A Three-Pass Electrofishing Removal Strategy Is Not Effective for Eradication of Prussian Carp in a North American Stream Network

Aquatic invasive species (AIS) have significant negative economic and ecological consequences (e.g., Zipp et al. 2019) and are a major threat to freshwater biodiversity (Dudgeon et al. 2006; Reid et al. 2019). Despite much effort by scientists and managers, research is still needed to understand site- and AIS-specific ecology and management (Gozlan et al. 2010). This knowledge can help identify areas of concern and assess site-specific risk (Ricciardi and Rasmussen 1998), ultimately, to reduce the negative consequences of AIS.

Prussian Carp Carassius gibelio (also known as Gibel Carp) is an AIS introduced into western Europe from central Europe and Asia (Lever 1996; Elgin et al. 2014; Docherty et al. 2017), and it is a prolific invader of freshwater ecosystems throughout Europe and the Middle East (Kottelat and Freyhof 2007). Prussian Carp tolerates a wide range of environments, has high fecundity, and exhibits clonal reproduction via gynogenesis (Kırankaya and Ekmekçi 2013; Docherty et al. 2017), making this species well suited to be invasive (Copp et al. 2005). Prussian Carp was recently confirmed in western North America (Elgin et al. 2014), and there is concern for its spread (Docherty et al. 2017) and negative effects on native aquatic biota (Ruppert et al. 2017). To our knowledge, no one has attempted to physically control Prussian Carp in a North American waterbody, particularly early in its invasion history. Many control projects aim to extirpate or eradicate AIS (i.e., remove every individual) from invaded waterbodies (Gozlan et al. 2010; Britton et al. 2011). There are, however, two important issues associated with the use of eradication as a tool: efficacy (i.e., complete eradication is difficult) and required effort (i.e., monetary and ecological costs are high; Gozlan et al. 2010; Britton et al. 2011). It is unknown whether control by means of removal is even possible in western North America for Prussian Carp.

The physical removal of captured individuals from a waterbody (sometimes called “cropping” or “harvesting”; Britton et al. 2011) is a method that has been used in North America for controlling and containing AIS. This removal strategy differs from eradication because the goal is not to completely remove the AIS from a system but to suppress the AIS to invaded stream reaches by limiting its population size. Typically, practitioners use electrofishing and netting to capture and euthanize individuals. This strategy has been used effectively in several incidences, including the removal of invasive Brook Trout Salvelinus fontinalis by using gill nets in the Sierra Nevada (Knapp and Matthews 1998) and the removal of Common Carp Cyprinus carpio by using a combination of boat electrofishing and radio-telemetered “Judas” fish from lakes in Minnesota (Bajer et al. 2011; Bajer and Sorensen 2012). There are, however, issues that limit efficacy of cropping, including the use of size-selective gear that limits the capture of smaller fish (Knapp and Matthews 1998; Neilson et al. 2004) and increased numbers of juvenile fish due to the lack of competition from removed adult fish (Ludgate and Closs 2003). Generally, however, data are lacking to determine whether removal is effective across a range of fish species and sites and whether specific biotic variables can predict success. Therefore, in the context of adaptive management, it is important to test whether a removal strategy for Prussian Carp will be effective in western North America, despite past failures for other species (e.g., Black Bullhead Ameiurus melas; Cucherousset et al. 2006).

Our objectives were to 1) determine the efficacy of the three-pass removal method for eliminating Prussian Carp in several streams in southern Alberta, Canada; and 2) assess whether the number of individuals removed, the location of the site in reference to the invasion front, or both influenced removal success (change in catch per unit effort [ΔCPUE] between summer and fall sampling periods). The removal method consists of three passes of capturing and removing fish in a closed stream reach by using backpack electrofishing. We hypothesized that removal would be most effective at sites with fewer Prussian Carp present and at sites farther from the invasion front (i.e., farther to the original invasion location).

The study was completed in southern Alberta, an area that is undergoing invasion by Prussian Carp (Elgin et al. 2014; Docherty et al. 2017). This area has a variety of land-use activities such as rangeland, residential land, industrial land, oil and coal deposits, and extensive croplands (Red Deer River Watershed Alliance 2008, 2009). We sampled sites located along tributaries of the Red Deer River, including Rosebud River, Ghost Pine Creek, Kneehills Creek, Lonepine Creek, Michichi Creek, West Michichi Creek, and Threehills Creek (Figure 1). We selected sites based on their invasion history between 2000 and 2014: early introduction (Prussian Carp have been present for 7–12 y), recent introduction (Prussian Carp have been present for 4–6 y), or none previously present (no reports of Prussian Carp within the past 3 y; Ruppert et al. 2017). Control sites were not possible in this study as provincial legislation requires euthanizing captured Prussian Carp. Site-specific environmental data are provided in Table S1 (Supplemental Material).

We used a three-pass removal method to capture and remove Prussian Carp (Zippin 1956; Zippin 1958; Carle and Strub 1978) during the summer (June 22–August 2, 2017). The three-pass removal method consists of three passes of a stream reach by using backpack electrofishing and the capture and removal of any invasive fish captured. At each site, we placed block nets (mesh size = 1/8 in. [0.32 cm]) at both ends of a 150-m reach to ensure a closed system during the sampling day. We used a Smith-Root LR-24 backpack electrofisher (see Table S1 for settings [varied by site]; Smith-Root Inc., Vancouver, WA, USA) and a netter to complete three passes. During each pass, we used consistent effort and a systematic pattern of zig-zagging from one side of the river to the other. We then enumerated the catch, measured each Prussian Carp for length, and euthanized all captured Prussian Carp. We repeated the sampling in a similar manner after a month (September 22–30, 2017). Statistically, we used a two-sample Kolmogorov–Smirnov test to determine whether length distributions of Prussian Carp differed between the two sampling events and a chi-squared goodness of fit test to determine whether the number of juvenile fish captured at each site differed between the two sampling events. During the postremoval sampling event, we only did a single electrofishing pass because Jones and Stockwell (1995) found that a single-pass removal method can sufficiently explain most of the variability in abundance between sites, which was the goal for the postremoval sampling.

We used three metrics to assess the effectiveness of the removal: 1) percent depletion for Prussian Carp captured across all sites, 2) site-specific decrease in abundance of Prussian Carp across the three passes, and 3) site-specific CPUE between the removal and postremoval sampling events. We used the number of sites that reached depletion divided by the total number of sites at which each species was present to calculate percent depletion. We defined depletion as sites having two or fewer individuals captured in the third pass. We used the percent decrease across passes for all sampled sites to calculate decrease in abundance across passes. This metric includes sites that did not fully reach depletion but did result in lower catchability by the third pass. To determine decrease in abundance across passes, we compared the maximum value between the first and second pass to the third pass. Finally, we calculated the change in Prussian Carp CPUE between removal and postremoval sampling events as the difference between the number or individuals per electrofishing second sampled during the first pass only for removal and postremoval sampling events. We used a paired t-test and a Cohen's d calculation to determine whether CPUE differed between the two sampling events and to what extent CPUE differed, respectively.

In addition to the three metrics used to determine the effectiveness of the removal method, we also wanted to evaluate removal success by both number of individuals present and invasion front (objective 2). We defined the number of individuals present as the Prussian Carp population estimate calculated using the removal method (equation 1). We used linear regressions to determine the relationship between change in CPUE and number of individuals present at each site and the relationship between change in CPUE and the distance of the sampled site to the invasion front. We used R for all analyses (R Core Team 2017).

We found Prussian Carp at every sampled site (Table 1) and, in total, we removed 4,292 Prussian Carp. We also captured more individuals during the single-pass postremoval sample (3,274 individuals) than during the three-pass removal sampling event (mean = 668 per pass). Prussian Carp ranged in size from 15 to 255 mm (Figure 2), and the length distributions differed convincingly between the two sampling periods (two-sample Kolmogorov–Smirnov test: D = 0.25; P < 0.001). We also found convincing evidence that smaller fish were present more often during the postremoval sampling, as 70% of fish were less than 60 mm (assumed to be young of year; Sarı et al. 2008; Şaşı 2008), compared with 47% of fish during the removal sampling (chi-square goodness of fit: χ² = 427.4; df = 1; P < 0.001). Overall, there is strong evidence that mean CPUE during the removal sampling event was significantly lower than the postremoval period (0.024 ± 0.026 individual/s vs. 0.100 ± 0.122; paired Student's t-test; t = −3.8249; df = 29; P < 0.001, Cohen's d = 0.862) with sampling period having a large effect on CPUE and sites differing in CPUE by less than 1 standard deviation. Removal success moderately related to distance to the initial introduction site, as success declined 0.2% per river kilometer away from the initial introduction site (linear regression: slope = −0.002 [±0.001], y-intercept = 0.567 [±0.197]; F = 6.285; df = 1,28; P = 0.0183; Figure 3).

During the removal period, we did not deplete all sites to zero Prussian Carp. Detection probability by distance sampled indicated that 50 and 100% detection probability occurred by 30 and 150 m, respectively (Figure 4). However, depletion of Prussian Carp occurred at only 16.7% (n = 5/30) of sites with a mean decrease in abundance across all passes of 50.4%.

Our use of the three-pass removal method for Prussian Carp from various tributaries in southern Alberta was not successful because Prussian Carp were only depleted at 5 of 30 sites (Britton et al. 2008, 2011). In fact, the average decrease in abundance was approximately half of fish present, all sites had Prussian Carp several weeks postremoval, and CPUE increased fivefold following several weeks postremoval. Although three-pass electrofishing was not found to be effective at removing Prussian Carp from the sampled tributaries, removal success was related to distance from the initial introduction site, suggesting that targeted removal at the edge of the invasion front may be more successful. Overall, future control efforts for Prussian Carp in North America should consider more broad-scale actions such as using rotenone on the entire stream network (Gozlan et al. 2010).

Despite our unsuccessful attempt to use the three-pass removal method to remove nonnative species, it has been shown to be effective for other species in different situations (Thompson and Rahel 1996; Peterson et al. 2008; Shepard et al. 2014), and these studies may highlight plausible explanations for the lower effectiveness observed in our study. For example, densities of nonnative Brook Trout in three small Rocky Mountain streams were reduced from 11.3 to 0.6, 3.4 to 0.3, and 2.3 to 0.2 individuals/m², respectively. Thompson and Rahel (1996) used a similar technique to that used in our study but also used additional passes later in the year. Although Brook Trout were still found in the streams at low levels, recruitment was determined to be nonexistent after 1 to 2 y (Thompson and Rahel 1996). Likewise, Brook Trout were successfully eradicated in other small Rocky Mountain streams when the authors used an adaptive approach. For example, the authors used repeated electrofishing passes during the year, taking advantage of autumn spawning and winter aggregations (Shepard et al. 2014). Similarly, in a lacustrine system, Common Carp were successfully removed when the deployment of barriers was designed based on identified spawning aggregations following a fish movement study using radiotelemetry (Taylor et al. 2012). These studies highlight that prior knowledge on species-specific space use and use of an adaptive approach where practitioners can vary gear and sampling time may have higher likelihood of successful Prussian Carp removal from tributaries in western North America when compared to our methods. Future work is certainly warranted given the known negative ecological impacts of Prussian Carp (Docherty et al. 2017; Ruppert et al. 2017).

Our low efficiency may be due to Prussian Carp being inherently difficult to remove because of their unique life-history characteristics. Prussian Carp notably reproduce asexually through gynogenesis (Kottelat and Freyhof 2007; Elgin et al. 2014), meaning a single female fish can cause the population to persist and even proliferate (Docherty et al. 2017). In this study, we show that not only was removal not successful but also, overall, we captured 464% more Prussian Carp during the postremoval sampling compared with the original removal event. There was a clear recruitment effect as we caught more than 3,000 young-of-year Prussian Carp during the postremoval sampling. These juvenile fish were likely spawned either before the first sampling period and were undetectable during the three-pass electrofishing, were offspring spawned following the first sampling period from adult fish that we did not remove, or were fish that migrated to the tributaries in between the sampling periods. Any one, or a combination, of these explanations is plausible. Prussian Carp tend to spawn in the spring and early summer and undergo multiple bouts of reproduction annually (Şaşi 2008), and young-of-year Prussian Carp are known to migrate up to 2 km upstream (and 85 km downstream; Slavik and Bartoš 2004). Furthermore, this species has a strong propensity to invade new habitats, as streams sampled in 2014 that had no Prussian Carp (see Ruppert et al. 2017) were found to have Prussian Carp in our study. Admittingly, we were limited by the area that could be effectively covered during the study periods and were not able to remove fish from all connected waterbodies. Fish were able to recolonize sites, a finding known to have occurred in a small lowland stream in Poland where presence of Prussian Carp varied annually, with some years having zero individuals and other years having hundreds of individuals (Penczak 2015).

Another problematic aspect of attempting to remove Prussian Carp from western North America is that the fish have entered lotic systems, allowing them to disperse between lentic environments and spread beyond their origin (Docherty et al. 2017; Ruppert et al. 2017). Successful removal of any AIS is more likely if the waterbody in question is small, easily accessible, and closed (Kolar et al. 2010). Extensive streams systems, such as the study system, are problematic and it is costly to remove AIS, often requiring sufficient funding and human resources (e.g., Donlan and Wilcox 2007). Most removal techniques work best in lentic systems (Kolar et al. 2010), where the application of a piscicide (e.g., rotenone) or dewatering and disinfection have been shown to work well (see Britton et al. 2008). Furthermore, removal of AIS has a higher likelihood of being successful if streams are frequently monitored and new individuals are rapidly identified (Simberloff 2008). Using removal as a method to control invasive species is often not considered early in the process as data are limited and practical guidance for resource managers is unavailable (Edwards and Leung 2009). Although removal may be a useful technique for some species if considered early enough, life-history characteristics paired with our results indicate that Prussian Carp removal may be difficult.

We deemed our removal method to be unsuccessful for Prussian Carp in southern Alberta; however, some recommendations can be made to assist with future removal of Prussian Carp. First, we recommend collecting data to understand movement patterns and important sites for spawning and overwintering. This information will hopefully make removal easier by identifying where most Prussian Carp are present. Second, reducing the spread of Prussian Carp throughout the stream networks (including irrigation canals) is likely paramount. Frequent electrofishing passes in the streams furthest from the invasion origin may help limit the likelihood that fish will continue to move upstream and entice fish to move downstream where practitioners using different gear types (e.g., trap or seine nets) could remove aggregated fish. Third, consider removal of fish from the lentic systems (e.g., piscicide or dewatering) to reduce propagule pressure on the streams and then target streams during periods when fish might aggregate in pools during dry periods. Fourth, we show that removal success relates to distance from the initial invasion. This result suggests that removal may be beneficial to curb the spread of Prussian Carp by prioritizing cropping closest to the leading edge of the invasion front. Fifth, because Prussian Carp are a high-impact AIS, it would be cost effective to work with biological researchers to identify physiological tools to reduce or eliminate reproduction, cull Prussian Carp by using a virus or disease (e.g., herpes virus has been shown useful in eradicating Prussian Carp; Daněk et al. 2012), or design species-specific barriers (e.g., olfactory cues; Noatch and Suski 2012). Given the fast spread of Prussian Carp (Docherty et al. 2017), their negative impacts to native freshwater diversity (Ruppert et al. 2017), and their potential to survive in most of continental United States (USFWS 2012) and Canada (Mackey et al. 2019), we believe it is prudent to invest in further activities to reduce their spread and removal from the region to avoid significant ecosystem consequences.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author for the article.

Table S1. Environmental (dissolved oxygen [DO], pH, water temperature, conductivity, turbidity) and electrofishing (volts, frequency) data collected during summer and fall 2017 in the Red River Watershed, Alberta, Canada. Data are provided for each sampling site. Site abbreviations (see Figure 1) are as follows: Ghostpine Creek (GC), Kneehill Creek (KC), Lonepine Creek (LC), Michichi Creek (MC, Rosebud River (RR), Spruce Creek (SC), Three Hills Creek (TC), and West Michichi (WM).

Found at DOI: https://doi.org/10.3996/JFWM-20-031.S1 (19 KB XLSX).

Table S2. Depletion survey data collected during summer and fall 2017 in the Red River Watershed, Alberta, Canada. The total number of Prussian Carp Carassius gibelio captured during each pass and the calculated population size and catchability estimates have been provided, with standard errors and associated 95% upper and lower confidence intervals. Site abbreviations (see Figure 1) are as follows: Ghostpine Creek (GC), Kneehill Creek (KC), Lonepine Creek (LC), Michichi Creek (MC, Rosebud River (RR), Spruce Creek (SC), Three Hills Creek (TC), and West Michichi (WM).

Found at DOI: https://doi.org/10.3996/JFWM-20-031.S2 (20 KB XLSX).

Table S3. Total length of each Prussian Carp Carassius gibelio caught at each site and during both summer and fall sampling periods in 2017. Fish were caught in the Red River Watershed, Alberta, Canada at various sites (see Figure 1) by using three-pass electrofishing: Ghostpine Creek (GC), Kneehill Creek (KC), Lonepine Creek (LC), Michichi Creek (MC, Rosebud River (RR), Spruce Creek (SC), Three Hills Creek (TC), and West Michichi (WM).

Found at DOI: https://doi.org/10.3996/JFWM-20-031.S3 (107 KB XLSX).

Table S4. Percent differences in Prussian Carp Carassius gibelio captured between fall and summer sampling in 2017 (i.e., eradication success) as well as distance to invasion front for each site. Fish were caught in the Red River Watershed, Alberta, Canada at various sites (see Figure 1) by using three-pass electrofishing: Ghostpine Creek (GC), Kneehill Creek (KC), Lonepine Creek (LC), Michichi Creek (MC, Rosebud River (RR), Spruce Creek (SC), Three Hills Creek (TC), and West Michichi (WM).

Found at DOI: https://doi.org/10.3996/JFWM-20-031.S4 (18 KB XLSX).

Table S5. Catch per unit effort (CPUE) for each site during summer and fall 2017 sampling period. We defined (CPUE) as number of fish caught during the first pass divided by total seconds of electrofishing. Fish were caught in the Red River Watershed, Alberta, Canada at various sites (see Figure 1) by using three-pass electrofishing: Ghostpine Creek (GC), Kneehill Creek (KC), Lonepine Creek (LC), Michichi Creek (MC, Rosebud River (RR), Spruce Creek (SC), Three Hills Creek (TC), and West Michichi (WM).

Found at DOI: https://doi.org/10.3996/JFWM-20-031.S5 (18 KB XLSX).

Reference S1.[USFWS] U.S. Fish and Wildlife Service. 2012. Prussian Carp (Carassius gibelio). Ecological risk screening summary. Web version – 8/14/2012. Washington, D.C.: U.S. Fish and Wildlife Service.

Found at DOI: https://doi.org/10.3996/JFWM-20-031.S6 (342 KB PDF); also available at https://www.fws.gov/injuriouswildlife/pdf_files/Carassius_gibelio_WEB_8-14-2012.pdf.

Funding for this project was provided by Alberta Health (grant 007282). Thomas Novak, Chen Xin Kee, Tyana Rudolfsen, Karling Roberts, and Zachary Hammond helped with fieldwork. Thomas Novak also assisted with geographic information system analyses. The article was greatly improved by several anonymous reviewers and the Associate Editor. Data used in this study have been included as part of the Supplemental Material.

Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.