maunder and Punt 2013;Paul 2013). Additionally, we elected
to use the Bayesian framework because it offered more modeling
flexibility and allowed for the accounting of the full range of
uncertainties related to all models and parameter values (Punt
and Hilborn 1997). Standard open population models were not
fit because surveys were not conducted in consecutive years
throughout the study.
A statistical approach known as data augmentation was used
to simplify Bayesian abundance estimation. Data augmentation
reparameterizes a standard closed capture–recapture model into
an occupancy-style model (Royle 2009;Kéry and Schaub 2012).
This parameterization removes the abundance parameter Nfrom
the model as an estimated parameter, which is useful because it
fixes the dimension of the parameter space and thus enables the
use of approaches such as Markov chain Monte Carlo (MCMC;
Royle and Dorazio 2008). Data augmentation removes Nfrom the
model by marginalizing over a Bin(M,
) prior distribution for N,
where Mis an arbitrarily large number of unobserved individuals
with all-zero capture histories, and
is an inclusion probability to
be estimated from the data (Royle 2009). The capture histories of
Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)
2 Can. J. Fish. Aquat. Sci. Vol. 00, 0000
Published by NRC Research Press
Can. J. Fish. Aquat. Sci. Downloaded from by UNIV CALGARY on 07/16/18
For personal use only.
individuals observed in the field are then augmented by the cap-
ture histories of the all-zero “potential” individuals, which are
said to be present in the population according to a Bernoulli prior
with inclusion probability
. This approach implies that the mar-
ginal prior for Nis Uniform(0, M) and hence that the expectation
of Ncan be recovered as the derived variable
Mthrough the
estimation of
, any relevant capture probabilities, and a Ber-
noulli latent variable (Royle et al. 2014).
The single-year data augmentation scheme described above was
expanded to a multisession framework by introducing a year-
specific subscript tto all parameters and values (e.g.,
, and
). Abundance in each year N
was assumed to be Poisson-
distributed with year-specific mean
(1) NtPoisson(
This multisession framework and Poisson assumption formed the
basis for three separate models, as the models could either be
used to estimate year-specific abundances (as in eq. 2 below;
model 1) or regress the population trend by fixing the trend on the
first year of data (as in eqs. 3 and 4below; models 2 and 3, respec-
tively). Two trend models were fit to account for potential issues
with field sampling: model 2 estimated a trend for years when the
surveys occurred in identical locations (i.e., 2003–2008), while
model 3 regressed a trend across the entire study period.
Model 1 was parameterized as a generalized linear model of the
0tare fixed effects that represent independent abundance
estimates for each year (i.e., 2003, 2005, 2007, 2008, 2011, 2012, and
2013; Royle et al. 2014). Model 2 regressed a trend through abun-
dance estimates during the first 4 years of data collection (i.e.,
2003, 2005, 2007, and 2008):
was fixed at a single intercept,
is a coefficient describ-
ing the log-linear abundance trend, and t* is a baseline year that
centers the trend. Model 3 regressed a trend term through abun-
dance estimates during 2003–2013:
Fig. 1. Map showing locations sampled via multipass electrofishing in the Lower Bow River during 2003–2013 (upper insert: location of the
study area in Alberta, Canada). Note that sampling locations in 2003, 2005, 2007, and 2008 were identical and are represented by the section
of river delineated with vertical black bars (sampling location “1”). Sampling locations were subsequently changed to stratified random
locations: 2011 = “2”, 2012 = “3”, 2013 = “4.” The Bearspaw Dam is indicated by a star ().
Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)
Cahill et al. 3
Published by NRC Research Press
Can. J. Fish. Aquat. Sci. Downloaded from by UNIV CALGARY on 07/16/18
For personal use only.
is a covariate controlling for the total length (km) of the
river sampled in each year. Additionally, all models featured con-
straints between the year-specific inclusion parameters
intercept terms (i.e.,
), and
can be fixed as
so that the intercepts are estimable (Royle et al. 2014;Kéry and
Royle 2015).
Capture probabilities (p) were modeled using a generalized lin-
ear mixed model of the same form for all three models. For each
year t, four occasion-specific capture probabilities jand iindivid-
ual random effects (i.e., p
) were modeled. Capture probabilities
were modeled as additive fixed effects based on the capture occa-
sion (i.e.,
) and the addition of a random logistic-normal term
that represented individual heterogeneity in capture probability,
(6) logit(pi,j,t)
i,tNormal(0, SD
tNormal(0, 1.25)
rawi,tNormal(0, 1)
is a year-specific random effect on capture probability
drawn from an informative prior distribution. The specific prior
values of
were not of primary interest to this study, but these
priors did alter the variance around the abundance estimates.
Data did not contain enough information to allow use of an unin-
formative prior for
Normal(0, 10)), and simpler ver-
sions of these models without individual heterogeneity terms
failed to pass goodness-of-fit tests because of overdispersion in the
individual encounter history data. Consequently, we were left
with the choice to use simpler but potentially ill-fitting models or
to employ an informative prior and obtain models that passed
goodness-of-fit tests. The latter approach was chosen, as it seemed
precautionary to err on the side of incorporating more uncer-
tainty in the models. Values for the informative priors ensured
that capture probabilities did not go to zero and that N
did not go
to infinity, so that the upper limits of the derived density esti-
mates (abundance·km
) corresponded to densities observed in
other fluvial wild stock rainbow trout fisheries (Vincent 1996).
rawi,taltered the individual random effects to a noncen-
tered parameterization, which manipulated the geometry of the
posterior and improved numerical performance (Betancourt and
Girolami 2013;Monnahan et al. 2017).
Estimates of abundance N
were derived as
(10) Ntnt
where n
is the number of individual fish captured in year t, and
represents the sum of M
independent Bernoulli trials with trial-
specific success probability equal to
4(1 pi,j,t)
4(1 pi,j,t)(1
Phrased differently, abundance of fish in each year was calculated
as the sum of the number of fish captured in that year and the sum
of M
Bernoulli random deviates describing fish that were present
in that year, but never physically captured. A standardized index
of rainbow trout density among years (N_km
) was computed by
dividing the estimates of N
by the length (km) of river sampled in
year t. Lastly, the percent decline per year for models 2–3 was
derived as
(12) Trend (%)(e
11) × 100
where the Trend term represents the per year percent change in N
during 2003–2008 and 2003–2013 (denoted as Trend
and Trend
Priors for all additional parameters were set at vague or weakly
informative values, and the sensitivity of the results to priors was
tested during model development. All predictor variables were
centered and standardized prior to analysis (Gelman 2008). For
each model, each year’s observed capture history data was aug-
mented by M
= 8000, which implied N
Uniform(0, 8000) priors,
and visual inspection of the results showed this bound was suffi-
ciently large so as to not influence estimates of abundance or the
corresponding derived densities (see also Royle et al. 2014).
Occasion-specific capture probabilities for all models were given
flat logit(
)Uniform(0, 1) priors. For all three models, intercept
0were given normal priors centered at zero, with a
prior standard deviation (SD) of 10 to represent a substantial
amount of ignorance regarding the values of
on the log scale.
The remaining regression coefficients (
for models 2 and 3,
for model 3) were given normal priors centered at zero with an SD
of 1. Priors for these remaining regression coefficients were
weakly informative, as our predictor variables were also on the
unit scale (Gelman et al. 2015), and while these priors did not
truncate estimates of
for either model, they did substan-
tially improve the numerical performance. Final model versions
were arrived at after fitting several alternative models that either
(i) failed to pass posterior predictive checks and (or) (ii) provided
identical (or nearly identical) posterior distributions for key pa-
rameters. Specifically, we fit models that assumed abundances
were distributed as a negative binomial random variable, but
these models produced nearly identical results for abundances
and the trend term as the simpler Poisson model. We also fit
simpler mark–recapture models that featured a single capture
probability for each year and models that featured occasion-
specific capture probabilities for each year without individual het-
erogeneity terms, and while these models produced equivalent
results as our final models, they did not pass our posterior predic-
tive checks (see below).
Our models were challenging to fit using traditional Bayesian
methods. For example, we originally attempted to fit our models
in JAGS (Plummer 2003), but MCMC chains mixed poorly and
simulations were slow. We circumvented these issues by fitting
models in Stan, which employs an efficient class of MCMC algo-
rithms based on Hamiltonian Monte Carlo (Stan Development
Team 2016). Stan proved useful for this analysis and data set when
coupled with the noncentered parameterization of the random
effects (see also Monnahan et al. 2017) and was more efficient than
analogous programs run in Jags (i.e., judged as the number of
effective samples per unit time). Thus, we echo Monnahan et al.
(2017) and recommend Stan as a powerful tool for ecological mod-
Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)
4 Can. J. Fish. Aquat. Sci. Vol. 00, 0000
Published by NRC Research Press
Can. J. Fish. Aquat. Sci. Downloaded from by UNIV CALGARY on 07/16/18
For personal use only.
Bayesian models were implemented using RStan in R (B. Carpenter
et al. 2017;Stan Development Team 2016;R Core Team 2017), and
we provide both R and Stan code to simulate and analyze data for
model 3 at the lead author’s personal website (available from We ran
ten chains for each model, discarded the first 5000 draws as burn-
in, and retained the next 5000 draws with a thinning rate of 1 to
summarize posterior distributions. Starting values were jittered
for each model, and chains were started at different random num-
ber seeds. Additionally, chains were inspected visually for mixing,
Gelman–Rubin statistics (R
ˆ) were used to test for convergence to a
stable distribution among chains, and tree-depth plots were ex-
plored to evaluate step size and performance of the no-U-turn
sampler using ShinyStan (Gelman et al. 2014;Gabry 2015). We also
ensured that models featured no divergent transitions during
sampling, which is a diagnostic that indicates whether results
from Stan are valid (Betancourt and Girolami 2013;Monnahan
et al. 2017). Model fit was evaluated using posterior predictive
pvalues (Kéry and Royle 2015), the development of which is de-
scribed in Appendix A.
Results from all three models suggest that the abundance of
rainbow trout ≥250 mm FL declined during 2003–2013. Model 1
showed that while year-specific estimates of abundance (i.e., N
and derived values of density (i.e., N_km
) were highly variable,
both declined during 2003–2008 and 2003–2013 (Table A1 and
Fig. 2, respectively). Density values from the independent inter-
cepts model declined more sharply during 2011–2013 than during
2003–2008 (Fig. 3). The highest derived values for rainbow trout
density occurred in 2003 (median = 414 fish·km
; 95% credible
interval (CRI): 290–713 fish·km
), while the lowest density esti-
mates occurred in 2013 (median = 72 fish·km
; 95% CRI: 57–
111 fish·km
;Fig. 2). Trend
was highly variable, but 88% of the
posterior mass for this derived variable was <0 (median = –5.5%
per year; CRI: –14.2% to 4.3% per year; upper panel Fig. 4). Values of
(2003–2013) were less variable than those of Trend
2008), but featured a similar median estimate for the percent
change per year, and 99% of the posterior mass was <0 (median =
–6.6% per year; CRI: –10.5% to –2.6% per year; lower panel Fig. 3).
Posterior predictive checks indicated that no major deviations
from the modeling assumptions existed for all models (i.e., poste-
rior predictive pvalues 0.18– 0.33; Fig. 4), and R
ˆstatistics suggested
convergence to a stable distribution among chains for all param-
eters in each model (i.e., all R
ˆ=1;Tables A1–A3). Estimates of year-
and occasion-specific capture probabilities were low for all three
models (i.e.,
typically ≤0.10; Appendix Tables A1–A3). Addition-
ally, values of
ranged from 0.03 to 0.39 for all models, which
indicated that the abundance estimates were not constrained by
our choices for the year-specific data augmentation variables M
(Converse and Royle 2012;Appendix Tables A1–A3).
Despite uncertainty in both abundance and trend estimates,
these results indicate that rainbow trout in the LBR likely declined
during both assessment periods. The median posterior estimates
for a 10-year standardized population trend were –43.0% using
estimated from the 2003–2008 data and –50.0% using
estimated from the 2003–2013 data. Since the size criteria
for tagged fish (≥250 mm FL) corresponds approximately to the
sizes of mature rainbow trout (Rhodes 2005), the estimated de-
cline likely represents a large reduction in the number of mature
individuals in this population. For example, The Committee on
the Status of Endangered Wildlife in Canada (COSEWIC) guide-
lines state that “in cases where the declines or its causes are un-
known”, a decline in the number of mature individuals ≥30%
or ≥50% over a period of 10 years would qualify a population
as “threatened” or “endangered”, respectively (COSEWIC 2018).
Rainbow trout in the LBR are naturalized (Gilmour 1950) and
hence do not qualify for listing as per COSEWIC guidelines; none-
theless, these criteria demonstrate the magnitude of the decline
documented here.
The pattern of decline in adult abundance in the LBR appears
similar to declines in other high-profile rainbow trout fisheries
caused by whirling disease. Whirling disease was first detected in
the Bow River in 2016 and subsequent monitoring detected the
parasite throughout the Bow River watershed in February 2017
(Canadian Food Inspection Agency 2017). Although timing of its
first occurrence in the province is unknown, extensive testing for
Myxobolus cerebralis from 1997 to 2001 was negative in the province
Fig. 2. Posterior distributions of rainbow trout density in the Lower Bow River, Alberta, Canada, during 2003–2013. Results are from model 1.
Boxplots designate the median (horizontal black line), the 25th and 75th percentiles (lower and upper box boundaries), the upper and lower
whiskers that extend to 1.5 times the interquartile range, and any remaining “outliers” (circles). Note that floods occurred in 2005 and 2013.
Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)
Cahill et al. 5
Published by NRC Research Press
Can. J. Fish. Aquat. Sci. Downloaded from by UNIV CALGARY on 07/16/18
For personal use only.
Fig. 3. Posterior distributions showing the percent change per year in rainbow trout abundance in the Lower Bow River, Alberta, Canada,
during 2003–2008 (upper panel) and 2003–2013 (lower panel). Percent change per year was calculated as a derived variable in the trend model
Fig. 4. Predictive replicate discrepancy statistics versus observed test statistics for three multisession models fit using the no-U-turn sampling
algorithm in Stan. Posterior predictive pvalues were calculated as the proportion of points above the dashed 1:1 line. Left panel: independent
intercepts model. Middle panel: 2003–2008 trend model. Right panel: 2003–2013 trend model.
Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)
6 Can. J. Fish. Aquat. Sci. Vol. 00, 0000
Published by NRC Research Press
Can. J. Fish. Aquat. Sci. Downloaded from by UNIV CALGARY on 07/16/18
For personal use only.
(Fish and Wildlife Policy, Alberta Environment and Parks, Edmon-
ton, Alberta, unpublished data). Therefore, the best available evi-
dence suggested that whirling disease arrived in the LBR
sometime during 2002–2016. Young-of-year and juvenile rainbow
trout are particularly susceptible to whirling disease, which can
cause recruitment failures (Markiw 1992;Walker and Nehring 1995;
Vincent 1996). However, the impact of whirling disease on wild
rainbow trout populations is highly variable (Bartholomew and
Reno 2002). Management responses to whirling disease varied,
but have included actions such as stopping or limiting stocking of
infected fish into natural waters, emergency bag limit reductions
for recreational anglers, and educational campaigns for the public
(Modin 1998;Nickum 1999;Nehring 2006).
Floods in 2005 and 2013 represent another potential contribut-
ing factor to the LBR decline. Although both floods caused exten-
sive damage, peak flows in the LBR during the 2013 flood were
more than two times those observed during the 2005 flood (Veiga
et al. 2015). These flows in 2013 substantially restructured fish
habitat throughout the LBR (City of Calgary 2018). Previous studies
that have assessed the impact of severe flood events on trout
abundances have found responses ranging from dramatic de-
creases in trout density (Jowett and Richardson 1989;Kitanishi
and Yamamoto 2015) to no change between pre- and postflood
measurement periods (George et al. 2015). High flow events lead-
ing to food supply limitations and subsequent population collapse
were responsible for boom-and-bust population cycles in a tailwa-
ter rainbow trout fishery below the Glen Canyon Dam in Arizona
(Korman et al. 2017). However, this mechanism seems unlikely in
the LBR, as substantial wastewater inputs have increased biologi-
cal productivity (Sosiak 2002;Askey et al. 2007), and these inputs
have remained relatively constant throughout the study period
(see Taube et al. 2016). Intriguingly, abundance declined in both
2005 and 2013 according to postflood surveys (Fig. 3), perhaps
because adult fish were displaced downstream. However, the
mechanism of downstream displacement was unlikely to have
been responsible for the entire decline observed, since declines
also occurred in non-flood years.
High angling effort on the LBR may also have contributed to the
observed decline via postrelease mortality. At present, no esti-
mates of catch-and-release mortality exist in the LBR during the
high effort summer fly-fishery. However, meta-analyses showed
that catch-and-release mortality of rainbow trout ranged from
3% to 9% when fish were captured once via fly-fishing at water
temperatures similar to those observed in the LBR during July–
September (i.e., 10–22 °C; Bartholomew and Bohnsack (2005);
Hühn and Arlinghaus 2011; however see Boyd et al. 2010). The best
available creel survey on the LBR estimated that 49 700 rainbow
trout were captured and released by anglers during July–September
2006 in a 50 km section of the LBR (Council and Ripley 2006), and
our highest estimate of abundance in this section of the river is
20 700 catchable-size trout (Fig. 3; 414 fish·km
× 50 km). Given
these numbers, the exploitation rate for this fishery was poten-
tially 0.07–0.22 for the summer fishery in 2006 (calculated as dead
fish/abundance; Ricker 1975). These values may be high and could
result in a constant mortality policy that does not scale with de-
creases in population size (see also Roughgarden and Smith 1996).
This approximation ignores the potential for sublethal effects,
such as interactions between the number of times a fish is captured
and released and mortality rate (Bartholomew and Bohnsack 2005;
Pope et al. 2007) and increased susceptibility to disease (Pickering
and Pottinger 1989).
A number of potential concerns exist regarding these data and
associated analyses, but we suggest that the trend documented
was unlikely to be caused by these issues. For instance, a key study
limitation was that no block nets were used during data collec-
tion, which was precluded by the large size of the LBR. Conse-
quently, the estimated population sizes may be biased, as these
were generated using closed population models. Furthermore,
the direction and magnitude of this bias depends on the behavior
of animals along the periphery of the study area (Van Katwyk
2014). However, it is unclear how changes in fish behavior along
the periphery of these relatively large study sites could result in
the consistent population decline observed across the entire study
period. Similarly, the decision to use a multisession modeling
framework ignored the possibility that an individual fish oc-
curred in more than 1 year (Royle et al. 2014), and we acknowledge
that these models feature less ecological resolution than the outputs
that would be generated via open population capture–recapture
models (see Korman et al. 2017). The approach used here, how-
ever, was warranted because rainbow trout are short-lived (i.e.,
maximum age of 7 years; Rhodes 2005) and have high instanta-
neous total mortality (van Poorten and Post 2005), which indi-
cated that there was likely high population turnover among years.
Additionally, these data did not support the use of an open popu-
lation model, as only nine total recaptures occurred across years.
Similarly, no estimates of instantaneous tag loss exist, even
though it may have been important (McFarlane et al. 1990;Walsh
and Winkelman 2004;Vandergoot et al. 2012). However, tagging
crews used the same protocols throughout the study period, and
hence differences in instantaneous tag loss among years are un-
likely to explain the consistent decline observed. While we ac-
knowledge some weaknesses in the available data, it is unlikely
that there was a consistent resultant “year” effect (i.e., consistent
changes in animal behavior along the periphery of our study area
or with tag loss differences among years). Lastly, the modeling
approach we used incorporated both process and observation
variability in its assessment of trout abundance, and while we did
not explicitly model population resilience, we postulate the mag-
nitude of decline over two to three generations should not be
ignored in the hope that natural variability returns the popula-
tion to a previous state.
The issues facing the LBR fishery and documented here are
germane to many inland recreational fisheries, which are typified
by poor or infrequent surveillance-style monitoring (Lester et al.
2003;Nate et al. 2003). For example, fundamental data limitations
precluded the ability to determine the cause(s) of the estimated
decline, despite compelling evidence that a decline occurred
(Figs. 2 and 3). A paucity of reliable age and length data, along with
inconsistent mark–recapture surveys, led to an inability to attri-
bute the trend observed to reductions in juvenile or adult survival
and hence to causal variables acting specifically on these life
stages. As a result, the available data did not support the construc-
tion of explicit models to test among hypotheses responsible for
the decline. Although additional information can reduce the
number of alternative explanations for a decline (see Venturelli
et al. 2014), calls to collect more data often fail to recognize key
funding and personnel constraints limiting biologists (Canessa
et al. 2015). Approaches that explicitly account for the marginal
costs of data collection pursuant to management objectives and
constraints could provide a strategic framework for triaging mon-
itoring resources (e.g., value of information analysis; Hansen and
Jones 2008). Similarly, targeted monitoring that views data collec-
tion as a component of a broader structured decision-making pro-
cess, rather than as a lone activity, would maximize the utility of
the data for distinguishing among competing hypotheses and
help guide management actions in the face of uncertainty
(Nichols and Williams 2006). Thus, improving the scientific defen-
sibility of inland fisheries management via strategic monitoring
and experimental management is necessary (McAllister and
Peterman 1992;Post 2013;Hansen et al. 2015).
The threats facing the LBR fishery include stressors that are
both within and beyond the control of local managers, which is
similar to the SOS concept forwarded by S.R. Carpenter et al.
(2017). Specifically, whirling disease and large flood events are
uncontrollable from a management standpoint, with high fishing
effort remaining as the lone stressor that managers can address.
Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)
Cahill et al. 7
Published by NRC Research Press
Can. J. Fish. Aquat. Sci. Downloaded from by UNIV CALGARY on 07/16/18
For personal use only.
For instance, managers can attempt to reduce catch-and-release
mortality through indirect means such as gear restrictions or
warm-weather closures (Cooke and Schramm 2007;Boyd et al.
2010), and indeed, the latter have occurred in the LBR in recent
years when mean water temperatures exceeded 20 °C (i.e., full
river closures in 2015 and voluntary restrictions 2017 where an-
glers were advised rather than required to not fish). Similarly,
programs seeking to educate anglers on proper catch-and-release
techniques may be helpful (Adams 2017). However, we suggest
these may be patchwork solutions to managing release mortality,
given that the LBR is a popular open-access fishery that flows
through a densely populated urban center.
Angler effort restrictions may represent an important manage-
ment option for maintaining rainbow trout in the LBR. Active
management policies that seek to reduce fishing effort are often
socially unacceptable (Schueller et al. 2012) and hence have rarely
been implemented in inland recreational fisheries (Pereira and
Hansen 2003). An approach that has been implemented with suc-
cess, albeit in big-game wildlife management systems, is access
limitation via lottery systems (Boxall 1995;Scrogin et al. 2000).
Similar limited-entry management systems have been recom-
mended and implemented in coastal sport fisheries (Cox et al.
2002;Abbott and Wilen 2009), and a harvest-tag program (but not
effort limitation) is currently used to manage Alberta’s high-effort
walleye (Sander vitreus) fisheries (Sullivan 2003). Empirical evalua-
tions of case studies such as these represent an important area for
future work.
The LBR fishery demonstrates the complexities of maintaining
inland recreational fisheries in the presence of multiple stressors.
A key take-away from this study was that a combination of these
stressors might have pushed rainbow trout in the LBR on an un-
sustainable trajectory during 2003–2013. We identified the most
plausible causes of the rainbow trout decline and suggest these
probably acted in concert to affect the patterns observed, similar
to sustainability issues caused by multiple stressors in other in-
land recreational fisheries (Hansen et al. 2015). All three hypothe-
ses warrant further investigation. Despite ambiguity in the causal
factors responsible for the decline, we posit that managers of the
LBR are left with few options short of active effort restrictions to
maintain the fishery.
We acknowledge the substantial effort that went into data col-
lection by biologists and technicians. C. Cahill acknowledges cor-
respondence with A. Royle and M. Kèry regarding multisession
models and with D. Gwinn regarding Bayesian goodness-of-fit
tests. H. Itô contributed code regarding the noncentered random-
effects parameterization. C. Cahill and K. Wilson are supported
by Vanier Canada Graduate Scholarships, and S. Mogensen and
A. Cantin are supported by NSERC doctoral scholarships. M. Faust,
A. Cameron, and two reviewers greatly improved earlier versions
of this manuscript.
Abbott, J.K., and Wilen, J.E. 2009. Rent dissipation and efficient rationalization
in for-hire recreational fishing. J. Environ. Econ. Manage. 58(3): 300–314.
Adams, A.J. 2017. Guidelines for evaluating the suitability of catch and release
fisheries: lessons learned from Caribbean flats fisheries. Fish. Res. 186: 672–
680. doi:10.1016/j.fishres.2016.09.027.
Allen, M.S., Ahrens, R.N.M., Hansen, M.J., and Arlinghaus, R. 2013. Dynamic
angling effort influences the value of minimum-length limits to prevent
recruitment overfishing. Fish. Manage. Ecol. 20(2–3): 247–257. doi:10.1111/j.
Askey, P.J., Hogberg, L.K., Post, J.R., Jackson, L.J., Rhodes, T., and Thompson, M.S.
2007. Spatial patterns in fish biomass and relative trophic level abundance in
a wastewater enriched river. Ecol. Freshw. Fish, 16(3): 343–353. doi:10.1111/
Barber, I., Hoare, D., and Krause, J. 2000. Effects of parasites on fish behaviour: a
review and evolutionary perspective. Rev. Fish Biol. Fish. 10(2): 131–165. doi:
Bartholomew, A., and Bohnsack, J.A. 2005. A review of catch-and-release angling
mortality with implications for no-take reserves. Rev. Fish Biol. Fish. 15(1–2):
129–154. doi:10.1007/s11160-005-2175-1.
Bartholomew, J.L., and Reno, P.W. 2002. The history and dissemination of whirl-
ing disease. In Whirling disease: reviews and current topics. American Fish-
eries Society, Bethesda, Md. pp. 3–24.
Beamish, R.J., McFarlane, G.A., and Thomson, R.E. 1999. Recent declines in the
recreational catch of coho salmon (Oncorhynchus kisutch) in the Strait of Geor-
gia are related to climate. Can. J. Fish. Aquat. Sci. 56(3): 506–515. doi:10.1139/
Betancourt, M., and Girolami, M. 2013. Hamiltonian Monte Carlo for hierarchi-
cal models. arXiv preprint. arXiv:1312.0906.
Boxall, P.C. 1995. The economic value of lottery-rationed recreational hunting.
Can. J. Agric. Econ. 43(1): 119–131. doi:10.1111/j.1744-7976.1995.tb00111.x.
Boyd, J.W., Guy, C.S., Horton, T.B., and Leathe, S.A. 2010. Effects of catch-and-
release angling on salmonids at elevated water temperatures. North Am. J.
Fish. Manage. 30(4): 898–907. doi:10.1577/M09-107.1.
Canadian Food Inspection Agency. 2017. Confirmed detections of whirling
disease Alberta 2016 [online]. Available from http://www.inspection.
2016/eng/1473443992952/1473443993551 [accessed June 2017].
Canessa, S., Guillera-Arroita, G., Lahoz-Monfort, J.J., Southwell, D.M.,
Armstrong, D.P., Chadès, I., Lacy, R.C., and Converse, S.J. 2015. When do we
need more data? A primer on calculating the value of information for applied
ecologists. Methods Ecol. Evol. 6(10): 1219–1228. doi:10.1111/2041-210X.12423.
Carpenter, B., Gelman, A., Hoffman, M., Lee, D., Goodrich, B., Betancourt, M.,
Brubaker, M.A., Guo, J., Li, P., and Riddell, A. 2017. Stan: a probabilistic
programming language. J. Stat. Softw. 76(1). doi:10.18637/jss.v076.i01.
Carpenter, S.R., Brock, W.A., Hansen, G.J.A., Hansen, J.F., Hennessy, J.M.,
Isermann, D.A., Pedersen, E.J., Perales, K.M., Rypel, A.L., Sass, G.G.,
Tunney, T.D., and Vander Zanden, M.J. 2017. Defining a safe operating space
for inland recreational fisheries. Fish Fish. 18(6): 1150–1160. doi:10.1111/faf.
City of Calgary. 2018. Fish compensation program [online]. Available from http://
aspx?redirect=/fishcompensation [accessed February 2018].
Converse, S.J., and Royle, J.A. 2012. Dealing with incomplete and variable detect-
ability in multi-year, multi-site monitoring of ecological populations. In
Design and analysis of long-term ecological monitoring studies. Edited by
R.A. Gitzen, J.J. Millspaugh, A.B. Cooper, and D.S. Licht. Cambridge University
Press, Cambridge, UK. pp. 426–442. doi:10.1017/CBO9781139022422.025.
Cooke, S.J., and Schramm, H.L. 2007. Catch-and-release science and its applica-
tion to conservation and management of recreational fisheries. Fish. Man-
age. Ecol. 14(2): 73–79. doi:10.1111/j.1365-2400.2007.00527.x.
COSEWIC. 2018. COSEWIC wildlife species assessment: quantitative criteria and
guidelines [online]. Available from
[accessed February 2018].
Council, T., and Ripley, T. 2006. Bow River sport fish population monitoring,
2003 and 2005. Alberta Conservation Association and Alberta Fish and Wild-
life, Lethbridge and Calgary, Alta.
Cox, S.P., Beard, T.D., and Walters, C. 2002. Harvest control in open-access sport
fisheries: hot rod or asleep at the reel? Bull. Mar. Sci. 70(2): 749–761.
Cox, S.P., Walters, C.J., and Post, J.R. 2003. A model-based evaluation of active
management of recreational fishing effort. North Am. J. Fish. Manage. 23(4):
1294–1302. doi:10.1577/M01-228AM.
Crowe-Swords, P. 2016. The economic importance of recreational river use to the
city of Calgary. Calgary River Users Alliance.
Elsayed, E., Faisal, M., Thomas, M., Whelan, G., Batts, W., and Winton, J. 2006.
Isolation of viral haemorrhagic septicaemia virus from muskellunge, Esox
masquinongy (Mitchill), in Lake St Clair, Michigan, USA reveals a new sublin-
eage of the North American genotype. J. Fish Dis. 29(10): 611–619. doi:10.1111/
j.1365-2761.2006.00755.x. PMID:17026670.
Gabry, J. 2015. Shinystan: interactive visual and numerical diagnostics and pos-
terior analysis for Bayesian models. R package version.
Gelman, A. 2008. Scaling regression inputs by dividing by two standard devia-
tions. Stat. Med. 27(15): 2865–2873. doi:10.1002/sim.3107. PMID:17960576.
Gelman, A., Carlin, J.B., Stern, H.S., and Rubin, D.B. 2014. Bayesian data analysis.
Chapman & Hall/CRC, Boca Raton, Fla.
Gelman, A., Lee, D., and Guo, J. 2015. Stan: a probabilistic programming lan-
guage for Bayesian inference and optimization. J. Educ. Behav. Stat. 40(5):
530–543. doi:10.3102/1076998615606113.
George, S.D., Baldigo, B.P., Smith, A.J., and Robinson, G. 2015. Effects of extreme
floods on trout populations and fish communities in a Catskill Mountain
river. Freshw. Biol. 60(12): 2511–2522. doi:10.1111/fwb.12577.
Gilmour, W.M. 1950. A study of the lower Bow River trout with special reference
to taxonomy. Department of Biological Sciences, University of Alberta,
Edmonton, Alta.
Glaser, S.M., Fogarty, M.J., Liu, H., Altman, I., Hsieh, C., Kaufman, L.,
MacCall, A.D., Rosenberg, A.A., Ye, H., and Sugihara, G. 2014. Complex dy-
namics may limit prediction in marine fisheries. Fish Fish. 15(4): 616–633.
Godwin, S.C., Dill, L.M., Reynolds, J.D., and Krkošek, M. 2015. Sea lice, sockeye
Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)
8 Can. J. Fish. Aquat. Sci. Vol. 00, 0000
Published by NRC Research Press
Can. J. Fish. Aquat. Sci. Downloaded from by UNIV CALGARY on 07/16/18
For personal use only.
salmon, and foraging competition: lousy fish are lousy competitors. Can. J.
Fish. Aquat. Sci. 72(7): 1113–1120. doi:10.1139/cjfas-2014-0284.
Hansen, G.J.A., and Jones, M.L. 2008. The value of information in fishery man-
agement. Fisheries, 33(7): 340–348. doi:10.1577/1548-8446-33.7.340.
Hansen, G.J.A., Gaeta, J.W., Hansen, J.F., and Carpenter, S.R. 2015. Learning to
manage and managing to learn: sustaining freshwater recreational fisheries
in a changing environment. Fisheries, 40(2): 56–64. doi:10.1080/03632415.
Hatai, K., and Hoshiai, G. 1992. Mass mortality in cultured coho salmon
(Oncorhynchus kisutch) due to Saprolegnia parasitica Coker. J. Wildlife Dis. 28(4):
532–536. doi:10.7589/0090-3558-28.4.532.
Hilborn, R. 2016. Correlation and causation in fisheries and watershed manage-
ment. Fisheries, 41(1): 18–25. doi:10.1080/03632415.2016.1119600.
Hilborn, R., and Walters, C.J. 1992. Quantitative fisheries stock assessment:
choice, dynamics and uncertainty. Rev. Fish Biol. Fish. 2(2): 177–178. doi:10.
Holling, C.S. 1973. Resilience and stability of ecological systems. Annu. Rev. Ecol.
Syst. 4(1): 1–23. doi:10.1146/
Hubert, W., and Quist, M. 2010. Inland fisheries management in North America.
American Fisheries Society, Bethesda, Md.
Hühn, D., and Arlinghuas, R. 2011. Determinants of hooking mortality in fresh-
water recreational fisheries: a quantitative meta-analysis. Am. Fish. Soc.
Symp. 75: 141–170.
Johnson, S.C., Bravo, S., Nagasawa, K., Kabata, Z., Hwang, J., Ho, J., and Shih, C.
2004. A review of the impact of parasitic copepods on marine aquaculture.
Zool. Stud. 43(2): 229–243.
Johnston, F.D., Post, J.R., Mushens, C.J., Stelfox, J.D., Paul, A.J., and Lajeunesse, B.
2007. The demography of recovery of an overexploited bull trout, Salvelinus
confluentus, population. Can. J. Fish. Aquat. Sci. 64(1): 113–126. doi:10.1139/f06-
Jones, M.L., and Hansen, G.J. 2014. Making adaptive management work: lessons
from the past and opportunities for the future. In Future of fisheries:
perspective for emerging professionals. Edited by W.T. Taylor, A.J. Lynch, and
N.J. Leonard. American Fisheries Society, Bethesda, Md. pp. 443–450.
Jowett, I.G., and Richardson, J. 1989. Effects of a severe flood on instream habitat
and trout populations in seven New Zealand rivers. N. Z. J. Mar. Freshw. Res.
23(1): 11–17. doi:10.1080/00288330.1989.9516335.
Kéry, M., and Royle, J.A. 2015. Applied hierarchical modeling in ecology: analysis
of distribution, abundance and species richness in R and BUGS. Volume 1:
prelude and static models. Academic Press. doi:10.1016/C2013-0-19160-X.
Kéry, M., and Schaub, M. 2012. Bayesian population analysis using WinBUGS: a
hierarchical perspective. Academic Press. doi:10.1016/C2010-0-68368-4.
Kitanishi, S., and Yamamoto, T. 2015. The effects of severe flooding on native
masu salmon and nonnative rainbow trout in the Atsuta River, Hokkaido,
Japan. J. Freshw. Ecol. 30(4): 589–596. doi:10.1080/02705060.2015.1015058.
Korman, J., Yard, M.D., and Kennedy, T.A. 2017. Trends in rainbow trout recruit-
ment, abundance, survival, and growth during a boom-and-bust cycle in a
tailwater fishery. Trans. Am. Fish. Soc. 146(5): 1043–1057. doi:10.1080/00028487.
Lester, N.P., Marshall, T.R., Armstrong, K., Dunlop, W.I., and Ritchie, B. 2003. A
broad-scale approach to management of Ontario’s recreational fisheries.
North Am. J. Fish. Manage. 23(4): 1312–1328. doi:10.1577/M01-230AM.
Link, W.A., and Barker, R.J. 2009. Bayesian inference: with ecological applica-
tions. Academic Press.
Lumsden, J.S., Morrison, B., Yason, C., Russell, S., Young, K., Yazdanpanah, A.,
Huber, P., Al-Hussinee, L., Stone, D., and Way, K. 2007. Mortality event in
freshwater drum Aplodinotus grunniens from Lake Ontario, Canada, associated
with viral haemorrhagic septicemia virus, Type IV. Dis. Aquat. Org. 76(2):
99–111. doi:10.3354/dao076099. PMID:17760383.
Lynch, A.J., Myers, B.J.E., Chu, C., Eby, L.A., Falke, J.A., Kovach, R.P.,
Krabbenhoft, T.J., Kwak, T.J., Lyons, J., and Paukert, C.P. 2016. Climate change
effects on North American inland fish populations and assemblages. Fisher-
ies, 41(7): 346–361. doi:10.1080/03632415.2016.1186016.
Markiw, M.E. 1992. Experimentally induced whirling disease I. Dose response of
fry and adults of rainbow trout exposed to the triactinomyxon stage of
Myxobolus cerebralis. J. Aquat. Anim. Health, 4(1): 40–43. doi:10.1577/1548-
Maunder, M.N., and Punt, A.E. 2013. A review of integrated analysis in fisheries
stock assessment. Fish. Res. 142: 61–74. doi:10.1016/j.fishres.2012.07.025.
McAllister, M.K., and Peterman, R.M. 1992. Experimental design in the manage-
ment of fisheries: a review. North Am. J. Fish. Manage. 12(1): 1–18. doi:10.1577/
McFarlane, G., Wydoski, R.S., and Prince, E.D. 1990. External tags and marks.
Am. Fish. Soc. Symp. 7: 9–29.
Modin, J. 1998. Whirling disease in California: a review of its history, distribu-
tion, and impacts, 1965–1997. J. Aquat. Anim. Health, 10(2): 132–142. doi:10.
Monnahan, C.C., Thorson, J.T., and Branch, T.A. 2017. Faster estimation of Bayes-
ian models in ecology using Hamiltonian Monte Carlo. Methods Ecol. Evol.
8(3): 339–348. doi:10.1111/2041-210X.12681.
Nate, N.A., Bozek, M.A., Hansen, M.J., Ramm, C.W., Bremigan, M.T., and
Hewett, S.W. 2003. Predicting the occurrence and success of walleye popula-
tions from physical and biological features of northern Wisconsin lakes.
North Am. J. Fish. Manage. 23(4): 1207–1214. doi:10.1577/M02-097.
Nehring, R.B. 2006. Colorado’s cold water fisheries: whirling disease case histo-
ries and insights for risk management. Colorado Division of Wildlife, Aquatic
Wildlife Research.
Nehring, R.B., and Walker, P.G. 1996. Whirling disease in the wild. A fresh
approach to stock assessment. Fish. Mag. 21(6): 28.
Nichols, J.D., and Williams, B.K. 2006. Monitoring for conservation. Trends Ecol.
Evol. 21(12): 668–673. doi:10.1016/j.tree.2006.08.007. PMID:16919361.
Nickum, D. 1999. Whirling disease in the United States. Trout Unlimited’s Cold-
water Conservation Fund, Boulder, Colo.
Ormerod, S., Dobson, M., Hildrew, A., and Townsend, C. 2010. Multiple stressors
in freshwater ecosystems. Freshw. Biol. 55(s1): 1–4. doi:10.1111/j.1365-2427.2009.
Paukert, C.P., Glazer, B.A., Hansen, G.J.A., Irwin, B.J., Jacobson, P.C.,
Kershner, J.L., Shuter, B.J., Whitney, J.E., and Lynch, A.J. 2016. Adapting in-
land fisheries management to a changing climate. Fisheries, 41(7): 374–384.
Paul, A.J. 2013. Environmental flows and recruitment of walleye (Sander vitreus)
in the Peace–Athabasca Delta. Can. J. Fish. Aquat. Sci. 70(2): 307–315. doi:10.
Pereira, D.L., and Hansen, M.J. 2003. A perspective on challenges to recreational
fisheries management: summary of the symposium on active management
of recreational fisheries. North Am. J. Fish. Manage. 23(4): 1276–1282. doi:10.
Pickering, A.D., and Pottinger, T.G. 1989. Stress responses and disease resistance
in salmonid fish: effects of chronic elevation of plasma cortisol. Fish Physiol.
Biochem. 7(1–6): 253–258. doi:10.1007/BF00004714. PMID:24221779.
Plummer, M. 2003. JAGS: A program for analysis of Bayesian graphical models
using Gibbs sampling. In Proceedings of the 3rd International Workshop on
Distributed Statistical Computing. Vol. 124. Vienna, Austria. p. 125.
Pope, K.L., Wilde, G.R., and Knabe, D.W. 2007. Effect of catch-and-release angling
on growth and survival of rainbow trout, Oncorhynchus mykiss. Fish. Manage.
Ecol. 14(2): 115–121. doi:10.1111/j.1365-2400.2007.00531.x.
Post, J. 2013. Resilient recreational fisheries or prone to collapse? A decade of
research on the science and management of recreational fisheries. Fish. Man-
age. Ecol. 20(2–3): 99–110. doi:10.1111/fme.12008.
Post, J.R., Sullivan, M., Cox, S., Lester, N.P., Walters, C.J., Parkinson, E.A.,
Paul, A.J., Jackson, L., and Shuter, B.J. 2002. Canada’s recreational fisheries:
the invisible collapse? Fisheries, 27(1): 6–17. doi:10.1577/1548-8446(2002)027
Post, J.R., Mushens, C., Paul, A., and Sullivan, M. 2003. Assessment of alternative
harvest regulations for sustaining recreational fisheries: model development
and application to bull trout. North Am. J. Fish. Manage. 23(1): 22–34. doi:10.
Post, J.R., van Poorten, B.T., Rhodes, T., Askey, P., and Paul, A. 2006. Fish entrain-
ment into irrigation canals: an analytical approach and application to the
Bow River, Alberta, Canada. North Am. J. Fish. Manage. 26(4): 875–887. doi:
Punt, A.E., and Hilborn, R. 1997. Fisheries stock assessment and decision analy-
sis: the Bayesian approach. Rev. Fish Biol. Fish. 7(1): 35–63. doi:10.1023/A:
R Core Team. 2017. R: A language and environment for statistical computing.
R Foundation for Statistical Computing, Vienna, Austria.
Rahel, F.J. 2016. Changing philosophies of fisheries management as illustrated
by the history of fishing regulations in Wyoming. Fisheries, 41(1): 38–48.
Rhodes, T. 2005. The immediate and short term impacts of catch-and-release
angling on migrating and pre-spawning condition rainbow trout (Oncorhynchus
mykiss) in the Bow River, Alberta. Department of Biological Sciences, Univer-
sity of Calgary, Calgary, Alta.
Ricker, W.E. 1975. Computation and interpretation of biological statistics of fish
populations. Bull. Fish. Res. Board Can. 191: 382.
Roughgarden, J., and Smith, F. 1996. Why fisheries collapse and what to do about
it. Proc. Natl. Acad. Sci. 93(10): 5078– 5083. doi:10.1073/pnas.93.10.5078. PMID:
Royle, J.A. 2009. Analysis of capture–recapture models with individual covari-
ates using data augmentation. Biometrics, 65(1): 267–274. doi:10.1111/j.1541-
0420.2008.01038.x. PMID:18422793.
Royle, J.A., and Dorazio, R.M. 2008. Hierarchical modeling and inference in
ecology: the analysis of data from populations, metapopulations and com-
munities. Academic Press. doi:10.1016/B978-0-12-374097-7.X0001-4.
Royle, J.A., Chandler, R.B., Sollmann, R., and Gardner, B. 2014. Spatial capture–
recapture. Academic Press. doi:10.1016/C2012-0-01222-7.
Schueller, A.M., Fayram, A.H., and Hansen, M.J. 2012. Simulated equilibrium
walleye population density under static and dynamic recreational angling
effort. North Am. J. Fish. Manage. 32(5): 894–904. doi:10.1080/02755947.2012.
Scrogin, D., Berrens, R.P., and Bohara, A.K. 2000. Policy changes and the demand
for lottery-rationed big game hunting licenses. J. Agric. Resour. Econ. 25(2):
Seber, G.A.F. 1982. The estimation of animal abundance. Macmillan Publishing
Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)
Cahill et al. 9
Published by NRC Research Press
Can. J. Fish. Aquat. Sci. Downloaded from by UNIV CALGARY on 07/16/18
For personal use only.
Sosiak, A. 2002. Long-term response of periphyton and macrophytes to reduced
municipal nutrient loading to the Bow River (Alberta, Canada). Can. J. Fish.
Aquat. Sci. 59(6): 987–1001. doi:10.1139/f02-071.
Stan Development Team. 2016. Rstan: the R interface to Stan.
Statistics Canada. 2016. Census profile, 2016 census [online]. Available from
Count&SearchText=calgary&SearchType=Begins&SearchPR=01&B1=All [ac-
cessed February 2018].
Sugihara, G., May, R., Ye, H., Hsieh, C., Deyle, E., Fogarty, M., and Munch, S. 2012.
Detecting causality in complex ecosystems. Science, 338(6106): 496– 500. doi:
10.1126/science.1227079. PMID:22997134.
Sullivan, M.G. 2003. Active management of walleye fisheries in Alberta: dilem-
mas of managing recovering fisheries. North Am. J. Fish. Manage. 23(4):
1343–1358. doi:10.1577/M01-232AM.
Taube, N., He, J., Ryan, M.C., and Valeo, C. 2016. Relative importance of P and N
in macrophyte and epilithic algae biomass in a wastewater-impacted oligo-
trophic river. Environ. Monit. Assess. 188(8): 494. doi:10.1007/s10661-016-5493-y.
Van Katwyk, K.E. 2014. Empirical validation of closed population abundance
estimates and spatially explicit density estimates using a censused popula-
tion of North American red squirrels. University of Alberta, Edmonton, Alta.
van Poorten, B.T., and Post, J.R. 2005. Seasonal fishery dynamics of a previously
unexploited rainbow trout population with contrasts to established fisher-
ies. North Am. J. Fish. Manage. 25(1): 329–345. doi:10.1577/M03-225.1.
Vandergoot, C.S., Brenden, T.O., Thomas, M.V., Einhouse, D.W., Cook, H.A., and
Turner, M.W. 2012. Estimation of tag shedding and reporting rates for Lake
Erie jaw-tagged walleyes. North Am. J. Fish. Manage. 32(2): 211–223. doi:10.
Veiga, V.B., Hassan, Q.K., and He, J. 2015. Development of flow forecasting mod-
els in the Bow River at Calgary, Alberta, Canada. Water, 7(1): 99–115. doi:10.
Venturelli, P., Bence, J., Brenden, T., Lester, N., and Rudstam, L. 2014. Mille Lacs
Lake Walleye blue ribbon panel data review and recommendations for future
data collection and management. Minnesota Department of Natural Re-
Vincent, E.R. 1996. Whirling disease and wild trout. Fisheries, 21(6): 32–33.
Walker, P.G., and Nehring, R.B. 1995. An investigation to determine the cause(s)
of the disappearance of young wild rainbow trout in the upper Colorado
River. Middle Park, Colorado. Colorado Division of Wildlife Report, Denver,
Walsh, M.G., and Winkelman, D.L. 2004. Anchor and visible implant elastomer
tag retention by hatchery rainbow trout stocked into an Ozark stream. North
Am. J. Fish. Manage. 24(4): 1435–1439. doi:10.1577/M03-246.1.
Walters, C.J. 1986. Adaptive management of renewable resources. Macmillan
Publishers Ltd.
Warren, D.R., Ernst, A.G., and Baldigo, B.P. 2009. Influence of spring floods on
year-class strength of fall- and spring-spawning Salmonids in Catskill moun-
tain streams. Trans. Am. Fish. Soc. 138(1): 200–210. doi:10.1577/T08-046.1.
Welcomme, R. 2008. Inland fisheries: ecology and management. John Wiley &
Appendix A. Methods used to evaluate goodness-
of-fit via posterior predictive pvalues for the
multisession models used in our analyses
We evaluated our models using posterior predictive pvalues
(Gelman et al. 2014). Seber (1982) provided an asymptotic dis-
crepancy statistic based on the
distrubtion for a single closed
capture–recapture model assuming multinomial sampling:
(A.1) T
where x
is the number of individual fish with capture history
is the expected number of animals with that capture history
calculated at the parameter estimates, and
represents the set of
capture histories excluding the all-zero history of “0000”. Link
and Barker (2009) extended this statistic to a Bayesian posterior
predictive check, such that e
is calculated for each draw (h)ofthe
posterior. We added an additional summation symbol to the sta-
tistic described by Link and Barker (2009) to evaluate the global fit
of the multisession models for all years considered in our analy-
ses. We calculated our observed test statistics as
(A.2) Th
where e
hcorresponds to the expected number of animals with
capture history parameter
based on the parameter estimates for
posterior draw h. We then compared Th
obs with a replicate test
statistic calculated as
(A.3) Th
rep e
where xh
rep is drawn from the posterior predictive distribution for
(Link and Barker 2009). The posterior predictive pvalue for
each model was then calculated as the proportion of Th
rep >Th
(Gelman et al. 2014), and we judged pvalues > 0.05 and <0.95 as
indications of sufficient fit (see also Royle et al. 2014).
Appendix Tables A1–A3 appear on the following pages.
Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)
10 Can. J. Fish. Aquat. Sci. Vol. 00, 0000
Published by NRC Research Press
Can. J. Fish. Aquat. Sci. Downloaded from by UNIV CALGARY on 07/16/18
For personal use only.
Table A1. Marginal posterior summaries of parameters (2.5%, 50%,
and 97.5% percentiles, posterior standard deviation (SD)) and diagnos-
tic values (Neff = number of effective samples, R
ˆ= Gelman–Rubin
statistic) from the independent intercepts multisession model 1 fitted
to Lower Bow River rainbow trout data during 2003–2013.
Parameter 2.5% 50% 97.5% SD Neff R
0.16 0.20 0.39 0.06 2998 1
0.15 0.19 0.29 0.04 5342 1
0.15 0.21 0.34 0.05 4176 1
0.12 0.15 0.23 0.03 3810 1
0.15 0.18 0.25 0.03 3823 1
0.07 0.09 0.16 0.02 7360 1
0.03 0.04 0.06 0.01 5750 1
02003 7.13 7.49 8.04 0.23 4348 1
02005 7.07 7.34 7.74 0.17 6097 1
02007 7.08 7.42 7.92 0.21 4725 1
02008 6.85 7.11 7.51 0.17 4171 1
02011 7.08 7.28 7.60 0.13 4205 1
02012 6.27 6.63 7.14 0.22 8221 1
02013 5.52 5.79 6.23 0.18 6248 1
1256 1789 3083 479.62 3520 1
1181 1542 2287 283.01 5264 1
1199 1672 2749 407.29 4134 1
950 1217 1827 226.03 3733 1
1200 1458 1995 203.47 3723 1
532 759 1259 188.93 7262 1
257 326 501 0.86 5289 1
0.02 0.29 0.82 0.22 2945 1
0.01 0.23 0.70 0.19 3634 1
0.01 0.29 0.81 0.22 2892 1
0.01 0.28 0.80 0.21 2992 1
0.01 0.24 0.70 0.19 2179 1
0.01 0.28 0.80 0.21 5045 1
0.02 0.39 0.99 0.26 4676 1
0.02 0.05 0.07 0.01 4690 1
0.02 0.04 0.07 0.01 4684 1
0.03 0.06 0.09 0.02 4586 1
0.02 0.04 0.06 0.01 4799 1
0.04 0.07 0.10 0.01 6204 1
0.04 0.07 0.10 0.01 6084 1
0.03 0.06 0.08 0.01 6285 1
0.04 0.07 0.10 0.02 5971 1
0.03 0.06 0.10 0.02 4670 1
0.03 0.06 0.09 0.02 4655 1
0.04 0.06 0.09 0.02 4558 1
0.02 0.05 0.07 0.01 7573 1
0.06 0.11 0.15 0.02 4121 1
0.07 0.12 0.16 0.02 4128 1
0.05 0.09 0.13 0.02 4360 1
0.01 0.05 0.07 0.01 6092 1
0.05 0.09 0.11 0.02 4414 1
0.06 0.09 0.12 0.02 4430 1
0.09 0.14 0.17 0.02 4268 1
0.05 0.08 0.10 0.01 4624 1
0.01 0.04 0.07 0.01 6534 1
0.04 0.08 0.12 0.02 8591 1
0.04 0.08 0.13 0.02 8489 1
0.05 0.10 0.15 0.03 8261 1
0.06 0.13 0.19 0.03 7038 1
0.07 0.15 0.21 0.04 6911 1
0.10 0.20 0.28 0.05 6515 1
0.06 0.12 0.19 0.03 7039 1
Table A2. Marginal posterior summaries of parameters (2.5%, 50%,
and 97.5% percentiles, posterior SD) and diagnostic values (Neff =
number of effective samples, R
ˆ= Gelman–Rubin statistic) from the
multisession trend model 2 fitted to Lower Bow River rainbow trout
data during 2003–2008.
Parameter 2.5% 50% 97.5% SD Neff R
0.16 0.22 0.32 0.04 5311 1
0.17 0.20 0.24 0.02 4650 1
0.15 0.18 0.22 0.02 4183 1
0.13 0.17 0.23 0.02 4009 1
6.96 7.19 7.50 0.14 4172 1
–0.68 –0.25 0.19 0.22 4745 1
1311 1760 2532 312.61 5300 1
1320 1579 1958 162.66 4688 1
1182 1414 1787 155.09 4289 1
1054 1325 1815 155.09 3976 1
0.01 0.27 0.71 0.19 3897 1
0.01 0.24 0.61 0.16 4370 1
0.01 0.20 0.56 0.15 5676 1
0.01 0.36 0.81 0.21 3074 1
0.03 0.05 0.07 0.01 5945 1
0.03 0.04 0.06 0.01 5974 1
0.04 0.06 0.09 0.01 5515 1
0.02 0.04 0.06 0.01 5949 1
0.05 0.07 0.09 0.01 6954 1
0.05 0.07 0.09 0.01 6913 1
0.04 0.05 0.07 0.01 7410 1
0.05 0.07 0.09 0.01 6902 1
0.06 0.08 0.10 0.01 6513 1
0.05 0.08 0.10 0.01 6437 1
0.07 0.10 0.13 0.01 5944 1
0.02 0.05 0.07 0.01 7540 1
0.06 0.10 0.13 0.02 4077 1
0.07 0.10 0.15 0.02 4017 1
0.05 0.08 0.11 0.02 4147 1
0.01 0.05 0.07 0.02 5993 1
Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)
Cahill et al. 11
Published by NRC Research Press
Can. J. Fish. Aquat. Sci. Downloaded from by UNIV CALGARY on 07/16/18
For personal use only.
Table A3. Marginal posterior summaries of parameters (2.5%, 50%,
and 97.5% percentiles, posterior SD) and diagnostic values (Neff =
number of effective samples, R
ˆ= Gelman–Rubin statistic) from the
multisession trend model 3 fitted to Lower Bow River rainbow trout
data during 2003–2013.
Parameter 2.5% 50% 97.5% SD Neff R
0.18 0.23 0.30 0.03 4158 1
0.17 0.20 0.24 0.02 4105 1
0.15 0.17 0.20 0.01 3767 1
0.14 0.16 0.19 0.01 3458 1
0.15 0.18 0.25 0.02 2950 1
0.06 0.07 0.10 0.01 4145 1
0.04 0.06 0.09 0.01 4194 1
6.91 7.04 7.19 0.07 3871 1
–0.83 –0.51 –0.20 0.16 4031 1
–1.00 –0.69 –0.39 0.16 3831 1
1423 1810 2413 252.17 4181 1
1324 1579 1954 160.73 4158 1
1209 1385 1621 104.82 4061 1
1133 1285 1492 91.45 3695 1
1211 1464 1971 197.43 2457 1
462 589 793 85.53 4320 1
338 466 683 89.02 4108 1
0.02 0.28 0.70 0.19 3485 1
0.01 0.24 0.61 0.17 3770 1
0.01 0.19 0.52 0.14 6504 1
0.02 0.32 0.67 0.18 3608 1
0.01 0.25 0.69 0.19 2351 1
0.01 0.18 0.55 0.15 8843 1
0.24 0.82 1.29 0.26 3407 1
0.03 0.05 0.07 0.01 4860 1
0.03 0.04 0.06 0.01 5087 1
0.04 0.06 0.08 0.01 4718 1
0.03 0.04 0.06 0.01 5115 1
0.05 0.07 0.09 0.01 5615 1
0.05 0.07 0.09 0.01 5552 1
0.04 0.05 0.07 0.01 5960 1
0.04 0.07 0.09 0.01 5305 1
0.06 0.07 0.10 0.01 8200 1
0.06 0.08 0.10 0.01 7995 1
0.08 0.10 0.13 0.01 7205 1
0.02 0.05 0.07 0.01 7540 1
0.08 0.10 0.13 0.01 5119 1
0.09 0.11 0.14 0.01 5082 1
0.06 0.09 0.1 0.01 5371 1
0.01 0.05 0.07 0.02 6012 1
0.05 0.08 0.11 0.01 3032 1
0.06 0.09 0.12 0.02 2870 1
0.09 0.14 0.17 0.02 2717 1
0.05 0.08 0.10 0.01 3513 1
0.01 0.04 0.07 0.01 3701 1
0.07 0.10 0.14 0.02 6675 1
0.07 0.11 0.15 0.02 6653 1
0.09 0.13 0.18 0.02 6140 1
0.04 0.08 0.14 0.03 4172 1
0.04 0.09 0.15 0.03 4148 1
0.06 0.12 0.21 0.04 4025 1
0.03 0.07 0.13 0.03 4137 1
Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)
12 Can. J. Fish. Aquat. Sci. Vol. 00, 0000
Published by NRC Research Press
Can. J. Fish. Aquat. Sci. Downloaded from by UNIV CALGARY on 07/16/18
For personal use only.
... Similarly, at the regional scale with many fisheries for the same species (e.g., multiple lakes in a district), lower regional effort results in greater stock abundance, lower proportions of fisheries being overfished or collapsed, and higher angler catch rates (Hunt et al. 2011;Matsumura et al. 2019). Furthermore, limiting entry via a license cap may be necessary to reduce the human contribution to target stock decline when uncontrollable environmental stressors are present (Cahill et al. 2018), to achieve regional maximum sustainable yield while preventing localized collapses (Parkinson et al. 2004;Post and Parkinson 2012;Matsumura et al. 2019), or to create a safe operating space in which the fishery will remain sustainable under dynamic conditions (Carpenter et al. 2017). ...
... Regionally, managers can implement combinations of regulations that all achieve biological goals but offer different angling experiences in each fishery, thereby providing opportunities tailored to each specialized angler group (van Poorten and Camp 2019). It is important to note that the least socially acceptable option of limiting entry may be required to salvage a fishery Cahill et al. 2018) if pervasive effort continues to maintain fishing mortality at an unsustainable level even after implementation of particularly restrictive regulations and/or aggressive habitat enhancement and stocking programs. Ultimately, managers must decide whether to take a more proactive approach, in which stricter regulations are implemented early to prevent or mitigate future stock decline, or a more reactive approach, in which progressively stricter regulations are implemented in response to ongoing stock decline; in fisheries where anglers have minimal ability to select among species, the reactive approach may incur substantially higher costs if persistent bycatch of an imperiled species eventually necessitates a fishing closure inclusive of highly-valued target species (Anderson et al. 2013). ...
... The most biologically effective regulations in commercial fisheries, those establishing a total allowable catch (Anderson et al. 2019), are largely absent in recreational fisheries (see exceptions in Hansen et al. 1991;Tsehaye et al. 2016). Although limited entry is increasingly suggested as a solution to excessive effort in recreational fisheries (e.g., Cox et al. 2003;Schueller et al. 2012;Post & Parkinson 2012;Cahill et al. 2018), this strategy often fails to meet biological goals in commercial fisheries because economic motivations induce increases in individual effort and investments in gear enabling higher catchability of target species (Eigaard et al. 2014;Anderson et al. 2019). A recreational limited entry program could overcome the issue of increases in individual effort by not only limiting the number of anglers but also the number of fishing days per angler; however, this would likely face strong opposition due to further reduction in fishing opportunity. ...
Full-text available
Recreational fishing is practiced by ~ 350 million people globally, and while it historically has been thought to have minimal ecological impact relative to commercial fishing, numerous recreational fisheries have recently declined or collapsed. The potential for recreational fishing to contribute to ecological decline, as well as the incentives of recreational anglers that are distinct from those of commercial fishers, highlights the need for greater understanding of recreational fisheries regulatory options. To aid managers in the decision-making process, we conduct the first comparative review of all seven major approaches to recreational fisheries regulation: harvest size restrictions, harvest quantity restrictions, spatial management, temporal restrictions, accessibility restrictions, rights-based management, and gear restrictions. We provide a synthetic guide for students and practitioners covering how these regulations can benefit target stocks, their potential limitations in achieving sustainability, and angler perceptions of their relative effectiveness and behavioral impositions. Considering the strengths and weaknesses of each strategy, we identify three key fishery metrics that together can guide selection of a suitable combination of regulations that will achieve the requisite biological outcome without restricting angler behavior more than is necessary. With this perspective, we reflect on uncertainties that complicate informed and effective, recreational fisheries regulation.
... shrubs are either Water 2021, 13, 990 3 of 14 dead or severely degraded. The fish population in the lower Bow River fell by an average of 46.5% during 2003-2013 in response to the potential stressors of whirling disease, flooding, and release mortality [36]. The Alberta Institute for Wildlife Conservation in collaboration with site management attempted to rewild ASCCA with the native beaver species; a pair of beavers was introduced on 18 May 2016 [37] but not to Pine Creek. ...
... shrubs are either dead or severely degraded. The fish population in the lower Bow River fell by an average of 46.5% during 2003-2013 in response to the potential stressors of whirling disease, flooding, and release mortality [36]. The Alberta Institute for Wildlife Conservation in collaboration with site management attempted to rewild ASCCA with the native beaver species; a pair of beavers was introduced on 18 May 2016 [37] but not to Pine Creek. ...
Full-text available
Beaver dam analogues (BDAs) are becoming an increasingly popular stream restoration technique. One ecological function BDAs might help restore is suitable habitat conditions for fish in streams where loss of beaver dams and channel incision has led to their decline. A critical physical characteristic for fish is stream temperature. We examined the thermal regime of a spring-fed Canadian Rocky Mountain stream in relation to different numbers of BDAs installed in series over three study periods (April–October; 2017–2019). While all BDA configurations significantly influenced stream and pond temperatures, single- and double-configuration BDAs incrementally increased stream temperatures. Single and double configuration BDAs warmed the downstream waters of mean maxima of 9.9, 9.3 °C by respective mean maxima of 0.9 and 1.0 °C. Higher pond and stream temperatures occurred when ponding and discharge decreased, and vice versa. In 2019, variation in stream temperature below double-configuration BDAs was lower than the single-configuration BDA. The triple-configuration BDA, in contrast, cooled the stream, although the mean maximum stream temperature was the highest below these structures. Ponding upstream of BDAs increased discharge and resulted in cooling of the stream. Rainfall events sharply and transiently reduced stream temperatures, leading to a three-way interaction between BDA configuration, rainfall and stream discharge as factors co-influencing the stream temperature regime. Our results have implications for optimal growth of regionally important and threatened bull and cutthroat trout fish species.
... These systems face an increasing number of threats to their resilience and sustainability, including overexploitation (Post et al. 2002;Embke et al. 2019), habitat degradation due to global (e.g., climate change) and local (e.g., watershed development, shoreline alterations) change , species invasions, and shifts in fish community dominance due to unbalanced fisheries practices (e.g., cultivation by preferentially harvesting some species, while practicing catch-and-release for others; Walters and Kitchell 2001;. The ability of fish populations to maintain resiliency in the face of multiple stressors is regulated by their compensatory response to increased mortality or reduced population productivity (i.e., surplus production ;Ricker 1975;Goodyear 1980;Cahill et al. 2018). The absence of compensation results in lost population productivity and poor recovery potential, necessitating increased expansion of stocking and/or use of restrictive regulations to maintain populations (Shertzer and Prager 2007). ...
... We compared the utility of five different plausible stockrecruitment models to assess the strength of evidence for density-dependent dynamics in Walleye recruitment across lakes. Because of the benefits of improved precision in fitting complex models using hierarchical approaches and examination of the posterior distributions of parameters of interest, all models were fit using Bayesian inference (Myers and Mertz 1998;Cahill et al. 2018Cahill et al. , 2020. Recruitment was modeled in log-space to improve model convergence and allow for comparability of model fit across models. ...
Inland fisheries face increasing threats to their sustainability. Despite speculation that depensation may exacerbate the effects of stressors on population resiliency, depensation has not been empirically explored in freshwater fisheries. Declining productivity of walleye (Sander vitreus) populations in northern Wisconsin foreshadows an underlying change in naturally reproduced juvenile walleye survival. We used long-term stock and recruitment data from lakes in the Ceded Territory of Wisconsin (CTWI) to quantify density-dependent trends in juvenile walleye survival and tested for the prevalence of depensation using the q parameter of Liermann and Hilborn (1997). Of 82 walleye populations evaluated, about half exhibited depensatory recruitment. An analysis of the global q for all populations examined suggested that the posterior probability of depensatory dynamics was about 0.89. In addition, there were few clear cases of compensation - most populations exhibited weak density-dependence. The general lack of strong compensatory recruitment across walleye populations could leave these stocks vulnerable to stressors and unresponsive to rehabilitation. We present multiple lines of evidence to suggest that depensation is a plausible phenomenon explaining declines in CTWI walleye populations and may be implicated in other invisible collapses of freshwater fisheries.
... In our case study, we applied benefit transfer to reservoir fishing in Nebraska and estimated the economic value of different policy scenarios. These scenarios-controlling effort and increasing catch rates-were selected because they represent novel strategies that have been suggested as potentially useful for overcoming challenging management issues, like catch rate hyperstability (Cahill et al. 2018;Feiner et al. 2020) or maintaining high satisfaction without impairing stock density (Camp et al. 2015;. As managers shift from reservoir-scale to landscape-scale approaches that recognize the heterogeneity inherent among angler preferences and reservoir dynamics, interdisciplinary approaches like benefit transfer can provide necessary tools for monitoring and evaluation. ...
Full-text available
Economic assessments are rarely applied to inland recreational fisheries for management purposes, especially when compared to fish, habitat, and creel assessments, yet economic assessments can provide critical information for management decisions. We provide a brief overview of economic value, key terminology, and existing economic techniques to address these issues. Benefit transfer, a technique used to measure economic value when an original analysis is not practicable, is conducted by drawing on existing estimates of economic value in similar contexts. We describe an application of benefit transfer to measure the economic value of several recreational fisheries in Nebraska, USA. We examine two approaches to benefit transfer—value transfer and function transfer—which we demonstrate estimate similar economic values for fishing site access but substantially different economic values for catch rate improvements at some reservoirs. We encourage agencies that are responsible for inland recreational fisheries management to consider economic assessment, especially benefit transfer, as a critical tool in the management toolbox.
... We fitted all models in Stan, which implements Hamiltonian Monte Carlo via the No-U-Turn sampling algorithm and is more efficient than standard MCMC for many ecological models (Monnahan et al. 2017;Cahill et al. 2018). We fitted a total of 220 Bayesian state-space stock assessment models (i.e., 55 lakes x 2 structural assumptions about stock-recruitment relationships x low and high ϕ values = 220 total fits). ...
... Combining information on exploitation with abundance estimates can help to expose population-level effects of fishing (Jennings and Blanchard 2004). Having reliable estimates of abundance is also important because it allows managers to assess population trends, particularly in response to management actions or environmental stressors (Farrell and Werner 1999;Cahill et al. 2018). ...
Muskellunge Esox masquinongy fisheries have increased in abundance and popularity in numerous southern and mid‐Atlantic rivers. However, the paucity of information for these fisheries has limited biologists’ ability to effectively manage these resources. We utilized data from simultaneously conducted fishery dependent tag‐return and fishery independent capture‐recapture studies to estimate catch‐and‐release exploitation and abundance of Muskellunge in the James River, Virginia. During winter electrofishing surveys in 2016–2019, we tagged 747 individual Muskellunge with dart tags and passive integrated transponders. Thirty‐seven percent of tags were returned by anglers and 33% of tagged Muskellunge were recaptured at least once during subsequent electrofishing surveys. Eighty‐eight percent of angled fish were caught by individuals targeting Muskellunge, and only 1.1% of returned tags were from fish that were harvested. Estimates of catch‐and‐release exploitation (u), using a Brownie dead‐recovery model, differed between the upper (u = 0.57 ± 0.09) and lower (u = 1.00 ± 0.10) reaches of our study area. Similarly, densities of adult (≥762 mm) Muskellunge differed between these two areas (upriver = 0.80/ha, downriver = 0.50/ha), but were comparable to densities observed in lakes and at least one other river. The greatest site‐specific density (4.69/ha) was observed at a 1.6 km reach below the lowest dam, which concentrated fish and restricted upstream movement. Given the high rates of catch‐and‐release exploitation in the James River and growing popularity of Muskellunge angling in southern waters, investigations into the effects of this level of angling on Muskellunge populations (e.g., trophy potential, abundance), will provide managers with the information needed to develop comprehensive Muskellunge management strategies.
Full-text available
Abstract—Walleye (Sander vitreus) populations in Alberta, Canada collapsed by the mid-1990s and were a case study in the paper Canada’s Recreational Fisheries: The Invisible Collapse? Here we fit age-structured population dynamics models to data from a landscape-scale monitoring program to assess Walleye population status and reconstruct recruitment dynamics following the invisible collapse. Assessments indicated that populations featured low F_msy values of approximately 0.2-0.3 under conservative assumptions for the stock-recruitment relationship but that many populations were lightly exploited during 2000-2018. Recruitment reconstructions showed that recovery from collapse in 33/55 lakes was driven in part by large positive recruitment anomalies that occurred during 1998-2002. Additionally, 15/55 lakes demonstrated cyclic recruitment dynamics. The documented recruitment anomalies and cyclic fluctuations could be due to environmental effect(s) or cannibalism, and experimentation is likely necessary to resolve this uncertainty. These findings contribute new information on the recovery dynamics of Walleye following the invisible collapse, and demonstrate the effectiveness of coupling traditional fisheries science models with broad-scale monitoring data to improve understanding of population dynamics and sustainability across landscapes.
Full-text available
This study describes some of the characteristics of the spawning runs of rainbow and brown trout from Lake Otamangakau, into Te Whaiau Stream. More than 55,000 fish undergoing spawning migrations were enumerated in a fish trap between 1994 and 2013. During that period, the spawning run size increased for both species, but particularly for rainbow trout, which exhibited a concurrent decline in average weight. The timing of the runs was consistent across both species, sexes, and maturation statuses but was affected by the average weight of the fish and the run size. The level of iteroparity revealed that 40% and 55% of the runs were made up of previous spawner rainbow and brown trout, respectively. Detections of PIT-tagged fish at the trap indicated that rainbow and brown trout could spawn up to six and eight times, respectively. Remarkably, repeat spawning fish of both species and sexes returned within a few days of the same date each year over consecutive spawning events. Rainbow and brown trout grew substantially after spawning up to the 3rd and 4th spawning, respectively. The results of this study are relevant for informing management of the popular Lake Otamangakau trout fishery, in particular achieving a balance between satisfactory CPUE and large fish size.
Growing interest in apps to collect recreational-fisheries data requires that relationships between self-reported data and other fisheries data are evaluated, and that potential biases are assessed. This study compared results from a mobile-phone application and website for anglers (MyCatch) to results from three types of fisheries surveys – 1 provincial-level mail survey, 2 creel, and 17 gillnet surveys. Results suggest that an app/website can (i) recruit users that have a broad spatial distribution that is similar to conventional surveys, (ii) generate data that capture regional fishing patterns (2218 trips on 289 lakes and 90 streams/rivers), and (iii) provide catch rate estimates that are similar to those from other fisheries-dependent surveys. Some potential biases in app users (e.g., urban bias) and in the relative composition of species caught provincially were identified. The app was not a suitable tool for estimating fish abundance and relative community composition. Our study demonstrates how apps can/cannot provide a complementary data-collection tool for recreational-fisheries monitoring, but further research is needed to determine the applicability of our findings to other fisheries contexts.
Primary demographic attributes such as somatic growth, mortality, size- and age-at-maturity, fecundity, and recruitment drive the structure of fish populations, and these processes are also influenced by density-dependent and -independent processes. Here I study multiple populations of unexploited (i.e. assumed equilibria), lentic Brook Trout to describe variation in population structure. Studying unexploited populations gives fisheries managers and scientists an understanding of the intrinsic variation in systems absent from harvest pressure. First, I identify how much primary demographic attributes vary among populations, then attempt to attribute the observed variation to a suite of nested models containing terms for ecosystem productivity (total phosphorus concentration), climate (growing degree days or GDD), and density (fish biomass per hectare) in 9 lakes. I found that (1) populations varied substantially in somatic growth parameters (two-fold), natural mortality (three-fold), age-at-maturity (three-fold), length-at-maturity (two-fold) and recruitment (three-fold), (2) growth early in life was negatively correlated with density (r = -0.58), but maximum length was positively correlated with GDD (r = 0.61), and (3) spawning stock density was negatively correlated with recruitment (r = -0.57), but positively correlated with GDD (r = 0.55). I then experimentally harvested these previously unexploited populations with size selective gillnets by removing 30-70 % of the largest individuals from 5 of the 9 populations over 2 consecutive years. I tested the compensatory response to harvest with a BACI analysis, where I looked to see if absolute growth rate, size- and age-at-maturity, reproductive investment, and recruitment compensate for fisheries harvest. I found (1) strong evidence of recruitment compensation, (2) that overall (i.e. site-wide) stock-recruitment relationship was strongly density dependent and over-compensatory (i.e. a humped, Ricker type relationship), (3) positive but nonsignificant compensation in growth and age-at- iii maturity, and (4) no change in reproductive investment, but noted that populations may compensate for reproductive capacity in other ways (e.g. a combination of increased somatic growth and younger age-at-maturity). Comparing observed variation in unexploited populations’ demography with environmental variables helps fisheries managers and scientists understand intrinsic variability, and drivers of said variability; further exploiting these populations in an experimental fishery shows the initial mechanisms behind compensation. Examining both unexploited and responses to fisheries will help fisheries managers and scientists understand which populations can have their density reduced (via setting appropriate harvest rates), set realistic targets to recover populations, and increase understanding of the mechanisms that structure populations.
Full-text available
The Safe Operating Space (SOS) of a recreational fishery is the multidimensional region defined by levels of harvest, angler effort, habitat, predation and other factors in which the fishery is sustainable into the future. SOS boundaries exhibit trade-offs such that decreases in harvest can compensate to some degree for losses of habitat, increases in predation and increasing value of fishing time to anglers. Conversely, high levels of harvest can be sustained if habitat is intact, predation is low, and value of fishing effort is moderate. The SOS approach recognizes limits in several dimensions: at overly high levels of harvest, habitat loss, predation, or value of fishing effort, the stock falls to a low equilibrium biomass. Recreational fisheries managers can influence harvest and perhaps predation, but they must cope with trends that are beyond their control such as changes in climate, loss of aquatic habitat or social factors that affect the value of fishing effort for anglers. The SOS illustrates opportunities to manage harvest or predation to maintain quality fisheries in the presence of trends in climate, social preferences or other factors that are not manageable.
Full-text available
Stan is a probabilistic programming language for specifying statistical models. A Stan program imperatively defines a log probability function over parameters conditioned on specified data and constants. As of version 2.14.0, Stan provides full Bayesian inference for continuous-variable models through Markov chain Monte Carlo methods such as the No-U-Turn sampler, an adaptive form of Hamiltonian Monte Carlo sampling. Penalized maximum likelihood estimates are calculated using optimization methods such as the limited memory Broyden-Fletcher-Goldfarb-Shanno algorithm. Stan is also a platform for computing log densities and their gradients and Hessians, which can be used in alternative algorithms such as variational Bayes, expectation propagation, and marginal inference using approximate integration. To this end, Stan is set up so that the densities, gradients, and Hessians, along with intermediate quantities of the algorithm such as acceptance probabilities, are easily accessible. Stan can be called from the command line using the cmdstan package, through R using the rstan package, and through Python using the pystan package. All three interfaces support sampling and optimization-based inference with diagnostics and posterior analysis. rstan and pystan also provide access to log probabilities, gradients, Hessians, parameter transforms, and specialized plotting.
Bayesian statistics has exploded into biology and its sub-disciplines, such as ecology, over the past decade. The free software program WinBUGS and its open-source sister OpenBugs is currently the only flexible and general-purpose program available with which the average ecologist can conduct standard and non-standard Bayesian statistics. Bayesian Population Analysis Using WinBUGS goes right to the heart of the matter by providing ecologists with a comprehensive, yet concise, guide to applying WinBUGS to the types of models that they use most often: linear (LM), generalized linear (GLM), linear mixed (LMM) and generalized linear mixed models (GLMM). Comprehensive and richly-commented examples illustrate a wide range of models that are most relevant to the research of a modern population ecologist. All WinBUGS/OpenBUGS analyses are completely integrated in software R. Includes complete documentation of all R and WinBUGS code required to conduct analyses and shows all the necessary steps from having the data in a text file out of Excel to interpreting and processing the output from WinBUGS in R.
A guide to data collection, modeling and inference strategies for biological survey data using Bayesian and classical statistical methods. This book describes a general and flexible framework for modeling and inference in ecological systems based on hierarchical models, with a strict focus on the use of probability models and parametric inference. Hierarchical models represent a paradigm shift in the application of statistics to ecological inference problems because they combine explicit models of ecological system structure or dynamics with models of how ecological systems are observed. The principles of hierarchical modeling are developed and applied to problems in population, metapopulation, community, and metacommunity systems. The book provides the first synthetic treatment of many recent methodological advances in ecological modeling and unifies disparate methods and procedures. The authors apply principles of hierarchical modeling to ecological problems, including * occurrence or occupancy models for estimating species distribution * abundance models based on many sampling protocols, including distance sampling * capture-recapture models with individual effects * spatial capture-recapture models based on camera trapping and related methods * population and metapopulation dynamic models * models of biodiversity, community structure and dynamics * Wide variety of examples involving many taxa (birds, amphibians, mammals, insects, plants) * Development of classical, likelihood-based procedures for inference, as well as Bayesian methods of analysis * Detailed explanations describing the implementation of hierarchical models using freely available software such as R and WinBUGS * Computing support in technical appendices in an online companion web site.
Data from a large-scale mark–recapture study were used in an open population model to determine the cause for long-term trends in growth and abundance of a Rainbow Trout Oncorhynchus mykiss population in the tailwater of Glen Canyon Dam, AZ. Reduced growth affected multiple life stages and processes causing negative feedbacks that regulated the abundance of the population, including: higher mortality of larger fish; lower rates of recruitment (young of year) in years when growth was reduced; and lower rates of sexual maturation the following year. High and steady flows during spring and summer of 2011 resulted in a very large recruitment event. The population had declined 10-fold by 2016 due to a combination of lower recruitment and reduced survival of larger trout. Survival rates for trout ≥ 225 mm in 2014, 2015, and 2016 were 11%, 21%, and 22% lower than average survival rates between 2012 and 2013, respectively. Abundance at the end of the study would have been three- to five-fold higher had survival rates for larger trout remained at the elevated levels estimated for 2012 and 2013. Growth declined between 2012 and 2014 owing to reduced prey availability, which led to very poor fish condition by fall of 2014 (~0.9–0.95). Poor condition in turn resulted in low survival rates of larger fish during fall of 2014 and winter of 2015, which contributed to the population collapse. In Glen Canyon, large recruitment events driven by high flows can lead to increases in the population that cannot be sustained due to limitations in prey supply. In the absence of being able to regulate prey supply, flows which reduce the probability of large recruitment events can be used to avoid boom-and-bust population cycles. Our study demonstrates that mark–recapture is a very informative approach for understanding the dynamics of tailwater trout populations. Received 06 Dec 2016 accepted 31 Mar 2017 revised 20 Mar 2017
Applied Hierarchical Modeling in Ecology: Distribution, Abundance, Species Richness offers a new synthesis of the state-of-the-art of hierarchical models for plant and animal distribution, abundance, and community characteristics such as species richness using data collected in metapopulation designs. These types of data are extremely widespread in ecology and its applications in such areas as biodiversity monitoring and fisheries and wildlife management. This first volume explains static models/procedures in the context of hierarchical models that collectively represent a unified approach to ecological research, taking the reader from design, through data collection, and into analyses using a very powerful class of models. Applied Hierarchical Modeling in Ecology, Volume 1 serves as an indispensable manual for practicing field biologists, and as a graduate-level text for students in ecology, conservation biology, fisheries/wildlife management, and related fields. Provides a synthesis of important classes of models about distribution, abundance, and species richness while accommodating imperfect detection Presents models and methods for identifying unmarked individuals and species Written in a step-by-step approach accessible to non-statisticians and provides fully worked examples that serve as a template for readers' analyses Includes companion website containing data sets, code, solutions to exercises, and further information