-
https://cpw.cvlcollections.org/files/original/e38b4af92a28241f3266d5a99b50df6d.pdf
4fda0c2d681833f479e624d1286e69e0
PDF Text
Text
The research in this publication was partially or fully funded by Colorado Parks and Wildlife.
Dan Prenzlow, Director, Colorado Parks and Wildlife • Parks and Wildlife Commission: Marvin McDaniel, Chair • Carrie Besnette Hauser, Vice-Chair
Marie Haskett, Secretary • Taishya Adams • Betsy Blecha • Charles Garcia • Dallas May • Duke Phillips, IV • Luke B. Schafer • James Jay Tutchton • Eden Vardy
�North American Journal of Fisheries Management 39:819–848, 2019
© 2019 American Fisheries Society
ISSN: 0275-5947 print / 1548-8675 online
DOI: 10.1002/nafm.10320
FEATURED PAPER
Predicting Persistence of Rio Grande Cutthroat Trout Populations in an
Uncertain Future
Matthew P. Zeigler
New Mexico Department of Game and Fish, Fisheries Management Division, 1 Wildlife Way, Santa Fe,
New Mexico 87507, USA
Kevin B. Rogers*
Colorado Parks and Wildlife, Aquatic Research Section, Post Office Box 775777, Steamboat Springs, Colorado 80477
USA
James J. Roberts
U.S. Geological Survey, Colorado Water Science Center/Fort Collins Science Center, 2150 Centre Avenue, Building C,
Fort Collins, Colorado 80526, USA
Andrew S. Todd1
U.S. Geological Survey, Crustal Geophysics and Geochemistry Science Center, Box 25046, Mail Stop 964D,
Denver Federal Center, Denver, Colorado 80225, USA
Kurt D. Fausch
Department of Fish, Wildlife, and Conservation Biology, Colorado State University, Fort Collins, Colorado 80523, USA
Abstract
The Rio Grande Cutthroat Trout Oncorhynchus clarkii virginalis (RGCT) occupies just 12% of its ancestral
range. As the southernmost subspecies of Cutthroat Trout, we expect a warming climate to bring additional stressors
to RGCT populations, such as increased stream temperatures, reduced streamflows, and increased incidence of wildfire. We developed a Bayesian network (BN) model using site-specific data, empirical research, and expert knowledge
to estimate the probability of persistence for each of the 121 remaining RGCT conservation populations and to rank
the severity of the threats they face. These inputs quantified the genetic risks (e.g., inbreeding risk and hybridization
risk), population demographics (disease risk, habitat suitability, and survival), and probability of stochastic disturbances (stream drying risk and wildfire risk) in an uncertain future. We also created stream temperature and base
flow discharge models coupled with regionally downscaled climate projections to predict future abiotic conditions at
short-term (2040s) and long-term (2080s) time horizons. In the absence of active management, we predicted a decrease
in the average probability of population persistence from 0.53 (current) to 0.31 (2040s) and 0.26 (2080s). Only 11%
of these populations were predicted to have a greater than 75% chance of persisting to the 2080s. Threat of invasion
by nonnative trout had the strongest effect on population persistence. Of the 78 populations that are already invaded
or lacking complete barriers, 60% were estimated to be extirpated by 2080 and the remainder averaged only a 10%
chance of persistence. In contrast, the effects of increased stream temperatures were predicted to affect the future persistence of only 9% of the 121 RGCT populations remaining, as most have been restricted to high-elevation habitats
*Corresponding author: kevin.rogers@state.co.us
1
Present address: U.S. Environmental Protection Agency, Region 8, 1595 Wynkoop Street, Denver, Colorado 80202, USA.
Received April 21, 2018; accepted May 22, 2019
819
�820
ZEIGLER ET AL.
that are cold enough to buffer against some stream warming. Our BN model provides a framework for evaluating
threats and will be useful to guide management actions that are likely to provide the most benefit for long-term conservation.
Rio Grande Cutthroat Trout Oncorhynchus clarkii virginalis (RGCT) currently occupy only a small fraction of
their former range (Pritchard and Cowley 2006; Alves et
al. 2008). The introduction of nonnative salmonids and
anthropogenic habitat alteration are the primary drivers
restricting the distribution of RGCT (Pritchard and Cowley 2006; Penaluna et al. 2016). The subspecies currently
remains in only 12% of the 10,700 km of streams they historically occupied in the Rio Grande, Canadian River,
and Pecos River basins of New Mexico and Colorado
(Alves et al. 2008). Most remaining populations are isolated in high-elevation headwater streams, resulting in
decreased genetic diversity within populations and
increased genetic differentiation among them (Pritchard et
al. 2007a, 2009). Continued threats to the subspecies have
resulted in multiple petitions to list RGCT for protection
under the Endangered Species Act, with recent findings
focused on the consequences of a warming climate
(USFWS 2008, 2014).
Models that describe future conditions and species resilience are an integral part of species status assessments
performed by the U.S. Fish and Wildlife Service (USFWS;
USFWS 2016; Smith et al. 2018) and should be key elements to any long-term viability assessment for RGCT.
The population health index currently used for assessing
RGCT population viability (Hirsch et al. 2006; Alves et al.
2008) can characterize current conditions but is not capable of predicting future conditions that are expected to
change drastically in a warming climate (IPCC 2013).
Others have used models to predict the effect of rising
water temperatures on native trout (Williams et al. 2009;
Peterson et al. 2013; Roberts et al. 2013), but those studies
focused on more northerly trout taxa rather than RGCT,
for which the effects of a changing climate are expected to
be particularly intense (Brown et al. 2004; Cook et al.
2004; Dettinger et al. 2015).
To predict the likely consequences of a warming climate,
it is useful to organize threats facing RGCT into three
broad categories: (1) genetic risks, (2) population demographics, and (3) stochastic disturbance risks. Genetic risks
include consequences of small population size and introgressive hybridization. Population demographics result
from the many biotic and abiotic interactions that can
drive species persistence. Stochastic disturbance risk
includes abiotic events that may happen infrequently but
are also capable of driving local RGCT populations to
extinction.
The fragmented nature of the remaining isolated
RGCT populations (Alves et al. 2008) means that genetic
risks are an important consideration. Many RGCT populations are small (occupying less than 5 km of stream habitat; Alves et al. 2008) and likely vulnerable to inbreeding
depression (Allendorf and Luikart 2007; Whiteley et al.
2013; Robinson et al. 2017). Outbreeding depression is
also a concern, as Rainbow Trout O. mykiss (a sister
taxon) readily hybridize with Cutthroat Trout O. clarkii,
reducing their fitness (Muhlfeld et al. 2009a) and leaving
populations susceptible to extinction through hybridization
(Rhymer and Simberloff 1996; Muhlfeld et al. 2017).
Other taxa of commonly stocked nonnative Cutthroat
Trout (e.g., Yellowstone Cutthroat Trout O. c. bouvieri
and Colorado River Cutthroat Trout O. c. pleuriticus) also
hybridize readily with RGCT, infusing nonnative alleles
into otherwise pure populations (Pritchard et al. 2007b;
Bestgen et al. 2019). At present, management of RGCT
genetic risk is focused on maintaining pure stocks that are
free or nearly free of nonnative alleles (Allendorf et al.
2001), in part because the USFWS considered only RGCT
populations with less than 10% admixture as contributing
to the viability of the subspecies in recent listing decisions
(USFWS 2014).
Species persistence is mediated by both biotic and abiotic interactions (Araújo and Luoto 2007; Van der Putten
et al. 2010; Wenger et al. 2011) that drive population
demographics in RGCT as well. Invasions of nonnative
salmonids into native trout habitat continue to be among
the greatest threats to native salmonid diversity in North
America (Behnke 1992; Dunham et al. 2002; Penaluna et
al. 2016). Invasions can have strong and consistent direct
effects through biotic interactions (Peterson et al. 2004;
McHugh and Budy 2005; Roberts et al. 2017) but also
can be mediated by abiotic factors, such as stream temperatures (Rahel and Nibbelink 1999; de la Hoz Franco
and Budy 2005), which are expected to rise in the Rio
Grande basin in the future (Isaak et al. 2012; Zeigler et al.
2012). Flow regimes also affect biotic interactions (Fausch
et al. 2001; Williams et al. 2009; Wenger et al. 2011), and
they too are likely to change (Luce and Holden 2009;
Williams et al. 2009; Wenger et al. 2011). Given suitable
abiotic conditions, invasion of nonnative trout can also
facilitate the invasion of other deleterious organisms, such
as the parasite Myxobolus cerebralis, the causative agent
of whirling disease (Nehring and Walker 1996; Schisler
and Bergersen 2002), which is capable of reducing RGCT
�PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
population viability (Thompson et al. 1999; Ayre et al.
2014).
Finally, the magnitude and frequency of stochastic disturbance events in the southern Rocky Mountains are
expected to increase in the future. Droughts are predicted
to become more severe (Seager et al. 2007, 2013), and
wildfires are predicted to be larger and more frequent
(McKenzie et al. 2004; Westerling et al. 2006; Westerling
2016), as are associated debris flows (Gresswell 1999; Burton 2005). All will have negative consequences for Cutthroat Trout populations (Dunham et al. 2003, 2007;
Williams et al. 2009; Wenger et al. 2011) and should be
considered when evaluating whether a population will persist as the climate changes.
Predicting the persistence of individual RGCT populations is difficult because genetic risks, population demographics, and stochastic disturbance risks all combine in
complex ways to affect populations and are in turn affected
by a changing climate. Bayesian network (BN) models are
ideally suited for addressing this task because they are able
to integrate disparate sources of information with varying
degrees of empirical support to inform management decisions (Marcot 2017). In conservation biology, these models
can be used to predict the likelihood that individual populations will persist into the future by integrating the effects
of biotic and abiotic factors (Mace and Lande 1991;
Morita and Yamamoto 2002; Morris et al. 2002).
Our objectives for this research were to (1) estimate the
probability that each RGCT conservation population will
persist over short-term (to the 2040s) and long-term (to
the 2080s) time horizons by using a BN model and
(2) assess the relative influence of diverse threats to the
persistence of the subspecies. We consider this BN to be a
second-generation model, building on simpler first-generation models, such as those developed by Peterson et al.
(2008a) and Roberts et al. (2013, 2017) for other Cutthroat
Trout subspecies. Although the BN will require further
validation with field data, it provides a working model for
prioritizing restoration activities for RGCT populations
and will be useful as a template for application to other
native trout in western North America.
METHODS
Study Area and Rio Grande Cutthroat Trout Distribution
Data
Our study area included the entire current distribution
of RGCT in the Rio Grande, Pecos River, and Canadian
River basins of Colorado and New Mexico (Figure 1).
The subspecies’ historical range is divided into five
geographic management units (GMUs) organized by
major drainage basins, four of which currently support
populations of RGCT (Lower Rio Grande, Rio Grande
821
Headwaters, Pecos, and Canadian GMUs). The fifth
(Caballo GMU) harbors a single admixed population of
RGCT in Las Animas Creek (Alves et al. 2008) and was
not considered further in the present study. We limited
our analysis to RGCT “conservation populations” (sensu
UDWR 2000; Shepard et al. 2005; Alves et al. 2008),
which are those considered genetically unaltered or only
mildly admixed (<10%) with nonnative salmonids.
Although additional RGCT populations with over 10%
admixture exist across the species’ range, they have
reduced conservation value and were not recognized by
the USFWS in previous listing decisions (USFWS 2014).
Using data on the subspecies’ range provided from the most
current rangewide status report (Alves et al. 2008) supplemented with updated information provided by Colorado and
New Mexico state agencies, we identified 121 discrete RGCT
conservation populations. Barrier locations and population
extents (i.e., distance between the farthest upstream and
downstream records) were mapped in a GIS using stream
data layers from the National Hydrography Dataset Plus
version 2 (NHDPlus v2; 1:100,000 scale; www.horizon-syste
ms.com/nhdplus/NHDPlusV2_home.php).
Rio Grande Cutthroat Trout Population Estimates
Field data collected over the last decade from RGCT
conservation streams (197 surveys using backpack electrofishers) were obtained from the New Mexico Department of Game and Fish and from Colorado Parks and
Wildlife. Two-pass removal data were analyzed using
JOM (JakeOmatic) version 2.4 (Rogers 2006), whereas
three- and four-pass removal data (61 surveys) were analyzed using the Huggins (1989) estimator, with fish TL as
a covariate, as implemented in Program MARK (White
and Burnham 1999). To be consistent with earlier modeling efforts by Young et al. (2005), fish smaller than 75 mm
were excluded from these site-specific estimates to eliminate age-0 fry that had not yet recruited to the population
and that have low capture probabilities compared to larger fish (e.g., Saunders et al. 2011). Site-specific population
estimates (fish/km) were then extrapolated to the entire
occupied reach to estimate the population size of 75-mm
and larger fish. For populations occupying less than 10
km of stream, one survey was deemed sufficient to estimate population size, but multiple estimates were averaged when available. For streams with a 10-km or greater
occupied length, we only estimated population size if at
least two multiple-pass population surveys had been conducted (Young and Guenther-Gloss 2004). When adequate
survey data were not yet available, we used simple linear
regression to correlate occupied stream length (km) with
population size (sensu Young et al. 2005). Since Cutthroat
Trout population densities are reduced when sympatric
with nonnative salmonids (Peterson et al. 2004; Benjamin
and Baxter 2012; Al-Chokhachy and Sepulveda 2019), we
�822
ZEIGLER ET AL.
FIGURE 1. Map of the current (black lines) and historical (gray lines) distributions of Rio Grande Cutthroat Trout in Colorado and New Mexico.
Inset shows the location of the subspecies’ range in Colorado and New Mexico, including the five geographic management units (GMUs): Rio Grande
Headwaters (RGH), Lower Rio Grande (LRG), Canadian (CAN), Pecos (PEC), and Caballo (CAB). The inset map also shows the general locations
(black dots) of one extant population in a 2.5-km segment of Pinelodge Creek that persists in the southern portion of the Pecos GMU (six other
historical populations nearby were extirpated) and one historical population that was extirpated from the Caballo GMU.
developed separate regressions to describe allopatric and
sympatric RGCT populations (Figure S.1 available in the
Supplement in the online version of this article).
Predicting Future Environmental Conditions
Stream temperature modeling.— Spatially referenced
stream temperature records were obtained from a large
monitoring program focused on RGCT (Zeigler et al.
2013b); additionally, stream temperature data from university researchers (Harig and Fausch 2002; A. Harig and K.
Fausch, Colorado State University, unpublished data) and
state and federal agencies (New Mexico Environment
Department, Colorado Parks and Wildlife, and U.S. Forest
Service [Carson and Santa Fe National Forests]) were
obtained to maximize the temporal and spatial extent of
the data. We also included temperature records from
nearby high-elevation streams just outside of the Rio
Grande basin because there were few temperature records
from high-elevation streams within the basin (Roberts et al.
2013). The resulting database comprised 544 unique stream
temperature records spanning the years 2000–2013. The
annual mean temperature for the warmest 30 d (maximum
30-d average temperature [M30AT]) and the week with the
warmest peak temperatures (maximum weekly maximum
temperature [MWMT]) were calculated for each stream
temperature record.
We selected five covariates based on previous stream
temperature models (Isaak et al. 2010; Roberts et al. 2013)
�PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
to predict stream temperatures across the current range of
RGCT. Two covariates were based on air temperature
(air_M30AT and air_MWMT), and three were based on
geomorphic attributes (elevation, drainage area, and
aspect; Appendix 1). Using these covariates, we selected
the most parsimonious model to predict M30AT and
MWMT from three statistical approaches: multiple linear
regression, universal kriging (Stein and Corsten 1991), and
a spatial flow-routed model (spatial stream network
[SSN]; Peterson and Ver Hoef 2010; Isaak et al. 2014). We
assessed which statistical approach to use by measuring fit
with the leave-one-out cross-validation technique to calculate the average root mean square predictive error
(RMSPE; Power 1993). We then used the method with
the best fit (i.e., lowest RMSPE) to predict stream temperatures in each of the 838 individual stream segments
inhabited by the 121 RGCT populations. Stream temperatures in future time periods were predicted from a dynamically downscaled regional climate model (Hostetler et al.
2011). Details of the stream temperature modeling are presented in Appendix 1.
Base flow modeling.— Little information on current base
flow conditions in streams occupied by RGCT is available
because most stream gauges in the subspecies’ current
range are in low-elevation stream segments where RGCT
do not occur. To model summer base flow for each occupied stream segment, we assembled continuous streamflow
data from 29 stream gauges in New Mexico and Colorado
where flows were thought to be relatively free from
anthropogenic influences. We used these data to develop a
linear regression to predict the mean 30-d minimum discharge (M30MD; i.e., base flow conditions) from drainage
area for each stream segment (R2 = 0.57). This model was
then used to predict M30MD for all 838 individual stream
segments occupied by the 121 RGCT populations. The
final base flow prediction for each population was then
determined by averaging across all of the individual
stream segments that comprised each population. Details
of base flow modeling are presented in Appendix 1.
Bayesian network development.— A Bayesian network is
a graphical model that represents probabilistic relationships among variables (i.e., nodes) by incorporating existing knowledge and uncertainty. Each node within a BN is
comprised of discrete states, each resulting from unique
combinations of conditions from contributing nodes. The
probability of each state in a node is quantified in conditional probability tables (CPTs) for each node and for all
possible combinations of contributing nodes. Once fully
parameterized, a BN can determine the probability of
each state in the terminal node (e.g., probability of population persistence) using Bayes’ theorem and the conditions of all contributing nodes (Marcot et al. 2006).
Therefore, BNs provide a robust and transparent way to
synthesize published research with site-specific field data
823
and can incorporate expert knowledge where quantitative
research results are not available to predict probabilities
of outcomes in complex systems (Lee and Rieman 1997;
Marcot et al. 2006; Conroy and Peterson 2013).
We developed the BN for predicting the probability of
persistence of RGCT populations based on guidelines
from Cain (2001), Marcot et al. (2006), and Marcot (2012,
2017). The development proceeded in eight major steps to
create alpha- and beta-level models proposed by Marcot
et al. (2006), and the final BN was used to make initial
predictions of population persistence. In all, the five
authors of this paper met during six web conferences and
two in-person workshops to create, discuss, and refine the
model and communicated closely throughout the process
to update the model at each step. These steps focus on
developing a robust BN, but this model will need further
validation based on empirical data to ensure its reliability.
Step 1: literature review and development of the initial
directed acyclic graph.— We reviewed the available information on RGCT and other related Cutthroat Trout subspecies (e.g., Peterson et al. 2008b, 2013; Roberts et al.
2013) to identify important factors that affect population
persistence. We used this to develop an initial directed acyclic graph (DAG) that depicted important causal links
among variables (nodes) that influence RGCT persistence.
Parent nodes in the network influence child nodes downstream and ultimately the terminal node: the probability of
population persistence. This DAG was presented to the
RGCT Conservation Team (a consortium of state, tribal,
and federal agencies charged with managing the subspecies),
and their peer review comments were used to modify the
network structure—a step suggested by Marcot (2017).
Step 2: alpha-level Bayesian network model.— We developed an alpha-level BN model based on a combination of
field empirical data, models fit to data, peer-reviewed literature, unpublished reports, and expert knowledge (Appendix 2). Bayesian network models that are based in part on
expert knowledge are often more robust than those based
solely on case studies in peer-reviewed literature, which
usually come from a limited region or sample and can
lead to overfitting and biased models (Marcot et al. 2006;
Marcot 2017). The five authors, all of whom have conducted research on native Cutthroat Trout in the southern
Rocky Mountains and who represent two state agencies,
one federal agency, and one academic institution, followed
a formal process of direct quantitative elicitation using the
probability method to develop the BN (Conroy and Peterson 2013). For each node in the DAG, states were defined
to represent the variation relevant to predicting outputs,
and narratives were written to document the key mechanisms involved and the information used (see Supplement
available in the online version of this article).
The final DAG (Figure 2) and BN model highlight the
three components that are thought to drive the probability
�824
ZEIGLER ET AL.
of persistence for RGCT populations: genetic risks (Figure 3), population demographics (Figure 4), and stochastic
disturbance risks (Figure 5). The complete DAG and BN
consist of 37 nodes (20 parent nodes, 16 child nodes, and
a single output node) with 48 links (see Supplement for
detailed node definitions and conditional probabilities).
The title for each node is shown in bold below.
Genetic risks.— Given the isolated nature of the remaining RGCT populations, we consider the potential effects
of inbreeding (Inbreeding Risk) on persistence as influenced by time (Time Period) and the effective population
size (Ne; Figure 3). Time periods spanned 2010–2019 (current time period), 2040–2049 (2040s), and 2080–2089
(2080s). Effective population sizes (Effective Population
Size) were determined as a product of the estimated adult
population size (N; Adult Population Estimate) and the
ratio of Ne to N (Ne/N Ratio). Little information is available on Ne/N ratios for inland Cutthroat Trout, so we
used the approximate midpoint (0.25) of published Ne/N
ratios for other salmonids (Rieman and Allendorf 2001;
Palm et al. 2003; Jensen et al. 2005). Nevertheless, the BN
allows the input of any Ne/N ratio from 0 to 1 when more
informed ratios calculated with molecular methods (Whiteley et al. 2010, 2012, 2013) become available.
Introgressive hybridization with nonnative Rainbow
Trout and other nonnative subspecies of Cutthroat Trout
(Invasion and Hybridization Risk) is a major driver of
FIGURE 2. Top-level directed acyclic graph used to predict future
persistence of Rio Grande Cutthroat Trout populations. Boxes with thin
outlines represent the three directed acyclic graph components expected
to have a direct influence on population persistence (thick outline), the
final output of the Bayesian network. See Figures 3, 4, and 5 for the
nodes and links associated with the three components: Genetic Risks,
Population Demographics, and Stochastic Disturbance Risks.
RGCT decline and will continue to be an important threat
in the future (Pritchard and Cowley 2006; Alves et al.
2008; Muhlfeld et al. 2009a, 2009b, 2014). We reasoned
that the potential risk of invasion by nonnative Rainbow
Trout and Cutthroat Trout and subsequent hybridization
is influenced by the proximity of a hybridizing source population (Proximity of Hybridizing Source Population) and
the presence of a barrier (Barrier Presence) and that the
risk and prevalence of hybridization increase through time
(Time Period). Presence of a complete barrier or the
absence of a source population in the watershed does not
entirely eliminate the risk of hybridization because there
remains a small chance that nonnative trout could be illegally introduced (Rahel 2004; Fausch 2007).
Population demographics.— The potential for RGCT
population growth was assessed by determining how abiotic
conditions affect the survival of RGCT throughout their life
cycle (Figure 4). Our approach to determining suitable habitat (Potential Habitat Suitability) for each conservation
population was based on Occupied Stream Length, stream
size (i.e., Stream Wetted Width as mediated by Base Flow
Discharge), and M30AT, which influences reproduction and
growth (Roberts et al. 2013). We also reasoned that anthropogenic activities (Anthropogenic Influence), such as grazing, timber harvest, mining, and recreation, could reduce
habitat suitability (Realized Habitat Suitability) in an occupied stream fragment. Although anthropogenic activities
are reported in the rangewide status assessment (Alves et al.
2008), we also solicited input from management biologists
to determine whether habitat degradation resulting from
the listed activities was likely to have a benign or substantial
effect on each RGCT population (see Supplement).
Population growth rate is determined by the combined
inputs of Realized Fry-to-Age-2 Survival and Adult Demographics as well as Demographic Support (Figure 4). Different mechanisms drive survival in small versus large
RGCT, so we modeled them separately. For small fish,
Potential Fry-to-Age-2 Survival is mediated by Realized
Habitat Suitability (from above) and maximum temperature (MWMT) and is then affected by interactions with
nonnative trout competitors, such as Brook Trout Salvelinus fontinalis (Peterson and Fausch 2003; Peterson et al.
2004; Fausch et al. 2009) and Brown Trout Salmo trutta
(McHugh and Budy 2005, 2006; Al-Chokhachy and
Sepulveda 2019; Competitor Invasion and Biotic Influence
Risk). These biotic interactions with nonnative Brook
Trout and Brown Trout reduce the survival of fry to age2 fish (Realized Fry-to-Age-2 Survival; Wang and White
1994; Peterson et al. 2004; McHugh and Budy 2005) but
depend on the Proximity of Competitor Source Populations
and Barrier Presence. Furthermore, whirling disease
caused by M. cerebralis (Mc Infection Risk) affects only
fry-to-age-2 RGCT survival but depends on the proximity
of the source population carrying the parasite (Proximity
�PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
825
FIGURE 3. Directed acyclic graph of the Genetic Risks component used to predict future persistence of Rio Grande Cutthroat Trout populations.
Shaded boxes represent parent nodes, and unshaded boxes are child nodes. The parent node with a dashed outline (Barrier Presence) can be influenced
by management actions. Thin arrows are links between nodes, and both thick arrows are links directly to the output node, Population Persistence.
The ratio of effective population size (Ne) to total population size (N) is given by Ne/N.
of Mc Source) and can be mitigated by warmer water temperatures (El-Matbouli et al. 1999) and barriers to migration. In contrast to juvenile fish, Realized Habitat
Suitability and maximum temperature (MWMT) are the
only major abiotic factors that influence survival of adult
RGCT (Adult Demographics). Among biotic factors, whirling disease does not influence adult RGCT survival (Nehring and Walker 1996; Vincent 1996), but nonnative trout
have the potential to completely extirpate populations by
reducing the number of individuals that reach reproductive
age (Invasion Vortex) unless they are removed (Nonnative
Control).
Population growth is also expected to be influenced by
potential immigration of RGCT from nearby populations
(Demographic Support). However, most RGCT populations occur in isolated stream fragments and cannot
receive immigrants, so only populations that are hydrologically connected to occupied lakes—or populations that
are downstream of a separate population isolated by a
barrier—could potentially receive demographic support
(Pritchard and Cowley 2006; Pritchard et al. 2007a; Alves
et al. 2008). We assumed no influence of density dependence by conspecifics on population growth rates because
nearly all populations are at low densities (Figure 6), and
most are strongly influenced by abiotic and biotic factors
associated with nonnative trout and parasites (Kovach et
al. 2017; Roberts et al. 2017).
Stochastic disturbance risks.— The risk of extirpation
during disturbance events (Stochastic Disturbance Risks)
will increase through time (Time Period; Rieman and
Isaak 2010). We include two types of stochastic disturbances here: (1) the potential risk of wildfire and subsequent debris flows that, when combined, often eradicate
small populations (Propst et al. 1992; Rinne 1996; Gresswell 1999); and (2) stream drying (Figure 5). These risks
are assumed to be mediated through potential buffering
(Stochastic Buffer) by the amount of habitat occupied
(Occupied Stream Length) and connectivity among subpopulations (i.e., number of occupied tributaries; Population Connectivity). Longer occupied reaches with more
connected and occupied tributaries are reasoned to be buffered against these catastrophic disturbances, whereas
shorter reaches that are not connected to other occupied
tributaries are at greater risk (Peterson et al. 2008b;
Roberts et al. 2013). The combined threats of wildfire and
debris flow (Wildfire/Debris Flow Risk) for each population were estimated from a rangewide assessment of wildfire risk using FlamMap (Miller and Bassett 2013), and a
model of debris flow risk developed by Cannon et al.
(2010). Stream Drying Risk was assessed using a suite of
nodes including Evidence of Intermittency and Base Flow
Discharge. We also assumed that refugia provided by beaver ponds and large perennial pools in the occupied reach
would buffer against stream drying risk (Stream Drying
Refugia Availability).
Step 3: conditional probability tables.— We used a modified Delphi method (Clark et al. 2006; Conroy and Peterson 2013) to define the conditional probabilities for each
child node, representing all combinations of the states of
the parent nodes (i.e., CPT). To estimate conditional
probabilities for each child node, each author assigned all
probabilities independently from the other authors. These
were summarized, and the differences were discussed during a review session to clarify the model components and
their relations and to avoid any linguistic uncertainty
(Conroy and Peterson 2013). For several nodes, more
information was needed to resolve uncertainties, so we
consulted other literature and subject matter experts (see
Appendix 2 and Supplement). In particular, for the Mc
Infection Risk node we consulted three research scientists
�826
ZEIGLER ET AL.
FIGURE 4. Directed acyclic graph of the Population Demographics component used to predict future persistence of Rio Grande Cutthroat Trout
populations (Source Pop = Source Population). Shaded boxes represent parent nodes, and unshaded boxes are child nodes. The four parent nodes
with dashed outlines can be influenced by management actions, whereas the three in ovals are influenced by climate. Thin arrows are links between
nodes, and the thick arrow is a link directly to the output node, Population Persistence. Time Period represents current, 2040s, or 2080s and influences
the Myxobolus cerebralis (Mc) Infection Risk, as does the maximum 30-d average temperature (M30AT). Adult Demographics and Fry-to-Age-2
Survival are influenced by maximum weekly maximum temperature (MWMT).
to clarify the risk of whirling disease invasion to the persistence of RGCT populations across time periods,
depending on the proximity of the source, the presence of
a barrier, and water temperature. After evaluating this
new information, we independently revised our initial
CPT values. The final probability values were averaged
across experts to ensure that the uncertainty in the expert
knowledge was incorporated in the BN (Marcot et al.
2006; Conroy and Peterson 2013). The alpha-level model
was then compiled and evaluated with key combinations
of input values for parent nodes, adjusting the CPT in a
few cases to ensure that the model behavior matched
expectations, as directed by Marcot et al. (2006).
Steps 4 and 5: formal independent review.— To develop
the beta-level model used for prediction, our fourth step
was to subject the alpha-level model to formal independent peer review (see Marcot 2017). We sought independent reviews from three experts in BNs and salmonid
ecology and from two conservation geneticists. Two of
these five additional reviewers represented government
agencies, and three came from academic institutions. All
reviewers were asked to review whether (1) the overall
structure of the model (the DAG) met the intended goal
of predicting persistence for RGCT populations, (2) all
important habitat variables that influence RGCT survival
were included, and (3) the effects of nonnative trout—and
of catastrophic disturbances and their interactions with
habitat patch sizes—were accurately portrayed. The three
reviewers with expertise on BNs and salmonid ecology
provided wide-ranging reviews of the model, the node
�PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
827
FIGURE 5. Directed acyclic graph of the Stochastic Disturbance Risks component used to predict future persistence of Rio Grande Cutthroat Trout
populations. Shaded boxes represent parent nodes, and unshaded boxes are child nodes. The three parent nodes with dashed outlines can be influenced
by management actions, whereas the one in an oval (Base Flow Discharge) is influenced by climate. Thin arrows are links between nodes, and the
thick arrow is a link directly to the output node, Population Persistence.
descriptions, and the CPTs. The two conservation geneticists focused primarily on the nodes and CPTs in the
Genetic Risks component (Figure 3). After receiving the
written reviews, the fifth step consisted of all five authors
carefully evaluating the comments, after which we modified the model to address the shortcomings identified by
the reviewers.
Step 6: sensitivity analysis.— We performed a sensitivity
analysis to show the full range of influence of each input
node on the model output based on values from our data
set. We then used a best-case versus worst-case scenario
approach (Marcot 2012; Conroy and Peterson 2013) to
rank the relative influence of each input node on the probability of RGCT persistence. To do this, we used the best
and worst values for each input node while keeping all
other input nodes at default values (the observed mean for
each continuous input node and the middle value for each
categorical input node; Table 1) to estimate two cases of
persistence for each input node. We did not assess sensitivity to the Ne/N ratio because this value was fixed; however, in the future, molecular data can be used to estimate
this ratio directly.
Step 7: predictions of Rio Grande Cutthroat Trout population
persistence.— The seventh step involved compiling the final
beta-level model, predicting the probability of persistence
for each population to 2040 and 2080, and analyzing these
results in relation to management. The predicted future
environmental conditions (i.e., stream temperature and base
flow predictions) for each NHDPlus v2 stream segment (n
= 838) were estimated and then averaged over the occupied
stream length for each population (n = 121) to determine
the final values for each population during each time period.
Segments with MWMT exceeding 25.0°C (too warm to support RGCT; Zeigler et al. 2013a) were not included as
viable habitat in subsequent time periods. These stream segments were also subtracted from the total occupied stream
length (km), resulting in a loss of stream length for some
populations in future time periods. Predicted stream temperatures and base flows, along with the adjusted stream
lengths, were then used as inputs into the BN model to predict the probability of persistence for each RGCT population during each time period. Populations that were
predicted to have a persistence probability of zero during
one time period were assumed to be extirpated and were
excluded from consideration in future periods. All BN
development and analysis were performed using Netica version 4.16 (Norsys Software Corp., Vancouver, British
Columbia). A tutorial for downloading Netica and running
the model is available from http://cpw.state.co.us/cutthroattrout.
We examined regional trends in RGCT persistence by
aggregating the probabilities of persistence within each
GMU. We assumed that extinction of a population by
2080 would occur independently from other populations
in the same GMU, allowing us to multiply their respective
extinction probabilities. The difference of that product
from 1.0 represents the probability of at least one population persisting in each GMU by 2080 and, more importantly, whether RGCT would be extirpated from a given
GMU during that time frame.
Step 8: review by management biologists.— In the final
step, we presented our results on probability of persistence to
the RGCT Conservation Team at their annual meeting. Biologists were queried to assess whether predicted probabilities
�828
ZEIGLER ET AL.
FIGURE 6. The frequency of Rio Grande Cutthroat Trout conservation populations (n = 121) by occupied stream length (left panel) and population
density of 75-mm and larger fish per kilometer (right panel) used to inform the Bayesian network model. Populations in the right panel are
distinguished based on whether they are allopatric versus sympatric with nonnative salmonids.
of persistence were consistent with their knowledge of the
populations they managed. Managers from both Colorado
and New Mexico could not identify any substantive shortcomings. As such, no further modifications to the BN model
were made.
RESULTS
We identified 121 RGCT conservation populations distributed across four of the five historically occupied GMUs
(Figure 1), with 82% found in the two Rio Grande GMUs
(99 populations), only 10% found in the Pecos River drainage (12 populations), and 8% found in the Canadian River
drainage (10 populations). Median length of occupied
stream reaches was 6.3 km and ranged from 0.6 to 69.3 km
(Figure 6). Empirical population abundance data were
available for 50 of the 121 populations, whereas population
size was estimated from a model for the remaining 71 populations. Fifty-eight populations (48%) were sympatric with
nonnative Brook Trout or Brown Trout. Population densities of RGCT (≥75 mm) in allopatric stream reaches ranged
from 53 to 1,119 fish/km (Figure 6), with a median value of
�PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
TABLE 1. Default values for each node used in the Bayesian network
sensitivity analysis, based on data in the Rio Grande Cutthroat Trout
(RGCT) Rangewide Database, and observed best- and worst-case scenarios for each node.
Node
Stream Drying
Refugia Availability
Evidence of
Intermittency
Population
Connectivity
MWMT (°C)a
Demographic
Support
Proximity of
Competitor Source
population
Base Flow Discharge
(m3/s)a
Anthropogenic
Influencea
Nonnative Control
Wildfire/Debris Flow
Risk
Adult Population
Estimateb
Occupied Stream
Length (km)c
Proximity of
Myxobolus cerebralis
Source
M30AT (°C)a
Time Period
Barrier
Proximity of
Hybridizing Source
Population
Default state
Best/worst case
observed
scenarios
Uniform
probabilities
Uniform
probabilities
Uniform
probabilities
16.3
None
Present/none
15/25
Consistent/none
Near
Far/invaded
0.0515
0.18/0.0016
Significant:
0.12; minimal:
0.88
None
Uniform
probabilities
1,964
Minimal/
significant
None/yes
Strong/isolated
Annual/none
Low/high
100,000/0
6.3
69.3/0.6
Absent
Absent/infected
15.23
Uniform
probabilities
None
Far
15.0/7.5 and 21.0
Current/2080s
Complete/none
Absent/invaded
a
MWMT (maximum weekly maximum temperature) and M30AT (maximum
30-d average temperature) were calculated as the averages of all possible values
from the RGCT Rangewide Database.
b
Calculated using the Young et al. (2005) model, with stream length set to the
median occupied stream length of all populations in the RGCT Rangewide Database.
c
Calculated as the median of all possible values from the RGCT Rangewide
Database.
312 fish/km and a 90th percentile of 377 fish/km. In sympatric stream reaches, population densities ranged from 8 to
476 fish/km, with a median of 158 fish/km and a 90th percentile of 221 fish/km.
829
Predicting Future Environmental Conditions
Model selection.— The most parsimonious models for predicting both M30AT and MWMT included air temperature,
stream segment elevation, cumulative drainage area, and
aspect (Table A.1.1). The SSN statistical technique was
selected as the best for modeling both M30AT and MWMT
because it displayed the lowest prediction errors (Appendix 1).
Final models for both temperature metrics tended to overestimate lower temperatures and underestimate higher temperatures, but this effect was slight for M30AT (Figure A.1.1).
Future thermal conditions.— Small increases in M30AT
and MWMT were predicted to occur in RGCT-occupied
streams by 2080. Across GMUs, M30AT was predicted to
increase by only 0.66–0.72°C, and MWMT was predicted
to increase by only 0.45–0.47°C (Table A.1.2). Nearly all of
the RGCT populations (n = 118) were predicted to occupy
streams having segments with suitable temperatures for
recruitment and growth (M30AT = 9.1–19.0°C) in 2080
(Figure A.1.2). No populations were predicted to occupy
habitats estimated to be completely thermally unsuitable
(MWMT ≥ 25.0°C) for RGCT by 2080, although three
populations were predicted to lose some currently occupied
stream segments due to increased MWMT. The resulting
loss of thermally suitable habitat amounted to only 2.2 km
of the 1,145 km currently occupied. Nevertheless, 11
RGCT populations (9%) occupied streams that were predicted to become warm enough to reduce survival
(MWMT > 21°C) but not warm enough for the entire occupied stream reach to become thermally unsuitable.
Future flow conditions.— With most stream gauges
located downstream of occupied habitats throughout the
current range of RGCT, flow conditions within occupied
segments were likely overestimated in comparison to
actual base flow discharges (see Zeigler et al. 2013b for
measured conditions). Comparison of summer base flow
predictions from our model to actual conditions was
challenging since only point estimates from these small,
high-elevation streams were available. Nevertheless, we
predicted decreases in summer base flow discharges from
the current time period to the 2080s (Figure A.1.3). The
greatest shift was observed in the number of populations
in streams with moderate summer base flow discharge
(0.0291–0.1799 m3/s) that decreased to low discharge
(0.0017–0.0291 m3/s). In the current time period, 95 populations (78%) were in the moderate-discharge category,
but this number decreased to 77 populations (64%) by the
2080s. This decrease in discharge led to an increase from
24 populations (20%) in the low-discharge category for the
current time period to 43 populations (36%) in that category by the 2080s.
Bayesian Network Model
Predicted persistence of Rio Grande Cutthroat Trout
populations.— Persistence of individual RGCT populations
�830
ZEIGLER ET AL.
was highly variable (Figure 7), with probability values
ranging from 0.06 to 0.93 in the current time period and
from 0.00 to 0.90 predicted for the 2080s. Mean probability of persistence to 2080 was highest in the Rio Grande
Headwaters GMU (0.30) and lowest in the Pecos GMU
(0.20) and Canadian GMU (0.21). Median persistence
probabilities of RGCT populations were generally low
across all GMUs during the current time period (0.31–
0.61) and decreased in both the 2040s and 2080s to final
probability values ranging from only 0.00 to 0.13 (Figure
7). About half the populations were predicted to have low
probabilities of persistence because nonnative competitors
are already present (i.e., invaded) in the current time period (58 of 121 populations; Figure 8). Another 17% of
populations not currently invaded have nonnatives in close
proximity, with no barrier preventing their invasion (16 of
121 populations) or only a partial barrier delaying invasion (4 of 121 populations). Most populations that were at
risk of invasion owing to these conditions were predicted
to be extirpated by 2080, whereas most populations that
are not currently invaded and for which invasion is unlikely (i.e., due to complete barriers) were predicted to have
a probability of persistence greater than 0.50. Aggregating
the predicted probabilities of population persistence across
each GMU revealed that total extinction was unlikely by
2080 in each of the four occupied GMUs, with only a 3%
chance in the Canadian GMU, 1% in the Pecos GMU,
and less than 1% in the remaining two Rio Grande
GMUs.
Model sensitivity and behavior.— Based on the sensitivity analysis, the three most influential parent nodes driving
the probability of RGCT population persistence were
related to threats posed by nonnative trout competitors
(Figure 9). The best-case and worst-case scenarios showed
that predictions of population persistence were most sensitive to Barrier Presence, Nonnative Control, and Proximity
of Competitor Source Population. Proximity of Mc Source,
facilitated by the spread of nonnative trout, was the fifth
most influential parent node and was also capable of driving populations close to extirpation (i.e., persistence probability near zero). Of the top-five most influential nodes,
only the fourth (M30AT) was directly influenced by a
warming climate and did not involve nonnative invasions.
Two other climate-related nodes that were expected to
pose additional challenges to RGCT populations (Base
Flow Discharge and MWMT) appeared to have limited
influence (ranked 11th and 12th of the 17 parent nodes)
based on values in the data set.
DISCUSSION
Probabilities of persistence for individual RGCT conservation populations were extremely variable but were
generally higher in the two Rio Grande GMUs than in
the Pecos and Canadian River basins. In fact, just three
populations in each of the Pecos and Canadian GMUs
were predicted to have a probability greater than 0.50 of
persisting to 2080 in the absence of human intervention,
whereas two-thirds of the extant populations were predicted to have probabilities of persistence less than 0.10 in
the same time frame. Ten populations were predicted to
be extirpated by 2080 (eight in the Pecos GMU and two
in the Canadian GMU). The mean predicted probability
of RGCT population persistence across all populations
during the current time period was 0.53 (median = 0.50)
and decreased to 0.31 (median = 0.11) by 2040 and 0.26
(median = 0.03) by 2080. These low values are attributable
to the fact that half of the populations have already been
FIGURE 7. Median probabilities of persistence for the current time period, the 2040s, and the 2080s for Rio Grande Cutthroat Trout populations in
the Lower Rio Grande (n = 56), Rio Grande Headwaters (n = 43), Pecos (n = 12), and Canadian (n = 10) Geographic Management Units (GMUs).
Each box encompasses the first to third quartiles, and whiskers display the 10th and 90th percentiles for each GMU.
�PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
invaded by nonnative competitors and another 17% are in
close proximity to nonnatives, without barriers to protect
them from subsequent invasion. Populations with the
highest predicted future probabilities of persistence occupied large habitats, were not invaded by nonnatives, and
were protected by complete barriers.
Although the average predicted probability of persistence for individual populations was low and we predicted
that at least one population in each GMU will be extirpated by 2080, the probability that RGCT will be completely extirpated in any of the four GMUs where the
subspecies currently occurs (assuming independence in
the individual extinction probabilities) is remote. Even in
831
the Canadian GMU, where only 10 populations remain,
there is only a 3% chance for total extinction in that
GMU by 2080, even without human intervention. The
chances for total extinction are even lower for the Pecos
GMU (1%) and the two Rio Grande GMUs (<1% each).
However, if the genetic diversity currently represented in
these GMUs is to be preserved, additional management
actions will be required to sustain it.
Although our results indicate that RGCT are facing
multiple stressors acting at different time horizons, it is
clear that the most proximate threat is posed by invading
nonnative trout (Figure 9). The three factors to which persistence is most sensitive are all related to nonnative trout
FIGURE 8. Predicted persistence in the current time period and in the 2080s for Rio Grande Cutthroat Trout conservation populations in four
geographic management units. Populations are arranged in categories that indicate the status and influence of nonnative trout invasions.
�832
ZEIGLER ET AL.
FIGURE 9. Tornado diagram showing the sensitivity analysis of Rio Grande Cutthroat Trout (RGCT) population persistence to the range of values
for each Bayesian network (BN) node, including the maximum 30-d average temperature (M30AT), the maximum weekly maximum temperature
(MWMT), and proximity to the causative agent of whirling disease (Myxobolus cerebralis [Mc]). The bars show the ranges of model output for the
best-case and worst-case values of each input node in the RGCT database and abiotic predictions while holding all other input nodes at their default
values (source pop = source population).
invasions (barriers, nonnative control, and proximity to
competitors). Currently, 48% of the remaining RGCT
conservation populations (58 of 121) are already invaded
by Brook Trout and/or Brown Trout, and they have a
much lower probability of persisting to 2080 than populations in uninvaded streams. For another 37% (45 populations), nonnative trout are found within 10 km. One-third
of those (14 populations) are not protected by barriers to
upstream migration, leaving them highly vulnerable to
future invasion. The mean probability of persistence for
those 14 populations is predicted to decline from 0.40 to
less than 0.10 by the 2080s. These values are consistent
with other studies demonstrating that nonnative trout in
the southern Rocky Mountains can rapidly displace and
extirpate native Cutthroat Trout in these high-elevation
watersheds (Peterson and Fausch 2003; Peterson et al.
2004; Roberts et al. 2017). A warming climate may serve
to exacerbate those conditions, facilitating the spread of
warm-adapted invaders (Rahel and Olden 2008; Lawrence
et al. 2014), but even under current thermal regimes the
invasions of nonnative trout are sufficient to extirpate
individual RGCT populations.
The threat of rising stream temperatures will take
longer to manifest than invasions of nonnative trout.
The MWMTs are predicted to exceed 21.0°C for 11
RGCT conservation populations (9%) by 2080. Predicted
increases in stream temperature are smaller than those
predicted for the upper Colorado River basin and northern Rocky Mountains (Wenger et al. 2011; Roberts et al.
2013). However, our results indicate that an increase of
less than 1°C in stream temperature (a 0.69°C increase in
M30AT; or a 0.46°C increase in MWMT) by 2080 could
reduce persistence of these 11 populations even in the
absence of other stressors. Moreover, our stream temperature model tended to overestimate lower temperatures and
underestimate higher temperatures (Figure A.1.1), which
would lead to underestimating the effects of both temperature metrics on recruitment and growth. Although few
streams are likely to remain so cold as to prevent any
recruitment (M30AT < 8°C; Figure A.1.2), more streams
than we have predicted could warm enough to reduce
growth or survival, leading to a slight underestimate of
the negative effects of warming temperatures on future
RGCT populations. Given that the RGCT is the southernmost subspecies of Cutthroat Trout (Penaluna et al.
2016), we expected warming stream temperatures to be a
larger threat. However, similar to other subspecies of
inland Cutthroat Trout (Roberts et al. 2013, 2017), nonnative trout invasions have already restricted RGCT to the
highest elevation habitats (Alves et al. 2008), which are
currently cold enough to buffer against some stream
warming (sensu Isaak et al. 2016). Therefore, although
populations of RGCT are sensitive to temperature, under
current and projected future conditions their persistence is
much more likely to be affected by nonnative trout invasions.
Declining streamflows are also likely to pose problems
for RGCT populations over the long term. By 2080, we
expect about one-third of the conservation populations to
face flows that are low enough to threaten population persistence, even though levels are not expected to be low
enough to extirpate populations independent of other factors. Incorporating flow when assessing climate change
threats to salmonids is widely recognized as important
(Kovach et al. 2016) but is rarely accomplished (but see
Wenger et al. 2011). Accurate predictions of base flows for
�PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
specific reaches that support native trout are usually only
possible by using mechanistic hydrologic models, which
require extensive data and testing (e.g., Soil and Water
Assessment Tool; Jaeger et al. 2014). We encourage the collection of continuous streamflow data from high-elevation
sites within RGCT conservation population habitats to
further inform and reduce the uncertainty in predictions.
Site-specific information is essential to effectively plan
and implement conservation strategies. This is problematic
when assessing large-scale, long-term threats such as climate change because most of the relevant data are available at regional scales, so most analyses have been
conducted at these larger spatial scales (Williams et al.
2009; Wenger et al. 2011). Extensive site-specific data are
available for RGCT populations because an active conservation team has been overseeing recovery and management of this subspecies (Alves et al. 2008). This allowed us
to avoid making broad-scale assumptions about the effects
of crucial habitat factors and biotic interactions. Previous
studies have assumed that nonnatives were ubiquitous in
stream reaches downstream of Cutthroat Trout (Wenger
et al. 2011; Roberts et al. 2017) and have estimated population size solely from general relationships of abundance
with occupied stream length (Hilderbrand and Kershner
2000; Young et al. 2005). In contrast, we had information
on invasions by nonnative trout in each watershed, and
site-specific population surveys from the past decade were
available for calculating abundance estimates. Now that
the modeling framework has been established, the strength
of the BN lies in the ability to inform management for
similar but perhaps less well-studied species for which only
more regional data are available. However, it is the synthesis of both regional and extensive site-specific data that
allowed for a more comprehensive BN model with which
to predict population persistence.
Model Validation and Uncertainty
Our BN is a working model that represents a set of
hypotheses about factors influencing the persistence of
RGCT populations. It was developed using standard
methods recommended by key experts (Marcot et al. 2006;
Marcot 2012, 2017; Conroy and Peterson 2013), integrated
the best available information from empirical data and formal expert elicitation, and was subjected to independent
peer review. Nevertheless, the model requires validation to
reduce potential bias in the conditional probabilities on
which the sensitivity analysis and the model predictions are
based.
Although validation of the full model will have to wait
until persistence can be determined in 2080, four key relationships described by child nodes could be validated in
the current period: Invasion and Hybridization Risk, Potential Habitat Suitability, Competitor Invasion and Biotic
Influence Risk, and Mc Infection Risk. For example, for
833
Invasion and Hybridization Risk, 8 of 12 combinations of
the two relevant input states (Proximity of Hybridizing
Source Population and Barrier Presence) for the current
time period have conditional probabilities that are not
fixed at 1.0 for one output state, suggesting uncertainty
and an opportunity for further investigation (Supplementary Table S.3). One combination—the case in which the
invaders are near and the barrier is partial—exhibits the
most uncertainty. Therefore, a set of populations could be
selected at random from those that represent these eight
combinations, and the relevant data on incidence of invasion and hybridization could be monitored for the next
decade. The other three child nodes could be validated in
similar fashion, in each case selecting more replicates for
combinations where conditional probabilities are the most
uncertain.
Another important consideration is the uncertainty in
model predictions resulting from uncertainty in model
specification, the values of input variables, and the conditional probabilities assigned by experts to the combinations of node states. The greatest uncertainty for this BN
involves specifying which variables affect persistence (Figures 2–5) and how to combine them in the model. We
made extensive efforts to minimize this uncertainty when
developing the model; we did this by incorporating the
expert knowledge of five biologists who study Cutthroat
Trout ecology and conservation, subjecting the model to
formal peer review by five additional experts and three
reviewers for specific points, and soliciting reviews of the
initial DAG and predictions of the final model by biologists on the RGCT Conservation Team. In contrast,
nearly all of the states or values of parent nodes can be
measured with certainty (see Supplement), so they are not
a significant source of uncertainty for this BN.
We assessed the effect of variation among the five
experts in defining conditional probabilities by examining
child nodes affected by the five key parent nodes to which
the predictions of population persistence are most sensitive
(Figure 9). These include all four child nodes that can be
validated using data from the current period (described
above). Of those, two are each influenced by two of the
five key parent nodes, so for these we evaluated the variation among the five experts in assigning conditional probabilities. For Competitor Invasion and Biotic Influence Risk,
71% of the 108 conditional probabilities (see Supplementary Table S.5) assigned independently by the experts were
identical, and only 4% varied by an absolute value greater
than 10% on average (i.e., SD > 0.10). For Mc Infection
Risk, a topic under active investigation, there was less
information for parameterizing the conditional probabilities, and experts originally disagreed on 39% of them (n =
56 of 144 conditional probabilities; Supplementary
Table S.13). Therefore, we sought more information from
three scientists in the field of whirling disease research (R.
�834
ZEIGLER ET AL.
Nehring, G. Schisler, and E. Fetherman, Colorado Parks
and Wildlife, personal communication), discussed this
information, and arrived at a consensus for these combinations of states. Considering the remainder of the entries,
conditional probabilities assigned by the experts were
identical for 40% (n = 58 of the 144 entries) and varied by
over 10% for only 6% (n = 8). Therefore, we conclude that
variation among the five experts was, at most, a modest
source of uncertainty for predictions of RGCT population
persistence.
Conservation Strategies
Our model results indicate that invasions of nonnative
trout represent the primary threat to RGCT population
persistence, whereas other factors, such as temperature
and declining streamflows, are secondary. We identified
seven parent nodes (dashed outlines in Figures 3–5) that
can be influenced by management actions (see Supplement). Two of these (Barrier Presence and Nonnative Control) represented the most important attributes for
ensuring long-term persistence of RGCT populations in
our sensitivity analysis (Figure 9). Our results indicated
that the average probability of persistence to 2080 for
uninvaded RGCT populations is over eight times greater
with a complete barrier (0.66) than without (0.08), highlighting the importance of maintaining populations free
from nonnative trout and protecting them from subsequent invasion with man-made or natural downstream
barriers that ensure allopatry. Intentional fragmentation is
likely warranted in RGCT management to conserve biodiversity (Fausch et al. 2006, 2009; Rahel 2013), and existing
BN models can be used to help weigh the tradeoffs associated with barrier construction (Peterson et al. 2008b).
Prioritizing habitats that are predicted to warm slowly
and to have adequate streamflows in the future is also
important for ensuring long-term persistence of RGCT
populations. As such, management strategies that focus on
restoring riparian vegetation to buffer stream temperatures
against future climate warming should be considered. For
example, reduced livestock grazing (Anthropogenic Influence) allows for regrowth of riparian vegetation, increasing
shading and decreasing stream temperatures (Lawrence et
al. 2014). These efforts could also suppress invasions by
some nonnative species that thrive in warmer waters (Isaak
et al. 2015), thereby further reducing the negative effects on
Cutthroat Trout.
These strategies highlight a few actions that managers
can implement at the population level, but the BN can
also help to identify what can be done across GMUs to
protect unique genetic and morphological diversity. For
example, our results indicate that only three populations
in each of the Canadian and Pecos GMUs have a greater
than 0.50 probability of persisting by the 2080s. Therefore,
conservation in these two GMUs might focus on
reintroductions to other streams where conditions will support long-term persistence. Maintaining these peripheral
populations is important (Haak and Williams 2012), not
only for the genetic (Pritchard et al. 2007a, 2009) and
morphological (Bestgen et al. 2019) diversity they display
but also for the potential ecological diversity that is yet
unmeasured (e.g., thermal tolerance). In contrast, a different strategy may be needed for populations in the Rio
Grande Headwaters GMU (n = 43 populations), where 15
populations have an over 50% chance of persistence. Here,
effective conservation could include drainagewide monitoring of nonnative invasions, repatriating pure RGCT after
removal of nonnative trout from invaded habitats, and
restoring riparian vegetation to reduce rising stream temperatures and increase base flows. Overall, the model
described here will be useful in an adaptive management
framework (e.g., Williams 2011) to identify conservation
strategies that not only benefit specific populations but
also increase persistence of the subspecies throughout its
range.
Despite the broad distribution of RGCT populations
and robust recovery efforts, our results indicate that few
populations will exceed the 90% probability of persistence
criterion suggested by Mace and Lande (1991) by 2080
without human intervention. Active management by the
RGCT Conservation Team may be required to conserve
and replicate peripheral populations with unique genetic
and life history diversity (e.g., Pecos River drainage) as
well as to detect and counteract invasions in other GMUs.
In addition, our results highlight additional strategies
(e.g., riparian restoration) that can be explored to increase
the resilience of specific populations subject to rising temperatures and decreasing base flows brought about by a
warming climate.
The most effective strategy for maximizing probabilities
of persistence will likely be to seek out large watersheds
with intact habitat where connected populations of RGCT
can be sustained above natural or human-made barriers
that prevent invasions by nonnative trout and pathogens.
This strategy is currently being implemented through a
large public–private partnership between Turner Enterprises and the New Mexico Department of Game and
Fish, Colorado Parks and Wildlife, U.S. Forest Service,
and USFWS. When completed, that project will have
restored about 200 km of stream for RGCT in the Rio
Costilla watershed of Colorado and northern New Mexico
(K. Patten, New Mexico Department of Game and Fish,
personal communication). Identifying watersheds like the
Rio Costilla, where factors threatening current populations can be mitigated or eliminated, will be essential in
ensuring the long-term persistence of RGCT.
The BN developed to evaluate persistence of RGCT
populations represents a flexible modeling framework that
can easily incorporate new information as it becomes
�PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
available. Unlike earlier modeling efforts, this BN is also
well suited to accommodate modeled future conditions
brought about by a changing climate. Model results reinforced some prevailing thoughts on the relative jeopardy
of RGCT populations and helped to quantify the threats
that they will likely face over shorter- and longer-term
time horizons. Temperatures are projected to increase in
the desert Southwest more than in other areas of North
America (Cook et al. 2004; Dettinger et al. 2015), but the
BN model suggests that temperature is likely not the most
pressing concern for RGCT conservation, particularly in
the near future. Instead, invasion by nonnative salmonids
will remain the most proximate threat. Management
strategies that address and mitigate those invasions should
be a priority, and responses to other threats should be
considered once conservation populations have been
secured. Implementing the conservation strategies predicted to be most effective based on this model will help
to ensure that the native trout diversity distributed among
the remaining RGCT populations will persist, despite the
challenging conditions they face.
ACKNOWLEDGMENTS
We thank Robert Al-Chokhachy, Jeff Falke, Douglas
Peterson, David Stewart, and Andrew Whiteley for
detailed reviews of our BN model and Andrew Martin for
reviewing criteria for Ne and inbreeding risk. Five anonymous reviewers provided comments that improved the
manuscript. We also thank the RGCT Conservation
Team, particularly John Alves, Kirk Patten, Bryan Bakevich, and Ben Felt, for providing data that allowed for
the calculation of population estimates. Barry Nehring,
George Schisler, and Eric Fetherman provided helpful
information that enhanced the integrity of the nodes
involving M. cerebralis. A grant from the Southern Rocky
Mountain Landscape Conservation Cooperative provided
additional funding for this project and was administered
by Kevin Johnson. Any use of trade, firm, or product
names is for descriptive purposes only and does not imply
endorsement by the U.S. Government. There is no conflict
of interest declared in this article.
ORCID
Kevin B. Rogers
James J. Roberts
Kurt D. Fausch
https://orcid.org/0000-0001-9366-8162
https://orcid.org/0000-0002-4193-610X
https://orcid.org/0000-0001-5825-7560
REFERENCES
Al-Chokhachy, R., and A. J. Sepulveda. 2019. Impacts of nonnative
Brown Trout on Yellowstone Cutthroat Trout in a tributary stream.
North American Journal of Fisheries Management 39:17–28.
835
Allendorf, F. W., R. F. Leary, P. Spruell, and J. K. Wenburg. 2001. The
problems with hybrids: setting conservation guidelines. Trends in
Ecology and Evolution 16:613–622.
Allendorf, F. W., and G. Luikart. 2007. Conservation and the genetics of
populations. Blackwell Publishing, Malden, Massachusetts.
Alves, J. E., K. A. Patten, D. E. Brauch, and P. M. Jones. 2008. Rangewide status assessment of Rio Grande Cutthroat Trout (Oncorhynchus
clarkii virginalis): 2008. Colorado Division of Wildlife, Fort Collins.
Available: http://cpw.state.co.us/learn/Pages/ResearchRioGrandeCut
throatTrout.aspx. (March 2018).
Anderson, D. R. 2008. Model based inference in the life sciences: a primer on evidence. Springer, New York.
Araújo, M. B., and M. Luoto. 2007. The importance of biotic interactions for modeling species distributions under climate change. Global
Ecology and Biogeography 16:743–753.
Armour, C., D. Duff, and W. Elmore. 1991. The effects of livestock
grazing on western riparian and stream ecosystems. Fisheries 19(9):
9–12.
Ayre, K. K., C. A. Caldwell, J. Stinson, and W. G. Landis. 2014. Analysis of regional scale risk of whirling disease in populations of Colorado and Rio Grande Cutthroat Trout using a Bayesian belief
network model. Journal of Risk Analysis 34:1589–1605.
Bear, E. A., T. E. McMahon, and A. V. Zale. 2007. Comparative thermal requirements of Westslope Cutthroat Trout: implications for species interactions and development of thermal protection standards.
Transactions of the American Fisheries Society 136:1113–1121.
Behnke, R. J. 1992. Native trout of western North America. American
Fisheries Society, Monograph 6, Bethesda, Maryland.
Belmar-Lucero, S., J. L. A. Wood, S. Scott, A. B. Harbicht, J. A. Hutchings, and D. J. Fraser. 2012. Concurrent habitat and life history influences on effective/census population size ratios in stream-dwelling
trout. Ecology and Evolution 2:562–573.
Benjamin, J. R., and C. V. Baxter. 2012. Is a trout a trout? A range-wide
comparison shows nonnative Brook Trout exhibit greater density, biomass, and production than native inland Cutthroat Trout. Biological
Invasions 14:1865–1879.
Bestgen, K. R., K. B. Rogers, and R. Granger. 2019. Distinct phenotypes
of native Cutthroat Trout emerge under a molecular model of lineage
distributions. Transactions of the American Fisheries Society 148:
442–463.
Brown, T. J., B. L. Hall, and A. L. Westerling. 2004. The impact of
twenty-first century climate change on wildland fire danger in the
western United States: an applications perspective. Climatic Change
62:365–388.
Budy, P., S. Wood, and B. Roper. 2012. A study of the spawning ecology and early life history survival of Bonneville Cutthroat Trout.
North American Journal of Fisheries Management 32:436–449.
Burton, T. A. 2005. Fish and stream habitat risks from uncharacteristic
wildfire: observations from 17 years of fire-related disturbances on the
Boise National Forest, Idaho. Forest Ecology and Management
211:140–149.
Cain, J. 2001. Planning improvements in natural resources management:
guidelines for using Bayesian networks to support the planning and management of development programmes in the water sector and beyond. Centre for Ecology and Hydrology, Crowmarsh Gifford, Wallingford, UK.
Caissie, D. 2006. The thermal regime of rivers: a review. Freshwater
Biology 51:1389–1406.
Cannon, S. H., J. E. Gartner, M. G. Rupert, J. A. Michael, A. H. Rae,
and C. Parrett. 2010. Predicting the probability and volume of postwildfire debris flows in the intermountain western United States. Geological Society of America Bulletin 122:127–144.
Chapin, T. P., A. S. Todd, and M. P. Zeigler. 2014. Robust, low-cost
data loggers for stream temperature, flow intermittency, and relative
conductivity monitoring. Water Resources Research 50:6542–6548.
�836
ZEIGLER ET AL.
Clark, K. E., J. E. Applegate, L. J. Niles, and D. S. Dobkin. 2006. An
objective means of species status assessment: adapting the Delphi
technique. Wildlife Society Bulletin 34:419–425.
Coleman, M. A., and K. D. Fausch. 2007a. Cold summer temperature limits
recruitment of age-0 Cutthroat Trout in high-elevation Colorado streams.
Transactions of the American Fisheries Society 136:1231–1244.
Coleman, M. A., and K. D. Fausch. 2007b. Cold summer temperature
regimes cause a recruitment bottleneck in age-0 Colorado River Cutthroat Trout reared in laboratory streams. Transactions of the American Fisheries Society 136:639–654.
Conroy, M. J., and J. T. Peterson. 2013. Decision making in natural
resource management: a structured, adaptive approach. Wiley-Blackwell, West Sussex, UK.
Cook, E. R., C. A. Woodhouse, C. M. Eakin, D. M. Meko, and D. W.
Stahle. 2004. Long-term aridity changes in the western United States.
Science 306:1015–1018.
de la Hoz Franco, E. A., and P. Budy. 2005. Effects of biotic and abiotic
factors on the distribution of trout and salmon along a longitudinal
stream gradient. Environmental Biology of Fishes 72:379–391.
Dettinger, M., B. Udall, and A. Georgakakos. 2015. Western water and
climate change. Ecological Applications 25:2069–2093.
Dunham, J., S. B. Adams, R. Schroeter, and D. Novinger. 2002. Alien
invasions in aquatic ecosystems: toward an understanding of Brook
Trout invasions and their potential impacts on inland Cutthroat
Trout in western North America. Reviews in Fish Biology and Fisheries 12:373–391.
Dunham, J. B., A. E. Rosenberger, C. H. Luce, and B. E. Rieman. 2007.
Influences of wildfire and channel reorganization on spatial and temporal variation in stream temperature and the distribution of fish and
amphibians. Ecosystems 10:335–346.
Dunham, J. B., G. L. Vinyard, and B. E. Rieman. 1997. Habitat fragmentation and extinction risk of Lahontan Cutthroat Trout. North
American Journal of Fisheries Management 17:1126–1133.
Dunham, J. B., M. K. Young, R. E. Gresswell, and B. E. Rieman. 2003.
Effects of fire on fish populations: landscape perspectives on persistence of native fishes and nonnative fish invasion. Forest Ecology and
Management 178:183–196.
El-Matbouli, M., T. S. McDowell, D. B. Antonio, K. B. Andree, and R.
P. Hedrick. 1999. Effect of water temperature on the development,
release and survival of the triactinomyxon stage of Myxobolus cerebralis in its oligochaete host. International Journal for Parasitology
29:627–641.
Falcone, J. A., D. M. Carlisle, D. M. Wolock, and M. R. Meador. 2010.
GAGES: a stream gage database for evaluating natural and altered
flow conditions in the conterminous United States. Ecology 91:621.
Fausch, K. D. 2007. Introduction, establishment and effects of nonnative salmonids: considering the risk of Rainbow Trout invasion in
the United Kingdom. Journal of Fish Biology 71:1–32.
Fausch, K. D., B. E. Rieman, J. B. Dunham, M. K. Young, and D. P.
Peterson. 2009. Invasion versus isolation: trade-offs in managing
native salmonids with barriers to upstream movement. Conservation
Biology 23:859–870.
Fausch, K. D., B. E. Rieman, M. K. Young, and J. B. Dunham. 2006.
Strategies for conserving native salmonid populations at risk from
nonnative fish invasions: tradeoffs in using barriers to upstream movement. U.S. Forest Service General Technical Report RMS-GTR-174.
Fausch, K. D., Y. Taniguchi, S. Nakano, G. D. Grossman, and C. R.
Townsend. 2001. Flood disturbance regimes influence Rainbow Trout
invasion success among five Holarctic regions. Ecological Applications 11:1438–1455.
Fortin, M., and M. Dale. 2005. Spatial analysis: a guide for ecologists.
Cambridge University Press, Cambridge, UK.
Frankham, R. 2005. Genetics and extinction. Biological Conservation
126:131–140.
Gresswell, R. E. 1999. Fire and aquatic ecosystems in forested biomes of
North America. Transactions of the American Fisheries Society
128:193–221.
Haak, A. L., and J. E. Williams. 2012. Spreading the risk: native trout
management in a warmer and less-certain future. North American
Journal of Fisheries Management 32:387–401.
Harig, A. L., and K. D. Fausch. 2002. Minimum habitat requirements
for establishing translocated Cutthroat Trout populations. Ecological
Applications 12:535–551.
Helsel, D. R., and R. M. Hirsch. 1992. Statistical methods in water
resources. Elsevier, New York.
Hilderbrand, R. H. 2003. The roles of carrying capacity, immigration,
and population synchrony on persistence of stream-resident Cutthroat
Trout. Biological Conservation 110:257–266.
Hilderbrand, R. H., and J. L. Kershner. 2000. Conserving inland Cutthroat Trout in small streams: how much stream is enough? North
American Journal of Fisheries Management 20:513–520.
Hirsch, C. L., S. E. Albeke, and T. P. Nesler. 2006. Range-wide status of
Colorado River Cutthroat Trout (Oncorhynchus clarkii pleuriticus):
2005. Colorado Division of Wildlife, Denver.
Hostetler, S. W., J. R. Alder, and A. M. Allan. 2011. Dynamically
downscaled climate simulations over North America: methods, evaluation, and supporting documentation for users. U.S. Geological Survey Open-File Report 2011-1238.
Huggins, R. M. 1989. On the statistical analysis of capture experiments.
Biometrika 76:133–140.
IPCC (Intergovernmental Panel on Climate Change). 2007. Climate
change 2007: a synthesis report. IPCC, Geneva, Switzerland. Available: www.ipcc.ch. (July 2014).
IPCC (Intergovernmental Panel on Climate Change). 2013. Climate
change 2013: the physical science basis. Cambridge University Press,
Cambridge, UK.
Isaak, D. J., C. H. Luce, B. E. Rieman, D. E. Nagel, E. E. Peterson, D.
L. Horan, S. Parkes, and G. L. Chandler. 2010. Effects of climate
change and wildfire on stream temperatures and salmonid habitat in
a mountain river network. Ecological Applications 20:1350–1371.
Isaak, D. J., C. C. Muhlfeld, A. S. Todd, R. Al-Chokhachy, J. Roberts,
J. L. Kershner, K. D. Fausch, and S. W. Hostetler. 2012. The past as
prelude to the future for understanding 21st-century climate effects on
Rocky Mountain trout. Fisheries 37:542–556.
Isaak, D. J., E. E. Peterson, J. M. Ver Hoef, S. J. Wenger, J. A. Falke,
C. E. Torgersen, C. Sowder, E. A. Steel, M. Fortin, C. E. Jordan,
A. S. Ruesch, N. Som, and P. Monestiez. 2014. Applications of spatial statistical network models to stream data. WIREs Water [online
serial] 1:277–294.
Isaak, D. J., M. K. Young, C. H. Luce, S. W. Hostetler, S. J. Wenger,
E. E. Peterson, J. M. V. Hoef, M. C. Groce, D. L. Horan, and D. E.
Nagel. 2016. Slow climate velocities of mountain streams portend
their role as refugia for cold-water biodiversity. Proceedings of the
National Academy of Sciences of the USA 113:4374–4379.
Isaak, D. J., M. K. Young, D. E. Nagel, D. L. Horan, and M. C. Groce.
2015. The cold-water climate shield: delineating refugia for preserving
salmonid fishes through the 21st century. Global Change Biology
21:2540–2553.
Jaeger, K. L., J. D. Olden, and N. A. Pelland. 2014. Climate change
poised to threaten hydrologic connectivity and endemic fishes in dryland streams. Proceedings of the National Academy of Sciences of the
USA 111:13894–13899.
Jamieson, I. G., and F. W. Allendorf. 2012. How does the 50/500 rule
apply to MVPs? Trends in Ecology and Evolution 27:578–584.
Japhet, M., J. Alves, and T. Nesler. 2007. Rio Grande Cutthroat Trout
status review for Colorado. Colorado Parks and Wildlife, Denver.
Jensen, L. F., M. M. Hansen, J. Carlsson, V. Loeschcke, and K. L. D.
Mensberg. 2005. Spatial and temporal genetic differentiation and
�PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
effective population size of Brown Trout (Salmo trutta, L.) in small
Danish rivers. Conservation Genetics 6:615–621.
Kershner, J. L., B. B. Roper, N. Bouwes, R. Henderson, and E. Archer.
2004. An analysis of stream habitat conditions in reference and managed watersheds on some federal lands within the Columbia River
basin. North American Journal of Fisheries Management 24:1363–
1375.
Kovach, R. P., R. Al-Chokhachy, D. C. Whited, D. A. Schmetterling,
A. M. Dux, and C. C. Muhlfeld. 2017. Climate, invasive species and
land use drive population dynamics of a cold-water specialist. Journal
of Applied Ecology 54:638–647.
Kovach, R. P., C. C. Muhlfeld, R. Al-Chokhachy, J. B. Dunham, B. H.
Letcher, and J. L. Kershner. 2016. Impacts of climatic variation on
trout: a global synthesis and path forward. Reviews in Fish Biology
and Fisheries 26:135–151.
Lawrence, D. J., B. Stewart-Koster, J. D. Olden, A. S. Ruesch, C. E.
Torgersen, J. J. Lawler, D. P. Butcher, and J. K. Crown. 2014. The
interactive effects of climate change, riparian management, and a
nonnative predator on stream-rearing salmon. Ecological Applications
24:895–912.
Lee, D. C., and B. E. Rieman. 1997. Population viability assessment of
salmonids by using probabilistic networks. North American Journal
of Fisheries Management 17:1144–1157.
Luce, C. H., and Z. A. Holden. 2009. Declining annual streamflow distributions in the Pacific Northwest United States, 1948–2006. Geophysical Research Letters 36(16):L16401.
Mace, G. M., and R. Lande. 1991. Assessing extinction threats—toward
a reevaluation of IUCN threatened species categories. Conservation
Biology 5:148–157.
Marcot, B. G. 2012. Metrics for evaluating performance and uncertainty
of Bayesian network models. Ecological Modelling 230:50–62.
Marcot, B. G. 2017. Common quandaries and their practical solutions in
Bayesian network modeling. Ecological Modelling 358:1–9.
Marcot, B. G., J. D. Seventon, G. D. Sutherland, and R. K. McCann.
2006. Guidelines for developing and updating Bayesian belief networks for ecological modeling. Canadian Journal of Forest Research
36:3063–3074.
McHugh, P., and P. Budy. 2005. An experimental evaluation of competitive and thermal effects on Brown Trout (Salmo trutta) and Bonneville Cutthroat Trout (Oncorhynchus clarkii utah) performance
along an altitudinal gradient. Canadian Journal of Fisheries and
Aquatic Sciences 62:2784–2795.
McHugh, P., and P. Budy. 2006. Experimental effects of nonnative
Brown Trout on the individual- and population-level performance of
native Bonneville Cutthroat Trout. Transactions of the American
Fisheries Society 135:1441–1455.
McKelvey, K. S., M. K. Young, T. M. Wilcox, D. M. Bingham, K. L.
Pilgrim, and M. K. Schwartz. 2016. Patterns of hybridization among
Cutthroat Trout and Rainbow Trout in northern Rocky Mountain
streams. Ecology and Evolution 6:688–706.
McKenzie, D., Z. Gedalof, D. L. Peterson, and P. Mote. 2004. Climate change, wildfire, and conservation. Conservation Biology
18:890–902.
Miller, L. W., and S. Bassett. 2013. Rio Grande Cutthroat Trout wildfire
risk assessment. Nature Conservancy, Santa Fe, New Mexico.
Morita, K., and S. Yamamoto. 2002. Effects of habitat fragmentation by
damming on the persistence of stream-dwelling charr populations.
Conservation Biology 16:1318–1323.
Morris, W. F., P. L. Bloch, B. R. Hudgens, L. C. Moyle, and J. R.
Stinchcombe. 2002. Population viability analysis in endangered species recovery plans: past use and future improvements. Ecological
Applications 12:708–712.
Muhlfeld, C. C., S. T. Kalinowski, T. E. McMahon, M. L. Taper, S.
Painter, R. F. Leary, and F. W. Allendorf. 2009a. Hybridization
837
rapidly reduces fitness of a native trout in the wild. Biology Letters
5:328–331.
Muhlfeld, C. C., R. P. Kovach, R. Al-Chokhachy, S. J. Amish, J. L.
Kershner, R. F. Leary, W. H. Lowe, G. Luikart, P. Matson, D. A.
Schmetterling, B. B. Shepard, P. A. H. Westley, D. Whited, A.
Whiteley, and F. W. Allendorf. 2017. Legacy introductions and climatic variation explain spatiotemporal patterns of invasive hybridization in a native trout. Global Change Biology 23:4663–4674.
Muhlfeld, C. C., R. P. Kovach, L. A. Jones, R. Al-Chokhachy, M. C.
Boyer, R. F. Leary, W. H. Lowe, G. Luikart, and F. W. Allendorf.
2014. Invasive hybridization in a threatened species is accelerated by
climate change. Nature Climate Change 4:620–624.
Muhlfeld, C. C., T. E. McMahon, M. C. Boyer, and R. E. Gresswell.
2009b. Local habitat, watershed, and biotic factors influencing the
spread of hybridization between native Westslope Cutthroat Trout
and introduced Rainbow Trout. Transactions of the American Fisheries Society 138:1036–1051.
Nehring, R. B., P. Lukacs, D. V. Baxa, M. E. T. Stinson, L. Chiaramonte, S. K. Wise, B. Poole, and A. Horton. 2014. Susceptibility to
Myxobolus cerebralis among Tubifex tubifex populations from ten
major drainage basins in Colorado where Cutthroat Trout are endemic. Journal of Aquatic Animal Health 26:19–32.
Nehring, R. B., and P. G. Walker. 1996. Whirling disease in the wild:
the new reality in the intermountain west. Fisheries 21(6):28–30.
Novinger, D. C., and F. J. Rahel. 2003. Isolation management with artificial barriers as a conservation strategy for Cutthroat Trout in headwater streams. Conservation Biology 17:772–781.
O'Grady, J. J., B. W. Brook, D. H. Reed, J. D. Ballou, D. W. Tonkyn,
and R. Frankham. 2006. Realistic levels of inbreeding depression
strongly affect extinction risk in wild populations. Biological Conservation 133:42–51.
Palm, S., L. Laikre, P. E. Jordan, and N. Ryman. 2003. Effective population size and temporal genetic change in stream resident Brown
Trout (Salmo trutta, L.). Conservation Genetics 4:249–264.
Patten, K. A., R. Castell, S. Denny, J. Dominguez, E. Frey, R. Hansen,
J. Vega, and C. M. Wethington. 2007. Final report as required by
Federal Aid in Sport Fish Restoration Act. New Mexico Department
of Game and Fish, Federal Aid in Sport Fish Restoration, Project F69, Final Report, Santa Fe.
Penaluna, B. E., A. Abadía-Cardoso, J. B. Dunham, F. J. García-Dé
León, R. E. Gresswell, A. R. Luna, E. B. Taylor, B. B. Shepard, R.
Al-Chokhachy, C. C. Muhlfeld, K. R. Bestgen, K. Rogers, M. A.
Escalante, E. R. Keeley, G. M. Temple, J. E. Williams, K. R. Matthews, R. Pierce, R. L. Mayden, R. P. Kovach, J. C. Garza, and K.
D. Fausch. 2016. Conservation of native Pacific trout diversity in
western North America. Fisheries 41:286–300.
Peterson, D. P., and K. D. Fausch. 2003. Dispersal of Brook Trout promotes
invasion success and replacement of native Cutthroat Trout. Canadian
Journal of Fisheries and Aquatic Sciences 60:1502–1516.
Peterson, D. P., K. D. Fausch, J. Watmough, and R. A. Cunjak. 2008a.
When eradication is not an option: modeling strategies for electrofishing suppression of nonnative Brook Trout to foster persistence of
sympatric native Cutthroat Trout in small streams. North American
Journal of Fisheries Management 28:1847–1867.
Peterson, D. P., K. D. Fausch, and G. C. White. 2004. Population ecology of an invasion: effects of Brook Trout on native Cutthroat Trout.
Ecological Applications 14:754–772.
Peterson, D. P., B. E. Rieman, J. B. Dunham, K. D. Fausch, and M. K.
Young. 2008b. Analysis and trade-offs between threats of invasion by
nonnative Brook Trout (Salvelinus fontinalis) and intentional isolation
for native Westslope Cutthroat Trout (Oncorhynchus clarkii lewisi).
Canadian Journal of Fisheries and Aquatic Sciences 65:557–573.
Peterson, D. P., B. E. Rieman, D. L. Horan, and M. K. Young. 2014.
Patch size but not short-term isolation influences occurrence of
�838
ZEIGLER ET AL.
Westslope Cutthroat Trout above human-made barriers. Ecology of
Freshwater Fish 23:556–571.
Peterson, D. P., B. E. Rieman, M. K. Young, and J. A. Brammer. 2010.
Modeling predicts that redd trampling by cattle may contribute to population declines of native trout. Ecological Applications 20:954–966.
Peterson, D. P., S. J. Wenger, B. E. Rieman, and D. J. Isaak. 2013.
Linking climate change and fish conservation efforts using spatially
explicit decision support tools. Fisheries 38:112–127.
Peterson, E. E., D. M. Theobald, and J. M. Ver Hoef. 2007. Geostatistical modelling on stream networks: developing valid covariance matrices based on hydrologic distance and stream flow. Freshwater
Biology 52:267–279.
Peterson, E. E., and J. M. Ver Hoef. 2010. A mixed-model moving-average approach to geostatistical modeling in stream networks. Ecology
91:644–651.
Peterson, E. E., and J. M. Ver Hoef. 2014. STARS: an ArcGIS toolset
used to calculate the spatial information needed to fit spatial statistical models to stream network data. Journal of Statistical Software
[online serial] 56(2).
Power, M. 1993. The predictive validation of ecological and environmental models. Ecological Modelling 68:33–50.
Pritchard, V. L., and D. E. Cowley. 2006. Rio Grande Cutthroat Trout
(Oncorhynchus clarkii virginalis): a technical conservation assessment.
U.S. Forest Service, Rocky Mountain Region, Lakewood, Colorado.
Available: http://www.fs.fed.us/r2/projects/scp/assessments/riograndec
utthroattrout.pdf.
Pritchard, V. L., K. Jones, and D. E. Cowley. 2007a. Genetic diversity
within fragmented Cutthroat Trout populations. Transactions of the
American Fisheries Society 136:606–623.
Pritchard, V. L., K. Jones, and D. E. Cowley. 2007b. Estimation of
introgression in Cutthroat Trout populations using microsatellites.
Conservation Genetics 8:1311–1329.
Pritchard, V. L., J. L. Metcalf, K. Jones, A. P. Martin, and D. E. Cowley. 2009. Population structure and genetic management of Rio
Grande Cutthroat Trout (Oncorhynchus clarkii virginalis). Conservation Genetics 10:1209–1221.
Propst, D. L., J. A. Stefferud, and P. R. Turner. 1992. Conservation and
status of Gila Trout, Oncorhynchus gilae. Southwestern Naturalist
37:117–125.
R Development Core Team. 2016. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna.
Rahel, F. J. 2004. Unauthorized fish introductions: fisheries management
of the people, for the people, or by the people? Pages 431–443 in M.
J. Nickum, P. M. Mazik, J. G. Nickum, and D. D. MacKinlay, editors. Propagated fish in resource management. American Fisheries
Society, Symposium 44, Bethesda, Maryland.
Rahel, F. J. 2013. Intentional fragmentation as a management strategy in
aquatic systems. BioScience 63:362–372.
Rahel, F. J., and N. P. Nibbelink. 1999. Spatial patterns in relations
among Brown Trout (Salmo trutta) distribution, summer air temperature, and stream size in Rocky Mountain streams. Canadian Journal
of Fisheries and Aquatic Sciences 56(Supplement 1):43–51.
Rahel, F. J., and J. D. Olden. 2008. Assessing the effects of climate
change on aquatic invasive species. Conservation Biology 22:521–533.
Rhymer, J. M., and D. Simberloff. 1996. Extinction by hybridization and
introgression. Annual Review of Ecology and Systematics 27:83–109.
Rieman, B., and J. Clayton. 1997. Wildfire and native fish: issues of
forest health and conservation sensitive species. Fisheries 22(11):6–15.
Rieman, B. E., and F. W. Allendorf. 2001. Effective population size and
genetic conservation criteria for Bull Trout. North American Journal
of Fisheries Management 21:756–764.
Rieman, B. E., and J. B. Dunham. 2000. Metapopulations and salmonids: a synthesis of life history patterns and empirical observations.
Ecology of Freshwater Fish 9:51–64.
Rieman, B. E., and D. J. Isaak. 2010. Climate change, aquatic ecosystems, and fishes in the Rocky Mountain West: implications and alternatives for management. U.S. Forest Service General Technical
Report RMRS-GTR-250.
Rinne, J. N. 1996. Short-term effects of wildfire on fishes and aquatic
macroinvertebrates in the southwestern United States. North American Journal of Fisheries Management 16:653–658.
Roberts, J. J., K. D. Fausch, M. B. Hooten, and D. P. Peterson. 2017.
Nonnative trout invasions combined with climate change threaten
persistence of isolated Cutthroat Trout populations in the southern
Rocky Mountains. North American Journal of Fisheries Management
37:314–325.
Roberts, J. J., K. D. Fausch, D. P. Peterson, and M. B. Hooten. 2013.
Fragmentation and thermal risks from climate change interact to
affect persistence of native trout in the Colorado River basin. Global
Change Biology 19:1383–1398.
Robinson, Z. L., J. A. Coombs, M. Hudy, K. H. Nislow, B. H. Letcher,
and A. R. Whiteley. 2017. Experimental test of genetic rescue in
isolated populations of Brook Trout. Molecular Ecology 26:4418–
4433.
Rogers, K. B. 2006. JakeOmatic: data analysis software for fishery managers, version 2.4. Colorado Division of Wildlife, Fort Collins. Available: http://cpw.state.co.us/learn/Pages/ResearchAquaticSoftware.aspx.
(July 2019).
Running, S. W., R. R. Nemani, and R. D. Hungerford. 1987. Extrapolation of synoptic meteorological data in mountainous terrain and its
use for simulating forest evapotranspiration and photosynthesis.
Canadian Journal of Forest Research 17:472–483.
Saunders, W. C., K. D. Fausch, and G. C. White. 2011. Accurate estimation of salmonid abundance in small streams using nighttime
removal electrofishing: an evaluation using marked fish. North American Journal of Fisheries Management 31:403–415.
Schisler, G. J., and E. P. Bergersen. 2002. Evaluation of risk of high elevation Colorado waters to the establishment of Myxobolus cerebralis.
Pages 33–41 in J. L. Bartholomew and J. C. Wilson, editors. Whirling
disease: reviews and current topics. American Fisheries Society, Symposium 29, Bethesda, Maryland.
Seager, R., M. Ting, I. Held, Y. Kushnir, J. Lu, G. Vecchi, H. Huang,
A. Leetmaa, N. Lau, C. Li, J. Velez, and N. Naik. 2007. Model projections of an imminent transition to a more arid climate in southwestern North America. Science 316:1181–1184.
Seager, R., M. Ting, C. Li, N. Naik, B. Cook, J. Nakamura, and H.
Liu. 2013. Projections of declining surface-water availability for the
southwestern United States. Nature Climate Change 3:482–486.
Shepard, B. B., B. E. May, and W. Urie. 2005. Status and conservation
of Westslope Cutthroat Trout within the western United States. North
American Journal of Fisheries Management 25:1426–1440.
Shepard, B. B., L. M. Nelson, M. L. Taper, and A. V. Zale. 2014. Factors influencing successful eradication of nonnative Brook Trout from
four small Rocky Mountain streams using electrofishing. North
American Journal of Fisheries Management 34:988–997.
Shepard, B. B., R. Spoon, and L. Nelson. 2002. A native Westslope Cutthroat Trout population responds positively after Brook Trout
removal and habitat restoration. Intermountain Journal of Science
8:191–211.
Smith, D. R., N. L. Allan, C. P. McGowan, J. A. Szymanski, S. R. Oetker, and H. M. Bell. 2018. Development of a species status assessment process for decisions under the U.S. Endangered Species Act.
Journal of Fish and Wildlife Management 9:302–320.
Stapp, P., and G. D. Hayward. 2002. Effects of an introduced piscivore
on native trout: insights from a demographic model. Biological Invasions 4:299–316.
Stein, A., and L. C. Corsten. 1991. Universal kriging and cokriging as a
regression procedure. Biometrics 47:575–587.
�PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
Thompson, K. G., R. B. Nehring, D. C. Bowden, and T. Wygant. 1999.
Field exposure of seven species or subspecies of salmonids to
Myxobolus cerebralis in the Colorado River, Middle Park, Colorado.
Journal of Aquatic Animal Health 11:312–329.
UDWR (Utah Division of Wildlife Resources). 2000. Genetic considerations associated with Cutthroat Trout management: a position paper.
UDWR, Publication Number 00-26, Salt Lake City. Available: http://
cpw.state.co.us/cutthroat-trout. (October 2016).
USFWS (U.S. Fish and Wildlife Service). 2008. Status review for Rio
Grande Cutthroat Trout. Federal Register 73:94(May 14, 2008):
27900–27926.
USFWS (U.S. Fish and Wildlife Service). 2014. 12-month finding on a
petition to list Rio Grande Cutthroat Trout as an endangered or
threatened species. Federal Register 79:190(October 1, 2014):59140–
59150.
USFWS (U.S. Fish and Wildlife Service). 2016. USFWS species status
assessment framework: an integrated analytical framework for conservation, version 3.4. USFWS, Washington, D.C. Available: https://
www.fws.gov/endangered/improving_esa/ssa.html. (September 2018).
Van der Putten, W. H., M. Macel, and M. E. Visser. 2010. Predicting
species distribution and abundance responses to climate change: why
it is essential to include biotic interactions across trophic levels. Philosophical Transactions of the Royal Society B: Biological Sciences
365:2025–2034.
Ver Hoef, J. M., E. E. Peterson, D. Clifford, and R. Shah. 2014. SSN:
an R package for spatial statistical modeling on stream networks.
Journal of Statistical Software [online serial] 56(3).
Vincent, E. R. 1996. Whirling disease and wild trout: the Montana experience. Fisheries 21(6):32–33.
Wagner, E., R. Arndt, M. Brough, and D. W. Roberts. 2002. Comparison of susceptibility of five Cutthroat Trout strains to Myxobolus cerebralis infection. Journal of Aquatic Animal Health 14:89–91.
Wang, L., and R. J. White. 1994. Competition between wild Brown Trout
and hatchery Greenback Cutthroat Trout of largely wild parentage.
North American Journal of Fisheries Management 14:475–487.
Wenger, S. J., D. J. Isaak, C. H. Luce, H. M. Neville, K. D. Fausch, J.
B. Dunham, D. C. Dauwalter, M. K. Young, M. M. Elsner, B. E.
Rieman, A. F. Hamlet, and J. E. Rieman. 2011. Flow regime, temperature, and biotic interactions drive differential declines of trout
species under climate change. Proceedings of the National Academy
of Sciences of the USA 108:14175–14180.
Westerling, A. L. 2016. Increasing western U.S. forest wildfire activity:
sensitivity to changes in the timing of spring. Philosophical Transactions of the Royal Society B: Biological Sciences 371:20150178.
Westerling, A. L., H. G. Hidalso, D. R. Cayan, and T. W. Swetnam.
2006. Warming and earlier spring increases western U.S. forest wildfire activity. Science 313:940–943.
White, G. C., and K. P. Burnham. 1999. Program MARK: survival estimation from populations of marked animals. Bird Study 46(Supplement):120–138.
White, S. M., and F. J. Rahel. 2008. Complementation of habitats for
Bonneville Cutthroat Trout in watersheds influenced by beavers, livestock, and drought. Transactions of the American Fisheries Society
137:881–894.
Whiteley, A. R., J. A. Coombs, M. Hudy, Z. Robinson, A. R. Colton,
K. H. Nislow, and B. H. Letcher. 2013. Fragmentation and patch size
839
shape genetic structure of Brook Trout populations. Canadian Journal of Fisheries and Aquatic Sciences 70:678–688.
Whiteley, A. R., J. A. Coombs, M. Hudy, Z. Robinson, K. H. Nislow,
and B. H. Letcher. 2012. Sampling strategies for estimating
Brook Trout effective population size. Conservation Genetics 13:625–
637.
Whiteley, A. R., K. Hastings, J. K. Wenburg, C. A. Frissell, J. C. Martin, and F. W. Allendorf. 2010. Genetic variation and effective population size in isolated populations of Coastal Cutthroat Trout.
Conservation Genetics 11:1929–1943.
Williams, B. K. 2011. Adaptive management of natural resources—
framework and issues. Journal of Environmental Management
92:1346–1353.
Williams, J. E., A. L. Haak, H. M. Neville, and W. T. Coyler. 2009.
Potential consequences of climate change to persistence of Cutthroat
Trout populations. North American Journal of Fisheries Management
29:533–548.
Young, M. K. 1994. Mobility of Brown Trout in south-central Wyoming
streams. Canadian Journal of Zoology 72:2078–2083.
Young, M. K., and P. M. Guenther-Gloss. 2004. Population characteristics of Greenback Cutthroat Trout in streams: their relation to model
predictions and recovery criteria. North American Journal of Fisheries Management 24:184–197.
Young, M. K., P. M. Guenther-Gloss, and A. D. Ficke. 2005. Predicting
Cutthroat Trout (Oncorhynchus clarki) abundance in high-elevation
streams: revisiting a model of translocation success. Canadian Journal
of Fisheries and Aquatic Sciences 62:2399–2408.
Young, M. K., D. J. Isaak, K. S. McKelvey, T. M. Wilcox, M. R.
Campbell, M. P. Corsi, D. Horan, and M. K. Schwartz. 2017. Ecological segregation moderates a climactic conclusion to trout
hybridization. Global Change Biology 23:5021–5023.
Young, M. K., D. J. Isaak, K. S. McKelvey, T. M. Wilcox, K. L. Pilgrim, K. J. Carim, M. R. Campbell, M. P. Corsi, D. L. Horan, D. E.
Nagel, and M. K. Schwartz. 2016. Climate, demography, and zoogeography predict introgression thresholds in salmonid hybrid zones
in Rocky Mountain streams. PLoS ONE [online serial] 11:e0167711.
Zeigler, M. P., S. F. Brinkman, C. A. Caldwell, A. S. Todd, M. S.
Recsetar, and S. A. Bonar. 2013a. Upper thermal tolerances of Rio
Grande Cutthroat Trout under constant and fluctuating
temperatures. Transactions of the American Fisheries Society
142:1395–1405.
Zeigler, M. P., A. S. Todd, and C. A. Caldwell. 2012. Evidence of recent
climate change within the historic range of Rio Grande Cutthroat
Trout: implications for management and future persistence. Transactions of the American Fisheries Society 141:1045–1059.
Zeigler, M. P., A. S. Todd, and C. A. Caldwell. 2013b. Water temperature and baseflow discharge of streams throughout the range of Rio
Grande Cutthroat Trout in Colorado and New Mexico–2010 and
2011. U.S. Geological Survey Open-File Report 2013-1051.
SUPPORTING INFORMATION
Additional supplemental material may be found online
in the Supporting Information section at the end of the
article.
Appendix 1: Modeling Future Stream Temperatures and Flow
Evaluating the future persistence of Rio Grande Cutthroat Trout (RGCT) requires the prediction of temperature and flow conditions in a changing climate (Wenger
et al. 2011). Methods used to predict future environmental
conditions that serve as inputs into our Bayesian network
(BN) model are detailed here.
�840
ZEIGLER ET AL.
Stream Temperature Modeling
To model predictions of the maximum 30-d average
temperature (M30AT) and maximum weekly maximum
temperature (MWMT) required for the BN, we chose statistical modeling techniques because they can be easily
applied to the large region across which predictions were
needed. Three different statistical methods were compared
for their ability to accurately predict M30AT and MWMT
across the current range of RGCT. Predictions were made
for 838 individual stream segments that made up the 121
RGCT populations, and the predictions were then averaged
across the stream segments occupied by each population.
Spatially referenced stream temperature records were
obtained from a large monitoring program focused on
RGCT (Zeigler et al. 2013b). Additional data were collected
from university researchers (Harig and Fausch, unpublished
data) and state and federal agencies (New Mexico Environment Department, Colorado Parks and Wildlife, and U.S.
Forest Service [Carson and Santa Fe National Forests]) to
increase the temporal and spatial extent of stream temperature data. We also included temperature records from nearby
high-elevation streams just outside of the Rio Grande basin
because there were few temperature records from highelevation streams within the basin. The resulting database
comprised 544 unique stream temperature records spanning
the years 2000–2013. The M30AT and MWMT were calculated for each stream temperature record.
Complex combinations of atmospheric conditions, topography, stream discharge, and interactions with the streambed
influence stream temperature by altering heat gains and losses
(Caissie 2006). We selected five covariates based on previous
stream temperature models (Isaak et al. 2010; Roberts et al.
2013) to predict stream temperatures across the current range
of RGCT. Two covariates were based on air temperature (°C;
air_M30AT and air_MWMT), two were based on geomorphic
attributes (elevation and drainage area), and one was based on
landscape position (aspect; Table A.1.1).
To estimate air temperature, data were obtained from
Snow Telemetry (SNOTEL) sites (U.S. Department of
Agriculture, Natural Resources Conservation Service;
http://www.wcc.nrcs.usda.gov/snow) in New Mexico and
southern Colorado that had continuous data from 2000 to
2013. The sites selected are located at a range of high elevations (2,560–3,487 m) similar to those of the remaining
RGCT populations. The same acute (air_MWMT) and
chronic (air_M30AT) temperature metrics as used for
stream temperatures were calculated from daily air temperature records at each site. The closest SNOTEL site to each
stream temperature site was determined using Euclidian distance. Yearly air temperature metrics were then matched to
each thermograph record for each year of record. A lapse
rate of 6.5°C per 1,000 m was used to adjust for temperature differences between the elevation of the thermograph
and that of the SNOTEL station (Running et al. 1987).
The remaining covariates for each thermograph record
were extracted from databases using ArcGIS 10 (ArcMap
10.2; ESRI, Redlands, California). Cumulative drainage area
(drng_area) was calculated as the total cumulative watershed
area contributing to an individual stream segment using the
Spatial Tools for the Analysis of River Systems (STARS)
toolset in ArcMap 10.2 (Peterson and Ver Hoef 2014). Elevation at the location of the thermograph (elev_pt), the mean
elevation of the stream segment where the thermograph was
located (elev_seg), and the aspect at the location of the thermograph were extracted from a 30-m-resolution National
Elevation Dataset using GIS tools.
To assess future stream temperature conditions for
each RGCT population, three statistical modeling techniques were compared for their ability to predict M30AT
and MWMT from the selected covariates. For each temperature metric, we developed a set of seven a priori
models using different combinations of the five covariates
(Table A.1.1). The same seven models were evaluated for
M30AT and MWMT except that the air temperature
covariate used was matched with the stream temperature
metric modeled.
The simplest statistical technique, multiple linear regression (MLR), was used to determine the most parsimonious
model for prediction of both stream temperature metrics.
The most parsimonious model for each stream temperature
metric was selected based on information-theoretic methods
(Anderson 2008) using Akaike's information criterion corrected for small sample size (AICc; Table A.1.1). Residual
plots were examined to ensure that each model fit the
assumptions of linear regression. The variance inflation factor (VIF) was examined for each model to assess potential
problems with multicollinearity, with VIF values greater
than 10 indicating multicollinearity (Helsel and Hirsch
1992). The subset of covariates identified in the most parsimonious model for each stream temperature metric was
used in subsequent analyses to identify the best statistical
technique to predict stream temperatures.
Two other statistical techniques were selected for their
ability to account for spatial autocorrelation among the
data. Failing to account for spatial autocorrelation can produce biased parameter estimates and autocorrelated error
structures when modeling spatially dependent data (Peterson et al. 2007). One spatial statistical technique used was
universal kriging (Stein and Corsten 1991), with spatial
dependence between observations assessed using Euclidean
distance. The universal kriging method assumes that most
of the variability in the data is associated with spatial
dependency among the data (Fortin and Dale 2005). To
make predictions using the universal kriging method, we
used the covariates from the top model identified in the
MLR analysis along with the X- and Y-coordinates of each
site to create variograms and determine the best fit given the
spatial structure of the residuals.
�PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
The second spatial statistical modeling technique was
based on recently developed spatial flow-routed models,
which use along-channel distance to model autocovariance
(Peterson and Ver Hoef 2010; Isaak et al. 2014). We used a
mixed-model autocovariance structure with “tail-up” and
“tail-down” covariance components based on hydrologic
distances and a covariance component based on Euclidean
distance. Tail-up covariances are based on hydrologic distances between flow-connected sites, and tail-down covariances allow spatial correlation between flow-unconnected
sites. The STARS tool box for ArcGIS 10.2 (Peterson and
Ver Hoef 2014) was used to generate the spatial stream network (SSN) needed to implement the model. The SSN was
then imported into R version 2.15.1 (R Development Core
Team 2016), and models were run using the Spatial Stream
Network package (Ver Hoef et al. 2014) for R. We used
exponential tail-up, exponential tail-down, and exponential
Euclidean autocovariance components to model both
M30AT and MWMT predictions.
The best predictive statistical technique was determined
by using the leave-one-out cross-validation method to calculate the average root mean square predictive error
(RMSPE; Power 1993). Once identified, predictions for
three time periods were made for each of the 838 individual stream segments comprising the 121 RGCT populations. Time periods spanned 2010–2019 (current time
period), 2040–2049 (2040s), and 2080–2089 (2080s). All
stream temperature analyses were conducted in R version
2.15.1 (R Development Core Team 2016).
Base Flow Modeling
Little information on current base flow conditions in
streams occupied by RGCT is available because most stream
gauges in the subspecies’ current range are located on low-elevation stream segments downstream of RGCT populations.
The limited data on base flows for RGCT populations (Zeigler
et al. 2013b) were available at only a few locations for a given
stream fragment and may not accurately represent conditions
across the entire reach. To model summer base flow for each
occupied stream segment, we assembled continuous streamflow data from 11 U.S. Geological Survey (USGS) stream
gauges located in New Mexico that were determined to be relatively free of anthropogenic influences (i.e., diversions or
impoundments; Falcone et al. 2010). No USGS stream gauge
data were available for the Colorado portion of the RGCT's
range, so we obtained continuous streamflow data from 18
gauges operated by the Colorado Division of Water Resources
(http://www.dwr.state.co.us/Surfacewater/default.aspx), which
were also considered to be relatively free of anthropogenic
influences. Most stream gauges used were not located within
RGCT-occupied stream reaches but represented the best available streamflow data within the subspecies’ current range.
We used these data to develop a linear regression to
predict base flow conditions from drainage area for each
841
stream segment in the National Hydrography Dataset
Plus version 2 (NHDPlus v2; 1:100,000 scale; www.horizonsystems.com/nhdplus/NHDPlusV2_home.php) stream data
layer. First, the mean 30-d minimum discharge (M30MD)
was calculated for each gauge over the period 1990–2012.
We then used simple linear regression to model M30MD as
a function of drainage area (km2) at the stream gauge (R2 =
0.57). Similar to stream temperature predictions, this model
was then used to predict M30MD for the 838 individual
stream segments covering habitat occupied by the 121
RGCT populations. The final base flow prediction for each
population was then determined by averaging across all
individual stream segments that comprised each population.
Future Climate Projections
To assess the effects of future climate changes on the
persistence of RGCT populations, we used dynamically
downscaled high-resolution output (15- × 15-km grid cells)
from a regional climate model (RegCM3) and the Max
Planck Institute for Meteorology (MPI) ECHAM5 general
circulation model (GCM; Hostetler et al. 2011). Atmospheric conditions for the RegCM3 simulations were
based on the Intergovernmental Panel on Climate Change
(IPCC) Fourth Assessment Report (AR4) A2 emission
scenario (IPCC 2007), a high-emission scenario, which is
also similar to the RCP8.5 scenario in the IPCC AR5
(IPCC 2013). The MPI ECHAM5 GCM is sensitive to a
doubling of atmospheric CO2, equating to a 2–4°C air
temperature increase over North America (Hostetler et al.
2011). To examine the effects of future climate on the persistence of RGCT, the model outputs for air temperature
and total runoff were averaged over the same three time
periods: current, 2040s, and 2080s.
Air temperature and runoff outputs from the dynamically
downscaled climate projections were then used to predict
future stream temperature and base flow. The projected surface air temperature from the grid cell where each SNOTEL
site was located and extracted and the percentage change in
modeled air temperature from the average 2005–2010 values
to the averages for the current, 2040s, and 2080s time periods
were calculated for each SNOTEL site to avoid model bias.
Total runoff outputs from the dynamically downscaled climate projections were used along with ArcGIS 10.2 hydrology tools to estimate summer base flow at each stream gauge
site, with percentage differences calculated similar to those
for surface air temperature. The observed values for both air
temperatures and summer base flow were then multiplied by
the percentage changes to estimate values for the current,
2040s, and 2080s time periods. Estimated air temperatures
for each time period were used in our analyses. The estimated
percentage change in summer base flow at each gauge was
multiplied by predicted base flows from the summer base
flow model and then was used to estimate the future discharge for each RGCT stream segment.
�842
ZEIGLER ET AL.
Stream Temperature and Flow Models
Model selection.— Air temperature, segment elevation,
cumulative drainage area, and aspect were included in the
most parsimonious model (based on AICc) for predicting
both M30AT and MWMT (Table A.1.1). The SSN statistical technique was selected as the best for modeling both
M30AT and MWMT based on RMSPE (Figure A.1.1).
Prediction error for the SSN model for M30AT (0.9°C)
was lower than those of both the universal kriging model
(1.0°C) and the MLR model (1.9°C). The prediction error
for the SSN model for MWMT (1.6°C) was also lower
than that of the universal kriging model (1.8°C) and was
half that of the MLR model (3.2°C). Final models for
both temperature metrics tended to overestimate lower
temperatures and underestimate higher temperatures, but
this effect was very subtle for M30AT (Figure A.1.1).
Future thermal conditions.— Small increases in M30AT
and MWMT were predicted to occur in RGCT-occupied
streams by 2080. Across geographic management units,
M30AT was predicted to increase by only 0.66–0.72°C, and
MWMT was predicted to increase by 0.45–0.47°C (Table
A.1.2). The majority of RGCT populations (n = 118) were
predicted to occupy streams with M30AT suitable for
recruitment and growth (9.1–19.0°C) in 2080 (Figure A.1.2).
No populations were predicted to occupy habitats that would
become completely thermally unsuitable (MWMT ≥ 25.0°C)
for RGCT by 2080, although three populations were predicted to lose occupied habitat due to increasing MWMT.
FIGURE A.1.1. Observed versus predicted values from the three statistical models (multiple linear regression [MLR], universal kriging [UK], and
spatial stream network [SSN]) that were used to predict two stream temperature metrics: maximum 30-d average temperature (M30AT; °C) and
maximum weekly maximum temperature (MWMT). The black line shows the one-to-one relationship (Y = X), and the root mean square prediction
error (RMSPE; °C) is given for each model.
�843
PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
TABLE A.1.1. Results of top model covariate subset selection for the rangewide stream temperature model based on Akaike's information criterion
corrected for small sample size (AICc). Selected models are presented in bold italics. The maximum 30-d average temperature (M30AT) and the maximum weekly maximum temperature (MWMT) were predicted from ambient M30AT (air), stream segment elevation (elev_seg) or thermograph elevation (elev_pt), drainage area (drng_area), and aspect.
Covariate
Model
1
2
1
3
2
4
5
6
7
air_M30AT
air_M30AT
air_M30AT
air_M30AT
air_M30AT
air_M30AT
air_M30AT
elev_seg
elev_seg
elev_pt
elev_pt
elev_seg
elev_pt
1
2
3
4
5
6
7
air_MWMT
air_MWMT
air_MWMT
air_MWMT
air_MWMT
air_MWMT
air_MWMT
elev_seg
elev_pt
elev_seg
elev_pt
elev_seg
elev_pt
3
4
AICc
drng_area
drng_area
drng_area
drng_area
aspect
2,287
2,297
2,316
2,321
2,346
2,383
2,704
drng_area
drng_area
drng_area
drng_area
aspect
aspect
M30AT
aspect
MWMT
The resulting loss of thermally suitable habitat amounted to
only 2.2 km of the 1,145 km currently occupied. Nevertheless, 11 RGCT populations (9%) occupied streams that were
predicted to become warm enough to reduce survival
(MWMT > 21°C) but not warm enough for the entire occupied stream length to become thermally unsuitable.
Future flow conditions.— Due to the lack of stream
gauges located on high-elevation streams throughout the
current range of RGCT, flow conditions within occupied
streams were likely overestimated in comparison to actual
base flow discharges (see Zeigler et al. 2013b for measured conditions). Comparison of summer base flow predictions from our model to actual conditions is difficult
due to having only point estimates from these small,
2,827
2,837
2,849
2,857
2,913
2,934
2,997
high-elevation streams. Although summer base flow discharges were likely overestimated, we still predicted
decreases in summer base flow discharges from the current time period to the 2080s (Figure A.1.3). The greatest
shift occurred in the number of populations with summer
base flow in the moderate-discharge category (0.0291–
0.1799 m3/s) decreasing to the low-discharge category
(0.0017–0.0291 m3/s). In the current time period, 95 populations (78%) were in the moderate-discharge category,
but this number decreased to 77 populations (64%) by
the 2080s. This decrease in discharge resulted in an
increase from 24 populations (20%) in the low-discharge
category during the current time period to 43 populations
(36%) by the 2080s.
�844
ZEIGLER ET AL.
FIGURE A.1.2. Histogram of predicted maximum 30-d average temperature (M30AT; °C) and maximum weekly maximum temperature (MWMT)
for Rio Grande Cutthroat Trout populations in the current time period (black bars) and the 2080s (gray bars).
�PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
845
TABLE A.1.2. Mean maximum 30-d average temperature (M30AT),
maximum weekly maximum temperature (MWMT), and summer base
flow discharge for the current time period and the 2080s, and difference
between them, for each geographic management unit (GMU) of Rio
Grande Cutthroat Trout.
GMU
Lower Rio Grande
Rio Grande
Headwaters
Pecos
Canadian
Current
2080s
M30AT (°C)
11.75
12.41
12.09
12.82
11.39
12.08
13.61
14.32
MWMT (°C)
16.37
16.82
16.72
17.17
Change by
2080s
0.66
0.72
0.69
0.71
Lower Rio Grande
0.45
Rio Grande
0.46
Headwaters
Pecos
15.23
15.68
0.46
Canadian
19.02
19.49
0.47
Summer base flow discharge (m3/s)
Lower Rio Grande
0.0401 0.0347
−0.0054
Rio Grande
0.0573 0.0420
−0.0153
Headwaters
Pecos
0.0378 0.0330
−0.0048
Canadian
0.0599 0.0599
0.0000
FIGURE A.1.3. Histogram of summer base flow discharge categories
for Rio Grande Cutthroat Trout populations in the current time period
(black bars) and the 2080s (gray bars).
�846
ZEIGLER ET AL.
Appendix 2: Information Sources Used to Populate the Bayesian Network Model
TABLE A.2.1. Sources of information used to define Bayesian network model nodes for predicting persistence in Rio Grande Cutthroat Trout populations. Parent nodes and child nodes (shown in italics) are grouped by sub-directed acyclic graph and are arranged alphabetically (Ne = effective population size; N = total population size; M30AT = maximum 30-d average temperature; MWMT = maximum weekly maximum temperature; Mc =
Myxobolus cerebralis). Sources of evidence used to define states or values for parent and child nodes (designated S) as well as conditional probabilities
for child nodes (P) are given. Key publications are listed when either peer-reviewed or non-peer-reviewed literature was used to define node values,
states, or conditional probabilities.
Node
Expert
knowledge
Field
observations
or data
Models
fit to
data
S
S
Peer
reviewed
literature
Non-peerreviewed
literature
Key publications
Genetic risks
Adult Population
Estimate
Barrier Presence
S
S
Ne/N Ratio
S
S
Proximity of
Hybridizing
Source
Population
Time Period
Effective
Population Size
S
S
S
S
S
S
S
S
Inbreeding Risk
P
S
Invasion and
Hybridization
Risk
P
S, P
Anthropogenic
Influence
Barrier Presence
S
Alves et al. 2008
Novinger and Rahel 2003; Fausch
et al. 2006, 2009
Rieman and Allendorf 2001; Jensen
et al. 2005; Belmar-Lucero et al.
2012; Roberts et al. 2013
Alves et al. 2008; Muhlfeld et al.
2009a, 2009b, 2017; McKelvey et al.
2016; Young et al. 2016, 2017
NA
Frankham 2005; Allendorf and
Luikart 2007; Alves et al. 2008;
Roberts et al. 2013
Frankham 2005; Allendorf and
Luikart 2007; Jamieson and
Allendorf 2012; Whiteley et al. 2013
Pritchard and Cowley 2006; Muhlfeld
et al. 2009b, 2017; Young et al. 2016,
2017
Population demographics
Base Flow
Discharge
Demographic
Support
S
S
S
Novinger and Rahel 2003; Fausch
et al. 2006, 2009
S
S
S
M30AT
S
MWMT
S
Nonnative
Control
Occupied Stream
Length
S
S
S
Rieman and Dunham 2000; Alves
et al. 2008; Peterson et al. 2008b;
Fausch et al. 2009
Bear et al. 2007; Coleman and Fausch
2007a, 2007b; Zeigler et al. 2013a
Roberts et al. 2013; Zeigler et al.
2013a
Shepard et al. 2002, 2014; Peterson
et al. 2004, 2008a
�847
PREDICTING RIO GRANDE CUTTHROAT TROUT PERSISTENCE
TABLE A.2.1. Continued.
Node
Proximity of
Competitor
Source
Population
Proximity of Mc
Source
Time Period
Adult
Demographics
Competitor
Invasion and
Biotic Influence
Risk
Invasion Vortex
Population
Growth Rate
Potential Fry-toAge-2 Survival
Potential
Habitat
Suitability
Realized Fry-toAge-2 Survival
Realized Habitat
Suitability
Stream Wetted
Width
Mc Infection
Risk
Base Flow
Discharge
Stream Drying
Refugia
Availability
Evidence of
Intermittency
Occupied Stream
Length
Population
Connectivity
Time Period
Wildfire/Debris
Flow Risk
Expert
knowledge
Field
observations
or data
Models
fit to
data
Peer
reviewed
literature
S
Non-peerreviewed
literature
S
S
Key publications
Young 1994; Peterson and Fausch
2003; Alves et al. 2008
Alves et al. 2008
S, P
S
P
S, P
S, P
S, P
S
S
P
S, P
P
S, P
S, P
S, P
P
S
P
P
S
NA
Peterson et al. 2004, 2008b, 2014;
Zeigler et al. 2013a
Peterson et al. 2004; McHugh and
Budy 2005, 2006
Shepard et al. 2002; Peterson et al.
2004, 2008b
Stapp and Hayward 2002; Peterson
et al. 2008a, 2010
Budy et al. 2012; Zeigler et al. 2013a
S
S, P
Harig and Fausch 2002; Young et al.
2005; Coleman and Fausch 2007a,
2007b; Peterson et al. 2013
Peterson et al. 2004, 2008a; McHugh
and Budy 2005, 2006; Nehring et al.
2014
Armour et al. 1991; Kershner et al.
2004
Harig and Fausch 2002; Alves et al.
2008
Thompson et al. 1999; Schisler and
Bergersen 2002; Wagner et al. 2002;
de la Hoz Franco and Budy 2005;
Nehring et al. 2014
Stochastic disturbance risks
S
S
S
S
S
S
Japhet et al. 2007; Patten et al. 2007;
White and Rahel 2008
Chapin et al. 2014
S
S
S
S
S
Dunham et al. 1997; Hilderbrand
2003; Fausch et al. 2006, 2009
NA
Cannon et al. 2010; Miller and
Bassett 2013
�848
ZEIGLER ET AL.
TABLE A.2.1. Continued.
Node
Expert
knowledge
Field
observations
or data
Models
fit to
data
Peer
reviewed
literature
Stream Drying
Risk
Stochastic
Buffering
Stochastic
Disturbance
Risks
S, P
S
P
S, P
P
S
Population
Persistence
P
Output
S
Non-peerreviewed
literature
Key publications
Seager et al. 2007, 2013; Kovach et al.
2016
Rieman and Clayton 1997; Roberts
et al. 2013
Dunham et al. 2007; Roberts et al.
2013
O'Grady et al. 2006
�
Dublin Core
The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/.
Title
A name given to the resource
Journal Articles
Description
An account of the resource
CPW peer-reviewed journal publications
Text
A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text.
Dublin Core
The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/.
Title
A name given to the resource
Predicting persistence of Rio Grande Cutthroat Trout populations in an uncertain future
Description
An account of the resource
<span>The Rio Grande Cutthroat Trout </span><i>Oncorhynchus clarkii virginalis</i><span> (RGCT) occupies just 12% of its ancestral range. As the southernmost subspecies of Cutthroat Trout, we expect a warming climate to bring additional stressors to RGCT populations, such as increased stream temperatures, reduced streamflows, and increased incidence of wildfire. We developed a Bayesian network (BN) model using site-specific data, empirical research, and expert knowledge to estimate the probability of persistence for each of the 121 remaining RGCT conservation populations and to rank the severity of the threats they face. These inputs quantified the genetic risks (e.g., inbreeding risk and hybridization risk), population demographics (disease risk, habitat suitability, and survival), and probability of stochastic disturbances (stream drying risk and wildfire risk) in an uncertain future. We also created stream temperature and base flow discharge models coupled with regionally downscaled climate projections to predict future abiotic conditions at short-term (2040s) and long-term (2080s) time horizons. In the absence of active management, we predicted a decrease in the average probability of population persistence from 0.53 (current) to 0.31 (2040s) and 0.26 (2080s). Only 11% of these populations were predicted to have a greater than 75% chance of persisting to the 2080s. Threat of invasion by nonnative trout had the strongest effect on population persistence. Of the 78 populations that are already invaded or lacking complete barriers, 60% were estimated to be extirpated by 2080 and the remainder averaged only a 10% chance of persistence. In contrast, the effects of increased stream temperatures were predicted to affect the future persistence of only 9% of the 121 RGCT populations remaining, as most have been restricted to high-elevation habitats that are cold enough to buffer against some stream warming. Our BN model provides a framework for evaluating threats and will be useful to guide management actions that are likely to provide the most benefit for long-term conservation.</span>
Bibliographic Citation
A bibliographic reference for the resource. Recommended practice is to include sufficient bibliographic detail to identify the resource as unambiguously as possible.
<a href="https://doi.org/10.1002/nafm.10320" target="_blank" rel="noreferrer noopener">https://doi.org/10.1002/nafm.10320</a>
Creator
An entity primarily responsible for making the resource
Zeigler, Matthew P.
Rogers, Kevin B.
Roberts, James J.
Todd, Andrew S.
Fausch, Kurt D.
Subject
The topic of the resource
Rio Grande Cutthroat Trout
<em>Oncorhynchus clarkii virginalis</em>
Bayesian network model
Extent
The size or duration of the resource.
30 pages
Date Created
Date of creation of the resource.
2019-10
Rights
Information about rights held in and over the resource
<a href="http://rightsstatements.org/vocab/InC-NC/1.0/" target="_blank" rel="noreferrer noopener">In Copyright - Non-Commercial Use Permitted</a>
Format
The file format, physical medium, or dimensions of the resource
application/pdf
Language
A language of the resource
English
Is Part Of
A related resource in which the described resource is physically or logically included.
North American Journal of Fisheries Management
Type
The nature or genre of the resource
Article