**Development of Carbon Dioxide Barriers to Deter Invasive Fishes: Insights and Lessons Learned from Bigheaded Carp**

#### **Cory D. Suski**

Department of Natural Resources and Environmental Sciences, University of Illinois, 1102 S Goodwin Ave, Urbana, IL 61801, USA; suski@illinois.edu; Tel.: +1-217-244-2237

Received: 2 June 2020; Accepted: 4 August 2020; Published: 13 August 2020

**Abstract:** Invasive species are a threat to biodiversity in freshwater. Removing an aquatic invasive species following arrival is almost impossible, and preventing introduction is a more viable management option. Bigheaded carp are an invasive fish spreading throughout the Midwestern United States and are threatening to enter the Great Lakes. This review outlines the development of carbon dioxide gas (CO2) as a non-physical barrier that can be used to deter the movement of fish and prevent further spread. Carbon dioxide gas could be used as a deterrent either to cause avoidance (i.e., fish swim away from zones of high CO2), or by inducing equilibrium loss due to the anesthetic properties of CO2 (i.e., tolerance). The development of CO2 as a fish deterrent started with controlled laboratory experiments demonstrating stress and avoidance, and then progressed to larger field applications demonstrating avoidance at scales that approach real-world scenarios. In addition, factors that influence the effectiveness of CO2 as a fish barrier are discussed, outlining conditions that could make CO2 less effective in the field; these factors that influence efficacy would be of interest to managers using CO2 to target other fish species, or those using other non-physical barriers for fish.

**Keywords:** invasive species; bigheaded carp; biodiversity; behavior; physiology; toxicity; avoidance

#### **1. Background**

The transport of species beyond their native range represents a major global problem. The arrival of an invasive species can lead to the suppression of native populations through competition, the introduction of pathogens, predation, hybridization, and disruptions to habitats and ecosystem function [1–3]. Invasive species are therefore believed to be the second most important driver of species extinctions after habitat loss [4], and can lead to billions of dollars in economic costs [1,5]. More importantly, the decrease in biodiversity that invasive species cause can threaten human health and well-being [1,6–8]. Freshwater environments are experiencing declines in biodiversity disproportionately large relative to other biomes [9,10], and invasive species are one reason for this decline [1,11]. Studies have suggested that almost 40% of North American freshwater and diadromous fishes are imperiled [12], and the pace at which freshwater fish are becoming imperiled exceeds other vertebrates, and appears to be accelerating [12,13]; invasive species are a key factor contributing to these declines [4,12]. The rate at which humans have been introducing species beyond their native ranges has also accelerated over the past hundred years, driven primarily by the growth in global trade and mobility [3,14]. More importantly, models suggest that the transport of invasive species around the planet is likely to increase in the future [15,16].

While the eradication of an invasive species is theoretically possible, the unfortunate reality is that, once a species is introduced into an area, its removal is often impossible. For an invasive species to be successfully eradicated, a number of conditions must be met. These conditions include: proper planning and establishing lines of authority, a commitment to complete the eradication effort

in terms of resources and enthusiasm, the biology of the target species must be amenable with the entire population of the target species put at risk, the target population must be removed faster than it can reproduce, the target species must be detectable at low densities, and efforts must be made to prevent re-invasion (possibly through restoration activities) [17,18]. These conditions are easiest to meet for isolated, small populations with low reproductive rates and poor dispersal capabilities, often for terrestrial vertebrates, with plants and aquatic species proving more challenging [17,18]. Thus, owing to the challenges associated with eradication efforts, the literature is rich with examples of failed attempts to extirpate invasive species, despite efforts that have extended over many years [18–20]. In some situations, the goal of completely removing an invasive species can be considered controversial as eradication can be costly, unlikely to succeed, and may result in considerable damage to non-target organisms and the environment [17,18,20]. Owing to the obstacles associated with eradication, a common outcome following the invasion of a species is "maintenance management", whereby the goal of elimination is abandoned, and the invader is simply controlled to a density that is deemed tolerable and allowed to persist [18,19]. Therefore, to avoid this sustained presence of an invasive species and perpetual "maintenance management", a more cost-effective, and meaningful approach to invasive species management is to prevent the arrival of an invasive species prior to invasion [1,21,22], or deter the secondary spread of invaders should they arrive [23].

#### **2. Bigheaded Carp**

Carps from the family Cyprinidae have been introduced outside of their native range for centuries. Bigheaded carps [24], and particularly bighead carp (*Hypophthalmichthys nobilis*) and silver carp (*H. molitrix*), have been introduced widely as phytoplankton control organisms in commercial aquaculture ponds and sewage lagoons owing to their large size and ability to efficiently filter phytoplankton and zooplankton from the water column [25]. Following transport to the United States for use as a biological control agent, floods allowed bigheaded carp to escape into the Mississippi River where they have spread throughout the basin [25], undeterred by locks, dams or other flood control structures [26], and are currently one of the most abundant species in portions of the Illinois River [27]. More importantly, bigheaded carp have had documented negative impacts on aquatic ecosystems [25]. Silver carp, for example, can consume detritus and bacteria [28], and reduce the size and abundance of both phytoplankton and zooplankton [25,27,29–31]. As a result, studies have shown that populations of bigheaded carps can result in reduced condition and abundance of native planktivorous fishes [32,33], as well as a reduction in the abundance of adult sport fish that compete with bigheaded carps at the larval and juvenile stages [34]. Owing to their abundance, mobility and impacts on receiving ecosystems, a tremendous amount of resources have been devoted to the suppression, removal and eradication of bigheaded carp from the Illinois River for almost a decade [35]. While efforts to date have been successful at reducing population sizes by removing millions of kilograms of fish through contract harvesting and agency collections [35], populations of bigheaded carp still remain throughout the Illinois River, necessitating suppression efforts to prevent the expansion of the population.

#### **3. Chicago Area Waterway System**

Bigheaded carp have direct access into the Great Lakes Basin from the Mississippi Basin due to the presence of the Chicago Area Waterway System (CAWS). The CAWS is a series of human-created canals and channels, completed in the early 1900s, that breached the continental divide between the two basins. The CAWS was constructed for the purpose of removing both sewage effluent and stormwater runoff from Chicago, coupled with allowing the passage of commercial shipping vessels to move from the Great Lakes to the Gulf of Mexico [36–38]. At present, the only means of deterring the movement of bigheaded carp through the CAWS from the Mississippi basin into the Great Lakes (beyond extensive suppression/harvest efforts) is a trio of electric barriers near Romeoville, IL, USA, constructed in 2002 [37]. Silver carp and bighead carp currently are over 60 km from Lake Michigan [39], so the effectiveness of these electric barriers at stopping bigheaded carp from passing

has not been tested explicitly. However, numerous investigations have documented that these barriers are subject to problems and deficiencies that could allow the passage of bigheaded carp. For example, Dettmers et al. [40] showed that a number of fish confined to cages did not become immobilized when dragged through the barrier alongside steel-hulled barges, Sparks et al. [41] showed that an adult common carp (*Cyprinus carpio*) outfitted with an acoustic telemetry tag was able to traverse the electric barrier (possibly associated with a passing barge), while Evans and Brouder [42] showed that fish can move through the electric barriers if they are trapped between barges. Parker et al. [43] used stationary sonar deployed within the barriers and showed small fish were able to move through the electric fields independent of the presence of barges. In addition, electricity loses effectiveness when applied to small fish [44], the electric barrier is prone to maintenance shut downs, floods and power loss [38], and no non-physical barrier is effective at stopping 100% of fish [45]. Mitigation measures have been proposed to redesign shipping locks to reduce the possibility of the exchange of invasive species between the Mississippi and the Great Lakes basins. The plan to modify locks will cost billions of dollars, take a decade or more to complete, and has yet to start, leaving the Great Lakes vulnerable to the passage of bigheaded carp through the CAWS for the foreseeable future [46]. The consequences, should bigheaded carps traverse the electric barriers and enter the Great Lakes, are not known and are difficult to predict [47–51]. The consensus is that an invasion of bigheaded carps would not be beneficial, however, making the containment of carp within the Mississippi Basin a critical priority for stakeholders. To supplement existing suppression efforts and increase redundancy and effectiveness at preventing movement or spread through the CAWS, additional barrier technologies would be valuable, ideally a technology that will permit the passage of barges and the downstream transport of wastewater through the CAWS.

Based on the above background, the goals of this review are to (1) outline the development of zones of carbon dioxide gas (CO2) as a non-physical barrier to deter the movement of invasive fishes, with a particular focus on two bigheaded carps: silver carp and bighead carp, and (2) highlight internal and external factors that mediate the performance of CO2 as a non-physical barrier, either increasing or decreasing its effectiveness as a barrier for invasive fish passage. When taken together, this review will not only share the origins of CO2 as a fish barrier, but also help researchers think about ways to improve performance and maximize the ability of different barrier technologies to deter the spread of invasive fishes.

#### **4. Carbon Dioxide in the Atmosphere**

The idea that CO2 could be used as a fish barrier is rooted in Earth's history and the evolution of fishes. Billions of years ago, CO2 levels in the Earth's atmosphere were high, and O2 was low [52]. As photosynthesizing bacteria on the planet became more abundant, the composition of gasses in the atmosphere changed such that the relative level of O2 increased and the level of CO2 declined [52,53]. This change in atmospheric oxygen was concurrent with metabolic evolution that increased reliance on oxidative phosphorylation that uses oxygen as a final electron acceptor resulting in more efficient metabolism, coupled with the production of CO2 as a waste product [52]. Thus, organisms developed the ability to sense environmental gasses, including CO2, and respond by either avoiding CO2-rich areas that might impair energetic processes, or possibly being drawn to CO2-rich areas if they provide an energetic advantage [53]. Bacterial and fungal pathogens, for example, can sense environmental CO2 associated with hosts and alter growth or life cycles to maximize virulence [53]. Hawkmoths (*Manduca sexta*, Lepidoptera: *Sphingidae*) use floral CO2 emissions to quantify food source profitability and the amount of nectar in flowers [54], while honey bees (*Apis mallifera*) actively fan their hives to remove CO2 wastes, and the number of individuals fanning correlates positively with CO2 levels inside the colony [55]. Carbon dioxide excreted by vertebrates is used by mosquitoes (*Aedes* spp.) as a signal of a potential host [56,57], while *Drosophila* will avoid CO2, likely as a signal that rotting fruit is a poor food source [58]. For many terrestrial vertebrates, CO2 is detected by chemoreceptors in the blood stream and brain stem to regulate breathing [59], while, more specifically, mammals detect

of CO2 in the air with free nerve endings of the trigeminal system [60]. Together, concentrations of environmental CO2 can be a source of ecologically relevant information, and the ability to detect and respond to CO2 as a stimulus has persisted across kingdoms.

#### **5. CO2 and Fish Physiology**

Carbon dioxide has a pronounced effect on fishes, resulting in a host of physiological and behavioral responses when concentrations above species-specific set-points are experienced. Fish predominantly sense ambient CO2 using peripheral chemoreceptors, largely in the gills, that respond to CO2 tension in the water, not changes in pH; some evidence does exist for the presence of internal CO2 sensors, but the location of these sensors has not been well-defined [61]. When fish are placed in a high carbon dioxide environment, CO2 passively diffuses into the fish down its concentration gradient, and arterial CO2 equilibrates with environmental CO2 within minutes, resulting in an internal acidosis [62,63]. Over time, this pH imbalance is corrected as fish uptake HCO3 − from the environment (in exchange for Cl−) and excrete H<sup>+</sup> (in exchange for Na<sup>+</sup>) [63]. Owing to this influx of CO2, hypercarbic environments cause an elevation of the general stress response [64–66], a drop in blood pH [67], a loss of ions [68], and, ultimately, equilibrium loss and anesthesia (Stage 2 or Stage 3) [64,67,69,70]. At present, the exact mechanism(s) responsible for the loss of equilibrium and the anesthetic impacts of carbon dioxide have not been well defined, but are believed to result from the movement of CO2 across the blood-brain barrier, which alters brain pH and an impairs brain electrical activity [71,72]; additions of H<sup>+</sup> or HCO3 − alone will not result in anesthesia for fish [71]. In addition to these physiological changes, studies have documented behavioral changes exhibited by fish in high CO2 environments including hyperventilation, coupled with a reduction in heart rate, likely to facilitate CO2 excretion [61]. Together, exposure to water with elevated concentrations of CO2 has been shown to result in both physiological and behavioral changes to fish.

#### **6. CO2 and Fish Behavior**

A number of past studies have highlighted the propensity for fish to voluntarily swim away from areas of high carbon dioxide, laying the foundation for the use of CO2 as a fish deterrent. Avoidance reactions are considered the first line of defense for fish that encounter adverse stimuli, and poor water quality can quickly induce a behavioral response that causes fish to depart an area and seek out improved water, presumably to minimize energetic costs [73,74] (Figure 1).

Over a century ago, Shelford and Allee [75] designed a simple experiment to observe the behavior of nine fish species when placed individually in a raceway containing a gradient of CO2, ranging from approximately 2–88 cubic centimeters of CO2 per liter of water. Shelford and Allee [75] showed that, upon entering zones of elevated CO2, some fish started surface ventilations, while others displayed a coughing or yawning reaction coupled with increased ventilation rates. In addition, Shelford and Allee [75] reported that some fish would enter the area of high CO2, stop suddenly, and then move backwards as if they had "encountered a sheet-rubber wall", and all fish spent less time in areas of high CO2 relative to areas with lower CO2. Powers and Clark [76] used a laboratory gradient tank design similar to Shelford and Allee [75] and showed that both brook trout (*Salvelinus f. frontinalis*) and rainbow trout (*Salmo gairdnerii iridus*) also avoided water that had received "very slight" additions of CO2 (a drop of approximately 0.4 pH units). This pattern was also confirmed by Collins [77] who showed that individual alewife (*Polumbus pseudoharengus*) and glut herring (*P. aestivalis*) (likely river herring) migrating upriver both avoided water with elevated CO2 that exceeded 0.3 ppm, independent of pH changes. Bishai [78] showed that juvenile brown trout (*Salmo trutta*) and Atlantic salmon (*Salmo salar*) demonstrated a stronger avoidance response to a pH change caused by CO2 relative to a pH change caused by hydrochloric acid. Jones et al. [79] noted that individual arctic char (*Salvelinus aplinus*) will avoid concentrations of CO2 that exceed 50 μmol/L, Ross et al. [80] showed that brook trout and blacknose dace (*Rhinichthys atratulus*) would avoid water with ≥2% CO2, while Clingerman et al. [81] reported that intentional elevations of CO2 to 60 mg/L in an aquaculture would induce avoidance

behavior in groups of rainbow trout (*Oncorhynchus mykiss*), thereby facilitating harvest and collection in recirculating tanks. Finally, both Bernier and Randall [64], as well as Yoshikawa [82], revealed that rainbow trout exhibited a "violent" struggle upon being exposed to water maintained at 36–350 mmHg CO2, while Clingerman et al. [81] indicated that rainbow trout in aquaculture tanks demonstrated "chaotic" swimming when CO2 levels were increased to 35–60 mg/L. Thus, for over 100 years, studies have documented that many fish species will avoid areas of elevated CO2 once a threshold is reached, providing the proof of concept that CO2 could be a potential non-physical barrier for invasive fishes.

**Figure 1.** Flow chart showing the possible reactions of fishes to a potentially adverse environmental stimulus, such as an area of elevated carbon dioxide [73].

Despite suggestions from past work that CO2 could induce avoidance behaviors and act a barrier to the movement of bigheaded carp, a key unknown was the threshold CO2 level that should be targeted to induce avoidance. While the response of fishes to high concentrations of CO2 when applied as an anesthetic appeared to be consistent [67,83], and the physiological responses of fishes to general hypercarbia had been well-defined [63], relatively less was known about the thresholds or "inflection points" that cause the onset of disturbances (i.e., a dose-response curve), and if those threshold concentrations were consistent across species and life stages. For example, Ross et al. [80] exposed book trout, slimy sculpin (*Cottus cognatus*) and blacknose dace to four levels of CO2 (0%, 1.4%, 2.8% and 5.1%) for either one or 24 h and noted differences in physiological responses both across species and across exposure durations, suggesting species-specific responses to CO2 exposure. To address this need and define concentrations that induced onset of disturbances, Kates et al. [66] exposed bluegill (*Lepomis macrochirus*), largemouth bass (*Micropterus salmoides*), silver carp (>450 mm) and bighead carp (>700 mm) to two different concentrations of CO2 (30 mg/L and 70 mg/L) for three hours and showed that, 30 mg/L CO2 (approximately 2000 μatm CO2) had minimal physiological or behavioral impacts, but a three hour exposure to 70 mg/L CO2 (approximately 50,000 μatm CO2) resulted in a drop in ventilation rates, and an increase in irregular behaviors such as erratic swimming, twitching and escape attempts for silver carp and bighead carp [66]. One of the challenges with the study by Kates et al. [66], however, was that adult bigheaded carp were used, which provided little evidence in support of how small fish, those presumably less vulnerable to the existing electric barriers in Romeoville, IL, would respond to CO2. In an effort to better define the allometric response of fish to CO2 exposure, Dennis et al. [84] exposed larval and juvenile (73 mm) silver carp and bighead

carp to 120 mg/L CO2 (approximately 42,000 μatm CO2) for 30 and 60 min. Results from this study were similar to previous work with adult fish, in that exposing larval silver carp and bighead carp to 42,000 μatm CO2 for 30 min resulted in an increase in the mRNA coding for genes associated with the stress response (*Hsp70* and *c-fos*) [84] (Figure 2).

**Figure 2.** Relative expression of c-fos mRNA extracted from the gill tissue of juvenile bighead carp (**a**), silver carp (**b**), bluegill (**c**), and largemouth bass (**d**) exposed to a two hypercarbic treatments. Relative mRNA expression of juvenile fish that had an exposure duration of 30 min are shown in black bars, while white bars show the mRNA expression of juvenile fish exposed for 60 min. Horizontal lines denote a significant CO2 concentration effect across exposure durations within a species. Dissimilar letters indicate significant differences between bars within a species. Data are mean ± SE, calculated relative to the expression of the reference gene (i.e., either *18s* or *ef1-a*). For clarity, data are expressed relative to the mean of juvenile fish exposed to ambient water conditions [84].

Thus, when results from these two studies are taken together, data suggest that thresholds of approximately 42,000 μatm CO2 (70–120 mg/L) induce a suite of physiological and behavioral responses for a range of sizes of silver and bighead carp consistent with stress or discomfort, providing a target in the development of a non-physical barrier for fish.

#### **7. CO2 and Physiological Responses**

Following the identification of putative thresholds that would induce behavioral and physiological disturbances, studies on CO2 barriers shifted to quantify aspects of avoidance (Figure 1). Despite the research mentioned above that indicated a pattern of fish avoiding zones of elevated CO2, there were suggestions in the literature that avoidance responses may be variable across species. Ross et al. [80], for example, showed that individual slimy sculpin did not avoid zones of elevated CO2 and preferred to rest in place when confronted with hypercarbia, while Summerfelt and Lewis [85] noted that CO2 concentrations from 3.0–9.7 mg/L did not repel green sunfish (*Lepomis cyanellus*) in a graded laboratory tank. Early work with CO2 avoidance and bigheaded carp was conducted by Kates et al. [66] who used a "shuttle-box" apparatus in a laboratory to show that individual adult silver carp (460 mm) would

voluntarily swim away from CO2 once concentrations were elevated to approximately 120 mg/L CO2, although there was considerable variation around this mean value (Figure 3) (Table 1).

**Figure 3.** Concentration of CO2 at which largemouth bass, bluegill, and silver carp displayed either an agitated activity (surface ventilations, twitching, or elevated swimming activity) (**a**) or movement out of high CO2 environment to a lower CO2 environment (**b**) during the course of avoidance trials [66].

Dennis et al. [84] later used juvenile silver carp and bighead carp (67 mm and 71 mm, respectively) and the shuttle-box apparatus, and, again, showed that individual fish would voluntarily swim away from zones of elevated CO2. The concentration of CO2 required to induce avoidance in this series of tests averaged approximately 180–220 mg/L CO2, and, again, the variance around the mean was considerable (Figure 4). The success of these laboratory trials led to work at larger settings, including Donaldson et al. [86] who showed that a number of fish species, including silver carp and bigheaded

carp, released into a 4000 m3 outdoor pond in groups of 5–10 avoided zones of elevated CO2 elevated to approximately 30,000 μatm (60 mg/L), and Cupp et al. [87] who used a two-channel, outdoor raceway (approximately 60 m3) with flowing water and showed that CO2 levels of approximately 30,000–40,000 μatm (~75 mg/L) would deter the movement of both silver and bigheaded carp (278 mm and 212 mm, respectively) when tested in groups of 10 (Table 1). Cupp et al. [88] showed that CO2 deployed at the mouth of an outflow structure draining a backwater lake could reduce the abundance of shoals of mixed fish species by 70–95% at low water flows once a target threshold of 100 mg/L was reached. Finally, Hasler et al. [89] worked in a 12 m long indoor swim flume and showed that bighead carp (145 mm) in shoals of 3 would avoid CO2 in water flowing at approximately 15 cm/seconds (equivalent to 1 body length per second), and CO2 levels in this study were approximately 190,000 μatm. When considered together, these studies used a number of environments (indoor, outdoor, static water, flowing water) to demonstrate that a range of sizes of invasive silver and bighead carp, including small fish presumed to be less vulnerable to electricity, would voluntarily swim away from zones of elevated CO2 once a threshold of approximately 70,000 μatm (100 mg/L) was reached, providing support for the use of CO2 as a non-physical barrier.

**Figure 4.** Concentration of CO2 at which juvenile bluegill, largemouth bass, silver carp and bighead carp displayed avoidance behaviors [84].


**Table 1.** Summary of studies quantifying CO2 thresholds that caused avoidance within the framework of generating a non-physical barrier for silver carp and bighead carp. Data have been approximated from figures whe.2e it was not clearly outlined in text of the citation. Units are left in the format that was used during publication.

#### **8. CO2 as a Potential Fish Barrier**

When the general anesthetic properties of CO2 exposure [67] were combined with results from field and lab avoidance trials, there was a considerable amount of evidence to suggest that zones of elevated CO2 could deter the spread of invasive bigheaded carps. More specifically, a CO2 barrier could be deployed in one of two different ways. First, CO2 could be deployed as a "fence" or wall with the goal of inducing *avoidance* behaviors in fishes, exploiting the fact that fish voluntarily swim away from areas of high CO2 once a threshold had been reached. For example, CO2 could be used to confine carp in backwater areas [90] preventing access to turbulent, high-velocity water flowing river environments used for spawning [91], or at a choke-point in a river (e.g., shipping lock) to stop movement. A CO2 barrier deployed in this way could be temporary (e.g., deployed only during summer or during harvest), or for longer periods of time. Secondly, zones of CO2 could be deployed to intentionally induce equilibrium loss for fish, taking advantage of the ability of fish to *tolerate* CO2 as an anesthetic. Again, an application of this kind could be temporary (i.e., deployed at specific times of the year) or longer-term (e.g., added to a shipping lock) [46].

#### **9. Questions from Avoidance Data**

While the concept of using CO2 as a barrier based on either avoidance or a tolerance has support from a number of studies, there were several puzzling trends in the data, which generated questions and presented challenges related to possible deployment. For example, shuttle-box work by Kates et al. [66] showed that avoidance thresholds for individual adult silver carp and bighead carp spanned from approximately 50 mg/L to 160 mg/L (Figure 3) (Table 1). Subsequent shuttle-box work by Dennis et al. [84] with juvenile fishes, showed that avoidance thresholds ranged 6-fold, from approximately 50–300 mg/L (Figure 4). This variation in avoidance is further complicated by work from outdoor ponds showing avoidance occurred for groups of bighead carp, but CO2 never exceed 60–75 mg/L [86,87]. Questions therefore arose related to the source of this variation (i.e., Is there inter-individual variation? What is the nature of the differing test environments? Is this variation inherent in how animals respond to zones of CO2?), the potential for inter-individual differences in tolerance, and the effectiveness/consistency of CO2 across time periods or environments. Thus, it was difficult to make recommendations to managers on target thresholds necessary to achieve an effective CO2 barrier, or to predict possible changes in barrier effectiveness, without a more thorough understanding of the response of fish to CO2 barriers. A series of studies were therefore conducted to quantify endogenous and exogenous factors that influenced the avoidance and tolerance of fishes to elevated CO2 in hopes of refining this technology, providing stronger, more definitive recommendations to managers on target thresholds for CO2 barriers, and improving the likelihood of long-term performance of CO2 as a non-physical barrier.

#### **10. Factors Influencing the Avoidance of CO2**

Several different endogenous and exogenous factors have the potential to influence the avoidance response of CO2 in the context of a non-physical barrier (Table 2). For instance, in recent years, it has become apparent that fish consistently differ from each other in behavior, often termed "personality"; some individuals are more bold than others, some are more active, and some are more likely to explore novel areas [92]. Invasive round goby at the leading edge of their range, for example, were shown to be more bold and willing to explore novel areas than individuals from established, core populations [93], and it is therefore plausible that individual differences in personality could be manifesting in inter-individual differences to CO2 avoidance [94] (Figures 3 and 4). More importantly, personality differences covary with characteristics such as the response to stressors, life span and growth rate through the pace-of-life continuum [92]. Therefore, if the response to an environmental stress and avoidance thresholds are mediated through behavior (e.g., proactive vs. reactive coping styles [92,95]), due to links between personality, life history and fitness [96], if target levels for an avoidance barrier are too low and fish of a particular behavioral type are able to pass, this could translate to population-level shifts in phenotypes, possibly changing avoidance thresholds for a population. In exploration of this concept, Tucker et al. [97] showed that aspects of personality (e.g., activity and boldness) did not influence CO2 avoidance in individual bluegill, with fish of all personality types avoiding CO2 at a threshold of approximately 67,000 μatm. In addition, Tucker and Suski [98] showed that social personality in bluegill (e.g., sociability, clustering with conspecifics and conspecific aggression) also did not influence CO2 avoidance thresholds or the order that fish avoided CO2. Related to this, past work has shown that food deprivation can alter the behavior of fish through plastic or flexible changes, with animals deprived of food taking more risks and becoming more active, likely as they search for food [99,100]. Interestingly, Suski et al. [101] showed that nine days of food deprivation did not influence CO2 avoidance thresholds for individual largemouth bass; fish that had been fed and fish that had been deprived of food, both avoided high CO2 at thresholds of approximately 70,000 μatm. However, Tucker et al. [97] showed that CO2 avoidance in individual largemouth bass was influenced by artificial activation of the stress axis as fish that received an intraperitoneal injection of cortisol (hydrocortisone 21-hemisuccinate) required 45% more CO2 to induce avoidance behavior relative to fish that did not receive an artificial elevation of the stress axis. Many initial studies of CO2 avoidance [66,84] were conducted on individual fish, but shoals have a number of benefits for fish including predator vigilance and food detection, resulting in a calming effect and a reduced response to environmental stressors [102]. Tucker and Suski [98] showed pronounced differences in CO2 avoidance thresholds for individual fish relative to shoals, with groups of bluegill choosing to swim away from CO2 at significantly lower thresholds than individual bluegill; interestingly Tucker et al. [97] also showed that shuttling thresholds were not repeatable within individuals. Allometry is known to influence a number of characteristics of fish including metabolism and survival, but intraspecific differences in CO2 avoidance thresholds across size categories is not clear. When avoidance thresholds for small and large bighead carp are compared across Kates et al. [66] and Dennis et al. [103], small fish appear to require higher CO2 thresholds to induce avoidance. These results, however, were obtained in separate studies, not in a single investigation, so inter-study differences may have been a complicating factor. The quantity of CO2 necessary to induce avoidance in round goby (*Neogobius melanostomus*) [104], silver carp and bighead carp [105] correlated positively with water temperature (range from 5–25 ◦C), such that more CO2 was required to induce avoidance at high temperatures for all three species tested. Note that both Cupp et al. [104] and Tix et al. [105] did not acclimate fish at test temperatures for a period of two to three weeks as is common [106,107], with holding times listed at two to six days, which could have influenced these results. Together, a number of factors have been shown to influence the threshold of CO2 required to induce avoidance behaviors, which have implications for the application of CO2 as an avoidance barrier to deter the movement of invasive fishes (Table 2).


**Table 2.** Factors influencing the thresholds of CO2 required to induce avoidance behaviors in fishes.

#### **11. Factors Influencing CO2 Tolerance**

Similar to work with avoidance, a number of studies have been carried out to quantify inter-individual differences in CO2 tolerance, as well as potential mechanisms for any differences (Table 3). Importantly, Hasler et al. [108] showed that, for largemouth bass, CO2 tolerance not only varied across individuals, with some fish losing equilibrium in high CO2 sooner than others, but also that the individual response to high CO2 was repeatable within individuals, suggesting potential for this to be a heritable trait that can be acted upon by natural selection [109]. In general, tolerance to CO2 is a function of the interaction of exposure time × concentration, further mediated by temperature [64,69,70,110–112]. More specifically, a brief exposure to a high concentration of ambient CO2, or an extended exposure to lower concentrations of CO2, will both result in equilibrium loss, provided that the concentration of ambient CO2 is sufficient to passively diffuse into the bloodstream of the fish [64,70,110,112]. Owing to reduced respiratory and metabolic rates at low temperatures, fish typically require additional time at lower temperatures before anesthetic effects are realized relative to high temperatures [112]. Indeed, this has been demonstrated for CO2 as both Fish [69] and Gelwicks et al. [110] used study designs where individual fish were transferred to containers of CO2 at a target concentration and showed decreased time to equilibrium loss at high temperatures, suggesting that fish are more sensitive to CO2 at high temperatures. Interestingly, both Cupp et al. [104], and Tix et al. [105], showed that, when CO2 was continually added to a test tank, round gobies [104], silver carp and bighead carp [105] all required higher concentrations of CO2 before equilibrium loss occurred when animals were at high temperatures relative to low temperatures, suggesting that fish were more tolerant to CO2 at higher temperatures. There are three potential explanations for the discrepancies across these studies. Firstly, differences across studies could be due to experimental animals, as Fish [69] and Gelwicks et al. [110] worked with salmonids, while Tix et al. [104] and Cupp et al. [105] used round goby and bigheaded carp. Secondly, Tix et al. [105] and Cupp et al. [104] applied CO2 to fish continuously until equilibrium loss occurred, while Fish [69] and Gelwicks et al. [110] pre-treated tanks of water with CO2 to a set point and added fish. Finally, both Cupp et al. [104] and Tix et al. [105] did not acclimate animals to each test temperature for extended periods of time, and, rather, animals were first held at 12 ◦C and then transferred to the test temperatures for 24–144 h prior to testing, which may have influenced their response to CO2

exposure [106,107]. Clingerman [81] showed that, when CO2 level was held constant, large rainbow trout were more likely to lose equilibrium than small rainbow trout in aquaculture tanks, suggesting an increased tolerance for smaller fish. Tucker et al. [97] showed that aspects of personality (e.g., activity and boldness) did not influence CO2 tolerance in bluegill, with fish of all personality types requiring similar durations of time to induce equilibrium loss when exposed to 123,000 μatm. Hasler et al. [108] showed that tolerance to CO2 was influenced by the metabolic phenotype of largemouth bass, and fish with higher anaerobic performance, quantified as time to become exhausted when burst swimming, required less time to lose equilibrium when exposed to high CO2, and also that aerobic aspects of metabolic phenotype (i.e., standard metabolic rate, aerobic scope) did not influence tolerance to carbon dioxide. Suski et al. [101] showed that largemouth bass that had been deprived of food for 14 days required 25% longer exposure to high CO2, relative to fish that had been fed over this 14 day period, thereby demonstrating an increased tolerance to CO2 from food deprivation. Together, tolerance to high CO2 can vary due to a number of endogenous and exogenous factors and should be considered should CO2 be deployed to deter the movement of invasive fishes (Table 3).


**Table 3.** Factors influencing the tolerance of CO2, indicated by loss of equilibrium.

#### **12. Management Implications**

There are a number of potential non-physical barriers that can be deployed to prevent the spread of invasive fishes, including bubble screens, sound or electricity, each with particular strengths and weaknesses [45]. A non-physical barrier that uses zones of elevated carbon dioxide to deter fish movements has a number an advantageous as a chemical control tool relative to other technologies as it has few human health concerns, can be applied in a carbon neutral fashion using repurposed CO2 (i.e., harvesting waste CO2 destined to be released into the atmosphere), is relatively inexpensive and readily available, can be deployed with relatively little infrastructure, and residual CO2 does not persist in the environment [113]. Carbon dioxide was recently registered with the United States Environmental Protection Agency (USEPA) as a pesticide for use as a deterrent of bigheaded carp under the name Carbon Dioxide—Carp (EPA Registration Number 6704-95). Dennis et al. [103] held largemouth bass at 21,000 μatm (13 mg/L) CO2 for almost two months and showed no decline in avoidance thresholds, suggesting that acclimation to the presence of high CO2 is not likely. Most important, the avoidance response of fishes to environmental CO2 appears to be canalized, demonstrated by virtually all fish species tested, while CO2 tolerance is repeatable and consistent [108], giving CO2 a number of advantages as a non-physical fish barrier as a tool to deter invasive fishes.

Results from the studies listed above have a number of implications for the deployment of CO2 as a non-physical barrier and can be used to minimize the likelihood of unintentional fish passage, while also helping minimize waste CO2 and reduce deployment costs. For example, for a CO2 barrier deployed with the intention of causing avoidance, it is important for managers to consider the context in which the barrier is deployed. More specifically, although not repeatable within individuals, avoidance of CO2 has been shown to be consistent across virtually all species tested when CO2 pressures reach approximately 30,000–60,000 μatm (60–100 mg/L). However, avoidance thresholds will likely be lower for fish in shoals (rather than individual fish) but will increase if fish are experiencing stress (independent of food availability), such as chronic hypoxia or environmental pollution. Finally, studies suggest that higher concentrations of CO2 may be required to induce avoidance at warmer water temperatures (summer) relative to cooler conditions (Table 2). It should be noted that Schneider et al. [114] showed that CO2 did not impair either the burst or sprint swimming performance of largemouth bass until thresholds of 100,000 μatm were reached (approximately 150 mg/L), well in excess of thresholds required to induce avoidance, suggesting that, if fish choose to challenge a CO2 barrier and burst through it, the barrier will likely not impair swimming performance. When considered together, a number of factors should be considered to ensure maximum effectiveness should CO2 be used in the field to deter the movement of invasive fishes (Table 2).

If a CO2 barrier is deployed with the intent of stopping fish via equilibrium loss (tolerance), aspects of individual fish need to be considered as these factors can influence effectiveness. At present, the relationship between exposure time × CO2 concentrations that results in equilibrium loss for most species has not been defined, so these data would need to be collected to help guide management targets, and owing to individual variation in the loss of equilibrium time for fish [108], a large number of fish would need to be assessed to quantify a range of equilibrium loss times. In general, however, small fish, and individuals that had been deprived of food, would be expected to have improved tolerance in high CO2 relative to larger, well-fed individuals. The role of environmental temperature has not been clearly defined, but studies suggest that a longer exposure time may be required at lower water temperatures and in periods of low food availability (e.g., winter) (Table 3).

#### **13. Future Work**

At present, there are five areas that should be the focus of future studies to improve the performance and efficacy of CO2 as a non-physical barrier. Firstly, additional work should focus on defining differences in both avoidance and tolerance thresholds across fish of different sizes; this is particularly important given the possibility that electricity as a barrier may lose effectiveness against small fish [43]. Currently, work that quantifies avoidance and tolerance thresholds across a range of sizes of fishes, within a single study with consistent methods, has not occurred. Owing to the likelihood that a CO2 barrier would be encountered by fish of a range of sizes, the ability to confidently predict the response of different sized individuals to either a tolerance or avoidance application of CO2 is critical. Secondly, the exact parameters of the time × concentration interaction to induce equilibrium loss associated with a tolerance-focused barrier have not been defined extensively, and would need to be parameterized across target species before tolerance barriers could be developed and/or implemented. Ideally, this work would be conducted across a range of temperatures. Thirdly, work should be conducted that pairs CO2 barriers with additional stimuli (e.g., deploy a sound barrier and CO2 barrier concurrently as in Ruebush et al. [115], or use CO2 as part of a bubble curtain rather

than compressed air). No non-physical barrier is 100% effective at stopping all fish [45], but a CO2 barrier paired with a second stimulus (e.g., light or sound barriers) could synergistically improve the overall effectiveness of each barrier, increasing the potential to deter invasive fishes across a range of conditions. Penultimately, efforts need to occur to quantify the logistics of CO2 deployment, including cost estimates, deployment feasibility and infrastructure requirements to assist with future planning efforts. The design for deploying a CO2 barrier will vary across sites and situations, but efforts to share costs and strategies across successful applications will help improve deployment efficiency and ensure success across locations. Finally, owing to the unavoidable reductions in pH that occur with zones of elevated CO2, work should continue to quantify the environmental impacts [116,117], consequences for non-target organisms (e.g., mussels [118,119]; native fishes [120–122]; crayfish [123]) and strategies for CO2 off-gassing. Work to both mitigate CO2 applications, coupled with efforts to predict possible impacts to non-target organisms or the receiving environment, will help improve the likelihood of a successful application. Together, work to address these 5 concerns will not only help improve the effectiveness of CO2 as a non-physical barrier, but also will help minimize unintended environmental consequences and improve the efficiency of CO2 as a tool to deter invasive fishes.

#### **14. Conclusions**

Invasive species represent a significant threat to global biodiversity, and models suggest that the rate of introduction of invasive species will likely accelerate in the future [16]. For fishes in North America, bigheaded carp represent a current threat to the Mississippi ecosystem, and there is potential for them to gain access to waterbodies in the eastern portion of the continent should they pass through the Chicago Area Waterway System (CAWS) into the Great Lakes basin. Carbon dioxide (CO2) is a naturally occurring compound that provides ecologically-relevant information to a host of taxa. A number of different studies, conducted across a range of conditions, have demonstrated that zones of elevated carbon dioxide gas can be an effective non-physical barrier to deter the spread of invasive fishes. More specifically, fish will voluntarily swim away from zones of high CO2 once a target threshold has been reached, or else equilibrium loss will occur due to the anesthetic properties of CO2, providing two different mechanisms by which carbon dioxide can deter fish movement. This response has been documented for a number of taxonomically diverse species of fish, and also across a range of sizes spanning from larvae to adults. In addition, unlike physical barriers, a CO2 barrier can be deployed without requiring the construction of permanent structures that can modify water flow or boat traffic. Several internal and external factors can influence the response of fishes to CO2, making them more effective, or less effective (e.g., fish experiencing stress will require additional CO2 to induce avoidance relative to non-stressed individuals; shoals of fish require less CO2 to induce avoidance relative to solitary individuals) and need to be considered when defining target thresholds should CO2 be deployed in the field. Additional studies to define effective deployment strategies at large scales, cost, and impacts to the receiving environment should continue as CO2 barriers grow in popularity and field applications. Work is currently ongoing to develop other non-physical barriers to deter invasive fishes (e.g., sound, electricity, strobe lights), and the lessons learned and experiences described here from CO2 can serve as potential considerations to refine the application of other barrier technologies to increase their effectiveness. Together, with continued exploration and testing, it is hoped that barrier technologies can be further developed to prevent the spread of invasive fishes and protect freshwater biodiversity.

**Funding:** This research was funded by the Illinois Department of Natural Resources, and the United States Geological Survey, through funds provided by the United States Environmental Protection Agency's Great Lakes Restoration Initiative.

**Acknowledgments:** A number of individuals have contributed to field and laboratory work that have generated the data contained in this review, including Caleb Hasler, Jennifer Jeffrey, Kelly Hannan, John Tix, Eric Schneider, Emi Tucker, Madison Philipp, Ian Bouyoucos, Christa Woodley, Cody Sullivan, Jason Romine, Dave Smith, Steve Midway, Aaron Cupp, Jon Amberg, Mark Gaikowski, Clark Dennis, Matt Noatch, Dan Kates, Shivani Adhikari, Michael Donaldson, and Adam Wright. Jake Wolf fish hatchery provided fish for experiments, and staff at the Illinois River Biological Station facilitated the collection of carp. Caleb Hasler provided comments on an early version of this review.

**Conflicts of Interest:** The author declares no conflict of interest.

#### **References**


#### *Fishes* **2020**, *5*, 25


© 2020 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Biocontrol of the Common Carp (***Cyprinus carpio***) in Australia: A Review and Future Directions**

#### **Kenneth A McColl \* and Agus Sunarto**

CSIRO-Health and Biosecurity, Australian Animal Health Laboratory, PO Bag 24, Geelong, VIC 3220, Australia; agus.sunarto@csiro.au

**\*** Correspondence: kenneth.mccoll@csiro.au

Received: 22 April 2020; Accepted: 29 May 2020; Published: 2 June 2020

**Abstract:** Invasive pest species are recognized as one of the important drivers of reduced global biodiversity. In Australia, the 267 invasive plant, animal and microbial species, established since European colonization in the 1770s, have been unequivocally declared the most important threat to species diversity in this country. One invasive pest, the common carp (*Cyprinus carpio*), has been targeted in an integrated pest management plan that might include cyprinid herpesvirus 3 (CyHV-3) as a potential biocontrol agent. The species-specificity of the released virus (and of field variants that will inevitably arise) has been assessed, and the virus judged to be safe. It has also been hypothesised that, because the virulence of the CyHV-3 will likely decline following release, the virus should be used strategically: initially, the aim would be to markedly reduce numbers of carp in naive populations, and then some other, as yet uncertain, complementary broad-scale control measure would knock-down carp numbers even further. Brief results are included from recent studies on the modelling of release and spread of the virus, the ecological and social concerns associated with virus release, and the restoration benefits that might be expected following carp control. We conclude that, while further work is required (on the virus, the target species, environmental issues, and especially the identification of a suitable broad-scale complementary control measure), optimism must prevail in order to ensure an eventual solution to this important environmental problem.

**Keywords:** biocontrol; Australia; common carp; *Cyprinus carpio*; cyprinid herpesvirus 3; safety; efficacy; modelling; risks

#### **1. Introduction**

In 2013, in his comprehensive book on invasive species, Simberloff [1] suggested that biological invasions are (along with climate change and habitat destruction) one of the great anthropogenic threats to global diversity. Subsequently, a United Nations-backed panel of scientists (representing over 130 nations) produced a report in 2019 that unequivocally identified invasive species as one of the important drivers of the global decline in numbers, and frequent extinction, of native animal and plant species. The report of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Sciences (IPBES) [2] noted that at least 680 species of vertebrates, alone, have been lost due to human actions taken since 1500.

The IPBES Report suggested that there were five direct drivers of reduced global biodiversity: (1) changes in the use of land and sea, (2) direct exploitation of the plants and animals of the world, (3) climate change, (4) pollution, and (5) invasive pest species. There has been a 70% increase in numbers of the latter since 1970 across 21 countries where detailed records were maintained. At about the same time that the IPBES Report was released, an Australian group declared [3] that invasive species were, in fact, the major threat to species diversity in Australia followed by modifications to ecosystems and agriculture. Furthermore, they found that this hierarchy of threats was consistent for

almost all native plants and animals in Australia, the only exceptions being native fish where pollution replaced agriculture as the third major threat.

It is estimated that, of the thousands of exotic species that have arrived since European colonization of Australia in the 1770s, 267 have become genuine invasive pest species (207 plants, 57 animals, and three microbial pathogens) [3]. Concomitant with this influx of invasive species, at least 93 native species of plants and animals have officially become extinct with the demise of many others either recognised informally, or likely to have gone unrecorded [3]. Australian inland water communities have been forced to contend with cane toads (*Rhinella marina*) and around 43 known invasive freshwater fish species including eastern gambusia (*Gambusia holbrooki*), goldfish (*Carassius auratus*) and perhaps the most disliked of all, the common carp (*Cyprinus carpio*). Surprisingly, of about 300 species of Australian freshwater fish that are recognised in 59 families, none are known to have become extinct since European colonization although there is evidence of regional extinctions, and recovery actions have probably saved several species from extinction [4]. In addition, different federal and state bodies have listed 74 freshwater species as 'threatened' [4]. It is likely that invasive fish species are, directly or indirectly, associated with the dire status of many Australian native freshwater fish.

*C. carpio* (known simply as 'carp' in Australia) was probably first introduced to waters around Sydney, Australia in 1908 (earlier records possibly being confused with goldfish, *Carassius auratus*) [5,6]. However, it was not until the 1960s that they were recognized as a serious invasive pest species, particularly in theMurray-Darling Basin (MDB), a regulated river system that covers 14% of the continent on the eastern side of the nation [4]. Carp comprise up to 90% of the fish biomass in parts of the MDB, and it is recognized that they are responsible for a deleterious cascading effect on the aquatic environment: they uproot and consume aquatic vegetation which increases the turbidity of the water. This change then leads to further reductions in aquatic vegetation, invertebrate communities, aquatic birdlife, and native fish [7].

In the early years of the new millennium, the Australian Federal Government began a program to develop innovative measures for the control of several important terrestrial and aquatic invasive pest species. An integrated pest management (IPM) plan was developed for carp in Australia, and, following the recognition of cyprinid herpesvirus 3 (CyHV-3; also known as koi herpesvirus) in Israel and the USA in 1998 [8], a research program was initiated to investigate the potential of this virus as a biological control (biocontrol) agent within the IPM plan for carp. This review summarizes the work already completed, and it also identifies the outstanding requirements before CyHV-3 could be considered as a biocontrol agent in Australia.

#### **2. The Essential Information Required for Potential Viral Biocontrol of Carp**

Since the 1950s, Australia's use of two different viruses for rabbit biocontrol has demonstrated many generic lessons for future viral biocontrol programs of invasive vertebrates [9]. In broad terms, these lessons indicate the necessity for an understanding of the biology of both the targeted pest species and the putative biocontrol virus. For carp control in Australia, in particular, a vast amount of information has already accumulated on carp biology in this country, although key pieces of information were still required. However, as an exotic (or foreign animal disease) virus for Australia, specific information about CyHV-3 in Australian conditions was almost non-existent. Table 1 summarizes the essential additional information required to not only understand carp biology and ecology in this country, but also the far greater needs to provide insights into CyHV-3 activity. Much of this information on the host and the virus has now been acquired through government-sponsored research programs ("Knowns" in Table 1), but deficiencies in our knowledge still exist ("Unknowns"). Due to space considerations, the following sections briefly focus on recent additions to our knowledge about carp and CyHV-3, especially for Australia's needs.





#### *2.1. Carp Biology in Australia*

#### 2.1.1. Distribution Models of Carp in Australia and Biomass Estimates

A great deal of excellent work has been conducted on carp biology since the 1960s when the species was first recognized as an important pest in Australian waterways; as examples, see [7,10–13]. However, while estimates of carp biomass in the MDB have been proffered in the past, any national carp biocontrol program would require a modern, more extensive and more accurate approximation. Working across numerous jurisdictions, Stuart et al. [14] used catch-based models to generate "heat maps" that depicted the biomass and spatial distribution of carp throughout the waterways of south-eastern Australia in 2011 and 2018. It was already known that when carp exceed a threshold density of 80–100 kg/ha, detrimental ecological impacts may occur [11]. Stuart et al. [14] found that modelled carp biomass exceeds this threshold across large areas of south-eastern Australia and therefore is consistent with the view that carp may have landscape-scale impacts manifested by the decline of water quality, native flora, fauna biodiversity, and recreational values. In short, carp represent a serious threat to freshwater ecosystems.

The 2011 and 2018 biomass estimates [14] were based on static spatial mapping. However, because carp populations can respond rapidly to changes in hydrological conditions, these estimates cannot be applied to future scenarios when a biocontrol virus might be released. Therefore, Stuart et al. [14] recommended the use of a dynamic model to provide future estimates of carp biomass, taking into account a variety of possible hydrological scenarios. Todd et al. [15] undertook dynamic modelling using an established carp population model [13] and the static biomass estimate for 2018 [14]. They then provided a range of estimates for the biomass of carp for 2023 in four regions of south-eastern Australia. Their results highlighted the variability of populations with differing hydrological and ecological conditions, and this, in turn, emphasized the advantage of dynamic modelling: it provides managers with a current estimate of carp populations in different locations, which then assists managers when considering where to release a biocontrol virus, and also where to focus clean-up operations to ameliorate the impacts of large numbers of dead carp.

#### 2.1.2. Genomic and Transcriptomic Map of Carp in Australia

Using variability in 14 microsatellite loci, Haynes et al. [5] studied the population genetics of carp at each of 34 locations throughout the MDB. They confirmed the presence of the four recognized strains in Australia: Boolarra, Yanco, koi and Prospect, despite Prospect being originally restricted to Sydney (outside of the MDB). They also concluded that there was significant genetic structuring of carp that was associated with barriers to dispersal. In fact, they divided the MDB into 15 management units, each unit based on man-made or natural barriers to dispersal of carp. They noted that, while invasive species often show decreased levels of genetic diversity in a new location, some actually have similar, or greater, diversity due to the invasives being introduced a number of times from different sources. This apparently applies to carp in the MDB which have high levels of genetic diversity (with multiple strains in all regions). They warned that the 15 management units should be interpreted with caution because fish-ladders may increase connectivity.

While the work of Haynes et al. [5] has been valuable in providing a preliminary understanding of the population genetics of carp in Australia, there is a dire need for a genomic study of this pest species in the MDB. This would provide a level of information on the targeted pest that is commensurate with our knowledge of the Indonesian strain of CyHV-3 that has been identified as a biocontrol virus in Australia (see Section 2.2). There are at least two immediate needs that could be addressed by a genomic study of carp in Australia: information on the virome of carp in this country, and an understanding of the polymorphisms in some immune response genes that may be critical in determining the virulence of CyHV-3 (see Section 2.4).

#### *2.2. Viral Epidemiology*

Since 1998, virulent CyHV-3 has been identified in many countries, but there is no evidence for its presence in Australia [16]. Although the origin of the virus remains problematic, it appears to have arisen in recent decades, possibly from avirulent variants in Europe [17], but there is also evidence for unusual variants in New York State [18] and Oregon [19] in the USA. Other molecular studies [20] also support the idea that an avirulent variant(s) of CyHV-3 has been present in *C. carpio* for tens of thousands of years although Kopf et al. [21] highlighted two major assumptions that perhaps cast some doubt on this time-frame—firstly, that the evolutionary rate of CyHV-3 has been constant, and secondly, that this rate for an alloherpesvirus from an exothermic host is similar to an alphaherpesvirus from an endotherm.

Under permissive conditions, CyHV-3 can cause 70–100% mortality in juvenile and adult *C carpio* [22–24]. However, larvae less than 1 cm in length are completely resistant to infection due to the protective effect of skin mucus. Larvae gradually become susceptible with increasing size, culminating in complete susceptibility when they are longer than 2 cm [25–27]. Only very low doses of virus are required for infection [28], the main portal for both infection and excretion being the skin [29]. Once excreted from an infected fish, a virus survives in the aquatic environment for only about three days, regardless of water temperature [30]. While most of these epidemiological data have been acquired from overseas studies, there is little reason to expect major differences under Australian conditions. However, due to biosecurity concerns with CyHV-3, this view has not been proven because field trials are yet to be conducted in this country.

An Indonesian strain of CyHV-3 (the C07 isolate) will potentially be used as the biocontrol virus in Australia. The full genome sequence has been determined [31], revealing that the gene layout is very similar to CyHV-3-U (a US reference genome) although 310 genetic variations between the C07 strain and the reference genome were identified. Phylogenetic analysis inferred from comparisons of whole-genome sequences revealed that the Indonesian isolate is more closely related to a Japanese isolate within the Asian lineage than to isolates within the European lineage.

Being a herpesvirus, CyHV-3 is likely capable of inducing latent infections in surviving carp, but while initial studies have demonstrated that low-temperature persistent infections are possible [32,33], unequivocal latent infections are yet to be demonstrated [34].

#### *2.3. Safety of the Virus*

Two of the most important lessons from Australia's earlier work on rabbit viral biocontrol have been the necessity for assessing both the safety and efficacy of any potential biocontrol virus. 'Safety' is about species-specificity, not only of the virus isolate selected for potential release into the environment, but also of any future generations of the virus that may evolve genetic changes (mutations or recombination) following release in the field (see Section 2.6).

The most compelling evidence for the specificity of CyHV-3 is that viral-induced disease has never been reported anywhere in the world in any species other than *C. carpio* since CyHV-3 was first recognized. This includes species in polyculture systems with carp. It is an observation that has often been ignored by critics of the virus, but its importance should never be overlooked. Importantly, this broad observation includes humans whose fears of infection can also be allayed by several observations: there has been no evidence of adverse effects on humans working in CyHV-3-affected carp farms. The two closely-related viruses, CyHV-1 and -2, are not known to infect humans. There was no evidence of infection in CyHV-3-challenged mice (selected as a representative mammal in non-target species susceptibility trials) [16]. More generally, there is no evidence for any fish virus causing disease in humans [35]. These findings were corroborated by Roper and Ford [36] who, in addition, recommended that the "psychosocial effects" on human health of a mass fish kill should be investigated.

For other potential targets, numerous laboratories have suggested that many species could become infected by CyHV-3, but without causing disease. In summary, the susceptibility of 24 to 25 species of fish was tested [22,37–42], and, in all cases, there was no evidence of disease. While CyHV-3 genomic DNA was detected in 10 of 15 species of fish that were exposed to acutely- or latently-infected carp, only a small proportion of each species was supposedly infected, none showed clinical signs of disease, and only low copy numbers of CyHV-3 DNA were found [39]. Similar results were found for plankton, mussels and crustaceans [43,44]. In none of these cases, however, was there an attempt to demonstrate CyHV-3 mRNA as an indicator of virus replication, a necessary corollary of infection. It should be noted that even though one study [45] did use an RT-PCR, ostensibly to demonstrate replication of CyHV-3, their work was flawed technically in that neither primer in their RT-PCR was designed in separate viral exons nor over splice junctions. It is likely that their primers were actually detecting residual contaminating genomic DNA from virus rather than viral mRNA [16].

Yuasa et al. [46] eventually developed an RT-PCR for CyHV-3 that allowed differentiation of genomic DNA from the mRNA of replicating virus. This allowed a definitive laboratory study on the susceptibility of the following non-target species (NTS) to CyHV-3 [16]: 13 native Australian fish species, introduced rainbow trout (*Oncorhynchus mykiss*), native lamprey ammocoetes (*Mordacia mordax*), domestic chickens (*Gallus gallus domesticus*), laboratory mice (*Mus musculus*), a freshwater crustacean (*Cherax destructor*), two species of frogs (*Litoria peronii* and *Lymnodynastes tasmaniensis*), and two reptilian species (*Intellagama lesueurii* and *Emydura macquarii*). When challenging each of these NTS, CyHV-3 was given the best chance of causing disease through the use of immature, susceptible NTS that were exposed, by immersion and/or intraperitoneal inoculation to 100–1000 times the dose of virus required to infect a carp.

All challenged NTS were subjected to clinical, gross pathological and histopathological examinations, and to PCR testing (using a screening qPCR, and the specific RT-PCR [46] to re-examine any qPCR-positive samples). While low copy numbers of CyHV-3 DNA were found in occasional samples by qPCR, all such samples were negative for viral mRNA by the RT-PCR suggesting that the weakly-positive qPCR results were, in fact, due to low-level contamination events during processing of samples rather than to the presence of replicating virus. Thus, it was concluded that no evidence could be found for infection, let alone disease, in any of the NTS. Boutier et al. [47], however, offered alternative interpretations. Firstly, they suggested that "technical issues" were not addressed (although they provided no specific details on what these issues might be), and, secondly, that the deaths in NTS could have been due to a non-replicative pathogenesis such as may occur in herpesvirus latent infections of non-natural host species [48]. The latter is an interesting suggestion, but seems to overlook two important observations: (1) latency, although likely to occur in carp surviving infection with CyHV-3, has not actually been demonstrated yet in the host species, let alone a NTS [34], and (2) a productive infection, the necessary precursor to a latent infection, has not even been demonstrated in any NTS. For example, McColl et al. [16] did not find clinical signs of disease, histological lesions or any evidence of an early productive infection (in the form of viral transcripts) in any of their NTS inoculated with CyHV-3 despite examining many NTS at early and later stages following inoculation.

Kopf et al. [21], while accepting that adverse effects of the virus on native species are "highly improbable", were, however, still loathe to absolve CyHV-3 of all potential threat. They suggested that native species could be "asymptomatic carrier(s)" or transmitters of the virus. Again, this ignores the fact that to be a carrier, a non-target native species must first be infected, a claim that has never been properly demonstrated (see above). Furthermore, claims for all but very short-term transmission by non-carp species were refuted by the elegance of the simple experiments with CyHV-3 on goldfish [49]. The final argument by Kopf et al. [21], that sub-lethal infections in immunocompromised NTS be investigated, has, indirectly, already been addressed by McColl et al. [16] through the use of immature fish (with incompletely developed immune systems) in their susceptibility studies on NTS.

In summary, we believe that a robust standard protocol is required for future susceptibility testing of NTS, and we propose the following: (1) time-course sampling to demonstrate an increase or decrease of viral DNA concentration in a viral-exposed NTS, (2) using both qPCR and RT-qPCR for detecting viral DNA and mRNA, respectively, (3) attempting virus isolation from any NTS with clinical signs of disease, and (4) using histopathological examination on moribund NTS. Molecular testing, including next generation sequencing if available, should also be considered for exclusion of other known or unknown pathogens.

The results of the NTS experimental work [16] were complemented by the findings from a North American study of natural outbreaks of CyHV-3 in carp [50]. At each outbreak, no disease was observed in any co-habitating species, even in native cyprinids, thus attesting to the species-specificity of the virus. However, there are two important criticisms of other observations in the North American work: (1) there was no attempt to look for the presence of any pre-existing, potentially cross-reactive viruses that might confer protection on carp. In particular, there have been no reported serological, PCR or next-generation sequencing studies on any carp populations in North America, and (2) only one of the outbreak sites offered the opportunity for direct fish-to-fish transmission by means of dense aggregates of carp. In the Thresher et al. study [50], there appeared to be few, if any, equivalents of the limited numbers of densely populated carp breeding sites distributed throughout the MDB. Thresher et al. [50] also claimed, probably correctly, that mass die-offs would not be expected for a herpesvirus that is in equilibrium with its natural host. CyHV-3, however, is a pathogen that has only recently been recognised [8], possibly because it has only recently arisen [20]. Therefore, it has not yet had time to come into equilibrium with its host, in which case, high mortalities are probably not unexpected.

#### *2.4. E*ffi*cacy of the Virus*

Determining the 'efficacy' of a potential biocontrol virus is slightly more complex than determining its 'safety' because the former depends on two variables, 'transmissibility' and 'virulence'. For a biocontrol virus, 'transmissibility' is defined as the ability of the virus to establish infection in new hosts, and is often measured by the basic reproduction number, R0, the average expected number of cases produced by a single case (in a population where all individuals are fully susceptible). 'Virulence' is a measure of the severity of the disease caused by the virus, not necessarily measured simply by mortality. For example, in the classical studies of the MYXV, Fenner and Woodroofe [51] established five grades of virulence that were based on a combination of both survival time following infection, and mortality.

Using Australia's two rabbit biocontrol viruses as examples, Di Giallonardo and Holmes [52] demonstrated that, while there are invariably strong selection pressures for transmissibility, this has been achieved for MYXV and RHDV by selection in the field of virus strains of intermediate and high virulence, respectively. These quite different paths suggest that, for any particular virus, it is not always easy to predict the outcome of the complex relationship between transmissibility and virulence. There have been no direct studies on how CyHV-3 achieves maximal transmission, but observations on viral epidemiology, particularly the virus–host interaction, may encourage the formulation of two hypotheses [53].

Firstly, the observations that CyHV-3 is excreted at low titre into an aquatic environment, and then only survives for about three days outside its host [30,54], suggest that direct transmission of virus between carp is likely to be much more important than indirect transmission via the aquatic environment. Furthermore, given that carp are highly sensitive to infection [28], and that the skin is the main portal of both infection and excretion of CyHV-3 [29], a reasonable hypothesis is that direct skin to skin contact between an infected and an uninfected fish, even if transient, is the most likely form of transmission. Such contact would likely disrupt the skin mucus layer which would enhance virus entry [26,27]. Clearly, carefully designed transmission experiments are required to test this hypothesis.

Having become infected, a viraemia develops in the carp, and the virus localizes in various tissues [55]. The adaptive immune response of the fish develops slowly [56–58], likely allowing survival of some infected hosts with potential latent infections although, as already mentioned, latency has not yet been proven unequivocally [34]. Nevertheless, assuming it does indeed occur in surviving fish (as it does in the hosts of all known herpesviruses), then recrudescence of acute

infections will also occur during periods when infected fish are stressed. In Australia, massive aggregations of carp occur at annual breeding events, and such aggregations are known to induce stress and immunosuppression [59]. This, in turn, implies that annual breeding would not only allow reactivation of CyHV-3 infections, but would also favour transmission of virus by direct skin-to-skin contact of the densely aggregated fish. These observations then suggest a second hypothesis: that long-term transmission of CyHV-3 in Australian conditions may be favoured by the natural selection of low virulence strains of CyHV-3 that would allow survival of some latently-infected fish which, in turn, would lead to multiple periods of recrudescence and transmission of virus to naive fish during annual breeding events. It was postulated earlier that selection pressures may change as the density of carp declines [53], but, on reflection, this possibility may be of little importance if, indeed, most transmission occurs at densely aggregated breeding sites. The latter sites will likely form regardless of the total number of carp in the river systems because, at least in Australia, carp seem to be irresistibly attracted to these sites at certain times of the year [60]. So, while the total area of any particular breeding site may decline, the density of fish will probably remain high.

A legitimate question that arises because of the second hypothesis is that, if transmissibility drives the selection of low virulence strains of CyHV-3, can the virus be an effective biocontrol agent in Australian waters? Perhaps the answer may be found in lessons from past viral biocontrol programs involving rabbits in Australia [9]. Field experience with both MYXV and RHDV has revealed that a virus, alone, will not control the targeted invasive pest species. In fact, to be effective, biocontrol viruses must be complemented by other broad-scale control measures, a fact that has been emphasized many times from the outset of the carp biocontrol program in Australia. The use of such measures is not an admission of failure in the proposal to use CyHV-3 as a biocontrol agent; rather, it is an argument for the use of the virus in a carefully designed IPM program. The virus would markedly reduce numbers of carp in naive populations, providing the opportunity for complementary measures to then substantially knock down carp numbers even further (see Section 2.7).

The second factor affecting the efficacy of the virus, virulence, may be difficult to determine. While Fenner andWoodroofe [51] used standard inbred laboratory rabbits for their work on the virulence of MYXV, no equivalent line of carp is available in Australia to allow a standard test of the virulence of different isolates of CyHV-3, nor, indeed, to determine R0. However, the virulence of the C07 isolate of CyHV-3 has been demonstrated in numerous studies on carp collected from all over south-eastern Australia for example, [16,61,62]. Further studies are required to test the virulence on carp collected from throughout the entire MDB.

Boutier et al. [47] expressed a number of reservations about the use of CyHV-3 as a biocontrol agent for carp in Australia. They contended that natural resistance of some carp (due to resistance-conferring polymorphisms in immune genes) and of carp-goldfish hybrids could lead to rapid proliferation of resistant phenotypes. Access to genomic and transcriptomic maps of Australian strains of carp would help to address the question of immune genes, while assessing the future importance of carp-goldfish hybrids would likely require modelling work and a better understanding of the current numbers of these hybrids in the MDB.

Boutier et al. [47] also proposed that phylogenetic studies suggested that CyHV-3 may already be present in carp in Australia, just as it may have long been present, without expressing virulence, in carp populations around the world [20]. Studies on 849 carp samples from nine sites throughout the MDB in Australia, utilizing a nested PCR (with primary and nested primers aligning perfectly with sites in the DNA polymerase gene of CyHV-1, -2, and -3), failed to reveal evidence for any known or undescribed cyprinid herpesviruses [63]. Nevertheless, this is recognized as only a preliminary study, and a more definitive virome study (using a next generation sequencing approach) from a similar sample of carp is essential to corroborate the PCR work.

Finally, in assessing the likely efficacy of CyHV-3 in Australia compared with natural overseas outbreaks [21], it is important to recognize a critical difference between the two situations: whereas most of the world is consumed with controlling outbreaks of CyHV-3 disease, Australia would aim to *enhance* the spread of the disease (and then to augment the effect of the virus with complementary control measures). To this end, it is essential that we have a deep understanding of the epidemiology of the disease under Australian conditions. Kopf et al. [21] state that "Lake Biwa (in Japan) and Blue Springs Lake (in the USA) are not good models for Australian conditions if the virus was (sic) released", but, nevertheless, most of their criticism of Australian activities is based on findings from these overseas outbreaks. Based on the known biology of the virus and of carp in Australia, we have proposed two hypotheses that account for virus pathogenesis and transmission under Australian conditions; aspects of these hypotheses have informed the work on epidemiological modelling.

#### *2.5. Epidemiological Modelling of Virus Release and Spread*

A large multi-disciplinary team developed four inter-related models, namely hydrological, habitat suitability, carp demographic, and epidemiological models, with the intention of informing any future staged release of CyHV-3 in the MDB [64].

The hydrological model focussed on the water temperature and connectivity of waterways for five diverse catchments in south-eastern Australia. It concluded that, while CyHV-3 will generally be effective from Spring through Autumn throughout south-eastern Australia, a staged release of virus would demand precise estimations of water temperature prior to release in any particular catchment to ensure conditions were permissive for virus activity. However, water temperature alone was insufficient for determining the time of virus release. It was also found that the major environmental factors influencing the distribution and abundance of carp in south-eastern Australia, and the manner in which these factors interacted with each other, were also essential in selecting a time for virus release.

This conclusion was reached through a habitat suitability workshop that utilized expert opinion within the context of a Bayesian belief network (BBN). The BBN identified river flow and water temperature as the two essential parameters determining the suitability of a habitat for adult and sub-adult carp, and both were rated as medium to high for most habitats throughout the study period. On the other hand, waterway inundation and connectivity, the essential habitat suitability factors for an abundance of larvae and young-of-year (YOY) stages, were rated poorly in most habitats during the study period. Population abundance for YOY stages, in fact, depended on a relatively small number of dense aggregations of juveniles and adults that occur in transiently flooded wetlands throughout the MDB (so-called 'recruitment hotspots'). Through the use of conversion factors guided by expert opinion, habitat suitability rankings were converted to biomass density estimates (the latter validated by recently acquired data [14]). These estimates would then allow CyHV-3 to be used in those areas where the population density of carp was approximately 80–100 kg/ha, the level at which detrimental ecological impacts may occur [11].

The biomass densities from the habitat suitability model were then used to develop a full spatio-temporal population projection (or demographic) model of carp population dynamics in which carp metapopulations were resolved into six age-stage classes (eggs, larvae, early YOY, late YOY, sub-adults, and adults). This demographic model, in turn, was integrated into a CyHV-3 epidemiological model that allowed the prediction of mortality and suppression of the subpopulations following a hypothetical release.

The epidemiological model was a variation on the standard SEIR transmission model. It included susceptible (S), exposed (E) and infectious (I) classes, but the usual recovered (R) class was replaced with classes more likely to represent a typical herpesvirus, such as CyHV-3. Thus, latent (L) and recrudescent (Z) classes were introduced, with Z representing second and subsequent infections following repeated reactivation of the virus in latently infected carp. The modelling predicted that, without recrudescence, introduction of the virus would be associated with a single mortality event in carp. However, in the more likely event of latency and recrudescence, there would be an ongoing and lasting suppression of carp populations in all catchments with reductions being to approximately 40% of the pre-release population. The impact of the virus would be sufficient to reduce carp populations in many MDB waterways to below the damage threshold of 100 kg/ha for at least 10 years. A further notable prediction

was that seasonal losses would be mainly in immature carp, the mortality in the adult population being less by at least an order of magnitude (and, therefore, possibly not easily observed).

Boutier et al. [47] suggested that, at various times of the year, there could be vast tracts of water in the MDB where temperatures may be non-permissive for virus replication. These concerns have long been noted [9]. However, the modelling report [64] specifically noted that "tailoring the release of the virus to the particularities of each catchment" would be important, especially in those areas where there was a very narrow window of permissive Spring temperatures. The use of other complementary control measures (see Section 2.7) could also be important in some areas. Boutier et al. [47] also suggested the importance of 'behavioural fever' in fish as a response to infection. There is no question that this phenomenon works for individual fish, but it has not prevented mass mortalities of carp in thermally variable natural aquatic environments overseas, and it would be unlikely to do so in Australia either (particularly if a virus were to be released in relatively homogeneous shallow breeding grounds of carp).

#### *2.6. Evolution of the Released Virus*

While laboratory work and field observations strongly suggest that current strains of CyHV-3 are highly specific for *C. carpio* (and some hybrids [65,66]; see Section 2.3), the question remains about the likelihood of genetic changes in current field isolates causing a future expansion of the host range. There is no direct evidence bearing on this question for CyHV-3, but there are pertinent lessons from Australia's past experience with viral biocontrol of rabbits. Evolutionary studies [67,68] have shown that DNA viruses can, indeed, mutate (although at a much lower rate than RNA viruses). However, while this may potentially allow spill-over events or host-jumps, such events for herpesviruses occur on timescales of millions of years, and, when they occur, they are invariably into taxonomically closely related species. It is reassuring then that, while there are a number of introduced cyprinids in Australia, there are no native cyprinids. Furthermore, the most closely related native Australian species (native catfish) are insusceptible to infection with CyHV-3 [16], and, again, it should be emphasized that viral-induced disease has never been reported anywhere in the world in any species other than *C. carpio*.

Field observations on rabbit biocontrol viruses in Australia lend support to these evolutionary studies [9]. Mutations are known to have occurred in the field in both the myxoma virus (MYXV, a DNA virus present in Australia for over 60 years) and in rabbit haemorrhagic disease virus (RHDV, an RNA virus, over 20 years) [69,70], but there is no evidence that either has jumped into another species. As noted by Di Giallonardo and Holmes [71], the overall conclusion is that "host-jumps to nontarget species are not an inevitable consequence of viral evolution". As a result, all observations, whether on current isolates or potential future field variants of CyHV-3, encourage the view that the chance of cross-species transmission is very small.

#### *2.7. Broad-Scale Control Measure(s) to Complement the Virus*

Saunders et al. [72] found that there have only been three major instances where viral pathogens have been used successfully against vertebrate pest species, namely MYXV and RHDV against rabbits in Australia and feline panleukopenia virus against cats on a South African offshore island. In reviewing the lessons from these attempts at viral biocontrol of invasive vertebrates [9], it was noted that in each case complementary measures were required for sustained control or eradication of the pest species. While these measures actually included supplementary regional controls, ideally broad-scale controls would be identified and implemented. A number of regional control measures have long been implemented for carp, including commercial harvesting, electrofishing, carp traps, fishing competitions, predator stocking, poisoning, and environmental controls [7,73,74]. However, these are generally ineffectual in the long-term.

The development of broad-scale control strategies that will deliver persisting declines in carp numbers has been more problematic. Wedekind [75] reviewed possible genetic biocontrol technologies that could be used on carp populations in Australia. He broadly classified them as those that involve genetic engineering (including 'daughterless' carp and gene-drive technologies) and those that do not (including 'Trojan Y chromosome' techniques).

Of the techniques involving genetic engineering, an early major investment was made into 'daughterless' carp technology in Australia [76]. The underlying principle was that, by using an RNAi approach that suppressed expression of the female differentiating genes, natural carp populations in Australia would be biased towards all-male populations. Although initially very promising, the approach gradually fell out of favour because modelling revealed that it would take many decades to exert an impact, the corollary being that very large numbers of these genetically modified fish would need to be added to waterways annually for many years in order to force a sex bias in natural populations. Currently, a gene-drive approach that would be lethal to female offspring, or leave them infertile, is not considered a safe option for carp [75], nor indeed for any biocontrol program [77]. In the future, however, gene-drive technology may become a universal approach to controlling invasive pest species, although many modifications to current technology will be required for this approach to become acceptable.

Approaches that do not involve genetic engineering include the use of Trojan Y fish [78]. This method relies on treating young male carp with a female sex hormone, oestrogen, resulting in genetic males (with XY chromosomes) that have female sex characteristics, including the ability to produce eggs. The latter have an XY constitution, and when fertilized by a normal male, they produce a preponderance of male offspring. Extensions of this basic approach can lead to stock populations of YY individuals being produced for release into wild populations [79]. Although this strategy would require the costly regular addition of modified fish to natural populations of carp for a number of decades, it is considered the most appropriate current technique, particularly if combined with measures to increase the survival and fecundity of the manipulated carp [75].

In the immediate future, perhaps new, more virulent strains of CyHV-3 may prove to be a useful complementary measure. A situation may develop that is analogous to the commercial chicken industries where the herpesvirus, Marek's disease virus (MDV), has been an ongoing threat for many decades. Strains of MDV of increasing virulence have evolved due to the use of imperfectly immunizing vaccines. Similar vaccines have been used in carp aquaculture to protect farmed carp from outbreaks of CyHV-3 for about a decade, and it is hypothesised that they too may lead to the evolution of more virulent strains of CyHV-3 that could be used as the next-generation biocontrol viruses in Australia [53].

In summary, a number of options for a broad-scale complementary control exist (Table 2). Each has their strengths and weaknesses, but currently, we are in complete agreement with Boutier et al. [47] and Kopf et al. [21] in declaring that the ideal broad-scale complementary measure(s) has not yet been identified. However, as previously mentioned [34], new genetic options continue to appear [80], providing optimism that the ideal measure will soon be developed. Until then, it would be unwise, even wasteful, to release CyHV-3 into the Australian environment.



#### *2.8. Ecological Concerns*

Kopf et al. [21] raised the possibility of "broad ecological risks of unintended and perverse outcomes from biocontrol with CyHV-3". They suggested a number of potential problems that could arise as a result of mass mortality, and subsequent decomposition, of carp following release of virus. Australia's National Carp Control Program [81] undertook a number of studies to address such concerns.

Beckett et al. [82] conducted a comprehensive ecological risk assessment of the consequences of the proposed release of CyHV-3 in a variety of aquatic settings including transient wetlands, river systems, lakes and other water bodies. Impacts on water quality were considered most likely in locations characterised by a high carp biomass and low water flow such as occurs during carp breeding in transient wetlands. It was suggested that risks to native fish and birds, in particular, could be avoided by releasing the virus during high-flow seasons, or by the partial removal of carp from waterways prior to the release of virus. Whether this would be a practical option for the release strategy would need to be considered. If not, then mitigation strategies might include physical removal of carp carcasses from affected areas, or the use of water regulation to flush not only carcasses but also cyanobacterial blooms from affected areas. These strategies would likely also reduce the risk of outbreaks of botulism in native species, although both a literature review and field experience in south-eastern Australia (where there have been many fish kills due to blackwater events) suggest that botulism is unlikely to have much practical importance anyway. The loss of many juvenile carp from wetlands treated with CyHV-3 raised the spectre of prey-switching by piscivorous waterbirds. While the potential impact on native fish and other species must certainly be considered, insufficient research has been conducted in Australia to allow firm conclusions on the many potential interactions. It is, however, noteworthy that a number of earlier studies on prey-switching have revealed complex dynamics between predators and prey species, but little cause for long-term concern about the survival of native species [83–85].

Finally, Beckett et al. [82] noted that residual uncertainty necessarily remains because it is not possible to predict, with confidence, the epidemiology of CyHV-3 in Australia. The extent of carp mortality, more or less than predicted, is a key uncertainty. Similarly, the impacts of low dissolved oxygen levels (DO) on the many and varied native aquatic species can never be certain, although water quality modelling studies [86] suggested that dangerously low DO was only likely to be a problem where carp biomass was high and there was a concomitant severely compromised (or absent) water flow. A similar situation is predicted for widespread cyanobacterial blooms.

In broad terms, other very important ecological considerations are, firstly, the clean-up procedures for carp following a mass mortality due to CyHV-3, and, secondly, the potential waste utilisation of the subsequent large masses of dead carp. A literature review revealed very limited information about clean-up processes following a fish-kill [87], and therefore, not surprisingly, most of the documented responses were of a reactive nature. However, the Atlantic region of Canada is one of the few locations that does have a well-documented clean-up procedure. Silva et al. [87] declared that the biomass of carcasses and the location of the fish-kill should be important determinants, among others, of the extent of the clean-up operation, particularly if the affected waterway is part of a town water supply (in which case the suggested importance of a cost-benefit analysis for the operation almost seems paradoxical). Globally, most fish-kills have relied on landfills for the disposal of carcasses.

Tilley et al. [88] investigated alternative methods of disposal, and found that composting methods, that are able to use even severely degraded material, are likely to be the best option on a large commercial scale. However, flexibility and scalability of the process would also allow small scale operations in remote regions up to larger scale operations by councils or smaller commercial organizations. A large-scale rendering option at a meat rendering facility was also shown to be possible, although only fish carcasses < 24 h post-mortality would be acceptable for processing.

#### *2.9. Social Risks*

Zhang et al. [89] undertook a risk assessment to determine public perceptions about the ecological and social risks associated with the proposed use of CyHV-3 as a biocontrol agent. They conducted wide-reaching qualitative and quantitative surveys of the Australian public, focussing in the former on the general public, and in the latter on those in urban settings versus those living near major waterways. An important finding was that people who live in the MDB and who are closely connected to the river system were more likely to accept the need for carp control while still retaining some reservations about aspects of the process. Overall, the studies emphasized the importance of early, effective communication programs in order to allay the concerns of various communities.

#### *2.10. Restoration Benefits from Carp Control*

Finally, Kopf et al. [21] questioned whether Australia could expect to see any ecological restoration benefits from carp control. Casual observation of affected waterways has long suggested that carp must be exerting a profound negative ecological impact, especially in those regions of the MDB where they account for up to 90% of the biomass.

However, the very comprehensive study conducted by Nichols et al. [90] relied on more than casual observation. Assuming the proposed biocontrol program would be successful in reducing carp numbers, an expert elicitation study was conducted on the expected medium- (5–10 years) to long-term (beyond 10 years) ecological consequences of a reduced carp population in Australia. The study addressed the effects on ecosystems as a whole, along with effects on the animal and plant components, and, on water quality.

In summary, the study found that Australia's waterways are complex ecological systems, and they will almost certainly continue to degrade if nothing is done to control carp. On the other hand, if carp populations could be sustainably reduced by 70–100%, experts believe there would likely be clear long-term ecosystem benefits [90]. The same experts also emphasised that carp are not the only ecological stressor, and that other widespread environmental problems must also be addressed. However, even under ideal conditions where all stressors are identified and controlled, the experts agreed that, rather than restoration of a degraded system to its original state, an unexpected new aquatic ecosystem may be generated.

#### **3. Final Comments and Conclusions**

Australia's experience with MYXV and RHDV for rabbit biocontrol has taught us a great deal about the principles of viral biocontrol for any invasive pest vertebrate species. Perhaps the most important lessons are that it is essential to have a deep understanding of the biology of both the targeted pest species and any potential biocontrol virus. Equally important is the lesson that a viral biocontrol agent, alone, can never be expected to completely eradicate an invasive pest species; to be successful, biocontrol agents must be complemented by other broad-scale control measures in an integrated pest management program.

Is there a need for a contingency plan for any proposed biocontrol virus in the event that, following release, it fails to work as expected? Assuming all the necessary precautionary studies have been conducted, particularly on the safety, efficacy, and epidemiology of the virus prior to its release, then the most likely reason it would be judged a 'failure' is that, after an initial burst of mortality, it apparently became ineffectual. This was the experience with MYXV in Australia in the 1950s. However, in later years, it was shown that, even as rabbit mortality due to MYXV declined, the virus still held rabbit numbers below the level at which they caused environmental damage [72]. Then, with the advent of effective complementary control measures, there has been sustainable control of rabbits for approximately 60 years in Australia, and, even in the absence of eradication, biocontrol of rabbits has delivered significant economic benefits to Australia [91]. Assuming that appropriate

additional controls are found to complement CyHV-3, we would anticipate a similar trajectory for carp control in Australia.

Our world currently faces myriad local and global challenges including, but not restricted to, climate change, overpopulation and loss of species biodiversity. In Australia, the control of invasive pest species, including carp, sits comfortably in this list of challenges. We need to do something about carp to improve the quality of our waterways in this country, and it is only reasonable and rational debate that will form the foundation of future decisions and actions. CyHV-3 appears to be a rare opportunity to control carp, although, as mentioned earlier, we recognise the importance of identifying a broad-scale control measure to complement the future activity of the virus in Australia. To release the virus prior to implementation of such a complementary measure would be very unwise.

However, placing a temporary embargo on the use of the virus is not to endorse inactivity. As a marine biologist recently noted in a general commentary, "unrelenting doom and gloom in the absence of solutions is not effective. Social scientists have known for decades that large problems without solutions lead to apathy, not action" [92]. We must all recognise our current progress and successes so that in 5–10 years we can all take pride in the contribution we made to carp control in Australia.

**Author Contributions:** K.A.M. and A.S. combined to write and edit this paper. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was partly funded by the Invasive Animals Co-operative Research Centre (Australia) through Freshwater Projects 4.F.7 and 3.W.1.

**Acknowledgments:** We thank to the Invasive Animals Co-operative Research Centre (Australia) through Freshwater Projects 4.F.7 and 3.W.1. The Australian National Carp Control Program, under the leadership of Matt Barwick at the time, provided encouragement for our work, and funded many other projects in the carp control program. Jawahar Patil (University of Tasmania, Taroona TAS 7053), Maciej Maselko (Macquarie University, North Ryde NSW 2109) and Mark Tizard (CSIRO Health and Biosecurity, Geelong VIC 3220) contributed substantially to Table 2.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Geographic-Scale Harvest Program to Promote Invasivorism of Bigheaded Carps**

#### **Wesley W. Bouska 1, David C. Glover 2, Jesse T. Trushenski 3, Silvia Secchi 4, James E. Garvey 5,\*, Ruairi MacNamara 6, David P. Coulter 5, Alison A. Coulter 5, Kevin Irons <sup>7</sup> and Andrew Wieland <sup>5</sup>**


Received: 17 July 2020; Accepted: 24 August 2020; Published: 1 September 2020

**Abstract:** Invasive bigheaded carps, genus *Hypophthalmichthys*, are spreading throughout the Mississippi River basin. To explore the efficacy of a consumer-based market (i.e., invasivorism) to manage them, we developed a conceptual model and evaluated three harvest approaches—direct contracted removal, volume-based incentives ("fisher-side" control), and set-quota harvest ("market-side" control). We quantified the efficacy of these approaches and potential population impact in the Illinois River. Contracted removal was effective for suppressing small populations at the edge of the range but cannot support a market. "Fisher-side" removals totaled 225,372 kg in one year. However, participation was low, perhaps due to reporting requirements for fishers. The "market-side", set-quota approach removed >1.3 million kg of bigheaded carp in less than 6 months. Larger, older fish were disproportionately harvested, which may hinder the ability to suppress population growth. Total density declined in one river reach, and harvest may reduce upstream movement toward the invasion fronts. With sufficient market demand, harvest may control bigheaded carp. However, lack of processing infrastructure and supply chain bottlenecks could constrain harvest, particularly at low commodity prices. Given the geographical scale of this invasion and complicated harvest logistics, concerns about economic dependence on invasivorism that encourage stock enhancement are likely unmerited.

**Keywords:** invasivorism; bigheaded carp; commercial fishing; *Hypophthalmichthys*; Illinois River

#### **1. Introduction**

Invasive species threaten biodiversity worldwide [1], costing \$120 billion USD annually in the United States [2]. Removal programs may control invasives [3]. For example, humans routinely overharvest fish stocks [4–6]. Thus, controlled harvest may help control invasive populations of fish and other taxa [7–10]. Yet, factors such as time, effort, and expense often limit success [11–13]. Whereas government assistance is necessary to control invasive species that have low market value such as sea lamprey, *Petromyzon marinus* (Linnaeus; [14]), fishes with commercial value could be marketed to reduce financial burden on

government agencies. This consumer-based control of invasive species is popularly called invasivorism. Potential candidates for invasivorism are silver carp *Hypophthalmichthys molitrix* (Valenciennes) and bighead carp, *H. nobilis* (Richardson), collectively known as bigheaded carp. These species invaded the lower Mississippi River basin of the US in the 1970s, expanded northward, are now more abundant in the Illinois River than anywhere else globally [15–17], and may invade the Laurentian Great Lakes via Lake Michigan. Establishment of bigheaded carp in the Great Lakes may jeopardize fisheries valued at \$7 billion USD per year [18,19].

Nearly a decade ago, the U.S. Army Corps of Engineers (USACE) explored options to prevent interbasin transfer of aquatic nuisance species (ANS) between the Mississippi River and Great Lakes basins, with bigheaded carp being a primary species of concern and the Chicago Area Waterway System (CAWS) the primary focus area. The CAWS contains five aquatic pathways [20], one of which, the Chicago Sanitary and Shipping Canal (CSSC), is the only permanently open connection between the basins, with Lake Michigan being the recipient Great Lake. The CSSC has previously allowed movement of ANS between the basins [21] and is near the edge of the bigheaded carp range, which is approximately 80 km south downstream in the upper Illinois River [22].

Removal of bigheaded carp is included in all eight management strategies for stopping interbasin movement of these fish [20], and non-structural control plus harvest is the only strategy that can be initiated immediately. Since 2010, contracted removal of greater than 3200 tons of bigheaded carp has occurred near the CAWS in the upper Illinois River, where commercial harvest is prohibited [23]. Removal at this range edge is expected to be agency funded for the foreseeable future [24] and has likely prevented upstream range expansion toward Lake Michigan [22]. In the lower Illinois River from where bigheaded carp in the upper Illinois River derive, commercial harvest is legal. If reliable moderate- to high-value markets can be developed for bigheaded carp in the lower river, exploitation should remain high and reduce upstream migrants via invasivorism. Although the likelihood of reducing bigheaded carp to extinction in such a large, open system is low, the capacity for population suppression and reducing further expansion may be high.

The idea of harvesting near the center of the invading population to reduce densities at range edges is supported by modeling that assesses the influence of harvest and other control measures in lower river reaches where population densities are high. The Spatially Explicit Asian Carp Population (SEAcarP) model [24] links movement probabilities among river pools or reaches with demographic responses to harvest removal to predict the likelihood of population density declines at the edge of the species' range (see Erickson et al. [25] for similar approach with grass carp). The model has been applied to the Illinois River using movement data from Coulter et al. [22], predicting that increasing mortality of bigheaded carp in the lower river will effectively reduce densities at the upstream invasion front.

Stimulating market demand to accomplish control via invasivorism may seem like a simple task. Bigheaded carp are valued food fish in much of the world. With native, wild bigheaded carp stocks threatened or extirpated [26], global demand is now primarily met by aquaculture, with these species being among the most cultured fish in the world [27]. Globally, over 5.3 million tons of silver carp are cultured annually, primarily in China, India, Bangladesh, Iran, the Russian Federation, and Cuba [27]. There may be a high demand for bigheaded carp from the U.S., since consumers in countries such as China are willing to pay a premium price for wild-caught fish [28], and may perceive U.S.-sourced fish as being of a higher quality than cultured products.

Although recent surveys have suggested that there is potential domestic consumer demand for bigheaded carp as food [29], most US markets are for rendered carp products (meals and oils), as ingredients in livestock and aquaculture feeds [30–32] and as hydrolyzed fertilizers. The current supply of bigheaded carp in US rivers is not a limiting factor in the growth of the industry, but rather the lack of processing plants and reliable domestic markets plus access to existing international markets to monetarily compensate commercial fishers. A well-developed fishery infrastructure does not exist in the central area of the invasion in the US. Thus, a fishery must be built to implement control via invasivorism.

To determine how a long-term, self-maintained fishery may be developed to control the bigheaded carp invasion, we first developed a broad conceptual supply chain model for economic development of a controlled harvest fishery to promote invasivorism in the Illinois River system, throughout the invaded US, and more broadly for any invaded ecosystem where market-driven consumption is an option. We then evaluated three harvest strategies against invasive bigheaded carp in the upper and lower Illinois River along the conceptual supply chain continuum, quantified the resulting exploitation, and examined economic factors affecting removal by harvest.

The harvest strategies evaluated during 2010 through 2012 were (1) an ongoing contracted harvest program in the upper Illinois River, (2) a "fisher-side" incentives program that offered select commercial fishers progressive economic rewards for participating (i.e., sharing harvest data) and harvesting increasing amounts of bigheaded carp for direct-consumption markets, and (3) a "market-side" incentives program that set a quota-based harvest of bigheaded carp for indirect-consumption markets of fish meal. Harvest removal of fish from populations is often biased toward larger, older individuals, which affects population responses [10]. We hypothesized that a removal effort of this large, geographic-scale magnitude should impact bigheaded carp population demographics including abundance, size structure, and age structure within the study area in the lower Illinois River, and stimulate direct and indirect consumption markets for bigheaded carp, with implications for basin-wide population dynamics and reduced risk to the Great Lakes.

#### **2. Conceptual Model**

The commercial fishing industry in the Illinois and Mississippi River basins is significantly reduced from its former past [16], where harvest of mussels and fish peaked in the early 1900s. Much of the fish harvest in this region was for local consumption and as consumer demand waned, river water quality declined, and sedimentation increased, fish harvest as a source of protein and income declined by the 1930s. For a fishery to develop to effectively reduce bigheaded carp from both free flowing and pooled reaches of these and other river systems, industry infrastructure must be reestablished. We propose a simplified conceptual model of bigheaded carp supply, transport, processing, and demand in flowing and non-flowing waters of the US Midwest (Figure 1).

The model identifies major components of a developing fishery for bigheaded carp. The first primary bottleneck to commercial-level invasivorism is the fishers and fishing fleet (Figure 1). The size of the fishing crews, training, time investment, and many other fixed and variable costs influence the reliability and effectiveness of fishing bigheaded carps. At this juncture, fishing is largely conducted by part-time crews with limited gears, because of a lack of funds for purchasing and maintaining boats and equipment. These crews fish for natives and carp, mostly for local businesses or contracted fishing in the upper Illinois River, maintaining local expertise to drive potential harvest expansion. However, with no reliable market for bigheaded carps, capital investment in large, sustained fishing operations will not occur.

The supply of bigheaded carps in the Illinois River and other systems is driven by myriad factors that vary with water body type and location (Figure 1). Lohmeyer and Garvey [33] conducted a recruitment assessment of bigheaded carps in pooled and unpooled reaches of the Upper Mississippi River, finding that recruitment was low but consistent in unpooled reaches and higher and variable in pools (also see Chick et al. [34]). Recruitment variability will affect the reliability of harvest and supply. Harvest will vary with location because of the logistics of reaching fishing locations and transporting fish back to boat landings. Unpooled rivers typically have fewer access points and longer travel times due to their longitudinal geomorphology. Ensuring that harvested bigheaded carps enter a market that provides an economic benefit is necessary to create a system where invasivorism is feasible. Currently, most bigheaded carps removed from the Illinois River system are sold directly to ethnic or local fish markets from the fishers or collected at no cost by fish processors following agency-contracted removal efforts in the upper Illinois River (see [32]). Harvested bigheaded carps also may be transported by fishers to larger processors, although the number of processors in the

region is limited and often geographically distant from harvest areas, making this a costly and often prohibitive option. A second bottleneck to developing market-based invasivorism is ensuring that a steady supply of fresh bigheaded carp is procured by fishers and transported to processors in a way that is safe for the consumer, economical, and efficient.

**Figure 1.** Conceptual model of a fishery for bigheaded carps in the rivers and lakes of the invaded south and north central United States. Economic factors affecting fishers will influence the species composition and biomass of bigheaded carp available to the supply chain. Transportation and markets will drive the level of harvest. Each component along the supply chain has costs (− sign) and economic gains (+) that determine how much fish are removed from the environment.

Providing bigheaded carp-derived products in ways that maximize their value is the goal of wholesalers and processors, where there is clearly global demand for bigheaded carp products. However, costs of setting up local processing facilities is high and potentially risky, while transporting whole fish to existing processors on the coasts or overseas is logistically difficult, involves federal regulation, and is ultimately costly (Figure 1). Creating local demand while developing export and processing facilities is likely the most economically feasible model for establishing harvest as control. However, most investors are wary of supporting such facilities without assurances of dependable supplies of high-quality bigheaded carp from the rivers, which is currently limited by undeveloped fishing capacity, uncertainties about regional fish production, government red tape, and a virtually non-existent transport network.

This conceptual model is not intended to be exhaustive, but it does provide several areas where investments or support may help develop a market for removing bigheaded carp at areas of high density. In the following sections, we describe an effort to stimulate fishing at the "fishing end" and at the "market demand" end of the bigheaded carp supply chain (Figure 1), and the collection of bigheaded carp population demographics through subsampling of commercially harvested carps at the processing plant, and fishery-independent sampling in the field before and after the harvest programs were implemented.

#### **3. Materials and Methods**

The two programs we created to stimulate harvest in the lower Illinois River were compared to the ongoing contracted fishing program in the upper Illinois River [35]. Commercial fishing does not occur in the upper Illinois River at the edge of invaded range, so market-driven removal is not an option in this region (Figure 2). In the lower Illinois River, commercial harvest has occurred for more than a century and has the potential to develop into a control method via invasivorism (Figure 2).

**Figure 2.** Map of the Illinois River study areas and processing plants during the 2011–2012 study period. Note that Griggsville became non-operational at an unknown time since the study concluded.

#### *3.1. Training, Certification, and Incentives-Based Approach*

To accomplish the goals of the "fisher-side" commercial fishing strategy in the lower Illinois River, a stakeholders' meeting was held in Grafton, IL, in the period 20–21 September 2010, during which academic researchers, regulatory authorities, commercial fishers, fish processors, marketers, and distributors discussed using commercial harvest of bigheaded carp in the lower Illinois River as a means of augmenting contracted fishing occurring in the upper river (IISG 2010). Key findings that shaped the design of our removal experiment included the need to (1) improve safety and quality of harvested bigheaded carp as food, (2) properly brand and find markets for the product [29,32], (3) provide financial support for commercial fishers, (4) form a public–private partnership to stimulate harvest, and (5) use associated data to inform an adaptive management framework [36].

We initiated the pilot-scale training, certification, and incentives-based approach to support harvest of bigheaded carp from the lower three reaches of the Illinois River in 2011 to augment the contracted fishing in the upper river (Figure 2). Names of licensed commercial fishers were obtained from the Illinois Commercial Fishing Association (ICFA), and participants were selected by lottery. The training related to (1) safe handling of bigheaded carp for consumption in foreign and domestic markets, (2) licensing and safe operation of commercial fishing vessels, (3) biosecurity practices to prevent transmission of aquatic nuisance species and pathogens, and (4) coordination and sharing of bigheaded carp harvest data with stakeholders.

In addition to receiving \$0.42 USD/kg for their catch, incentives for participating fishers included reimbursement for the cost of two ICFA memberships (i.e., fisherperson and deckhand; \$100 USD total), the annual Illinois commercial fishing license fee (\$35 USD), and net tags required for commercial fishing in Illinois waters (\$250 USD). Furthermore, participating fishers received \$1000 USD to offset fuel costs after harvesting 22,680 kg of bigheaded carp and \$3000 USD to offset gear purchase/repair/replacement costs if they harvested a total of 45,359 kg. For their catch to be eligible for incentives, participating fishers had to report the date/time, location (using a provided handheld GPS), and species composition of their catch. Data had to match information recorded on processor receipts. Only bigheaded carp caught from the Illinois River were considered, and all fish had to be sold to processors for human consumption.

#### *3.2. Set-Quota Harvest Approach*

The quota commercial fishing strategy ("market-side") program was conducted in spring 2012, whereby a set-quota fishing effort was implemented to explore the biological and ecological effects of increased bigheaded carp harvest. Following a competitive bidding process, a third-party logistics company (Select Logistics Network, Inc., Clinton, IL, USA) and a local fish processing facility (Big River Fish Company, Pearl, IL, USA) were selected to coordinate the harvest and processing of bigheaded carp at a price of approximately \$0.42 USD/kg to fishers, from the lower three reaches of the Illinois River in order to yield up to 453,600 kg of dried fish meal. The bigheaded carp removed through this approach were not eligible for incentives and were processed into fish meal by Protein Products, Gainesville, FL, USA.

#### *3.3. Field and Processor Subsampling*

Fisheries are typically highly selective for sizes, ages, and species and thus may have unique impacts on the populations. We visited processing plants approximately every two weeks during the period 1 February 2012 through 8 May 2012 while contracted fishing was occurring. In order to characterize bigheaded carp population demographics within the commercial catch during each biweekly sample, we randomly selected up to 100 silver carp and 100 bighead carp that were harvested from each of the lower three reaches of the Illinois River (depending on availability) and recorded total length (TL) and weight data. Post-cleithra were removed from up to five fish per species per 50 mm length group per river reach for age determination.

We sampled the Illinois River using standardized, fishery-independent sampling to compare bigheaded carp population metrics. We recognized that the sampling was insufficient to detect a fishing impact in such a short time with limited effort. However, we were able to assess annual variability in stock and potential harvest selectivity. Sampling occurred in August 2011 before contracted harvest began and August 2012 after harvest concluded. Sampling was conducted along the main channel of the Illinois River at four fixed locations within each of the three lower pool reaches, as well as nearby backwater lakes and side channels (Figure 2). Pulsed-DC electrofishing transects (Smith-Root GPP 5.0 electrofisher; 15 min each), with two netters, were conducted along each main channel and backwater site during the day. Captured fish were euthanized by immersion in 300 ppm tricaine methanesulfonate (MS-222) until opercular movement ceased. All fish were weighed and measured (total length), and post-cleithra were collected from a subsample of ten silver carp per 10 mm length group per reach; due to smaller sample sizes, post-cleithra were removed from all collected bighead carp.

To determine age, post-cleithra were sectioned transversely across the center with a 1.5 amp diamond-blade low-speed isomet saw (Buehler, Lake Bluff, IL, USA) following Johal et al. [37]. Two independent readers used side illumination from an MI-150 fiber optic light (Dolan-Jenner Industries, Boxborough, MA, USA); if disagreement between readers could not be resolved, the sample was omitted. Age distributions were developed for the entire dataset using an age–length key [38].

Electrofishing catch per unit effort was compared between 2011 and 2012 for silver and bighead carp by reach and for all reaches combined using paired *t*-tests with each site being treated as the unit of replication. All statistical analyses were conducted using SAS 9.2. (SAS Institute, Cary, NC, USA). An alpha level of 0.05 was used to judge statistical significance.

#### **4. Results**

#### *4.1. Training, Certification, and Incentives Approach*

Although the fishers harvested bigheaded carp from the Illinois River (Figure 2), the "fisher-side" data collection and fish removal goals of the 2011 incentives approach were not fully achieved. Despite the promise of incentive payments, commercial fishers infrequently reported data such as fishing location (GPS coordinates). Of the ten commercial fishers enrolled in the incentives program, only three fully participated, providing GPS locations and dates related to their fishing efforts. These individuals harvested a combined total of 225,372 kg of bigheaded carp from the Illinois River, and received a total of \$8000 USD in incentive payments. Although they technically fulfilled the obligations of the incentives program, review of the data revealed the GPS coordinates provided were typically locations of boat access ramps from which they launched, not the precise locations of harvest. This provided little detailed information other than from which river reach fish were harvested. Harvest effect in the Illinois River was likely minimal due to the low harvest quantities.

#### *4.2. Set-Quota Harvest and Processor Subsampling Approach*

The "market-side" fishing approach, with a goal of harvesting enough bigheaded carp to yield an open order of 453,600 kg of dried fish meal, was successful in meeting removal and data goals. Between 25 January 2012 and 11 June 2012, commercial fishers harvested 805,878 kg of bigheaded carp from the Alton reach, 223,910 kg from the La Grange reach, and 276,559 kg from the Peoria reach, for a total of 1,306,346 kg, which yielded enough dried meal to meet the contracted goal.

#### *4.3. Bigheaded Carp Standardized Sampling Catch Rates Pre- and Post-Harvest*

Mean silver carp electrofishing CPUE for the three lower reaches of the Illinois River combined had a 2011 rate of 100.4 fish/h (SE = 14.6) and 2012 rate of 81.0 fish/h (SE = 17.4). This difference was not significant (t17 = 1.60; *p* = 0.128). Silver carp mean CPUE declined from 2011 to 2012 in the Alton reach by nearly half (t5 = −3.77; *p* = 0.01; Table 1), but was not significantly different for the La Grange reach (t6 = −1.53; *p* = 0.18) or the Peoria reach (t4 = 0.33; *p* = 0.76; Table 1). Mean bighead carp CPUE was not different from 2011 to 2012 for the lower three reaches of the Illinois River combined (t19 = 1.12; *p* = 0.28) or among reaches (t6 ≤ 1.05; *p* ≥ 0.34). Overall bighead carp CPUE was 2.9 fish/h in 2011 (SE = 2.3) and 0.3 fish/h in 2012 (SE = 0.2) among all reaches.


**Table 1.** Standardized densities (mean fish electrofished per hour; catch per unit effort, CPUE) of silver carp in each reach pool of the Illinois River before (2011) and after (2012) harvesting occurred.

#### *4.4. Size and Length at Age*

Comparisons of length frequency distributions and length at age from commercially harvested and electrofished silver carp in 2011 and 2012 demonstrated the size selectivity of commercial gears. The length frequency histogram comparing commercial harvest to standardized sampling showed a bimodal distribution of commercially caught silver and bighead carp (Figure 3). Examination of the age frequency of silver carp showed over 25% of commercially harvested silver carp were age 5 and older, compared to just over 7% of silver carp collected during standardized sampling (Figure 4), with harvested fish having larger length at age than fish in the standard samples.

**Figure 3.** Percent frequency of total lengths of (**a**) silver carp and (**b**) bighead carp sampled with standardized gear (upper panels) and harvested by fishers (**c**,**d**) in the lower Illinois River during 2012.

**Figure 4.** Percent frequency of ages of (**a**) silver carp and (**b**) bighead carp sampled with standardized gear and harvested by fishers (**c**,**d**) in the lower Illinois River during 2012.

#### **5. Discussion**

Our "experimental" approach toward stimulating invasivorism in the lower Illinois River yielded important information about how to effectively and economically remove large quantities of bigheaded carp and potentially other invasive species. There are three examples for comparison.

First, the completely subsidized, decade-long contracted removal program in the upper Illinois River is an example of one extreme where market demand for fish does not drive harvest, although harvested bigheaded carps are acquired for free by processors. As we noted earlier, this approach has likely prevented density increases in the upper river, although societal, taxpayer costs are high and bigheaded carp densities continually rebound as fish immigrate from downstream [17]. Along the supply chain model we developed, this approach is completely decoupled from fish availability, market needs, or economic costs and benefits of fishing. Because processors obtain fish at no cost, there is no need to develop a consistent demand for fish. Without contracted removal at the range edge, populations will build in the upper river, greatly increasing the probability of movement toward the CAWS and potentially into Lake Michigan and the other Great Lakes.

The second approach we conducted at the "fisher side" of the model was considered unsuccessful. In the lower Illinois River, when it came to enlisting fishers to assist in the collection of data, the incentive levels did not substantially increase fishing, with our participants not complying. Non-compliant fishers may have placed a greater value on their proprietary information (i.e., fishing locations and methods) than what they could gain from the incentives program. Providing incentives without confirming information about fishing location, effort, and species composition may lead to potential misinformation about the source of the fish, compromising the efficacy of control programs where targeted harvest is necessary. Any successful fisheries management program requires significant buy in and compliance by fishers [39], and this was not accomplished with this approach. In summary,

for this incentives program to work, stakeholders need to work together to reduce impediments to compensation and information sharing, which will require considerable effort, time, and resources.

The third approach we conducted at the "market side" of the conceptual model was considered a success, with our removal goal achieved within 6 months. Creating a fish meal demand using a logistics company to arrange transport and processing allowed fishers to quickly deploy and meet the market price we set. The processor was required by contract to ensure that the source of catch by fishers was recorded accurately, allowing us to match catch data with field-derived sampling. Presumably, this scenario shows that, if demand increased with prices similar to those we set (\$0.42 USD/kg), consistent fishing would occur. At the time, prices for bigheaded carp commodities were approximately half of what we set (\$0.26 USD/kg) for fish intended for rendering or other industrial purposes (personal communication, Lisa McKee, Big River Fish Co.; personal communication, Gray Magee Jr., CEO, American Heartland Fish Product LLC, Grafton, IL, USA). Fishers also likely preferred the increased income (i.e., they were paid directly by the processor) over the paperwork and documentation necessary to meet incentive benchmarks set in the "fisher-side" approach (i.e., funding required application).

The experimental large-scale removal approaches we instituted during the first half of 2012 were at that time the primary market for bigheaded carp on the Illinois River. The State of Illinois had contracted with China to export over 13.6 million kg annually for direct consumption [40]; however, this volume was not exported due to the lack of fish processing infrastructure and logistic difficulties in transportation and distribution. At the time of the project, "market side"-driven harvest appeared to stimulate the expansion of the bigheaded carp fisheries in the region, at least temporarily. Big River Fish Company subsequently relocated to a larger facility in Griggsville, IL. Upon reopening in 2013, they purchased significant amounts of bigheaded carp for direct consumption markets in China (personal communication, Lisa McKee, Big River Fish Company). Despite this promising start, this processor appeared to have closed permanently by 2020. Schaefer Fisheries in northern Illinois (Thomson, IL, USA, Figure 2) purchased bigheaded carp for domestic direct consumption markets, and for use in liquid organic fertilizer [41]. Direct consumption purchases by this processor have declined since a fire burned the original building in 2015 and reduced capacity for manufacture of bigheaded carp food products such as hotdogs and extruded meats [41,42]. A rendering facility operated by American Heartland Fish Products LLC opened in Grafton, IL near the confluence of the Illinois and Mississippi Rivers in May 2014 (Figure 2). This facility was processing up to 27,000 kg of bigheaded carp per day into fish oil and fish meal until noxious odor violations led to its closure (personal communication, Gray Magee Jr., CEO, American Heartland Fish Product LLC). Other processors and distributors have opened in the region since the early 2010s, with varied success and capacity [32]. Despite continued interest among stakeholders in elevating market demand for bigheaded carp from the Illinois River and other inland waters, harvest has fluctuated rather than increased in the lower Illinois River since 2010 (Figure 5). Likely contributing to the lack of increased harvest, market prices have remained relatively unchanged since the inception of this project [32].

Population modeling suggests that, to deplete bigheaded carp populations in the Illinois River, all age classes must be targeted for removal [10]. Although a large amount of bigheaded carp biomass was removed in 2012, no large-scale removal of bigheaded carp under 500 mm total length occurred. Based on processor subsampling and informal surveys, we determined that a limited number of commercial fishers appeared to use seines or other gears which would harvest all sizes of bigheaded carp. The majority of fishers used gill, trammel or hoop nets that target larger fish. Not only were commercial fishers harvesting larger, older fish, compared to standardized sampling, but they were also harvesting the largest fish within younger age classes. This is supported by the lack of a bimodal distribution in the age-frequency histogram of commercially harvested fish. As an alternative to the collapse approach, the SEAcarP model suggests that market-based fishing in the lower river will effectively reduce densities of carp near the edge of the invaded range in the upper river [24]. This allows for more effective contracted removal efforts at the range edge and reduces the probability of a breach through barrier systems.

**Figure 5.** Total biomass in kg of bigheaded carp harvested from the lower Illinois River during the 2010–2018 period, derived from Illinois Department of Natural Resources records [43].

#### **6. Conclusions**

Bigheaded carp standing stock and biomass will vary among river reaches and years as a function of variable recruitment, emigration, immigration, harvest impacts, and environmental conditions. In our study, silver carp densities in the Alton Pool did decline, whereas they did not in the other reaches, perhaps because Alton Pool was in proximity to the contracted processor and received greater than 60% of the total harvest. However, without control reaches, more intensive sampling, and multiple years, it is impossible to infer a direct, causal relationship. With sufficient market demand, commercial harvest may control bigheaded carp. However, lack of processing infrastructure and supply chain bottlenecks could constrain harvest, particularly at low commodity prices. Whether a commercial harvest approach to fighting bigheaded carp will be successful can only be assessed over time. Any such evaluation must consider that once a nuisance species becomes an economic resource or a part of local culture, it may no longer be considered a nuisance, but an asset. This could result in pressure to maintain the species or even expand its range to uninvaded regions [9,44], causing a paradox for managers trying to restrict or prevent the spread of certain invasives or mitigate their effects on native species and ecosystems (see Settle et al. [45], and a related discussion on socioeconomic feedbacks of invasive lake trout control in Yellowstone Lake [3]). Given the large, geographical scale of this invasion and complicated harvest logistics, concerns about economic dependence on invasivorism that encourage stock enhancement are likely unmerited.

While harvest-driven extirpation is unlikely, more realistic reduction and control benchmarks might be achieved, although the nature of such benchmarks has yet to be fully articulated. Identifying bigheaded carp density thresholds that would lessen their impacts on native ecosystems or reduce the risk of density-dependent upstream movement toward the Great Lakes or other uninvaded regions are options. Once population goals are determined, management agencies must monitor market prices relative to population densities. If harvest can achieve significant decreases in bigheaded carp density, flexibility in the fishery to move to other river reaches where bigheaded carps are abundant or switch to alternative stocks would be necessary to maintain stability. It is likely that higher (possibly subsidized) prices for bigheaded carp would be needed to compensate for the greater effort necessary to maintain

harvest levels in a declining abundance scenario. Nonetheless, partially subsidized fishing may still be a more cost-effective, efficient, and publicly acceptable means of bigheaded carp control. Agencies must be prepared to provide the economic flexibility and stability necessary in a widespread, complex region such as the Mississippi River basin to ensure that removal by harvest remains robust to maintain control through time without creating unintended dependencies.

**Author Contributions:** Conceptualization, W.W.B., D.C.G., J.T.T., S.S., J.E.G., R.M., and K.I.; methodology, W.W.B., D.C.G., J.T.T., S.S., and J.E.G.; software, W.W.B., D.C.G., and J.E.G.; validation, W.W.B., D.V.G., J.T.T., S.S., J.E.G., R.M., D.P.C., A.A.C., K.I., and A.W.; formal analysis, W.W.B. and J.E.G.; investigation, W.W.B.; resources, W.W.B.; data curation, W.W.B. and A.W.; writing—original draft preparation, W.W.B.; writing—review and editing, W.W.B., D.C.G., J.T.T., S.S., J.E.G., R.M., D.P.C., A.A.C., K.I., and A.W.; visualization, W.W.B.; supervision, J.E.G.; project administration, J.E.G.; funding acquisition, J.E.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Illinois Department of Natural Resources.

**Acknowledgments:** We would like to thank the staff and graduate students that assisted with data collection and processing, Paul Hitchens for his assistance with logistics and harvest data compilation, and Kristen Bouska for creating the map figure and helpful manuscript edits. The findings and conclusions in this article are those of the author(s) and do not necessarily represent the views of the U.S. Fish and Wildlife Service.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*

## **Numeric Simulation Demonstrates That the Upstream Movement of Invasive Bigheaded Carp Can Be Blocked at Sets of Mississippi River Locks-and-Dams Using a Combination of Optimized Spillway Gate Operations, Lock Deterrents, and Carp Removal**

**Daniel Patrick Zielinski 1,\* and Peter W. Sorensen <sup>2</sup>**


**Abstract:** Invasive bigheaded carp are advancing up the Upper Mississippi River by passing through its locks-and-dams (LDs). Although these structures already impede fish passage, this role could be greatly enhanced by modifying how their spillway gates operate, adding deterrent systems to their locks, and removing carp. This study examined this possibility using numeric modeling and empirical data, which evaluated all three options on an annual basis in both single LDs and pairs under different river flow conditions. Over 100 scenarios were modeled. While all three approaches showed promise, ranging from 8% to 73% reductions in how many carp pass a single LD, when employed together at pairs of LDs, upstream movement rates of invasive carp could be reduced 98–99% from current levels. Although modifying spillway gate operation is the least expensive option, its efficacy drops at high flows, so lock deterrents and/or removal using fishing/trapping are required to move towards complete blockage. Improved deterrent efficacy could also offset the need for more efficient removal. This model could help prioritize research and management actions for containing carp.

**Keywords:** integrated pest management; model; hydraulic; acoustic deterrent; invasive fish; conservation

#### **1. Introduction and Mini-Review**

The spread of invasive fish has contributed to the extirpation of many species of fish as well as a loss of biodiversity and ecosystem integrity across the globe [1–3]. When eradication is not possible, as is almost always the case [4,5], containment is the only option [2,3]. In rivers, containment can be complicated by the presence of migratory native fishes and flooding. Developing ways to selectively control the upstream movement of invasive fish has challenged North American fisheries managers since the turn of the 19th century, when the common carp, *Cyprinus carpio*, and sea lamprey, *Petromyzon marinus*, [6–8] became abundant. Only a few solutions have been identified, and none for large rivers where testing options are expensive and difficult. These complexities make numerical simulations of control options a valuable tool. Here, we use numerical models to evaluate three control options for invasive bighead carp, *Hypophthalmichthys nobilis*, and silver carp, *H. molitrix*, (collectively known as bigheaded carp) at the locks-and-dams (LD) they must pass to move upstream in a large river. Our findings describe several promising ways that a targeted and integrated approach can effectively control an important invasive fish. In this introduction, we review the bigheaded carp problem, Mississippi River LDs, and three ways to control bigheaded carp at these choke points; we then outline our study objectives and approach before proceeding to the methods.

**Citation:** Zielinski, D.P.; Sorensen, P.W. Numeric Simulation Demonstrates That the Upstream Movement of Invasive Bigheaded Carp Can Be Blocked at Sets of Mississippi River Locks-and-Dams Using a Combination of Optimized Spillway Gate Operations, Lock Deterrents, and Carp Removal. *Fishes* **2021**, *6*, 10. https://doi.org/10.3390/ fishes6020010

Academic Editor: Maria Angeles Esteban

Received: 13 March 2021 Accepted: 24 March 2021 Published: 26 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

#### *1.1. The Bigheaded Carp Problem*

Recently, bighead carp and silver carp from Asia have become a serious problem in the Mississippi River Basin of North America [9]. Bigheaded carp were introduced to Arkansas from Asia in the 1960s, escaped into the Mississippi River [10] and continue to invade the upper reaches of the Mississippi River Basin. These species are large (>20 kg), microphagous filter-feeding fish that compete with native planktivorous fish for food, driving reductions in their abundance, size and condition, while altering food webs [11–13]. Additionally, silver carp can jump up to 3 m out of the water, interfering with recreational boating [14]. Bigheaded carps reproduce in areas of flowing water and have semi-buoyant eggs that require long stretches of flowing water to hatch and recruit, making the pools between LDs a good place to control and remove adults because LDs restrict fish movement [15–17]. Carp are also sensitive to sound, making them susceptible to being blocked with acoustic (non-physical) deterrent systems [18–20]. Finally, bigheaded carps are not particularly strong swimmers [21], so their movement through LDs is open to manipulation, especially in systems with multiple LDs that create impassible water velocities.

Bigheaded carp presently comprise the majority of the fish biomass in many areas of the Mississippi River Basin, although they have yet to establish themselves in either the headwaters of the Mississippi River or the Laurentian Great Lakes. While carp passage into the Great Lakes is currently protected by an electrical barrier in the Illinois River [17], the headwaters of the Mississippi River remain unprotected because they are wide and prone to flooding and thus cannot support a simple electrical barrier, so new approaches at LDs are sought.

#### *1.2. Mississippi River Locks-and-Dams*

The Upper Mississippi River (UMR) is regulated by a series of 29 LDs operated by the US Army Corps of Engineers (USACE) and are named (and numbered) in a sequential fashion from north to south (Figure 1). Nearly all LDs have both a navigational lock and a gated spillway system. The USACE operates these structures in a manner that permits navigation while protecting the structures from erosion/scour by limiting water velocities. Spillway gates are seated at the bottom of the river and progressively raised to pass water and regulate water depth, but in so doing, create water velocities underneath them that fish may struggle to overcome. As LD spillway gates are lifted, the velocity of water passing underneath them is reduced, dropping to a minimal value when/as they come out of the water entirely (a condition known as "open-river"). In contrast, flow in navigational locks is very low to allow boats (and fish) to pass, but access is regulated by miter gate opening and the locks are a relatively small (~10%) part of most dams. Together, spillway gates and locks inhibit upstream fish passage. However, their effects on fish vary: some LDs exert large effects on fish passage and some very little—depending on their design, local river conditions, spillway gate operations (e.g., the number, location, and opening height of each gate), and the fish species (fish swimming ability varies greatly) and their size.

Several LDs whose spillways rarely open fully are known to greatly reduce upstream fish passage of native migratory species including lake sturgeon, *Acipenser fulvescens*, and paddlefish, *Polyodon spathula*, [22–24] as well as invasive species including both bigheaded carps [25] and common carp [26]. Notably, the swimming abilities of carps are very similar when size is considered [21,26]. While some migratory fishes have disappeared from the Upper Mississippi River (UMR) since LDs were installed, analyses of the current fish population structure suggest that LDs likely have little effect on the remaining populations of native fishes [27], although their effects on newly arriving invasive carp appear quite substantial. The abilities of fishes to surmount spillway gates varies with environmental conditions that include water velocity, water temperature, fish species, fish size, and physiological condition, LDs that experience open-river conditions less frequently are more likely, on average, to impede upstream fish movement [28]. Many LDs in upper regions of the UMR experience open-river conditions far less often than those in the lower portion of the river (Table 1, [29]). While some LDs have overflow systems that operate during high

flow (Table 1), many do not, or they could be screened, and thus these LDs can be used in carp control.

**Figure 1.** Location of locks-and-dams (LDs) in the upper portion of the Upper Mississippi River (UMR). See Table 1 for details on LDs.

Importantly, LDs influence each other and the fish that pass through them, synergizing the ability of each to impede overall fish movement upstream; although this has not been well studied. Further, it is likely that adjacent (consecutive and proximate) LDs could have greater influence on bigheaded carp populations more than other LDs separated by great distances because bigheaded carp require 50–100 km of turbulent open river to reproduce successfully [30]. Of course, short pools (50–100 km) also create excellent opportunities for fisheries managers to sample, catch, and remove carp that might pass the LD immediately below them.

#### *1.3. Options to Control Carp Passage at Locks-and-Dams*

Three good options exist to control carp at LDs: the spillway gates, the lock, and the pools above LDs into which fish must pass and where capture is possible. Of these, the spillway gates are of singular importance because they typically comprise 90–95% of the structure size and are at least partially open most of the time. Adjusting spillway gate openings is a good option to reduce carp passage. Its potential has been shown by both modeling [28] and descriptions of fish passage from fish tracking studies [26,31–33], the latter showing a strong correlation between spillway gate opening, water velocity, and passage. Numerical modeling at two relatively typical Mississippi River LDs, LD 2 and LD 8, has shown that fish passage through their gated spillways is dependent on hydraulic conditions that include velocities that exceed 5 m/s at lower gate openings through which very few fish can pass [26,28,34]. Further, we have developed a numeric fish passage model (FPM) that uses three-dimensional water velocities found around LD spillways gates to determine whether and/or how fish with known swimming abilities can (and do) swim through gates with different settings and river flow [28]. FPM simulations have also shown that the spillway gate operations presently used by the USACE can result in slightly unbalanced flow regimes at LDs, and thus create regions of low velocity that fish (carp) can swim through. Remarkably, this validated FPM describes ways ("optimized operating conditions") that spillway gate settings can be re-balanced to reduce carp passage, sometimes by as much as 50–75% [28,34]. As these modifications reduce scour, they have proven to be acceptable to the USACE [28]. Thus, modifying/optimizing spillway gate operations to balance water velocities at LDs when they are not in an open-river condition has great potential to restrict upstream carp passage at little to no cost.

**Table 1.** Summary of locks-and-dams (LDs) in the Upper Mississippi River (UMR). The percent time spent in open-river was calculated from historical records between 1970–2000 [29]. Δ River km is the distance (pool size) between that LD and the next one upstream [29]. The Upper and Lower Saint Anthony Falls Dam (upstream of LD 1 and lacking an operational lock) and Chain of Rocks Lock (downstream of LD 26) differ structurally from LDs 1–26 and are not included. The final column indicates whether the lock-and-dam has an additional uncontrolled overflow spillway that functions during high flow conditions. Consecutive LDs that experience open-river conditions less than 5% of the time are shaded. Two LDs that do not go into open-river (0%), because they do not have spillway gates are also shown (LD 1, LD 19).


A second option to control carp passage at LDs is to add deterrent systems to the lock chambers. LD lock chambers are designed to support barge navigation and thus have little measurable flow, making them well suited to these systems. Upstream fish passage through open lock chambers has been observed in the summer months for a number of fishes, including bigheaded carps [25,26,31]. Non-physical deterrent systems that use sound, or sound paired with other stimuli (i.e., air bubbles, strobe lights, carbon dioxide), are presently being developed for use in these systems [18,35–42]. Sound is favored because it is safe and, similarly to all ostariophysians, bigheaded carp have a wider hearing range and lower hearing threshold than many native fish. Laboratory tests using a variety of sound signals [37–41] and sound coupled with air-bubble curtains [36,38] have documented deterrent efficiencies between 75–97%. A test of a cyclic sound coupled with an air curtain and light (a bio-acoustic fish fence or "BAFF") blocked 95% of all carps in a creek, but further testing is required [42]. The effects of sound could be taxon-specific.

A third option to control bigheaded carp is fish removal in pools upstream of LDs. Removal is especially feasible in short pools where sampling to gauge effectiveness is reasonable and bigheaded carp may also be unable to reproduce. Carp removal could be achieved through subsidized targeted removal or possibly commercial ventures [43,44]. In

the Illinois River, contracted harvest of bigheaded carp has been used successfully since 2010 to help reduce propagule pressure on the USACE electric barrier at Chicago [45]. Bigheaded carp are typically removed using short-set large-mesh gill and trammel nets. The gear used in the Illinois River selects for larger fish, and removal has been effective at decreasing the density of bigheaded carp populations restricted by a downstream LD [43,44,46].

#### *1.4. Introduction to This Study*

In the present study, a stochastic size-structured fish passage model (S-FPM) was developed to examine the potential for controlling bigheaded carp passage by blocking upstream passage using different combinations of modified gate operations, acoustic deterrence at navigational locks, and carp removal across pairs of consecutive UMR LDs. This model simulated passage of carp to examine ways it could be reduced. It examined many options at both single and consecutive LDs using known carp passage rates, monthly river discharge, several levels of lock deterrent systems, fish size, and different levels of removal. Our overarching goal was to determine whether and how an integrated approach to control bigheaded carp might be reasonable in the UMR and if so, what factors might best contribute to its efficacy. To address this, we asked several related questions: (1) What gains might be realized by managing bigheaded carp passage at two adjacent LDs versus just one?; (2) What benefits might be realized by modifying spillway gate operations at one or two LDs to reduce carp passage at different river flows?; (3) What are the benefits of adding a non-physical deterrent(s) to either individual or pairs of LDs and how do they compensate for increased fish passage at high flow?; (4) What additional benefits might carp removal schemes have on carp control?; and (5) How might these three options be employed together as part of an integrated pest management scheme? We focused on silver carp as it is the species of greatest concern and worked in sequential fashion, combining factors as we went to examine synergistic effects at varying river flows, the effects of which on spillway gate passage are complex but important. Lessons from silver carp should nevertheless apply to bighead and common carp as they have similar biologies. Possible effects of carp population size-structure and behavioral drive to attempt spillway gate passage were also examined. We use changes in fish passage rate as our metric, given the absence of data on silver carp population size in the upper reaches of the UMR.

#### **2. Methods**

The S-FPM was created to simulate and estimate annual upstream passage rates of bigheaded carp through either one or two LDs in the UMR. This model included 6 categories of variables including: (1) whether a single or a paired set of LDs is being managed; (2) local environmental variables (e.g., river flow); (3) carp population sizestructure; (4) carp behavior/passage route; (5) carp passage rates at spillway gates and locks (and effects of deterrents on them); and (6) estimated effects of carp removal on overall passage rate (Table 2, see below). Over 100 scenarios were modeled using empirical data from LD 8, a relatively typical UMR LD (Table 1) which has 15 spillways gates and 1 operational lock. First, we describe the model, then the variables it uses, and then how it was deployed.

#### *2.1. Model Framework*

The S-FPM evaluates fish passage as a consequence of a series of junctions at either a single or two consecutive LDs (i.e., pairs of LDs with one located immediately upstream of the other so they synergize each's actions). While doing so it uses fish movement rates and route selection (i.e., the path fish pursue while swimming upstream) informed by both field data and fish spillway passage indices for LDs calculated using our fish passage model (FPM). Specific variables used in the S-FPM model include (Table 2): environmental conditions; fish population and size-structure; fish behavior—upstream movement and route selection; fish behavior—passage indices and deterrence; and carp removal. Fish passage at LD spillway gates is considered using our FPM which considers fish swimming performance with respect to species and size, as well as water velocities at specific LD spillway gates as informed by LD structure and river flow using computational fluid dynamics (CFD) [28]. When possible, silver carp data (e.g., swimming performance) were used but when not available, data were used from the closely related common carp (e.g., passage rates through spillways gates of LD).

**Table 2.** List of stochastic size-structured fish passage model (S-PFM) variables and values. Where available, variable mean +/− standard deviation is provided. Variables are categorized by italicized section headings and further described in the methods. Data derived from common carp are noted with an (\*), otherwise data come from silver carp. *Q* is river discharge. *P*\_*AUS*, *P*\_*BUS* are the proportions of fish that move upstream at LD A and LD B. *P*\_*AL*, *P*\_*BL* and *P*\_*AS*, *P*\_*BS* are the proportions of upstream swimming fish to atempt passage through the lock and spillway gates at each LD, respectively. *PIlock*, *PIspill* are the passage indices at the lock and spillway gates. *At* is the number of passage attmpts made per month at the spillway gates. *D* is the efficiency of a deterrent system inhibiting passage through the lock chamber. *R* is the proportion of fish removed from the intermediate pool.


The S-FPM model employs two LDs (LD A and LD B) and they have the same spillway gate operations, a realistic scenario because most UMR LDs have nearly identical structural components (Figure 2). LD 8 is used as the base conditions for each, which is reasonable because its design is typical of most LDs and it is also well studied [28]. Both LD A and LD B are associated with pools: Pool A is downstream of LD A, Pool B is located between LD A and LD B, and Pool C is located upstream of LD B. In the model, carp start in Pool A. and the S-FPM calculates passage rates of carp moving from Pool A to Pool C each month for a year (which thus includes seasonal effects). Each month a proportion of fish moves upstream (*P*\_*AUS*, *P*\_*BUS*) and then attempt to pass through one, or both LDs. While doing so, each upstream swimming carp is assigned to one of three routes: the spillway gate (*P*\_*AS*, *P*\_*BS*), the navigational lock (*P*\_*AL*, *P*\_*BL*), or both spillway and lock (*P*\_*AS*+*L*, *P*\_*BS*+*L*). The combined route of spillway and lock gives fish the opportunity to pass through either the lock or spillway (a scenario observed at LD8 [33]), while the other routes limit to just one route. The likelihood of passage through the lock chamber is modelled using mean passage rates of common carp observed at LD 8, while passage through the spillway gates has been determined using the fish passage index (FPI) previously calculated by Zielinski et al. [28]. Individual carp that pass either route (*P*\_*Apass*, *P*\_*Bpass*) then move into the upstream pool and those in Pool B are subjected to the passage model again whereas those in Pool C remain upstream of LD B. Fish that do not either move upstream or attempt to do so and are blocked by either LD A or LD B's

spillway gates (*B*\_*AS*, *B*\_*BS*) or lock chamber (*B*\_*AL*, *B*\_*BL*) return to their pool of origin and undergo the passage model the subsequent month (if/when the model simulation allows for future attempts- we tested 1–5 attempts). Those carp that pass LD A and are found in Pool B are also then subject to possible removal (*R*). River flow (i.e., discharge), proportion of upstream movement, route selection, and passage indices were updated monthly in the model. The total number of fish from each size class within each pool was recorded monthly and divided by the initial population size to determine the proportion of fish passing each LD (the percent). Finally, the number/proportion of carp eventually found in Pool C represents the proportion that passed both LDs while the combined proportion of fish in Pool B and Pool C reflect the proportion passing a single LD (LD A). The model was coded in Matlab (Mathworks, MA, USA) (Figure 2).

**Figure 2.** Schematic representation of the stochastic size-structured fish passage model (S-FPM). The model uses a silver carp population with five size classes that are released in Pool A (downstream).

#### 2.1.1. Environment

Opportunities for fish passage at LDs vary with spillway gate openings and these were determined by river discharge. The S-FPM model examined 4 hydrologic scenarios in the UMR and which we describe as monthly exceedance values. Monthly exceedance discharge is equal to the median monthly discharge that is exceeded for some percentage of the time. Exceedance was calculated from 30 years of river discharge at LD 8; we identified 50%, 25%, 5% and 1% exceedance discharge values [47]. In our case (LD 8), a 50% and a 25% exceedance discharge condition does not require the spillway gates to be fully open anytime during the year, while the 1% exceedance discharge condition requires LD 8 to operate in open-river conditions 7 months of the year (Figure 3).

#### 2.1.2. Fish Population and Size Structure

Each simulation used a population of 200,000 numeric silver carp from 4 size classes (50,000 carp per size class). This number was selected to minimize variance between model runs (the variance was calculated to be less than 0.5% for each size class at 50,000 fish). A size-structured approach was used because swimming performance, is influenced by fish size [21]. Each run of the model was initialized with a population of carp being placed into Pool A, which was assigned a body length from one of four 100 mm size classes based on data from either the UMR or Wabash River where carp have been established longer and are larger [48] (Table 3). Because most silver carp in the UMR have a total length of less than 600 mm, a size class whose swimming abilities are not known, the size distribution of carp used in the model was adjusted so that the smallest carp was 600 mm. This likely led to conservative (artificially high) estimates of passage rate as small fish cannot swim as fast as larger fish. For model simulations, the proportion of each size class of carp found within each pool was multiplied by the length-frequency percentage of a given population distribution (Table 3) to produce relevant size-specific results. The UMR population size-structure was used as the default in the S-FPM reported in the results although the impact of fish size-structure on the model was calculated for reference (see Supplemental data, Figure S1).

**Figure 3.** Monthly discharge at LD 8 based on 50%, 25%, 5%, and 1% exceedance durations between 1972–2000 [47]. Open-river conditions start when discharge is greater than 2718 m3/s.

**Table 3.** Length-frequency distributions of silver carp (by 100 mm length increment) from the Upper Mississippi River (UMR) and Wabash River [48].


\* 73% of the UMR silver carp population has a total length ≤ 500 mm.

2.1.3. Fish Behavior—Upstream Movement and Route Selection

Telemetry studies have shown that upstream movement rates of carp vary seasonally [49,50] and that carp take different paths through LDs [26,31,33] with carp moving upstream more vigorously in the spring than in summer and fall. Our S-FPM used seasonal upstream movement rates, and assumed fish did not move between November and February (Table 2). The proportion of each size class within each pool that was selected to move upstream (*P*\_*AUS*, *P*\_*BUS*) was randomly assigned from a normal distribution with a mean equal to the mean upstream movement measured by Coulter et al. [49] with a standard

deviation of 25% of the mean. All individuals were then assigned a movement indicator (*R*1) from a uniform random distribution (0–100) each month. In Pool A, individuals with *R*<sup>1</sup> ≤ *P*\_*AUS* moved upstream to challenge LD A. Any individuals that passed LD A were then assigned a new movement indicator once they entered Pool B and the selection process repeated itself.

Just as different numbers of bigheaded carp could move upstream (or not) in the river and our model, they could also choose different paths or routes, with some carp following the river's edge to encounter a lock, others moving to the center of the channel and encountering a spillway gate, and others demonstrating a mixed approach that included both options. Our model considered these three possibilities using available data. An ongoing study using acoustic telemetry is assessing the movement and passage of common carp at LD 8 [33] and we used its findings. Briefly, data collected in 2019 from over 100 transplanted, tagged common carp downstream of LD 8 found that 7.3% of all adult common carp approached only the lock chamber, 27% approached only the spillway gates, and the remainder explored both options. These values were employed and the proportions of upstream moving carp selected to move towards the lock chamber (*P*\_*AL*, *P*\_*BL*) or the spillway gates (*P*\_*AS*, *P*\_*BS*) were randomly assigned from a normal distribution with a mean and standard deviation derived from the common carp data collected by Whitty et al. [33]. All individuals moving upstream were assigned a route indicator (*R*2) from a uniform random distribution (0–100) each month. In Pool A, individuals with *R*<sup>2</sup> ≤ *P*\_*AL* attempted to pass through the lock chamber and individuals with *R*<sup>2</sup> ≥ 1 − *P*\_*AS* attempted passage through the spillway gates. All remaining, unassigned fish attempted passage through both the lock chamber and spillway gates. Individuals passing LD A are assigned a new route indicator and the route selection process repeated for LD B.

#### 2.1.4. Fish Behavior—Passage Indices and Deterrence

The likelihood of any fish (carp) passing through a lock chamber is dependent on a combination of opportunity and behavior. In contrast, the likelihood of them making through spillway gates is driven by opportunity, behavior, and swimming performance. Both were modeled for a single LD and consecutive LDs. First, we discuss passage rate at the spillway gates, then locks.

#### Spillway Gate Passage

Fish (carp) passage through spillway gates is dependent on several variables including fish species, size, behavior, and gate opening/water velocity (i.e., gate operations). To estimate the likelihood of carp passing through LD spillway gates, we used the fish passage indices (FPI) we developed earlier [28] for silver carp at LD 8. Briefly, the FPI was calculated using a FPM which pairs high-resolution water velocity data at specific gate settings with known fish swimming performance data to predict if, when, and where fish could pass through a hydraulic structure assuming the fish follows the path of least resistance (a conservative assumption) [28]. This FPM and its resultant FPIs have been validated in tracking studies of common carp at LD 2 [26] and LD 8 [33].

To create estimates of spillway passage, the S-FPM used FPI values [28] to assign a spillway passage index (*PIspill*) at both LD A and LD B that was based on river discharge and fish length (we used data from the UMR and another location, see below). We calculated FPI for silver carp assuming both base (current/historical) gate operations and gate operations modified and optimized to restrict carp passage for five river discharges including open-river conditions [28]. We used linear interpolation to calculate the spillway passage index at intermediate discharges and the nearest value for discharges outside of the range [28] (Figure 4). Individuals assigned to the spillway route were then assigned a spillway passage indicator (*R*3) from a uniform random distribution (0–100) each month. In Pool A, individuals with *R*<sup>3</sup> ≤ *PIspill* successfully passed through LD A spillway gates while all remaining fish were blocked. Any individuals that passed LD A and were assigned to the spillway route were assigned a new spillway passage indicator and the spillway passage process repeated. The spillway passage index calculated for the S-FPM included an attempt variable (*At*) that allowed fish to challenge the spillway gates multiple times per month. Based on the average number of attempts observed by silver carp at Starved Rock Lock-and-Dam [25], the model assumed each fish following the spillway gate route was assigned 2 passage attempts per month. Over the 8-month period of our model, any given fish could attempt to pass through the spillway gates up to 16 times. Simulations using 1 and 5 attempts per month were also run to evaluate how attempt rate impacts passage estimates (Supplemental data, Figure S1).

**Figure 4.** Spillway passage index for silver carp with a total length of 600–900 mm at LD 8 under base and modified gate operations [28]. The solid black line indicates the mean passage index and shaded area is the standard deviation. The passage index is calculated at 635, 1250, 1475, 2325, and 2720 m3/s (open-river). Note, the different y-scales for each total length.

#### Lock Passage and Deterrents

Fish can only swim upstream through a lock chamber when a boat is locking through it and its miter gates are open. To do so, fish must enter the lock chamber, an area of high noise and turbulence, and their success in passing appears to be low. For instance, the rate of passages relative to the number of passage attempts was found to be 7% for silver carp at LD 26 [31] and 5% for common carp at LD 8 [33]. In our model, the lock passage index (*PIlock*) at each LD was randomly assigned from a normal distribution with a mean and standard deviation from empirical data collected by Whitty et al. [33] and Tripp et al. [31]. As reported, passage rates [31,33] were measured relative to the number of passage attempts, so the lock passage index does not need to explicitly simulate multiple passage attempts through the lock chamber (i.e., passage rate is expected to be ~5% regardless of the number of attempts). Individuals assigned to the lock chamber route were assigned a lock chamber passage indicator (*R*4) from a uniform random distribution (0–100) each month. In Pool A, individuals with *R*<sup>4</sup> ≤ *PIlock* successfully passed through the lock chamber while all remaining carp were blocked. Individuals passing LD A were then assigned a new lock chamber passage indicator and the lock passage process repeated at LD B.

Of course, base passage rates through locks can, in theory, be reduced by adding deterrent systems to them. We included the possibility that a deterrent will be developed and successfully implemented for use in LD(s) in our model. Due to uncertainty in the specifics of the deterrent type and efficacy, we examined the impact of adding deterrents at one or both locks with several efficiencies: 0%, 25%, 50%, 75%, and 100%. Deterrent values were based on those already measured in the field and laboratory for acoustically based systems [36–42].

#### Lock and/or Spillway Passage

Finally, our model allowed for the possibility that some carp will attempt to move upstream using a combination of both locks and/or spillway gates (e.g., fish assigned to the spillway + lock chamber route). Each month these individuals were assigned both a lock chamber and spillway passage indicator. Similar to fish assigned to just the spillway route, fish were allowed multiple attempts to pass the spillway gate per month (if appropriate). If passage criteria were satisfied for either the lock chamber or spillway gates, that carp was deemed to have passed that LD.

#### 2.1.5. Carp Removal

Physical removal of fish is commonly used to control populations of invasive species [43,51]. This approach is already being successfully employed in the Illinois River to control bigheaded carp using contracted commercial fishers [43,45,46]. Simulations using the Spatially Explicit Asian Carp (SeaCarP) model estimate 40% of the population needs to be harvested to reduce the risk of introduction into the Great Lakes, and it is possible this is presently being achieved in some areas [17] where the population seems to be constant. Several fishing techniques have been developed for this purpose and are still being improved including the "Modified Unified Method" from China [17]. We included the possibility of removal in our model as *R* (removal) and assign it values 0%, 5%, 10% and 40%. Each month, all individuals that move into Pool B were assigned a removal indicator (*R*5) from a uniform random distribution (0–100). Individuals with *R*<sup>5</sup> ≤ *R* were then removed from the population. The likelihood of removal was the same for all sizes of fish in Pool B.

#### *2.2. Model Simulation*

Over 100 simulations were run to assess the individual and combined impacts of modified spillway gate operation, lock deterrence, and removal on silver carp passage rates through single and consecutive LDs (Table 4). For each simulation, we tracked the number of fish passing both LD A and LD B individually and the annual proportion of fish passing each structure was calculated by dividing the total passed by the initial population size. The proportion of carp passing LD A was the total passage rate expected at one LD and the proportion passing LD B is the total passage at two consecutive LDs. Modeling proceeded in 4 steps so we could systematically evaluate the role of different variables in a step-wise fashion with each variable (management action) being added to the previous case. We started by exploring the roles of the simplest management option, modified gate operation. First, passage rates during either base (current as determined from USACE historical records) or modified spillway gate operations to block silver carp were calculated and then compared at different flow (exceedance) scenarios. Second, the impact of adding non-physical (acoustic) deterrent(s) with several efficiencies to LD lock(s) were examined using modified spillway gate operations. Third, the impact of employing carp removal in the intermediate pool (Pool B) on overall annual passage was examined in combination with varying levels of lock deterrence, including none and assuming modified spillway gate operations. All cases used the carp size structure measured in the UMR distribution [48] while carp were allowed to attempt to pass twice a month, per expected values. After completing these runs at different flow (exceedance) conditions, we examined the average annual effects of several combinations of variables across all exceedance values expected in a year. We did this to evaluate the overall effects of individual variables. Finally, we assessed the impact of population size structure and spillway gate passage attempt rate assuming modified spillway gate operations, no lock deterrence, and no removal (Supplementary data, Table S3). A total of 104 simulations were run to accommodate all

iterations over four hydrologic scenarios (Table 4), the results of which are presented in Supplemental data (Tables S1 and S2).

**Table 4.** List of unique model simulations. Brackets indicate range of values used in each simulation. Each simulation provides the annual proportion passing either a single or pair of two consecutive LDs.


#### **3. Results**

#### *3.1. Effects of Managing Carp Using Consecutive LDs*

Our model suggested that approximately 18.1% of silver carp of the size presently found in the UMR can be expected to pass a single typical LD under base (historical) spillway gate operating conditions during the course of a typical year with this rate increasing to 22.4% at high flows (Figure 5, Table S1). When two LDs were considered instead of a single LD, this rate dropped by 85% across all simulated flows to approximately 2.7% (Table S1). The effects of managing carp at two adjacent LDs locations were multiplicative.

**Figure 5.** The proportion of silver carp passing a single LD (black lines) or two consecutive LDs (grey lines) under base spillway passage conditions (solid lines) or modified spillway gate conditions as calculated by our FPM (dashed lines).

#### *3.2. Effects of Managing Carp by Modifying LD Spillway Gate Operations*

Modifying spillway gate function at a single LD had notable effects, reducing the proportion passed by approximately 11% at an exceedance of 50% for one LD but dropping to only 2% at an exceedance of 1% when the river is mostly in open-river conditions (Figure 5, Table S2). When the effects of modifying spillway gate operations on passage through consecutive LDs was considered, the overall proportion passing two LDs decreased by about 88% to an overall value of only 1.5–3.7%. Notably, while consecutive LDs may be expected to go into open-river at similar times, they were unlikely to be identical and if the distance between them small, reproduction may be unlikely. Modifying spillway gate operations was thus especially beneficial at pairs of LDs that rarely go into openriver, but other options probably should be considered for the later scenario at higher flows (exceedances).

#### *3.3. Effects of Adding a Non-Physical Deterrent to One or Both Locks*

Adding a deterrent to the lock chamber of one or both LDs operating their spillway gates in a modified manner was very effective, especially when pairs of LDs were considered (Table 5, Figure 6 and Table S2). At a single LD, lock deterrence systems that were more than 50% effective reduced the number of silver carp that could pass to less than 10%. If a deterrent with 100% efficacy was used, the value dropped to 2% at the 50% exceedance flow, and to less than 10% at the 1% exceedance flow when gate operations were modified (Table S2). When two LDs were considered, each with a deterrent in the lock, the annual proportion of silver carp passing was less than 2% for a deterrent only 50% effective overall under all deterrence levels and modified gate operations. Notably, the relative impact of a lock deterrent on fish passage was relatively unaffected by flow conditions.

**Table 5.** Summary of the estimated effects of pairing LDs, modifying their gate operation, adding a deterrent to the lock chamber and removing carp in the intermediate pool between them on the overall annual passage rates of the silver carp population with the size structure presently found in the UMR [48]. The annual proportion of carp passed is averaged across all four flow scenarios. Percent reduction is calculated relative to the proportion passed at a single LD (1 LD) under base gate operations conditions.


**Figure 6.** The proportion of upstream swimming silver carp passing through either a single LD (black lines) or two consecutive LDs (grey lines) equipped with nonphysical deterrents of different efficacies and using modified spillway gate operations.

#### *3.4. Effects of Carp Removal in the Intermediate Pool*

Adding carp removal to a control scheme while utilizing modified gate operations and a deterrent had additional effects on reducing passage. Effects were multiplicative with a removal rate of 40% without lock deterrence reducing overall annual passage by 93% compared to a single LD with base spillway gate operations (Figure 7, Table 5 and Table S2).

**Figure 7.** Proportion of silver carp passing two consecutive LDs equipped with non-physical deterrent systems with different efficacies and whose intermediate pool was subjected to carp removal.

#### *3.5. Overview of the Averaged Combined Effects of Multiple Management Options*

Lastly, we calculated average annual carp passage rates when all exceedance values were considered. These showed that when pairs of adjacent LDs were considered, only 3% of all carp attempting to pass can be expected to do so with 2 attempts, versus 20% at 1 LD (Table 5). Modifying gate operations drops this value to 2.5% (88% drop from one LD with base spillway operations). If a 50% effective deterrent is added to two LDs the average value decreased to 1.3% and if the deterrent increases to 100% effective, the proportion passed drops to 0.6% (a 97% decrease vs. nothing occurring at one LD, the current situation). The addition of carp removal together with lock deterrents had the greatest impact on reducing silver carp passage rates. The best-case scenario reduced silver carp passage to only 0.4% and required 100% lock deterrence paired with 40% removal in the pool (Figure 7). Notably, several levels and types of carp removal and lock deterrence achieved the same level of passage reduction. For example, the annual passage rate at consecutive LDs could be reduced to less than 1% by pairing 10% removal rate with a lock deterrent with as little as 50% efficacy, even when exceedance values approached 1%. If the deterrent was close to 100% effective, values decreased by about half again (Table 5).

#### **4. Discussion**

Our simulations demonstrate that upstream passage of invasive silver carp in the UMR can be reduced to only 1–2% of current rates through an integrated approach that uses consecutive LDs and some combination of three tractable control techniques. These include reducing passage using spillway gate adjustment, adding non-physical deterrents to lock chambers, and removing carp from the intermediate pool. While modification of the spillway gate operation could occur with no modification to infrastructure, both lock deterrents and carp removal are likely to be costly, although they do not need to be highly efficient (i.e., 50% efficacy might suffice) to drive over 90% reductions in carp passage. Remarkably, carp control appears possible even during high river flows with an approach that employs pairs of strategically selected LDs. All of the control measures we describe can be implemented.

We believe that our simulations are reasonable because they are based on empirical data (ex. exceedance values, known gate settings, velocities, fish passage routes and swimming abilities) and a validated fish passage model that was designed to provide conservative overestimates of actual passage [28]. It is also promising that silver carp telemetry data suggest this species does not challenge LDs repeatedly [31,32]. The recent documented movement of significant numbers of adult bigheaded carp through both LD 19 [52] and LD 8 [53] attests to an urgent need to reduce bigheaded carp passage rates below the conditions currently existing at LD 8. A 50% reduction in passage rates seems possible using a single control option, while a 90% or greater reduction to an overall rate of just 2% appears attainable if both a deterrent and carp removal is used, even during times of high flow and need only be moderately effective (25%). Previous suppositions that carp can only be stopped at systems that lack operating gates [32] appear overly simplistic, which is important because only 2 of the 29 LDs in the UMR do not have bottom mounted spillway gates.

The most significant finding of our study is likely that bigheaded carp should not be managed at single LD, as has been the practice, but at pairs of LDs close to each other that rarely experience open-river conditions. Fortunately, three such locations exist in the UMR: LD 14–15, LD 7–8, and LD 4–5 (see management section below) (Table 1). Across all hydrologic scenarios, the cumulative impact of adding a second LD resulted in an average decrease in carp upstream passage of 85% compared to passage at a single LD. These LDs need to be located close to each other (50–100 km) to be effective, prevent spawning, and facilitate monitoring as well as possible removal.

Likely our next most important finding is that modifying LD spillway gate operations to reduce passage can be highly effective on an annual basis and would come at little cost because the predictive models have been developed and validated [26,28]. Simply modifying gate operations at a single LD decreases carp passage by about 8% overall. Multiplicative effects are expected if operations are optimized at two locations. Importantly, the modifications to gate operations we propose are safe for navigation and LD structural integrity as they do not induce additional scour [28]. While promising for both carp

control and LD operations, the benefits of modified gate operations are restricted to the period when LD(s) are operating under controlled conditions (e.g., non open-river), so additional control options such as adding a lock deterrent and removing carp must be considered. Notably, the S-FPM model results we describe are conservative and estimate the upper limit of passage rates owing to our conservative assumption of fish behavioral drive and our assumption that carp can find the most efficient way upstream [28]. The hydrologic scenarios we considered were also conservative because the possibility of average monthly discharge surpassing the 1% exceedance flow for 12 consecutive months is low. For example, the 1% exceedance discharge conditions that would require LD 8 to operate in open-river conditions for 10 straight months, or 83% of the year, actually occur less than 5% of the time [29].

The third most important finding of this study is that a single approach to controlling carp is unlikely to suffice: an integrated approach is needed. Together the three options we described synergize with each other's activities, especially at times of high flow when passage through the spillway gates is high. By using all three options, none of them needed to be singularly effective. Ideally, three options would be implemented but two might suffice if used strategically.

The addition of non-physical deterrent systems to LDs had a notable effect on overall system efficacy that persisted during high river flows even if not highly effective. Typically, non-physical deterrents can be expected to reduce overall annual silver carp passage by about 5% even if the deterrents are only 25% efficient, and close to 20% overall if 100% efficient, the efficiency presently suggested for a BAFF [38,42]. If deterrents were used in two locks, the effects would be multiplicative at all flows. Notably, a BAFF guides fish away from the lock openings so it could be paired with a trap to remove carp as well as capture native species below the LD for possible movement upstream (although see [27]). A BAFF operating at 100% efficiency could thus drive a removal rate of about 20%, compensating for the cost and effort of running a removal program and supplement native fish conservation. Some level of species-specificity which might permit native fish passage may also be possible with acoustic deterrents, such as the BAFF, because carp are especially sensitive to sound [19,37,38,40–42]. Other types of deterrent systems that use CO2 [36] could be considered, but the would not be species-specific. A BAFF is presently being tested at a LD on the Kentucky River and shows promise [54]. Deterrents appear likely to be a necessary component of an invasive carp control system and their continued development is encouraged.

Even modestly effective carp removal efforts would also be helpful in an integrated carp control program, especially if implemented in pools between paired, managed LDs. Removal would amplify the effects of modified gate operation and deterrents. Further, if a deterrent is not implemented, removal would be necessary, especially at times of high river flow when carp passage will be high. While the actual efficacy of carp removal is presently unknown, and numerous reports suggest it is low, it has adequately prevented the spread of adult silver carp further up the Illinois River [43,44]. Several techniques have been developed and improvements are being made to the "modified unified method" [17]. Notably, carp removal is likely to be especially effective in small pools where it would also limit possible spawning success, the ultimate objective of most fish control strategies [3]. The choice of LDs and the pool between them will be very important for removal strategies, and even modestly effective removal strategies, as low as 5%, would be beneficial. Admittedly, removing carp when there are low densities is difficult and may require use of radio-tagged Judas fish or perhaps eDNA [55,56]. Removal year-round is exceedingly labor-intensive, difficult [17], and expensive (Illinois spends more than a million dollars on this annually [53]). If less than 5% efficiency is realized in a UMR pool then a deterrent will be needed. More work on quantifiable removal options is needed. In any case, it is clear that an integrated approach using multiple control options at multiple LDs is highly desirable.

Our model also evaluated the importance of fish size on passage rates. Large silver carp, such as those found in the Wabash River appear nearly twice as likely to pass (Figure S1, Tables S3 and S4). The behavior of these fish is also important; an increased number of passage attempts significantly increased passage across all hydrologic conditions. For example, fish that attempted spillway passage 5 times per month had nearly a 2-fold increase in passage (Figure S1, Tables S3 and S4). This result is consistent with findings of others [57]. Fortunately, there is good reason to consider that the average attempt rate of bigheaded carp may not be higher than 5 attempts, although this requires study.

Our model has some notable strengths and weaknesses. Most important, as described above, our model assumptions are conservative and likely produce overestimates of passage. Indeed, they are based on empirical data and consistent with the slow upstream spread of bigheaded carp—over 10 years to pass LD 19 [52]. River flows are unlikely to be as consistently high as we modeled. Further, bighead carp are less likely to pass than silver carp based on their swimming performance [21]. Nevertheless, our model does have some uncertainties. First, we do not know the efficiencies of non-physical deterrents at LDs [38,42]. Second, the efficacy and size-selective nature of removal in rivers is unknown. Our model also does not account for fish population demographics.

#### **5. Summary**

This study clearly demonstrates that silver carp and likely other carps can be effectively (98%+) blocked at select pairs of LDs if they are operated in tandem and employ multiple approaches including modified gate operation, lock deterrents, and carp removal. These options could be used in multiple ways and need not be 100% efficient. Further information and improvement can come once an integrated control scheme is put into place.

#### **6. Management Recommendations and Future Directions**

It is reasonable to consider controlling invasive bigheaded carps at LDs in the UMR. Control strategies should employ pairs of LDs that are close to each other and rarely experience open-river conditions and at least two of the three options we have described. This could be extremely effective and economical. As the likelihood of carp passage increases with fish size, so does the chance of their reproducing, efforts should be timely. Three pairs of UMR LDs meet the criteria for successful control, but LD 4–LD 5 seem to have special promise because silver carp have not moved beyond them yet and bred, they are very rarely in open-river, Pool 5 is small, and they resemble LD 8 so their hydraulics are understood [28,34]. Ideally gate operations will be modeled and optimized. As with common carp control, developing ways to monitor carp abundance will be critical to success [4,5]. Detailed studies of carp movement around and through LDs will be extremely helpful as would further modeling efforts to improve model precision to guide carp control in UMR and elsewhere in the basin [58].

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/fishes6020010/s1, Figure S1: Impacts of population size on passage, Table S1: Carp passage rates with and without gate modifications at different exceedances; Table S2: Carp passage rates with gate modifications at different control options at different exceedances; Table S3: Carp passage rates with gate modifications and different numbers of attempts using fish the size of those in the UMR; Table S4: Carp passage rates with gate modifications and different numbers of attempts using fish the size of those in the Wabash River.

**Author Contributions:** Conceptualization, P.W.S. and D.P.Z.; methodology, D.P.Z.; validation, P.W.S. and colleagues; formal analysis, D.P.Z.; resources: P.W.S.; data curation, D.P.Z.; writing—original draft preparation, D.P.Z.; writing—review and editing, P.W.S.; funding acquisition, P.W.S. Both authors have read and agreed to the published version of the manuscript.

**Funding:** The data described in this research was funded by the Legislative Citizens Commission for Minnesota Resources.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data are available from the authors upon request.

**Acknowledgments:** We thank the Sorensen lab group for generating the data described herein as well as the USACE and USFWS and MN DNR for administrative and field help. Jeff Whitty's help is especially acknowledged.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Review* **Achieving Sea Lamprey Control in Lake Champlain**

**Bradley Young \*, BJ Allaire and Stephen Smith**

U.S. Fish and Wildlife Service, Essex Junction, VT 05452, USA; bj\_allaire@fws.gov (BJA.); stephen\_j\_smith@fws.gov (S.S.)

**\*** Correspondence: bradley\_young@fws.gov; Tel.: +1-802-662-5304

**Abstract:** The control of parasitic sea lamprey in Lake Champlain has been a necessary component of its fishery restoration and recovery goals for 30 years. While adopting the approach of the larger and established sea lamprey control program of the Laurentian Great Lakes, local differences emerged that shifted management focus and effort as the program evolved. Increased investment in lamprey assessment and monitoring revealed under-estimations of population density and distribution in the basin, where insufficient control efforts went unnoticed. As control efforts improved in response to a better understanding of the population, the effects of lamprey on the fishery lessened. A longterm evaluation of fishery responses when lamprey control was started, interrupted, delayed, and enhanced provided evidence of a recurring relationship between the level of control effort applied and the measured suppression of the parasitic sea lamprey population. Changes in levels of control efforts over time showed repeatedly that measurable suppression of the parasitic population required effective control of 80% of the known larval population. Understanding the importance of assessment and monitoring and the relationship between control effort and population suppression has led to recognition that a comprehensive, not incremental, approach is needed to achieve effective control of sea lamprey in Lake Champlain.

**Keywords:** population suppression; lampricide; fishery restoration; sea lamprey; Lake Champlain

#### **1. Introduction**

Sea lamprey (*Petromyzon marinus*) parasitism is a limiting factor to both the restoration [1] and recovery [2] of fish populations in Lake Champlain. The preferred host species of the lake include lake trout (*Salvelinus namaycush*), land-locked Atlantic salmon (*Salmo salar*), and lake sturgeon (*Acipensir fulvescens*). While sea lamprey do parasitize other species in the lake, the parasitic load on these species and the level of induced mortality place sea lamprey more in the role of a predator than parasite. The origin of sea lamprey in Lake Champlain has been the subject of debate. A series of genetic studies [3–5] concluded that they were endemic to the lake and likely remnants of the Champlain Sea, when the basin was contiguous with the Atlantic Ocean in following the last glacial event approximately 10,000 years ago. Eshenroder [6,7] challenged the assumptions of the genetics models using historical collections and canal construction timelines to propose that sea lamprey entered Lake Champlain through the New York State canal system when it joined the Hudson River to Lake Champlain through a series of connections during the end of the 19th century. Regardless of origin, Atlantic-native sea lamprey have proven to be a nuisance species in Lake Champlain and incompatible with its freshwater hosts.

Lake Champlain was historically home to lake trout and landlocked Atlantic salmon populations [8–10]. During 19th-century industrialization, the damming of tributaries and deforestation degraded riverine habitat [9]. This loss of habitat in concert with over-exploited fisheries led to the extirpation of native stocks of both species between approximately 1850 and 1900 [10]. A programmatic effort to restore these native species and introduce other salmonids began in 1973 with the formation of the Lake Champlain Fish and Wildlife Management Cooperative (Cooperative); the Cooperative comprises the

**Citation:** Young, B.; Allaire, BJ; Smith, S. Achieving Sea Lamprey Control in Lake Champlain. *Fishes* **2021**, *6*, 2. https://doi.org/ 10.3390/fishes6010002

Received: 6 December 2020 Accepted: 28 December 2020 Published: 26 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

New York Department of Environmental Conservation, the Vermont Fish and Wildlife Department, and the U.S. Fish and Wildlife Service. In 1977, the Cooperative set goals to reestablish a lake trout and Atlantic salmon fishery and establish a rainbow trout (steelhead) fishery by 1985 [11]. As efforts to improve the fishery moved forward, parasitic sea lamprey populations surged. It became clear to the Cooperative that meeting fishery restoration and recovery goals would require efforts to suppress sea lamprey population.

In developing a program to control Lake Champlain sea lamprey, the Cooperative followed the existing Laurentian Great Lakes (Great Lakes) model [12,13] in establishing three fundamental management components. First, basin-wide assessments determine densities and distributions of larval sea lamprey and direct selection and implementation of control efforts. Second, as part of an integrated pest management approach, both chemical and physical control methods target larval and adult life history stages. Because the larvae of Lake Champlain consistently spend four years maturing in tributaries before emigrating to the lake as parasites, four year classes can be eliminated effectively once every four years using lampricides (selective piscicides) applied to tributaries and their associated deltas [14]. The active ingredient of the liquid and bar formulations of lampricide applied to rivers is 4-nitro-3-(trifluoromethyl)phenol (IUPAC nomenclature) and commonly referred to as TFM. The active ingredient of the lampricide applied to deltas in a granular formulation and occasionally applied to rivers in a liquid formulation as a synergist with TFM is 5-chloro-*N*-(2-chloro-4-nitrophenyl)-2-hydroxybenzamide (IUPAC nomenclature) and commonly referred to as niclosamide. All formulations of these lampricides are restricted-use pesticides and manufactured solely for application by designated federal and state government agencies. While manufacturers have refined product formulations at times, the two active ingredients used for controlling sea lamprey have remained the same for the entirety of the control program.

Second, physical control methods such as dams, temporary barriers, and screens serve to block and trap migrating adults before they reach habitat suitable for spawning [15–17]. The program benefits from dams on ten tributaries (labeled 1, 2, 4, 15, 16, 17, 19, 21, 24, 25; Figure 1) built for purposes other than lamprey control where they serve to limit the length of river accessible to adult sea lamprey migrating upstream to spawn. The program uses temporary, seasonally-installed barriers on seven tributaries that block adult sea lamprey during their spring spawning season (April–June), but are removed for the other nine months of the year (labeled 10, 11, 19, 20(2), 22, 26; Figure 1). These temporary barriers include traps which allow adult sea lamprey to be removed and killed and other aquatic species to be removed and passed above the barrier. The effectiveness of physical control methods in Lake Champlain varies from 100% with large hydropower dams to occasionally 0% with small temporary barriers subject to failure when overcome by high water events. When feasible, lampricides are a more effective, reliable, and consistent method of control. However, especially where lampricide use is restricted, physical control methods have a role in the program.

Third, we monitor and evaluate control efforts using a wounding rate index [18,19] to track changes in the frequency of lamprey parasitism on host species of interest. The wounding rate index is not a direct measure of parasitic lamprey abundance. Characteristics of host species and their population dynamics affect it in ways that are difficult to quantify [20,21]. Despite the limitations of the wounding rate index to provide direct point estimates of abundance, its consistent and standard usage over time [18,22] in Lake Champlain and the Great Lakes provides opportunities to compare general trends in the relationship between sea lamprey and host abundances.

The population dynamics of sea lamprey and the effort necessary to suppress their population has been studied and modeled by Great Lakes researchers to develop and refine their control program [23]. However, understanding the stock recruitment relationship of sea lamprey has proven challenging because of the uncertainty associated with density independent recruitment and compensation from density dependent survival [24–26]. Defining and establishing consistent or standardized levels of control effort required to

achieve desired levels of lamprey population suppression have therefore also proven difficult. While long-term successful suppression of populations has been satisfactory for decades, Jones and Adams [27] propose that population eradication remains possible. We share these interests and seek to add experience from the Lake Champlain sea lamprey control program to further the understanding of how lamprey populations respond to increasing levels of control effort. The smaller scale of Lake Champlain and availability of a 30-year data set present an opportunity to consider these dynamics in ways that may lead to new insights as lamprey control efforts continue to evolve.

#### **2. Management Phases**

When the Cooperative evaluated progress toward its fishery restoration goals in 1985, approximately half of the lake trout and Atlantic salmon collected were found to be the target of sea lamprey parasitism as measured using the standardized wounding rate index [18,22]. Experience from the Great Lakes and Lake Champlain fishery data showed that efforts to restore these salmonid species would not be successful without suppression of the sea lamprey population. In 1990, the Cooperative began an 8-year experimental control program (ECP) under the guidance and in coordination with the Great Lakes program. At that time, assessments documented larval lamprey in 19 tributaries [28]. The ECP used lampricide to control populations in 13 tributaries (labeled 1, 4–8, 10, 13–15, 17, 22, 23; Figure 1) while trapping migrating adults on three others [29]. The ECP was designed as a pilot program to determine whether the model of sea lamprey control used in the Great Lakes could be applied to Lake Champlain to suppress the lamprey population. After eight years, the evaluation of both sea lamprey suppression and fishery responses led the Cooperative to pursue further and continuing sea lamprey control to support its fishery goals [29].

To transition from the ECP to a long-term control program (LTCP), a federal Environmental Impact Statement (EIS) was required. The process of writing and approval of this document took three years. Several groups opposed the use of lampricide and filed lawsuits challenging the EIS. The Cooperative ultimately received approval of the EIS in 2001 [30] and made plans to resume the control of sea lamprey in 2002 as the LTCP began. The period (1998–2001) between the ECP and LTCP has been termed the partial control program (PCP). During that time, lampricide treatments remained on schedule in New York where available state funds effectively extended the ECP there. Continued treatment of Vermont tributaries required federal funds that remained unavailable until approval of the EIS. During the PCP, the nine New York tributaries treated during the ECP remained controlled while of the four Vermont tributaries treated during the ECP, two were trapped (labeled 22, 23; Figure 1), and two were left uncontrolled (Table 1). At the time, the Cooperative believed that a reduction in control efforts during PCP would sustain some lesser level of population suppression, but would avoid surrendering all progress made during the ECP.

With the EIS in place to begin the LTCP, lampricide treatments resumed in Vermont in 2002 and continued in New York. Sea lamprey wounding rates of 25 per 100 lake trout and 15 per 100 Atlantic salmon were set as goals that the Cooperative believed could support fishery restoration goals [30] based on experience from the Great Lakes and the ECP [29]. Although the EIS enabled the LTCP to proceed, issues on individual rivers resulted in further permitting challenges. As the LTCP resumed in 2002, assessments documented larval lamprey populations in 20 tributaries in need of control [30]. Of those, nine were treated with lampricide and five were trapped (Table 1) [30]. As work progressed toward meeting the requirements for the inclusion of new and existing lamprey-producing tributaries, sea lamprey control efforts languished and wounding rates climbed higher until implementation of a more comprehensive approach. The LTCP authorized by the EIS [30] has continued to the present day. As the program progressed and incorporated experience to affect changes and improvements, the LTCP of 2020 has grown and now documents

larval lamprey in 26 tributaries, controlled presently using 19 lampricide treatments and five barriers with traps.

**Figure 1.** Lake Champlain and its 26 tributaries with currently known larval sea lamprey populations, controlled as indicated. The lake map represents the red-colored region on the inset United States map. Lamprey-producing subordinate tributaries controlled concurrently with mainstem tributaries are not included in counts.


**Table 1.** Historical levels of chemical (lampricide) and physical (barriers with traps) sea lamprey control efforts used on Lake Champlain. Lamprey-producing subordinate tributaries controlled concurrently with mainstem tributaries are not included in counts. ECP = experimental control program; PCP = partial control program; LTCP = long-term control program.

> These three differing periods of Lake Champlain lamprey control unintentionally provided insight into the relationship between the lamprey population of Lake Champlain and the effort required to suppress it. Over 30 years, lamprey densities have fluctuated in individual tributaries, populations have expanded to new tributaries, control efforts have adjusted and sometimes been delayed, and technological advancements have enabled new and improved approaches to control. When viewing these programmatic adaptations over the long term, patterns emerged that lead to a better understanding of how to successfully control sea lamprey in Lake Champlain.

#### **3. Management Review**

The Lake Champlain lamprey control program presents opportunities for reviewing both short- and long-term population responses to control efforts. The lamprey control program set forth as a management initiative, not an ecological experiment [11,28,30]. Variables were not controlled, measures were coarse, and replication exists only in the form of a time series. Despite the inability to apply statistical models or tests, some general trends and patterns emerged over time that demonstrate relationships and emphasize aspects of the program in ways that will inform managers when making decisions in years to come.

By 2005, the Cooperative began questioning why the level of control effort was not showing the same type of anticipated response in lamprey reduction seen during the experimental program. With the implementation of the LTCP in 2002, fishery managers anticipated similar positive results based on the responses seen during the experimental program. The Cooperative distilled explanations for why wounding rates reached record highs in the period 2004–2006 into three general categories that led to further investigations into the need for: (1) control of additional known sources of larval production, (2) locating and controlling unknown sources of larval production, and (3) improved lampricide treatments that reduce the number of residual (surviving) larvae. While most agreed that the reason for rising wounding rates was some combination of these three, determining where to focus efforts required more data.

At the inception of the experimental program, it was logical and appropriate to believe that because Lake Champlain had a parasitic sea lamprey problem consistent with what the Great Lakes program manages, that if the Cooperative implemented the same control techniques and methodologies as the Great Lakes successfully employed, Lake Champlain would experience the same positive results. The immediate responses seen during the experimental program did appear to validate that approach as application of standard lamprey control techniques showed an expected reduction in sea lamprey wounds (Figure 2) [29]. As time passed, it became increasingly clear that while the general approach to sea lamprey control was capable of working in Lake Champlain as it does in the Great Lakes, there were nuanced differences that had not been recognized or accounted for and undermined existing assumptions.

**Figure 2.** Sea lamprey wounding rates on lake trout and Atlantic salmon measured as number of wounds per 100 fish [18]. The vertical double lines separate the periods before lamprey control (Pre), the 8 year experimental control program (Experimental), the period of partial control (Partial), and the long-term control program (Long-term). Non-standardized collection effort among years for Atlantic salmon wounding data, prior to the Long-term program, led to grouping and averaging available data across the years 1985–1992 and 1993–1998 to approximate and reflect the time-lag responses to lamprey parasitism during the Pre and Experimental control periods, respectively. Horizontal dashed lines indicate the management goals for lake trout (25) and Atlantic salmon (15).

The three areas of concern shared a common need for more assessment and monitoring data. Enhancement of existing control actions was a simpler, more direct, a more convenient solution, and might ultimately prove necessary. However, such determinations required a more detailed understanding of the density and distribution of the larval population in the basin and site-specific measures of control efficacy. The Great Lakes program affirmed this need for enhanced assessment and its critical importance as they also placed attention on assessment in developing more effective control strategies [31,32]. Any broadly applied attempts focused on increasing existing control efforts were unlikely to address all remaining sources of lamprey production that contributed to the parasitic population. With increased attention paid, Lake Champlain assessment and monitoring developed into a more systematic approach, where quadrennial surveys provided comprehensive coverage of all tributaries in the basin for the detection of new and emerging larval populations. Implementation of standardized surveys that both preceded and followed every lampricide treatment became a permanent method for determining effectiveness. Regular surveys on

tributaries with barriers and traps verified the effectiveness of the method used at each site. The data gained from these increased assessment and monitoring efforts provided new insights and helped to isolate the reasons that the LTCP was not matching the success of the ECP.

#### *3.1. Discontinuity*

#### 3.1.1. Partial Control

During the four years of the PCP, lake trout wounding rates rose sharply from 33 in 1998 to 77 in 2002 (Figure 2). The Cooperative expected that a reduction in control effort during the PCP would result in higher wounding rates, but the resurgence of lamprey during this period to even higher wounding levels on lake trout than seen prior to the ECP was unexpected (Figure 2). A rebound effect appeared underway that partial control efforts failed to slow or lessen. While treatments continued on the nine New York tributaries treated during the ECP, the PCP did not include delta treatments previously associated with four of those tributaries during the ECP. We cannot quantify the contribution of those untreated deltas, but it amounted to further reduction in the cumulative control effort expended during the PCP. The four Vermont tributaries controlled during the ECP were not disproportionately large lamprey producers, based on larval population survey data. In fact, larval survey data indicate the nine treated New York tributaries accounted for more than a commensurate 69% of the total larval production among the 13 tributaries treated during the ECP. In light of the success of the recent ECP, expectations were that partial control efforts would produce partial population suppression and, at the very least, keep the lamprey population from returning to previous levels. Yet despite the treatments conducted in New York and attempts to trap two (labeled 22, 23; Figure 1) of the four Vermont tributaries treated during the ECP, the lake trout wounding rate incline that began during the PCP in 1998 continued to rise through the early years of the LTCP (Figure 2). The Cooperative did not track Atlantic salmon wounding data during the PCP, but once those measures resumed in 2003, they showed the same sharp increase in wounding rate as seen for lake trout (Figure 2).

#### 3.1.2. Delayed Control

After the approval of the EIS, some local citizens continued to express concern and objection to the use of lampricide to control sea lamprey on the Poultney River (Figure 1). Through engaged conversation, the Cooperative chose to negotiate an agreement to delay chemical control for five years on that river. The agreement led to the creation of Federal Advisory Committee Act (FACA) group whose charter was to work toward alternative methods to control sea lamprey that circumvented the use of lampricides on the Poultney River. The 5 year delay ended in 2007 at which time no feasible alternatives had emerged that could effectively control the larval population estimated at over 163,000 in 2006. The wounding rates for both lake trout and Atlantic salmon in 2006 had reached a record high point (Figure 2) and led the Cooperative to proceed with application of lampricide to the Poultney in 2007. Following that treatment, wounding rates that had remained elevated even after the start of the long-term program in 2002, began to decline (Figure 2). The decline could not be attributed solely to the treatment of the Poultney River because other program improvements were also underway. The program treated the Winooski River with lampricide for the first time in 2004, making it the largest treated Vermont tributary at that time. The level of control effort applied across the basin was rising which included bringing the Poultney River back into the program. The end of the Poultney 5 year delay and other initiatives begun in 2006 marked a turning point, as seen in Figure 2.

#### *3.2. Enhanced Assessment and Monitoring*

From 1990 through 2005, we evaluated lampricide treatment effectiveness primarily by counting visible lamprey mortality the day following a treatment. Those observations provided evidence of dead lamprey that validated reasonable assumptions of treatment effectiveness. However, when wounding rates remained higher than expected and without a clear cause, we questioned whether qualitative observations of dead larvae following treatments missed quantitative measures of actual treatment effectiveness. To evaluate that, we added a new regular aspect to the assessment program in 2006. The summer following each fall lampricide treatment, assessment crews began performing post-treatment assessments for comparison to pre-treatment assessments.

Post-treatment surveys provided a new way to evaluate and understand the effectiveness of treatments. We found that measurements of treatment effectiveness based on the comparison of pre- and post-treatment assessments were a more nuanced and river-specific metric than previously understood. Perhaps the most surprising finding was that when we observed relatively large numbers of dead lamprey following some treatments, we occasionally found substantial numbers of larvae that simultaneously survived those same treatments. This led to further investigations and refinements in lampricide application approach and methodology.

When looking into the reasons that some treatments were highly successful and others were not, we discovered multiple factors that contributed to varying levels of larval lamprey survival during some treatments. We understood and addressed variables affecting dose, alkalinity, pH, seasonality, and stream-specific requirements based on toxicity testing. We were also aware of and accounted for variables affecting exposure, discharge, attenuation, dilution, channel morphology, and others. We found that the ineffective treatments were not the result of program-related miscalculations or technical errors. Instead, river-specific characteristics had been missed which required applying a more nuanced control approach to each river.

After evaluating all lampricide-controlled tributaries, post-treatment assessments revealed that most treatments had indeed been successful. However, some did show a consistent presence of residual larvae following treatments. The Ausable River and Putnam Creek (labeled 7, 13; Figure 1) provide two examples of how we identified and corrected ineffective treatments. The Ausable River has a mean annual discharge of 715 cubic feet per second (CFS), making it the second-largest New York tributary to Lake Champlain. Larval lamprey population estimates have averaged more than 600,000 over the past 15 years, not including its associated delta population, thereby ranking the Ausable as the largest producer of sea lamprey in Lake Champlain. Assuming that recruitment of larvae to the parasitic population of the lake is density independent [26] and similar to that of other tributaries in the basin, ineffective treatments there yield more considerably more net lamprey production than would ineffective treatments in smaller and lower populated tributaries. This recognition reemphasized that the importance and consequences of successful lampricide treatments increased as the size of the larval lamprey population increased.

We found that two factors in the Ausable were responsible for its insufficient treatment effectiveness. One was river morphology on the day of treatment. The Ausable splits into two mouths near its terminus. Large portions of the larval population reside in each mouth. Depending on the discharge of the river on the day of treatment, or changes in channel morphology from year to year, we found that disproportionate volumes of the mainstem followed one mouth or the other. Under ideal conditions, lampricide reaches both mouths in volumes proportional to their channel volumes. When conditions are not ideal, one mouth becomes a disproportionate route for lampricide-treated water traveling downstream and leads to sub-lethal lampricide exposure for the population of the mouth receiving lower flow.

To address this in the short term, selected portions of river that received sub-lethal doses during the 2006 and 2014 fall treatments received supplemental retreatments in the following springs (2007 and 2015). Increased secondary applications of backwaters and the addition of a supplemental downstream lampricide application point, contingent on discharge at the time of treatment, also improved delivery of lethal doses of lampricide to lamprey infested habitat. These additional steps used during lampricide applications are

common, but the need to place additional application points along the river are usually obvious and arranged when first designing a treatment. The new development here was using the post-treatment assessment as a tool to identify a problem that was not otherwise recognized. For over a decade prior, presence of dead larvae following treatments served as sufficient indication of successful treatments of the Ausable. It was not until resolute postassessment surveys identified the presence of residual survivors and their locations that we were able to isolate and address the issue. When first performed on the Ausable in 2011, that post-treatment assessment revealed the 2010 treatment had been 47% effective. Since that time, following improvements to application methodology, post-treatment assessments showed that the 2014 treatment and 2015 supplemental retreatment were cumulatively responsible for raising treatment effectiveness to 94%. The 2019 post-assessment survey of 2018 treatment found that it successfully eliminated 72% of the larval population.

Putnam Creek presented a much different set of circumstances. Despite being smaller with a mean annual discharge of 80 CFS, the abundant preferred habitat of this tributary provides conditions that support a larval population consistently estimated at over 150,000 during the last 15 years. Treatment monitoring data consistently showed that lampricide concentration and other water chemistry parameters fell within the bounds of successful treatments. However, once we started post-assessment surveys, we discovered despite seeing numerous dead larvae following treatments, there were often still large numbers of residual larvae the following year. Because larvae distribute themselves and drift over time [33], identifying any specific point sources leading to treatment survival proved difficult. Following several investigations, we discovered groundwater influence was the likely source of residual larvae in Putnam Creek. A portion of the river is in an area where groundwater routinely seeps from the banks. Through an additional series of spatial measurements in the channel using a temperature probe, we found a groundwater sublayer present within the channel sediment as well. Though this groundwater was not a substantial contributor to the overall discharge of Putnam Creek and did not affect measured treatment concentrations, we believe it provided microrefugia to sediment-dwelling larval lamprey. Fresh water recharge from below the sediment water interface countered the lethal treatment concentration present above that interface during treatments resulting in a net sub-lethal and survivable exposure in that section of the tributary. We have not yet developed a way to fully negate groundwater influence that leads to treatment residuals, but detecting its presence and extent have allowed treatments to be fine-tuned to better address the specific areas now presumed to provide lamprey with refuge during treatments.

Before post-treatment assessments began in 2006, numerous dead lamprey led managers to believe treatments were successful. The examples in the Ausable River and Putnam Creek show how that assumption led to incorrect expectations that control effort equated to control effectiveness. Until we quantified and monitored control effectiveness with additional assessment effort, remaining sources of production like these two tributaries went unnoticed. There are additional tributaries in the basin found to need river-specific adjustments to treatment strategy as well. The importance of post-treatment assessments and the way they inform control effectiveness has led us to make them a standard part of the control program.

#### *3.3. Aggressive Programmatic Expansion*

As post-treatment assessments revealed sources of residual larval lamprey production in need of additional attention, the Cooperative sought to eliminate lamprey populations in additional tributaries where assessments had detected their presence. The EIS from 2001 had prescribed plans for how and where to control the population as part of the LTCP. Unfortunately, it did not include provisions for the addition of newly colonized tributaries in the future. As efforts continued to control all known sources of production, the detection of new larval populations led to the need for a federal Environmental Assessment (EA) in 2008 to authorize the inclusion of the Lamoille River and Pond Brook in Vermont and Mill Brook in New York into the LTCP [34]. Continued annual larval assessments later found

new populations of lamprey in the Little Chazy River and Rea Brook in New York (Figure 1). To keep pace with larval lamprey colonization, these two tributaries warranted production of another EA in 2018 that added them to the LTCP [35]. Following yet another detected new colonization, a third EA added Hoisington Brook in New York (Figure 1) to the LTCP in 2019 [36]. Experience from the PCP, when partial control allowed the population to expand, factored heavily into the decision to maintain an aggressive approach to addressing all sources of production. While EA's were required to expand the LTCP and deliver control at those locations, another tributary identified in the original EIS [30] showed new evidence of an emerging population. The LaPlatte River (Figure 1) did not warrant control throughout the ECP and LTCP, but when surveys showed an emerging population, the Cooperative chose to initiate lampricide treatments there in 2016.

Morpion stream and the Pike River in Québec (Figure 1) were both known sources of production, but as Canadian tributaries to Lake Champlain, they are not subject to jurisdiction under the EIS issued by the United States. Requests made to Québec provincial officials to treat both tributaries with lampricide were not successful, leaving both as uncontrolled sources of lamprey production. Through a long process of evaluating potential alternatives to using lampricide, an innovative seasonally-removable, modular screen barrier structure was designed and installed in Morpion Stream 2014. Morpion stream is approximately 10 m wide and up to 1.5 m deep at the barrier site. Each spring, prior to lamprey spawning season, seven flow-through screen modules are set into place on a concrete base laid into the sediment. Each module is composed of 5 m height aluminum frame containing a bottom-hinged screen. Each screen locks in place upright using a float barrel mechanism that lifts during flood conditions to release the top of the screen to pivot on its bottom hinge and fall flat and flush to the sediment. This feature prevents debris buildup or extreme flows from turning the flow-through screen barrier into a dam that would flood surrounding lands. When locked in place and operating, the 13-mm spaced grates on each screen block lamprey from migrating upstream, but allow the river to flow through with minimal impoundment upstream of the barrier. The barrier is also angled between banks which naturally directs sea lamprey searching for passage into a trap where they are collected. This design is a unique solution to blocking sea lamprey in a river where discharge is too high to use small-scale (channel width < 5 m) barrier solutions and where lampricide usage is prohibited. While being a smaller tributary to the larger Pike River, the larval population of Morpion Stream has been estimated as high as 135,000 and warrants control. Following installation of the barrier, larval population estimates have averaged under 50,000 with recent technical improvements expected to result in additional declines. The Pike River remains uncontrolled and is a known producer of sea lamprey. At this time, we have no options available to control lamprey there. Its size and migratory non-target species concerns preclude consideration of a barrier or lampricide. As new technologies are developed, we hope to find an agreeable form of control to use in the Pike River in the future.

#### **4. Discussion**

After 30 years of perspective since the start of the ECP, factors that influenced the longterm success of sea lamprey control in Lake Champlain have been recognized, addressed, and used to steer decisions on where to focus resources efficiently. Periods of discontinuance resulted in a disproportionate population resurgence. Insufficient attention to assessment and monitoring led to misinformed assumptions. Control techniques executed soundly and according to plan suffered from cryptic sources of unaccounted variation. The examples presented do not provide particularly novel or noteworthy management actions. The fine-tuning of sea lamprey control has been ongoing in the Great Lakes for over 70 years [12,13]. However, with 26 lamprey-producing tributaries among a watershed with 226, Lake Champlain may offer a scale where the dynamics between sea lamprey parasitism and its effects on the fishery produce detectable effects among a relatively few sources of lamprey production. With lamprey found in 450 of the 5400 tributaries of the

Great Lakes [13], changes in individual streams become less pronounced and detectable in a control program of nearly 20 times the scale. When considering the many changes to control efforts over the length of the Lake Champlain program, when they occurred, and their various effects on the lamprey population, we have formed two conclusions that we believe offer insight into sea lamprey control efforts into the future.

First, we assert that the relationship between control effort and population reduction is non-linear based on the measurements of wounding rates following changes in the control program. The wounding rate index is not a direct measure of the parasitic population and cannot be used to make empiric estimations of that relationship. Attempts to understand the relationship are therefore limited to general qualitative observations of this relationship rather than quantitative descriptive models. Even with that limitation, we believe the wounding rate data can reflect changes in trends and serve in part as a lesser surrogate measure of relative abundance to indicate when substantial changes in the lamprey population occur or are sustained over time. The relationship between control effort and population suppression appears to follow an inverse sigmoidal relationship depicted in the conceptual diagram in Figure 3. The long-term wounding rate data (Figure 2) show instances during the PCP and during the start of the LTCP when wounding rates failed to decline until additional and more effective control was administered to sources of lamprey production. If the relationship was linear, then some fractional decline should have been detected during the PCP when 69% of the tributaries controlled during the ECP continued to be treated. As the LTCP began and added tributaries to the original 13 treated during the ECP, there was an expectation of decline that failed to materialize for the first five years of the LTCP. When additional assessments and monitoring began in 2006 and led to improvements in control effectiveness, along we the resumption of delayed treatments and inclusion of new ones, the benefits of control efforts began to exceed costs as represented by point 1 on Figure 3.

**Figure 3.** Conceptual representation of the relationship between the Lake Champlain sea lamprey population and efforts to control it. Point 1 indicates where benefits of population suppression begin to exceed the costs of control efforts. Point 2 indicates the beginning of diminishing returns where the costs of additional control efforts yield limited additional population suppression benefits.

Second, we found that achieving a conceptual 50% reduction point in wounding rate requires considerably more than a 50% control effort. Thus, not only is the relationship non-linear, it also skews toward the need for a disproportionately higher level of control effort to achieve desired reductions in lamprey populations. To suppress the lamprey population into the region between points 1 and 2 on Figure 3, control efforts needed to address more than 80% of the known sources of larval sea lamprey production in the basin. That same required level of control effort appeared consistently and repeatedly during the ECP, PCP, and the LTCP (Table 1). We do not suggest that the observed percentages

constitute specific numeric management benchmarks, but we do think the long-term data reflect the existence of a threshold for required control effort, below which measurable reductions in the Lake Champlain lamprey population cannot be achieved.

Evaluation of sea lamprey control efforts is indirect where the larval and adult life history stages receive control while assessment of those efforts focuses on the juvenile (parasitic) stage. This indirect evaluation prevents immediate determinations of population suppression measures corresponding to applied effort. The cumulative effects (changes in wounding rates) of individual control efforts are also not observable until at least one year following implementation. These control program characteristics make comparisons between sea lamprey and other fish species controlled by removal and assessment of the same life history stages tenuous when looking for common relationships between control effort and population response. So while other long-term invasive fish control programs have modeled and quantified relationships between direct species removal and measured population responses [37–39], we hesitate in seeking to relate our findings to theirs because of the differences in target species life histories, niche, and control and assessment methodologies.

Relating our findings to other sea lamprey control efforts are complicated by scale and management focus. The Great Lakes program has historically used different measures and models [40–42] to prioritize their allocation of limited resources to achieve the greatest benefits across their larger scale. Lake Champlain differs in that limits to control have historically been the result of socio-political issues rather than limited resources. This difference and the 20× smaller scale enables the Lake Champlain program to control a higher proportion of its lamprey-producing tributaries. Currently, the Great Lakes regularly controls 166 of their 450 (37%) lamprey-producing tributaries with lampricides [13]. Those 166 do represent a large portion (more than 37%) of the total basin-wide larval population, yet it compares to 19 of 26 (73%) lampricide treated tributaries in Lake Champlain. Despite the apparent advantage Lake Champlain has in proportional control effort, lamprey wounding rates on lake trout in lakes Superior, Michigan, Huron, and Erie have remained under 20 since at least the year 2000 and under five in Lake Ontario since 1985 [43]. This compares to Lake Champlain lake trout wounding rates that have remained above the management target of 25 since recording began in 1982 (Figure 2). There are many presumed reasons for this [20,21], yet aside from the causes, the differences in response relative to control highlights how control effort and population responses can differ widely between two similar programs that focus on the same target species. This leads us to conclude that our specific findings may have limited applicability to Lake Champlain or similar, smaller watersheds.

#### **5. Conclusions**

The differing phases of Lake Champlain sea lamprey control over 30 years offered an occasion to evaluate trends and anomalies during periods of cessation, adjustment, and improvement. With 26 current lamprey-producing tributaries in the basin, the potential for each to exert influence on the population forces managers to remain vigilant in assessing larval population densities and distributions. It also requires validation that implemented control efforts meet management expectations. We learned that fractional efforts do not correspond to fractional reductions and that the minimum effort required to successfully control sea lamprey falls closer to the maximum end of the range. In recent years, we began referring to our approach as "comprehensive" to imply that we have come to realize the need to address all sources of lamprey production in the basin. Ignoring even a few or one source of lamprey production can negate gains that have taken years to achieve.

As sea lamprey control continues to serve as a tool to facilitate the restoration and recovery of native fish stocks in Lake Champlain, further refinement of current methodologies and the development of new approaches are both needed to ultimately meet the management targets for sea lamprey population suppression. Continued reliance on thorough larval assessment is critical to keeping pace with expanding colonization of new tributaries. We also look to shift the assessment of the parasitic population from exclusive reliance on

wounding rates to a more inclusive and direct measure of lamprey abundance used by the Great Lakes [44,45]. Having established the level of assessment and control effort required to achieve and sustain population suppression during the previous three decades of the Lake Champlain sea lamprey control program, we expect further reductions during the ensuing fourth decade to require additional approaches, not just additional effort.

**Author Contributions:** Conceptualization, B.Y., BJA., and S.S.; methodology, B.Y., BJA. and S.S.; formal analysis, B.Y., BJA. and S.S.; investigation, B.Y., BJA. and S.S.; writing—original draft preparation, B.Y.; writing—review and editing, BJA. and S.S.; visualization, B.Y., BJA. and S.S.; supervision, B.Y.; project administration, B.Y. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding. The Lake Champlain Sea Lamprey Control Program is federally funded by the U.S. Department of the Interior and the U.S. State Department. The States of New York and Vermont contribute in kind by providing staff and equipment to participate in control activities.

**Institutional Review Board Statement:** Not Applicable.

**Informed Consent Statement:** Not Applicable.

**Data Availability Statement:** Data evaluated here are products of the United States Fish and Wildlife Service, New York State Department of Environmental Conservation, and the Vermont Fish and Wildlife Department and are available upon request from the authors.

**Acknowledgments:** We would like to collectively recognize and thank the over 100 employees from the New York State Department of Environmental Conservation, Vermont Fish and Wildlife Department, and U.S. Fish and Wildlife Service, from leaders to seasonal field technicians, who all contributed to the success of the Lake Champlain sea lamprey control program over the past 30 years. We thank United States Senator Patrick J. Leahy (Vermont) for his leadership and commitment in advocating for the program over its course. The authors thank the United States Fish and Wildlife Service for supporting this programmatic review effort and allowing us the time needed to compile, review, evaluate, and publish this work. Finally, we thank the families who endure the long and unusual schedules and travel of those who work to control sea lamprey and for their continued support.

**Conflicts of Interest:** The authors declare no conflict of interest. The Federal and State government entities who funded the work presented here had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. The findings and conclusions in the article are those of the authors and do not necessarily represent the views of the USFWS.

#### **References**


## *Article* **Yellowstone Lake Ecosystem Restoration: A Case Study for Invasive Fish Management**

**Todd M. Koel 1,\*, Je**ff**ery L. Arnold 1,**†**, Patricia E. Bigelow 1, Travis O. Brenden 2, Je**ff**ery D. Davis 3, Colleen R. Detjens 1, Philip D. Doepke 1, Brian D. Ertel 1, Hayley C. Glassic 4, Robert E. Gresswell 5, Christopher S. Guy 6, Drew J. MacDonald 7, Michael E. Ruhl 7,**‡**, Todd J. Stuth 8, David P. Sweet 9, John M. Syslo 4,§, Nathan A. Thomas 1, Lusha M. Tronstad 10, Patrick J. White <sup>1</sup> and Alexander V. Zale <sup>6</sup>**


Received: 17 May 2020; Accepted: 9 June 2020; Published: 12 June 2020

**Abstract:** Invasive predatory lake trout *Salvelinus namaycush* were discovered in Yellowstone Lake in 1994 and caused a precipitous decrease in abundance of native Yellowstone cutthroat trout *Oncorhynchus clarkii bouvieri.* Suppression efforts (primarily gillnetting) initiated in 1995 did not curtail lake trout population growth or lakewide expansion. An adaptive management strategy was developed in 2010 that specified desired conditions indicative of ecosystem recovery. Population modeling was used to estimate effects of suppression efforts on the lake trout and establish effort benchmarks to achieve negative population growth (λ < 1). Partnerships enhanced funding support, and a scientific review panel provided guidance to increase suppression gillnetting effort to >46,800 100-m net nights; this effort level was achieved in 2012 and led to a reduction in lake trout biomass. Total lake trout biomass declined from 432,017 kg in 2012 to 196,675 kg in 2019, primarily because of a 79% reduction in adults. Total abundance declined from 925,208 in 2012 to 673,983 in 2019 but was highly variable because of recruitment of age-2 fish. Overall, 3.35 million lake trout were killed by suppression efforts from 1995 to 2019. Cutthroat trout abundance remained

below target levels, but relative condition increased, large individuals (> 400 mm) became more abundant, and individual weights doubled, probably because of reduced density. Continued actions to suppress lake trout will facilitate further recovery of the cutthroat trout population and integrity of the Yellowstone Lake ecosystem.

**Keywords:** adaptive management; cutthroat trout; ecosystem restoration; nonnative fish suppression; national park; lake trout; native species recovery; *Oncorhynchus*; predatory fish invasion; *Salvelinus*; trophic cascade; wilderness preserve

#### **1. Introduction**

Apex predatory fishes introduced to freshwaters of the United States Intermountain West are invasive because they can spread within lakes or through interconnected river networks and pose a high risk to native species [1,2]. Native fish species richness in this region is naturally low and made up largely of non-predatory guilds [3]. Although introduced predatory fish often provide enhanced sport fishing opportunities [4], they prey upon vulnerable natives resulting in reductions of native species abundance or complete extirpation [5–7]. Predation losses in some areas have contributed to listing of native fishes as threatened or endangered under the Endangered Species Act (ESA) [8–10]. Introduction of a novel apex predator to a freshwater ecosystem may also result in cascading changes whereby inverse patterns in abundance, productivity, or biomass of populations or communities emerge across links in the aquatic food web [1,11]. Given that invasive predatory fishes have been introduced to all large lakes and rivers in the Western United States [12–14], mitigating negative effects from these introductions is a widespread problem faced by resource managers.

Because complete eradication or containment is generally not feasible in large aquatic systems [15,16], programs have been implemented to suppress invasive fish populations and relieve predation pressure on sympatric native species or desired, introduced sportfish populations [17–19]. Suppression programs for other waters are being contemplated [20–22]. However, proposed suppression programs are often challenged by constituents of the popular nonnative fisheries that have become established [23,24] or are complicated by presence of ESA-listed species that might be harmed by the suppression actions [25]. Lack of species-selective removal methods [15] and uncertainty in outcomes of the removal programs are common obstacles. In addition, because complete, system-wide eradication of an invasive fish is probably unattainable, a long-term commitment is required to maintain suppression actions (and funding to support them) to ensure the invasive population does not rebound [26].

The natural variation in abiotic conditions and complexity of biotic interactions within large aquatic ecosystems make outcomes of conservation actions to suppress invasive fish uncertain. These uncertainties may be accounted for, however, if an active adaptive management strategy is adopted [27,28]. In taking this approach, conservation actions are treated as deliberate, large-scale experimental manipulations and the results of these actions increase knowledge about the system and decrease uncertainty in management outcomes. Alternative approaches are incorporated into the monitoring design and evaluated as experimental treatments with expectations (hypotheses) in outcomes [29–31]. Future management decisions are adjusted based on new knowledge about the resource being managed [32]. Because introduced apex predatory fish directly result in loss of prey fish and indirectly force altered, cascading interactions throughout food webs [33], their removal is predicted to allow recovery of the prey species and a return of food-web interactions and other ecosystem services to their natural state [34]. An active adaptive management approach allows for incorporation of the response uncertainties within these complex aquatic ecosystems.

Further adding to the complexities of predicting ecological responses to invasive fish removal are anthropogenic alterations within watersheds. Agriculture, cattle grazing, mining, power generation, timber harvest, and urbanization are common disturbances in the United States Intermountain West. Multiple, interacting invaders may also occur and contribute to altered ecological interactions and the complexity of responses to management actions [35,36]. Understanding the effects of predatory fish introduction and assessing outcomes that are specifically driven by suppression actions are challenging because of these concurrent, confounding factors. The majority of these challenges are minimal within large federally-protected wildlands in the United States, including national parks and wilderness areas, where habitats are strictly preserved to support fish life history, diversity, population persistence, intact food webs, and natural ecological function. Ecological recovery of populations is more likely in areas with relatively little anthropogenic disturbance and few other invaders [36] than where confounding anthropogenic factors exist. Studies assessing the long-term benefits of invasive, predatory fish suppression in protected natural areas may therefore be more informative than those in more complex, anthropogenically confounded systems elsewhere.

#### **2. Study Area and Focal Species**

Yellowstone Lake is a large aquatic system on the Yellowstone Plateau (2357 m in elevation) with a highly protected watershed (>3200 km2) located within Yellowstone National Park and the Bridger-TetonWilderness ofWyoming (Figure 1, Video S1). As such, invasive fish are the only large-scale impact *sensu* [37] on the lake; its waters remain physically and chemically pristine. Yellowstone Lake is the largest alpine (above 2000 m) lake in North America and has a surface area of 34,000 ha, 239 km of shoreline, mean depth of 48 m, maximum depth of 137 m, and volume of 1.5 <sup>×</sup> 1010 m3 [38,39]. Powerboat access is limited to only two locations, at Bridge Bay and Grant marinas, and most of the shoreline lies in protected (federally proposed) wilderness. Thermal structure of the lake is typically unstable with a weak and variable thermocline at a depth of 12–15 m during July to September (Figure A1). Surface water temperatures rarely exceed 18 ◦C [40,41]. Specific conductance is typically <100 μS/cm [42]. The lake freezes over by late December and can remain frozen until late May (Figure 2) or early June. In winter, ice about 1 m thick under deep (>1 m) snow covers much of the lake except where shallow water covers active hot springs. Roads are not cleared of snow, and access to the lake during winter is restricted to over-snow vehicles. These logistical constraints and safety risks preclude work during winter months.

The fish assemblage in Yellowstone Lake includes only two natives, Yellowstone cutthroat trout (see Table A1 for all scientific names) and the less abundant minnow, longnose dace. Ancestral Yellowstone cutthroat trout, hereafter cutthroat trout, are thought to have accessed the upper Yellowstone River and Yellowstone Lake from the upper Snake River via natural connections across the Continental Divide [43,44] following glacial recession about 14,000 years ago [45]. Cutthroat trout then evolved as the sole salmonid and dominant fish within the lake and its connected river network. During spring (May–July), cutthroat trout spawn in up to 68 tributaries around Yellowstone Lake, move downstream to spawn in the Yellowstone River below Fishing Bridge, or make long-distance spawning migrations upstream into the remote headwaters of the upper Yellowstone River [46]. Because of their lacustrine-adfluvial life history strategy [40,47,48], they transport lake-derived nutrients into numerous tributary streams [49] where they are important prey for grizzly bears [50], black bears [51], and numerous avian predators [52,53]. Although some cutthroat trout fry may remain in the natal stream for 1–2 years, most move into Yellowstone Lake within several months of hatching. In the lake, juvenile cutthroat trout are pelagic and feed on zooplankton [54]. Adults occupy the epilimnion at depths < 20 m but are most frequently found in the littoral zone where they feed on benthic macroinvertebrates and zooplankton. Because cutthroat trout are commonly found in shallow waters of Yellowstone Lake, they are preferred prey of river otters [55], osprey, bald eagles, and several colonial waterbirds [41,56–58].

**Figure 1.** Yellowstone Lake within Yellowstone National Park in Northwestern Wyoming, USA, indicating locations of long-term gillnetting assessment sites for annual lake-wide monitoring of cutthroat trout and lake trout, historical gillnetting assessment sites that were sampled for cutthroat trout (prior to 2010), tributaries visually surveyed for spawning cutthroat trout each spring, and verified lake trout spawning sites [59]. Although 14 lake trout spawning sites are known in Yellowstone Lake, others probably exist [60]. Prevailing southwest (247◦) winds and lake fetch during the autumn spawning period may preclude successful spawning along the eastern shore [61,62].

Along with their ecological importance, cutthroat trout are also historically significant. Early explorers of the Yellowstone Lake area noted their unique beauty and abundance. Soon after the establishment of Yellowstone National Park in 1872, cutthroat trout played an important role locally for subsistence, and nationally for recreation, as anglers were drawn from the Eastern United States for the angling experience [63,64]. Initial activities of the newly formed U.S. Commission on Fish and Fisheries focused on Yellowstone Lake. With the development of methods to propagate and move fish species, 310 million cutthroat trout eggs were shipped during 1901–1956 across the United States and elsewhere [65]. Other non-native fishes were introduced to Yellowstone Lake, presumably by anglers, including lake chub, longnose sucker, and redside shiner [65]. Although rarely

studied, these fishes were new additions to the food web and likely altered the aquatic and terrestrial ecosystems by feeding on plankton and macroinvertebrates [54,66] and serving as prey for piscivorous birds [56,67] and mammals [55]. There was no evidence these fishes negatively affected the native cutthroat trout [40,68,69]. Yellowstone National Park prohibited stocking non-native fish into park waters as early as 1936 [70].

By the 1950s, following half a century of liberal angler harvest and egg collections by the U.S. Bureau of Sport Fisheries, abundance of the cutthroat trout population of Yellowstone Lake was declining and showing imminent signs of collapse [40]. Numbers of cutthroat trout migrating into tributary spawning streams were declining, and lakewide angler catch rates were low. A paradigm shift in National Park Service (NPS) management then occurred to one with an ecological basis [71] resulting in a redefinition of the role of cutthroat trout in Yellowstone Lake. Following restrictions on angler harvest and closure of the egg-collection operations, the population rebounded in the 1960s and 1970s and became so abundant that >70,000 cutthroat trout were counted spawning in a single spawning tributary (Clear Creek; Figure 1) during the spring of 1979 [72]. Biologists estimated 3.5 million (95% CI: 1.9–11.2 million) cutthroat trout (>350 mm total length) inhabited Yellowstone Lake at that time, and the consumers of these fish, such as bears, otters, ospreys, and bald eagles, were numerous near the lake. The ecosystem reflected its natural, pre-Euro-American condition [71].

**Figure 2.** Yellowstone Lake in Yellowstone National Park on 26 May 2019. Yellowstone Lake is the largest alpine (above 2000 m) lake in North America and has a surface area of 34,000 ha, 239 km of shoreline, mean depth of 48 m, maximum depth of 137 m, and volume of 1.5 <sup>×</sup> 1010 <sup>m</sup><sup>3</sup> [38,39].

#### **3. Predatory Fish Invasion and Initial Management Response**

The perception of Yellowstone Lake as a secure refuge for cutthroat trout changed abruptly on 30 July 1994, when a nonnative lake trout was caught from the lake by an angler on a guided fishing trip [73]. Additional lake trout were caught soon afterwards causing grave concern, because their potential to negatively affect native trout had previously been well-documented in other large lakes in the Western United States (e.g., Lake Tahoe) [6,7]. An NPS press release dated 11 August 1994, described the discovery of lake trout in Yellowstone Lake, outlined ecological consequences that could

result from establishment of this highly piscivorous, invasive fish species, and offered a US\$10,000 reward for information leading to the arrest and conviction of the person(s) responsible for illegally stocking the fish. The NPS immediately implemented a must-kill regulation to prevent angler-caught lake trout from being returned to Yellowstone Lake alive. An illegal stocking of lake trout was assumed because natural movement into Yellowstone Lake from waters of the upper Snake River (in which they had previously been established) was not thought possible [74]. Regardless of the mode of introduction, lake trout were present and were already well on their way to establishing themselves as a new apex predator in Yellowstone Lake.

The native range of lake trout in North America includes Alaska, Canada, the Great Lakes, and parts of New England [75]. In their native range, lake trout fill an important ecological niche as an apex predator in food webs of lakes [76] and support valuable fisheries [77]. Lake trout are a deep-water dwelling, cold-adapted (<10 ◦C) predatory species that do not serve as an ecological substitute for cutthroat trout in Yellowstone Lake. Lake trout spawn within the lake and do not use tributary streams, making them inaccessible to native piscivorous avian and terrestrial wildlife. Additionally, lake trout can be extremely long-lived (30+ years, if unexploited) [78], grow longer than any other charr, can weigh more than 27 kg [44], and are capable of capturing prey at least half their body length (Figure 3) [79]. Lake trout in Yellowstone Lake mature at an earlier age (males age 4) than other populations in the Western U.S., probably because of their fast growth rates [7]. Fecundity is high, with a 5-kg female capable of producing 6000–8000 eggs in a single spawning event [79,80]. In addition, the early life history (pre-recruit) survival of lake trout in Yellowstone Lake is estimated to be 4–6 times greater than in their native range (discussed below) [81].

**Figure 3.** Total lengths-at-age (*Lt*) of native cutthroat trout (YCT; n = 1350) and invasive lake trout (LKT; n = 6387) collected by gillnetting on Yellowstone Lake during 1998 to 2019, with von Bertalanffy growth functions. Asymptotic mean lengths of cutthroat trout and lake trout were 622 mm (95% CI: 600–648) and 883 mm (95% CI: 869–898), respectively. Cutthroat trout with a maximum age of 14 years had a mean predicted length of 604 mm and lake trout with a maximum age of 25 years had a mean predicted length of 854 mm

The NPS convened a panel of experts from throughout the United States and Canada to assess the consequences of lake trout presence in Yellowstone Lake in 1995 [82]. At that time, in the absence of knowledge of the behavior and habits of lake trout in Yellowstone Lake, the primary recommendations of the science panel were to develop a program for limiting their expansion coupled with careful monitoring and application of adaptive management strategies [83]. The 1995 panel concluded that, despite a high level of uncertainty, the probability of eliminating lake trout was low and that the introduced predator would reduce the cutthroat trout population in Yellowstone Lake (Figure 4). At the same time, the group suggested that lake trout abundance could, with a high degree of probability, be limited by initiating an aggressive control program using gillnetting. Because complete eradication of lake trout was unlikely, a long-term commitment would be required to control lake trout abundance. It was agreed that the cutthroat trout population would decline even if lake trout could be suppressed, but a lake trout suppression program could reduce the expected loss of cutthroat trout by 50% or more. Most of the information needed to increase the effectiveness of initial control measures could be obtained from the control program itself, but some modification of the existing mid-September cutthroat trout gillnetting assessment program would be required to also evaluate changes in the lake trout population [83].

**Figure 4.** Twelve cutthroat trout from the stomach of a lake trout (approximately 3 kg) gillnetted immediately following ice-off from Yellowstone Lake in May 2017. During 8 months of the year (mid-October through mid-June) there is no thermal cause for separation of lake trout from cutthroat trout. Predation pressure by lake trout under the ice, which has never been studied, is likely high.

Lake trout suppression began in 1995 primarily by control (targeted) gillnetting at depths > 20 m to avoid cutthroat trout, and secondarily by experimental (exploratory) gillnetting designed to gain information on lake trout distribution seasonally throughout the lake. By 1999, two key spawning areas in the West Thumb of Yellowstone Lake had been identified by telemetry—Carrington Island and Solution Hump—where high concentrations of lake trout occurred during autumn (Figure 1). By 1999, nearly 15,000 lake trout were killed by gillnetting, with progressively more netted each year despite limited NPS resources (Figure A3). Retrospective population modeling estimated that 10,000 or more lake trout were probably in Yellowstone Lake in 1994 when they were first discovered [79]. Several years of bioenergetics research provided estimates of lake trout predation on the cutthroat trout population [79,84], and it became apparent that a dedicated program was required to curtail further lake trout population growth. However, enough funding to support such an effort had not been obtained.

Yellowstone National Park biologists then successfully competed for and were awarded an NPS Natural Resource Management grant in 1999 specifically for the development of a lake trout suppression program in 1999–2001. The supplemental funding allowed hiring a seasonal NPS crew dedicated to suppressing lake trout. With these new resources and redirected park funds, the park developed a comprehensive program for suppression of lake trout that consisted of intensive control gillnetting, gillnetting for annual monitoring, and focused gill netting of spawning lake trout. This program resulted in the removal of more than 340,000 lake trout by 2008. However, as the enhanced suppression program continued, gillnet catch rates continued to increase, suggesting that lake trout population abundance was continuing to increase and the population was expanding spatially across Yellowstone Lake. In addition, long-term monitoring of the cutthroat trout population suggested a concurrent decline to levels lower than ever previously recorded [85]. After more than a decade of sustained lake trout gillnetting and no evidence of cutthroat trout recovery, a comprehensive scientific appraisal of the on-going program was warranted.

#### **4. Development of a Conservation Strategy that Embraces Uncertainty**

The lake trout-induced collapse of cutthroat trout became increasingly apparent to the public during the 2000s as angler catch rates declined severely on Yellowstone Lake and the Yellowstone River both upstream and downstream of the lake (Figure 1) [46,51]. The loss was particularly noteworthy because cutthroat trout were the only sport fish available to anglers in this ecosystem. The decline also affected backcountry outfitters and anglers seeking the migratory, spawning fish in distant reaches of the upper Yellowstone River south of Yellowstone Lake within the Bridger-Teton Wilderness of Wyoming [46]. In addition to the precipitous decline of cutthroat trout, the novel piscivore altered plankton assemblages within the lake and reduced nutrient transport to tributary streams [49,86]. Effects extended across the aquatic-terrestrial ecosystem boundary as grizzly bears and black bears in the area necessarily sought alternative foods [41,87]. Nest density and success of ospreys greatly declined, and bald eagles shifted their diet to compensate for the loss of cutthroat trout [41,52]. An urgent need for action to reverse these declining trends was recognized by resource managers, scientists, and a wide range of constituents of Yellowstone Lake fishery and wildlife resources.

#### *4.1. Scientific Review Panel*

Yellowstone National Park requested assistance from an independent scientific review panel in August 2008 to critically evaluate the effectiveness of the lake trout suppression program in Yellowstone Lake [88]. The panel was tasked with evaluating the effectiveness of the lake trout suppression program, reviewing emerging technological opportunities for suppressing lake trout, and providing alternatives for the future direction of the program in the context of the primary mission of the NPS, which is to preserve unimpaired the natural and cultural resources and values of the National Park System for the enjoyment, education, and inspiration of this and future generations. To that end, the panel sought to ensure the long-term persistence of native cutthroat trout and the natural function of the Yellowstone Lake ecosystem.

The panel concluded that suppression gillnetting during 1995 to 2008 had not curtailed lake trout population growth, and that the cutthroat trout population had declined severely; however, the cutthroat trout population was not completely lost and the ecosystem could be restored with immediate, aggressive action. Because overharvest had caused collapse of lake trout populations throughout their native range, the panel thought that intensified suppression gillnetting could drive the lake trout population of Yellowstone Lake into decline. Although specific gillnetting effort benchmarks that would result in lake trout decline could not be determined, the panel recommended a doubling of the 28,000 units of annual effort (unit of effort = 100-m net nights) expended at that time. Because an immediate increase in gillnetting effort of that magnitude was beyond the capacity of NPS resources, the panel recommended incorporation of private sector (commercial, professional) gillnetters, an approach which was successful for lake trout suppression on Lake Pend Oreille in

Northern Idaho [89]. The panel also recommended reinitiating lake trout telemetry studies to determine movement patterns, locate spawning habitats, and inform the suppression gillnetting. Additionally, development of novel suppression alternatives to gillnetting and experimentation to assess their effectiveness was also supported by the panel. Specific recommendations of the 2008 scientific review panel, 15 years after initial lake trout discovery, were to:


#### *4.2. Planning and Environmental Compliance*

These major actions to suppress lake trout and restore the Yellowstone Lake ecosystem required increased funding, heightened support by partners and stakeholders, detailed long-term planning, and National Environmental Policy Act (NEPA) compliance. The recommendations of the scientific review panel provided a well-defined need for fund-raising to support the increased efforts. The park supported incorporation of private sector gillnetters during a limited, pilot phase during 2009 to 2010 to determine their feasibility for operations on Yellowstone Lake. Hickey Brothers Research, LLC, a company from Door County, Wisconsin, with extensive commercial fishing experience on Lake Michigan, was awarded the NPS contract to initiate the expanded gillnetting. Combined with the continued NPS operations, the enhanced gillnetting effort resulted in eradication of nearly a quarter million lake trout during 2009 to 2010. Yellowstone National Park concurrently completed an environmental compliance process evaluating potential effects on park resources by an increased suppression program. Input from the public was sought on alternative management actions that would ensure the long-term recovery of cutthroat trout and restoration of natural ecosystem function. A Native Fish Conservation Plan/Environmental Assessment (EA) was made available for review and comment on 16 December 2010 [90]. Development of the plan included scientific review of current conservation efforts, projected changes in native fish status given known threats, a review of relevant emerging science and technology, and public and stakeholder input received during a public scoping process. The plan that emerged identified the following goals:


The plan proposed to conserve native fish from threats of lake trout and other nonnative species, disease, and climate-induced environmental change. It provided guidance for managing fisheries and aquatic resources over the following two decades. The plan described in detail the development of an adaptive management strategy (Figure 5) for implementing large-scale removal of lake trout on Yellowstone Lake by NPS netting crews and incorporation of private sector, contract netters, and called for the development and implementation of robust monitoring and continued scientific review through collaboration with partners. Assumptions made in the selection of reasonable alternatives for the Yellowstone Lake ecosystem included:




**Figure 5.** Adaptive management (ADM) strategy conceptual model for Yellowstone Lake ecosystem restoration. Desired condition (**A**), conservation actions (**B**), quantitative responses (**C**), and performance metrics (**D**) are defined in the text and listed in Table 1 [90].

Because > 100,000 lake trout were being killed annually by 2010, the plan included an alternative analysis of the marketing and sale (or the donation) of gillnetted lake trout for human consumption. The thought was that proceeds from the sale of lake trout could be used to supplement the increased funding needed for the program. However, a large portion of the catch was not suitable for human consumption because the soak times of suppression gillnets were often 7 nights. In addition, handling time and care, including holding fish on ice and transport, would greatly reduce time available for suppression gillnetting. Gillnetting effort would decrease by about an estimated 50% if lake trout were processed for human consumption, requiring a doubling of the number of boats and crews to maintain the same gillnetting suppression effort. Markets and food banks were far from Yellowstone Lake, resulting in high shipment costs. Moreover, the enabling legislation for Yellowstone National Park does not allow for the sale of its natural resources. Sale or donation of lake trout was rejected because of these significant issues. The lake trout carcasses were to instead be deposited in deep (>65 m) regions of the lake to retain their nutrients within the lake ecosystem.

#### *4.3. Conceptual Ecosystem Model and Hypothesized Linkages*

A conceptual model (Table A3) was developed to assist in identifying issues confronting the Yellowstone Lake ecosystem and to clarify which aspects of the ecosystem would likely respond as a result of management actions. The conceptual model illustrated the complex relations among agents of change, stressors on native fish, and ecosystem responses. Agents of change were sources of stressors on native fish when they operated outside the range of natural variability; they included natural processes and events as well as human activities. Ecosystem responses were defined as measurable and detectable changes or trends in the quality or integrity of ecosystem structure, function, or processes.

Agents of change appropriate to the Yellowstone Lake ecosystem were organized into five broad categories (Table A3):


The degree to which each agent of change contributed to a problem was considered and a list of potential stressors on native fish was compiled. Each stressor was matched to general management issues within the 2006 NPS Management Policies [91]. The preliminary list of ecosystem responses was also grouped into five broad categories (Table A3):


For example, the lake trout invasion was an agent of change within the Yellowstone Lake ecosystem that resulted in stressors on native fish in the form of fewer cutthroat trout recruited to the spawning population, direct mortality, predation losses, and loss due to competition/displacement (Table A3). These stressors can also result in ecosystem responses such as changes in nutrient transport, primary and secondary production, fish functional roles and life history strategies, and impacts on avian and terrestrial fish consumers. The conceptual model was not intended to represent a comprehensive account of the entire ecosystem but rather was a framework implicating known or hypothesized agents of change that stress native fish and result in negative ecosystem responses. The goal of the model was to illustrate relationships between and among agents of change and key ecosystem processes and variables. It served to demonstrate the complexity of interactions within the Yellowstone Lake ecosystem, many of which are unknown. Multiple agents of change can lead to multiple stressors, resulting in multiple ecosystem responses.

The underlying hypotheses for the preferred alternative of the Native Fish Conservation Plan were that a gillnetting-driven reduction in lake trout would result in cutthroat trout recovery, and this recovery would, in turn, result in positive responses by piscivorous wildlife (Table A3). In addition, the cascading changes within the lake that followed the cutthroat trout decline (e.g., shifts in zooplankton, phytoplankton, and nutrient transport) would revert to pre-lake trout conditions.

#### *4.4. Desired Conditions*

Primary, secondary, and tertiary desired conditions for the Yellowstone Lake ecosystem were described in the Native Fish Conservation Plan (Table 1) [90]. Complete eradication of lake trout, the most significant agent of change, was the primary desired condition for Yellowstone Lake. However, the secondary condition was initially set as the management target because available lake trout suppression methods were incapable of achieving the primary desired condition. The tertiary condition would become the management target if implementation of conservation actions did not achieve the secondary desired condition. Failure to achieve at least the tertiary condition would be considered a failure to meet the objectives of the plan. Although cutthroat trout are expected to naturally recover following lake trout decline, conservation actions of all desired conditions included ensuring spawning tributary connectivity to Yellowstone Lake during drought years [51,90,92] and reintroduction (stocking) of cutthroat trout to tributaries lacking use by spawners if deemed necessary to maintain the tertiary desired condition.

#### 4.4.1. Primary Desired Condition

The primary desired condition was characterized by cutthroat trout restored to pre-lake trout abundances, and free from all stress by lake trout. This condition would be achieved by a 100% eradication of lake trout or a suppression of lake trout to the point where the species had no measurable impact on the ecology of Yellowstone Lake. Quantitative responses to characterize this condition would include full recovery of cutthroat trout abundance to the averages observed during the five years prior to lake trout discovery (1987–1991; 40 per 100-m net night during long-term gillnetting assessments (relative abundance monitoring, see below); 60 observed during visual spawning surveys; angler catch rate of 2.0 per hour; Table 1). Performance metrics were a lake trout population growth rate (λ) ≤ 0.75; catch-per-unit-effort (CPUE) = 0.01 per 100-m net night during long-term gillnetting

assessments; and angler catch rate < 0.05 per hour. Lake trout abundance would be extremely low and difficult to detect in this condition.

#### 4.4.2. Secondary Desired Condition

The secondary desired condition would be characterized by restoration of cutthroat trout to abundances present during the early stages of lake trout invasion, indicating significantly reduced lake trout stress on cutthroat trout. This condition would be achieved by significantly reducing lake trout abundance in Yellowstone Lake. Quantitative responses to characterize this condition would include recovery of cutthroat trout abundance to the averages observed during the five years following lake trout discovery (1995–1999; 26 per 100-m net night during long-term gillnetting assessments; 40 observed during visual spawning surveys; angler catch rate of 1.5 per hour; Table 1). Performance metrics were a lake trout population growth rate (λ) ≤ 0.85; CPUE = 0.1 per 100-m net night during annual long-term gillnetting assessments; and angler catch rate < 0.1 per hour.

#### 4.4.3. Tertiary Desired Condition

The tertiary desired condition would be characterized by cutthroat trout restored to abundances during the later stages of lake trout invasion, indicating moderately reduced lake trout stress on cutthroat trout. This condition would be achieved by slightly reducing lake trout abundance in Yellowstone Lake. Quantitative responses to characterize this condition would include maintaining cutthroat trout abundance at the average observed prior to lake-wide expansion by lake trout (2001–2005; 12 per 100-m net night during long-term gillnetting assessments; 20 observed during visual spawning surveys; angler catch rate of 1.0 per hour; Table 1). Performance metrics were a lake trout population growth rate (λ) ≤ 0.95; CPUE = 0.5 per 100-m net night during annual long-term gillnetting assessments; and angler catch rate < 0.5 per hour.

#### **5. Stakeholder Involvement and Fundraising to Support Conservation Actions**

Following completion of the Native Fish Conservation Plan/EA [90], a Finding of No Significant Impact (FONSI) was signed by the NPS Intermountain Region Director in June 2011. That year, Yellowstone National Park entered into a 5-year contract with Hickey Brothers Research, LLC, to increase lake trout gillnetting suppression effort lakewide. In addition, research to improve suppression efficiency began in earnest as adult lake trout were surgically implanted with acoustic tags to determine broad scale movement patterns and locate key spawning sites. A plan with clearly articulated objectives and benchmarks for Yellowstone Lake ecosystem restoration and feedback from annual scientific panel reviews provided strength for acquiring funding to support the program. Missing however, was a process to better incorporate anglers, conservation groups, and the general public in on-the-ground actions to conserve native fish as described in the plan.

#### *5.1. Yellowstone Fly Fishing Volunteer Program*

The *Yellowstone Fly Fishing Volunteer Program* was initiated in 2002 to acquire information about fish populations throughout the park without requiring Yellowstone biologists to travel to sample the populations themselves using electrofishing or other sophisticated gear [93]. The volunteers fly-fished to gather and archive information and biological samples that park biologists would otherwise not be able to collect. In addition to providing valuable data, samples, and assistance to the fisheries program, volunteer fly fishers have played an important role with the public by interacting positively with park biologists and the public and demonstrating their passion for native fish and the importance of protecting these species. Volunteer fly fishers have promoted an understanding of the Yellowstone Lake ecosystem restoration program and generated greater awareness of the current issues facing Yellowstone's native fish. These passionate and informed supporters have been an important contribution to the success of our program.

#### *5.2. Yellowstone Lake Workgroup*

A consortium of conservation groups met with NPS officials in 2011 with the intent of becoming partners in addressing the threats to the Yellowstone Lake fishery and ecosystem. From that meeting and subsequent discussions, a Memorandum of Understanding (MOU) was developed that formalized a cooperative relationship among participants to ensure the ecosystem was protected, maintained, and managed to achieve established goals. Signatories to theMOU were Trout Unlimited National, Wyoming Council, Montana Council, and Idaho Council; National Parks Conservation Association; Greater Yellowstone Coalition; Yellowstone Park Foundation (subsequently named Yellowstone Forever); and Yellowstone National Park. These stakeholders began meeting semi-annually and created a formal *Yellowstone Lake Workgroup* that acted as a sounding board to review lake trout suppression activities, population monitoring activities and trends, telemetry research results, new suppression technologies, and other fisheries-related science. They also initiated positive public outreach and education and authored publications directed at the general public and potential donors, including a publication with answers to frequently asked questions about the science supporting management of Yellowstone Lake [94]. The group has responded to public concerns about fish conservation actions when applicable. The *Yellowstone Lake Workgroup* has been actively involved in fundraising and has raised over US\$1 million to directly support cutthroat trout restoration in Yellowstone Lake. The majority of the funds have been spent on (1) telemetry studies to determine lake trout seasonal movement patterns and location of spawning areas, (2) studies assessing the reproductive potential, cycles, and timing of lake trout spawning, and (3) studies to identify and optimize alternative suppression technologies aimed at lake trout embryos. In addition, members of the *Yellowstone Lake Workgroup* have provided volunteer labor to support the work on Yellowstone Lake.

#### *5.3. Yellowstone Forever Fund-Raising Partnership*

As the need for lake trout suppression effort increased, so did the need for increased funding to support it. Funding for the Yellowstone Lake ecosystem restoration was≤ US\$500,000 until 2009, when it was increased to support expanded suppression effort by contracted gillnetters. *Yellowstone Forever*, which was the official fund-raising partner of Yellowstone National Park, approved a grant in 2012 to initiate strong, annual support of the program. This support, funded by private donations, increased total program funding (donated and NPS) from US\$1,000,000 in 2011 to more than US\$2,000,000 by 2013 (Figure 6A). These funds allowed a rapid increase in suppression gillnetting effort by the contracted crews to meet (and surpass) annual gillnetting benchmarks, while concurrently enhancing long-term monitoring, population modeling, and applied research to improve program efficiency.

**Figure 6.** Costs, effort, and catch of the Yellowstone Lake ecosystem restoration program, 1995–2019, including (**A**) the total program cost (US \$Millions) of National Park Service (NPS) and contracted operations, gillnetting effort unit (100-m net nights) benchmarks estimated by statistical-catch-at-age modeling to achieve λ < 1 (2009–2019) or to reduce lake trout (LKT) total abundance to 100,000 fish (2017–2019), and the actual total annual gillnetting effort applied, (**B**) total numbers of lake trout killed by small and large mesh gillnets, catch-per-unit-effort (CPUE) by all mesh sizes combined, and total biomass of lake trout gillnetted, and (**C**) cutthroat trout proportion (%) of total gillnet catch and number of bycatch in small and large mesh gillnets.

#### **6. Historical Development of the Gillnetting Program**

Gillnets have remained the primary lake trout suppression gear on Yellowstone Lake throughout the first 25 years of the program because they were the preferred gear in commercial fisheries and could overharvest lake trout populations in their native range [83,95]. At first (1995–1999), lake trout distribution or abundance in Yellowstone Lake was poorly understood, crews were inexperienced in population suppression gillnetting, and boats lacked specializations for lifting and processing long gillnets. Summer (June–September) gillnet sets were short (< 200 m length) and mostly lifted by hand from relatively small (6–8 m) research and monitoring vessels. Despite these constraints, NPS crews annually processed an average of 1164 100-m net nights and killed an average of 2965 lake trout (14,823 total lake trout killed during 1995–1999, mean CPUE = 2.1 per 100-m net night, mean biomass caught = 0.07 kg/ha; Figure 6B) with nearly all of the effort focused in the West Thumb region, where the original discovery of lake trout occurred and catch rates were highest (Figure 7).

**Figure 7.** Spatial expansion and increase in intensity of suppression gillnetting effort on Yellowstone Lake during years of rapid lake trout population growth (1999, 2004, and 2009) and years where distribution of gillnets and overall effort was great enough to curtail population growth (i.e., λ < 1; 2014 and 2019). Darker shades of red indicate lake areas with greater intensity of gillnetting effort.

National Park Service crews increased gillnetting effort and total lake trout killed after acquisition of a large (10 m) enclosed gillnetting boat in 2001. Both the new boat and an existing boat (8 m) were outfitted with hydraulic net lifters, thereby greatly increasing their ability to retrieve long gangs of nets (550 m). Crews used sonar and detailed lake bathymetry [96] to locate concentrations of lake trout and extended gillnetting throughout the possible field season from late-May through mid-October in 2000–2004 to increase gillnetting effort more than 10-fold throughout the Breeze Channel and into the Main Basin of Yellowstone Lake (Figure 7). An average of 12,675 100-m net nights were processed during this period resulting in the kill of 17,072 lake trout annually (85,861 total lake trout killed during 2000–2004, mean CPUE = 1.7 lake trout per 100-m net night, mean biomass caught = 0.28 kg/ha; Figure 6B).

Gillnet effort and lake trout killed increased again during 2005–2009 as experienced NPS crews focused solely on suppression. Netting effort expanded across Yellowstone Lake (Figure 7) to follow the increasing lake trout population that, despite the 10+ years of control efforts, was experiencing exponential growth. A pilot study of contracted gillnet services initiated in 2009 brought a third specialized gillnetting vessel to the lake and increased gillnetting effort by 3 weeks in 2009. An average of 21,769 100-m net nights were processed and 68,761 lake trout were killed annually (343,807 total lake trout killed during 2005–2009, mean CPUE = 3.3 lake trout per 100-m net night; mean biomass caught = 1.05 kg/ha; Figure 6B).

The gillnetting effort was increased during 2010 to 2014 to a level that curtailed further lake trout population growth [81] as a result of (1) completion of the Native Fish Conservation Plan/EA, (2) estimation of gillnetting benchmarks by population modeling (see below), (3) growing support of stakeholders, (4) significant private donor funding support, and (5) guidance by the scientific review panel. Five large, specialized gillnetting vessels expanded effort and spatial distribution of gillnets across Yellowstone Lake (Figure 7). Contiguous gillnet lengths were increased to >3 km by the contract crews and were set in a serpentine pattern along bottom contours to maximize catch of lake trout, which swam parallel to net panels. The increased effort focused on large adult lake trout, especially during the autumn spawning period, while not reducing effort targeting smaller lake trout [97] (Figure 6B). The increased focus on adult lake trout along with telemetry of acoustic-tagged fish (see below) resulted in the discovery of several additional lake trout spawning sites that were subsequently targeted [59,98]. During this period of rapid effort expansion, an average of 48,073 100-m net nights were processed and 249,466 lake trout were killed annually (1,247,332 total lake trout killed during 2010–2014, mean CPUE = 5.6 per 100-m net night; mean biomass caught = 3.62 kg/ha; Figure 6B).

Large deepwater trap nets were set for extended periods (months) in fixed locations in 2010–2013 to complement gill netting and maximize capture of large adult lake trout (and minimize cutthroat trout bycatch) [42]. The trap nets also captured live lake trout for critical telemetry studies (see below). Although the trap nets caught nearly 33,000 additional adult lake trout during a total of 2810 net nights over four years, the fish were of the same size classes caught by large-mesh gillnets. Moreover, the trap nets, which were 9–15 m high, had complex leads > 300 m long, were held in place by heavy (50 kg) anchors, and required highly trained contracted crews to set and maintain. Use of the trap nets was discontinued after 4 years because the time and cost of their use were high relative to that required to achieve similar catches with gill nets.

The suppression program was expanded during 2015 to 2019 with the purchase of an additional NPS gillnetting vessel resulting in a total of six large specialized boats (Figure A4) and crews (NPS and contract) with substantial fishing experience (Video S2). Although all size classes of lake trout were targeted, effort continued to be focused on removal of large, adult lake trout. Net inventories were increased greatly to accommodate the increase in effort and included a broad range of mesh sizes to target the changing population and maximize catches throughout the suppression season. Twine (monofilament) diameters were reduced, which increased catch efficiencies. Gillnets were distributed across most of the lake that was <60 m deep, the depths that had proven to be most productive (Figure 7). Proportionally, effort continued to be highly focused on the West Thumb, Breeze Channel, and Main Basin regions near Frank Island where catches remained the highest (Figure 1). Experienced, professional gillnetting crews processed an average of 88,124 100-m net nights resulting in an annual average kill of 331,783 lake trout and a reduced CPUE (1,658,917 total lake trout killed during 2015 to 2019, mean CPUE = 3.8 lake trout per 100-m net night; mean biomass caught = 3.77 kg/ha; Figure 6B).

Bycatch of cutthroat trout has occurred throughout the gillnetting program. Both lake trout and cutthroat trout occupy shallow-water habitats during spring and autumn despite spatial segregation

during summer stratification (late-July through mid-September) when lake trout seek cold, deep lake areas. Targeted gillnetting of lake trout while avoiding cutthroat trout bycatch has therefore been challenging during spring and autumn. Immediately following the discovery of lake trout during 1995–1997, 68%–87% of annual catches were cutthroat trout (Figure 6C). As lake trout abundances increased and crews developed better methods to target them, the bycatch of cutthroat trout declined to <15%. Annual bycatch increased to about 20,000–30,000 cutthroat trout beginning in 2012, concurrent with rapid expansion of gillnetting effort and cutthroat trout recovery. Bycatch makes up only about 6%–10% of the total catch but is an unfortunate consequence of the use of gill nets as the primary lake trout suppression tool. However, the number of cutthroat trout saved by the killing of 100,000s of lake trout annually far exceeds bycatch losses given that each lake trout consumes an estimated 41 cutthroat trout annually [79].

#### **7. Lake Trout Population Modeling and Gillnetting E**ff**ort Benchmarks**

The most robust approach for evaluating the success of the lake trout suppression program was a combination of long-term monitoring and population modeling [99–101]. We estimated lake trout abundances and mortality through time by integrating gillnetting effort, harvest data, and standardized monitoring data (long-term gillnetting assessments, described below) in a statistical catch-at-age (SCAA) assessment model [80,81,102]. The results of the SCAA model were used to forecast the amount of gillneting effort required to achieve a given level of mortality [103,104]. Our goal, established in 2010, was to reduce the abundance of lake trout to their mid-1990s levels (about 100,000 fish), when they probably had little effect on the native cutthroat trout population [90]. However, a major uncertainty in reaching that goal was the amount of gillnetting effort needed. Population modeling and analyses of lake trout suppression data collected over several years were used to address this question and to assess suppression program success. The three important metrics assessed were (1) total annual mortality, (2) population abundance, and (3) population growth rate (λ) of lake trout.

Annual gillnetting effort benchmarks were estimated iteratively using the SCAA model beginning in 2009, when 29,000 100-m net nights were predicted to be needed to reduce λ < 1 within 5 years (Figure 6A). This initial benchmark was estimated using age-0 and age-1 survival rates from the native range of lake trout [80]. Subsequent analyses of local data indicated that pre-recruit survival rates in Yellowstone Lake were much higher than in the native range and the model was adjusted accordingly. The benchmark increased to 57,000 units in 2011 and stabilized at 75,000 units from 2016 to 2019. Because annual lake trout harvests remained high, we estimated the effort required to reduce lake trout abundance to the goal of 100,000 fish identified in the Native Fish Conservation Plan. This target ranged from 90,000 to 110,000 100-m net nights between 2017 and 2019, respectively, which appeared to be converging with estimates of effort required to maintain λ<1 (Figure 6A). Fortunately, the combined NPS and contracted suppression gillnetting crews were able to achieve > 90,000 effort units in those years.

#### **8. Lake Trout Population Response to Suppression Gillnettting**

Annual monitoring metrics indicated that suppression gillnetting successfully reduced lake trout abundances (ages 3+) and biomass. The catch rate of lake trout in annual long-term gillnetting assessments declined from 2011 to 2019 (*p* = 0.032), with a high of 4.9 per 100-m net night in 2014 and a low of 2.0 in 2018 (Table A4, Figure 8), but remained above established desired conditions (Table 1). Large adult lake trout (> 400 mm) consistently made up 13% to 29% of the catch. Proportions of the catch ≤ 280 mm (primarily age-2 individuals) were <50% in all years except 2019 when small lake trout made up 63% of the catch (Figure A5). Small, juvenile lake trout consistently represented a large proportion of the annual catch, suggesting that recruitment remained strong.

**Figure 8.** Catch-per-unit-effort (CPUE) of lake trout during annual long-term gillnetting assessments on Yellowstone Lake, 2011–2019. The blue line represents a simple linear regression model with 95% confidence intervals (dashed lines). Numbers within parentheses are the upper and lower 95% confidence limits of the slope parameter estimate, indicating a temporal decline in the CPUE response variable because the interval does not include zero.

Total lake trout abundance estimates (age 2 and older, at the beginning of the year, derived from the SCAA model, Table A5) declined (*p* = 0.038) from 925,208 in 2012 to 673,983 in 2019 (Figure 9A) after suppression gillnetting effort increased sufficiently to curtail lake trout population growth (i.e., λ < 1). However, high among-year variation in estimated total abundance (R<sup>2</sup> = 0.54) was apparently driven by highly variable recruitment of age-2 fish to the population. Abundances of age-2 lake trout during 2012 to 2019 ranged from a low of 318,640 in 2013 to a high of 480,961 in 2015 and no significant relationship existed between abundance and year (*p* = 0.834, Figure 10A). Abundances of age 3–5 (*p* = 0.011) and age 6+ (*p* < 0.001) lake trout declined from 416,814 and 57,722 in 2012 to 197,681 and 12,345 in 2019, respectively (Figure 10B,C). These declines resulted in a 54% decrease in total lake trout biomass (*p* < 0.001) from 432,017 kg in 2012 to 196,675 kg in 2019 (Figure 9B).

The lake trout population growth rate (λ) from 2017 to 2018 was 0.75 (95% CI: 0.65–0.85), which met our primary desired condition for this performance metric (Table 1) [81]. However, it grew to 1.18 (95% CI: 0.95–1.40) from 2018 to 2019 because of high age-2 recruitment. Even though total abundances of older lake trout declined between 2012 and 2017, abundance of age-2 lake trout increased by 69% between 2018 and 2019 (Figure 10A) because of the high year-class strength of the 2017 cohort, which was detected in 2019 after it recruited to the gear. This increase in recruitment, despite reductions in adult abundances, indicated a compensatory response by the lake trout population. Lake trout length at maturity did not change between 1995 and 2019; female lake trout matured at 515 mm (95% CL: 503–525) and male lake trout matured at 431 mm (95% CL: 423–444) (Figure A6). However, relative condition (Kn) of large lake trout (400+ mm), increased during this period, from 102.8 during 1995–1999 to 111.8 during 2005–2019 (Table A6, Figure A7). Nevertheless, estimated total egg production declined from 51.1 million in 2010 to only 15.8 million in 2017 (Figure 11). Accordingly, age-2 recruitment was maintained despite reductions in abundances of adults and egg production.

**Figure 9.** Total (**A**) abundance and (**B**) biomass of age-2 and older lake trout at the start of the year from 2012 through 2019 estimated using a statistical catch-at-age (SCAA) model [81]. Blue lines represent simple linear regression models with 95% confidence intervals (dashed lines). Numbers within parentheses are the upper and lower 95% confidence limits of the slope parameter estimate, indicating temporal declines in these response variables because intervals do not include zero.

Mandatory angler-harvest of lake trout, which was implemented immediately upon lake trout discovery, is substantial at about 5% of the total killed by all methods (angling and suppression gillnetting) combined in recent years. Moreover, angler harvest is additive to total mortality and comes at no cost to the program. Annual angler harvest increased from an estimated 500 lake trout in 1995 to 2900 in 1999 and averaged about 17,000 after 2002. Angler catch rate declined from 0.5 fish per hour in 2012 to 0.2 per hour in 2019 but remained above desired conditions for this performance metric (<0.05 lake trout per hour; Table 1). Historically, more than 50% of lake trout caught by anglers were large (>450 mm), but only 32%–34% were >450 mm during 2017 to 2019, reflecting the reduced abundance caused by suppression gillnetting. Anglers have consistently caught fewer lake trout than cutthroat trout in Yellowstone Lake.

**Figure 10.** Abundances of (**A**) age-2, (**B**) age-3 to age-5, and (**C**) age-6+ lake trout at the start of the year from 2012 through 2019 estimated using a statistical catch-at-age (SCAA) model [81]. Blue lines represent simple linear regression models with 95% confidence intervals (dashed lines). Numbers within parentheses are the upper and lower 95% confidence limits of the slope parameter estimate. There was no temporal change in age-2 abundance because the interval includes zero. Age-3 to age-5 abundance and age-6+ abundance declined significantly.

**Figure 11.** Age-2 lake trout abundances during 1998 to 2017 and the number of spawned eggs that produced these recruits two years previously estimated by a statistical catch-at-age (SCAA) assessment model [81]. Different colors represent 5-year periods of the lake trout suppression program described in text. Although fewer eggs were produced annually since lake trout population growth was curtailed in 2012, similar (or often greater) abundances of age-2 lake trout were recruited to the population, suggesting a compensatory response.

#### **9. Cutthroat Trout Population Response to Lake Trout Suppression**

The average cutthroat trout CPUE (during historical gillnetting assessments; Figure 1) was 37.8–48.7 per 100-m net night in the late-1980s and 41.9 in 1994, the year lake trout were first discovered (Figure 12A). Cutthroat trout CPUE then declined to 19.3 by 2004 (average of 8% reduction per year) following more than a decade of predation pressure by lake trout. The lowest lake-wide cutthroat trout gillnetting CPUE was 10.9 in 2010, and other monitoring metrics (see below) also reached a minimum.

Cutthroat trout abundance declined precipitously until suppression efforts reached sufficient levels to reduce lake trout abundances in 2012 [41]. The number of cutthroat trout caught during annual long-term gillnetting assessments varied subsequently, with mean CPUE ranging from a low of 12.5 per 100-m net night in 2011 to highs of 27.3 and 26.4 in 2014 and 2018, respectively (Table A4, Figure 13). These CPUEs met established secondary desired conditions for cutthroat trout (CPUE > 26; Table 1; Figure 12A). Size structure of the cutthroat trout population also varied during this period. The proportion of the long-term gillnetting assessment catch ≤ 280 mm (primarily age-2 individuals) was extremely low (16%) in 2011 when lake trout predation pressure was high (Figure A8). Although large (> 400 mm) individuals continued to dominate the cutthroat trout population during 2012 to 2019 when abundances of age 3+ lake trout declined, cohorts of smaller cutthroat trout subsequently became a more common component of the population, indicating greater avoidance of lake trout predation.

**Figure 12.** Cutthroat trout quantitative response variables monitored to assess the effects of conservation actions in the adaptive management strategy for Yellowstone Lake included the (**A**) catch-per-unit-effort (100-m net night) during within-lake gillnetting assessments, (**B**) observed during visual surveys of spawning streams, and (**C**) caught per hour by lake anglers, 1985–2019. Primary (1◦), secondary (2◦), and tertiary (3◦) desired conditions are from the Native Fish Conservation Plan (Table 1) [90].

**Figure 13.** Catch-per-unit-effort (CPUE) of cutthroat trout during annual long-term gillnetting assessment monitoring on Yellowstone Lake, 2011–2019. The red line represents a simple linear regression model with 95% confidence intervals (dashed lines). Numbers within parentheses are the upper and lower 95% confidence limits of the slope parameter estimate, indicating no temporal change in the CPUE response variable because the interval includes zero.

Spawning cutthroat trout abundances decreased significantly following the lake trout invasion as judged by visual surveys of 11 spawning streams conducted annually since 1989 (Figure 1) [51,105]. An average of 74 cutthroat trout was observed during each stream survey in 1990, compared to ≤1 cutthroat trout during 2004 to 2010 (Figure 12B). Subsequently, spawning cutthroat trout increased slightly to a mean of 7.5 and 6.2 per survey in 2016 and 2019, respectively. Abundance increased in Little Thumb Creek, a tributary in the West Thumb near Grant (Figure 1), where more than 50 cutthroat trout were seen during a single visit in 2013, and more than 100 were seen during visits in 2014 and 2015. The number of fish observed in Little Thumb Creek after 2016 was about 80% of the total fish counted each spring at all of the visually-assessed spawning tributaries combined. Although the increased number of fish observed in Little Thumb Creek is encouraging, counts remained far below the primary or secondary desired conditions (means of 60 and 40 spawning cutthroat trout, respectively, observed per visit to all 11 of the visually-assessed spawning tributaries combined).

An estimated 68,000 cutthroat trout were caught by Yellowstone Lake anglers in 2019 at a catch rate of 0.9 fish per hour (Figure 12C). This catch rate was below the primary (2.0) and secondary (1.5) desired conditions (Table 1), which were the catch rates experienced during the 1980s and early 1990s. However, the average size of cutthroat trout caught in 2019 (440 mm) was much larger than caught prior to the lake trout invasion (380 mm). Although fewer fish were caught by anglers, the quality of the fish (from an anglers perspective) was much greater than in the past.

Lake trout predation was associated with a long-term shift in cutthroat trout lengths from dominance by small (100–280 mm) and midsized (290–390 mm) individuals to dominance by large individuals (400+ mm) in annual gillnetting assessments. The mean CPUE of small and midsized cutthroat trout declined from 18.6 per 100-m net night and 15.1, respectively, in the 1980s to just 6.9 and 3.9, respectively in the 2010s (Table A7, Figure 14A). Concurrently, the mean CPUE of large cutthroat trout nearly doubled, from 7.5 in the 1980s to 14.6 in the 2010s. Lake trout also caused increases in individual weights and condition of cutthroat trout. Although the average weight of small cutthroat trout slightly declined, the average weight of midsized and large cutthroat trout increased from 408.0 g and 682.8 g, respectively, in the 1980s to 463.4 g and 1418.6 g, respectively, in the 2010s (Table A7, Figure 14B). Relative weights (condition factors) of individual cutthroat trout also increased during this period. Mean relative weights of small, midsize, and large cutthroat trout were 58.8, 56.5, and 55.8, respectively, in the 1980s and increased to 68.4, 70.4, and 67.7, respectively, in the 2010s (Table A7, Figure 14C). Lower densities of cutthroat trout with higher individual weights and conditions should have higher fecundity [106], which should aid further recovery.

**Figure 14.** Mean (**A**) catch-per-unit-effort (100-m net night), (**B**) individual weight (g × 100), and (**C**) relative weight of each of three length groups (mm) of cutthroat trout from annual gillnetting assessments during each decade (1980–2019) on Yellowstone Lake. Means marked with the same letters (a–c) within each length group indicate no statistical difference among decades (α = 0.05).

#### **10. Ecological Response to Lake Trout Suppression**

Lake trout induced stress on cutthroat trout caused trophic shifts over four decades across multiple trophic levels both within and outside of Yellowstone Lake [41]. Hypothesized outcomes of lake trout reduction and cutthroat trout recovery were the return of trophic levels to natural, pre-lake trout conditions (Table A3). We therefore monitored components of the aquatic and terrestrial ecosystems to document the cascading changes that occurred after lake trout suppression.

The introduction of lake trout added a fourth trophic level [86] and resulted in cascading interactions within the aquatic food web of Yellowstone Lake, including shifts in cutthroat trout prey consumption and the biomass and individual lengths of zooplankton. When cutthroat trout were abundant in the 1980s, large-bodied cladocerans made up 80% of their diet [79]. After cutthroat trout declined in the 2000s, cladocerans made up only 11% of their diet, and cutthroat trout more frequently consumed amphipods, which made up 79% of their diet by 2011 [107]. After cutthroat trout abundance declined and predation on large zooplankton was reduced, the biomass of small-bodied copepods declined and the biomass of large-bodied zooplankton increased within the pelagic zone [41]. In addition, the average length of large-bodied zooplankton increased. Increased grazing by the dominant large-bodied cladocerans following the decline of cutthroat trout resulted in lower phytoplankton biomass and increased water clarity [41]. Although zooplankton and phytoplankton communities can be early indicators of ecological change (because of rapid turnover rates), we did not observe any shifts of plankton to a pre-lake trout condition resulting from the suppression program.

Lake trout invasion caused substantial indirect effects [108] that extended to native avian and terrestrial animals, such as ospreys, eagles, and bears, because spawning cutthroat trout were an important high-energy food for them. Ospreys are obligate piscivores that cannot switch to alternative food sources in the absence of fish. Osprey nest densities declined concurrently with declines in cutthroat trout prey, from an average of 38 nests during 1987 to 1991 to 11 during 2004 to 2008 [52] and only 2 during 2015 to 2019 [41]. Nesting success during 1987 to 1991 averaged 59% but declined to zero during 2008 to 2011 when no young ospreys were fledged from Yellowstone Lake nests. An average of 13% of osprey nests successfully fledged young during 2015 to 2019. We did not observe an increase in osprey nest density resulting from the suppression program, but expected a lagged response to increased cutthroat trout abundance.

Four to six bald eagle nests were typically present on Yellowstone Lake during the 1960s and 1970s. The nest count increased to an average of 11 during 2004 to 2008 [52] but then declined to 8 nests by 2015 to 2019 [41]. A steady long-term decline in bald eagle nest productivity occurred over two decades concurrently with the lake-wide decline in cutthroat trout. During 1985 to 1989, 56% of bald eagle nests on Yellowstone Lake successfully fledged young; however, nest success declined to zero in 2009 when prey fish abundance was low. Bald eagles are opportunistic feeders and increased consumption of alternative prey, including scavenging carnivore-provided carcasses and winterkill. Bald eagles were also observed preying on common loons and trumpeter swan cygnets, which have declined recently in Yellowstone National Park.

Grizzly and black bears are opportunistic feeders with flexible diets that consumed other foods available in the Yellowstone Lake area when cutthroat trout abundance was low [51,87]. No bear activity was found on surveyed spawning streams in 2008, 2009, or 2011 after cutthroat trout declined [41]. Compared to estimates obtained from 1997 to 2000, the number of grizzly bears visiting spawning streams a decade later (2007–2009) decreased by 63%, and the number of black bears decreased by 64 to 84% [109]. Bear activity on spawning tributaries subsequently increased in response to the slight recovery of spawning cutthroat trout [41].

#### **11. Applied Research to Inform Decision Making**

Management actions to restore the Yellowstone Lake ecosystem were supported by science through a strong program of applied research. Primary university partners were Michigan State University, Montana State University, University of Montana, University of Vermont, University of Wyoming, and Utah State University. Agency research partners included Montana Department of Fish, Wildlife and Parks, U.S. Geological Survey, U.S. Fish and Wildlife Service, U.S. Department of Agriculture, and Wyoming Game and Fish Department.

Research identified Lewis Lake or other waters in the upper Snake River system as the source of lake trout introduced to Yellowstone Lake [110,111] and identified the potential impacts the new predatory trophic level could have on the cutthroat trout and the ecology of the lake and tributary spawning streams [79,86,112]. As the cutthroat trout population declined, additional causal factors were investigated including *Myxobolus cerebralis* Hofer, 1903 (the causative agent of whirling disease), which was discovered in cutthroat trout from Yellowstone Lake in 1998 [113,114] and caused localized losses of cutthroat trout in Pelican Creek and the Yellowstone River downstream (Figure 1) [115–118]. Environmental factors influencing variation in cutthroat trout year class strength were also investigated. Persistent drought conditions were considered a strong contributing factor leading to cutthroat trout decline from the 1980s to the 2000s [51,92]. Studies also documented large-scale migrations, spawning locations (natal origins), and lake-wide movements of cutthroat trout [46,119].

After more than a decade of suppression gillnetting did not stop lake trout population growth, SCAA modeling of the lake trout population was conducted in the late 2000s to estimate lake trout demographics and establish gillnetting effort benchmarks that would result in λ < 1 [80,102]. An annual long-term monitoring protocol (long-term gillnetting assessment) was developed with sufficient power to detect changes in lake trout abundances over time [120]. Mark-recapture of tagged lake trout estimated population abundances and independently validated modeled lake trout abundance estimates [42]. The population model was updated annually and was critical for evaluating the efficacy of lake trout suppression relative to Native Fish Conservation Plan performance metrics [81,90].

Identification of lake trout spawning areas and movement patterns was critical for increasing the efficiency of the suppression program. Spawning habitat models, based on wave energy theory and geomorphology, suggested conditions for lake trout spawning habitat are patchy and exist in <4% of the lake [62]. Acoustically-tagged "Judas" fish were used to document seasonal movement patterns, habitat use (including spawning), and guide gillnetting efforts. For example, telemetry of lake trout conducted over several years (2011–2016) using fixed array receivers revealed extensive movements throughout the lake, important migration corridors, and spawning habitats [98,121]. Active mobile telemetry (boat-mounted receivers) further refined lake trout spawning site locations [60] and provided gillnetting crews with real-time locations and depths of lake trout aggregations, which increased gillnetting catches [122]. Subsequently, putative lake trout spawning sites were searched for presence of gametes, and all verified sites were investigated using remotely operated vehicles or scuba diving to delineate substrate boundaries, estimate substrate sizes and depths, and document substrate types [59]. The early life history of lake trout in Yellowstone Lake was assessed to better understand potential vulnerabilities to alternative (complementary) suppression methods that target embryos or fry [123].

Ongoing research focused on the development of alternative suppression methods that target embryos on spawning sites and complement gillnetting (see below). Because the diets of lake trout and cutthroat trout shifted to dominance by amphipods following the cutthroat trout decline [107], research was conducted to better understand lake ecology related to large-scale carcass deposition (about 300,000 carcasses per year in lake areas deeper than 65 m) and treatment of spawning sites with carcasses and organic pellets during autumn (see below). Ongoing efforts are underway to increase efficiency of monitoring cutthroat trout spawning abundances in tributaries using eDNA technology [124], detect congregations of lake trout using airborne lidar [125], and to further document effects of the lake trout and suppression actions on the lake and associated terrestrial ecosystems, including indirect effects on bears, birds, and river otters [41].

#### **12. Need for Complementary Methods that Target Multiple Life Stages**

The integrated pest management (IPM) approach to controlling invasive species uses a variety of suppression methods to target multiple life stages of invaders to maintain abundance levels below those causing harm [126–128]. The IPM approach has been most widely used in terrestrial systems to control agricultural pests [129,130]; however, a growing body of scientific evidence suggests that control of multiple life stages is required to suppress invasive fish populations over long time scales [131–133]. The suppression of sea lamprey in the Laurentian Great Lakes is an example of IPM in which chemical treatments, pheromone attractants, migration barriers, and other methods have been successfully used in combination [134–136]. The Yellowstone Lake gillnetting program curtailed lake trout population growth, but required extremely high levels of effort [41,81]. Because lake trout population growth rates are most sensitive to changes in age-0 survival [18,137,138], we developed and experimentally assessed methods to reduce it, with the intent of implementing an IPM approach on Yellowstone Lake.

The search for suppression methods that could complement gillnetting began at Montana State University in 2004 by a mechanical and chemical engineering senior design team [139] and continued in 2008 with a comprehensive literature review [140]. Later, the effectiveness of high-pressure water [42], electricity [141], suction-dredging [142], tarping (suffocation), and use of lake trout carcasses [143,144] and organic (plant-based) pellets [59,145,146] were evaluated for increasing lake trout embryo mortality *in situ* on Yellowstone Lake spawning sites. Additionally, chemical compounds (rotenone, sodium chloride, calcium carbonate, gelatin), sedimentation, fish carcasses, and organic (plant- and fish-based) pellets were evaluated in the laboratory [147]. To date, the most promising method for increasing lake trout embryo mortality was to intentionally degrade interstitial water quality. Treatment of spawning sites with lake trout carcasses or organic (soy and wheat gluten) pellets (Figure A9) induced organic decomposition and decreased dissolved oxygen concentrations at the substrate surface and 20 cm below the surface [59,144,145]. Biological oxygen demand of the decomposing organic materials caused dissolved oxygen to decline to 0 mg/L immediately after treatments and caused high embryo mortality within 200 h.

Fourteen lake trout spawning sites have been located over the past 25 years by gillnetting, sonar, telemetry, and shoreline visual surveys [59]. These sites vary in size, depth, substrate type, and thermal characteristics (Figure 15), and are generally located near western shorelines in areas of relatively low fetch during the autumn spawning period (when prevailing winds are from the southwest, Figure 1). Because only a few weeks are available to safely work on Yellowstone Lake following the peak of lake trout spawning each autumn, we expanded the embryo suppression research to include a comprehensive treatment of a spawning site with organic pellets by helicopter (with long line and seeder/spreader) to better understand the logistical constraints that may be faced when attempting large-scale, multi-site applications in the future. During an October 2019 experimental treatment, all of the rocky substrate at the Carrington Island spawning site (0.5 ha; Figures 1 and A9) was treated with 18,000 kg of organic pellets in less than one day (Video S3). Fry traps placed at Carrington Island in spring 2020 captured no lake trout fry, indicating all lake trout embryos were likely killed by the treatment and recruitment from the site was completely eliminated. Relative to the expansive lake areas intensively gillnetted over a 22-week season (>60 km of gill nets set daily), lake trout embryo suppression targets relatively small sites during a period of 2–3 weeks in autumn where the majority of a future year class is concentrated. Broad-scale application of pellets in autumn may reduce lake trout recruitment and enhance population suppression as part of an IPM approach targeting multiple lake trout life stages because the area of the 14 verified spawning sites is only 11.4 ha (0.03% of lake surface area) [59].

September 2017–May embryos require degree days (dd) laboratory [147]. hatching are within shaded gray. Although cooling temperatures slow development [148], embryos fertilized by 20 September may hatch prior to Yellowstone Lake becoming ice covered in late-December, many months earlier than within their native range in the Great Lakes [149].

#### **13. Discovery of Nonnative Cisco in Yellowstone Lake**

A cisco (likely *Coregonus artedi* Lesueur, 1818), not native to Yellowstone Lake, was caught during lake trout gillnetting operations north of Stevenson Island during August 2019 (Figure 1). The cisco was a live immature female, age-3, caught in 50 m depth. Otolith microchemistry analysis indicated it was probably hatched in Yellowstone Lake, meaning that parents and siblings are probably present. Someone illegally introduced them because no possible natural pathway exists for this species to reach Yellowstone Lake. The nearest possible source populations are in Tiber and Ft. Peck [150] reservoirs in the Missouri River drainage of Northern Montana, at travel distances of 560 km (6+ hrs) and 720 km (7+ hrs), respectively. We will implement monitoring for cisco population expansion using multimesh gillnetting of the lake pelagic zone, sampling for larvae, sampling for cisco eDNA, and by examining the stomachs of gillnetted lake trout. Cisco coevolved with lake trout and are their preferred prey in their native range within the Laurentian Great Lakes. If cisco become abundant in Yellowstone Lake, they could compete directly with cutthroat trout for zooplankton and other food resources, while providing additional prey for lake trout. Genetic analyses are being completed to confirm the fish caught in 2019 as *C. artedi* and possibly determine the source population for the illegal stocking.

#### **14. Discussion**

Yellowstone National Park is the site of successful restoration programs for iconic wildlife populations of American bison, gray wolves, grizzly bears, and trumpeter swans. Restoring cutthroat trout and the natural ecology of Yellowstone Lake, however, has been the most challenging restoration effort faced in the park's long history. The park was able to implement a suppression program that killed 3.35 million lake trout because scientific evidence strongly supported it, a must-harvest angler regulation was applied, numerous press releases and reports highlighted the urgent need, and a strong constituency for the lake trout fishery (that could have opposed suppression) never formed [151]. Moreover, the mission of the NPS is to preserve unimpaired the natural and cultural resources and values of the national park system for the enjoyment, education, and inspiration of this and future generations [152]. Cutthroat trout supported a fishery at Yellowstone Lake with historical importance and underpinned what had been the most intact naturally-functioning ecosystem in the lower 48 states. Therefore, allowing invasive predatory lake trout to persist and further degrade these nationally significant resources was not acceptable.

#### *14.1. Suppressing an Invasive Population Below Carrying Capacity*

Yellowstone Lake was perfectly suited for invasion and proliferation of predatory lake trout. The lake offered large expanses of deep, unoccupied habitat, devoid of co-evolved enemies [153], and rich in available resources [154] such as cutthroat trout forage that facilitated their establishment and expansion. Lack of thermal segregation (Figure A1) exacerbated interactions between lake trout and cutthroat trout throughout much of the year (mid-October through mid-June). Most of the lake (64%; 21,810 ha) is <60 m deep [155], depths at which lake trout thrive. The lake is mesotrophic and more productive than most lakes within the native range of lake trout. Lake trout in Yellowstone Lake therefore grew rapidly [7], matured early, and had high fecundity [80] (r-selected traits) [156,157]. Suppression of this invasive lake trout population, which never reached carrying capacity (K) [80,158], was therefore problematic. Managers use a total biomass yield threshold of 0.5 kg ha−<sup>1</sup> yr−<sup>1</sup> to avoid population collapse of lake trout fisheries in oligotrophic lakes in their native range [77,159]; yield densities of 95% of 145 populations in North America were <3.42 kg ha−<sup>1</sup> yr−<sup>1</sup> [95]. We did not drive the population into decline until biomass yield (suppression) reached 4.4 kg ha−<sup>1</sup> yr−<sup>1</sup> in 2012 (Figure 6B). Therefore, gillnetting effort applied during the first 17 years (1995–2011; 68%) of the 25-year suppression program was insufficient to force the lake trout population into decline (i.e., λ < 1; Figure 6B).

Shifts in habitat use by the expanding lake trout population probably increased their survival and enhanced the complexity of our gillnetting suppression efforts. Although gillnetting effort greatly increased during 1995 to 2011, lake trout abundance increased in advance and the population expanded lake-wide (Figure 7). The lake trout population probably used the most productive habitats early in the invasion [160] because the species is adaptable and able to colonize new environments that satisfy their basic habitat requirements [161]. Discrete 'islands' of rocky substrates provided spawning opportunities to the expanding population as it radiated outward from the West Thumb. Island biogeography theory predicts that habitat use is driven by a balance between colonization and extirpation [162]. Predation is a potent force driving species sorting along environmental gradients in freshwater habitats [163]. High predation risk in one habitat may cause a shift to another habitat where risk is lower. Our 'predatory' gillnetting efforts on spawning sites discovered early in the program (e.g., Carrington Island) to target spawning adults may have forced straying and pioneering of new spawning sites in more remote locations of the lake that were unknown or not targeted. Site fidelity and selection pressure would cause greater use of the most successful spawning habitats by each successive generation. Successful spawning at a site may have been caused by better habitat quality (e.g., favorable thermal characteristics; Figure 15) or simply our lack of awareness of it, which precluded suppression gillnetting there.

Spatially disproportionate suppression effort may have afforded lake trout refuge in remote regions of Yellowstone Lake. Suppression gillnetting crews actively sought lake trout and placed gillnets to maximize catches throughout each season. These efforts expanded unhindered across Yellowstone Lake as the lake trout population expanded (Figure 7) but were limited in the Flat Mountain, South, and Southeast arms, which are within proposed wilderness (Figure 1); 30% (6650 ha) of the 21,810 ha of Yellowstone Lake most suitable for gillnetting (<60 m deep) is within the proposed wilderness. Moreover, large lake trout move into the arms to prey upon juvenile cutthroat trout that emigrate from the upper Yellowstone River and other tributaries during late summer. Boat speeds within the arms were restricted to < 8 km hr−1, and only nonmotorized boats (kayaks and canoes) were allowed in the far southern ends of the arms (including the delta of the upper Yellowstone River). Gillnetting in the arms was also deterred by long travel times and low catches, such that in 2013 only 14% of the total lake-wide gillnetting effort was applied there. The proposed wilderness was functioning as an aquatic (freshwater) protected area [164–166] limiting harvest of lake trout. We therefore completed a wilderness Minimum Requirements Analysis (MRA) in 2014 to suspend boating rules in the arms and allow aggressive targeting and gillnetting of lake trout lake-wide.

#### *14.2. Why are Lake Trout Resilient to Suppression Gillnetting on Yellowstone Lake?*

High early life history survival may buffer the effects of suppression gillnetting on the lake trout population. Sustainable exploitation of a fishery requires a reproductive surplus that can be removed [167–170]; harvest that exceeds this surplus can cause population collapse [171–173]. Survival of lake trout in their native range is regulated *sensu* [174] during early life stages [175,176] but survival of pre-recruits in Yellowstone Lake may be 4–6 times higher [81] thereby requiring a 67% increase in gillnetting mortality at later stages to reduce population abundance [80,81]. The unoccupied habitat of Yellowstone Lake may provide lake trout a juvenile-survival advantage similar to that afforded to spawning common carp, arguably the most harmful invasive fish in the world, in predator-free, winterkill-prone shallow lakes [177,178]. Interstitial embryo predators such as sculpin *Cottus* spp. and crayfish *Orconectes* spp. [179,180] and fry predators such as rock bass and yellow perch [181] that are common in the native range of lake trout do not exist in Yellowstone Lake, which is naturally species-depauperate because of its isolation and elevation. Larval lake trout can therefore stay on spawning sites later into the summer, feed more, and achieve greater maximum lengths before dispersing [123]. The lack of predation on pre-recruit lake trout in Yellowstone Lake may afford the population an 'ecological release' [81,182], thereby buffering it against our suppression efforts.

High mortality of suppressed age classes may enhance survival of pre-recruit lake trout and add to population resilience through an an overcompensatory response to gillnetting mortality. Subjecting a life stage of a population to mortality can increase the abundance of other life stages and the total population [174]. Density-dependent processes can thereby confound removal efforts [183], resulting in positive population-level effects [132]. For example, intensive removal of age-0, juvenile, and adult smallmouth bass for 7 years from a north-temperate lake in New York, USA, reduced population biomass but increased population abundance, primarily by increasing juvenile abundance [184]. Size-selective mortality (i.e., uneven mortality across life stages) *sensu* [185] can elicit a similar response. For example, a large decrease in abundance of adult Eurasian perch resulted in a corresponding increase in juveniles [186]. Reduced competition among adult survivors increased somatic and reproductive growth, and juvenile survival was higher after release from cannibalism, collectively resulting in overcompensation. Similarly, the age composition of the Yellowstone Lake population shifted to predominantly younger fish as we increased gillnetting effort and targeted adult lake trout. Age-2 fish composed 26%–43% of total abundance during 1998 to 2004, but increased to 48%–55% during 2014 to 2018 [81] and 69% in 2019. Large-mesh gillnetting for 8 years (2012–2019) successfully reduced adult (age 6+) lake trout abundance by 79% and reduced total population biomass. However, abundance of age-2 fish did not change appreciably (Figure 10A). In Yellowstone Lake, per capita recruitment of lake trout at low levels of spawner abundance (Figure 11), pre-recruit survival [81], and maturation of age-4 and older fish (Figure A6) are all high. All of these characteristics can result in overcompensation whereby population abundance increases in response to harvest [183,185]. Moreover, variation in adult biomass (gradient from high to low, 2012–2019) may have gradually reduced competition, increased fecundity, or allowed a shift from alternate-year to annual spawning. Cannibalism of embryos and other early life stages may also be reduced, further enhancing pre-recruit survival. Although the actual mechanisms are unknown, such compensatory responses may impede attempts to curtail lake trout population growth in Yellowstone Lake.

#### *14.3. Transition to Suppression of Multiple Lake Trout Life Stages*

The realization that high survival of pre-recruit lake trout may offset increased mortality of older age classes has heightened interest in an IPM approach targeting multiple life stages with complementary suppression methods on Yellowstone Lake. Specifically, intense treatment of lake trout spawning sites with carcasses or organic pellets may mimic habitat degradation in their native range to increase mortality of embryos or fry, or both, and thereby decrease recruitment, especially as fewer adults spawn at fewer sites. Location and characterization of primary spawning sites and assessment of the quantity and quality of embryo-deposition habitat and hatching success will be critical. Methods that focus on early life-history stages to disrupt and reduce lake trout reproduction and recruitment should prove effective if spawning continues to be concentrated in shallow (<20 m) lake areas.

Considerable uncertainty and built-in time lags deter significant reductions in suppression gillnetting effort. Currently, an estimated 95,000–100,000 gillnetting effort units will be required annually for 5 years to achieve a 90% probability of reducing lake trout abundance to 100,000 fish [81]. However, we expect that a combination of gillnetting and embryo suppression will probably be used to maintain the lake trout population below target levels, and population-level effects of embryo suppression cannot be distinguished from the effects of on-going gillnetting that targets lake trout adults. The SCAA model estimates that 55,000 effort units will be required to maintain suppression after the target abundance of 100,000 fish is met, but the estimate includes uncertainty, and any resurgence in lake trout (caused by premature or excessive reduction of suppression gillnetting) would not be detected until they recruit to our monitoring and suppression gillnets at age 2. A management correction would not occur until the following year, giving the lake trout population a full 3 years of recovery. Therefore, a reduction of suppression gillnetting should only be made with extreme caution and vigilant monitoring.

#### **15. Conclusions**

Lake trout are being harvested from Yellowstone Lake at a greater rate than ever before anywhere on Earth [187]. This ecosystem restoration program illustrates that predatory fish invasions can be managed and controlled over large areas, even if total eradication may not be feasible. The process, however, requires a long-term commitment, is laced with uncertainty, and requires a great deal of collaboration. Program development, learning, demonstrating need, and building capacity to implement suppression actions at a large scale all require considerable time. Our adaptive management approach allows the program to move forward and implement conservation actions despite uncertainty in outcomes. Continuous learning from assessments and feedback obtained during annual science panel reviews are used to adjust lake trout suppression or other actions to progress towards desired conditions. This approach is used due to the varied environments and stressors (e.g., whirling disease and drought) impacting cutthroat trout in Yellowstone Lake, and the fact that some uncertainty exists in the possible responses by cutthroat trout and lake trout to future management actions. For example, although science-based findings indicate that lake trout population growth in Yellowstone Lake has been curtailed, uncertainty remains in the estimates of the number of years that high levels of suppression will need to be maintained to reduce the population to target levels (100,000 fish). Similarly, the rate of cutthroat trout recovery after the population is released from overriding lake trout impacts is also uncertain, as are the responses of avian and terrestrial wildlife. In the future, due to their use of shallow lake areas and dependence upon tributary streams, cutthroat trout may be more greatly harmed by climate-induced change than lake trout, which solely inhabit the comparatively stable, deep lake environment. The presence of cisco as a new, additional invader that functions as prey for lake trout (co-occurring exotic prey and exotic predator) *sensu* [35] further complicates matters. Lake trout suppression will become even more critical as these new threats emerge. Modeling of the cutthroat trout population is currently being conducted to better understand demographics and potentially refine objectives to include measures of stock biomass because lake trout predation resulted in a shift in cutthroat trout size structure to dominance by large fish (Video S4). Performance metrics will continue to be refined and monitored to track system responses to lake trout suppression, and the results will continue to be used to make adaptations and adjust management actions each year.

#### **16. Materials and Methods**

#### *16.1. Lake Trout Suppression Netting*

Up to six boats (Figure A4) were used to capture lake trout with sinking gillnets during late-May to mid-October 1995–2019 [97,188]. Suppression netting consisted of small-mesh (25 to 38 mm) and large-mesh (44 to 76 mm) bar measure gill nets targeting lake trout at depths typically greater than 20 m to reduce cutthroat trout bycatch. Nets were set shallower than 20 m at known spawning locations during peak spawning activity in autumn. Gill net soak time was typically three to four nights. Annual effort (effort unit = 100-m net per night) was 249 units in 1995 and increased to highs of 97,397 units and 96,971 units in 2018 and 2019, respectively (Figure 6A). Trap nets were also used during 2010 to 2013 to target large lake trout (i.e., >450 mm) [42,188]. Four to 10 trap nets were deployed at fixed locations throughout Yellowstone Lake each year. The netted lake trout were cut to puncture air bladders and then returned to deep (>65 m) regions of Yellowstone Lake.

#### *16.2. Cutthroat Trout and Lake Trout Gillnet Assessment*

Within Yellowstone Lake, cutthroat trout population metrics and individual characteristics (e.g., relative abundance, size structure, body condition) were assessed by standardized gillnetting programs. In mid-September during 1980 to 2010, gillnet surveys were conducted at 11 fixed sites throughout the lake (historical gillnetting assessment; Figure 1) [72,106]. At each site, five sinking experimental gill nets were set overnight perpendicular to shore in shallow water. Nets were set 100 m apart with the near-shore end about 1.5 m deep. Nets were 1.5 m in height and 38 m length, consisting of 7.6 m panels of 19- to 51-mm bar measure.

In 2011, a new protocol (long-term gillnetting assessment) was developed and implemented through 2019 to encompass monitoring of both cutthroat trout and lake trout. During the long-term gillnetting assessment program, 24 sites throughout the lake were sampled each year with a split-panel design to maximize spatial coverage and power for detecting temporal change [120]. Thirty-six sites were originally selected randomly with 12 designated to be revisited each year and the remaining 24 split into two panels of 12 sites each that were revisited every other year on an alternating basis. The sampling occurred after establishment of the lake thermocline during early August with a total of six experimental gill nets per site (Figure 1). At each site, a small-mesh and large-mesh sinking gill net were set overnight at each of three depth strata [epilimnion (3 to 10 m), metalimnion (10 to 30 m), and hypolimnion (>40 m)]. Small-mesh gill nets were 2 m in height and 76 m length, consisting of 13.7-m panels of 19- to 51-mm bar measure. Large-mesh gillnets were 3.3 m in height and 68.6 m length, consisting of 13.7-m panels of 57- to 89-mm bar measure. Gill nets were set perpendicular to shore and nets within a stratum were set parallel 100 m apart. Only the gillnets set in the shallow stratum were used to assess cutthroat trout. Gillnets at all three depth strata were used to assess lake trout.

Both the mid-September historical gillnetting assessment and August long-term gillnetting assessment were conducted for a period of 4 years to ensure that cutthroat trout mean CPUE and size structure were similar and a continuous dataset among the years of both monitoring programs could be compared, 1980–2019 (Figure 12). Discontinuing the historical gillnetting assessment of cutthroat trout in 2010 allowed additional time for NPS crews to focus on lake trout suppression activities during the critical autumn spawning period.

Relative weight for individual cutthroat trout was calculated using the equation Wr = (W/Ws)\*100, where W = measured weight and Ws = standard weight predicted from a cutthroat trout (lentic) standard weight equation log10(Weight) = a + b \* log10(Length), where a = −5.192 and b = 3.086 [189]. One-way analysis of variance (ANOVA) was used to compare mean CPUE, individual weight, and relative weight for each length class among the four decades (α = 0.05). If a statistical difference among decade means was detected, a post hoc Tukey's honestly significant difference multiple comparison procedure was used to test for differences between decade means. Data were manipulated using the "dplyr" package [190] in Program R and analyzed using Program R [191].

Relative condition (Kn) for individual lake trout was calculated using the equation Kn=(W/W')\*100, where W = measured weight and W' = predicted weight of a fish of the same length from a lake trout average weight-length equation log10(Weight) = a + b \* log10(Length), where a = −5.589 and b = 3.210 [95]. One-way analysis of variance (ANOVA) was used to compare mean Kn for each length class among the five time periods (α = 0.05). If a statistical difference among means was detected, a post hoc Tukey's honestly significant difference multiple comparison procedure was used to test for differences between means of each time period.

Lake trout length at 50% maturity (L50) was estimated from lake trout (female n = 1766; male n = 2812) captured in gillnets in Yellowstone Lake during 1996 to 2019. Maturity stages (i.e., immature or mature) were assigned macroscopically, proportion of mature lake trout by 10 mm length bins was calculated, and sex-specific logistic regression models were fit to the proportional data. Length at 50% maturity was estimated from the model's inflection point and confidence intervals (95%) of L50 were estimated with bootstrapping using the bootCase() function from the "car" package [192] in program R [191]. Data were manipulated using the "dplyr" package [190] in Program R and analyzed using Program R.

#### *16.3. Cutthroat Trout Tributary Spawner Assessment*

Visual surveys for spawning cutthroat trout and bear activity were conducted annually during 1989 to 2019 on 9 to 11 tributaries located along the western side of Yellowstone Lake between Lake and Grant (Figure 1) [51,105]. Spawning reaches were delineated on each tributary, and the standardized reaches were walked in an upstream direction once each week from May to July. Observed cutthroat trout were counted, and the activity by black bears and grizzly bears was estimated by noting the presence of scat, parts of consumed trout, fresh tracks, and/or bear sightings. The average number of spawning cutthroat trout observed per visit was obtained by dividing the total observed (in all 9 to 11 tributaries combined, through the entire spawning period) by the number of surveys conducted.

#### *16.4. Cutthroat Trout and Lake Trout Angler Catch*

Because of its remote location, largely roadless (proposed wilderness) shoreline, and the short period of time (approximately 4 months annually) that local supportive facilities are open (campgrounds, gas stations, marinas), the angler effort on Yellowstone Lake is extremely low as compared to other large lakes in the Western U.S. We estimated angler effort and success via a report card distributed to all anglers when purchasing a special use permit for fishing in the park [72]. Annually, approximately 4000 anglers (10% of all park anglers) have voluntarily completed and returned cards for analysis. More than 9000 anglers fished Yellowstone Lake in 2019.

#### *16.5. Lake Trout Population Modeling and Gillnetting E*ff*ort Benchmarks*

The Yellowstone Lake lake trout population is assessed annually using an integrated SCAA assessment model that incorporates time-series of data from both suppression netting and long-term gillnetting assessment programs [80,102]. The SCAA model is age structured and encompasses ages ranging from age 2 to age 17, with the last age class an aggregate group including all fish age 17 and older. Age 2 is the age of recruitment in the SCAA model because younger age lake trout are not frequently captured in suppression or assessment gillnets. The SCAA model uses time-series of observations from the suppression gillnet program, the suppression trapnet program that ran from 2010 to 2013, and the long-term gillnetting assessment. The SCAA model generates predictions of abundances at age of the lake trout population based on model-based estimates of recruitment levels, abundances-at-age in the first assessment year, and underlying mortality levels for the different fishery components in operation and assumed natural mortality levels. Conditional on the predicted abundances at age, the SCAA model predicts suppression and assessment netting harvest and age-composition of harvest, which are compared to observed values. Predictions from the SCAA model can then be combined with other population descriptors (e.g., length at age, length-weight relationships, maturation at age, fecundity) to estimate the stock-recruitment relationship for the lake trout population, total and age-specific fishery yield, total and age-specific population biomass. In combination, the analyses and modeling results provided a robust time series prediction of the lake trout population, which in turn, can be used to gauge the success of the suppression program in decreasing lake trout abundance in Yellowstone Lake and strategize future efforts.

Annually, the SCAA model is used to estimate the amount of gillnetting effort required to cause an abundance decline in the lake trout population (i.e., λ < 1) and, in recent years (2017–2019), to achieve an abundance goal of 100,000 lake trout. These estimates have been critical for the restoration program as they have dictated the numbers of crews, boats, nets, and other gear (and therefore funding) needed to achieve suppression targets. The amount of housing required for the crews, which is extremely limited in the Yellowstone Lake area, was also driven by the suppression targets established from the assessment modeling. Annually, information gained from monitoring and suppression gillnetting was used to lengthen the time series of the data components that feed into the SCAA model. In turn, the annual gillnetting effort benchmarks evolve as the model updates the most recent estimates of abundances and mortalities (Figure 6A).

#### *16.6. Monitoring for Ecological Response*

A goal of the Yellowstone Lake ecosystem restoration is to restore the natural ecological role of native cutthroat trout. The lake trout-induced stress on cutthroat trout caused trophic shifts over the past four decades across multiple trophic levels both within and outside of Yellowstone Lake [41]. Hypothesized outcomes of lake trout reduction and cutthroat trout recovery is that these altered trophic levels will revert to their natural, pre-lake trout conditions (Table A3). To document the cascading changes that may occur due to lake trout suppression, we monitor several components of the aquatic and terrestrial ecosystems.

Aquatic ecological monitoring occurred during ice-free seasons and included measurements of zooplankton density, biomass, and size from samples collected at four lake regions (Main Basin, West Thumb, South Arm, and Southeast Arm; Figure 1) [41]. Phytoplankton biomass was estimated using chlorophyll a, and light transmission was measured using a Secchi disk in West Thumb. The thermal structures of Yellowstone Lake (e.g., isotherm depths) were measured in the West Thumb using a multiparameter sonde (Hydrolab Surveyor). Temperature was also measured routinely at the lake's surface. Lake surface levels, ice-on and -off dates, and outlet discharge (Yellowstone River at Fishing Bridge) were also obtained annually.

Avian and terrestrial consumers of cutthroat trout were annually monitored to document potential recovery. The number of breeding pairs and nesting success was determined for bald eagle and osprey populations each breeding season by surveying all forested areas up to 1 km from the Yellowstone Lake shoreline, connected tributaries, and forested islands using a fixed-wing Super Cub airplane [41]. Bear use of spawning cutthroat trout was documented during visual surveys for spawning cutthroat trout (as described above) on 9 to 11 tributaries located along the western side of Yellowstone Lake. Activity by black bears and grizzly bears was estimated by noting the presence of scat, parts of consumed trout, fresh tracks, and/or bear sightings along spawning stream corridors.

#### *16.7. Permits and Ethical Aspects*

This study was performed under the auspices of Montana State University Institutional Animal Care and Use Protocol 2018-68. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

**Supplementary Materials:** The following are available online, Video S1 (Native cutthroat trout and the Yellowstone Lake Ecosystem, http://doi.org/10.5281/zenodo.3820758); Video S2 (Gillnetting invasive lake trout http://doi. org/10.5281/zenodo.3829258); Video S3 (Organic pellet application to Carrington Island spawning site http: //doi.org/10.5281/zenodo.3829479); and Video S4 (Angling for restored native cutthroat trout http://doi.org/10.5281/ zenodo.3829613).

**Author Contributions:** Conceptualization, T.M.K., M.E.R.; methodology, J.L.A., P.E.B., C.R.D., P.D.D., B.D.E., R.E.G., D.J.M., T.J.S., L.M.T., A.V.Z.; formal analysis, T.M.K., P.E.B., T.O.B., C.S.G., J.M.S., N.A.T.; investigation, J.L.A., P.E.B., C.R.D., P.D.D., B.D.E., R.E.G., D.J.M., T.J.S., L.M.T., A.V.Z.; supervision, P.J.W.; project administration, T.M.K., P.J.W.; funding acquisition, T.M.K., J.D.D., D.P.S.; writing—original draft preparation, T.M.K.; writing—review and editing by all authors. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Yellowstone Forever, grant number G-022; George B. Storer Foundation; Greater Yellowstone Coalition; Idaho Council of Trout Unlimited; Montana Council of Trout Unlimited; Montana Fish, Wildlife and Parks; Montana State University; National Parks Conservation Association; National Park Foundation; University of Wyoming; U.S. Fish and Wildlife Service; U.S. Geological Survey; Whirling Disease Initiative, Montana Water Center; Wyoming Council of Trout Unlimited; Wyoming Game and Fish Department; Wyoming Wildlife and Natural Resource Trust; and the U.S. National Park Service, Yellowstone National Park.

**Acknowledgments:** Special thanks to the nearly five hundred seasonal NPS biological science technicians, Student Conservation Association (SCA) interns, contract gillnetting captains and crewmembers, and long-term volunteers that contributed to restoration of Yellowstone Lake over the past 25 years, This project greatly benefitted from management by Wayne Brewster, Jennifer Carpenter, David Hallac, Lynn Kaeding, Daniel Mahony, James Selgeby, S. Thomas Olliff, and John Varley, with strong support by park superintendents Michael Finley, Suzanne Lewis, Daniel Wenk, and Cameron Sholly. Partners advancing restoration science include Lindsey Albertson, Julie Alexander, Michelle Briggs, Kerry Gunther, Billie Kerans, Dominique Lujan, Thomas McMahon, Silvia Murcia, Alex Poole, James Ruzycki, Douglas Smith, Kole Stewart, and Jacob Williams. Along with NPS and contract staff the SCA program (www.theSCA.org) has provided immeasurable support for recovering cutthroat trout. The Rocky Mountains Cooperative Ecosystems Studies Unit (www.cfc.umt.edu/cesu) facilitated numerous agreements supporting critical research. We thank all Scientific Review Panel members 1995–2019—recent members include Michael Hansen, Michael Jones, Christopher Luecke, Ellen Marsden, Patrick Martinez, Jason Stockwell, Jack Williams, and Daniel Yule. Christopher Downs provided comments that greatly improved this manuscript. We greatly thank Allison Klein for producing all manuscript figures.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results

**Dedication:** This case study is dedicated to the memory of Jacqueline J. Koel, loving mother of Todd Koel, who sadly lost her decade-long battle with Parkinson's disease during preparation of this manuscript.

#### **Appendix A**

**Table A1.** Common and scientific names of birds, fishes, and mammals referred to in text for locations within Yellowstone National Park and elsewhere.


<sup>1</sup> Likely *C. artedi*, confirmation by genetic analysis in process. <sup>2</sup> Subspecies designation.


*Fishes* **2020** , *5*, 18

> **Table A2.** Public support played an important role in the Yellowstone

> Lake ecosystem restoration

> program. Outreach occurred via multiple sources using a


**Table A2.** *Cont.*






*Fishes* **2020** , *5*, 18

#### **Table A3.** *Cont.*



**Table A4.** Mean catch-per-unit-effort (CPUE, 100-m net night) of cutthroat trout and lake trout during annual long-term gillnetting assessments on Yellowstone Lake, 2011–2019, with lower (Lwr) and upper (Upr) 95% confidence limits (CL).

**Table A5.** Mean annual total abundance and biomass, and abundances of age-2, age-3 to age-5, and age-6+ lake trout at the start of the year from 2012 through 2019 with lower (Lwr) and upper (Upr) 95% confidence limits (CL) estimated using a statistical catch-at-age (SCAA) model [81].



**Table A6.** Mean relative condition (Kn) with lower (Lwr) and upper (Upr) 95% confidence limits (CL) for three length classes (mm) of lake trout captured in Yellowstone Lake during periods of lake trout population growth (1995–1999, 2000–2004, and 2005–2009) and periods of population decline (2010–2014 and 2015–2019).

**Table A7.** Mean catch-per-unit-effort (100-m net night), individual weight (g), and relative weight with lower (Lwr) and upper (Upr) 95% confidence limits (CL) of each of three length groups (mm) of cutthroat trout from annual gillnetting assessments during each decade (1980–2019) on Yellowstone Lake.


#### **Appendix B**

**Figure A1.** Depths of isotherms (◦C) measured using a multiparameter sonde (Hydrolab Surveyor) in the West Thumb of Yellowstone Lake during a portion of the ice-free period, 2012–2015. The thermal structure of Yellowstone Lake is typically unstable with a weak and variable thermocline in July, August, and September. During nine months of each year, there is no thermal cause for segregation of invasive lake trout and native cutthroat trout. Only the upper 30 m are shown for better resolution of surface water temperatures.

**Figure A2.** Native cutthroat trout accessed the upper Yellowstone River and Yellowstone Lake from the upper Snake River via natural connections across the Continental Divide following glacial recession about 14,000 years ago. Cutthroat trout then evolved as the sole salmonid and dominant fish within the lake and its connected river network.

**Figure A3.** Invasive lake trout are a large-bodied, long-lived, and cold-adapted predatory species that became inadvertently introduced to Yellowstone Lake and were first discovered in 1994. They then became established as a new apex predatory trophic level within the lake. Because they are deep-water dwelling and do not use tributary streams, they are inaccessible to piscivorous avian and terrestrial wildlife and do not serve as an ecological substitute for native cutthroat trout.

**Figure A4.** Lake trout gillnetting boats on Yellowstone Lake included (**A**–**C**) National Park Service *Freedom, Hammerhead,* and *Cutthroat,* and (**D**–**F**) Hickey Brothers Research, LLC. *Kokanee, Patriot,* and *Northwester*.

**Figure A5.** Length-frequency distributions of lake trout sampled during annual long-term gillnetting assessments on Yellowstone Lake with total number sampled (*n*), percentage ≤ 280 mm (*n*1), and percentage ≥ 400 mm (*n*2) each year, 2010–2019.

**Figure A6.** Proportions of mature female (n = 1766) and male (n = 2812) lake trout captured in Yellowstone Lake from 1996–2019 in 10-mm length bins with estimated length at which 50% of fish were mature (L50; black symbols at inflection points) and 95% confidence limits (in parentheses).

**Figure A7.** Relative condition (Kn) for three length classes (mm) of lake trout captured in Yellowstone Lake during periods of lake trout population growth (1995–1999, 2000–2004, and 2005–2009) and periods of population decline (2010–2014 and 2015–2019). Same letters (a–c) indicate no statistical difference in mean Kn between the five time periods for each length class. Over the past 25 years, the relative condition of large lake trout (400+ mm) has increased. Relative condition of smaller size classes did not appreciatively change.

**Figure A8.** Length-frequency distributions of cutthroat trout sampled during annual long-term gillnetting assessments on Yellowstone Lake with total number sampled (*n*), percentage ≤ 280 mm (*n*1), and percentage ≥ 400 mm (*n*2) each year, 2010–2019.

(**A**)

(**C**)

**Figure A9.** (**A**) The angular-rock substrate surrounding Carrington Island in the West Thumb is prime lake trout spawning habitat in Yellowstone Lake. (**B**) Scuba divers inspect the spatial coverage of lake trout carcasses, and (**C**) organic (soy and wheat gluten) pellets spread by helicopter to induce decomposition, reduce dissolved oxygen concentrations, and increase mortality of lake trout embryos.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
