Theodore Castro-Santos, Kevin B. Mulligan, Micah Kieffer, Alexander J. Haro

U.S. Geological Survey-Leetown Science Center, S.O. Conte Anadromous fish Research Center, One Migratory Way, Turners Falls, MA, 01376, USA

ABSTRACT

Keywords:

fish passage

River herring

Blueback herring

Plunge pool

Survival

1.Introduction

Downstream passage at hydroelectric dams poses several risks to diadromous fish species. Passage through turbines can lead to blade strike or other injuries (Brown et al., 2014; Ferguson et al., 2008;Richmond et al., 2014), but passage through alternative routes can also injure fish (Cada, 1998; Cada et al., 2006; Heisey et al., 1996). Risks do not begin or end with the passage event however: pre-passage delays incurred while fish seek alternate routes can reduce survival and fitness(Castro-Santos & Haro, 2003; Castro-Santos & Letcher, 2010; McCormick et al., 1998), and disorientation associated with passing through hydraulically energetic environments like turbines and spillways can lead to elevated predation risk (Odeh et al., 2002). Although these risks are widely recognized, actual data quantifying them can be rare. Such data are needed if managers are to make informed decisions that optimally balance fishway design criteria with the operational constraints of hydroelectric facilities.

The most common solution to downstream passage is to bypass fish around the turbines through some form of spillway or sluice. The spilled water passes over the dam and into a plunge pool that is designed to dissipate the kinetic energy of the flow and minimize stress and injury to the fish (Castro-Santos & Haro, 2006). This approach requires that flow be diverted through the bypass, and this flow is typically not recovered for generation, meaning that potential generating capacity is lost.Furthermore, although a well-designed plunge pool will minimize contact by the fish with solid structures, other hazards remain. Turbulence and shear forces in plunge pools can be severe, causing mechanical injury or death (Deng et al., 2005; Johnson et al., 2003; Neitzel et al.,2004). In some cases, this diverted flow can lead to supersaturation of gases in the tailwater area, which can lead to embolisms and other fitness-reducing injuries (Geldert et al., 1998). Concerns over minimizing flow dedicated to spill have led to important advances in downstream fishway design that have both improved passage rates while simultaneously reducing losses to generation (Adams et al., 2014;Haro et al., 1998; Johnson et al., 1995). Despite these advances however, few data exist that quantify how discharge and plunge pool configuration affect subsequent survival of downstream migrants.

Of the studies that do address this issue, most focus on migratory salmonids, particularly from the Columbia River basin (Johnson et al.,2003). On the East Coast of North America, however, and in many other rivers throughout the Northern Hemisphere (Navodaru & Waldman,2003; Waldman, 2003), anadromous clupeids are the dominant migratory species. These are known to be more fragile than salmonids and suffer greater mortality when subjected to conditions typically found in plunge pools (Neitzel et al., 2004). Although populations may be relatively insensitive to dam-induced mortalities among juveniles (Boreman& Friedland, 2003), mortality incurred from dam passage remove juveniles from the prey base and other important ecological roles in both rivers and coastal ecosystems (McDermott et al., 2015). Because of this,there is a pressing need for data on the effects of plunge pool configuration on survival of downstream migrant juvenile clupeids. This study addresses this need and describes results of a series of experiments that quantify survival of juvenile blueback herring (Alosa aestivalis) subjected to a simulated downstream bypass scaled to match common conditions found in coastal rivers of Northeast USA, with particular focus on plunge pool depth and discharge quantity.

2.Methods

2.1.Experimental apparatus

Facility. Experiments were performed at the U.S. Geological Survey’s S.O. Conte Anadromous fish Research Center (CAFRC). With its three main flow-through experimental flumes, the CAFRC facility was explicitly designed to test fish passage with live, actively migrating fish,and has the capacity to pass discharges ＞10 m/s (Castro-Santos & Haro,2006). Experiments were conducted in the facility’s east flume, which measures 3 m wide by 6 m deep by 40 m long. Flow is diverted from an adjacent hydroelectric power canal on the Connecticut River and discharges to the river downstream of the Turners Falls Dam (rkm 191).

Experimental bypass structure. We created an experimental testing structure to quantify survival of wild downstream-migrating juvenile blueback herring passing over a simulated weir crest and through a plunge pool (Figs. 1 and 2). The experimental bypass weir (1.2 m wide ×1.1 m long) was mounted atop steel bulkheads that allowed for variable height of the weir crest above the flume floor. The plunge pool consisted of the concrete floor and walls of the flume (3 m wide ×9.1 m long),delimited at the downstream end by a 2.3 m tall flip gate which was hinged at the bottom and raised and lowered using an electric hoist. This allowed us to regulate the depth of the plunge pool, while maintaining a constant drop of 3 m between the water surface above the weir crest and the plunge pool surface (Fig. 2A-C).

Downstream of the flip gate a screen (12 m long ×2 m tall) made of perforated aluminum plate (50% open with 5 mm openings) was set at a 10angle to the flow (Fig. 2 E). Water and fish flowed over the weir crest, through the plunge pool, and the excess flow was discharged through this screen, guiding the fish after they exited the plunge pool towards a collection pen (Fig. 2 E&F). This pen was a welded aluminum cube (2.1 m on each side) frame, faced with 3 mm nylon netting and a tapered floor that led to a ringed fyke and a weighted, 17 L transport tub(Fig. 2 G&H). At the end of each trial this collection apparatus was raised and lowered within the test flume using an electric hoist.

2.1.1.Hydraulic data

The use of the flip gate to control the plunge pool depth resulted in different hydraulic conditions (e.g. depth, velocity, and turbulence)downstream of the gate for each treatment. To better understand these conditions, three-dimensional velocity components were measured in the study area downstream of the plunge pool. Data points were located along the face of the guidance screen (Locations 1-3; Fig. 1), at the collection pen entrance (Location 4), and within the collection pen(Location 5). The velocity was recorded at 20% and 80% of the total water depth at twelve unique x- and y-coordinate positions. Velocity data along the face of the guidance screen were used to estimate the impinging (i.e., perpendicular to the screen) and sweeping (i.e., parallel to the screen) velocity vectors.

Fig. 1.Schematic of the experimental apparatus at the USGS Conte Anadromous fish Research Center, including locations of velocity measurements (Table 2). Note that locations 1-3 are along the screen, 4 is at the entrance to the recovery pen, and 5 represents mean conditions within the recovery pen. Block arrows indicate direction of flow and fish movement.

2.1.2.Trials and treatments

When permitting downstream fish passage in the United States,regulatory agencies recommend the ratio of height of drop: plunge pool depth of

＞

4:1 and a minimum plunge pool depth of 120 cm(USFWS, 2017). A plunge pool volume: discharge ratio of

＞

10 : 1 is also recommended (Odeh & Orvis, 1998). We selected a suite of four treatments to assess whether common designs at low-head dams provide optimal protection for downstream migrating herring. All trials were configured with a 3 m drop from the head pond nappe to the mean water surface level in the plunge pool. Three treatments were performed with a discharge of 1.7 m/s and plunge pool depths of approximately 60, 120, or 180 cm. Henceforth these conditions are referred to as ‘High-Q-060’, ‘High-Q-120’, and ‘High-Q-180’ (Table 1).The tested condition that most closely approximates typical conditions at low-head dams is the High-Q-120 treatment. A fourth, low- flow treatment was selected to evaluate the effects of a much higher pool volume: discharge ratio (‘Low-Q-120’: 0.3 m/s discharge and 120 cm pool depth; Table 1).

Table 1Treatment conditions for all trials. Trial is presented as date (in yymmdd format) followed by a letter indicating the order in which it was conducted. Condition indicates the nominal treatment condition and matches values in Table 2. ‘Control’ indicates whether a given trial was an experimental (T) or control (C) test; Ntot is the total number of juvenile blueback herring introduced to the trial; Q indicates discharge; and headpond and plunge pool indicate the depth (in cm) of the headpond and plunge pool, respectively.

Table 2Hydraulics data by treatment. Data are mean downstream flow velocities for the plunge pool cross-section (). Sweeping (parallel to screen) and impinging(perpendicular to screen) velocities are shown for locations 1-3 (Fig. 1). Mean longitudinal velocities are presented at the entrance of the collection pen (location 4).Velocities for locations 1-4 were taken at 20% and 80% depths, as measured from the flume floor. Relative discharge (m3/ s-) passing through the screen (% Q) as well as the mean velocities passing through the screen openings (mean pore velocity) were calculated for the entire discharge screen. At Location 5, mean velocity magnitude as well as turbulence intensity (TI) and turbulent kinetic energy (k) are average values taken at 16 locations within the collection pen.

Table 3Output of Cox Regression model of treatment effects on mortality rate. Treatments are named according to Table 1, with n the High-Q-120 condition being the null case against which other treatments and controls are compared. Coefficients indicate the effect of each treatment on the ln(hazard): a positive value indicates an increased mortality rate relative to the standard condition (Treatment High-Q-120); a negative value indicates a reduced mortality rate. Mortalities are presented as ‘events’, and fish that survived to the end of the experiment are considered ‘censored’.

At the beginning of each trial, approximately 100 fish were brailed from pre-trial holding tanks and transferred to a custom-designed portable injection hopper (0.33 mvolume: 110 cm long ×60 cm wide ×50 cm deep (Fig. 2D); brailing is a netting technique whereby fish are crowded into a non-porous container (such as a bucket) and transferred to another tank-this method avoids physical contact between the fish and netting or any other solid surface to help minimize scale loss and pathogen transfer). The hopper tank had a tapered floor fitted with a flush-mounted discharge pipe with a 30elbow to direct fish with the flow. Water was retained within the hopper using a standpipe inserted into the discharge pipe; when removed this allowed fish to be rapidly and simultaneously injected into the experimental bypass at the flow nappe just upstream of the weir crest. The hopper was mounted on a wooden frame, allowing it to ride in the bed of a pick-up truck for transport between holding tanks and the flume building(a distance of about 200 m), and an electric crane was used to hoist the hopper tank assembly from the truck and position it several cm above the injection site. Controls were identical to treatments except that instead of being introduced to the water surface above the weir, fish were introduced to the water surface outside of the plunge pool, just downstream of the flip gate. In this way control specimens were exposed to the downstream screening and collection system, but not to the weir,drop, and plunge pool.

Fig. 2.Photographs of experimental apparatus (Fig. 1) showing bypass weir at (A) High-Q-60; (B) High-Q-120; (C) High-Q-180; and (D) Low-Q-120. Panel D also shows the injection hopper in place at the moment that fish are released into the flow. Panel E shows the guidance screen looking downstream toward the collection pen; Panel F shows the same screen with the primary out flow in the foreground. Panels G and H show the collection pen, including the tub and juvenile herring during a trial.

Fig. 3.Kaplan-Meier survivorship curves (Kaplan & Meier, 1958) showing proportion of fish surviving under each treatment. Each trial is represented by a single curve, with treatment conditions (two trials per treatment) indicated with solid lines and controls (one trial per treatment) indicated by dashed lines. Follow-up observation times were ≥96 h; all fish were released at the same time, however, so early trials had greater follow-up times than the last trial (Table 1).

During trials fish typically passed rapidly through the plunge pool and into the collection pen. To ensure that all fish were recovered, at the end of each trial the flip gate was lowered to the floor, staff entered the upstream section of the structure and used a push seine to crowd remaining fish into the collection pen. After all specimens were successfully corralled into the collection pen, the pen was raised, and fish were guided through the ringed fyke into the collection tub. The tub was fitted with a lid, hoisted out of the flume, and transported back to the outdoor holding tanks where the fish of each trial cohort were gently released from the tub into a holding tank for post-trial monitoring.Throughout the entire process care was taken to avoid contact between the fish and nets, and final placement into the holding tank was performed by submerging the transport tub into the tank and allowing fish to volitionally swim out as the tub was removed.

All trials were performed over the course of three days (September 10-12, 2018; temperature range: 21.0C-23.5C). Trials lasted 30 min;each of the four treatments was tested twice and was also matched with a control trial, resulting in a total of 12 trials. Each treatment pair and their associated control were conducted consecutively on a single day(timing of controls was randomized relative to treatments). Weather,temperature, and river conditions were stable throughout the experiments.

2.2.Study animals

2.2.1.Capture

All juvenile blueback herring used for this study were collected during a single night (September 6, 2018; 26C), via purse seine from Weathers field Cove, a setback of the Connecticut River located at rkm 80, just downstream of Hartford, Connecticut. More than 1600 specimens were collected after sunset from 3 purse-seine hauls (Devine et al.,2018). Pursed specimens were transferred with minimal net contact into aerated buckets in a companion vessel, motored to shore, and transferred into two 600 L aerated tanks mounted on a pick-up truck and filled on-site with river water. Highway transport from the capture site to CAFRC (120 km) took roughly 80 min. Upon arrival at CAFRC,200-300 individuals were transferred into each of 6 pre-trial holding tanks, again using brailing and buckets to minimize contact with nets.The holding tanks were round,fiberglass tanks (180 cm diameter ×75 cm depth) supplied with flow-through ambient Connecticut River water drawn from the power canal (16-18 L/min). Water was introduced into the tank through custom-built manifolds that created a gentle clockwise current. A 5.1 cm diameter internal standpipe covered with a protective mesh prevented specimens from escaping through the out flow. An additional small-mesh nylon net secured over the top of each occupied holding tank prevented leaping escapes. The southern half of each tank was also covered with a black plastic sheet to provide a shade refuge.

2.2.2.Post-trial fish handling

Holding tanks for post-trial recovery and monitoring were identical to those used for pre-trial holding. Each trial cohort was assigned to a separate tank; fish were introduced as described above, and were then held and monitored for ≥96 h. Upon completion of the experiments fish were released to the Connecticut River just downstream of the laboratory outfall (rkm 191).

Throughout the pre- and post-trial holding periods, fish were visually inspected at least once every 4 h from 0500 h-2300 h, with a maximum interval of 6 h between inspections during the overnight hours. Any erratically swimming or dead fish were removed immediately,measured, and time of death was recorded. All herring were fed daily ad libitum with 1 mm commercial fish pellets, which they took readily.

Note that throughout the handling process great care was taken to minimize handling and contact with nets. The only time fish actually contacted a net was during the initial collection and when removing dead fish. All other transfers were done with fish fully submerged in water.

2.3.Analysis

Mortality rates were assessed using survival analysis methods, which have been shown to provide more precise measurement of treatment effects than simple binary,fixed-time responses (Castro-Santos & Haro,2013; Hosmer & Lemeshow, 1999). Start time for each post-trial period began at the end of the trial and time of death was measured from this time. Follow-up was ≥96 h for all trials. At the end of the experiment survivors were counted and released with release time considered a censored observation for all survivors.

Effects of treatments and their associated controls on mortality rates were tested using Cox proportional hazards regression (Castro-Santos &Haro, 2003; Hosmer & Lemeshow, 1999; Therneau & Grampsch, 2000).Using this method, the sampling unit is the individual fish, and the dependent variable is the log of the hazard rate (i.e. the instantaneous mortality rate); independent variables were the treatment condition and the associated control nested within treatments. Holding tank (and hence trial) was included as a random effect using the frailty approach(Therneau et al., 2003). To facilitate comparisons of treatments withcriteria recommended by the U.S. fish and Wildlife Service (USFWS,2017) the 120 cm plunge pool with 1.7 m/s (High-Q-120) was included as the null condition and effects of all treatments were compared against this condition.

We tested for non-proportionality of hazards (mortality rates) using Schoenfeld residuals. Sometimes mortality rates can vary over time, and the response to a given treatment may not remain constant. This might occur, for example, if mortality increases following a given treatment but then returns to a baseline that is not different from other treatments.When this occurs it violates the assumptions of the Cox’s proportional hazards approach and must be controlled for (Hosmer & Lemeshow,1999). All statistical tests were performed using R software (version 3.5.1: R CoreR_Core_Team 2018; package ‘survival’: Therneau, 2015).

3.Results

A total of 1655 juvenile blueback herring were subjected to one of the four conditions described above (Tables 1 and 2). During three of the test conditions fish were seen to enter the collection pen immediately following injection into the system, with little evidence of retention within any of the plunge pool conditions. The one exception to this was the low- flow condition, where fish had to be seined downstream out of the plunge pool and into the holding pen under both treatment and control conditions.

There were only two immediate mortalities: these both occurred on the second trial (the control treatment for the High-Q-180 condition as presented in Table 1) and were caused by handling error during transfer to the holding tanks and so were not included in the analysis. Subsequent mortalities were also low for all treatments, with the maximum final mortality of 19% (Fig. 3). Because most fish were released at the end of the study the actual follow-up times varied from 96 h (for the High-Q-120 condition) - 143 h (for the High-Q-180 condition).

Cox regression indicated that significant differences existed among the tested conditions (Table 2). The most dramatic effects occurred among treatments, with greatest mortality associated with the shallow plunge pool, and lowest mortality associated with the low- flow condition (Table 3, Fig. 3). There was no significant difference between treatment and control groups for both the null condition and low- flow treatments. The one exception to this was the deep plunge pool, which had reduced mortality among controls relative to the treatment groups.

Examination of the Schoenfeld residuals indicated that the proportionality assumption was met, meaning there was no evidence of a change in mortality rate relative to the treatment condition. Importantly, this indicates that the elevated mortality associated with the shallow condition persisted throughout the entire monitoring period.

The elevated mortality associated with the shallow condition occurred for both treatment and control groups. This is important,because it suggests that mortality was not caused only by exposure to the drop, impact, and subsequent turbulence present within the plunge pool.It is important to recognize, however, that in order to produce this shallow condition the flip gate was lowered all the way to the floor and so the turbulence and high flow velocities persisted downstream along the full length of the diffusion screen. Thus, the control groups varied by treatment and included some mortality associated with the treatment conditions.

The hydraulic data support this interpretation. Sweeping velocities were generally greater than impinging velocities but varied along the length of the screen (Table 2, Figs. 1 and 2). At the upstream end of the guidance screen, however, impinging velocities were greatest, and exceeded sweeping velocities under the High-Q-060 and High-Q-120 conditions. The ratio of impinging:sweeping velocities increased from downstream-upstream and was probably even greater upstream of the uppermost measurement location. Note also that negative sweeping velocities were present at 20% depth at the upstream-most measurement location. This suggests that fish that occupied the lower part of the water column were more likely to spend a greater amount of time there and so were more likely to be subjected to impinging flows, particularly at the High-Q-60 and High-Q-120 conditions.

Flow rate into the collection pen was similar among the three high flow treatments but lower at the low- flow condition. Under all conditions both velocity and turbulence were low in the collection pen,meaning that fish that entered the collection pen were subject to quiescent conditions.

4.Discussion

Although hazards to downstream migrant fishes posed by hydroelectric dams have long been anticipated (Bell & Delacy, 1972; Brett et al., 1958; Rogers, 1892) actual field data that explicitly account for contributions of the various mechanisms of injury and associated structures are rare (Bevelhimer et al., 2019; Brown et al., 2014; Castro-Santos & Haro, 2006; Pracheil et al., 2016). This arises for the simple reason that dams produce high-energy environments that make study of individual animals challenging (but not impossible: Baumgartner, Reynoldson, & Gilligan, 2006; Brown et al., 2014). Here, we have used a large-scale laboratory to recreate bypass (spillway) conditions commonly found at low-head dams and we related these directly to fish mortality during ≥96 h following downstream passage.

Overall, survival through this structure was greater than expected: in all but the shallowest condition 96-h survival was ＞90%, and for the low- flow treatment this exceeded 97%. This compares favorably with what has been found for salmonids subjected to plunge pools (Bell &Delacy, 1972; Johnson et al., 2003), but is not consistent with other studies that have found alosine clupeids to be more fragile than salmonids (Neitzel et al., 2004). Clupeids have deciduous scales, and may be particularly vulnerable to contact with solid surfaces and/or diseases that might be transferred through contact with those surfaces or with nets during handling (Farmer et al., 1998; Shackleton, 1988). Moreover those few studies that have examined survival of clupeids subjected to plunge pool-like conditions have been reductionist in nature, simulating particular forces such as shear and turbulence, which may not faithfully represent the conditions that fish actually experience when passing through these structures (Neitzel et al., 2004; Deng et al., 2005; Cada et al., 2006.)

Although overall survival was better than expected, some mortality did occur, and this was affected by treatment conditions. Greatest survival was associated with the low- flow treatment, and worst survival was associated with the high- flow, low-depth plunge pool treatment.This is similar to a recent study on inland species subjected to low-head plunge pools (Bestgen et al., 2018). The low-depth treatment also had elevated mortality among controls; indeed, only the deep, high- flow condition and the low- flow condition had negligible mortality associated with the controls. This is likely a consequence of the challenges of producing the shallow condition-the flip gate was lowered to the floor,meaning that turbulent conditions were not contained within the plunge pool, but instead propagated downstream through the discharge area.Moreover, owing to the shallower depths, impinging velocities were greatest under this condition. Although no impingement or obvious injury were observed it is possible that exposure to and contact with this screen might have contributed to the elevated mortality (Swanson et al.,2005). This constitutes a flaw in our experimental design, which can be easily rectified in future studies by 1) maintaining a constant height of the bypass weir; 2) matching this with varying plunge pool depths by constructing a variable height floor within the plunge pool; and 3)maintaining a constant depth in the discharge screen and collection pen.

It is important to recognize that, although elevated mortality was observed among the control group for this treatment, this does not mean that the discharge area was the only contributing factor. Mortality was not immediate, but instead remained elevated for the entire follow-up period. Such a chronic elevated mortality rate might result from various factors, and so this does not preclude risks incurred during passage, it merely suggests that conditions within the plunge pool were not the only source of elevated mortality. This has important implications for understanding survival studies in both laboratory and field environments, with as-yet unresolved consequences for the design and analyses of these studies (Dubois & Gloss, 1993; Perry et al., 2012;Zydlewski et al., 2016). In addition to the structural modifications to design mentioned above, future studies should also consider including additional replicates among both controls and treatment trials to better quantify the random effects associated with each trial.

The foregoing discussion notwithstanding, it is important to note that the low- flow condition had markedly better survival than the High-Q-120 condition, which corresponds most closely with the pool volume:bypass discharge ratio recommended by Odeh and Orvis (1998) at low-head dams. These data indicate that existing designs offer less than optimal protection for downstream migrating clupeids and suggest that further study is needed to better understand how these conditions can be improved.

The fact that elevated mortality rates for all but the low- flow treatment persisted for at least 96 h raises important concerns for survival studies generally. It is common practice to monitor post-passage survival for only 24-48 h (e.g. Bestgen et al., 2018; Johnson et al., 2003; Mathur,Heisey, McGrath, & Tatham, 1996), and to assume that all treatment-induced mortality occurs within this period of time. In these studies, mortality is typically treated as a binary response (survival vs.death), rather than the result of continuous, and possibly chronic processes (Dubois & Gloss, 1993). This means that studies that are prematurely terminated will tend to underestimate overall mortality rates and will also have lower power to detect differences in these rates. This results in an elevated Type II error rate, with consequent failure to identify and accurately quantify risks associated with downstream passage(Castro-Santos & Haro, 2013; Zydlewski et al., 2016). By applying well-established survival analysis methods, coupled with extended monitoring periods and improved study designs, future researchers will be able to substantially improve our understanding of downstream passage survival.

CRediT author statement

Theodore Castro-Santos: Conceptualization, Funding acquisition,Methodology, Supervision, Investigation, Formal analysis, Visualization, Writing - original draft. Kevin B. Mulligan: Conceptualization,Data curation, Formal analysis, Writing - original draft. Micah Kieffer:Data curation, Writing - original draft. Alexander J. Haro: Conceptualization, Data curation, Writing - review & editing.

Acknowledgements

This work was funded through a collaborative agreement between USGS and Albany Engineering Corporation., Albany, NY (AE) (agreement #18ENLB500001). James Besha Sr. (AE) provided context and guidance on the design parameters for the project, which were then further developed in coordination with Jesus Morales (U.S. fish and Wildlife Service). We extend our thanks to various members of Adrian Jordaan’s laboratory at the University of Massachusetts Amherst(especially graduate students Matthew Devine and Lian Guo) who provided essential assistance in collections of study animals, without which this project would not have been possible. Likewise, John Noreika,Stephen Walk, Samuel Parker, and Kevin Molongoski of the USGS S.O.Conte Anadromous fish Research Center provided vital support in design, construction, and execution of experiments. Any use of trade,firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S Government. Animal care and handling was performed following approved protocols under USGS IACUC#09069.