Laboratory Evaluation of the Survival of Fish Impinged on Modified Traveling Water Screens

Abstract

Traveling water screens (TWSs) modified for fish protection can be used at power plant cooling water intakes to reduce the injury to and mortality of impinged fish. Existing biological efficacy data show that postimpingement survival is highly variable by species and has improved over time as screen designs have incorporated more fish-friendly features. Data with the improved screen designs were largely absent for many freshwater species prior to this evaluation. The mortality, injury, and scale loss rates of 10 species of freshwater fish impinged and recovered with a modified Ristroph-style TWS were evaluated in a laboratory flume. Fish were impinged at a 0.3-, 0.6-, or 0.9-m/s approach velocity. Over 13,000 fish were tested in more than 100 replicates during the study. Mortality rates did not exceed 5% for any species and velocity tested. Despite a general trend toward increasing mortality at higher velocities, velocity was only a significant factor in the mortality of Bluegills Lepomis macrochirus. Injury and scale loss rates were low for most of the species tested, although they were more variable than the observed rates of mortality. There was a trend toward lower mortality, injury, and scale loss in larger fish. In all cases in which fish length was a significant factor, the pattern of decreasing mortality, injury, and scale loss as fish increased in length was consistently observed. These results indicate that TWSs modified for fish protection have the potential to prevent impingement losses at power plant intakes.

Received June 28, 2013; accepted October 25, 2013

Section 316(b) of the Clean Water Act requires that the “location, design, construction, and capacity of cooling water intake structures (CWISs) reflect the best technology available for minimizing [the] adverse environmental impact” resulting from their operation. Adverse environmental impact may occur through the impingement of aquatic organisms on intake screens or the entrainment of early life stages through the 9.5-mm (3/8-in) mesh of a traveling water screen (TWS) that protects the CWIS and the cooling water system from debris (USEPA 2011a). Although recent analysis suggests that the impact of impingement and entrainment losses on fish populations and communities is limited compared with that from overfishing, habitat destruction, pollution, and invasive species (Barnthouse 2013), there is still interest in reducing losses due to impingement and entrainment through improvements in screening technologies.

Conventional TWSs are used at over 83% of U.S. power plants (USEPA 2011b) to prevent debris from damaging critical components of the cooling water system. Screen baskets fitted with coarse mesh (typically about 9.5 mm square) are mounted on continuous loop chains and rotated through the water vertically. These screens are located upstream of circulating water pumps, steam condensers, and other sensitive equipment. As the screen rotates out of the water, debris and fish impinged on the screen mesh are removed with a high-pressure water spray and either collected for disposal or returned to the source water body. In general, these screens are rotated only occasionally when debris accumulates on them. Conventional TWSs do not include any fish-friendly features and do not mitigate the impacts of impingement (USEPA 2011b).

Beginning with the implementation of the Clean Water Act, some TWSs were modified to improve the survival of fish during their impingement on the screens and spray-wash removal. The first modifications to TWSs were made at Virginia Electric Power's Surry Station in 1976 (Allen et al. 2012). The Ristroph screen, named for its designer, J. D. Ristroph, had a screen basket equipped with a water-filled lifting bucket to hold collected organisms as they were carried up with the rotation of the screen (Allen et al. 2012). Laboratory and field studies by Fletcher (1990) identified the buffeting of fish within the fish bucket as a major source of underwater injury. Fletcher (1990) recommended several measures to lessen injury and scale loss among fish collected by TWSs, notably (1) redesign of the fish buckets to reduce turbulence and thus prevent fish from striking the screen structure; (2) modifications to the leading edge of the bucket to reduce hydrodynamic turbulence and prevent fish from escaping the bucket; and (3) utilization of smooth-top woven mesh to minimize descaling when fish contact the screens.

The Salem Generating Station on the Delaware River (Township, New Jersey) made improvements to its TWS to increase postimpingement survival that incorporated many of the features identified by Fletcher (1990). In particular, this station used a nonmetallic fish basket developed by Envirex, Inc. (Ronafalvy et al. 2000). This basket was made of a composite polymer material rather than stainless steel. The added strength of the composite material allowed the use of smaller-gauge wire, which increased the open area, reduced through-screen velocity, and decreased the water force upon impinged fish. The baskets incorporated 6.4-mm×12.7-mm rectangular smooth-top mesh and had a fish bucket with a curved-up leading lip to minimize turbulence in the bucket (Ronafalvy et al. 2000). These baskets were modified in width to fit the test screen and evaluated in this study.

Traveling water screens modified for fish protection are designed to operate continuously to reduce impingement time. As each bucket passes over the top of the screen, fish are rinsed into a collection trough by a low-pressure spray-wash system. Once collected, the fish are transported back to a safe release location in the source water body. Modified TWSs have the potential for wide-scale application in all regions of the country and in all water body types to meet section 316(b) requirements (USEPA 2011c).

Evaluations of the latest generation of modified TWSs have generally shown improved survival over previous screen designs (Allen et al. 2012). The results of impingement survival studies from the full-scale application of modified TWSs have been variable, depending on species-, life stage-, and site-specific factors. Such variability has made it difficult to estimate effectiveness prior to screen installation. For this reason, laboratory studies would be better, because they can control for factors such as water quality and temperature, fish health, and hydraulic variability. Varying those factors that could contribute to postimpingement mortality (e.g., approach velocity) may help determine the underlying mechanisms of postimpingement survival and identify those factors that could be optimized in the field to improve survival at water intakes.

Regulatory attention has been placed on the velocity of water approaching an intake as a major factor influencing impingement (USEPA 2011a). Approach velocity is the speed of water flowing perpendicular to the upstream TWS face and is generally measured as close as practical to the screen face or calculated from the cross-sectional area and design intake flow (EPRI 2000). Many retrofit designs to modify TWSs call for reducing the approach velocity on the assumption that impingement mortality decreases directly with velocity. Quantifying the importance of approach velocity for postimpingement survival would allow the most appropriate (and perhaps least costly) civil/structural or operational modifications to existing CWISs and TWSs.

The objectives of this study were to determine the effectiveness of a modified TWS, specifically (1) to characterize the rate at which impingement occurs and determine whether there is a variation in survival based on the time swimming prior to impingement; (2) to quantify the effect of approach velocity on species-specific fish survival; (3) to assess the postimpingement survival of previously untested freshwater species; (4) to document the type and frequency of injuries that occur following removal; and (5) to investigate the effect of fish length on postimpingement survival.

METHODS

Test facility.—A modified TWS was installed in a 208-m³ (208,198-L) recirculating steel flume equipped with a 150-hp (1 hp = 746 W) bow thruster (U.S. Electrical Motors, Model 638057) to produce water velocities up to 0.9 m/s. The flume was also equipped with a bag filter system to remove suspended particulates and maintain water clarity. The section of the flume where the testing was performed had a maximum depth of 2.0 m and a width of 1.8 m. About 1.2 m upstream of the modified TWS, an isolation screen was used to prevent tested fish from swimming upstream and away from the test area. The isolation screen was articulated so that at the end of a treatment fish swimming upstream in the test area could be crowded toward the modified TWS. A collection trough was positioned immediately downstream of the modified TWS to collect fish removed from the screen by the low-pressure spray-wash. At a power plant, this collection trough would be the beginning of a fish return system that would transport fish back to the source water body. Here, the collection trough (1.6 m long×45.7 cm wide×20.3 cm deep) was modified with an end plate and a 1-mm perforated overflow that maintained water depth at 10.2 cm and allowed the collected fish to be held between collections. Fish that were unsuccessfully transferred to the collection trough or passed by the side or boot seals of the screen were collected in the downstream collection net (Figure 1).

FIGURE 1. Schematic diagram of the test flume, including the isolation screen, traveling water screen, collection trough, and collection net.

The modified TWS tested was shorter than most full-scale screens in operation at CWISs. The water depth and ceiling height constrained the screen to be 2.4 m tall (1.8 m between the centerlines of the top and bottom screen sprockets). The screen baskets were identical to those used previously by Ronafalvy et al. (2000). These baskets were 1.2 m wide and were equipped with 0.64-cm×1.3-cm rectangular, smooth-top mesh. The screen wire diameter was 2 mm (14 gauge).

The screen was designed to rotate at speeds ranging from 2.4 to 10.4 m/min. A rotation speed of 2.4 m/min was selected for testing to maximize the duration of impingement while keeping the screen continually rotating.

The boot section (bottom) of the screen was equipped with a brush to prevent fish from passing underneath the screen. The screen had a fish and debris spray-wash–return system, which included spray-wash headers (one external and two internal pairs), a neoprene rubber flap seal, and a return trough on the downstream side of the screen. Unlike modern screen designs, a single debris–fish trough was used because laboratory height limitations prohibited the use of dedicated fish and debris troughs. Apart from the absence of a fish trough and the shorter height of the pilot-scale screen, the screen was identical to screens that would be installed at CWIS, i.e., the basket dimensions, gaps between the baskets, spaces between the baskets and the screen frame, spray-wash nozzle orientation and water pressure, and the distance between the flap seal and the screen face met manufacturer's full-scale design specifications for U.S. Filter (now Siemens) screens.

Test fish.—Ten species were selected based on their availability and lack of postimpingement survival data with modified TWSs. The test species included Fathead Minnow Pimephales promelas and White Sucker Catostomus commersonii (Harmon Brook Farms, Canaan, Maine); Bigmouth Buffalo Ictiobus cyprinellus, Bluegill Lepomis macrochirus, Largemouth Bass Micropterus salmoides, Channel Catfish Ictalurus punctatus, and Freshwater Drum Aplodinotus grunniens (Osage Catfish Hatchery, Osage Beach, Missouri); and palmetto bass (White Bass Morone chrysops×Striped Bass M. saxatilis), Golden Shiner Notemigonus crysoleucas, and Yellow Perch Perca flavescens (Delmarva Aquatics, Smyrna, Delaware). Fathead Minnows and White Suckers were netted from wild populations by the supplier; all other species were pond reared. Fish were transported to the laboratory either by tank truck or by overnight delivery in coolers in water with supersaturated dissolved oxygen and ice packs.

Fish holding facility.—The fish holding facility consisted of a 32,000-L recirculating system. Fish were held in eight cylindrical tanks (either 800 or 1,600 L) connected to a central pool. Water was pumped through water treatment filters before being returned to the holding tanks. Bag filters and an activated charcoal filter were used to remove particulates (solid waste material, chlorine, and other impurities). An ultraviolet light sterilizer and a fluidized bed (sand) biofilter were used to control bacteria and soluble waste products. The holding facility was equipped with a chiller to maintain water temperatures. Dissolved oxygen, temperature, and salinity were monitored daily. Salinity levels were maintained at approximately 5,000 mL/L to reduce the occurrence of fungal growth, which is common in freshwater aquaculture systems. Hardness, alkalinity, and ammonia levels were monitored weekly and chemically adjusted as needed. Water changes (5–20% of total holding water volume) were performed as needed. It was not possible to control water temperatures within the flume, so holding facility water temperatures were maintained within 2°C of that of the flume and ranged from 10°C to 20°C throughout the study period, which is within the normal range for these species.

A circular flow pattern was maintained in the fish holding tanks to keep fish active and reduce their contact with the tank walls, which could cause scale loss. Fish were fed age-, size-, and species-specific diets, which included live and dried brine shrimp Artemia spp. and solid, commercially available pelleted feed. Fish physiology and behavior were qualitatively assessed daily to screen for external signs of disease, fungus, and infection by parasites.

Velocity measurements.—Velocity measurements were recorded at the beginning of the experiments using a Swoffer (Model 2100) flowmeter to verify that the flume operating conditions produced the desired approach velocities with a relatively uniform flow distribution upstream of the TWS. The Swoffer meter measures velocities in the 0.03–7.5-m/s velocity range with 1% accuracy. For each test condition, approach velocities were measured at three locations and three depths along a transect that was approximately 1.2 m upstream of the screen face. This location was the closest to the screen face that was practical to access from the water's surface. Because the geometry of the flume approaching the screen was uniform, the velocities measured in this location would closely approximate the velocities immediately upstream of the screen face. Velocity measurements were used to develop a predicted bow thruster output curve, such that bow thruster revolutions per minute (rpm) could be used to set the approach velocity for each replicate.

Test parameters.—Seven parameters were measured for tested fish: postimpingement mortality immediately after exposure to the TWS and after 24 and 48 h of holding, rate of injury, percentage of scales lost, swimming prior to impingement, and fish length (to correlate with other parameters).

Fish were exposed to approach velocities of 0.3, 0.6, or 0.9 m/s. These velocities were selected based on an analysis of the approach velocities at existing CWISs (EPRI 2002). Data from 88 CWISs using once-through cooling were used in the analysis. These facilities were selected for analysis because the reports provided by power companies had the necessary information to calculate approach velocities. The calculated velocities were based on CWIS design information, and it was assumed that the intakes were operating at design intake flow (all pumps operational) at the expected low water depth. Thus, the calculated velocities represent the expected highest velocity approaching the TWSs. The intakes evaluated were located nationwide and represent all major water body types (freshwater lakes and rivers, estuaries, oceans, and the Great Lakes). This analysis indicated that the mean approach velocity at existing plants is 0.30 m/s, with a median of 0.27 m/s and a range of 0.09–0.82 m/s. Since mortality was hypothesized to increase with approach velocity, the approach velocities selected for testing erred on the side of being greater than those observed in the field to eliminate the possibility of underestimating the effects of velocity.

For each species, five replicates at each of the three approach velocities were employed. Species were tested one at a time or in pairs. On each day of testing, three treatment replicates and one handling control were completed. Individual species were tested alone or with one other species to form a species group. Species groups were created to reduce the time required to complete testing in order to ensure more consistent length frequencies across all replicates within each test condition. Care was taken to select compatible species to form the species groups. Each species or species group was tested sequentially, such that all replicates were run for a single species or species group prior to testing the next species or species group. A randomized design was used to determine the order of the test conditions within each species or species group. On each day of testing, a handling control group for each species was tested to separate the mortality associated with handling (removal from holding facility, marking, counting into test groups, and introduction to and removal from the test flume) from the mortality associated with impingement and removal from the screens. Handling control groups were randomly assigned a time slot on each test day to ensure that handling control groups were not all tested at the same time each day.

During testing, the screen was rotated at 2.4 m/min and the spray-wash pressure was 68.9 kPa. At actual CWISs, rotation speeds are variable and facilities may use a single speed, multiple speeds (e.g., “high” and “low”), or variable speeds (with variable frequency drives that allow operation over a wide range of speeds). The 2.4-m/min speed used here is relatively slow and was chosen to maximize the duration of impingement while continuously rotating the screen. A spray-wash pressure of 68.9 kPa is within the range used at CWISs for low-pressure fish sprays.

Fish marking.—Test fish were marked a minimum of 24 h prior to testing. A pneumatic marking gun was used to mark 300 treatment and 100 control fish of each species. The marking system used compressed CO₂ to inject biologically inert, micro-encapsulated photonic dye into the base of individual fins. Species- and size-specific adjustments to injection pressure and dye volume were made to optimize the visibility of the marks. Four colors and three fin locations were used to provide 12 unique marks, which allowed the resulting combination of color and fin location to be used once per week (based on four replicates per day×three testing days per week). Uniquely marked release groups allowed fish to be identified to the replicate from which they were released, regardless of when they were collected. Mark retention was high; of the 14,366 fish collected by impingement during testing, only 60 fish (0.4%) did not have discernible fin marks.

Test procedures.—Once the test facility was started with a rotation sped of 2.4 m/min and 68.9 kPa spray-wash pressure, the approach velocity was set initially at 0.15 m/s. The isolation screen that confined fish to the area immediately upstream of the TWS was lowered into place. A floating net collection pen was placed at the discharge of the return trough. The knife-gate valve that prevented fish from sluicing into the net collection pen from the fish return trough was confirmed to be in the closed position.

For every treatment replicate, 100 fish of each species were introduced just upstream of the TWS in the test enclosure. Once the fish were swimming normally (approximately 30 s), the velocity was increased rapidly to the test velocity. Fish that impinged on and were washed from the screen were held in the fish collection trough and subsequently sluiced into a floating net pen (a new one for each time interval) through the knife-gate valve, which was opened at set intervals of 15, 60, and 120 min. When conducting a control replicate, 100 marked fish of each species being tested were released into the fish return trough. Once the fish were oriented to the flow, the knife-gate valve at the discharge was opened and the fish were allowed to sluice into a net pen.

Immediately following each collection (treatment or control), fish were classified into one of three categories: (1) live (swimming normally with no signs of injury, orientation problems, or abnormal behavior), (2) stunned (struggling, swimming on their sides, floating belly-up but alive, bleeding, or displaying wounds, missing body parts, severe abrasions, or lacerations), or (3) dead (no vital signs, no body or opercular movement, and no response to gentle probing).

Following the 120-min collection, the water velocity was reduced to 0.3 m/s and the mechanical crowder was raised to move fish that were still swimming upstream of the screen into the screen collection buckets. At the end of each collection event, any live or stunned fish (treatment or control replicates) were transferred back to the holding facility in individually marked net pens and held for latent impingement mortality assessment.

Fish that were entrained through the screen or that did not successfully transfer to the collection trough were removed from the downstream collection net once per day at the end of testing. These fish were enumerated and their fork lengths were measured to the nearest millimeter (total length for Freshwater Drum).

Fish condition (live, stunned, or dead) was monitored immediately after collection and at 24 and 48 h following impingement. Dead fish and all fish at the end of 48 h were examined for external injuries and percent scale loss. Any fish unable to maintain equilibrium 48 h after testing was considered dead. External injuries were recorded by type (bruising or hemorrhaging, disease or fungus, eye damage, lacerations or tears, severed body, fin damage, or predation) using methods and scale loss categories similar to those reported by Neitzel et al. (1985) and Basham et al. (1982); the scale loss categories were as follows: <3%, 3–20%, 20–40%, and >40%.

To ensure consistent evaluations, on the first day of injury and scale loss assessment with each species tested, each biologist examined 20 sample fish. Injury and scale loss assessments were compared among biologists, discrepancies were discussed, and common evaluation criteria were developed by consensus. Subsequently, any fish observed with questionable injuries were cross-checked by a second biologist.

Data analysis.—Descriptive statistics of percent mortality ([cumulative number of fish dead during the 48-h observation period/number of fish collected]×100), percent injury ([cumulative number of fish with injuries during the 48-h observation period/number of fish collected]×100), and percent scale loss in each of the four categories ([cumulative number of fish observed {by category} over the 48-h collection period/number of fish collected]×100) were calculated. Ninety-five percent confidence intervals for mortality and injury were calculated using the normal approximation of the binomial distribution from the descriptive statistics for mortality and injury rates. In addition, data were analyzed by logistic regression using SAS software (SAS 2013). Four possible independent variables were included in the analysis: approach velocity, collection time (15, 60, and 120 min after introduction and at time of crowding), observation time (initial, 24 h, and 48 h), and fish length.

Preliminary logistic regression models contained all four variables. However, the impingement responses were not sufficiently distributed over two of these variables (collection time and observation time) to allow a simultaneous analysis. In the case of collection time, at lower velocities fish tended to remain upstream of the screen until becoming crowded. Conversely, at higher velocities, the majority of fish tended to be impinged and collected during the first 15 min. Thus, for any given velocity, the majority of fish fell into only one collection category. Similarly, very little mortality was observed during the initial and 24-h observation periods. In all cases, this lack of differentiation between collection time and observation time prevented the logistic regression models from converging on reliable estimates. Therefore, the model design was reduced to include only velocity and fish length as variables.

Under the theory of generalized linear models (McCullagh and Nelder 1989), a simple one-variable linear expression can be replaced by a linear expression containing many independent variables. These variables can be continuous (such as fish length) or categorical (such as discrete velocity treatments). Maximum likelihood was the preferred estimation procedure because it could be implemented when there was only one subject for each unique combination of independent variables. With maximum likelihood estimation, the statistical significance of each independent variable was assessed through a chi-square test, which can be thought of as analogous to the F-test used in linear regression.

Thus, the final regression model used was where p is the proportion responding (i.e., mortality, injury, or scale loss) at a given level of velocity, collection, and fork length; α is the intercept; velocity_i is a term to model the velocity treatment (i = 0, 3); collection_j is a term to model the collection effect (j = 1, 2, 3); and β is the coefficient for the effect of fork length on the log-odds.

The analysis of scale loss was based on an extension of the logistic regression to include a multiple-category response variable (McCullagh and Nelder 1989) in which the categories were strictly ordered (1 = <3%, 2 = 3–20%, 3 = 20–40%, and 4 = >40%). For each treatment, the distribution of observations among the categories can be viewed as a cumulative distribution. The probabilities predicted by the model are the cumulative probabilities of the scale loss categories, i.e., p1 = Pr(≤1) = Pr(1); p2 = Pr(≤2) = Pr(1 or 2); p3 = Pr(≤3) = Pr(1, 2, or 3); and p4 = Pr(≤4) = Pr(1, 2, 3, or 4) = 1.0. Since p4 is by definition equal to 1.0, the model need only predict the first three. The standard logistic model was formulated for each of these three probabilities: where p_ijk is the cumulative probability for loss category i, treatment j, and fish k; β_0i is the intercept value for the first three scale loss categories; and β_1j is the parameter for each velocity j; (the control is modeled by the intercept and this parameter models the offset from the control for each velocity treatment); β₂ is a coefficient for the effect of length; and L_k is fork length of fish k, e is the base of the natural logarithms.

In this model, the term β₀ models the increase in probability from one cumulative probability to the next, and the term β₁ models a shift of the cumulative distribution from one velocity treatment to the next. This feature causes the predicted probabilities for one velocity treatment to be proportional to the predicted probabilities for another. That is, there is a constant odds ratio across scale loss categories between a pair of velocity treatments. For this reason, this model is sometimes called the proportional odds model.

The logistic regression model results are hierarchical, such that if the model is not significant and fails to converge to a reliable estimate, the length and velocity components are not significant regardless of their P-values.

RESULTS

A total of 163 treatment and control replicates were generated over 33 d between May 11 and August 24, 2005. Over 19,000 fish were tested (n = 19,401). Of these, 13,009 (67.1%) were used for data analysis. Of the remaining fish, a majority (22.7% of the total number released) were collected downstream of the screen (i.e., were entrained or passed between the flap seal and the screen during transfer from the screen to the fish return trough). Others were either unaccounted for (1.7%), did not have an identifiable mark (0.3%), or were collected in a subsequent replicate (8.4%). Because fish collected in subsequent replicates were exposed to multiple velocities and crowding events, they were not included in the analyses.

Observed postimpingement mortality was low under all conditions tested (<5%). Greater mortality rates were observed at higher velocities, but this effect was significant only for Bluegills (P = 0.001; Table 1). Injury rates were more variable by species, with significantly higher injury rates (P < 0.05; Table 2) being observed at higher velocities for many species. Velocity was also a significant indicator of scale loss for the six species (Bigmouth Buffalo, Bluegill, Freshwater Drum, Fathead Minnow, Golden Shiner, and White Sucker) with successful logistic regression models (P < 0.05; Table 3). Fish length was an important factor in mortality, injury, and scale loss. In all cases in which length was a significant factor, there was a trend toward decreased mortality, injury, and scale loss as fish grew larger.

TABLE 1. Summary of pertinent P-values from the maximum likelihood estimates for the mortality regression models for all species tested. Asterisks denote significant differences at P < 0.05; down arrows indicate that mortality decreased significantly with increasing fish length and plus signs that treatment fish exhibited significantly greater mortality than control fish.

Species	Model	Length	Velocity	0.3 m/s	0.6 m/s	0.9 m/s
Bigmouth Buffalo	0.823
Bluegill	<0.001*	<0.001* ↓	0.0005*	0.8860	0.0004* +	0.0601 +
Channel Catfish	0.634
Freshwater Drum	0.190
Fathead Minnow	0.215
Golden Shiner	<0.001*	<0.001* ↓	0.724
Palmetto bass	0.992
Largemouth Bass	0.055	0.077		0.117
White Sucker	<0.001*	<0.001* ↓	0.742
Yellow Perch	0.039*	0.005* ↓	0.314

TABLE 2. Summary of pertinent P-values from the maximum likelihood estimates for the injury regression models for all species tested. See Table 1 for additional details.

Species^a	Model	Length	Velocity	0.3 m/s	0.6 m/s	0.9 m/s
Bigmouth Buffalo	0.023*	0.004* ↓	0.268
Bluegill	<0.001*	<0.001* ↓	0.139
Bluegill excluding CA	0.001*	0.007* ↓	0.008*	0.003* +	0.394	0.659
Channel Catfish	0.020*	0.070	0.030*	0.010* +	0.971	0.766
Freshwater Drum	<0.001*	<0.001* ↓	0.996
Fathead Minnow	0.011*	0.192	0.015*	0.009* +	0.304	0.582
Fathead Minnow excluding CA	0.004*	0.096	0.009*	0.003* +	0.329	0.810
Golden Shiner	0.577
Palmetto bass	0.993
Largemouth Bass	<0.001*	0.224	<0.001*	0.276	0.001* +	0.047* +
White Sucker	<0.001*	<0.001* ↓	0.595
Yellow Perch	0.166

^aCA = conspecific aggression.

TABLE 3. Summary of pertinent P-values from the maximum likelihood estimates for the scale loss regression models for all species tested. See Table 1 for additional details.

Species	Model	Length	Velocity	0.3 m/s	0.6 m/s	0.9 m/s
Bigmouth Buffalo	<0.001*	<0.001* ↓	<0.001*	0.278	<0.001* +	<0.001* +
Bluegill	<0.001*	<0.001* ↓	<0.001*	0.134	<0.001* +	<0.001* +
Freshwater Drum	<0.001*	<0.001* ↓	0.001*	0.732	0.001* +	0.002* +
Fathead Minnow	0.003*	0.783	0.001*	<0.001* +	0.545	0.629
Golden Shiner	<0.001*	<0.001* ↓	<0.001*	<0.001* +	<0.001* +	<0.001* +
Palmetto bass	0.376
Largemouth Bass	0.313
White Sucker	<0.001*	<0.001* ↓	<0.001*	0.487	<0.001* +	<0.001* +
Yellow Perch	0.075

Time to Impingement

For most species, impingement times were either short for high velocities or long for low velocities. For most species it appears that there was a threshold between 0.3 and 0.6 m/s at which fish could no longer maintain their position upstream of the screen. However, for other species a substantial number of fish were able to swim at 0.6 m/s for the duration of testing. At 0.3 m/s, most fish were collected during the crowding, indicating that the fish were capable of swimming for 120 min at the target velocity. At 0.9 m/s, most fish were collected 15 min after introduction, indicating that they were unable to swim at this velocity for extended periods (Table 4). This highly skewed distribution of fish collected either at low velocity–long collection time or high velocity–short collection time prevented the use of collection time as a variable in the logistic regression models.

TABLE 4. Percentages of fish collected by species, collection time, and velocity treatment. At the end of the 2-h test, free-swimming fish upstream of the traveling water screen were moved toward the screen with a mechanical crowder; these fish were held and evaluated separately and designated as “crowded” in this table.

		0.3 m/s		0.6 m/s		0.9 m/s
Species	Collection time	n	%	n	%	n	%	Control n
Bigmouth Buffalo	15 min	3	1.2	33	35.1	132	93.0
	1 h	3	1.2	3	3.2	5	3.5
	2 h		0.0		0.0	3	2.1
	Crowded	239	97.6	58	61.7	2	1.4
	Control							295
Bluegill	15 min	66	19.4	225	68.6	317	100.0
	1 h	11	3.2	17	5.2		0.0
	2 h	17	5.0	10	3.0		0.0
	Crowded	246	72.4	76	23.2		0.0
	Control							397
Channel Catfish	15 min	48	23.3	71	48.0	174	70.7
	1 h	58	28.2	26	17.6	40	16.3
	2 h	33	16.0	13	8.8	17	6.9
	Crowded	67	32.5	38	25.7	15	6.1
	Control							490
Freshwater Drum	15 min	15	6.2	46	20.0	203	99.5
	1 h	6	2.5	9	3.9		0.0
	2 h	9	3.7	10	4.3		0.0
	Crowded	213	87.7	165	71.7	1	0.5
	Control							512
Fathead Minnow	15 min	164	41.7	307	89.5	295	99.7
	1 h	25	6.4	16	4.7		0.0
	2 h	36	9.2	9	2.6	1	0.3
	Crowded	168	42.7	11	3.2		0.0
	Control							493
Golden Shiner	15 min	80	21.9	350	93.6	385	99.0
	1 h		0.0		0.0	4	1.0
	2 h	3	0.8		0.0		0.0
	Crowded	282	77.3	24	6.4		0.0
	Control							485
Palmetto bass	15 min	13	5.3	16	6.6	270	75.8
	1 h	5	2.1	19	7.9	45	12.6
	2 h	13	5.3	26	10.8	21	5.9
	Crowded	212	87.2	180	74.7	20	5.6
	Control							498
Largemouth Bass	15 min	4	1.0	215	59.9	344	95.3
	1 h	1	0.3	5	1.4	4	1.1
	2 h	9	2.3	11	3.1	1	0.3
	Crowded	380	96.4	128	35.7	12	3.3
	Control							501
White Sucker	15 min	196	52.5	273	80.5	362	96.8
	1 h	25	6.7	22	6.5	9	2.4
	2 h	17	4.6	7	2.1	3	0.8
	Crowded	135	36.2	37	10.9		0.0
	Control							459
Yellow Perch	15 min	18	15.0	118	75.6	159	98.1
	1 h	10	8.3	10	6.4	3	1.9
	2 h	5	4.2	7	4.5		0.0
	Crowded	87	72.5	21	13.5		0.0
	Control							498

TABLE 5. Mean ± SE FL, total number of fish tested (n), percent dead 48 h after testing, and 95% confidence intervals (CIs), by species and velocity. Confidence intervals were calculated using the normal approximation of a binomial distribution. Freshwater Drum, lacking forked tails, were measured to total length.

		0.3 m/s			0.6 m/s			0.9 m/s			Control
	FL		% Dead			% Dead			% Dead			% Dead
Species	(mm)	n	at 48 h	95% CI	n	at 48 h	95% CI	n	at 48 h	95% CI	n	at 48 h	95% CI
Bigmouth Buffalo	72.1 ± 0.26	245	0.0	0.0–0.2	94	0.0	0.0–0.5	142	0.0	0.0–0.4	295	0.3	0.0–1.1
Bluegill	47.6 ± 0.19	340	0.9	0.0–2.1	328	4.6	2.2–7.0	317	2.2	0.4–4.0	397	1.5	0.2–2.8
Channel Catfish	64.2 ± 0.17	206	1.0	0.0–2.6	148	0.0	0.0–0.3	246	0.4	0.0–1.4	490	0.0	0.0–0.1
Freshwater Drum	84.7 ± 0.22	243	0.0	0.0–0.2	230	0.0	0.0–0.2	204	0.5	0.0–1.7	512	0.2	0.0–0.7
Fathead Minnow	54.5 ± 0.16	393	2.3	0.7–3.9	343	3.2	1.2–5.2	296	1.7	0.1–3.3	493	1.2	0.1–2.3
Golden Shiner	74.3 ± 0.31	365	1.6	0.2–3.0	374	1.3	0.0–2.6	389	1.5	0.2–2.8	485	1.2	0.1–2.3
Palmetto bass	60.6 ± 0.14	243	0.4	0.0–1.4	241	0.0	0.0–0.2	356	0.3	0.0–1.0	498	0.0	0.0–0.1
Largemouth Bass	98.5 ± 0.30	394	1.0	0.0–2.1	359	2.2	0.5–3.9	361	2.8	1.0–4.6	501	0.8	0.0–1.7
White Sucker	79.8 ± 0.35	373	4.3	2.1–6.5	339	4.7	2.3–7.1	374	4.5	2.3–6.7	458	3.1	1.4–4.8
Yellow Perch	49.7 ± 0.27	120	0.8	0.0–2.8	156	2.6	0.0–5.4	162	1.2	0.0–3.2	498	1.0	0.0–2.0

Mortality

The 48-h mortality data indicate that mortality for all species was less than 5% regardless of approach velocity (Table 5). A summary of the pertinent P-values from the species models is presented in Table 1.

The mortality rates for six species (Bigmouth Buffalo, Channel Catfish, Freshwater Drum, Fathead Minnow, palmetto bass, and Largemouth Bass) did not exceed 3.2%. For these species, mortality rates near or equal to zero in some treatments could not be fit by the logistic regression models in order to determine statistical significance among the test variables (Table 5).

The mortality rates for the remaining four species (Bluegill, Golden Shiner, White Sucker, and Yellow Perch) were less than 4.7% (Table 5). The logistic regression models for mortality were significant for all four species (P ≤ 0.039), which allowed the estimation of significance for the test variables. For Bluegills, mortality among the velocity and control treatments ranged from 0.9% at 0.3 m/s to 4.6% at 0.6 m/s (Table 5). The 0.3-m/s treatment was not significantly different from the control while the 0.6-m/s treatment was significantly greater (P < 0.001; Table 1). The length effect was significant (P < 0.001; Table 1), indicating a reduced likelihood of mortality as Bluegills increase in length. Mortality among the velocity treatments with Golden Shiners ranged from 1.2% in the control to 1.6% in the 0.3-m/s treatment (Table 5). These values were not significantly different (P = 0.724; Table 1). There was evidence of a length effect, with the likelihood of mortality decreasing as fish length increased (P < 0.001; Table 1). Mortality among the velocity and control treatments for White Suckers ranged from 3.1% in the control to 4.7% in the 0.6-m/s treatment (Table 5). These differences were not significantly different (P = 0.742; Table 1). There was evidence of a reduced likelihood of mortality as fish increased in length (P < 0.001; Table 1). Mortality among the velocity treatments for Yellow Perch ranged from 0.8% in the 0.3-m/s treatment to 2.6% in the 0.6-m/s treatment (Table 5). These differences were not significant (P = 0.314; Table 1). There was evidence of a reduced likelihood of mortality as fish increased in length (P = 0.005; Table 1).

Injury

The injury data indicate that the rate of injury for most species was low regardless of approach velocity (Table 6). Fish showing clear signs of fungus or other disease were recorded as diseased and included in the analysis.

TABLE 6. Percent injury by species and velocity.

Species	Velocity (m/s)	Percent injured
Bigmouth Buffalo	0.3	1.6
	0.6	1.1
	0.9	3.5
	Control	1.4
Bluegill	0.3	6.8
	0.6	7.3
	0.9	4.4
	Control	4.8
Channel Catfish	0.3	3.4
	0.6	0.0
	0.9	0.4
	Control	0.6
Freshwater Drum	0.3	0.0
	0.6	0.9
	0.9	0.0
	Control	1.0
Fathead Minnow	0.3	23.4
	0.6	18.1
	0.9	13.9
	Control	15.0
Golden Shiner	0.3	2.7
	0.6	4.6
	0.9	2.6
	Control	3.7
Palmetto bass	0.3	0.0
	0.6	0.0
	0.9	0.3
	Control	0.0
Largemouth Bass	0.3	0.8
	0.6	6.1
	0.9	3.9
	Control	1.6
White Sucker	0.3	28.7
	0.6	25.4
	0.9	22.5
	Control	20.0
Yellow Perch	0.3	1.7
	0.6	1.9
	0.9	0.0
	Control	1.0

The two wild-caught species (Fathead Minnow and White Sucker) had much higher rates of injury. However, the rates of injury for these two species were also much higher among control fish, indicating that they were not exclusively the result of screen exposure. The most common injury was fin damage. Other types of injuries and their frequencies for all species combined are shown in Figure 2.

Fish that were swimming and behaving normally during the initial assessment but that were found dead, descaled, or missing eyes or fins after 24 or 48 h were assumed to have died from injuries resulting from conspecific aggression (CA) rather than interaction with the screen. For two of the species evaluated (Bluegill and Fathead Minnow), a substantial amount of CA injury occurred while fish were held in pens during the latent impingement mortality assessment (48 and 11 observations, respectively). To investigate the possibility that CA injuries were influencing the outcome of the injury regression analysis, a second regression was completed that eliminated fish with CA injuries. In the case of Fathead Minnows, eliminating CA injuries did not affect the outcome of the model. In the case of Bluegills, the overall velocity treatment statistic was significant (P = 0.008), with the 0.3-m/s treatment exhibiting significantly greater injury than the control (P = 0.003; Table 2).

The injury rates for Golden Shiners, palmetto bass, and Yellow Perch did not exceed 4.6% under any of the test conditions (Table 6). The logistic regression models for injury for these three species were not significant (model P > 0.05; Table 2), preventing a determination of statistical significance among the test variables. Velocity was not a significant predictor of injury for four of the seven species that had meaningful logistic regressions (Bigmouth Buffalo, Bluegill, Freshwater Drum, and White Sucker; P > 0.05; Table 2). In four of the five comparisons in which velocity was a significant factor in predicting injury (Bluegills excluding CA injuries, Channel catfish, and Fathead Minnows with and without CA injuries), only the 0.3-m/s treatment resulted in greater injury than the control treatment. For Largemouth Bass, by contrast, the 0.6-m/s and 0.9-m/s treatments showed significantly greater injury than the control treatment (P < 0.05; Table 2), while the 0.3-m/s treatment did not.

In general, larger fish exhibited fewer injuries than smaller fish and had significantly lower injury rates among four of the species tested (Bigmouth Buffalo, Bluegill, Freshwater Drum, and White Sucker; P < 0.05; Table 2).

Scale Loss

While scale loss was low for most species, the relationship between approach velocity and scale loss was more consistent among species than were mortality and other types of injury. Most species exhibited an increase in scale loss at higher velocities and a decrease with increasing fish length (Table 3). Species-specific scale loss fell into three distinct groups. The first group, which exhibited very little scale loss, included Freshwater Drum, palmetto bass, Largemouth Bass, and Yellow Perch. For these species, more than 95% of the fish exhibited scale loss of 3% or less (Table 7). More than 90% of the Bluegill control fish also exhibited 3% or less scale loss, but for treatment fish the fraction ranged from 82% to 89%. The second group, which exhibited much greater scale loss, included Bigmouth Buffalo and Golden Shiner. For these species, more than 50% of the fish exhibited scale loss greater than 3%, even control fish. The third group, which exhibited intermediate scale loss, included Fathead Minnow and White Sucker. The percentage of fish exhibiting scale loss greater than 3% in this group ranged from 15% to 51% among treatment fish and 14–19% among control fish.

TABLE 7. Percent scale loss category by velocity and species.

		Velocity
Species	Scale loss (%)	0.3 m/s	0.6 m/s	0.9 m/s	Control
Bigmouth	<3	42.0	16.0	29.6	48.5
Buffalo	3–20	45.3	58.5	21.1	38.0
	20–40	9.4	18.1	15.5	10.2
	>40	3.3	7.5	33.8	3.4
Bluegill	<3	88.8	84.8	82.3	91.7
	3–20	8.5	7.9	7.6	2.5
	20–40	1.5	2.7	6.9	2.8
	>40	1.2	4.6	3.2	3.0
Freshwater	<3	99.6	95.7	96.1	99.4
Drum	3–20	0.4	3.9	3.9	0.4
	20–40	0.0	0.0	0.0	0.0
	>40	0.0	0.4	0.0	0.2
Fathead	<3	76.6	84.8	85.1	86.4
Minnow	3–20	19.3	12.2	11.5	10.1
	20–40	3.8	2.0	3.4	1.6
	>40	0.3	0.9	0.0	1.8
Golden	<3	24.7	23.0	15.9	48.3
Shiner	3–20	40.0	38.5	31.1	40.4
	20–40	27.1	23.3	28.0	9.5
	>40	8.2	15.2	24.9	1.9
Palmetto	<3	97.9	99.2	96.9	98.2
Bass	3–20	2.1	0.8	2.8	1.6
	20–40	0.0	0.0	0.3	0.2
	>40	0.0	0.0	0.0	0.0
Largemouth	<3	99.0	99.2	97.5	98.6
Bass	3–20	0.5	0.8	1.1	1.2
	20–40	0.5	0.0	0.8	0.2
	>40	0.0	0.0	0.6	0.0
White	<3	72.9	59.3	48.9	81.3
Sucker	3–20	17.2	20.7	25.4	13.3
	20–40	7.5	15.0	20.1	3.5
	>40	2.4	5.0	5.6	2.0
Yellow	<3	100.0	98.1	98.2	99.8
Perch	3–20	0.0	0.0	0.6	0.2
	20–40	0.0	1.9	1.2	0.0
	>40	0.0	0.0	0.0	0.0

FIGURE 2. Distribution of injury types for the 886 fish with injuries.

The logistic regression models for scale loss did not yield statistically significant associations for three of the nine species evaluated (palmetto bass, Largemouth Bass, and Yellow Perch; P > 0.05; Table 3). Channel Catfish, a species that lacks scales, were excluded from this analysis. Velocity was a significant predictor of scale loss for Bigmouth Buffalo, Bluegill, Freshwater Drum, Fathead Minnow, Golden Shiner, and White Sucker (P < 0.05; Table 3) but not for palmetto bass, Largemouth Bass, or Yellow Perch (P > 0.05; Table 3). For four of the six species for which velocity had a significant effect on scale loss (Bigmouth Buffalo, Bluegill, Freshwater Drum, and White Sucker), the effect was significant at 0.6 m/s and 0.9 m/s but not at 0.3 m/s (P < 0.05; Table 3). Only Golden Shiners exhibited significant scale loss at all three treatment velocities (P < 0.001; Table 3). Surprisingly, Fathead Minnows showed significantly more scale loss at 0.3 m/s than at 0.6 m/s and 0.9 m/s (Table 3).

Length was a significant factor in predicting scale loss for five of the six species with reliable logistic regressions (Bigmouth Buffalo, Bluegill, Freshwater Drum, Golden Shiner, and White Sucker); the exception was the Fathead minnow (P = 0.783; Table 3). For all five species exhibiting significant length effects, the coefficient for length was positive and significant (P < 0.05; Table 3), indicating that larger fish tended to have less scale loss.

DISCUSSION

Postimpingement mortality was low for all of the species and velocities tested (<5%), indicating that modified TWSs with nonmetallic fish baskets have the potential to limit impingement mortality at CWISs. While fish mortality was low at all velocities, higher rates of injury and scale loss were observed at 0.6 and 0.9 m/s than at 0.3 m/s.

Fish length played an important role in mortality, injury, and scale loss. In all cases in which fish length was a significant factor, the pattern of decreased mortality, injury, and scale loss as fish increased in length was observed. Thus, modified TWSs offer a higher level of protection to larger fish.

Despite a general trend toward decreasing mortality at lower velocities, velocity was only a significant factor in mortality for Bluegills. This suggests that modified screens can be used at CWISs with approach velocities greater than 0.3 m/s. It is unclear why Bluegills exhibited greater mortality at 0.3 m/s than at 0.6 or 0.9 m/s. Further research would be required to determine the underlying mechanism for this. Despite the Bluegill results, visual observations indicated that velocities greater than 0.3 m/s create more turbulence in the fish lifting buckets, which may account for the greater injury and scale loss observed at higher velocities.

Injury rates were generally low but highly variable across species, indicating that the injury of fish in the field will be determined largely by species composition and abundance at the individual CWIS. However, the two species with the highest injury rates (White Sucker and Fathead Minnow) also exhibited greater handling injury among the control fish, suggesting that the individuals from these species were not as healthy as those from the other species tested or that these two species are more susceptible to injury as a result of handling. Although obviously injured fish were culled prior to testing, these two species (which were captured from wild populations) may have had higher injury rates from living in the wild or as a result of handling during their capture.

Scale loss was highly species specific, with some species (Largemouth Bass and Yellow Perch) experiencing almost no scale loss and others (Bigmouth Buffalo and Golden Shiner) experiencing substantial levels of scale loss even among the control fish. It appears that susceptibility to scale loss is associated with scale structure. Although there was considerable variability, fish with cycloid scales (Bigmouth Buffalo, Fathead Minnow, Golden Shiner, and White Sucker) exhibited greater scale loss than species with ctenoid scales (Bluegill, Freshwater Drum, palmetto bass, Largemouth Bass, and Yellow Perch).

There were limitations to testing the modified TWS in the laboratory. Ideally, the screen height would have matched those typically used at CWISs, thereby allowing rotation speeds and durations of impingement comparable to those at CWISs. The height of the TWS used involved a trade-off between screen rotation speed and the duration of impingement; we could either match the rotation speeds used at CWISs and reduce the duration of impingement or decrease the rotation speed and maintain the duration of impingement that would occur at CWISs. Because the water depth in the test flume was limited to 1.5 m, the maximum duration of impingement for fish during the tests was about 40 s (assuming the minimum screen rotation speed of 2.4 m/min and that became fish impinged at the deepest point in the flume). Therefore, it was not possible to evaluate longer durations of impingement while maintaining continuous screen rotation, a feature hypothesized as being important to fish survival. The impingement of fish at 2.4 m/min indicates that there was a behavioral component to the impingement process that would have been lost if the screens had been held stationary in order to maximize the duration of impingement. In many cases, fish would hold their position in the flume and fall back to the screen surface until their tails touched the screen face. The tail touching would, in turn, stimulate the fish to swim forward. This “tail tapping” often occurred several times before the fish entered the bucket. Once in the calm conditions in the bucket, fish ordinarily remained there. Because the investigation sought to evaluate the entire screening process—from impingement to the spray-wash and finally transfer to the return trough—it was deemed necessary to continuously rotate the screen. In addition, since modified TWSs are designed to rotate continuously in the field, continuous rotation in the laboratory is more representative of actual field conditions.

The experimental protocol called for a rapid increase in velocity after fish were introduced into the flume. It is possible that the rapid acceleration caused fish to become disoriented and behave abnormally. However, underwater observations made during preliminary testing indicated that this velocity acceleration was less disruptive to swimming fish than introducing them directly into the higher velocities (0.6 and 0.9 m/s). Since some of the fish may have been impinged during the 30-s period of flow acceleration, some of the velocity effects that we observed may have been a result of rapid acceleration rather than exposure to a constant velocity. Since the mortality, injury, and scale loss rates were low and this interval was less than 0.5% of the total test period (2 h), the data were analyzed as though fish were exposed to a constant velocity.

It might be assumed that site-specific environmental factors (e.g., debris, water quality, water temperature, turbulence, etc.) will impact fish condition and possibly increase postimpingement mortality, injury, and scale loss at CWISs using modified TWSs. Previous research has identified several factors that influence the postimpingement survival of organisms. These factors include the time between screen washes (e.g., Chase 1975; King et al. 1978; Tatham et al. 1978), salinity (e.g., Palawski et al. 1985; Kane et al. 1990; Bowser and Buttner 1991), debris loading (Landry and Strawn 1974), water temperature (e.g., Lifton and Storr 1978; Loar et al. 1979; Lankford 1997), and preexisting disease in the impinged population (e.g., Baker 2007; Knight 2008). In addition, transit through a fish return system back to the source water body may contribute to mortality. However, a 3-year laboratory study evaluating the survival of fish passing through a fish return system indicates that length of fish return (the linear distance between the point of entrance to and exit from a fish return system), water velocity, roughness, drop height, and the presence/absence of debris had minimal impact on survival (EPRI 2010, 2013).

Our results indicate that modified TWSs should result in low impingement mortality and reduce adverse environmental impacts on the species that we tested. It should be pointed out that this study only evaluated velocity and its effects on survival in a controlled laboratory environment. Other biotic, abiotic, and plant operational characteristics of real-world situations could impose additional stress on aquatic organisms and increase the mortality of wild fish. Laboratory evaluation is only one step in the standardized process established by the American Fisheries Society's Bioengineering Section for the effective evaluation and application of fish passage and protection technologies (AFS–BES 2000). Now that laboratory evaluations have demonstrated the potential for TWSs modified for fish protection to protect juvenile and adult freshwater fish, the next logical step toward wide-scale application would be to conduct a prototype evaluation at a field location with species similar to those tested in the laboratory. Such an evaluation would allow researchers to assess the biological efficacy of modified TWSs under the operational and environmental conditions present in a real-world setting. There is accumulating evidence that the fish impinged at modified TWSs are disproportionately ones that were already diseased or injured (e.g., Baker 2007; Knight 2008). Therefore, we suggest that in situ impingement survival testing begin with hatchery-reared fish prior to evaluating the screens with wild fish. Hatchery-reared fish have been used previously in evaluating modified TWS when low numbers of wild fish were impinged (e.g., Bigbee et al. 2011).

ACKNOWLEDGMENTS

A very special thank you to the late Ned Taft, former president of Alden Research Laboratory. He provided value input, guidance, and support throughout the project. These studies were undertaken, in part, as a master's thesis requirement for Jonathan Black at the University of Massachusetts–Amherst (UMass). Thanks go to all the members of the thesis committee: Alex Haro (USGS, Conte Anadromous Fish Laboratory), who chaired the committee and provided perceptive guidance throughout this project; Martha Mather (research associate professor, UMass, and assistant unit leader, Massachusetts Cooperative Fish and Wildlife Research Unit); and Bernie Morzuch (professor, UMass) for their consideration and recommendations. We thank everyone from the Alden Research Laboratory for their input, support, hard work, and helpful reviews of the manuscript. Daniel Davis was involved in every aspect of this study, and his critical advice was invaluable. Brian McMahon developed the facility design and was instrumental in its construction. Tom Cook, Greg Allen, Bob Richmond, Kevin Simes, Brian O’Coin, Jeremy Sinkus, Michael Briggs, and Michael Comeau all assisted in the design and construction of the facility. Steve Amaral and Ray Tuttle assisted in the development of the study design and shared their experience in overcoming logistical hurdles. Daniel Giza, Timothy Sullivan, and Anthony Quinta put in long hours marking, testing, and assessing fish. Nate Olken, Erica Matys, and Eric Grady also assisted in testing. Funding for this study was provided by the Electric Power Research Institute (EPRI). Doug Dixon of EPRI provided frequent and extensive support without which this study would not have been possible. We would also like to thank Dave Michaud (WE Energies); Igor Cherko and Sonya Kim (Omaha Public Power District); Robert Reider (DTE); and Michael Peters, Brad Foss, and John Thiel (Dairyland Power Cooperative) for their comments on the draft manuscript.