Survival and Behavioral Effects of Exposure to a HydrokineticTurbine on Juvenile Atlantic Salmon and Adult American Shad

Theodore Castro-Santos & Alex Haro

Received: 17 September 2012 /Revised: 18 June 2013 /Accepted: 11 July 2013# Coastal and Estuarine Research Federation (outside the USA) 2013

Abstract This paper describes a series of experiments designedto measure the effect of exposure to a full-scale, vertical axishydrokinetic turbine on downstream migrating juvenile Atlanticsalmon (N=175) and upstream migrating adult American shad(N=208). Controlled studies were performed in a large-scale,open-channel flume, and all individuals approached the tur-bine under volitional control. No injuries were observed, andthere was no measurable increase in mortality associated withturbine passage. Exposure to the turbine elicited behavioralresponses from both species, however, with salmon passingprimarily over the downrunning blades. Shad movement wasimpeded by the device, as indicated by fewer attempts ofshorter duration and reduced distance of ascent up the flume.More work should be performed in both laboratory and fieldconditions to determine to what extent these effects are likelyto influence free-swimming fish.

Keywords Hydroelectric . Migration . Fish passage .

Injury . Turbine . Behavior . Telemetry . Salmon . Shad

Introduction

Hydroelectric power development has been historicallyproblematic for the migration, passage, and restoration ofdiadromous and other riverine migratory fishes. Dams aretypically required to maintain the hydraulic head necessaryto efficiently drive turbines, and these dams pose barriers tomovement in both upstream and downstream directions.Downstream migrants are confronted with additional risks,

incurring injuries and mortality as they pass through turbinesand other routes (Bell and Kynard 1985; Bickford andSkalski 2000; Ferguson et al. 2006); even delays associatedwith passage in either direction can reduce fitness (Castro-Santos and Letcher 2010; McCormick et al. 2009). Fishwaysand bypass structures can provide safe passage routes, butthe structures are costly and their performance is often poor.Because of these and related factors, hydropower develop-ment associated with dams is often blamed for the decliningpopulations of migratory and other riverine fish species.

Recently, there has been renewed interest in hydrokineticturbines. These devices are deployed in rivers or tidal zoneswhere the kinetic energy of flowing water drives the turbineswithout requiring construction of dams or other obstacles. In theabsence of dams, fish might pass these structures simply byswimming around them, thereby avoiding what is widely per-ceived as the primary environmental impact of hydroelectricgeneration.

Questions remain, however, as to whether such devices areindeed safe for fish passage or for aquatic communities. Evenwithout a dam, the potential still exists for fish and other aquaticorganisms to be injured by moving turbine blades. Mechanicalinjury is not the only concern, however, and avoidance behav-iors hold their own risks. For example, fish may refuse to passthe structure, in which case access to habitat may still beblocked. Sometimes fish will pass a structure, but at a reducedrate (Castro-Santos and Haro 2003). Such a structure stillconstitutes a barrier, and the resulting migratory delay can havesubstantial consequences. For example, when populations offish concentrate above or below barriers, they may becomeattractive to predators, suffer energy depletion, disease risk,etc. (McLaughlin et al. 2012). Also, delays can alter run timingand prevent fish from accessing essential habitat during keytime windows (McCormick et al. 1998; McCormick et al.2009). Because of the potential importance of behavioral ef-fects, studies of interactions between hydrokinetic devices andfish should not be limited to immediate mechanical injury:

Communicated by Wayne S. Gardner

T. Castro-Santos (*) :A. HaroUSGS-Leetown Science Center, S.O. Conte Anadromous FishResearch Center, P.O. Box 796,One Migratory Way, Turners Falls, MA 01376, USAe-mail: tcastrosantos@usgs.gov

Estuaries and CoastsDOI 10.1007/s12237-013-9680-6

Turning Vane/ RetentionScreen

Raised Floor Turbine

FLOW

Smoltrelease area

Shad Entering From Holding Area

Staging Area

0 5 mOverhead Lights

FLOW

Raised floor

Turbine

Wall-mounted video cameras

Cameras below floor

Hydrophones

PIT antenna

PIT antenna

0 5 m

Ramp RampFLOW

c

b

a

Estuaries and Coasts

avoidance and delay behaviors should also be quantified andtheir consequences assessed (Castro-Santos et al. 2009; Castro-Santos and Haro 2009). Here, we summarize a series of exper-iments that were designed to characterize mechanical injury,avoidance behaviors, and delay of migratory fish passing ahydrokinetic device in a large-scale, semi-controlled laboratorysetting.

Methods

The Flume and Turbine

The experiments described in this study were performed inthe flume facility at the S.O. Conte Anadromous Fish Re-search Center (Conte Lab), located on the Connecticut Riverin Turners Falls, MA, USA (Fig. 1). This is a flow-throughfacility, capable of passing up to10 m3 s−1 of river waterthrough the flumes. The flow was diverted from an adjacent

power canal and returned to the river downstream of theassociated hydropower dam.

We tested live, migratory Atlantic salmon (Salmo salar)smolts and adult American shad (Alosa sapidissima) passingthrough one of these flumes outfitted with a functional hy-drokinetic turbine (EnCurrent model ENC-005-F4; New En-ergy Corp., Inc., Calgary, AB, Canada; Fig. 1c.). This was avertical axis-type turbine capable of producing 5 kW ofpower in flow velocities of 3 m s−1. The turbine measured1.52 m diameter with rotor height of 0.76 m. Given that theflumes at the Conte Lab measure 3.05 m wide and that flowdepths of 1.21 m were required to efficiently drive theturbine, actual flow velocities averaged only about2.25 m s−1, producing a power equivalent of approximately3 kW (Table 1) at a total discharge of 8.2 m3 s−1. Turbineblades rotated at about 60 rpm. This is a realistic conditionfor many locations where these units are designed to bedeployed, and so, was deemed acceptable for biologicaltesting.

Because we were interested in volitional behavior, it wasnecessary to create velocity zones both upstream and down-stream of the turbine test area that were low enough to allowfish to voluntarily approach and pass it. This was accom-plished by raising the floor of the flume by 60 cm for adistance of 10 m upstream and 10 m downstream of theturbine. The greater depth upstream and downstream of thisraised floor caused velocities to be reduced in those sectionsby approximately 40 % (Fig. 1a; Table 1). A larger area wasalso provided downstream to serve as a recovery area forAtlantic salmon smolts and as a resting and staging area forthe upstream migrant American shad. This area consisted ofa large screened corral measuring 6.1 m wide by 6.1 m longadjacent to the test flume (Fig. 1). Flow was dischargedthrough a set of screens and gates 10.3 m wide by 2.1 m tall.The screen immediately downstream of the turbine testing

Table 1 Trial test conditions and sample sizes for Atlantic salmon smolts and American shad exposed to treatment (turbine in) and control (turbine out)conditions

Species Turbine Number Fork length Date range Temperature(°C)

Upstream flowvelocity (m s−1)

Flow depth (m)

Upstream Downstream

Salmon smolts In 119 183±13 May 13–May 19 11.1–14.5 1.89±0.13 1.44±0.01 1.29±0.02

Out 56 183±12 May 13–May 18 10.8–14.4 2.38±0.07 1.21±0.02 1.27±0.02

Adult shad In 134 411±39 May 26–June 09 20.6–24.5 1.89±0.02 1.44±0.02 1.29±0.02

Out 74 410±40 May 26–June 09 20.0–23.9 2.38±0.17 1.21±0.01 1.27±0.01

Dates and temperatures are presented as ranges; and fork length, velocity, and depth are presented as trial means and SDs among trials. Flow velocity istaken 2.45 m upstream of the turbine and corresponds to the “Upstream” panels of Fig. 3. Flow depth measurements were taken 2.45 m upstream and2.45 m downstream of the turbine hub. Discharge was held nearly constant for both species at 8.30 m3 s−1 (turbine in) and 8.78 m3 s−1 (turbine out)

Fig. 1 a Test flume facility at the Conte Lab in plan view (upper panel)and elevation view (lower panel). Note the placement of the turbine, aswell as the release and staging locations: shad staging area was also therecovery area for smolts. Lower panel shows the elevation view of theraised floor and inlet and outlet structures. b Detail of the test area withlocations of turbine, cameras, hydrophones, and PIT antennas (heavylines). Hydrophones were placed on walls at alternating heights of30 cm (open circles) and 80 cm (closed circles) above the floor—thiscreated the optimal conditions for two-dimensional positioning. Forsmolt tests, no hydrophones were placed at the downstream location;instead, the uppermost hydrophones were moved to the downstreamlocation for shad tests. c EnCurrent model ENC-005-F4, vertical axishydrokinetic turbine (elevation view).Heavy lines indicate the floor andwalls of the test flume. The turbine blades were 76 cm tall by 152 cmdiameter, and the device was mounted 15 cm above the floor. The waterlevel shown was upstream of the turbine while running (Table 1); thewater level at the same location was 1.21 m with the turbine out

R

Estuaries and Coasts

flume was built on a curve with a 3-m radius; this created asweeping cross-flow that kept fish from being impingedthere after passing through the turbine. Screens wereconstructed of galvanized steel, with 1.0 cm clear openingto allow for maximum flow while minimizing the risk ofescapement or impingement.

The flume was illuminated with six 400-Wmercury vaporlamps placed 2.5 m above the water surface. These wereconfigured in such a way as to provide uniform lightingaround the turbine and to avoid strong shadows from theturbine and associated mounting hardware. Light intensity atthe water surface was 5.7–7.9 klx, or about the intensity of anovercast day. The intent was to avoid startling the fish whileproviding sufficient illumination for the video monitoringsystem (see the succeeding sections).

Experiments were performed using a treatment (turbinein) and control (turbine out) design. For the treatmentcondition, the turbine was mounted with the lower portionof its blades 15 cm above the floor and the upper portion126 cm below the water surface (Figs. 1c, 2, and 3). Notethat the total swept area of the turbine was 1.15 m2, or24.3 % of the flume cross-section (4.76 m2). Under thecontrol condition, discharge was held approximately con-stant, but the turbine was removed (Fig. 2c; Table 1). Stag-ing and recovery area conditions were similar under bothconditions: velocity in the upstream staging area was 9 %lower under treatment conditions relative to controls anddepth in the downstream area was identical for both condi-tions. This means that any treatment effects could be attrib-uted to the turbine itself and its immediate effects on flumehydraulics. Telemetry and monitoring systems remained inplace under both treatment and control conditions, allowingfor direct comparison of movement patterns with the tur-bine present and absent.

Instrumentation

Because of the novel nature of this study, we used severalmethods to monitor the passage of fish past the turbine (testconditions) or the unimpeded flume (control condition), know-ing that it was likely that not all monitoring methods would beeffective. Passive integrated transponder (PIT) telemetry wasused to monitor gross movements up or down the flume, videocameras monitored passage by the turbine itself, and an inte-grated hydrophone array and acoustic tracking system (ModelHTI-290; Hydroacoustic Technology Inc., Seattle, WA;hereafter termed the acoustic telemetry system) monitoredmovements in two-dimensional (horizontal) space with amean temporal resolution of 220 ms and mean spatialresolution of ±5 cm.

A single PIT antenna was used to monitor downstreammovements of Atlantic salmon smolts; this was primarily toreference passage times to allow for the identification of

smolts as they passed the video cameras (see the succeedingparagraphs). For upstream migrants (shad), a total of fourantennas were used, allowing for quantification of the distanceof ascent and delays as shad approached the turbine location.

Video cameras were deployed below the false floor, angledupward through clear acrylic panels to provide a ventral per-spective of fish as they passed the turbine. The field of viewwas often obscured by bubbles trapped below the acrylic,however, and cameras were later moved above the floor toprovide a lateral perspective of the fish.

The acoustic telemetry system integrated input from anarray of hydrophones that recorded the difference in arrivaltimes of acoustic transmissions from each tag as they passedthrough the array. This information was used to triangulate atwo-dimensional position for the tag at each transmissiontime. Eight hydrophones provided input to the acoustic te-lemetry system (Fig. 1b). These were positioned upstreamand downstream of the turbine to provide optimum two-

Fig. 2 Test flume with turbine in (a, b) and out (c). Visible in a are theturbine, lights, PIT antennas, cameras, and hydrophones (see the sche-matic in Fig. 1). Note that the turbine created some head differential,which affected the flow velocities in those zones (Table 1; Fig. 3)

Estuaries and Coasts

dimensional coverage of fish as they approached and passedthe test area. Hydrophones were placed at alternating heightsof 30 and 80 cm above the false floor. For smolt tests, fourhydrophones were placed upstream of the turbine: two in linewith the turbine and two downstream of the turbine. Thisprovided optimal coverage of the upstream end as smoltsapproached the turbine and were situated to maximize ourability to detect behavioral responses to the turbine beforepassing it. For shad, the two most upstream hydrophoneswere moved to the downstream end of the array, providingbetter coverage of the shad as they approached the turbinefrom the downstream direction.

Flow velocities were also monitored continuously through-out each run using acoustic Doppler current profilers (ADCP;Sontek Argonaut, model SL3000; Sontek/YSI, San Diego,CA, USA) deployed 2.45m upstream and 2.45m downstreamof the turbine location. Velocities were measured in 10 dis-crete cells, eachmeasuring 0.28m long. Cells were distributedlaterally and uniformly across the flume channel, and veloci-ties were recorded every 60 s. Representative velocity condi-tions were also recorded at several locations along the flume,creating full, two-dimensional profiles of flow velocity towhich test animals were subjected.

For each of these systems, PIT, video, acoustic telemetry,and ADCP, clocks on the associated instruments were syn-chronized to the nearest second at the beginning of each trial.This allowed for later comparisons and verification among thevarious types of data.

Study Animals

Because hydrokinetic devices such as the EnCurrent modelENC-005-F4 are meant to be deployed in locations withanadromous migrant fishes, we wanted to explore the effectson both the upstream migrant (adult) and downstream mi-grant (juvenile or smolt) phases. Some of the proposed sitinglocations, like the Yukon and Mackenzie rivers, have

important populations of migratory salmonids, so our firstchoice was to select a salmonid species. Because our labo-ratory discharges directly to the Connecticut River, however,we are unable to test non-native fish that might escape andcolonize the river or transmit disease. Atlantic salmon areavailable in this system, but because this is a populationunder restoration, only hatchery-reared juveniles were avail-able for testing. For this reason, we used Atlantic salmonsmolts as our representative species for the juvenile lifestage. The Connecticut River also has a large native popula-tion of anadromous American shad. Adults of this species arelarge, averaging 435 mm in length, or about the adult size ofmany large salmonids. Shad are also susceptible to handling,which makes them a good indicator species—any injury thatwould harm an adult salmonid would almost certainly have agreater effect on American shad. Furthermore, shad areknown as a “nervous” fish, one that is easily deterred frompassing obstacles or conditions that might be perceived asunnatural (Larinier and Travade 2002). This is also a usefulcharacteristic because it means that behavioral effects of theturbine would likely be easier to observe in shad than insome other species. Thus, shad were chosen as a surrogatespecies for adult salmonids and other anadromous fish, pro-viding conservative estimates of both injury and behavioraleffects of the turbine.

Atlantic Salmon Smolts

Atlantic salmonwere obtained from the Dwight D. EisenhowerNational Fish Hatchery in Pittsburgh, VT, USA andtransported by truck to the Conte Lab. The combined distribu-tions of date and length indicate that all or nearly all individualshad transformed to smolts at the time of transport and testing(McCormick et al. 1998). Upon arrival, smolts were immedi-ately transferred to 2-m-diameter round tanks, where they wereheld and fed to satiation twice daily. Two days after arrival,feeding was withheld, and all smolts were anesthetized andtagged with 23-mm half-duplex PIT tags (TIRIS model RI-TRP-WRHP; Texas Instruments, Dallas, TX, USA; Castro-Santos et al. 1996). At this time, smolts were divided equallyinto two new 2-m-diameter tanks and allowed to recover. Twodays after tagging, smolts from one of these tanks were trans-ferred to a 23-m-long open-channel swim chamber (Haro et al.2004; Castro-Santos 2005). This chamber, originally designedfor studying sprinting performance, was outfitted with a low-velocity staging area downstream. Flows were regulated suchthat the flume maintained a depth of 50 cm and a mean flowvelocity of 0.5 m s−1. These conditions were provided to givethe smolts opportunity to exercise and swim in an open-channel environment, and so hopefully, be better able to swimat speeds representative of wild fish when exposed to theturbine test arena. Smolts were held in this chamber for theduration of the study (7–19 days, mean=12 days). Throughout

Fig. 3 Tag attachment methods for smolts (upper image) and adultshad (lower image)

Estuaries and Coasts

this holding period, smolts were fed twice daily and monitoredfor mortality. Only healthy individuals were used for testing.

Trials were performed between 13May and 19May 2010.The number of smolts in each trial was limited (N=12–26,mean=20) to limit total handling and post-exposure holdingtime. One treatment and one control condition were tested oneach test day. On the day of a test, smolts were randomlyselected from the exercise channel and transferred to theupstream end of the flume facility. Once test flows wereestablished, the fish were individually tagged with acoustictelemetry tags, which had been outfitted with steel loops andsuture threads for this purpose (Fig. 4). Tags were set totransmit at a very rapid rate (four to five transmissions persecond). This transmission rate limited tag life, and in orderto maximize the sample, size tags were activated andattached just before beginning each test. This timing alsomeant that anesthesia could not be used when taggingsmolts as it would likely have affected their ability torespond to the turbine. Instead, smolts were restrainedwithout anesthesia and tagged by passing the suture thread

through the skin just behind the dorsal fin and tying it offto the loop. This technique prevented the suture threadfrom cinching down on the skin and possibly ripping it—in this way, we simultaneously avoided injuring the fishand reduced the risk of losing the tags. Also, at this time,each fish was inspected visually for any signs of externalinjury (cuts, bruises, fin damage, etc.). This informationwas recorded and used for comparison with post-run con-dition assessments (see the succeeding paragraphs). Totalhandling time was typically <30 s.

After tagging, smolts were transferred to a recovery tankwhere they were held for 1–5 min before being released intothe test flume. Once they had recovered (as evidenced byupright swimming and active response to researchers),smolts were transferred one at a time to the test flume bybucket and released in a slack-water zone about 20 m up-stream of the turbine (Fig. 1a). Structures were placed in thiszone on the floor and walls to create flow refugia in whichsmolts were able to hide before volitionally entering theflume. In this way, we hoped smolts would approach the

Hei

ght a

bove

floo

r (m

)

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Distance from left wall, looking downstream (m)0.5 1.0 1.5 2.0 2.5

0.2

0.4

0.6

0.8

1.0

1.2

1.4

0.5 1.0 1.5 2.0 2.5

1.52.02.53.03.5

Upstream Downstream

Turbine

outT

urbinein

Flow velocity

Fig. 4 Time-averaged flow velocity contours looking downstream, taken 2.45 m upstream and 2.45 m downstream of the turbine hub.Upper panelsshow conditions with the turbine removed, and lower panels show conditions with the turbine running. Note the low-velocity zones upstream andimmediately downstream of the turbine and the high-velocity zones near the walls. This represents the wake shed by the turbine while running

Estuaries and Coasts

turbine under their own control and in a way that was as closeto the natural environment as could be achieved in ourlaboratory.

Once a trial was complete, and all the allotted smolts hadbeen released and passed the turbine or control condition,flow through the flume was reduced and smolts were col-lected with dip nets and transferred by bucket to 1 m holdingtanks for recovery. There they were fed ad libitum andmonitored several times daily for 48–96 h. Time of deathwas recorded to the nearest hour. Survivors were eithereuthanized or, when possible, released into the ConnecticutRiver to supplement ongoing restoration efforts there. Beforeeuthanasia or release, all smolts were visually inspected andany signs of injury were recorded.

Adult American Shad

Adult, actively migrating American shad were collectedfrom a fish lift at Holyoke, MA, USA and transferred bytruck to holding facilities at the Conte Lab. Five collectionswere performed between 25 May and 9 June 2010, with 41–68 shad per collection. The truck was outfitted with a 4.2-m3

round tank specifically designed for transporting shad, with arecirculating pump and supplementary oxygen provided at arate of 10 L min−1. Water for transport was treated with asimulated seawater solution diluted to 7.2 ppt. This solutionis commonly used for shad transportation and helps reducestress and disease associated with transport.

Upon arrival at the Conte Lab, shad were PIT-taggedintraperitoneally and deposited in groups of approximately20 into large flow-through holding tanks adjacent to andhydraulically connected with the flume facility (Burrowsand Chenoweth 1970). The following day, a subset of eachcollection was fitted with acoustic transmitters. The attach-ment differed from that used for smolts. In this case, theacoustic tags were fitted with #6 Aberdeen style gold-platedfishhooks coated with epoxy. This method allows for rapidtagging and detagging, so that tags could be used repeatedlyon successive experiments (Castro-Santos et al. 1996). Oncethis subset was tagged, all shad from a given holding pondwere “corralled” (seined, but without contacting the seine)into the staging area downstream of the test flume. A screensituated at the downstream end of the flume kept shad fromentering while flow levels were raised to the test condition.Once test conditions were established, the screen was raisedand all shad were allowed to enter and ascend the flumevolitionally. Throughout each trial, shad had free access tothe flume and the staging area. Often shad would ascend theflume, fall back downstream, and then hold in the stagingarea. In other cases, shad remained in the upstream end forthe duration of the trial. At the end of each trial, flows werereduced and shad were returned to the staging area andseined back into the holding ponds, where they were held

for 67–96 h and monitored several times a day for mortal-ities. For each mortality, PIT ID and time were recorded andthe animal was assessed for injuries. Survivors were likewiseinspected before release.

Analysis

Post-Trial Condition and Survival

Survival rates for both salmon and shad were compared usingKaplan–Meier survivorship curves and statistical comparisonsusing Wilcoxon and logrank tests (Allison 1995; Hosmer andLemeshow 1999; Kaplan and Meier 1958). These are well-established, nonparametric methods for comparing survivalrates for treatment and control animals and are superior tologistic and other forms of binomial comparison of twogroups because they explicitly include a time componentand allow for testing of differences in mortality over time.These methods are also robust against unequal time intervalsfor monitoring such as happens, for example, when laboratorypersonnel were absent overnight. Thus, the multiple observa-tions per day acted to improve the resolution of the tests, andthe comparatively longer gaps that typically occurred at nighthad little effect on the sensitivity of the tests. This techniquealso allowed us to include data from animals that were held forlonger post-trial periods. Furthermore, the two tests appliedhave different sensitivities, with the Wilcoxon test beingmore sensitive to differences in survival early in the timeseries (left side of the distribution) and logrank tests beingmore sensitive to the later part of the time series (right sideof the distribution).

Movement Behaviors

Video was recorded continuously throughout each trial byfour cameras interfaced with a multiple input digital videorecorder (Tyco model TVR-08025; Tyco Video, Boca Raton,FL, USA). Passage events were identified using PIT records(recorded separately), and video was reviewed for severalseconds before and after each recorded event. If a fish wasidentified, its position was documented, along with any ob-servations of strike, avoidance behavior, passage route, etc.

PIT data were compiled in a database containing the ID,location, and time to the nearest 0.01 s. Each smolt has only asingle exposure to the test flume. The data for shad were morecomplex. Here, it was possible to identify individual ascentattempts, and inmany cases, more than one attempt was made.Not all shad staged attempts, however. For each condition,proportion attempting was recorded and compared using lo-gistic regression between treatment and control conditions.Distributions of the number of attempts staged were comparedusing a Kolmogorov–Smirnov test. Because each antenna had

Estuaries and Coasts

a known location, it was also possible to use the PIT array toestimate the maximum distance of ascent (Haro et al. 2004).

Acoustic telemetry data were summarized as locationinformation on a horizontal plane, with position resolved tothe nearest centimeter every 220 ms (the mean transmissioninterval). Spatial resolution of the acoustic telemetry systemwas ±5 cm.

Results

A total of 175 salmon smolts and 208 adult American shadwere introduced to the flume structure (Table 1). Each indi-vidual was used only once. For both species, more individualswere subjected to the treatment condition (turbine in) than thecontrol (turbine out). This allowed us to improve our estimatesof survivorship for those individuals that were exposed to theturbine, while still including enough data from control fish forperforming statistical comparisons between treatment andcontrol groups.

In the case of the salmon smolts, flow velocities exceedthe swimming ability of the fish, so all individuals ultimatelypassed downstream. Typically, smolts passed the turbinewithin about 30 s of release, although a few individuals wereable to hold position upstream for as long as 90 s.

Flume conditions varied by condition, and only slightlyby trial (Table 1; Fig. 3). With the turbine in place andrunning, water was held back, creating a head drop acrossthe turbine. A zone of high-velocity flow occurred along thewalls downstream of the turbine, and a zone of low-velocityflow occurred immediately downstream of the turbine. Flowdownstream of the turbine was also visibly quite turbulent.Flow velocities in the upstream and downstream stagingareas were not measured. As mentioned previously, hydrau-lic conditions in the upstream staging area were sufficientlyenergetic that all smolts moved downstream shortly afterrelease. The downstream area was much more tranquil, how-ever, and smolts and shad could be easily observed duringthe trials, resting and holding station without any indicationof stress or fatigue. Moreover, no fish of either species wereimpinged on the discharge screens under either treatment orcontrol conditions, providing further evidence that the down-stream staging area provided suitable resting habitat.

Instrumentation

Performance of instrumentation differed between the twospecies. For the salmon smolts, only a single PIT antennawas in place. Of the 175 salmon smolts, only 89 weredetected by the PIT system (34 % with the turbine in; 86 %with the turbine out). This difference in detection efficiencymight have been caused by radio interference emitted by theturbine or by different orientation of the smolts as they

passed the antenna, which was located just upstream of theturbine. Detection efficiency of the PIT system was greaterfor shad, largely because there were four antennas instead ofjust one. Combined antenna efficiency of all antennas forboth conditions was 93 %. Antenna 3, which was located justupstream of the turbine, performed less well than the otherantennas with the turbine in (65 %) than with the turbine out(97 %). When antenna 3 was not included, the remainingantennas had an overall detection efficiency of 99 %. Thereduced efficiency on antenna 3 with the turbine in mighthave arisen from interference or from reduced exposure time.

Turbid conditions and bubbles obstructed much of thevideo; however, we were able to characterize spatial distri-bution of 33 smolts (29 with the turbine in and 4 with theturbine out; Fig. 5) and 14 shad (7 with the turbine in and 7with the turbine out) passing the turbine zone. Fortunately,however, the acoustic telemetry system provided excellentdata for many of the salmon smolts (N=85, or 49 % of thetotal). Only a subset of all the introduced shad carried acous-tic tags, and because entry was volitional, data were obtainedfrom a smaller sample (N=23).

Movement Behaviors and Survival

With the turbine running, 43 (72 %) smolts passed through,above, or beneath the swept area of the blades and 17 (28 %)passed around the outside of the blades. This is significantlygreater than a 50:50 ratio (binomial distribution, P<0.0001),despite the fact that the swept area of the blades only occupied50% of the flume width. This raises the possibility that smoltswere actively entrained or attracted to the turbine. However,an alternate explanation exists, which is that the smolts were

Distance from left wall (cm)

0 - 76 76-152 152-229 229-305Dis

tanc

e fr

om fl

oor

(cm

)

0-39

39-78

78-117

117-156

Fig. 5 Distribution of the passage locations for Atlantic salmon smoltspassing downstream through turbine (viewed in the downstream direc-tion, gray dashed line represents the swept area). Bubble size scaleswith number, with the largest bubble indicating eight individuals andthe smallest indicating single observations. White diamonds representobservations with the turbine removed, and black circles representobservations with the turbine in place. The turbine spun in a counter-clockwise direction, viewed from above, i.e., the right side of the panelwas associated with the downstream sweep of the turbine blades

Estuaries and Coasts

simply avoiding the walls, or perhaps being drawn to thecenter of the flume by the greater velocities present there(Fig. 3). This can be assessed by comparing treatment andcontrol conditions, and under the control condition, we alsoobserved a tendency of smolts to gravitate toward the center ofthe flume, with 15 (60 %) individuals passing in the turbinezone (with the turbine removed) and only 10 (40 %) passingoutside of that zone. This difference was not significantlydifferent from 50 % (P=0.167). Taken together, however,these data suggest that the turbine did not significantly affectthe likelihood of passing through the swept area (logisticregression, P=0.486). Thus, it is likely that the tendency topass down the center of the flume reflects either volitional orpassive avoidance of the walls and preference for the center ofthe flume. Despite the high incidence of turbine passage, weobserved no injuries to individual smolts following trials.Also, overall survival was high, with 48-h survival of98.3 % for treatment smolts and 96.4 % for controls (Fig. 6).This difference was nonsignificant (logrank, P=0.41;Wilcoxon, P=0.29). It is important to recognize, however,that this study was designed to identify strong effects. Giventhe observedmortality among controls, the power provided bythese sample sizes to detect 5 or 10 % increases in mortality ata 0.05 significance level was 0.225 and 0.517, respectively,meaning that negative results should be interpreted with cau-tion. The unbalanced design yielded some benefit, however, inthat it improved the precision of the total mortality estimate forsmolts exposed to the turbine. From the binomial distribution,

the 98.3 % survival estimate has a 95 % confidence intervalfrom 95.4 to 99.7 %.

The low mortality rate may be attributable in part to theroute through which most smolts passed the turbine. Videoanalysis suggests that smolts were disproportionately in-clined to pass over the top and around the downstream-sweeping side of the turbine blade when the turbine waspresent (Fig. 5; chi-square, P=0.05). Body orientation ofsmolts to the current was variable and about equally distrib-uted among upstream-oriented, downstream-oriented, orsideways as they passed the cameras. Also, the observedspatial distribution was less symmetrical than the velocityprofile (Figs. 3 and 5)—it is possible, therefore, that thispattern was affected by flow, but given the distribution, it islikely that volitional response to the turbine had some effecton passage route.

Assessment of the speed at which smolts moved relativeto the flow suggests that there was some ability to orient toand resist the current (Fig. 7). There was noticeable hesita-tion in the upstream zone for both treatment and control fish.This probably represents a response to the elevated floor andassociated flow acceleration. In the presence of the turbine,smolts were slightly slower in the approach zone than whenit was removed and slightly faster downstream as they exitedthe flume. These differences were nonsignificant, but areevocative of slight hesitation upstream of the turbine by

Days after Trial0 1 2 3 4

% S

urvi

val

0

20

40

60

80

100

0

20

40

60

80

100Atlantic salmon smolts

Adult American shad

Pr > Chi Square

Log-Rank: P = 0.407Wilcoxon: P = 0.293

Pr > Chi Square

Log-Rank: P = 0.383Wilcoxon: P = 0.255

Fig. 6 Survivorship curves for Atlantic salmon smolts and Americanshad exposed to turbine (solid) and control (dashed) conditions. Circlesindicate censoring, when survivors were either sacrificed or returned tothe river to continue their migration

Gro

unds

peed

(m

s-1

)

0

1

2

3

Upstream

Approach

Departure

Downstream

Turbine

Fig. 7 Ground speed of Atlantic salmon smolts as they passed down-stream through the flume under treatment (red) and control (blue)conditions. This figure partitions the flume into “Upstream” (>4 mupstream of the turbine hub), “Approach” (1–4 m upstream of theturbine hub), “Turbine” (1 m upstream to 1 m downstream of thehub), “Departure” (1–4 m downstream of the turbine hub), and “Down-stream” (>4 m downstream of the turbine hub). Columns are the meansand error bars are the SDs of ground speed. This entire range is abovethe horizontal portion of the elevated floor, and so, the reduced groundspeed upstream suggests either some response to the floor or pre-fatigueefforts to hold station in the rapid flow. The dashed lines show the flowmean velocity under each condition. Differences between groundspeeds for treatment and control conditions were nonsignificant for allzones, owing at least in part to the strong variability in ground speed.Mean ground speeds were consistently less than flow velocity, however,indicating that most smolts were resisting the current, backing down-stream as they passed the turbine zone

Estuaries and Coasts

some individuals, and perhaps escape behavior followingpassage. Several smolts exhibited rapid lateral movementsimmediately after passing the turbine. These suggested thatsome smolts experienced disorientation immediately follow-ing passage, which accounts for the slight drop in groundspeed in the departure zone (Fig. 7). It is important torecognize that, given the variability in behavior, it is possiblethat the observed patterns may have arisen by chance.

The behavior of American shad was very different fromsalmon smolts. Both PIT and acoustic telemetry systemsprovided detailed information on movements, with consis-tent results. By quantifying lags between detections, it waspossible to discriminate among attempts (Castro-Santos andPerry 2012). Shad staged more attempts when the turbinewas removed (mean±standard deviation [SD] number ofattempts=1.13±2.1 with turbine in and 1.80±2.8 with tur-bine out; Wilcoxon, P=0.125) and spent more time in theflume with the turbine out (median=56.7 s) than the turbinein (median=30.2 s; Wilcoxon, P<0.001). Although the dif-ference in attempt number might have arisen by chance,several shad staged large numbers of attempts with theturbine out. This created strong inequality of variance(P>F=0.002). Taken together, these results indicate thatshad were more willing to enter the flume and to expendgreater effort attempting to pass the high-velocity zone in theabsence of the turbine than with it present.

For those shad that did enter the flume, there was a cleareffect of the turbine on the distance of ascent, with more shadpassing the turbine location with the turbine removed thanwhen it was in and running (Fig. 8; Wilcoxon, P=0.004;logrank, P=0.734). Shad were more likely to arrive at an-tenna 4 once they passed the turbine. This may reflect avoid-ance of the turbine or improved swimming ability in therelatively lower flow velocities present upstream of the tur-bine when it was running. Reduced efficiency of antenna 3

with the turbine running was not sufficient to explain themagnitude of this difference, and the data indicate that thepattern was driven by behavior, not antenna efficiency. Withthe turbine running, shad were more likely to occupy thezone near the downrunning blades (P=0.02), although therewas broad overlap in distribution between the twoconditions.

As with the salmon data, posttest assessment of shadcondition yielded no evidence of strike injuries, and survivalof treatment and control groups was comparable (Fig. 6;Wilcoxon, P=0.126; logrank, P=0.413). Both treatmentand control groups suffered greater mortality than did thesalmon, especially after 2 days of post-trial observation (notethat shad are sensitive to confinement, and these animals hadbeen held for a total of >3 days). The observed mortalityrates were consistent with those observed among shad heldin these same facilities without any handling after beingtransported (Sullivan 2004). Interestingly, posttest mortalitywas actually greater among shad that staged few or noattempts (Cox’s proportional hazard regression, P<0.001).This suggests (a) that entry into the flume and exposure tothe turbine had no detrimental effect and (b) that the ob-served mortality reflected variability in condition of the fish,rather than the effect of being subjected to the turbine treat-ment. Given the observed mortality among controls, thepower to detect a 10 % increase in mortality after 48 h was0.59 with this sample size, but after 96 h, decreased to 0.29.Therefore, despite the apparent lack of treatment effect,negative results must be viewed with caution.

Discussion

The most striking result of this study is the apparent lack ofany injury or mortality incurred as a result of passing throughthe turbine zone for either species. Even conservative esti-mates of turbine-induced mortality indicate values <5 %.This is comparable with the expected survival through themost fish-friendly turbine designs currently in use, such assome Kaplan turbines, and is also comparable to experimen-tal units under development with the specific objective ofreducing harm to fish (Bell and Kynard 1985; Odeh 1999;Stier and Kynard 1986).

In order to definitively show lower mortality rates, studieswith much larger sample sizes must be conducted. The powerof tests on turbine mortality studies is important because itinforms us of the scale of likely effects. For example, asensitivity of 5 % may be sufficient if only a small numberof turbines were to be deployed on a large river system with astrong diversity of spawning habitat (and, hence, several dis-crete stocks of philopatric species like the salmonids). As aunit in a larger array of, say, thousands of similar devices (oralternatively for resident species that may have many

Fig. 8 Maximum distance of ascent as measured by PIT antennanumber (see Fig. 1b) with the turbine in (red) and out (blue). Data arepresented as percent failing at each antenna; associated numbers areindicated above each bar. Antennas are numbered moving from down-stream to upstream, i.e., flow moves from left to right, and turbine islocated between antennas 2 and 3

Estuaries and Coasts

exposures over a lifetime), this level of certainty could rapidlybecome unacceptable. On such a scale of exposure, samplesize for survival studies like this one would have to increasedramatically. This is particularly true for species like Ameri-can shad, which are sensitive to handling and holding. As themortality rate of the controls increases, the relative sample sizeneeded to detect effects also increases (Perry et al. 2012;Skalski 1998; Skalski et al. 2001). To detect a 5 % increaseinmortality relative to, say, 5%mortality among controls witha power of 0.8 using logistic regression, a sample size of 926would be required (Hosmer and Lemeshow 1989). The sur-vival analysis methods employed in this study (Hosmer andLemeshow 1999) are more powerful, however, and to achievea similar power over 4 days of follow-up would require asample size of only 436. This need for large sample sizes andcontrolled follow-up is one great advantage of laboratorystudies over field studies—handling effects and losses tofollow-up can be minimized. This means that laboratory stud-ies, particularly when designed to take advantage of the supe-rior power of survival analysis methods, can be far moreefficient at detecting survival effects than field studies thattreat mortality as a binomial response. A counterpoint to this isthat the smolts used in this study were of hatchery origin andthe flume environment is highly artificial. The turbine occu-pied a much larger proportion of the flume than would beexpected in a field situation. Also, actual behaviors of wildsmolts in a free-flowing river may differ from what was testedhere. Because of this, any conclusions drawn from this andother laboratory work should be viewed as preliminary andsubject to verification in the field.

Similar conclusions can be applied to the adult Americanshad. In this case, the fish were wild, and their behaviors maybe more representative of what one would expect in the field.Here again, though, the flume environment is highly artifi-cial and movements were constrained. The observed reluc-tance to pass the turbine may be less of an issue if it were tobe deployed in a larger river system, with more space above,below, and around the turbine through which fish could passunimpeded.

Field studies have their own problems, however. To date,we know of only one survival study performed on an in-riverhydrokinetic turbine (Normandeau Associates Inc. 2009).This study evaluated the survival of five fish species throughan axial flow turbine, but likewise found no significant inci-dence of injury-associated or turbine-associated mortality.Their studies were limited to survival, however, with noinformation on behavioral effects. Also, the study conductedby Normandeau Associates Inc. (2009) used balloon tags torecover fish. This method is useful because it eliminates theneed to use fixed netting to recover test and control fish. Themethod itself can affect behavior, however, and studiesperformed at conventional hydroelectric facilities have shownthat they can dramatically underestimate mortality associated

with turbine passage (Ferguson et al. 2006). Nevertheless, thistype of study can provide preliminary information, and theresults of the study conducted by Normandeau Associates Inc.(2009) are largely consistent with what we observed.

Because the Conte Lab is a flow-through facility, waterquality can be variable, and during these trials, video qualitywas poor. At least one other large laboratory exists that hasresolved this problem by using a large-scale closed-circuitflume (Electric Power Research Institute 2011). Comparablestudies performed at this facility using video analysis yieldedsimilar results to ours, with even fewer fish being struck byblades than would be predicted by chance alone. Moreover,because they had excellent water clarity, these researcherswere able to definitively document avoidance behaviors offish. Here again, the results were similar to what we observed,with fish holding station near the spinning turbine blades, butactively avoiding strike, and incurring minimal injury evenwhen strike occurred.

None of the foregoing studies quantified behavioral effects,however. Behavioral barriers are a concern because they createa situation in which fishmay avoid passage or reduce the rate ofpassage (i.e., increase the time required to pass). On the scale ofan individual turbine, such delays may be inconsequential, butat larger scales, withmany turbines deployed throughout a riversystem, cumulative effects could lead to reduced spawningviability, reduced access to habitat, and possibly increased riskof predation, disease transmission, etc. (Bickford and Skalski2000; Castro-Santos and Letcher 2010; Harris and Hightower2012). American shad are notorious for being reluctant to passstructures of many designs (Larinier and Travade 2002;Sprankle 2005), and these results should be viewed as a pre-liminary indication of possible effects on upstream adult mi-grant fishes. As with the salmon smolt data, any conclusionsdrawn from laboratory studies should be viewed as preliminaryand subject to verification in field settings. Furthermore, thelikely effects of deployed turbines in the field will vary as afunction of the number of units deployed and the scale andhydrography of the deployment location.

A final note of caution: these studies were performed ononly two species and were done under strong lighting condi-tions. Other species and life stages might have respondeddifferently to this turbine, and more data on a greater diversityof species would help define the scale of likely effects. Also,many riverine and migratory species are most active at night.Although we saw evidence that fish passing through thisturbine appeared not to suffer injury, it is an open questionas to whether the same would be true under low-light condi-tions. Further work is needed to address this question. Thisstudy has shown the difficulty of using video to monitormovements under turbid conditions that are common in na-ture, but has also shown the ability of advanced telemetrysystems to offset this challenge. Alternative video and acous-tic technology should be applied to see if they produce better

Estuaries and Coasts

imagery; infrared video might hold some promise as well.Regardless, additional study on other species, life stages,lighting, and hydraulic conditions would further advance theconclusions and broaden the relevance of this study.

Acknowledgments This work was supported by funding from the USDepartment of Energy, Canada Department of Fisheries and Oceans(DFO), Alaska Fish and Game and NewEnergy Corp. Special thanks toSteve Walk and John Noreika (USGS Conte Lab); Amy Teffer andMelissa Belcher (University of Massachusetts Amherst); Paul Jacobson(Electric Power Research Institute), Bruce Hannah (DFO); and BobMoll (New Energy Corp.).

Disclaimer Any use of trade, product, or firm names is for descriptivepurposes only and does not imply endorsement by the U.S.Government.

References

Allison, P.D. 1995. Survival analysis using the SAS system: A practicalguide. Cary: SAS Institute.

Bell, C.E., and B. Kynard. 1985. Mortality of adult American shad passingthrough a 17-megawatt Kaplan turbine at a low-head hydroelectricdam. North American Journal of Fisheries Management 5: 33–38.

Bickford, S.A., and J.R. Skalski. 2000. Reanalysis and interpretation of25 years of snake—Columbia River juvenile salmonid survival studies.North American Journal of Fisheries Management 20(200): 53–68.

Burrows, T.E., and H.H. Chenoweth. 1970. The rectangular circulatingrearing pond. The Progressive Fish Culturist 32: 67–80.

Castro-Santos, T. 2005. Optimal swim speeds for traversing velocitybarriers: an analysis of volitional high-speed swimming behaviorof migratory fishes. Journal of Experimental Biology 208:421–432.

Castro-Santos, T., A. Cotel, and P.W. Webb. 2009. Fishway evaluations forbetter bioengineering—An integrative approach. In Challenges fordiadromous fishes in a dynamic global environment, ed. A.J. Haroet al., 557–575. Bethesda: American Fisheries Society, Symposium 69.

Castro-Santos, T., and A. Haro. 2003. Quantifying migratory delay: Anew application of survival analysis methods. Canadian Journalof Fisheries and Aquatic Sciences 60: 986–996.

Castro-Santos, T., and A. Haro. 2009. Fish guidance and passage atbarriers. In Fish locomotion: An etho-ecological approach, ed. P.Domenici and R.W. Blake. Enfield: Science Publishers.

Castro-Santos, T., A. Haro, and S. Walk. 1996. A passive integratedtransponder (PIT) tagging system for monitoring fishways. Fish-eries Research 28: 253–261.

Castro-Santos, T., and B.H. Letcher. 2010. Modeling migratory bioener-getics of Connecticut River American shad (Alosa sapidissima):Implications for the conservation of an iteroparous anadromous fish.Canadian Journal of Fisheries and Aquatic Sciences 67: 806–830.

Castro-Santos, T. and Perry, R.W. 2012. Time-to-event analysis as aframework for quantifying fish passage performance. In TelemetryTechniques. Edited by N.S.Adams, J.W.Beeman, and J.Eiler.American Fisheries Society, Bethesda, MD. pp. 427–452.

Electric Power Research Institute. 2011. Evaluation of fish injury andmortality associated with hydrokinetic turbines. Palo Alto: EPRI.

Ferguson, J.W., R.F. Absolon, T.J. Carlson, and B.P. Sandford. 2006.Evidence of delayed mortality on juvenile Pacific salmon passingthrough turbines at Columbia River dams. Transactions of theAmerican Fisheries Society 135: 139–150.

Haro, A., T. Castro-Santos, J. Noreika, and M. Odeh. 2004. Swimmingperformance of upstream migrant fishes in open-channel flow: Anew approach to predicting passage through velocity barriers. Ca-nadian Journal of Fisheries and Aquatic Sciences 61: 1590–1601.

Harris, J.E., and J.E. Hightower. 2012. Demographic population model forAmerican shad: Will access to additional habitat upstream of damsincrease population sizes?Marine and Coastal Fisheries 4: 262–283.

Hosmer, D.W., and S. Lemeshow. 1989. Applied logistic regression.New York: Wiley.

Hosmer, D.W., and S. Lemeshow. 1999. Applied survival analysis. NewYork: Wiley.

Kaplan, E.L., and P. Meier. 1958. Nonparametric estimation fromincomplete observations. Journal of the American Statistical As-sociation 53: 457–481.

Larinier, M., and F. Travade. 2002. The design of fishways for shad.Bulletin Francais de la Peche et de la Pisciculture 135–146.

McCormick, S.D., L.P. Hansen, T.P. Quinn, and R.L. Saunders. 1998.Movement, migration, and smolting of Atlantic salmon (Salmosalar). Canadian Journal of Fisheries and Aquatic Sciences 55:77–92.

McCormick, S.D., D.T. Lerner, M.Y. Monette, K. Nieves-Puigdoller,J.T. Kelly, and B.T. Bjornsson. 2009. Taking it with you when yougo: How perturbations to the freshwater environment, includingtemperature, dams, and contaminants, affect marine survival ofsalmon. Challenges for Diadromous Fishes in A Dynamic GlobalEnvironment 69: 195–214.

McLaughlin R.L., E.R.B. Smyth, T. Castro-Santos, M.L. Jones, M.A.Koops, T.C. Pratt, and L.A. Vélez-Espino. 2012. Unintendedconsequences and trade-offs for fish passage. Fish and Fisheries.doi:10.1111/faf.12003.

Normandeau Associates, Inc. 2009. An estimateion of survival andinjury of fish passed through the Hydro Green Energy hydrokinet-ic system, and an characterization of fish entrainment potential atthe Mississippi Lock and Dam No. 2 hydroelectric project (P-4306), Hastings, Minnesota—Final report. Drumore: NormandeauAssociates, Inc.

Odeh, M. 1999. A summary of environmentally friendly turbine designconcepts. DOE/ID/13741. Idaho Falls: US Department of Energy,Idaho Operations Office.

Perry, R.W., T. Castro-Santos, C.M. Holbrook, and B.P. Sandford.2012. Using mark–recapture models to estimate survival fromtelemetry data. In Telemetry techniques: A user guide for fisheriesresearch, ed. N.S. Adams, J.W. Beeman, and J. Eiler, 453–476.Bethesda: American Fisheries Society.

Skalski, J.R. 1998. Estimating season-wide survival rates ofoutmigrating salmon smolts in the Snake River, Washington. Ca-nadian Journal of Fisheries and Aquatic Sciences 55: 761–769.

Skalski, J.R., J. Lady, R. Townsend, A.E. Giorgi, J.R. Stevenson,C.M. Peven, and R.D. McDonald. 2001. Estimating in-riversurvival of migrating salmonid smolts using radiotelemetry.Canadian Journal of Fisheries and Aquatic Sciences 58:1987–1997.

Sprankle, K. 2005. Interdam movements and passage attraction ofAmerican shad in the lower Merrimack River main stem.North American Journal of Fisheries Management 25:1456–1466.

Stier, D.J., and B. Kynard. 1986. Use of radio telemetry to determine themortality of Atlantic salmon smolts passed through a 17-MWKaplan turbine at a low-head hydroelectric dam. Transactions ofthe American Fisheries Society 115: 771–775.

Sullivan, T.J. 2004. Evaluation of the Turner’s Falls fishway complexand potential improvements for passing adult American shad. M.S.thesis, University of Massachusetts, Amherst.

Estuaries and Coasts