Protection and guidance of downstream moving fish with horizontal bar rack bypass systems.

During their life cycle, fish carry out distinct movements within rivers and migrate upstream and downstream to reproduce, to feed, and to shelter in refuge habitats. During downstream movements, they can incur severe injuries that may be lethal directly or indirectly over time when passing through hydropower plants or when being entrained at other water intakes. Horizontal bar rack bypass systems are a state-of-the-art technology to protect and guide downstream moving fish towards a reasonably safe corridor around water intakes. They have been in operation at multiple hydropower plants for more than a decade, but only little is known about the potential fish protection and guidance efficiencies and the fine scale reactions of different fish species when encountering such racks. To resolve this, systematic live fish laboratory tests were conducted under various hydraulic conditions involving a diverse assemblage of riverine fish species differing in their swimming behavior and morphology. Six riverine species, namely spirlin (Alburnoides bipunctatus), barbel (Barbus barbus), nase (Chondrostoma nasus), brown trout (Salmo trutta fario), Atlantic salmon (Salmo salar), and European eel (Anguilla anguilla) were tested with a rack consisting of foil-shaped bars, clear bar spacings of 15 and 20 mm, a horizontal rack angle of 30° to the flow direction, and a full depth open channel bypass. Variations in fish behavior were observed between different species and hydraulic conditions, but the results suggest that the guidance and protection efficiencies primarily depend on the ratio of the fish width to the clear bar spacing. Larger fish were well protected by the horizontal bar rack, while smaller fish frequently passed through the rack. New equations are proposed to estimate the protection and guidance efficiencies as a function of the clear bar spacing and the fish species' biometry, which is highly relevant to assess the effect of horizontal bar racks as fish protection measures prior to installation.

Introduction

Fish guidance structures at hydropower plants

On a global scale, hydropower is the most important renewable electricity source, with the best conversion efficiency, high energy payback ratios, and relatively low greenhouse gas emissions (IPCC, 2011). A disadvantage of hydropower plants (HPPs) is their multifaceted impact on river ecosystems, which includes among others hydropeaking, the disruption of the sediment transport continuum, altered flow conditions of river stretches through damming, the impairment of fish movements, or change in species compositions. Downstream moving fish may incur severe or even lethal injuries when passing through HPPs without safe downstream passages and through water intakes (Čada, 2001). To mitigate these impacts, different fish protection measures have been developed to protect and bypass downstream moving fish, including physical barriers, mechanical behavioral barriers, sensory behavioral barriers, and collection systems (Larinier and Travade, 2002; Turnpenny and O'Keeffe, 2005; USBR, 2006; Kriewitz, 2015; Schwevers and Adam, 2020). Physical barriers are typically either inclined with an inclination angle γ to the approach flow vector in longitudinal section (Fig. 1A) or angled with an angle α in plan view, where the bypass is located at the downstream rack end (Fig. 1B; Ebel, 2016). To evaluate the velocity fields of angled racks, the approach flow velocity vector can be converted to the rack-parallel and rack-normal flow velocity components V and V, respectively (Fig. 1B; Meister et al., 2020a).

Fig. 1

Different rack layouts with the average approach flow velocity U: (A) inclined rack and (B) angled rack with the bypass location indicated at the downstream rack end.

Horizontal bar racks (HBR) are angled racks (Fig. 1B) with horizontally aligned bars. They are often combined with a bypass at the downstream rack end to allow for safe fish downstream passage and are then referred to as horizontal bar rack bypass systems (HBR-BSs). Since the first installation of an HBR-BSs in 2006 in Germany (Ebel, 2016), more than 100 HBR-BSs were installed at small to medium-sized HPPs with a design discharge of Q < 120 m3/s (Meister, 2020). They are designed considering the target fish species' swimming capacity and size, such that the approach flow velocity is typically 0.40 m/s ≤ U ≤ 0.80 m/s, the horizontal approach flow angle 20° ≤ α ≤ 40°, and the clear bar spacing s ≤ 15 mm (Ebel, 2016). If designed accordingly, HBR-BSs cause only small rack head losses (Meister et al., 2020b) and are advantageous in operation as floating debris can be automatically flushed downstream through the bypass (Ebel, 2016).

Fish biometry and protection potential of physical barriers

Horizontal bar racks are designed to act primarily as physical barriers (Ebel, 2016). Their potential to physically protect downstream moving fish therefore depends on the fish species' biometry and the clear bar spacing of the rack. Different proportional metrics can be calculated from the following characteristic dimensions: w = absolute fish width [m], h = absolute fish height [m], and TL = total fish length [m], which is measured from the nose to the caudal fin. The relative fish width and the relative fish height are defined as w = w/TL and h = h/TL, respectively. Intuitively, the fish width is decisive for determining the clear bar spacing s of physical barriers with vertical bars, whereas the fish height is limiting for physical barriers with horizontal bars. However, fish are not static and can adjust their body axis and position in the water column. Hence, when approaching a barrier, they can turn to their sides, that is, rotate to the side by 90°, allowing them to pass through HBRs even if h > s. This is corroborated by monitoring campaigns, when fish of different species with h > s were caught with stow nets at the turbine outlets of HPPs with HBRs (e.g. Wagner et al., 2019). Rack passages with h > s were also observed in laboratory studies (e.g. de Bie, 2017), suggesting that fish do rotate to the side to pass through HBRs. However, despite these indirect observations from field and laboratory studies, little is known to what extent fish adjust their position in the water column and turn to their sides when approaching and eventually passing through HBRs. Due to this knowledge gap, Ebel (2016) suggested to design s of physical barriers independent of the orientation of the bars. Thereby, the critical total fish length TLcrit, defined with Eq. (1), is the minimal total fish length, above which fish are physically blocked at barriers. Table 1 lists w and h of selected fish species, which were derived from multiple monitoring campaigns and the corresponding TLcrit calculated with Eq. (1) for s = 15 mm. Although the fish biometry can vary between rivers and individuals, w and h can be approximated fairly well from TL with w and h. Racks with either horizontally or vertically aligned bars are therefore a physical barrier to fish with h ≥ w if s < w.

Table 1

Relative fish width w, relative fish height h, and TLcrit for s = 15 mm for selected fish species (data from Ebel, 2016; Wagner et al., 2019).

Fish species	wf,rel [−]	hf,rel [−]	TLcrit1 [mm]
Spirlin (Alburnoides bipunctatus)	0.09	0.20	167
Barbel (Barbus barbus)	0.11	0.17	136
Nase (Chondrostoma nasus)	0.11	0.24	136
Brown trout (Salmo trutta fario)	0.10	0.19	150
Atlantic salmon (Salmo salar)	0.10	0.18	150
Chub (Squalius cephalus)	0.12	0.21	125
European eel (Anguilla anguilla)	0.032	0.032	500

For s = 15 mm.

Accounts for the body flexibility of eels.

Hydraulic design criteria for fish guidance at HBR-BSs

The angle to the approach flow vector, flow velocities, velocity gradients, and turbulent flow structures at the bypass are important design parameters to effectively protect and safely guide downstream moving fish around water intakes with HBR-BSs. Angled racks like HBRs are typically arranged such that the rack-parallel flow velocity component V exceeds the rack-normal flow velocity component V (V/V > 1; NMFS, 1997, Courret and Larinier, 2008; Vowles et al., 2013; Fig. 1B). Thereby, downstream moving fish which follow the main flow are guided towards the bypass. Importantly, to avoid fish impingments, V should be smaller than the sustained swimming speed of fish (Turnpenny and O'Keeffe, 2005; Ebel, 2016). Furthermore, the hydraulic conditions at HBR-BSs are important for guiding fish effectively with little delay (Ebel, 2016). It is known that high turbulence levels, large-scale vortices, and large rates of spatial flow acceleration may hamper fish orientation, reduce their swimming stability, and result in avoidance reactions (Haro et al., 1998; Enders et al., 2012; Vowles et al., 2014; Silva et al., 2016; Silva et al., 2020). Bypass systems should therefore be designed with low turbulences, smooth flow transition from the rack to the bypass, and moderate rates of spatial flow acceleration. (Haro et al., 1998; Turnpenny and O'Keeffe, 2005; Kemp et al., 2006; Ebel, 2016; Beck et al., 2020). The rate of spatial flow acceleration towards the bypass is often described as the velocity ratio of the bypass inlet to mean approach flow velocity to VR = U/U (Ruggles and Ryan, 1964; Ducharme, 1972; USBR, 2006). For salmon smolts at Louvers, Ruggles and Ryan (1964), Ducharme (1972), and Turnpenny and O'Keeffe (2005), recommended VR ≥ 1.4, VR ≥ 1.0–1.5, and VR ≈ 1.4, respectively. On the basis of multiple studies, USBR (2006) suggested VR = 1.0–1.5 not only for salmon smolts, but also for brown trout and white catfish. NMFS (1997) proposed that U should be equal to or exceed the maximal flow velocity at the downstream rack end. By evaluating these studies, Ebel (2016) recommended 1.0 ≤ VR ≤ 2.0 for HBR-BSs. To achieve moderate spatial rates of flow acceleration towards the bypass, common design guidelines demand a minimal relative bypass discharge in the range 2% ≤ Q = Q/Q ≤ 5% for angled racks, where Q is the bypass discharge (Larinier, 1998; Ebel, 2016).

Recent live fish studies at HBR-BSs

De Bie (2017) and de Bie et al. (2021) studied the behavior of downstream moving barbel (Barbus barbus) and juvenile chub (Squalius cephalus) at an HBR-BS with a circular bar cross section and s = 10 mm. With the HBR with α = 30° and U = 0.36 m/s, 32.1% of the barbels with an average total length of =84.2 mm ± 7.1 mm entered the bypass, 29.5% passed through the rack, and 38.5% were inactive. This means that 52% of the active barbels used the bypass, which is called fish guidance efficiency in the following. The fish guidance efficiency of chub with =88.3 mm ± 7.5 mm was 70%. All tested fish could physically pass through the HBR (w < s), but only if they rotated to the side since h was larger than s. Although a non-negligible share of fish of both species was guided by the HBR into the bypass, many fish passed through the HBR with s = 10 mm, such that de Bie (2017) concluded that the investigated HBR-BS did not sufficiently guide juvenile cyprinids under the tested conditions.

In a similar study, de Bie et al. (2018) investigated the behavior of downstream moving juvenile chub with =108 mm at angled racks with α = 30° and horizontally aligned wedge-wire bars with s = 6 mm (TLcrit = 50 mm; Eq. (1) and Table 1). Despite the small fish size, the HBR was a physical barrier to all tested fish (w > s), such that no rack passages were observed. However, it was noted that fish avoided abrupt velocity gradients at the bypass inlet and that higher flow velocities increased the number of refusals. Berger (2018) conducted live fish laboratory experiments with hatchery-reared Atlantic salmon smolts (Salmo salar) and European eels (Anguilla anguilla) at HBR-BSs with rectangular bars for different approach flow angles (α = 30, 45, 55, 70°), clear bar spacings (s = 10, 18, 30 mm), and approach flow velocities (U = 0.4, 0.5, 0.7, 0.8 m/s). The lowest 50 mm of the HBRs close to the channel bottom were physically blocked by a beam. Only a few rack passages of salmon smolts with =160 mm were observed. The fish protection efficiency FPE, defined as the number of fish which did not pass through the rack over the total number of tested fish, was FPE = 100%, FPE ≈ 608/610 ≈ 99.7%, and FPE ≈ 173/182 ≈ 95% for s = 10 mm (TLcrit = 100 mm; Eq. (1) and Table 1), s = 18 mm (TLcrit = 180 mm), and s = 30 mm (TLcrit = 300 mm), respectively. Berger (2018) did not report the size distribution of the tested eels, but only mentioned that most eels were in the range of 600 mm ≤ TL ≤ 800 mm, while some individuals were as small as TL = 400 mm. None of the eels passed through the HBR with s = 10 mm (TLcrit = 333 mm; Eq. (1) and Table 1) and s = 18 mm (TLcrit = 600 mm), while 15% of the eels passed through the HBR with s = 30 mm (TLcrit = 1000 mm).

Kammerlander et al. (2020) did not study classical HBR-BSs as those investigated herein, but a “flexible fish fence”, where fish are protected by horizontally aligned steel cables with s = 10, 20 mm, α = 20, 40°, and U = 0.3, 0.5, 0.65 m/s. Their live fish experiments were conducted in a 20 m long and 2 m wide outdoor channel with a gravel bed and constant approach flow depth of h = 0.5 m. Chub, grayling (Thymallus thymallus), brown trout (Salmo trutta) and rainbow trout (Oncorhynchus mykiss) with 100 mm ≤ TL ≤ 200 mm were tested, where the response reactions rack passage, bypass passage, and headwater were distinguished. The latter includes all fish which did neither pass through the rack nor through the bypass. With s = 10 mm (83 mm ≤ TLcrit ≤ 100 mm; Eq. (1) and Table 1), only five out of 1346 individuals passed through the rack, leading to a fish protection efficiency of FPE ≈ 100% for all species. However, for s = 20 mm (167 mm ≤ TLcrit ≤ 200 mm), especially grayling were poorly protected, where three-quarter of all specimens passed through the flexible fish fence. Moreover, only 6%, 14%, and 37% of grayling, trout, and chub, respectively, entered the bypass for s = 20 mm, indicating poor guidance efficiencies or unfavorable conditions in the bypass.

Goals of the study

Although multiple HBR-BSs have been in operation for more than a decade and several live fish studies were carried out, little is known about the behavior of downstream moving fish when they encounter an HBR-BS and the protection and guidance efficiencies at these state-of-the-art fish downstream passage facilities. For example, it was unclear to which degree fish rotate to the side to pass through HBRs, which is a key information for choosing an adequate clear bar spacing. Up to now, it was not possible to estimate the fish guidance and protection efficiency for differently sized fish, which is of high importance as this information is needed to assess the effect of HBR-BSs prior to installation at hydropower plants. The goal of the present study is to close these research gaps by (a) briefly describing the hydraulics of HBR-BSs, (b) reporting the fish swimming behavior and species-specific guidance and protection efficiencies, (c) analyzing the effect of the fish species, size distribution, and the clear bar spacing on the former, (d) setting up a logistic regression model to analyze the overall effect of different parameters across species, and (e) analyzing typical behavior with a sector analysis. Together with the studies of Meister et al., 2020a, Meister et al., 2020b, which focus on the head losses and velocity fields of HBR-BSs, the present investigation provides the technical knowledge such that HBR-BSs can be successfully designed while accounting for local fish species' requirements, site-specific hydraulic conditions, cost-efficiency, and sustainable operation.

Materials and methods

Experimental setup

All experiments were conducted in a 30 m long, 1.5 m wide, and 1.4 m deep indoor laboratory channel with a horizontal bed at the Laboratory of Hydraulics, Hydrology and Glaciology (VAW) of ETH Zurich, Switzerland (Fig. 2). The water temperature of the closed water circuit, to which the channel and the holding tanks were connected, was controlled with a 51 kW cooling system of Climaveneta (NECS-W 0182), such that all live fish tests were carried out with water temperatures in the range of 12 °C ≤ T ≤ 16 °C. The approach flow depth was kept constant at h = 0.90 m for all experiments and the discharge could be regulated to control the approach flow velocity. A symmetric approach flow was achieved with a honeycomb flow straightener and two hard foam floaters. A 1.5 m long acclimatization compartment was installed at the channel inlet, such that the acclimatization compartment outlet was located 10.5 m upstream of the bypass inlet on the left channel side in flow direction (opposite of bypass inlet; Fig. 2A,D). The fish movements were recorded with five submerged cameras of the type acA2040-35gmNIR (Basler) with a maximal resolution of 2048 × 1536 px2, 185° fisheye lenses of type FE185C086HA-1 (Fujifilm), and a waterproof IP67 Orca S dome (autoVimation), which were installed upstream of the rack (Fig. 2A,D). The left channel wall (in flow direction) and the channel bottom were made of concrete, whereas the right channel wall was made of glass to allow for visual observations from the dark observation room. A perforated foil, imitating a concrete wall, was installed at the glass wall, such that fish could not see into the observation room and were not mirrored at the glass wall, while they could still be observed from the side (Fig. 2B). To ensure constant indirect light conditions for all experiments, white sheets were installed above the channel and illuminated with eight 1000 W halogen spot lights. A 10 cm wide concrete wall was installed at the downstream rack end to separate the bypass from the tailwater. The bypass was w = 0.25 m wide and expanded to w = 0.50 m in its downstream section, such that fish entering the bypass were not exposed to large flow velocities until the end of the experiment. The bypass discharge and flow depth were regulated with a separate flap gate independent of the approach flow conditions (Fig. 2D). Flow velocities were measured with an acoustic Doppler velocimeter (ADV, Nortek AS), mounted on a 3D traverse system. The HBR consisted of foil-shaped bars (Fig. 2C), which were connected with vertical threaded bars and spacers, allowing for investigating different clear bar spacings s.

Fig. 2

(A) Photo of the test facility in flow direction, (B) detailed view from the observation room through the glass wall with the perforated foil on three salmon parr, (C) detailed cross section of the investigated bars, and (D) sketch of the laboratory channel.

Experimental procedure

3D instantaneous flow velocities were measured with a down-looking ADV. The measurement volume height was 7 mm and the measurement duration was 30 s with a sampling rate of 200 Hz at each measurement location. All data were despiked following Goring and Nikora (2002) and Mori et al. (2007) with a minimal correlation and signal-to-noise ratio of 70% and 10 db, respectively.

Only wild fish were used in the experiments, which were caught with mild DC electrofishing (Grassl EL 64_II, 7.0 kW and Grassl ELT 60-II, 1.3 kW) in nearby rivers (details in supplementary materials), from where they were transported to the laboratory in aerated and temperature-stable tanks. As there are no self-sustaining salmon populations (Salmo salar) in Switzerland, they were caught from a stocked population in a natural river (Möhlinbach, details in supplementary materials). The river water in the transport tanks was diluted with water from the two holding tanks with a capacity of 3.5 m3 each, such that fish could slowly adapt to the laboratory conditions. Water temperature changes during acclimatization to the water from the fish holding tanks were below ΔT ≈ 1 °C/h and fish were not released into the holding tanks before the water temperature difference was below 1 °C. Each holding tank was equipped with two filter pumps and an additional air pump. It also contained tubes to provide shelter. A viewing window was installed in the top cover of the tanks, which allowed to expose the fish to the natural day-night rhythm. Fish were kept up to five days in the laboratory (details in supplementary materials), where their condition (physical appearance, natural behavior), the water quality (oxygen concentration, Hanna Instruments HI9147, pH), water temperature, and turbidity were monitored daily. Fish were not fed during their stay in the laboratory and different fish species were not simultaneously kept in the same compartment of the holding tanks. The maximum number of fish kept in one holding tank was 44. All experiments were conducted during daytime between 7:30 a.m. and 6:30 p.m. To reduce the number of fish, individuals were used in several experiments. Learning effects and distress were reduced by not testing individuals more than once per day with a maximum of four experimental rounds per individual and thereby always subjecting them to different rack and/or bypass configurations. After the last experimental round, all fish were set free in the same river reach where they were initially caught.

In previous live fish laboratory experiments, Flügel et al. (2015) and Albayrak et al. (2020) found that the group size (number of individuals tested in an experimental round) affected the fish behavior of some species. When tested individually or in small groups, spirlin showed no substantial differences, while barbels and brown trout were much more active when swimming in groups of three individuals. The experiments of the present study were therefore conducted with groups of three fish of the same species with the exception of two experiments (E7 and E13 in Table 2), where eels were tested individually. Limiting the number of fish per experiment to three allowed for analyzing the fish behavior individually for most experiments. At the beginning of an experiment, three fish were caught with a dip net from the holding tanks and released into a bucket or cuvette. A picture with a reference scale was taken to measure the total fish length TL. Afterwards, the fish were released into the acclimatization compartment of the channel (Fig. 2A,D). For experiments with U = 0.7 m/s and all experiments with nase, a brick was positioned in the acclimatization compartment, behind the rack, and in the bypass, offering fish a zone with reduced flow velocities. After an acclimatization time of 15 min, the brick was removed from the compartment, which was simultaneously opened, allowing for the fish to swim within the entire channel. Besides recording the fish movements with the submerged cameras, the fish behavior was observed and written down on manual protocols from the observation room. Each experiment was completed when all three fish either passed through the rack or the bypass, or after the maximal experimental duration of 30 min. After each experiment, fish were caught with a dip net and brought back to the holding tanks.

Table 2

Test program of the live fish tests, summarizing the different hydraulic conditions, fish dimensions, and number of specimens tested per configuration; U: mean approach flow velocity from continuity, U: flow velocity at the bypass inlet, VR: ratio of bypass inlet to mean approach flow velocity, s: clear bar spacing, TLmin, TLmax, , σ: minimal, maximal, average, and standard deviation of total fish length, N: total number of tested fish, n: number of active fish.

Test	Rack	Uo [m/s]	Uby,in [m/s]	VR [−]	sb [mm]	Fish species	TLmin–TLmax (TL¯, σTL) [mm]	N [−]	n [−]	n/N [%]
E0	No rack	0.50	0.60	1.2	–	Spirlin	85–119(99, 10)	21	21	100
E1	HBR	0.50	0.60	1.2	15	Trout	112–196(158, 19)	75	30	40
E2		0.50	0.60	1.2	20	Spirlin	81–107(96, 7)	33	33	100
E3						Barbel	103–208(163, 28)	45	41	91
E4						Nase	59–93(72, 9)	18	15	83
E5						Trout	91–187(139, 23)	36	27	75
E6						Salmon	96–149(119, 13)	24	24	100
E7						Eel	472–796(665, 106)	10	8	80
E8		0.50	0.70	1.4	20	Spirlin	87–129(105, 8)	27	24	89
E9						Barbel	110–163(135, 15)	24	19	79
E10						Nase	59–83(71, 6)	21	20	95
E11						Trout	99–210(140, 27)	42	28	67
E12						Salmon	91–134(110,10)	27	24	89
E13						Eel	413–828(649, 140)	7	7	100
E14		0.70	0.85	1.2	20	Spirlin	80–123(101,10)	33	33	100
E15						Barbel	80–202(155, 35)	24	23	96
E16						Trout	88–208(143, 26)	51	27	53
E17		0.70	1.00	1.4	20	Spirlin	78–113(95, 9)	24	22	92
E18						Barbel	75–190(143, 33)	21	7	33
E19						Trout	86–201(127, 28)	45	24	53

Parameter range and test program

The tested fish species, Alburnoides bipunctatus (spirlin), Barbus barbus (barbel), Chondrostoma nasus (nase), Salmo trutta fario (brown trout), S. salar (salmon), and Anguilla anguilla (European eel), are hereafter called by their common names (given in parenthesis above). Fish were tested under four hydraulic conditions. Two average approach flow velocities U (from continuity) were investigated, namely U = 0.5 and 0.7 m/s. The flow conditions in the bypass were quantified with the velocity ratio VR, defined as the ratio between the average approach flow velocity at the bypass inlet U and U, that is VR = U/U. Based on the literature recommendations described in Section 1.3, VR = 1.2 and 1.4 were investigated for both approach flow velocities. The experiments were conducted with the HBR with α = 30° and s = 20 mm, except for the test series E1, where trout were tested with s = 15 mm (Table 2). These parameters were chosen on the basis of previous research projects (e.g. Berger, 2018), experiences at monitoring campaigns (e.g. Zaugg and Mendez, 2018), hydraulic measurements at various HBR configurations (Meister et al., 2020a, Meister et al., 2020b), and general literature recommendations (Ebel, 2016). As a reference, spirlin were additionally tested without a rack installed in the channel. All test configurations are summarized in Table 2 with the minimal total fish lengths TLmin, maximal total fish lengths TLmax, average total lengths , and the corresponding standard deviations σ. The study focused on fish which could theoretically pass through the HBR (w < s). The parameter N describes the total number of tested fish, while the parameter n represents the number of active fish, which swam close to the HBR (definition in Section 2.4). Nase, eels, and salmon were not tested with U = 0.7 m/s to prevent exhaustion of nase as the individuals were small (Table 2) and due to the limited number of available eels and salmon.

Data analysis

General data analyses were conducted with the software MATLAB R2019b, while all statistical analyses were performed with the software R v.1.2.5033. The video recordings of the submerged cameras were used to extract the fish swimming tracks with a specifically designed Matlab-based fish tracking software, which is described in detail in Harby et al. (2020). The raw images were undistorted by accounting for the fisheye lens effect and the refraction at the air-glass interfaces (the glass dome housing). Each pair of cameras with an overlapping view was calibrated with a stereo calibration using a checkerboard. Finally, the local stereo camera coordinates were transformed to the global channel coordinates, by using 43 fixed reference points on the channel bottom. Moving objects like fish were then detected on each frame, and a motion-based multiple object tracking algorithm was used to group all detections, corresponding to the same object over time. The main challenges were to correctly filter out noise, which resulted from moving reflections at the glass wall or fish shadows on the channel bottom and to correctly merge individual tracks. 2D data of the fish center were retrieved for the whole observation area.

The fish swimming behavior was analyzed with the fish swimming tracks and the manually written protocols were used to verify the fish behavior for each experiment. Three main lines were defined, marking different x-coordinates, where the point of origin was set to the downstream rack end at the channel bottom (Fig. 3). Line 1 indicates the upstream end of the relevant experimental area (X = x/h = −8.3), that is the area which could be studied from the observation room (Fig. 3). Line 2 was defined 30 cm upstream of the rack head, that is X = −2.7 (Fig. 3), indicating if fish swam into the area affected by the HBR. Line 3 represents the downstream end of the relevant experimental area at X = 1.8 (Fig. 3). The channel was further divided into different sectors (Sec), where the sectors Sec1, Sec2, and Sec7 covered a 15 cm wide area adjacent to the channel walls and Sec5 a 15 cm wide area upstream of the rack (Fig. 3). On the basis of these lines, the following behaviors were distinguished for each fish, where the first behavior after the start of the experiment was considered in case of bypass and rack passages. The number of fish is indicated with the variable N and the corresponding subscript:

Fig. 3

Top view of the laboratory channel indicating lines 1–3 and definition of sectors 1–7 (y-axis distorted by factor 2).

Inactive fish: not crossing line 2 within the maximal experimental duration of 30 min.

Bypass passage without rack interaction (RI): the fish passed line 3, without swimming into sector 5.

Bypass passage with rack interaction: the fish passed line 3, after swimming through sector 5 (N).

Rack passage (Nrack), the fish passed through the HBR.

Refusal: the fish passed line 2, but neither entered the bypass (did not cross line 3) nor the HBR (Nref)

Inactive fish and fish which entered the bypass without rack interaction were excluded from all analyses, as they were likely not guided by the HBR. The total number of fish used for the analysis is defined as Ntot = N + Nrack + Nref. The fish guidance efficiency (FGE) and fish protection efficiency (FPE) were defined with Eqs. (2), (3), respectively. FGE and FPE can be calculated for each test configuration or as an average per species, by considering all tested specimens, independent of the hydraulic conditions (, ). The latter is especially useful for practical applications, where approach flow velocities are variable and not well known.

To analyze the fish swimming behavior and to compare different configurations with each other, the so-called residence coefficient R is introduced with Eq. (4). This coefficient is a measure for how much time all fish, tested at one specific hydraulic configuration, spent in a sector in comparison to the other sectors, while accounting for the different sector sizes. It was calculated for all active fish, that is, fish which crossed line 2 in Fig. 3. By normalizing the time a fish j spent in a sector i with the total time the fish j spent in any of the sectors 1–7, each fish was weighted equally, independent of how fast it swam downstream. The normalized residence coefficient R is defined with Eq. (5). A normalized residence coefficient of R = 1 means that all fish of the investigated configuration spent the entire time in the sector i, whereas R = 0 indicates that none of the fish of the investigated configuration entered the sector i.where i = sector number [−], j = fish number [−], n = number of fish tested for this configuration [−], t = time the fish j spent in sector i [s], t = t=total time the fish j spent in any of the sectors 1–7 [s], A = area of the sector i [m2], Atot = A= total area of all sectors 1–7 [m2].

Different statistical analyses were conducted to answer different research questions. Two-sided χ2-tests with a significance level of αsig = 0.05 were conducted to assess if different parameters, such as U and VR, significantly affected the fish behavior. To account for species-specific behavior, these tests were conducted for each species separately. The null hypothesis H0 states that there was no significant difference between the two tested configurations, whereas the alternative hypothesis H1 applies for significant differences. If two test series with a multinomial output are compared to each other, it is only possible to detect if the output differs significantly. No statements can be made about the individual output categories. Therefore, the multinomial output variable “fish reaction” (bypass passage, rack passage, and refusal) was converted to binary data (bypass passage: yes/no; rack passage: yes/no; refusal: yes/no) and χ2-tests were made for each of these categories. The effect of the parameters U and VR on the fish reaction was quantified with generalized linear models (GLM, logistic logit-regression), where the multinomial output variable fish reaction was converted to binary data. Three different GLMs of the general form shown in Eq. (6) were used to quantify which parameters affected the bypass passages, rack passages, and refusals.where OR = odds ratio, β0 = regression coefficient of the intercept, β = (β1, β2, …, β) = regression coefficients, i = parameter number, n = total number of parameters, X = data of parameter i. The odds ratio is defined by Eq. (7) and can be rewritten as Eq. (8) to calculate the probability p(X) given the independent parameter X = (X1, X2, …, X).

The model included the primary regression parameters U and VR. Besides, the number of times a fish was used for the experiments (usage) and the time of the day (morning/afternoon) could have affected the results. These parameters were also included in the GLMs as secondary parameters to improve the model fit. The parameters U (0.5 and 0.7 m/s), VR (1.2 and 1.4), time (morning: start of experiment until 1 p.m.) and afternoon (start of experiment after 1 p.m.) were modeled as factors, while the parameter “usage” was defined as a numerical variable. In all statistical analyses, each fish was considered as an independent data point as most fish within an experiment consisting of three individuals showed individual movement patterns. Yet, some fish species (e.g. spirlin) showed strong schooling behavior during downstream movement within the channel (which will partially result in pseudoreplication in the analysis), but to some extent different movement patterns at the rack and channel. To increase the statistical power of the GLMs, a complete randomization of the order of experiments was pursued, which could only partially be achieved due to various experimental limitations, such as the limited number of holding tanks and the availability of different fish species.

Results

Hydraulics of horizontal bar rack bypass systems

Fig. 4 shows the normalized streamwise flow velocities U/U in a horizontal plane close the channel bottom (Z = z/h = 0.14) for the HBR configuration with foil-shaped bars, s = 20 mm, U = 0.5 m/s, and VR = 1.4. The arrows, shown at each measurement location, indicate the direction and magnitude of the longitudinal and transversal flow velocities U and V, respectively. Fig. 4 shows that the HBR hardly affects the upstream velocity field. The bypass affects the velocity field only locally, despite a relative bypass discharge of Q = Q/Q = 16.9%, calculated from the end overfall at the bypass outlet based on the Poleni equation. The ratio of the rack-parallel V to rack-normal V flow velocity components 40 mm upstream of the HBR (measured rectangularly), was 1.7 ≤ V/V ≤ 1.9 for 0 ≤ Y ≤ 0.8 in Fig. 4, which is close to the theoretical value of V/V = cot(30°) = 1.73 derived from the vector decomposition by neglecting the small transversal flow velocity componenents V. The commonly used criterion V/V > 1 (cf. Section 1.2) is therefore fullfilled in the entire area upstream of the HBR, such that fish should be guided towards the downstream rack end. The relative bypass discharges for the configurations U = 0.5 m/s and VR = 1.2, U = 0.7 m/s and VR = 1.2, and U = 0.7 m/s and VR = 1.4 were 13.7%, 14.6%, and 16.8%, respectively. The hydraulic characteristics at the bypass center line (Y = −0.2, Z = 0.5) of all investigated configurations (cf. Table 2), including the normalized streamwise flow velocities U/U, the spatial velocity gradient in x-direction SVG = ∂U/∂x, and the turbulent kinetic energy TKE, are shown in Fig. 5.

Fig. 4

Flow velocities at the HBR-BS close to the channel bottom (Z = 0.14) with a relative bypass discharge of Q = 16.9%, foil-shaped bars, s = 20 mm, U = 0.5 m/s, and VR = 1.4.

Fig. 5

Characterization of the hydraulics at the bypass center line (Y = −0.2, Z = 0.5) for different hydraulic configurations with (A) the normalized streamwise flow velocity U/U, (B) the spatial velocity gradient in x-direction SVG, and (C) the turbulent kinetic energy TKE; the horizontal dashed lines in (A) highlight U/U = 1.2 and U/U = 1.4, while the vertical dashed lines indicate the bypass entrance at X = 0.

Upstream of the bypass inlet up to X ≈ −0.5, flow velocities were equal to the average approach flow velocity (U/Uo ≈ 1), and then increased to U/U = 1.2 and U/U = 1.4, corresponding to VR = 1.2 and VR = 1.4, respectively, at the bypass inlet (Fig. 5A). The spatial velocity gradient in x-direction was SVG < 0.1 m/s/m, except near the bypass inlet (−0.5 ≤ X ≤ 0.0), where it peaked at SVG = 0.8 m/s/m for U = 0.7 m/s and VR = 1.4 (Fig. 5B). For all tested hydraulic conditions, the turbulent kinetic energy TKE was small in the entire area upstream of the HBR and within the bypass (1 × 10−3 ≤ TKE ≤ 3 × 10−3 m2/s2; Fig. 5C).

Fish swimming behavior and species-specific guidance and protection efficiencies

The response reactions of all tested fish are summarized in Fig. 6 to Fig. 11. The guidance and protection efficiencies FGE and FPE, defined in Section 2.4, can be directly read from the x-axis, whereas the absolute number of active fish used for the analysis is indicated with white numbers. Videos of the typical fish swimming behavior of each species and examples of rack passages are provided as supplementary materials.

Fig. 6

Response reactions of spirlin without a rack installed (E0) and with the HBR for four different hydraulic configurations.

Fig. 11

Response reactions of European eel for different hydraulic configurations.

Spirlin (Alburnoides bipunctatus)

Behavior: Spirlin showed a distinct schooling behavior in the experiments and rarely swam individually. With only a few exceptions, most spirlin swam downstream with positive rheotaxis, which means that they were facing the oncoming current. Spirlin typically swam downstream in zigzag movements from one to the other wall and thereby used the whole channel width. Downstream of line 2, these zigzag movements occurred between the HBR and the right channel wall, which guided spirlin to the bypass entrance. Spirlin mostly swam close to the bottom and approached the water surface almost solely next to the channel wall. The movements of spirlin seemed very cautious and controlled and they reacted very sensitively to spatial flow accelerations at the bypass inlet. Spirlin swam close to the HBR and sometimes gently touched the rack with their caudal fins, but they hardly immersed their fins between the bars. Spirlin passing in between the bars, and thus through the rack, did so fast without hesitation.

Reaction: Spirlin were tested with four different hydraulic conditions and without a rack installed for reference, where the behavior was counted as a rack passage if fish crossed the axis, where the HBR was installed for the other experiments. With the exception of three spirlin, which swam into the bypass by chance, all other specimens passed the rack axis without an HBR installed with U = 0.5 m/s and VR = 1.2 (E0 in Fig. 6). For the same hydraulic conditions but with the HBR installed, all spirlin entered the bypass (E2 in Fig. 6), which was a statistically significant difference to the configuration without a rack (χ2 = 33.462, p < 0.001). An increased U led to significantly more rack passages (χ2 = 7.361, p = 0.007) and less bypass passages (χ2 = 8.478, p = 0.004; E8 in Fig. 6). Larger VR significantly reduced the number of bypass passages (χ2 = 17.495, p < 0.001), while the number of refusals increased (χ2 = 18.452, p < 0.001), indicating that spirlin react sensitively to flow accelerations. The configuration E17 performed worse than all other configurations, which might be a result of the combination of the larger approach flow velocity and larger velocity ratio, leading to large flow velocities at the bypass inlet of U = U·VR = 1.0 m/s. Averaged over all test configurations with an HBR installed, =78% and =92%.

Barbel (Barbus barbus)

Behavior: The most distinctive behavior of barbels was that they were frequently swimming close to the channel walls. Often, barbels swam towards the glass wall after they left the acclimatization compartment. They were then slowly moving downstream with positive rheotaxis and many of them entered the bypass without rack interaction and were therefore excluded from further analyses (cf. Section 2.4). Barbels swam almost exclusively very close to the bottom. Only in a few occasions were barbels observed swimming towards the water surface along the glass wall. Barbels often passively kept their position for minutes with little effort on the channel bottom. In some experiments, barbels showed a schooling behavior when approaching the HBR or the bypass, whereas in other experiments each specimen acted individually. When barbels swam towards the HBR, they often touched it only briefly with their caudal fins and moved then back to the glass wall. In a few occasions it was observed that barbels were guided perfectly from the upstream rack end along the rack to the bypass. Barbels did not only seek contact with the channel walls, but also with the bars of the HBR. It seemed that they sensed the flow between the bars with their caudal fins, so that it was not uncommon that half of their body was between the channel bottom and the lowest bar of the HBR. Barbels which passed through the rack did so actively, primarily with positive rheotaxis (see example video 10 in supplementary materials), but sometimes they needed multiple attempts, especially if they were so large that they were hardly fitting between the bars of the HBR.

Reaction: A total of 74 barbels swam into the bypass, of which only 30 individuals interacted with the rack, corresponding to 30/74 ≈ 41% (Fig. 7). Neither U nor VR affected the response reactions of barbels significantly (bypass passages U: χ2 = 0.062, p = 0.804; VR: χ2 = 1.850, p = 0.174. refusals: U: χ2 = 0.580, p = 0.446; VR: χ2 = 0.000, p = 1.000. rack passages U: χ2 = 1.074, p = 0.300; VR: χ2 = 1.764, p = 0.184), which means that these parameters did not affect the response reactions of barbels or that the sample sizes were too small to detect these differences. The sample size of barbels for all configurations except E3 was especially small as many specimens were inactive (cf. Table 2) or entered the bypass without rack interaction and are therefore not shown in Fig. 7. Averaged over all test configurations =65% and =72%.

Fig. 7

Response reactions of barbel for different hydraulic configurations.

Nase (Chondrostoma nasus)

Behavior: Similar to spirlin, most nase swam downstream in zigzag movements, making use of the full channel width. Also most of the nase swam primarily together as a school and only a few specimens acted individually. Nase preferred to swim close to the bottom, were rarely guided by the HBR, and often passed through the rack the first time they approached it. If they were protected at their first approach, they swam another zigzag movement and passed the rack at another attempt. Nase primarily approached the HBR with positive rheotaxis, always kept a minimal distance to the rack, and never touched it.

Reaction: In total, 39 nase were tested of which 35 crossed line 2. Not a single specimen entered the bypass, neither with nor without rack interaction, while all nase passed the rack with the exception of one individual (E10 in Fig. 8). The majority of nase which passed through the HBR (76%) did not swim close to the bypass inlet (X = −0.22), such that the poor protection cannot be explained by the hydraulic conditions at the bypass. Although nase were tested with two different configurations only, it is clear that the HBR with s = 20 mm offered no sufficient protection. However, the poor protection of nase might not be a species-specific aspect, but was likely related to the small TL of the tested specimens (=71 mm; Table 2). The average guidance and protection efficiencies of the tested nase were =0% and =3%.

Fig. 8

Response reactions of nase for different hydraulic configurations.

Brown trout (Salmo trutta fario)

Behavior: In comparison to most other tested fish species, trout changed their rheotaxis very frequently. They were generally passive, either not swimming out of the acclimatization compartment or doing so only after some time. This can be at least partially explained by the natural behavior of trout, which often show sedentary behavior within flowing water. In the channel, they sometimes kept their position for several minutes, which was often followed by very fast movements. Trout preferred to swim close to the bottom, but it was observed several times that they quickly swam up to the water surface, before diving down to the bottom again. Most trout swam individually, while a schooling behavior was observed in a few experiments only. Trout used their caudal fins to sense the velocities between the bars and sometimes kept their position in front of the rack for several minutes. When they attempted to pass through the HBR, they did so with little hesitation, whereby especially the larger specimens needed multiple attempts to pass between the bars.

Reaction: Neither U nor VR affected the behavior of trout significantly at the HBR with s = 20 mm (bypass passages U: χ2 = 1.698, p = 0.193; VR: χ2 = 0.000, p = 1.000. refusals: U: χ2 = 0.854, p = 0.356; VR: χ2 = 0.001, p = 1.000. rack passages U: χ2 = 0.404, p = 0.525; VR: χ2 = 0.014, p = 0.907). The average guidance and protection efficiencies for trout with s = 20 mm were =51% and =58%, while with s = 15 mm they were FGE = 88% and FPE = 96% (E1 in Fig. 9). For the same hydraulic conditions (U = 0.5 m/s, VR = 1.2), a reduction of the clear bar spacing from s = 20 mm to s = 15 mm led to significantly more bypass passages (χ2 = 5.224, p = 0.022) and significantly less rack passages (χ2 = 7.972, p = 0.005).

Fig. 9

Response reactions of brown trout at the HBR with s = 15 mm (E1) and at the HBR with s = 20 mm for different hydraulic configurations.

Salmon parr (Salmo salar)

Behavior: Most salmon parr swam individually and showed frequent rheotaxis changes, which was also observed for trout. Salmon parr also preferred to swim close to the bottom and did not favor the channel walls. Almost all salmon parr swam out of the acclimatization compartment within the experimental duration and were thus more active than trout. No distinct schooling behavior was observed for salmon parr, but in some experiments, two individuals swam together. When salmon parr touched the HBR with their caudal fins, they often responded with fast but short avoidance reactions, before touching the rack again. Most individuals which passed the rack used their caudal fins to sense the flow between the bars and then quickly passed through the rack without hesitation.

Reaction: Not a single salmon parr refused the HBR-BS, leading to FGE = FPE (Fig. 10). The FGE of the tested configurations ranged between 19% and 35% but the difference was not significant (χ2 = 0.640, p = 0.424), leading to ==27%.

Fig. 10

Response reactions of salmon parr for different hydraulic configurations.

European eel (Anguilla anguilla)

Behavior: Most eels swam downstream with negative rheotaxis, which means they were swimming head-first with the current. They often collided with the HBR, and tried to squeeze head-first through the rack. During these active rack passage attempts, most eels were guided along the HBR and entered the bypass. Only a few individuals approached the rack either with positive rheotaxis or in passive drift. Eels which swam actively with negative rheotaxis acted insensitively to velocity gradients at the bypass inlet. The majority of eels acted solitarily, swam close to the bottom, and showed a strong thigmotactic positive behavior. The latter means that eels had contact with either a channel wall or the HBR most of the time.

Reaction: Only a few eels were available for the experiments, which led to too small sample sizes for statistical tests. Only one rack passage and one refusal were observed, leading to =87% and =93% (Fig. 11).

Effect of fish species, size distribution, and clear bar spacing

Fish size distribution and rack passages

In total, 128 rack passages were observed and analyzed regarding the fish size, fish orientation (rheotaxis, rotation), and vertical rack passage location. The majority of tested fish was small enough to physically pass through the HBR (w < s). The HBR with s = 20 mm was a physical barrier (w > s) to 0% of spirlin, 22% of barbel, 0% of nase, 2% of trout, 0% of salmon, and 47% of eel. With s = 15 mm, the HBR was a physical barrier to 64% of trout. Table 3 summarizes all rack passages and quantifies the percentage of individuals, which laid to their sides to pass through the HBR, which means that they rotated to the side over the longitudinal axis. Two videos of such rotations are provided as supplementary materials (videos 9 and 10). Additionally, the percentage of fish, which passed through the HBR with positive rheotaxis, is specified in Table 3. The majority of rack passages occurred close to the channel bottom. Weighted-averaged over all species, 77% of all individuals passed through the HBR between the channel bottom and the lowermost bar, 20% between the lowermost bar and the second bar (counted from bottom to top), and only 4% above the second bar (Table 3).

Table 3

Number of rack passages per fish species with (A) the corresponding percentage of fish which rotated during rack passage, (B) passed through the rack with positive rheotaxis, and (C) the percentage of rack passage location in vertical direction.

	Spirlin	Barbel	Nase	Trout	Salmon	Eel	Total
Number of rack passage [−]	7	13	34	43	30	1	128
A) Rotation [%]	14	38	18	37	13	0	25
B) Positive rheotaxis [%]	100	92	85	67	100	0	84
C1) Passage below bar 1 [%]	57	77	91	56	97	100	77
C2) Passage between bars 1–2 [%]	43	23	6	37	3	0	20
C3) Passage above bar 2 [%]	0	0	3	7	0	0	4

Effect of fish width on guidance and protection efficiencies

A comparison of and revealed large variations across species. While the HBR with s = 20 mm offered essentially no protection for nase with =71 mm (=3%), more than 90% of the tested spirlin and eels were protected. To assess if these variations resulted from the different species or from the different tested body size distributions and therefore different fish widths, and were plotted as a function of the ratio of the fish width to the clear bar spacing w /s (Fig. 12). A linear trend was observed between nase, salmon parr, trout, barbel, and eel (R2 = 0.98, blue circles in Fig. 12). A least squares regression analysis led to Eqs. (9), (10).

Fig. 12

(A) and (B) as a function of w/s for all tested fish species with s = 20 mm and additionally with s = 15 mm for trout (Trout*, red triangle); Spirlin* was additionally calculated with h/s instead of w/s (red diamond-shaped symbol); the dashed line in (A) shows Eq. (9) and in (B) Eq. (10). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

The application limits are 0.36 ≤ w/s ≤ 1 (Eq. (9), Fig. 12A) and 1/3 ≤ w/s ≤ 1 (Eq. (10), Fig. 12B), which means that FGE = 0% and FPE = 0% for w/s ≤ 0.36 and w/s ≤ 1/3, respectively. As there is no reason to assume a positive correlation between w/s and FGE for w/s ≥ 1, it is proposed to presume constant FGE for w/s ≥ 1, that is, FGE = 90%. Fish with w/s > 1 are physically protected by the HBR, such that FPE = 100% for w/s > 1. Eq. (10) suggests that fish which are not physically blocked by the HBR (w/s < 1) are protected by a partial behavioral protection effect, which increases with increasing w/s. The vertical error bars in Fig. 12 mark the minimal and maximal FGE and FPE for the different tested configurations, while the horizontal error bars indicate w ± σ.

Spirlin were not considered for determining Eqs. (9), (10), because and were considerably larger than for all other species (red squares in Fig. 12). In contrast to all other species, the reactions of spirlin significantly varied between configurations (cf. Section 3.2), which led to the large error bars in Fig. 12. It is very likely that the better protection of spirlin is a species-specific result. Fig. 12 is based on the assumption that the fish width w and not the fish height h limits rack passages. When w/s is replaced by h/s for spirlin in Fig. 12 (red diamond-symbol), and correspond well with the prediction equations of the other fish species, indicating that the fish height h might limit rack passages for most spirlin. This assumption is supported by the detailed analysis of the rack passages, which revealed that spirlin hardly rotated to the side to pass through the HBR (cf. Table 3). The FGE and FPE of trout with s = 15 mm is shown with red triangles in Fig. 12, which also fits well to the proposed equations. The correlation between w/s and and in Fig. 12 indicates that it is possible that the differences in and between all tested species, except for spirlin, is mainly a result of the different tested size distributions.

Logistic regression models

Up to now, the effects of U and VR on FGE and FPE were assessed for each species individually, while the overall effect across species is assessed in this section with GLMs. The output of the GLMs considering the bypass passages, rack passages, and refusals as the dependent variable are summarized in Table 4.

Table 4

Regression coefficients β with the corresponding standard errors in parenthesis of the three GLMs with the dependent variables bypass passage, rack passage, and refusal; the reference categories are U = 0.5 m/s, VR = 1.2, and time = morning; ⁎ denotes p < 0.05, ⁎⁎p < 0.01, and ⁎⁎⁎p < 0.001; N = number of fish used in the model, AIC = Akaike Information Criterion, BIC=Bayesian Information Criterion.

	M1: bypass	M2: rack	M3: refusal
Intercept	1.797⁎⁎⁎	−2.427⁎⁎⁎	−2.446⁎⁎⁎
	(0.389)	(0.423)	(0.717)
Uo: 0.7 m/s	−0.210	−0.160	1.286⁎⁎
	(0.262)	(0.275)	(0.473)
VR: 1.4	−0.892⁎⁎⁎	0.596⁎	1.216⁎
	0.235	(0.242)	0.491
Usage	−0.339⁎⁎⁎	0.558⁎⁎⁎	−0.875⁎⁎⁎
	0.103	(0.111)	(0.251)
Time of day: afternoon	−0.730⁎⁎	0.566⁎	1.039⁎
	(0.246)	0.256	(0.472)
N	327	327	327
Log-likelihood	−210.409	−199.312	−68.304
AIC	430.817	408.624	146.609
BIC	449.767	427.574	165.559

The probability of a specific fish reaction can be calculated from the regression coefficients β for different configurations. The intercept represents the standard configuration, where all parameters are set to their reference categories, that is, U = 0.5 m/s, VR = 1.2, usage = 1, and time of the day = morning. In this case, the odds ratio of a bypass passage is OR = exp(β0) = exp(1.797) = 6.032 (Eq. (6)), which is equal to a probability of p(X) = OR/(1 + OR) = 86% (Eq. (8)). The approach flow velocity U did neither affect the bypass passages nor the rack passages significantly, but an increase of U significantly increased the odds of refusals by OR = exp(1.286) = 3.618 (Table 4). Only 8% of all fish refused the HBR-BS with the standard configuration (OR = exp(−2.446) = 0.087; p(X) = 0.087/(1 + 0.087) = 8%). For an approach flow velocity of U = 0.7 m/s, 24% of all fish refused the HBR-BS (OR = exp(−2.446 + 1.286) = 0.313; p(X) = 0.313/(1 + 0.313) = 24%). In other words, the rather small probability of a refusal for the standard configuration with U = 0.5 m/s of 8% increased to 24% for U = 0.7 m/s. This trend was observed for all species which were tested with U = 0.5 m/s and U = 0.7 m/s, namely spirlin, barbels, and trout (Fig. 6, Fig. 7, and Fig. 9). However, the effect was not significant on the species-specific analysis, as the species-specific sample sizes were rather small compared to the overall sample size. An increased velocity ratio VR led to significantly less bypass passages, but more rack passages and refusals (Table 4). For the standard configuration, an increase from VR = 1.2 to VR = 1.4 reduced the probability of a bypass passage from 86% to 71%, while the rack passages increased from 8% to 14% and the refusals from 8% to 23%. This trend was also observed for multiple species, but it was only significant for bypass passages and refusals for spirlin (cf. Section 3.2). The secondary parameters “usage” and “time of the day” were only introduced to enhance the model fit and should therefore not be interpreted.

Sector analysis

In this section, the effect of the different hydraulic configurations on the behavior of spirlin is analyzed with the sectors defined in Fig. 3. Similarly, differences across species are assessed for the configuration with U = 0.5 m/s and VR = 1.2.

Without a rack installed (E0), R≈R in Fig. 13, which indicates that spirlin were equally distributed between the right (Sec2) and the left channel walls (Sec1). They swam quickly through the rack plane (Sec5) and hardly entered the bypass (Sec6; cf. Fig. 6). For the same hydraulic conditions but with an HBR installed (E2), spirlin spent much less time in the upstream area which was unaffected by the HBR (Sec1-Sec3), while they were swimming often in front of the HBR (Sec5) and in the bypass (Sec6). These results point out the excellent guidance efficiency of this configuration (E2, cf. Fig. 6). With an increased VR (E8), spirlin spent more time searching for a downstream passage corridor in front of the rack (Sec5), while they spent less time in the bypass (Sec6). With constant U, an increased VR led to a larger probability of fish swimming along the right wall upstream of the bypass (E2 vs. E8 and E14 vs. E17; Sec2 and Sec7).

Fig. 13

Normalized residence coefficient R of spirlin for sectors Sec1 to Sec7 and some of the configurations defined in Table 2.

Across species, no clear trend could be identified for the use of the sectors close to the left (Sec1) and right (Sec2) channel walls (Fig. 14). However, eels often left the acclimatization compartment, which was installed on the left channel side (cf. Fig. 2), with negative rheotaxis, such that they subsequently spent more time close to the left channel wall. Eels, which are known for their thigmotactic positive behavior, avoided the sectors 3 and 4 in the channel center. For all fish species, the residence coefficient upstream of the rack R was larger than the residence coefficient of most other sectors, which means that most fish perceived the rack and spent comparably much time in front of it. Spirlin spent much time in sector 6, meaning that they were guided efficiently into the bypass (cf. Fig. 6). On the contrary, not a single nase entered the bypass, which led to R = 0 (cf. Fig. 8). R > R for spirlin, indicating that spirlin reacted sensitively to the flow acceleration at the bypass inlet. For barbel and trout, R ≈ R, suggesting that the downstream movements of these species were hardly slowed down by the acceleration at the bypass inlet. Similarly R ≈ R for nase, but not a single nase entered the bypass (cf. Section 3.2.3), indicating that nase which approached the bypass were not slowed down, but swam towards the rack which they subsequently passed. It has to be assumed that the bypass acceleration would have slowed down nase (R > R), if they were well protected by the rack.

Fig. 14

Normalized residence coefficient R of all tested fish species for sectors Sec1-Sec7 for the configuration U = 0.5 m/s and VR = 1.2.

Discussion

HBR-BSs are a state-of-the-art fish downstream passage technology (Ebel, 2016), but little was known about the swimming behavior as well as guidance and protection efficiencies of different fish species under various flow conditions. Up to now, it was also not possible to estimate the fish guidance and protection efficiencies of HBR-BSs for differently sized fish. The present study demonstrates that HBR-BSs, which are designed according to current design guidelines (e.g. Ebel, 2016; Meister, 2020), are well suited to protect downstream moving fish and to guide them to a bypass. The present study also demonstrates that fish of various species – at least to a certain degree – rotate to the side to pass through HBRs, such that fish are only physically blocked if w > s. Although variations were observed between different species and hydraulic conditions, the results of the present study suggest that the guidance and protection efficiencies primarily depend on the ratio of the fish width to the clear bar spacing w/s. Besides the physical barrier effect (w > s), HBRs also cause a partial behavioral guidance effect. The protection and the guidance efficiency increased linearly with w/s. Large fish with w/s > 1 were physically protected by the HBR, while no protection was observed for very small fish (w/s < 1/3). Quantifying the fish protection efficiency is of high relevance as it allows to better assess the effect of the implementation of HBR-BSs at hydropower plants. The presented equations allow for estimating the fish protection efficiency for a population with a given size distribution. In combination with turbine survival estimates it is now possible to estimate the total survival rate of downstream moving fish at a hydropower plant for the present situation and for the situation after the implementation of an HBR-BS (details in Meister, 2020).

Comparison with other relevant laboratory studies

The flow velocity measurements reported in the present study show that HBRs without overlays hardly affect the velocity fields, which is in line with previous studies (Meister et al., 2020a). It was also shown that the turbulent kinetic energy TKE was small in the entire area upstream of the HBR and within the bypass, which agrees with the findings of Szabo-Meszaros et al. (2018).

When the results of different live fish laboratory studies are compared to each other, it is crucial to carefully study the experimental setup, methods, and data analysis. An important point is to distinguish between fish which interacted with the rack and which did not. Thereby, passive fish which stayed in the headwater throughout the whole experiment can be distinguished from fish actively refusing the rack. This is especially important for inactive species such as trout. Additionally, if fish entered the bypass without rack interaction, it should be acknowledged that they were not guided by the rack. In contrast to the present investigation, it was not distinguished between bypass passages with and without rack interaction in the studies of Berger (2018) and Kammerlander et al. (2020). This may not be important for some species, while it can strongly affect the results of other species such as barbel, which preferred to swim along channel walls in the present study (cf. Section 3.2.2).

The results reported by de Bie (2017) and de Bie et al. (2021) lead to FGE = 52% and FGE = 70% for barbel and chub, respectively (cf. Section 1.4). For the parameters reported by de Bie et al. (2021) and the fish biometry by Schwevers and Adam (2020), the fish guidance efficiencies predicted with Eq. (9) are FGEpred = 62% and FGEpred = 65% for barbel and chub, respectively. Considering all uncertainties and differences between both studies, these values agree well. The validation with the study of de Bie et al. (2021) therefore indicates that Eq. (9) can be used for other fish species such as chub, HBRs with smaller s, and fish with smaller w. None of the fish tested by Bie et al. (2021) could pass through the HBR in regular swimming position as h > s. This supports the findings of the present study, where at least some specimens of all tested species except eel rotated to the side to pass through HBRs (cf. 1.4, 3.3.1). The HBR investigated by de Bie et al. (2018) was a physical barrier to all fish (w > s), such that FPE = 100%, which agrees with the present study. De Bie et al. (2018) also highlighted that abrupt velocity gradients triggered avoidance reactions. Similar to the present study, larger U increased the number of refusals.

While the vast majority of salmon was protected by the HBR even with s = 30 mm in the study of Berger (2018) (FPE = 95%; cf. Section 1.4), FPE was 27% for s = 20 mm for salmon in the present study (cf. Section 3.2.5). These variations can be partially explained by differences in the biology and handling (e.g. wild fish vs. hatchery-reared), hydraulics, and the experimental setup, which were assessed in detail in Meister (2020). The most important differences are therefore discussed only in the following. In the experiments of Berger (2018), the lowest 5 cm of the HBR were blocked by a beam, which acted as a bottom overlay. Berger (2018) highlighted that most salmon smolts preferred to swim close to the bottom or in areas with reduced flow velocities, that is, in front of the bottom overlay. It is therefore very likely that the bottom overlay strongly affected the number of rack passages in the study of Berger (2018). It is known from other studies, such as EPRI; DML (2001) and Albayrak et al. (2020), that bottom overlays significantly increase FGE in live fish laboratory studies at mechanical behavioral barriers. In the present study, all salmon parr passed through the HBR below the second lowest bar, which is equal to 50 mm above the bottom. It must therefore be assumed that the high guidance efficiencies of salmon smolts in the study of Berger (2018) were at least partly caused by the bottom overlay and not the HBR itself. Although HBRs are built with bottom overlays in prototype, the fish swimming depth cannot be directly transferred from laboratory to prototype (cf. Section 1.1). Especially in larger rivers, salmon smolts are known to migrate downstream close to the water surface (Thorstad et al., 2012; Ebel, 2016), such that they can hardly be effectively guided with bottom overlays. Another difference between the two studies is that Berger (2018) investigated HBRs with rectangular bars, while foil-shaped bars were used in the present study. It is difficult to assess the effect of the bar shape, because, to the best of our knowledge, studies are missing, that take into account the effect of the bar shape on the guidance efficiency, while keeping all other parameters constant. Rectangular bars lead to larger flow separation zones along the bars (Kirschmer, 1925; Meister et al., 2020b), which might trigger avoidance reactions and thereby increase fish protection. Additionally, the sharp edges and the long constricted section of rectangular bars potentially make rack passages more difficult. It is therefore possible that the differences between the present study and the investigation of Berger (2018) were also partially related to variations of the bar shape.

Kammerlander et al. (2020) demonstrated that the FPE of the flexible fish fence with s = 10 mm is almost 100% for trout, chub, and grayling with TL = 100–200 mm. Similarly, in the present study, FPE was 96% for trout with =158 mm at the HBR-BS with s = 15 mm (Section 3.2.4). It is more difficult to compare the results with s = 20 mm, because all fish which neither passed through the flexible fish fence nor entered the bypass were classified as “headwater” in the study of Kammerlander et al. (2020). It remains unknown how many fish swam close to the rack and showed avoidance reactions triggered by the rack itself. Instead of comparing FGE and FPE with different definitions, the percentage of all bypass passages from all bypass and rack passages is compared in the following. For trout with s = 20 mm, this percentage was 58% in the present study, while it was 25% in the study of Kammerlander et al. (2020). The median total length of trout tested with s = 20 mm was 135 mm in the present study, compared to the median of 154 mm in Kammerlander et al. (2020). It remains unclear if these differences are a result of the different rack type or variations in the experimental setup, but both studies found that barriers with horizontal bars with s = 20 mm cannot sufficiently protect trout with TL ≈ 150 mm. Similar to the present study, Kammerlander et al. (2020) described that most rack passages occurred close to the channel bottom. Although not explicitly mentioned in Kammerlander et al. (2020), from the fish dimensions and s, multiple fish of all tested species must have turned to their sides to pass the flexible fish fence, which can also be observed in the videos of the online supplement of Kammerlander et al. (2020).

Beck (2020) and Beck et al. (2020) tested recently developed curved-bar racks (CBR) with vertical bars. Their experiments were conducted in parallel to the present study, in the same laboratory channel with the same fish species and individuals, allowing for a direct comparison of the two rack types. With the CBR-BS, =79% for spirlin, =76% for barbel, =85% for nase, =38% for trout, =83% for salmon parr, and =27% for eel (Beck, 2020; Beck et al., 2020). The differences between both rack types were comparably small for spirlin, barbel, and trout. was higher with the CBR compared to the HBR for salmon (=83% vs. =27%) and nase (=85% vs. =0%). The result with nase is particularly interesting, as the tested individuals were small (TL < 100 mm) and the clear bar spacing at the upstream tip of the bars was s = 20 mm for the HBR and s = 50 mm for the CBR. This comparison shows that the behavioral protection effect of the CBR was larger than for the HBR for juvenile nase. The main difference between the two rack types is that the CBR induces stronger hydraulic cues in between the bars and slightly upstream of the rack (Beck et al., 2020). Juvenile nase seem to react very sensitively to these different flow regimes. In contrast to the HBR, eels were hardly protected by the CBR (=27%; Beck et al., 2020) compared to the HBR (=87%). This indicates that most eels were physically protected by the HBR and that eels did hardly react to the hydraulic cues of the CBR.

Limitations and challenges when upscaling to prototype

The reasons for the knowledge deficit related to HBR-BSs are the complexity of the problem; the large effort to study fish behavior on-site at HPPs and in live fish laboratory experiments, and the high cost of the field and laboratory equipment to obtain the required data. Within the present study, these research gaps are partially filled with knowledge gained from live fish laboratory experiments. However, the results of such experiments have to be carefully interpreted and some cannot be directly transferred from laboratory to prototype due to abiotic and biotic factors, which could not be mimicked with the current channel setup. This includes variations in environmental effects (e.g. turbidity, floating debris, changes in light conditions, unsteady discharge, and water temperature), diurnal rhythms, differences in the fish biology (e.g. stress from experimental handling and unnatural environment), and geometric restrictions (e.g. dimensions of channel, HBR, and bypass).

The HBR of the present study was only 2.3 m long. It is possible that the rack passage risk increases at longer racks, meaning that FGE and FPE might be smaller at longer HBRs than in the present laboratory study. The upstream channel width was limited to w = 1.50 m, meaning that the channel walls had a disproportionately high effect. It was shown that barbels swam primarily along the channel walls and therefore often had no rack contact in the present laboratory tests, which raises the question if barbel are a suitable species to assess the guidance efficiency of a rack in laboratory experiments. The flow depth was limited to h = 0.90 m, which means that fish which swam close to the bottom in the laboratory experiments, do not necessarily swim close to the river bottom at HPPs with larger approach flow depths. This may especially be the case for trout and salmon. As an example, the majority of rack passages occurred near the channel bottom, indicating that the transition from the bottom to the HBR is very important. Although it is likely that the guidance efficiency of fish guidance racks can be increased with bottom overlays (EPRI; DML, 2001; Albayrak et al., 2020), it should not be concluded from the present experiments that fish pass through all HBRs close to the bottom only.

The relative bypass discharge varied between 13% ≤ Q ≤ 17% in the present study and was therefore significantly larger than 2% ≤ Q ≤ 5%, which is typically recommended for HBR-BSs (cf. Section 1.2). However, we deem the bypass velocities at the inlet, which in the present study were kept similar to prototype conditions, more important than Q. Due to size limitations of the channel, Q could not be further reduced, without either narrowing the bypass, which would have made it unsuitable as a downstream passage corridor, or reducing the velocity ratio VR which is essential for fish guidance. The results can nonetheless be transferred to prototype situations, if the flow conditions are similar.

The present live fish experiments were conducted during daytime, while it is known that eels primarily migrate from dusk till dawn (Calles et al., 2012; Egg et al., 2017). Similarly, freshwater species, including barbel and brown trout, showed higher downstream movement activity at dawn and during the first hours of the night in monitoring campaigns (Adam et al., 2018; Zaugg and Mendez, 2018). Infrared or near-infrared light with typical wavelengths of 840–1200 nm, which are considered beyond the visible range of fish (Beach, 1978), was used in multiple experiments to study the fish behavior in the darkness (e.g. Beach, 1978; Kemp and Williams, 2009; Russon et al., 2010; Vowles and Kemp, 2012). However, as water absorbs infrared light stronger than normal light (Beach, 1978), these studies are often limited to small flow depth of h < 30 cm. Due to the missing visual cues, it is possible that the guidance and protection efficiency of HBR-BSs is smaller when fish are exposed to near darkness.

The proposed equations (Eqs. (9), (10)) have to be used with caution due to the discussed limitations, especially if conditions deviate strongly from the conditions in the present experiments. For example, FGE and FPE are expected to be significantly lower if no suitable bypass is available. If FGE and FPE are estimated with the presented equations, it is strongly recommended to use w/s. However, for some species such as spirlin, it might be permissible to replace w/s by h/s. The live fish laboratory tests conducted within the present work give valuable insights in the fish behavior at HBR-BSs, but they cannot replace extensive monitoring campaigns due to the discussed limitations.

Recommendations for future HBR-BS research

The Eqs. (9), (10) were developed for one size range per species. It is recommended that these equations are validated in future studies, where different size ranges per species are tested. Further experiments should be conducted for a wider parameter range, including very low approach flow velocities (Uo ≈ 0.2 m/s), night experiments, and different bar shapes. Besides rectangular and foil-shaped bars, it is recommended to investigate other shapes (e.g. similar to a tilde), which would induce larger turbulences and flow deflections. This would likely trigger avoidance reactions and would increase the partial behavioral avoidance effect. Future live fish studies should also focus on the flow conditions in the bypass, where the bypass is ideally modeled on a scale of 1:1. Last but not least, it is vital to conduct extensive monitoring campaigns at state-of-the-art HBR-BSs, where the number of bypass and rack passages are simultaneously quantified.

Conclusions

The hydraulics and fish guidance and protection efficiencies of horizontal bar rack (HBR) bypass system were systematically investigated in a large laboratory channel. The tested rack configurations had foil-shaped bars with a clear bar spacing of 15 and 20 mm, a horizontal rack angle of 30° to the flow direction, and a full depth open channel bypass. The live fish tests were conducted for four hydraulic conditions with six different fish species, namely spirlin, barbel, nase, brown trout, Atlantic salmon parr, and European eel, leading to the following key findings:

The HBR hardly affected the velocity field, leading to little turbulent kinetic energy and small spatial velocity gradients.

The bypass affected the velocity field only locally. Turbulent kinetic energy and the spatial velocity gradient in the full depth open channel bypass were small and flow velocities increased only slightly at the bypass inlet.

A moderate to large proportion of most fish species, including eel, were protected and efficiently guided along the HBR to the downstream rack end.

Some individuals of all tested species (except for eel) rotated to the side to pass between the bars. Hence, fish can only be fully protected, if both their body width w and height h are larger than the clear bar spacing s.

Besides this physical barrier effect, smaller fish of most species were partially guided into the bypass, suggesting that HBRs act as behavioral barriers to a certain extent. This partial behavioral guidance effect was quantified as a function of the ratio of the fish width to the clear bar spacing w/s.

The fish guidance and protection efficiencies varied across species, which for the most part can be explained by the ratio of the fish width to the clear bar spacing w/s for nase, salmon, trout, barbels, and eels. Given identical w/s, spirlin were better protected than all other species.

Considering all fish species, the approach flow velocity did neither affect the number of bypass passages nor rack passages for 0.5 m/s ≤ U ≤ 0.7 m/s, but it significantly increased the number of refusals. For 1.2 ≤ VR ≤ 1.4, an increased ratio of the bypass inlet to mean approach flow velocity led to significantly less bypass passages, but more rack passages and refusals. This is because many fish of some species hesitated to enter the bypass for larger VR, which increased the rack passage risk as they actively searched for an alternative downstream passage corridor. A suitable bypass at the downstream rack end is therefore essential to reduce the rack passage risk.

Fish passed through the HBR primarily within the first few bars above the bottom with an approach flow depth of h = 0.90 m. The transition from the HBR to the channel bottom is therefore crucial to avoid model effects in live fish laboratory experiments.

The findings of the present study contribute to a better understanding of the fish behavior at horizontal bar rack bypass systems. The proposed equations can be used to estimate the guidance and protection efficiencies of these state-of-the-art fish downstream passage facilities and thereby contribute to a sustainable use of hydropower. Nevertheless it is indispensable to conduct extensive monitoring campaigns to validate the proposed equations and to verify functional efficiency.

The following are the supplementary data related to this article.

Supplementary Video 1

Typical fish swimming behavior of spirlin at horizontal bar rack bypass systems.

Supplementary Video 2

Typical fish swimming behavior of barbel at horizontal bar rack bypass systems.

Supplementary Video 3

Typical fish swimming behavior of nase at horizontal bar rack bypass systems

Supplementary Video 4

Typical fish swimming behavior of trout at horizontal bar rack bypass systems

Supplementary Video 5

Typical fish swimming behavior of salmon parr at horizontal bar rack bypass systems.

Supplementary Video 6

Typical fish swimming behavior of European eel at horizontal bar rack bypass systems; Cam 1

Supplementary Video 7

Typical fish swimming behavior of European eel at horizontal bar rack bypass systems; Cam 2

Supplementary Video 8

Rack passage of an European eel at a horizontal bar rack

Supplementary Video 9

Rotation of a trout during rack passage at a horizontal bar rack

Supplementary Video 10

Rotation of a barbel during rack passage at a horizontal bar rack

Short explanation of the video files

Details of the fish sampling locations

Ethics

All live fish tests conducted within the present study met the ethical guidelines and legal requirements (Swiss animal welfare act) under permission from the canton of Zurich and the veterinary office (animal experimentation licenses No. 30383 and 31,339; laboratory animal husbandry license No. 180).

Funding

This work was supported by research and innovation program FIThydro (Fishfriendly Innovative Technologies for Hydropower) [grant number 727830] and the Swiss (SERI) [grant number 16.0153].

CRediT authorship contribution statement

Julian Meister: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing, Visualization, Project administration. Oliver M. Selz: Conceptualization, Methodology, Software, Validation, Investigation, Resources, Writing – review & editing, Supervision. Claudia Beck: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing – review & editing. Armin Peter: Conceptualization, Methodology, Resources, Writing – review & editing. Ismail Albayrak: Conceptualization, Methodology, Writing – review & editing, Supervision, Project administration, Funding acquisition. Robert M. Boes: Conceptualization, Methodology, Resources, Writing – review & editing, Supervision, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.