Evaluating the effects of an electric barrier on fish entrainment in an irrigation canal in Colorado

Entrainment of fish in irrigation canals is a source of mortality for sport and native fishes and can affect populations and species diversity. To reduce entrainment of wild trout, an electric barrier was installed on the South Canal, an irrigation ditch on the Gunnison River in western Colorado, USA. The objective of this study was to evaluate the effectiveness of the barrier by marking fish upstream of it and estimating fish populations in the canal downstream before and after the barrier was operational. Three groups of fish were tagged and released upstream of the barrier: fish from the canal, wild fish from the Gunnison River, and hatchery‐reared Rainbow Trout. Boat electrofishing was completed in the canal reach below the barrier, and population estimates were made with the Huggins Closed Capture Model. The estimated Brown Trout population of the canal declined following the installation of the barrier, but Rainbow Trout remained stable because of the entrainment of small fish and their growth and survival in the canal. A total of 288 tagged fish less than 300 mm and 4 fish greater than 300 mm were recovered below the barrier, representing 1.3% of all tagged fish. The electric barrier appears to successfully exclude a portion of adult‐sized trout from the irrigation canal, but smaller adult and sub‐adult trout can pass the barrier. The entrainment, growth, and survival of smaller fish maintain a reduced but stable population of fish in the canal, but fewer entrained mature fish is likely a benefit to the populations of the Gunnison River.


| INTRODUCTION
Many freshwater fishes must migrate substantial distances to fulfill specific life history requirements such as feeding, reproduction, and developmental growth (Schlosser, 1991). The ability to move within large river systems and their associated tributaries is vital for salmonid species like Rainbow Trout and Brown Trout which are known to move long distances and have diverse life histories (Behnke, 2002). Habitat connectivity is important for different life stages of stream fishes that use a diverse range of habitats (Fausch, Torgersen, Baxter, & Li, 2002). Construction of dams and water diversions can reduce that connectivity and fragment lotic ecosystems, fundamentally changing a river's temperature and flow regimes as well as nutrient cycles and sediment transport (Ward & Stanford, 1979). Additionally, dams and irrigation diversions are often a direct source of mortality to fishes that may reduce populations and impact biodiversity (Carlson & Rahel, 2007;Moyle & Williams, 1990).
Fish entrainment in irrigation canals is a large problem in the western United States and worldwide (Carlson & Rahel, 2007;Clothier, 1953;Kaeding & Mogen, 2020;Moyle & Williams, 1990;Roberts & Rahel, 2008). The loss of fish in irrigation canals has been shown to be a population sink for trout in Wyoming (Roberts & Rahel, 2008). Because of the limited swimming ability of juvenile fish, they are often disproportionately affected by water diversions, especially during high flow periods (Hooley-Underwood et al., 2018).
Water use in the western United States and the associated infrastructure is expected to increase in the future because of both population growth and a declining water supply related to climate change, further fragmenting aquatic habitats (MacDonald, 2010).
Fish deterrent systems can guide fishes away from sources of mortality such as hydropower turbines and irrigation canals but are also used to limit the spread of invasive species (Jones et al., 2021;Noatch & Suski, 2012). There are two main types of fish deterrent systems: physical barriers, and nonphysical (behavioral) guidance. Physical barriers include screens, netting, drop structures, and low-head dams (Noatch & Suski, 2012). Nonphysical guidance barriers use external stimuli to influence fishes' behavior to divert them from unwanted areas and includes electricity, strobe lights, acoustic deterrents, air bubble curtains, water velocity, chemical deterrents, pheromones, and magnetic fields (Jones et al., 2021;Noatch & Suski, 2012). Both physical and behavioral fish guidance systems have been successful in excluding fish from unwanted areas but they have different costs and benefits that are generally site dependent and highly variable (Kim & Mandrak, 2019).
The South Canal in southwest Colorado diverts an average of 444,793,557 m 3 (360,600 acre-feet) of water each year from the Gunnison River primarily for agricultural use, with 24.3 m 3 /s average daily discharge from March through November. The Gunnison River supports a highly utilized, Gold Medal classified trout fishery. Entrainment of fish in the South Canal has been documented for many years but the effect on the Gunnison River salmonid population was unknown. While entrained fish are lost from the Gunnison River, they retain some value to anglers who target the fish in the canal and in the Uncompahgre River lower in the irrigation system. Construction of a hydropower plant on the South Canal was expected to increase mortality of entrained fish, reducing their contribution to the sport fishery, so an electric fish barrier was installed at the diversion structure in 2012.
The objective of this study was to evaluate the overall effectiveness of the electric barrier in reducing fish entrainment. To accomplish this, fish population estimates were compared before and after the barrier was installed, and tagged fish were used to document movement across the barrier. The need for this research is demonstrated by the large number of fish that have been documented becoming entrained in irrigation ditches in Colorado and the potential effects on native and sport fish populations (Crowley & Ryden, 2019).

| Study area
This study was conducted on the South Canal of the Gunnison River near the town of Montrose in western Colorado (Figure 1). The canal begins at an elevation of 1997 m at the western portal of the Gunnison Tunnel (UTM NAD83 Zone 13, 268993, 4267300) and conveys water from the Gunnison River for irrigation and municipal use in the surrounding Uncompahgre Valley. The Gunnison River below the South Canal is a sixth-order stream with pool-riffle or pool-riffle/plane bed morphology in the Rocky Mountains. It has a mean annual discharge of 35.4 m 3 /s and a drainage basin area of 1605 ha.
From the diversion structure and barrier, the canal travels underground 9.2 km through the Gunnison Tunnel before emerging approximately 800 m above the hydropower plant ( Figure 1). There is a total of 12.4 km of earthen canal which contains the majority of fish that are entrained from the Gunnison River. The canal diverts water from

| Electric fish barrier
The fish barrier was constructed in 2012 and was operational before the 2013 irrigation season. It consists of a series of vertically suspended electrodes across the east portal of the Gunnison Tunnel.
The waterway at the barrier is 22.6 m wide, 4.9 m deep, and has water velocities between 0.2 and 0.7 m/s and conductivity generally less than 300 μs/cm. The system is powered by three Smith-Root 1.5 KVA pulsators with a maximum power output of 4.5 kW and is designed to operate with a frequency of 2 Hz, a pulse width of 0.005 s, and a field strength of 0.4 V/cm. The barrier was designed to exclude "brood stock" Rainbow and Brown Trout but the target size was not specified reach, while representative of the earthen sections of the canal, contains an unknown proportion of the total entrained fish. Fish in the study reach must have passed the electric barrier, navigated the 9.2 km tunnel, avoided entrainment in two small lateral canals, and survived the passage through the hydropower turbines to be detected by our sampling in the canal.

| Fish population estimates
The South Canal was sampled with mark-recapture electrofishing In March 2013, the canal was sampled to estimate the number of fish in the study reach before the barrier was operational and to reduce the numbers of fish as much as possible. Discharge was 0.6 m 3 /s and the canal consisted of two distinct habitat types, the concrete stilling basin and the earthen portion of the canal below, so the two areas were sampled independently. The entire stilling basin was sampled with a bag seine that was 15.2 m long and 1.8 m deep with 0.3175 cm mesh. Multiple seine hauls were made through the stilling basin so a multiple-pass removal population estimate could be made (Anderson, Burnham, & Otis, 1982;Zippin, 1958). Fish were held in a live pen and then measured for total length to the nearest millimeter. The portion of the canal below the stilling basin consisted of a shallow, slow-moving channel that was 1.1 km long and averaged 14.1 m wide. A sampling reach was randomly chosen in this portion of the study area that was 304.8 m long and block nets were used to ensure closure to satisfy the assumptions of the multiple-pass removal model (Anderson et al., 1982). Because population estimates made through electrofishing are commonly biased because of size selectivity of the gear, we took measures to ensure robust estimates (Riley & Fausch, 1992;Saunders, Fausch, & White, 2011). The data were analyzed in Program Mark with the Huggins Closed Capture Model (Huggins, 1989;White & Burnham, 1999). Capture probabilities were modeled with length as an individual covariate, similar to the approach described in Saunders et al. (2011). Four models were built by estimating capture probabilities using length, species, species + length, as well as a constant capture probability for all fish, identical to a Lincoln Petersen model (Seber, 1982). Akaike's information criterion corrected for small sample size (AIC c ) was used to rank model performance. To account for model selection uncertainty, population and parameter estimates were made by averaging AIC c weights across all four models (Burnham & Anderson, 2002). To compare fish population estimates from different years, we computed estimates and 95% confidence intervals with Program Mark. Statistical significance was evaluated at the α = .05 level by determining if an estimate fell within the 95% confidence interval of another estimate. The analysis of the mark-recapture data indicated that modeling capture probabilities by fish length was important in the South Canal to get robust population estimates ( Table 2). The exception was the March 2013 sampling, when the canal contained a lower number of fish and there was little variation in fish size after they had overwintered in an intermittent canal ( Figure 2). Then, a simple removal model with constant capture probability was favored, but models containing fish length still had 27% of the model weight. Modeled capture probabilities for the mark-recapture electrofishing estimates increased with fish size. For example, in October of 2014 the estimated capture probability of a 400 mm fish was approximately 2.9 times that of a 200 mm fish (0.37 vs. 0.13). Estimates using length to model capture probabilities were 23% higher (6-41%) than estimates from a simple Lincoln-Petersen model, and models containing length as a covariate had between 98% and 100% of the model weight across all sampling occasions. Our data support earlier findings that modeling capture probabilities by length is important for making unbiased fish population estimates with electrofishing.

| Tagged fish
In October 2013, after the first irrigation season in which the electric barrier was in use, a total of 248 CWT fish from 123 to 337 mm were documented passing through the barrier: mostly smaller hatcheryreared Rainbow Trout (n = 246, mean length 163 mm, in October,

| DISCUSSION
The electric barrier on the South Canal appears to serve its designed purpose of excluding larger fish from the irrigation system. Some fish in the Gunnison River are successfully passing the electric barrier and surviving the turbines, mostly smaller fish. Their growth and survival in the canal maintains a stable fish population that is significantly lower than before the barrier for Brown Trout (α = .05). This differential effect related to fish size was expected with electrical-based fish deterrence and reflects the established effects of pulsed DC current on fish in both deterrence and electrofishing (Noatch & Suski, 2012;Riley & Fausch, 1992).
The lack of a significant decline in the Rainbow Trout populations in the canal after the barrier is likely related to entrainment of small fish and higher than expected survival and growth of fish living in the canal. In 2014, 17% of all sampled fish (37% >350 mm) had been handled the previous October, denoted by the presence of a healed caudal punch scar. This indicates that there is fair to good overwinter survival in the canal. Growth of fish living in the study reach is also relatively high; CWT Rainbow Trout grew an average of 163 mm from age-1 to age-2. With high growth rates and annual survival, the large numbers of smaller fish that pass the barrier maintain a relatively stable population of fish in the study reach, even though large fish do appear to be excluded from the canal by the electric barrier.
The difference between species is likely due to two factors: larger size of age-0 Brown Trout when water is being diverted, and the potential spawning of Rainbow Trout in the study reach. Because Note: Included are Akaike information criterion corrected for small sample size (AIC c ), the number of model parameters (K), the difference in AIC c values (ΔAIC c ), and model weight (w i ).

F I G U R E 3 Length-frequency histogram of Rainbow Trout captured in the South Canal in October 2013
Brown Trout fry in the Gunnison emerge about 8-10 weeks earlier than Rainbow Trout, they are larger during their first summer (Nehring & Anderson, 1993). Because the barrier is size selective, Brown Trout fry are expected to be entrained at a lower rate than Rainbow Trout. The canal is first filled with water around April 1 of each year, just before Rainbow Trout spawn. Large numbers of age-0 Rainbow Trout were observed in the canal in July 2014. It is unknown if they were entrained fish from the Gunnison River or were spawned in the canal; both are likely. Brown Trout spawn throughout the month of October in the Gunnison River, and flows in the canal generally cease at the end of that month, with subsequent winter flows being intermittent. There is minimal, poor-quality spawning habitat for Brown Trout in the canal compared to that of Rainbow Trout, which spawn at higher flows in the canal that are stable or increasing. A combination of higher entrainment rates and better potential spawning success in the canal likely leads to higher numbers of small Rainbow Trout in the canal.
In this study, we estimated the minimum number of fish that passed the barrier, navigated the 9.2 km tunnel, avoided entrainment in two small lateral canals, survived passage through the hydropower turbines, and were detected in the sampling reach. We can expand those estimates to get an idea of total entrainment rates, but more The results of this study agree with those of other work, indicating that the effects of electric barriers (like electrofishing) is related to fish length as well as site-specific characteristics such as water velocities (Pugh, Monan, & Smith, 1970). The probability of deterring small salmonids with an electric barrier in that study decreased with increasing water velocity, and deterrence was not practical in veloci- Fish Recovery Program, 2012). In Wyoming, it was estimated that between 6300 and 10,400 fish encompassing 10 species were entrained in a single canal system (Roberts & Rahel, 2008). These three examples of large irrigation diversions represent a small fraction of diversions across the west and display not only the large magnitude of fish entrainment that can occur but also the lack of information on the problem overall.
With over 16,000 active irrigation ditches in Colorado and the documented loss of native and sport fish in various locations, the problem of fish entrainment appears to be widespread and potentially of large magnitude in some rivers. Screening canals with mechanical or electric barriers appears to reduce the entrainment but not eliminate it. Success is dependent on site-specific factors, the design of the barrier, as well as its maintenance and operation. Further research is necessary in Colorado and across western United States to document the extent of fish loss in irrigation canals and identify solutions. Closer cooperation between the engineers designing fish barriers and the biologists who manage the fisheries is encouraged, as the local nuances of the river environment and fish populations will dictate the ultimate effectiveness. Electric fish barriers like the one on the South Canal of the Gunnison River can help address the problem, but both site-specific approaches to fish entrainment as well as a larger, riverscape perspective of stream fish populations will be necessary to address it fully.

ACKNOWLEDGMENTS
We thank Jim Heneghan of Delta-Montrose Electric Association, Ron Harthan, Zack Hooley-Underwood, Eric Fetherman, George Schisler, and two anonomous reviewers. Funding was provided by Colorado Parks and Wildlife and the Federal Aid in Sportfish Restoration Program Grant F-237.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.