Mry L. Moser, Mtthew L. Keefer, Stephen C. Corett, Kinsey E. Frik,Christopher C. Cuill, Sen C. Tkley

aNorthwest fisheries Science Center, National Marine fisheries Service, National Oceanic and Atmospheric Administration, Seattle, WA, 98112, USA

bDepartment of fish and Wildlife Sciences, College of Natural Resources, University of Idaho, Moscow, ID, 83844-1136, USA

cOcean Associates, Arlington, VA, 22207, USA

dU.S. Army Corps of Engineers, Environmental Resources Branch, Portland, OR, 97204-3440, USA

ABSTRACT

fishways at hydroelectric dams are unnatural environments that typically present fish with channels of uniform depth, variable water velocity, and lack of cover. fish retention and overall passage success may be improved by providing rest and sheltering areas inside fishways. We examined adult Pacific lamprey Entosphenus tridentatus use of two specially designed fishway refuges at Bonneville Dam on the Columbia River (northwestern USA). These relatively small boxes(1.1 m × 0.4 m × 0.2 m) provided low-velocity, dark refuge from predation for Pacific lamprey, a largely nocturnal species. Refuge boxes were equipped with antennas to detect entry of lamprey tagged with passive integrated transponder (PIT) tags. We PIT tagged and released 3,247 lamprey, including 599 that were doubletagged with a PIT and a radio transmitter, over three study years (2012-2014). In each year, PIT-tagged lamprey detected at nearby fishway exits had used a refuge:12% in 2012, 28% in 2013, and 36% in 2014. Median residence time of PIT-tagged lamprey in the refuges for each year was 20.3 h, 16.6 h, and 13.1 h. Lamprey entrance into refuges peaked at 0300-0500 PDT, and they typically exited at around 2000 PDT, suggesting that refuges primarily functioned as shelter from daylight.Probability of refuge use increased with a morphological indicator of sexual maturity (distance between dorsal fins). In the radio-tagged group, refuge users were equally likely to pass Bonneville Dam as non-users. However, refuge users were less likely than non-users to be detected at sites upstream from the dam, perhaps owing to their maturation status or longer mean passage time through our study area (2.0 vs. 0.6 d for non-users). Refuges show promise for improving fish retention in fishways, particularly for nocturnal and/or small-bodied species that seek shelter from light or predation.

1.Introduction

Man-made obstacles to fish movement (dams, weirs, tide gates,culverts, etc.) result in fragmented habitats and population declines(Lucas & Baras, 2008). To mitigate for these effects, fishways are often constructed to offer fish a passage route around the obstacle (Clay,1995). These fishways are typically constructed of concrete or metal and are designed to accommodate the swimming performance of target species. fishway engineers also must factor in construction costs,maintenance requirements and space available (Clay, 1995).

The resulting fishways are often unnatural environments that present fish with both behavioral and physical challenges. fish can be forced into proximity with artificially high densities of conspecifics and other species, including potential predators (Agostinho, Agostinho, Pelicice, &Marques, 2012). In addition, fishways often feature high velocities and turbulence, are subject to man-made noise/vibration, can present fish with thermal barriers, and are often uniform in depth. Shallow fishways have unnatural lighting conditions, with high penetration of both natural light in daytime and artificial light at night. Many fish species are sensitive to these conditions, and after fishway entry, may avoid them by retreating downstream and abandoning migration (Castro-Santos, Cotel, & Webb, 2009; Haro & Kynard, 1997; Keefer, Caudill,Clabough, et al., 2013).

Early in the history of fishway design, the need to provide fish with refuge from high water velocity was recognized (Clay, 1995). Technical solutions included incorporation of baffles, resting areas, or turn pools into many fishway designs. More recently, efforts to provide velocity heterogeneity and reduce turbulence within fishways have been a design focus (Muraoka, Nakanishi, & Kayaba, 2017; Silva, Katapodis, Ferreira,& Pinheiro, 2012). Nature-like fishway designs further address the need for more familiar and less extreme hydraulic elements to promote passage for a wider variety of species.

While the need for hydraulic refuges is well established, providing cover where fish can shelter during technical fishway passage has not, to our knowledge, been tested. Refuge from light and/or predation in fishways may be particularly important for small-bodied species that are susceptible to predation/mutilation or the mere presence of potential predators. Agostinho et al. (2012) documented the concentration of predators in and around a Brazilian fishway and found that 36% of the prey species collected from the fishway had evidence of piscivore attacks. Providing sheltering areas for prey species in fishways could improve their retention.

Addition of fishway refuges could also promote passage of species that seek darkened, low velocity areas for resting. Artificial lighting in fishways is common and it could cause delays in passage or fishway exit back to the tailrace. Sea lamprey Petromyzon marinus seek refuge from light during daylight hours (Binder & McDonald, 2007). Adult Pacific lamprey Entosphenus tridentatus are also photophobic and assume a nocturnal pattern of activity in complex fishway environments (Keefer,Caudill, Peery, & Moser, 2013). In contrast, visual predators, like salmonids, are more likely to attempt fishway passage in the day when their sensory systems are most effective (Gowans, Armstrong, Priede, &Mckelvey, 2003; Keefer, Caudill, Peery, et al., 2013; Standen, Hinch,Healey & Farrell et al., 2002).

Pacific lamprey is a culturally and ecologically important anadromous species (Wicks-Arshack, Dunkle, Matsaw, & Caudill, 2018) that may benefit from the ability to shelter in dark, low-velocity refuges during fishway transit. Poor pre-spawning lamprey passage at fishways in the Columbia River catchment has contributed to decreased population abundance, resulting in limited harvest opportunities for Native Americans in historically-important upriver fishing areas (Luzier et al.,2011; Wicks-Arshack et al., 2018). For example, studies using radiotelemetry and passive integrated transponder (PIT) tags have revealed that only about one-half of the Pacific lamprey that approach Bonneville Dam (Columbia River km 235) are able to pass over (Keefer, Caudill,Clabough, et al., 2013; Keefer, Moser, Boggs, Daigle, & Peery, 2009;Moser, Ocker, Stuehrenberg, & Bjornn, 2002).

Most adult Pacific lamprey attempt to move upstream through fishways at night (Keefer, Caudill, Peery, et al., 2013). If they do not successfully pass at night, lamprey often move back downstream within the fishway the following day (Keefer, Caudill, Clabough, et al., 2013).These downstream movements often are not followed by re-ascensions of the fishway. In addition, Pacific lamprey are subject to predation by white sturgeon Acipenser transmontanus that hold station for extended periods inside the lower parts of some fishways at Columbia River dams(The Dalles Dam fishways, Parsley et al., 2007; Washington-shore fishway at Bonneville Dam, Kirk, Caudill, Johnson, Keefer, & Clabough et al., 2015). For these reasons, provision of refuges could help to retain lamprey in fishways and promote passage.

The objective of this study was to evaluate two prototype refuges designed and installed for adult Pacific lamprey at a Bonneville Dam fishway. The refuges were equipped with PIT antennas, and detections of PIT-tagged lamprey were used to indicate the number and timing of refuge entries and exits. We used these data and radiotelemetry to test the following hypotheses: 1) lamprey would use fishway refuges, 2)lamprey would use refuges for extended periods (＞1 h) and that refuges would help to retain lamprey in fishways during the day, 3) refuge use would result in higher rates of dam passage, and 4) refuge users would be more likely than non-users to be detected at sites upstream from the dam.

2.Methods

2.1.Refuge design and installation

During 2011, two refuge boxes were installed in the auxiliary water supply channel at Bonneville Dam Powerhouse 2 fishway. Each refuge was a weighted aluminum box (6061 alloy); 40.6 cm wide by 114.3 cm long by 17.8 cm high with 7 cm square openings at both ends (Fig. 1).Both refuges featured natural cobble substrate cemented to the bottom of the upstream end. A single half-duplex PIT antenna was fitted to the perimeter of each box and wired to a detector that logged the time and date of individual PIT tag detections.

The long axis of each box was oriented with the flow (Fig. 1), and we positioned one box along each wall of the auxiliary water supply channel at the Bonneville Dam Powerhouse 2 fishway (Fig. 2). This channel provides water to maintain hydraulic head in the fish ladder and is accessible to lamprey through a picketed lead (Fig. 2) but is not accessible to adult salmonids (Moser, Keefer, Pennington, Ogden, & Simonson, 2011). The channel was 1.5 m deep and 9.1 m wide; hence, the refuge openings oriented perpendicular to the channel bottom (49 cmeach) represented 0.07% of the channel’s cross-sectional area. Lamprey that enter this channel could exit upstream either by using a lamprey-specific fishway (described in Moser et al., 2011) or by returning to the main fish ladder (Fig. 2).

Fig. 1.Oblique views of refuge installed in a Bonneville Dam fishway.

Fig. 2.Inset of the study area at Bonneville Dam and location of refuges (black rectangles) in the auxiliary water supply channel at the Powerhouse 2 fishway. Also shown are the locations of radio antennas (stars) used to monitor movements of double-tagged lamprey. Diagram not to scale.

2.2.Pacific lamprey tagging

Adult Pacific lamprey were collected nightly during their spawning migration using passive traps set in a channel parallel to the main fishway (Fig. 2). After anaesthetizing the lamprey using 60 ppm eugenol, we measured total length (nearest 0.5 cm, all years), weight(nearest 1 g, all years) and distance between dorsal fins (dorsal distance,nearest 0.1 cm, 2014 only). A 4-mm incision was made using a BD Bard-Parker #11 surgical blade just off the ventral midline at a point even with the insertion of the anterior dorsal fin. A disinfected half-duplex PIT tag (3.6 mm ×32.0 mm, Oregon RFID, Portland, OR) was then inserted into the body cavity. fish were allowed to recover for at least 6 h and were released in the evening approximately 3 km downstream from the dam.

In each year, a portion of PIT-tagged lamprey was also implanted with an active radio or acoustic transmitter (i.e., double-tagged) using methods of Johnson et al. (2012). Double-tagged fish received an acoustic transmitter (Advanced Telemetry Systems, Isanti, Minnesota)in 2012 (n =153) and 2013 (n =197) for use in a concurrent study of lamprey behavior in dam tailraces and reservoirs (Noyes, 2013). In 2014, 599 double-tagged fish received a radio transmitter (Model NTC-4-2L, Lotek Wireless, Newmarket, Ontario).

Monitoring was conducted via an extensive array of PIT and radio antennas positioned at key locations throughout the fishway and at dams and tributaries further upstream (details in Keefer et al., 2009;Moser et al., 2011). In addition to PIT detectors integrated into the refuges, there were detectors at the lamprey-specific fishway and main fish ladder exits (Fig. 2) and at other locations within and upstream from the dam. Upstream PIT detection locations included at The Dalles (rkm 308), John Day (rkm 347) and McNary Dams (rkm 467) on the lower Columbia River, at dams in the Snake and upper Columbia rivers, and in some secondary tributaries. An extensive array of aerial and underwater coaxial-cable radio antennas was installed in the dam tailrace, fishways,and at upstream dams to allow high-resolution monitoring of double-tagged lamprey with radio transmitters (details in Keefer, Caudill, Clabough, et al., 2013). Detection data from acoustic monitoring sites in tailraces and reservoirs were not used in this study.

2.3.Data analysis

Pacific lamprey use of refuges was documented using detections at the refuge PIT antennas. PIT-tagged lamprey detected at each refuge were enumerated in each year, along with numbers detected upstream at the main ladder exit and the lamprey fishway exit (Fig. 2). In addition,detections by radio antennas (n =3) in the auxiliary water supply channel in 2014 (Fig. 2) allowed us to identify double-tagged individuals exposed to the refuges. For these double-tagged lamprey, we used PIT detections to distinguish refuge users from non-users.

A combination of radiotelemetry and PIT data was used to test for differences between double-tagged lamprey that did or did not use a refuge in 2014. A two-sample, two-tailed t-test that assumed unequal variances was used to compare the time (hours) double-tagged refuge users and non-users spent moving from the radio receiver at the auxiliary water supply channel entrance to a fishway exit (Fig. 2). A series of logistic regression models was used to assess morphological differences between refuge users and non-users. The five covariates evaluated included lamprey length, weight, girth, and dorsal distance, as well as release date. Fifteen models that included non-interacting combinations of the covariates were compared using Akaike’s information criterion(AIC, Burnham & Anderson, 2002).

Refuge use timing and duration were determined from PIT detections. These data were used to calculate the median time that lamprey were detected at the refuges and diel timing of entries and exits. Refuge residence was defined as the time from first PIT detection at a refuge antenna to the last detection for each individual. For lamprey that entered refuges multiple times, the residence times were added to give a total duration of refuge occupation. Medians of total duration were computed for each year.

PIT detections were also used to assess diel timing of refuge use. For this analysis, only lamprey that were detected at a refuge antenna for at least 1 h were included. The time of day for refuge entries and exits was tabulated for only the first refuge use by an individual. These refuge entries and exits were binned by hour. We used Rao’s spacing test of uniformity to test whether diel timing of refuge entries and exits was uniformly distributed, both within an individual refuge and across refuges (v3.3.4; R Development Core Team 2016).

Whether refuge use affected Bonneville Dam passage was assessed using logistic regression. In this fishway system, dam passage was equivalent to reduced downstream movement to the tailrace. The analysis was limited to fish double-tagged in 2014 that were detected on radio antennas inside the auxiliary water supply channel, which ensured that each lamprey was exposed to the refuges. The

a priori

model set included 16 models, each of which included a binary variable for refuge use (yes, no) plus the same combinations of traits described previously.The dependent variable was dam passage (yes, no).Whether refuge use affected lamprey detection upstream from Bonneville Dam was also evaluated using logistic regression. All tagged Pacific lamprey detected at either the main fishway or lamprey fishway exits were included in this analysis. All

a priori

logistic regression models included refuge use (yes, no) as a covariate, along with combinations of trait data. Because dorsal distance was not measured in 2012 or 2013,preliminary analyses were restricted to individual years, with eight models each in 2012 and 2013 and the aforementioned 16 models in 2014. Since dorsal distance was measured only in 2014, we did the analysis first for that year and found that dorsal distance was not significant. Based on this finding, we fit data from all three years with eight models.

3.Results

3.1.Pacific lamprey use of refuges

Lamprey refuges were operated continuously in summer and fall each year, from 30 May to 17 October 2012, 27 June to 23 October 2013, and 15 May to 30 October 2014. Across years, 1,022 individuals PIT-tagged lamprey were detected in the upper Powerhouse 2 fishway,including 362 (35%) that had been detected in a refuge and 660 (65%)that had not. Of the refuge users, 1%-2% used the same refuge multiple times (1 in 2012 and 3 in 2014). In addition, 6%-13% of refuge users used more than one refuge (5 in 2012, 9 in 2014 and 11 in 2014). In each year, 63%-74% of refuge users were eventually detected exiting a fishway into the forebay (Table 1). The annual proportion of lamprey detected at fishway exits that had used a refuge was 12% (25 of 207) in 2012, 28% (95 of 341) in 2013, and 36% (128 of 359) in 2014.

Table 1Number of PIT- and double-tagged Pacific lamprey released downstream from Bonneville Dam and the numbers and percentages detected in the upper Powerhouse 2 fishway and refuges (study area), at Powerhouse 2 fishway exits,and at dams or in tributaries upstream from Bonneville Dam.

In 2014, 217 double-tagged lamprey were detected in the Powerhouse 2 fishway study area, and 63 (29%) of these were detected inside a refuge. Among the 15 logistic regression models compared, the most parsimonious model included only dorsal distance. Lamprey with a smaller gap were substantially more likely to use a refuge than those with a larger gap (χ=9.70,

P

=0.002, Fig. 3). Four additional models had statistical support with ΔAIC ≤2; each included dorsal distance and either a size metric (length, weight, girth) or release date.

Fig. 3.Estimated probability of refuge box use was negatively associated with the distance (cm) between dorsal fins for double-tagged (PIT +radio) Pacific lamprey in 2014. Envelope indicates 95% confidence interval.

3.2.Refuge use timing and duration

The annual median time that individual lamprey stayed in a refuge ranged from 13.1 to 21.3 h (

n

=366, Fig. 4). In each year 14%-29% of the detection durations were less than 1 h (5 in 2012, 44 in 2013 and 51 in 2014). Maximum residence times for individual fish were 35 d (2012),41 d (2013), and 45 d (2014). As a consequence, double-tagged lamprey passage times through the study area (first detection in the auxiliary water supply channel to last detection at a fishway exit) were longer for refuge users (2.0 d) than for non-users (0.6 d;

t

=2.51, df =57,

P

=0.01).As predicted, the PIT-tagged lamprey showed a tendency to enter the refuges near dawn (south refuge,

Z

=320.6,

P ＜

0.001; north refuge,

Z

=226.3,

P ＜

0.001) and to exit the refuges near dusk (south refuge,

Z

=320.6,

P ＜

0.001; north refuge,

Z

=257.1,

P ＜

0.001) (Fig. 5). At both refuges, the distribution of entry times was significantly different from the distribution of exit times (high concentration

F

-test; south refuge,

F

=139.1,

P

=

＜

0.001; north refuge,

F

=43.77,

P ＜

0.001).Distributions of both entry and exit times did not differ between the north and south refuges (comparison between entries

F

=1.411,

P

=0.236; comparison between exits

F

=0.297,

P

=0.587).

Fig. 4.Frequency histograms of the combined amount of time individual PIT-tagged Pacific lamprey resided in refuges during each year of study (white bars =2012,gray bars =2013, black bars =2014).

3.3.Bonneville Dam passage

In 2014, 99 double-tagged lamprey were detected at radio antennas inside the auxiliary water supply channel. Of these, 54 (55%) were refuge users and 45 (45%) were non-users. The point estimate of dam passage success was higher for non-users (75.6%) than users (68.5%).However, the logistic regression model comparison (

n

=16 models)indicated that no covariates, including refuge use, were statistically significant predictors of dam passage success. The most parsimonious (i.e., ‘best’ model) included only refuge use (χ=0.60,

P

=0.440).

Fig. 5.Rose plots of north and south refuge entries (dark gray) and exits (light gray) by PIT-tagged adult Pacific lamprey for a 24-h clock.

3.4.Detection upstream from Bonneville Dam

In the logistic regression model comparison using 2012 data, the best model for predicting upstream detection included refuge use (χ=0.06,P =0.805), lamprey girth (χ=10.83, P ＜ 0.001), and release date(χ=5.91, P =0.015). Early migrants and those with larger girth were more likely to be detected upstream. In 2013, the best model included refuge use (χ=0.57, P =0.450) and lamprey weight (χ=5.95,P =0.015) and indicated heavier lamprey were more likely to move upstream. The best model in 2014 included refuge use (χ=5.59,P =0.018) and length (χ=12.05, P ＜ 0.001), with longer fish and nonusers more likely to move upstream. A total of 855 lamprey had length,weight, girth and release date data across years. fish length (χ=24.04,P ＜ 0.001) and refuge use (χ=4.0, P =0.046) were included in the best model. Lamprey use of a refuge was associated with a ～10% lower predicted probability of upstream detection, after accounting for length(Fig. 6).

Fig. 6.Logistic regression estimates of the probability of Pacific lamprey occurrence at sites upstream from Bonneville Dam in all years combined for refuge users (solid line) and non-users (dashed line) with 95% confidence intervals indicated. Includes all lamprey that were detected near the exit at the top of the Powerhouse 2 fishway at Bonneville Dam.

4.Discussion

Adult Pacific lamprey were clearly attracted to and made use of the refuges we tested. We estimated that up to 36% of PIT-tagged fish that passed Bonneville Dam via a Powerhouse 2 fishway had used a refuge.Lampreys naturally seek out structure and hide under rocks and debris during migration through rivers (Baker et al., 2017; Kelso & Glova,1993; Quintella, Andrade, Koed, & Almeida, 2004; Robinson & Bayer,2005). Refuges proved to be attractive surrogate structures for lamprey to shelter in during passage through a rectilinear fishway environment characterized by a lack of spatial complexity. In addition, refuge boxes were maintenance-free and did not result in debris accumulation or other operational risk.

Video monitoring could be used to obtain more accurate estimates of lamprey use of refuges and to further evaluate fish behavior in and near refuges. Actual refuge use was probably higher than our estimates due to the likelihood of PIT-tag collisions on refuge antennas. Tag collisions occur when two or more PIT-tag transmissions occur simultaneously,precluding either from being read correctly. Some PIT-tagged lamprey remained in a refuge for several weeks. During such periods, other PIT-tagged lamprey that briefly entered the refuge would potentially be missed due to tag collision. Also, PIT detections could not discriminate tagged lamprey inside a refuge from those passing nearby on the outside.Video could help to document this and provide insights into refuge use and behavior by untagged lamprey and other species.

Considering the potential for underestimating refuge use, tagging data clearly indicated that lamprey were seeking out the relatively small refuge openings. Refuge openings represented less than 0.07% of the channel cross-section area, yet approximately one-third of doubletagged lamprey that entered the study area were detected in a refuge.Lamprey were likely attracted to the dark interior of the refuge and its natural rock substrate and relatively low current velocities. In addition,refuges were positioned along the base of fishway walls (Fig. 2) where lamprey are often observed (Keefer et al., 2010; Kirk, Caudill, Johnson,Keefer, & Clabough, 2015). Lamprey orientation to substrate undoubtedly increased their likelihood of encountering a refuge entrance.

Refuge users exhibited significant morphological differences when compared to Pacific lamprey that entered the study area but did not use a refuge. In particular, individuals with a smaller gap between the first and second dorsal fins were more likely to use refuges than those with a larger dorsal distance, though we note that this trait was only measured in 2014. Dorsal distance in Pacific lamprey is positively correlated with time to spawning; that is, lamprey having a smaller gap are closer to sexual maturity (Clemens, 2011; Lampman et al., 2016). Lamprey closer to spawning may be more easily exhausted and/or more likely to seek refuge in darkened areas with natural substrate. Indeed, adult Pacific lamprey having shorter dorsal distance were shown to have lower passage success when experimentally exposed to burst swimming challenges (Kirk, Caudill, Tonina, & Syms, 2016). Refuges may facilitate passage of this segment of the fishway population, thereby increasing the biocomplexity of populations upstream (Agostinho, Pereira, Oliveira, Freitas, & Marques, 2007; Lucas & Baras, 2008; Poff, 1997).

In addition to finding and entering refuges, some lamprey used them for extended periods and use of the refuges may have reduced downstream movements and lamprey exits into the tailrace. In each year, one or two fish were detected in a refuge for over a month and over half of the animals detected entering a refuge used it for longer than 12 h. We predicted that most lamprey would remain in refuges during daytime and then proceed upstream the following evening. While lamprey did enter the refuges at dawn and leave near dusk, they often remained in the refuges for multiple days. This extended holding may have been needed to rest and recover from strenuous swimming through the fishways (Quintella et al., 2004) or to avoid perceived predators (Kirk et al.,2015).

Alternatively, extended holding by some lamprey may have been related to advanced maturation status; reflecting a shift from active migration to cryptic post-migration behaviors. Whether extended use of refuges had positive or negative fitness consequences is unknown.Future studies should seek to relate refuge use to underlying behavioral factors (exercise recovery, predator avoidance, pre-spawn refuge seeking, etc.). In addition,fitness consequences should be considered when evaluating fishway effectiveness, particularly for non-philopatric or facultative migrants like Pacific lamprey.

Refuge use did not obviously result in improved passage success of double-tagged adult Pacific lamprey at Bonneville Dam. Moreover,refuge occupation could delay migration and prevent timely arrival at the upstream-most spawning habitats. Alternatively, we speculate that refuges afford passage for a previously obstructed part of the population.We hypothesize that most refuge users were closest to spawning; a group most likely to overwinter and spawn in the farthest downstream reaches of the drainage, including unmonitored tributaries emptying into the Bonneville Dam reservoir. This theory was supported by evidence that refuge users were less likely than non-users to migrate to upstream sites in the year of tagging. The relatively poor swimming performance of adult lamprey with short dorsal distance and the disproportionately high abundance of sexually-mature adult Pacific lamprey immediately downstream from some dams suggests that ability to traverse obstacles decreases as lamprey reach final sexual maturation (Clemens, 2011; Kirk et al., 2016).

Specific bottlenecks to lamprey passage have been identified at Bonneville Dam fishways, and these are areas where installation of refuges might be most beneficial (Keefer, Caudill, Clabough, et al.,2013). Notable problem areas were in the lower fishway at junction pools where collection channels joined the over flow-weir sections of the ladders and through serpentine weir (i.e., vertical slot) sections near the tops of ladders, where radio-tagged lamprey had high turn-around rates.Judicious use of refuges could help to retain lamprey in problem areas or provide a collection mechanism for lamprey-specific fishways (Moser et al., 2011).

Testing of the refuges at Bonneville Dam showed promise for adult Pacific lamprey and potentially for other cryptic fishes that use fishways.Providing both natural and artificial structure for fish to shelter in rivers and streams is a common fish management technique (Abbe & Montgomery, 1996; Russell et al., 2008; Zalewski, Thorpe, & Naiman, 2001).Applying this concept to fishways, while retaining hydraulic and maintenance requirements, deserves further consideration.

The ultimate expression of providing refuges would take the form of nature-like fishways (Castro-Santos et al., 2009). Nature-like fishways use natural materials (logs, boulders, rock substrates) that introduce structural complexity along with heterogeneous hydraulics. Nevertheless, comparison of a nature-like bypass in Sweden to nearby natural streams indicated that fish retention in the bypass would benefit from addition of overhanging vegetation where fish could shelter (Tamario,Degerman, Donadi, Spjut, & Sandin, 2018). Hence, more complex structural elements may be needed in fishways to increase retention and reduce predation effects, particularly for species at risk. In the case of Pacific lamprey, this likely would involve construction of more continuous sheltering elements, or even a false floor, under which they could migrate without exposure to excessive light or predators.

Ethics statement

The methods used in this work were approved by the Institutional Animal Care and Use Committee at the University of Idaho, Moscow,Idaho.

Conflicts of interest

The authors have no conflicts of interest to declare.

Acknowledgements

We thank M. Hanks for providing technical assistance on all aspects of this research. E. Johnson, C. Noyes, D. Joosten, M. Kirk, and S. Lee helped with lamprey collection and tagging. The design, fabrication, and installation of the refuges would not have been possible without the exceptional skills and efforts of J. Simonson, J. Moser, G. Wolf, and B.Wassard. We also thank the rigging crew at Bonneville Dam for help with equipment installations. A. Traylor, B. Hausmann, I. Royer, B.Bissell, and N. Zorich of the U.S. Army Corps of Engineers provided assistance on many fronts throughout the project. J. Smith helped with statistical analysis. Administrative assistance was provided by D. Dey, P.McAteer, and S. Downing. M. Jepson and T. Clabough provided administrative and database support. J. Butzerin, D. Dey, and B. Sandford reviewed an early version of this manuscript. Funding for this work was provided by the U.S. Army Corps of Engineers, Portland District.