Introduction

The prospect of ecological transformation is a growing reality that is challenging how people perceive and manage landscapes (Schuurman et al. 2022). Ecological transformation can occur slowly over time from incrementally warming temperatures or long-term drought (Lake 2011) or immediately due to wildfire (Jager et al. 2021), biological invasion (Ricciardi et al. 2013), or other phenomena that can rapidly transform entire landscapes (Crausbay et al. 2022). Because these events have great potential to influence species (in some cases leading to local extirpations), species conservation strategies often include provisions that allow for connectivity or access to diverse conditions within landscapes to reduce threats from environmental variability or disturbance (Pickett and Thompson 1978; Schindler et al. 2015).

When the capacity of species to move across landscapes in response to changing environmental conditions is limited, it may be necessary to consider more active conservation measures or interventions (Cole et al. 2022). Among different forms of conservation, reintroductions of species to previously occupied habitats within their natural ranges have been attempted for many cases (Jachowski et al. 2016). Reintroductions may be especially important for groups of species with limited movement capacity, such as freshwater fish, where movement is constrained by hydrological connectivity (Fullerton et al. 2010; Dunham et al. 2016; Flitcroft et al. 2019). Although reintroductions offer much promise in these cases, they have proven variably successful in practice (Fischer and Lindenmayer 2000; Cochran-Biederman et al. 2015). These variable outcomes have motivated calls for more rigorous approaches for the design and evaluation of reintroduction decisions (Jachowski et al. 2016; Lamothe et al. 2019).

In this paper, we consider the conservation of threatened freshwater fish (bull trout, Salvelinus confluentus) in a landscape that is rapidly transforming through the influence of nonnative species, drought, and associated wildfires (Falke et al. 2015; Kovach et al. 2019). Bull trout is a coldwater specialist (Dunham et al. 2003; Isaak et al. 2015; Isaak and Young 2023) and often is restricted to occupying only the coldest available streams in the presence of nonnative species (Benjamin et al. 2016). Bull trout have been extirpated from much of their historical habitat (USFWS 2015a, b), which provides potential opportunities for reintroductions. We evaluated reintroduction opportunities in the upper Klamath River basin using a structured decision-making approach (Gregory et al. 2012; Conroy and Peterson 2013) that engaged local stakeholders in co-production of models specifically adapted to bull trout and conditions that influence their populations in each potential reintroduction location (Brignon et al. 2018; Benjamin et al. 2019). We also considered the consequences of reintroductions on populations of bull trout that could be used to source reintroduced fish, i.e., “donor populations” (Dunham et al. 2011).

In addition to addressing the question of reintroductions, we also addressed impacts of wildfire disturbance on donor populations of bull trout. This assessment was motivated by a massive wildfire that occurred during the study that potentially impacted several populations of bull trout. Because dramatic reductions in local populations of bull trout were anticipated, we considered supplementation (i.e., adding fish to existing populations affected by the fire) as a new alternative. With the reintroduction model formulated, we were able to consider real-time changes to the system during the study. Consequently, this study exemplifies the value of applying co-produced decision support models and adapting them to re-evaluate a system following a major transformational event. As such events become increasingly likely for many species and ecosystems, examples for evaluating them in the context of conservation decision alternatives are greatly needed (Magness et al. 2022).

Methods

Study area

The upper Klamath River basin is at the southern margin of the current range of bull trout (Dunham et al. 2008). Bull trout were once widely distributed throughout the upper Klamath River basin (Buchanan et al. 1997). Eight known populations of bull trout remain (Fig. 1). Declines in abundance and distribution have been attributed to habitat degradation, fragmentation, and invasion of nonnative species (USFWS 2015a, b). The remaining bull trout populations are primarily restricted to headwater habitats with limited migratory life history expression. Efforts by tribal, state, and federal government agencies are ongoing to improve habitat conditions (Buktenica et al. 2018) and control nonnative fish (especially brook trout, S. fontinalis; Buktenica et al. 2013; Banish et al. 2019).

Fig. 1
figure 1

Klamath River basin showing stakeholder-identified recipient streams for reintroduction of bull trout, streams bull trout currently occupy, perimeter of the three core areas, movement barriers in Sun Creek, and the extent of the Bootleg Fire

Full size image

The upper Klamath River basin is hydrographically disconnected from other basins supporting bull trout and recognized as one of six distinctive recovery units distributed across the species’ range in the coterminous United States (USFWS 2015a, b). Within the extent of the Klamath Recovery Unit for bull trout, major tributaries within this basin are further classified into three core areas, including Upper Klamath Lake, Sycan River, and the Upper Sprague River (USFWS 2015b; Fig. 1). Streams currently occupied by bull trout in these systems drain largely forested watersheds and are generally upstream of water diversions and groundwater pumping, which represents a growing influence on the hydrology of downstream ecosystems (Gannett et al. 2007). Precipitation and annual rates of groundwater discharge are generally higher in the Upper Klamath Lake Core Area, relative to the Sycan River and Upper Sprague River Core Areas (Gannett et al. 2007). This hydrological variability translates into more abundant and stable flows, as well as cooler temperatures in streams in the Upper Klamath Lake Core Area (Gannett et al. 2007; Benjamin et al. 2016). In July 2021, during the study, a massive wildfire known as the Bootleg Fire (Ager et al. 2022) occurred within the Sycan River and Upper Sprague River Core Areas, burning nearly 170,00 hectares, including basins with streams occupied by bull trout at the time.

Approach to co-production

Our approach to developing models to support evaluation of alternatives for a bull trout reintroduction followed commonly prescribed steps in stakeholder-driven decision-making processes (Gregory et al. 2012; Conroy and Peterson 2013). We refer to this as co-production because stakeholders were engaged regularly over a period of 10 months to develop each step of the decision-making process (Cooke et al. 2021). These steps included defining the decision situation, identifying fundamental objectives, specifying conservation alternatives, drafting an influence diagram, and interpreting results from a quantitative assessment tool created from the influence diagram. In contrast to some past examples (Dunham et al. 2011; Brignon et al. 2018; Benjamin et al. 2017, 2019), new situations emerged during the process due to refined specification of decision alternatives and occurrence of the Bootleg Fire. These led to modified objectives and updated assessments of conservation alternatives. For logistical and safety reasons (the COVID-19 pandemic), stakeholders’ decision support efforts involved virtual meetings held monthly from June 2021 to May 2022. Each meeting was 2–3 hours in length. Participating stakeholders included individuals from the Oregon Department of Fish and Wildlife, Oregon Department of Forestry, Trout Unlimited, The Nature Conservancy, Green Diamond Resource Company, U.S. Forest Service (Fremont-Winema National Forest), U.S. National Park Service (Crater Lake National Park), and U.S. Fish and Wildlife Service.

Decision situation and fundamental objectives

The decision situation, as described by stakeholders, was that “bull trout distribution and abundance is severely reduced; therefore, we need to explore options to establish one or more additional population(s).” With this decision situation articulated, stakeholders then identified two fundamental objectives: (1) maximize the number of adult bull trout in at least one stream selected for a reintroduction (the recipient stream) and (2) maximize the persistence of populations that could act as donors for a reintroduction. Stakeholders defined persistence as 100 adult bull trout. Recipient streams for reintroduction were identified by stakeholders based on two criteria. First, streams were included that ranked highly in reintroduction potential in a previous feasibility assessment for the Klamath River basin (USFWS 2020). This assessment was based on a scoring system following Dunham et al. (2011) and considered elements such as historical and contemporary presence, habitat suitability, and threats to bull trout (e.g., nonnative fish), as well as other ecological, economic, and social factors. Second, stakeholders considered their confidence or belief in reintroduction success within a stream. Stakeholders acknowledged that options may occur outside of core areas for potential reintroduction (e.g., assisted migration) but chose not to consider those. The time frames to realize two fundamental objectives were specified as 10 and 30 years. The 10-year time frame was identified because it is approximately when subsequent management decisions would be made, whereas 30 years allowed multiple generations of bull trout to manifest responses over the long term.

During the initial process in summer 2021, the Bootleg Fire burned upland and riparian habitat in most of the potential donor populations in the Upper Sprague River Core Area (Fig. 1). Following the fire and apparent reduction in bull trout abundance observed in exploratory surveys, stakeholders believed that the abundance of some bull trout populations was substantially reduced. Thus, stakeholders added a third fundamental objective: maximize adult bull trout following supplementation alternatives under a range of reductions in abundance of bull trout owing to the Bootleg Fire.

Conservation decision alternatives

Stakeholders identified five conservation decision alternatives to consider for consequences in recipient streams and donor populations (Fig. 1; Table 1). (1) Baseline, in which no fish would be added to recipient streams or taken from donor populations. (2) Artificial propagation by artificially spawning adults in captivity and releasing the progeny as fry. For a recipient stream, 10,000 fry would be released for the first five consecutive years. We assumed adults could be captured from a donor population and retained in the hatchery, and would reproduce year after year. Hence, 60 adults would need to be removed from a donor population and held in captivity. Because some mortality or variation in fecundity is likely, removal of adults from a donor stream would occur for 2 years. (3) Assisted rearing, which was defined as the collection of bull trout fry from a donor population, rearing the fry to juveniles in a hatchery or natural habitat (e.g., isolated pond or stream), and then releasing the juveniles into a recipient stream. Based on donor population sizes and ability to capture fry, stakeholders identified 500 fry could be captured and subsequently 500 juveniles released into a recipient stream for the first 2 years. (4) Translocation of 100 fry, 100 resident subadults, and 25 resident adults from a donor population to a recipient stream for two consecutive years. Life stages and individual numbers were identified to be achievable based on population size of potential donor populations and capture efficiency. (5) Hatch boxes, in which remote-site incubators (Shepard et al. 2021) would be used to release fry in recipient streams for five consecutive years. We assumed each incubator could hold 1000 eggs (Table S1; Howell 2021), for a total of 5000 eggs needed from a donor stock. For the hatch box alternative, eggs would be collected by manually spawning captured adults at stream-side, and then adults would be returned to the stream for each of the five years. Different penalties or discounts on survival and growth for each alternative were applied to account for release into the recipient stream, transportation, survival in captivity, and hatchery production (Table S1). Penalties were applied after individuals were added to a recipient stream. For example, 5000 eggs are added under the hatch box alternative, and 4000 fry emerge following a 20% penalty. Penalties were not applied to source populations.

Table 1 Conservation decision alternatives (strategy), life stage, number of individuals, and the number of consecutive years for bull trout reintroduced into recipient streams and taken from donor populations
Full size table

Model description and inputs

We used a life-stage-based demographic matrix model for bull trout, which was identical across all recipient and donor populations. To provide individual stream-specific context, life history expression and survival rates within the matrix model were adjusted based on stream-specific abiotic and biotic conditions.

Demographic model. Stakeholders agreed to use a demographic matrix model for bull trout that was similar to those used in previous bull trout reintroduction decision support efforts (Brignon et al. 2018; Benjamin et al. 2019). In the Klamath River basin, six life stages were considered to represent resident and migratory life histories. These stages were fry, juveniles, resident subadults, resident adults, migratory subadults, and migratory adults (Fig. 2). We define the migratory life stage as those individuals that move to exploit growth opportunities outside their natal streams and then return to spawn. For each of the recipient and donor populations, we used the same transition matrix (A) and parameter values (Table S1) for bull trout:

$${\varvec{A}}=\left(\begin{array}{cccccc}0& 0& 0& {F}_{4}& 0& {F}_{6}\\ {G}_{1}& {P}_{2}& 0& 0& 0& 0\\ 0& {rG}_{2}& {P}_{3}& 0& 0& 0\\ 0& 0& {G}_{3}& {P}_{4}& 0& 0\\ 0& {mG}_{2}& 0& 0& {P}_{5}& 0\\ 0& 0& 0& 0& {G}_{5}& {P}_{6}\end{array}\right)$$

where Fi is the fecundity of stage i, Gi is the probability of surviving and transitioning to stage i + 1, and Pi is the probability of surviving and staying in stage i. Once a juvenile fish transitioned to the resident (rG2) or migratory (mG2) subadult, we assumed it remained a resident or migratory individual for the remainder of its life; however, for this model, their progeny can express either life history. Fecundity for residents (F4) and migrants (F6) was the product of the number of eggs per female, proportion of females in the population, survival of eggs, and the probability of spawning. A density-dependent relationship was used to calculate the survival and transition of fry to juvenile bull trout (G1) (Benjamin et al. 2019)

$${G}_{1}=DI*\left\{1-exp\left[\frac{-\left({K}_{j}*{habitat}_{i}\right)}{DI*{N}_{ji}}\right]\right\}$$

where DI is the density-independent survival of fry, Kj is the carrying capacity of juveniles per river kilometer of suitable spawning and rearing habitat in stream i (habitati), and Nji is the abundance of juveniles in stream i. Model parameters used were from a previous demographic modeling study in a tributary in the Klamath River basin (Benjamin et al. 2017), and then modified by stakeholders.

Fig. 2
figure 2

Influence diagram for bull trout in the upper Klamath River basin. Dashed lines denote parameters that negatively influence (i.e., discount) life stage survival. Note that impacts of brook and brown trout include species occurrence and discount

Full size image

Stream-specific context. The model described above was used in all streams considered. Stakeholders identified four important threats that varied among streams: available spawning and rearing habitat, connectivity or migratory potential, occurrence of nonnative fish, and low (< 0.03 m3/s) summer discharge (Fig. 2; Table 2).

Table 2 Characteristics for recipient and donor populations including the length of spawning and rearing habitat (km; USFWS 2010), proportion of the bull trout population that could be or is migratory, presence (1) or absence (0) of brook or brown trout (Benjamin et al. 2016; Banish et al. 2019), and the probability of low flows (< 0.03 m3/s) occurring in a given year (Miller et al. 2018)
Full size table

The length (km) of suitable spawning and rearing habitat was included in the density-dependent function within the matrix model to estimate the carrying capacity of juveniles in a stream (Benjamin et al. 2017, 2019). We used habitat length from the U.S. Fish and Wildlife Service (USFWS 2010), which was then modified by stakeholders based on current knowledge of habitat quality and quantity. If no data on spawning and rearing habitat were available in a stream (e.g., Rock Creek), the stream length was considered and modified by stakeholders. Henceforth, we refer to this stream length as “habitat size.”

Migratory bull trout are often larger and more fecund than fish that reside in their natal habitat (Rieman and McIntyre 1993). This is consistent with migratory fish in the Klamath River basin, although the sizes of both migrants and residents is smaller relative to bull trout in other watersheds (Buchanan et al. 1997; Benjamin et al. 2017). Moreover, some streams may be disconnected from downstream migratory destinations. Stakeholders estimated the proportion of juveniles that could become migratory based on connectivity and local knowledge of each stream. In Sun Creek, human-made barriers exist (Fig. 1) that allow for downstream movement but not upstream movement. Adult bull trout are assisted in moving upstream past the lowest barrier set. The proportion of migrants for Sun Creek accounts for this assisted movement above the lower barrier.

Nonnative brook trout and brown trout (Salmo trutta) can have negative effects on bull trout through hybridization, competition, or predation (USFWS 2015a). Stakeholders chose to use penalties when nonnative trout occurred in a stream. When brook trout were present in a stream, a 10% reduction was made to each egg, fry, and juvenile survival of bull trout. We assumed this reduction would account for potential consequences of hybridization, predation, and competition. Brown trout typically occurred in downstream sections of some streams (Benjamin et al. 2016). When present, a 10% reduction in migratory subadult survival would occur to account for predation of this early migratory life stage by larger brown trout. Brook trout have previously been eradicated from Rock Creek (USFWS 2020), Threemile Creek (USFWS 2015b; Banish et al. 2019), and Sun Creek (Buktenica et al. 2013, 2018). Crater Lake National Park plans to eradicate brook trout from Annie Creek in sequential sections beginning in 2023, following an approach similar to Sun Creek in which a fish barrier will be installed, and migratory fish assisted above the barrier. Accordingly, for this model exercise, brook trout were considered absent in these four streams.

Given that drought is becoming more common worldwide in the face of climate change and of particular concern in the Klamath River basin (Dai 2013; Williams et al. 2022), stakeholders were interested in considering potential drought impacts in this assessment. Drought was accounted for by addressing low flows, defined as discharge below 0.03 m3/s. Discharge measures were not available for all streams in the Klamath River basin that bull trout occupy or that stakeholders considered as recipient streams for reintroductions. We estimated the annual probability of low flows occurring in each stream using discharge estimates from Miller et al. (2018). This data set estimated monthly discharge for all stream sections across the contiguous United States from 1950 to 2015 using random forest modeling. We used September flow assuming this would be a representative month of low-flow occurrence. The probability of a low-flow event occurring was estimated from the weighted average by section lengths in each stream. Stream specific probability of low flow was then incorporated into the model using a random binomial distribution to estimate the annual probability of low flow over the 30 years of model simulations. If a low-flow event occurred in a given year, stakeholders believed it would reduce survival of fry by 20%, juvenile by 30%, resident subadult by 40%, and resident adult by 50%. Migratory bull trout would occupy downstream locations during seasonal low flows and were assumed to be buffered from low-flow conditions in natal and juvenile rearing habitat (Rieman and Clayton 1997; Jager et al. 2021).

Stochasticity. Stochasticity was incorporated into the model in two different ways, demographic and environmental. Demographic stochasticity was assumed to account for annual variation in bull trout abundance owing to endogenous (demographic) and exogenous (environmental) factors not considered specifically within the model. Demographic stochasticity was incorporated after population projections for each year (Brignon et al. 2018). We used a binomial distribution to select the number of individuals surviving each year drawn from the number of individuals in life stage i and their survival parameter as the probability.

Environmental stochasticity was considered in terms of rare, extreme disturbances (e.g., high-intensity wildfire, debris flows) that could negatively impact bull trout. We assumed an event like this would have an annual probability of 0.05 using random binomial distribution within the model. If disturbance did occur, the abundance of fry, juveniles, resident subadults, and resident adults was reduced by 50%. Similar to low-flow conditions, we assumed migratory fish would be downstream and would not suffer consequences of stochastic events in natal and juvenile rearing habitat.

Model simulations and sensitivity

Simulations. For recipient streams, all starting values were zero were and adjusted as fish were added. For donor populations, we used a stable age distribution developed for bull trout in the Klamath River basin (Benjamin et al. 2017) and the estimated age 1+ population size of bull trout based on previous information (USFWS 2015b) as starting values. For each recipient stream, all alternatives were considered. For each alternative, reintroduced fish started at year 1 and continued each consecutive year for the duration identified by stakeholders (Table 1). Reintroduction penalties (Brignon et al. 2018) were applied each year of reintroduction.

For donor populations, we considered removal of bull trout from one stream. Stakeholders did discuss the possibility of partial take from multiple streams, donor and recipient streams to occur in the same core area, and unsuitable donor stock owing to low population size (e.g., Dixon Creek). We assumed that modeling removal of bull trout from one stream would evaluate the maximum reduction or worst-case scenario for each of the bull trout donor populations. Along with consequences for removing bull trout from Sun Creek as a whole, stakeholders were also interested in evaluating removal of bull trout from the two barriered sections in Sun Creek separately. This was done as an ad hoc analysis.

The occurrence and severity of the Bootleg Fire during initial meetings with stakeholders motivated them to explore supplementation options for streams within the fire perimeter that bull trout occupied. Supplementation, as used here, represents the addition of fish to populations affected by the fire. The concern was that the fire could have reduced the abundance of bull trout, but the magnitude of reduction was uncertain. As such, we evaluated supplementation options for these streams using the same reintroduction alternatives considered above (Table 1). We used four potential population reductions—25%, 50%, 75%, and 100%—for supplementation in streams affected by the fire. Bull trout life stage and numbers used in the reintroduction alternatives were simulated for each percentage of population reduction. We compared this to no effect of the fire (i.e., baseline) for these streams—bull trout were not added under the baseline scenario. Note that by considering fire in this context, the stakeholders were most focused on addressing population recovery from fire by means of supplementation and not addressing potential impacts of fire to habitat or other environmental factors (Jager et al. 2021).

Sensitivity. Sensitivity analyses were conducted to understand model and decision uncertainty. Sensitivity results are shown at year 10 and year 30 per stakeholder-defined time for short- and long-term evaluation of conservation alternatives. All sensitivity analyses were performed with 1000 simulations.

For model uncertainty, we used a global sensitivity analysis to rank the relative importance of each parameter (Table S1). A Latin hypercube sampling was used to simultaneously vary all parameters (McKay et al. 2000). This sampling approach divides each parameter into N bins, where N is the number of simulations, and a random parameter value is selected within each bin for each parameter. Bin combinations are then randomly assigned, which ensures adequate sampling across the range of each parameter. Random forests (Liaw and Wiener 2002) were then used to rank each parameter by importance to the number of adult bull trout for recipient streams and persistence for donor populations. We did not conduct a global sensitivity analysis for each stream; instead, information across all streams was used to inform parameter ranges. For environmental parameters (i.e., habitat length, migratory potential, nonnative occurrence, and low flows), we used the range of values across all streams (Table 2). Similarly, the range of reintroduction penalties across the alternatives was used. Demographic parameters were the same across all streams and varied by ±50% of the literature value (Table S1). A uniform distribution was used for all parameters.

For decision uncertainty, we used a series of one-way and two-way analyses. One-way sensitivity analyses vary one parameter between minimum and maximum values while keeping other parameters at their literature value. Similarly, two-way sensitivity varies two parameters. Response profiles (one-way) or policy plots (two-way) were created for the conservation alternatives for recipient and donor populations separately (Conroy and Peterson 2013). Response profiles were created for each parameter considered in the model, and policy plots were created for each combination of parameters.

Results

Reintroduction

Releasing 10,000 artificially propagated fry per year for 5 years into Annie Creek produced the largest number of adult bull trout among all simulated conservation alternatives and recipient streams (Fig. 3). We estimated the number of adult bull trout would be approximately 77 in year 10 and 78 in year 30. This was followed by artificially propagating bull trout for release into Rock Creek (68 adults in year 10, 51 in year 30). Annie Creek and Rock Creek were consistently the greatest adult bull trout abundance regardless of reintroduction strategies considered. Hatch boxes and translocation were similar to each other in numbers of adult bull trout for each recipient stream (27–31 adults in year 10, 26–33 in year 30), whereas captive rearing was marginally better than the baseline option (0 adult bull trout).

Fig. 3
figure 3

Mean number of adult bull trout from 1000 simulations for each reintroduction alternative and recipient stream

Full size image

Global sensitivity for reintroductions into recipient streams suggested a shift in the most influential model parameters between year 10 and 30 (Fig. 4). At year 10, the greatest relative importance was the reintroduction penalty, followed by the growth and survival of resident subadult to adult bull trout (G3) and impacts of low flows. At year 30, G3 and low flows had the greatest relative importance, followed by egg survival (G0). The growth and survival of juvenile to resident subadult bull trout (r.G2), the occurrence of brook trout, and habitat size were also important. All other parameters were relatively less important.

Fig. 4
figure 4

Relative importance of parameters based on global sensitivity analysis for recipient (a) and donor (b) streams at years 10 and 30

Full size image

The optimal decision was robust to the range of values considered (Fig. 5). At years 10 and 30, release of 10,000 artificially propagated fry into Annie Creek was consistently the optimal decision. This decision was closely followed by the same release strategy into Rock Creek at year 10 and when parameter values were at the low end of the range at year 30.

Fig. 5
figure 5

Examples of decision sensitivity for reintroduction into recipient streams for year 10 (left) and year 30 (right) of mean values from 1000 simulations. Representative one-way sensitivity response profiles (A) and two-way policy plots (B). Response profiles are limited to the top ten reintroduction alternatives and the baseline decision for reference. For the policy plots, gray represents Annie AP and pink represents NF Sprague AP. AP = artificial propagation, Tr = translocation, Ha = hatch boxes

Full size image

The policy plots suggested that two decisions could be optimal depending on the parameters combined (Fig. 5). Those decisions were the release of artificially propagated fry into Annie Creek or NF Sprague River. For example, at lower values of growth and survival from juvenile to resident subadult (r.G2) and resident subadult to adult (G3), releases into NF Sprague River appeared optimal, whereas at higher values of those parameters, releases into Annie Creek were optimal.

Potential donor

For streams that could be potential donor populations for reintroductions or supplementation, modeled persistence of bull trout varied among streams and over time (Fig. 6). Under baseline, model simulations suggested a consistently high (> 0.9; Sun Creek), low (0; Dixon Creek), or declining (remaining streams) probability of persistence of bull trout.

Fig. 6
figure 6

Mean probability of persistence of bull trout simulated for each donor population and conservation alternative of fish removal for a reintroduction effort

Full size image

The removal of bull trout for each of the various reintroduction strategies reduced persistence of bull trout populations at a faster rate and to lower values relative to baseline (Fig. 6). The hatch box strategy, which removed 5000 eggs/year for 5 years, had the greatest negative effect on the probability of persistence for donor populations, followed by artificial propagation (i.e., removal of 60 adults/year for 2 years). Captive rearing had little impact on the persistence of bull trout in the streams.

Parameters important to the persistence of bull trout in donor populations differed slightly between years 10 and 30 (Fig. 4). In the short term (year 10), parameters with the highest importance for persistence in donor populations were associated with stream-specific context (i.e., spawning and rearing habitat and probability of low flows). In the long term (year 30), persistence was influenced by G0 followed by stream-specific context parameters. Other parameters of importance in the sensitivity analyses were similar between donor and recipient populations (i.e., G3, r.G2). Regardless of the conservation alternatives simulated to remove bull trout, population persistence in Sun Creek appeared to be consistent. However, the discrete sections of Sun Creek separated by barriers had lower persistence under the baseline and range of subadult and adult removal relative to Sun Creek as a whole (Fig. S1).

Supplementation

Supplementation of bull trout in streams affected by the Bootleg Fire had the potential to maintain the populations if they were reduced (Fig. 7). Model simulations suggested that artificial propagation or hatch boxes could keep the populations at or near baseline levels, regardless of the percent of the population reduced by the fire. In contrast, supplementation using translocation or captive rearing may not conserve bull trout, particularly at higher percentages of reduction. Compared with baseline (0% reduction), declines in adult numbers were more frequent if fire impacts reduced the population by 50% or more.

Fig. 7
figure 7

Mean probability of persistence following potential reduction in bull trout population size caused by the Bootleg Fire for year 10 (left panel) and year 30 (right panel). A reduction of 0% is the number of adults under baseline scenario and no supplementation effort was simulated. A reduction of 100% is a complete extirpation. The same life stage and number of individuals as described in Table 1 were added under each reduction scenario

Full size image

Discussion

Although the primary focus of species conservation is management of extant populations, reintroductions to establish additional populations are becoming increasingly important (Jachowski et al. 2016). For species such as bull trout, where extant populations occur in a small fraction of historically occupied habitats, reintroductions can offer important opportunities to establish new populations and make progress toward species recovery (USFWS 2015). Reintroductions of bull trout have been evaluated and implemented across the species’ range (Dunham et al. 2016; Hayes and Banish 2017; Benjamin et al. 2019; Whitesel et al. 2022). Recently, considering the consequences of climate change and associated drought, wildfire, and other disturbances has become increasingly important for population persistence (Kovach et al. 2019; Bell et al. 2021; Isaak et al. 2022). Given the increasing prevalence of climate-related disturbances, reintroductions may reduce vulnerability to these events by virtue of distributing species across a broader landscape (Dunham et al. 2003). Here we discuss each component of our assessment, including the likelihood of persistence of extant or potential donor populations, likely success of reintroduction decision alternatives, and the prospects for supplementation as a measure to ensure persistence of extant populations affected by wildfire.

Persistence of donor populations of bull trout in the upper Klamath River basin was modeled with no removal (baseline) or with removal of individuals as donors to support a potential reintroduction or supplementation effort (Fig. 6). Of the eight populations simulated under baseline, only a single population, Sun Creek, exhibited a high probability (> 0.9) of persistence over the next 30 years. Removal of individuals to serve as donors further reduced the probability of persistence for all populations except for Sun Creek. The overall importance of Sun Creek to persistence of bull trout in the upper Klamath River basin highlights the value of previous investments in recovery of this population, which have included intensive removal of nonnative brook trout, construction of barriers to prevent re-invasion, and habitat restoration to increase the distribution, abundance, and connectivity of bull trout in the system (Buktenica et al. 2013, 2018). Without this effort and the robust population of bull trout produced by it, the upper Klamath River basin would lack viable opportunities for translocation to contribute to recovery of other habitats or populations. Previous assessments of bull trout reintroductions have been able to identify viable donor populations (Dunham et al. 2011; Galloway et al. 2016; Brignon et al. 2018; Mims et al. 2019), but there are many other portions of the species’ range where such opportunities may be limited by highly fragmented, small populations (USFWS 2015a; Howell 2018; Isaak et al. 2022). In these situations, recovery of at least one population or stronghold (Haak and Williams 2012) that can act as a viable donor may provide a first step toward recovery of multiple populations.

Among the potential recipient streams identified by stakeholders for reintroduction of bull trout in the upper Klamath River basin, those with more available habitat consistently ranked as likely to produce greater numbers of bull trout in our simulations. Model simulations indicated that Annie Creek consistently produced the strongest response (higher adult bull trout abundance) over 10- and 30-year time frames, relative to other candidate streams. Annie Creek was among the largest potentially suitable habitats available among those that do not currently support a donor population of bull trout. Habitat size is often associated with the presence of bull trout across landscapes (Rieman and McIntyre 1995; Dunham and Rieman 1999; Isaak et al. 2022), and such habitats offer more space to accommodate disturbances, such as wildfire, drought, or other events that can potentially extirpate populations in smaller habitats (Dunham et al. 2003). In our simulations, habitat size was the most important factor influencing longer-term (30-year) persistence, as indicated by the global sensitivity analysis (Fig. 4), reinforcing the importance of the demographic, as well as environmental resilience of larger habitats (Rieman and McIntyre 1993).

Among all streams simulated, sensitivity analyses indicated that stakeholder-specified influences of drought were influential on adult bull trout abundance at both 10- and 30-year time frames were sensitive to the stakeholders’ definition of drought (flow < 0.03 m3/s). For example, Camp and Corral Creeks were believed to have a high probability of low flow (0.91 and 0.79, Table 2), and simulations indicated very low numbers of adult bull trout produced by any reintroduction alternative. These two streams were also among the smallest in terms of length of suitable habitat available. Similarly sized streams (e.g., Rock and Fort Creeks) that were deemed insensitive to low flows produced much higher numbers of bull trout in most scenarios (Fig. 3). Although persistence of bull trout is often considered in terms of temperature and winter flooding (Wenger et al. 2011; Isaak et al. 2022), influences of low flows are not well understood. A recent analysis of species distributions in western Montana streams identified low flows as important drivers of the distribution of several trout species, but not bull trout (Bell et al. 2021). Across North America, increasing stream size is associated with larger body size in trout, including bull trout (Al-Chokhachy et al. 2022). Low flows can reduce available habitat, which in turn can reduce the size (Al-Chokhachy et al. 2022) and fecundity (Al-Chokhachy and Budy 2008) of adult bull trout. In the upper Klamath River basin, stakeholders assigned relatively low proportions of migratory individuals to potential recipient habitats (0.08–0.48, Table 2) because of concerns with habitat fragmentation, but increasing numbers of larger and more fecund females may increase population size (Dunham et al. 2022).

Among threats to bull trout, stakeholders widely agreed that nonnative trout were a primary concern. In the upper Klamath River basin, brook trout and brown trout are widely established (Benjamin et al. 2016; Table 2). In simulations at both 10- and 30-year timescales, influences of brook trout parameters were within the top 10 out of 25 parameters (Fig. 4). Influences of brown trout were less important, likely due in part to the model focusing their impacts on migratory life stages of bull trout, which were believed to represent only a small fraction of populations simulated herein. In more generic simulations when migratory proportions were higher, model results indicated possible coexistence of bull trout and introduced brook trout (Dunham et al. 2022). Overall, model sensitivities to different parameters indicate where potential management actions may be warranted. For example, in the upper Klamath River basin, both past (Buktenica et al. 2018; Banish et al. 2019) and future efforts mentioned by stakeholders in the basin focused on removal of brook trout to benefit bull trout in donor and potential recipient habitats. Stakeholders will need to evaluate the logistical feasibility of brook trout control or eradication in the streams identified in the current study.

As this study was being conducted in 2021, the Bootleg Fire burned through many streams supporting extant populations of bull trout. Although specific impacts of the fire are not fully known, the stakeholders asked to explore the potential for supplementation as a means of maintaining populations in five streams within the burn perimeter. Each of the four supplementation strategies that we evaluated produced similar outcomes among all five streams across the range of fire impacts assessed, and overall populations declined or stabilized at fairly low numbers (< 50 adults). Even if these populations persisted for a longer period, such low numbers are likely to lead to a substantial loss of genetic variability and inbreeding (Rieman and Allendorf 2001). As with many other species of trout experiencing wildfires in highly fragmented habitats (Dunham et al. 2003; Jager et al. 2021), the Bootleg Fire may have a long-lasting negative effect on bull trout that will be difficult to reverse without addressing the problem of fragmentation as well as other threats such as nonnative trout. These conclusions were based on simulated outcomes. Additional time and field data collection will be needed to better understand the actual impacts of the fire on these streams and the number of bull trout that may have survived in them.

The co-produced model in this study highlights a general pattern of suitable recipient streams for reintroduction and donor populations but requires considerations for managers prior to implementation of a conservation strategy. First, model inputs and results suggest that reintroduction efforts in the Upper Klamath Lake Core Area may be more optimal than the other core areas. Streams in the Upper Klamath Lake Core Area were less likely to exhibit low flows, which may result from relative greater precipitation and the connectivity to groundwater (Gannett et al. 2007; Risley 2019). The prevalence of groundwater discharge confers stable stream flow and cool water temperatures (Power et al. 1999), which could become increasingly important for cold-water-adapted species like bull trout, given regional stream warming projections (Isaak et al. 2015). Although results of this study indicate that a reintroduction effort is more feasible in the Upper Klamath Lake Core Area within the Klamath River recovery unit, bull trout in the Sycan and Upper Sprague Core Areas and additional management resources may be needed in these streams to prevent extirpation.

Second, managers and practitioners may wish to assess additional conservation actions. For example, reintroducing bull trout into a stream occupied by brook trout may lead to hybridization and bull trout displacement (Leary et al. 1993; Rieman et al. 2006). Prior to reintroduction, eradication of brook trout combined with an impediment that prevents recolonization and allows for migratory passage may be warranted (Buktenica et al. 2013; Banish et al. 2019). Moreover, these actions may be considered to enhance the likelihood of persistence in the donor populations.

Third, the logistics of accessing donor fish for each strategy and the effect on those populations are also important considerations. Sun Creek was suggested to be the most optimal source for bull trout; however, additional evaluation of impact on subpopulations in Sun Creek suggests that 50 adult or subadult bull trout may be the maximum number of donor individuals without greatly reducing population persistence in those reaches (Fig. S1). Fifty bull trout is less than the required number of donor fish for some conservation strategies like artificial propagation, but additional removal of donors for reintroduction into multiple streams is expected to further impact donor sources.

If the conservation alternatives modeled here are implemented, monitoring to identify the establishment of reintroduced populations and the consequences on donor populations could help to adaptively refine model inputs for future decision analyses (Jachowski et al. 2016). The sensitivity analysis identified the most important parameters that could be used to focus surveys. For example, the impacts of low flows on bull trout or the survival of resident subadults and adults were identified as highly important to model outputs. Better information for these parameters could be used to update model inputs for future evaluations of conservation actions. Many of the parameters in the model had a high level of uncertainty because they were based on bull trout populations in other watersheds (e.g., Bowerman 2013) or assumptions and beliefs of stakeholders. In the broader context of adaptive management (Lynch et al. 2022), addressing critical uncertainties revealed in this decision analysis application could improve future decisions within the system we studied herein, and perhaps other reintroductions as well (Hayes and Banish 2017).

This unique case study illustrates where reintroduction is likely to be successful and contribute to recovery of climate-sensitive species, such as bull trout. Broad-scale assessments of bull trout have identified factors that drive the species’ vulnerability to climate change (Bell et al. 2021; Isaak et al. 2022). In the state of Oregon, insights from these studies, along with other sources of information, have been assembled into a co-produced prioritization model to identify conservation units where investments in species recovery may be more likely to produce benefits (Brignon et al. 2022). This prioritization effort identified three major basins (Sprague River, Sycan River, upper Klamath Lake, Fig. 1) as ranking fourth, fifth, and 11th, respectively, in terms of their priority among 25 units considered (Brignon et al. 2022). By evaluating these three basins in more detail and considering reintroductions or supplementations in more detail (e.g., among populations within core areas, rather than among core areas), we found that Upper Klamath Lake had greater potential to support establishment of a new population via reintroduction. Furthermore, the results of this assessment itself were impacted dramatically by a major disturbance, the Bootleg Fire of 2021, even as the assessment was being conducted. Collectively, these findings and unexpected disturbances highlight the importance of active inquiry at all scales (range-wide to local) to provide support for agile decision-making in a world where the pace of increasing threats challenges the pace of management action.