Protection and Guidance of Downstream Moving Fish with Electrified Horizontal Bar Rack Bypass Systems

Abstract

Horizontal bar rack bypass systems (HBR-BS) are characterized by a horizontal bar rack (HBR) with narrow clear bar spacing of 10–20 mm and an adjacent bypass (BS) to efficiently protect and guide downstream moving fish at water intakes. The small bar spacing may lead to operational challenges, such as clogging and high head losses. This study investigated whether combining an HBR with a low-voltage electric field (e-HBR) allows one to increase the clear bar spacing while maintaining a high standard of fish protection and guidance efficiency. To this end, an HBR-BS with 20 mm bar spacing and an e-HBR-BS with 20 and 51 mm bar spacing were tested with spirlin (Alburnoides bipunctatus) and European eel (Anguilla anguilla) in a laboratory flume. The racks were electrified with 38 V pulsed direct current. The protection efficiency of the e-HBR with 51 mm was 96% for spirlin and 86% for eels, which are similar results to those of the HBR with 20 mm. Some eels passed through the e-HBR, but only when they were parallel to the rack. Fish injuries of variable severeness due to the electrification were observed. The results highlight the potential of hybrid barriers for the protection of downstream moving fish. However, fish injuries due to electricity may occur; and reporting applied voltage, electrode geometry, resulting electric field strength and the pulse pattern of the electrified rack setup is necessary to ensure comparability among studies and to avoid injuries.

1. Introduction

1.1. Protection and Guidance of Downstream Moving Fish

Fish move upstream and downstream within river systems for various reasons during their life cycles, such as finding suitable habitats for spawning or overwintering [1]. During downstream movements, they can incur severe or even lethal injuries when passing through hydropower plant (HPP) turbines [2] or when they are entrained at other water intakes. This has implications not just on the individual level but on the population and species levels [3]. For this reason, free movement of fish both in the downstream and upstream directions has to be restored by law in the European Union (see the European Water Framework Directive) and in Switzerland (see the Swiss Federal Waters Protection Act). Measures for protecting and bypassing downstream moving fish are typically classified as physical barriers (e.g., horizontal bar racks), mechanical behavioral barriers (e.g., louvers, vertical-oriented bar racks), sensory behavioral barriers (e.g., electric barriers, acoustic barriers, bubble screens), collection systems and so-called fish-friendly turbines and operations [4,5,6,7]. All of the barriers have to be combined with a bypass system (BS) to provide longitudinal connectivity. The main function of the BS is to attract, safely collect and transport fish and to return them undamaged to the river downstream of an obstacle. The main types of BS are: surface bypasses, bottom bypasses, both surface and bottom bypasses and full depth open channel bypasses, depending on where the entrance is situated [8].

Horizontal bar racks (HBR) are characterized by a small clear bar spacing of typically 10 mm ≤ s_b ≤ 20 mm, such that they are physical barriers to many fish [8,9]. Since the installation of the first horizontal bar rack bypass system (HBR-BS) with a specific focus on fish protection in 2001, HBR-BSs have become a state-of-the-art technology in central Europe for protecting and bypassing downstream moving fish [8,9]. The German [10] and Swiss Guidelines [11] for downstream migration recommend HBR for small-to-medium sized HPP and require a bar spacing of ≤20 mm. In most rivers, the main operational challenge of HBR-BSs with such small bar spacings is clogging by floating debris, which can significantly increase the head losses caused by HBRs [9,12]. The clogging probability at HBRs strongly depends on the clear bar spacing s${}_b$ [9], so that a large clear bar spacing is preferable from an operational point of view. It also reduces the impingement risk for fish, especially for small-sized fish with limited swimming capabilities [8]. However, recently conducted live fish tests with HBR-BSs involving a diverse assemblage of riverine fish species revealed that fish can only be fully protected with HBRs when the clear bar spacing is smaller than the fish width instead of the fish height as previously assumed [9,13].

Graduated field electric barriers are widely used in the US and well suited to prevent upstream movement of invasive species and to keep fish from entering dead ends, such as tailrace channels [14,15,16]. The use of electric barriers for safe downstream passage is more challenging. Whereas the flow direction works in favor of upstream barriers, fish will be swept through the area from which they are meant to be excluded if electric barriers cause immobilization to downstream moving fish [15]. Generally graduated field electric barriers for downstream blockage were found to be effective only for approach flow velocities <0.2 m/s [17,18]. This is much lower than approach flow velocities generally present at HPPs.

One challenge of electrical barriers is that large fish react to lower voltage gradients and are more susceptible to injuries than small fish due to the larger voltage differences over the lengths of their bodies (details in Section 1.2). This has been shown at a specific electric barrier, where small fish with total lengths of 3.5 cm ≤ TL ≤ 4.0 cm were protected but required large voltage gradients which were lethal for large fish [19] (as cited in [20]). Therefore, for upstream movement, graduated electric fields are used, such that fish can react to electric stimuli at different locations, depending on their perceptions of the voltage gradient. If graduated electric fields are used for downstream passage, it is very challenging to guide fish to a certain location such as a bypass, since this requires that all fish react to the electric field in the same way.

To guide downstream moving fish with electricity, the electric field has to be strong enough to trigger avoidance reactions, but injuries, narcosis and tetany should be prevented for all species and sizes [15]. A promising approach is to guide downstream moving fish with hybrid barriers [15], where a fish guidance rack is combined with an electric field. To the best of our knowledge, such a hybrid barrier, consisting of an HBR and a low-voltage electric field (e-HBR), has not yet been installed at any HPP. However, in recent years, several laboratory-based studies have been conducted with such a hybrid barrier or similar systems. Berger (2018) [21] tested several e-HBR configurations with European eel (Anguilla anguilla) in a 2 m wide laboratory channel with a 0.2 m wide bypass opening. The tested e-HBRs each consisted of rectangular bars that acted as the anode, and a metal mesh installed approximately 30 cm downstream of the e-HBR used as the cathode [21]. The NEPTUN system of the Procom company [18] was used to generate gated burst pulsed direct current (gpDC). At the HBR with $s_b$ = 30 mm, 15% of the N = 188 tested eels (40 cm ≤ TL ≤ 80 cm) passed through the HBR [21], which is referred to as the rack passage rate. With the electric field, two out of N = 190 eels passed through the rack, leading to a rack passage rate of only 1% [21].

Tutzer et al. (2021) [22] conducted live fish tests with the “fish protector”, a fish protection system using horizontally arranged steel cables. These were electrified with 80 V and gpDC. Brown trout (Salmo trutta), rainbow trout (Oncorhynchus mykiss), grayling (Thymallus thymallus) and chub (Squalius cephalus) were tested. Averaged over the horizontal approach flow angles $\alpha$ = 20° and 40° and all tested fish species, the mean rack passage rate was 5% for $s_b$ = 30 mm and 12% for $s_b$ = 60 mm without electrification of the cables. In contrast, for the electrified "fish protector" with $s_b$ = 30 mm and 60 mm, a mean rack passage rate of 2% was achieved.

Within a recent field study, a pumping station on the Danube River, Germany, was equipped with a “fish protector” with 50 mm gaps, and the fish behavior was observed with a sonar system for three days [23]. At the non-electrified fish protector, 3% of the fish that approached the rack passed through it, and at the electrified fish protector, 7% of the approaching fish passed the barrier. This seems to imply that the protection is not improved by electrification, which contradicts the findings of [22]. However, the authors considered the turning behavior or direction a fish kept swimming after approaching the rack a more relevant measure for their setup. In the non-electrified state, all fish swam parallel to the rack or slightly upstream after approaching it. In contrast in the electrified state, 72% of the fish turned away from the rack and swam back upstream. This indicates its suitability for fish protection, as the fish are less likely to pass the rack at a later time. However, the results cannot be generalized as the study was limited to a small sample size, did not distinguish between species-specific reactions and was limited to very low approach flow velocities of $U_o$ = 0.05 m/s.

To date, only guidance and protection efficiencies have been quantified in laboratory and monitoring-based studies of electrified hybrid barrier systems. Yet, it is essential that such studies also focus on possible electrification-related injuries that may be directly or indirectly lethal, and on a possible increase in the vulnerability to predation [24].

1.2. Fish Response to Electric Stimuli

The fish behavior in electric fields depends on various factors, such as field strength, water conductivity, fish conductivity, biometry, species, fish size, type of current, pulse length, pulse frequency and pulse pattern. An intuitive measure for fish reactions accounting for both the orientation of the fish within the electric field and the fish length is the “body voltage” $\Delta U_e$, which is the voltage potential over the fish’s length. If a fish is aligned perpendicularly with the equipotential lines (lines with a constant voltage), the body voltage will be large, and the body voltage will be small if a fish is aligned parallel to the equipotential lines [15]. Due to the high power demand of direct current and the large risk of injury of alternating current, electric barriers are electrified using pulsed direct current (pDC) either in the form of regular pulses repeated with a given frequency or gated burst pulsed direct current (gpDC). The latter are high-frequency pulse groups, which are repeated at regular intervals. Gated bursts lead to high catch efficiencies in electrofishing and require less power than pDC due to the intervals between the pulse groups. However, the level of injury is variable depending on the pulse pattern [15]. The common behavior zones of a DC electrical field can be classified according to Beaumont (2016) [15] as:

• Indifference zone, where fish behavior is not influenced by the electric field.
• Repulsion zone, where fish react by escaping or seeking refuge.
• Attraction zone, where fish are approaching the anode either due to anodic taxis (electric field stimulates the central nervous system and/or the muscles) or forced swimming (effect of the electric field on the autonomic nervous system).
• Narcosis zone, where fish are immobilized but their muscles are relaxed, and the fish can breath.
• Tetanus zone, where fish are immobilized, muscles contract strongly and breathing stops. It may take a fish several minutes to recover from this state, and this zone may lead to severe injuries.

Most knowledge about thresholds for fish reactions and suitable pulse patterns in electric fields is based on electrofishing studies (e.g., [15,24]), where galvanotaxis and fast narcosis are desirable. However, for downstream electric barriers, immobilization should be avoided. For pDC, the attraction and tetanizing voltage gradient thresholds found in the literature are 0.1 V/cm $\le E\le$ 0.2 V/cm and 0.5 V/cm $\le E\le$ 0.6 V/cm, respectively, where E is the electric field strength or voltage gradient. These thresholds may vary greatly depending on the fish species [15,25]. In moderate water conductivity (250–500 μS), voltage gradients of 0.1 V/cm $\le E\le$ 1.0 V/cm are considered suitable for electrofishing [15]. If not designed properly, fish exposed to electric fields may suffer severe injuries, which result from powerful convulsions of the musculature [24]. Potential injuries include spinal fractures and hemorrhages, which are often not externally obvious. To reduce the risk of injuries, it is generally recommended to use low frequency pDC (≤30 Hz) and large diameter electrodes [24]. At least one gpDC pattern was found to cause less injuries than higher frequency pDC [24]. Furthermore, the risk of injuries depends on the exposure time, length of the fish and the fish species [24].

To the authors’ knowledge, systematic studies on the reactions of fish to different waveforms and parameters of electric currents have only been undertaken for capture prone responses (anodic taxis or immobilization) targeted at electrofishing. Bird and Cowx (1993) [26] found frequencies of approximately 100 Hz to be most effective for trout and carp. Briggs (2019) [27] studied capture prone responses for grass carp (Ctenopharyngodon idella). It was shown that water temperature, frequency, pulse width and fish standard length all had significant effects on the threshold voltage gradient for immobilization. pDC with 60 Hz and a 4 ms pulse width was the most effective waveform (requiring the lowest voltage gradient) for immobilization of juvenile carp. For a frequency of 60 Hz and water temperatures of 12–25 °C, the immobilization threshold varied between 0.25 and 0.45 V/cm. The transferability of these results to electrified bar rack systems for downstream passage is difficult, because a low threshold for immobilization is not desirable. Furthermore, as highlighted by Snyder (2003) [24], lower frequencies than recommended by Briggs (2019) [27] and Bird and Cowx (1993) [26] should be used to reduce the risk of injuries. This is even more important for permanently installed electric barriers than for electrofishing. These findings highlight the effect of waveform parameters on fish behavior and the lack of knowledge of suitable waveforms for electric barriers for various European fish species.

1.3. Goals of the Study

The main goal of the present study was to assess whether the combination of an HBR-BS with a low-voltage electric field allows for increasing the clear bar spacing from the typically recommended 20 mm to 51 mm to reduce the clogging and impingement risk, while maintaining high fish protection and guidance efficiencies for spirlin (Alburnoides bipunctatus) and European eel (Anguilla anguilla). Furthermore, the fish behavior in front of an electrified rack was investigated and the advantages and disadvantages of the applied electrification setup were evaluated.

2. Materials and Methods

2.1. Experimental Setup

All experiments were conducted in a 30 m long, $w_o$ = 1.5 m wide and 1.2 m deep laboratory channel with a horizontal bed in the Laboratory of Hydraulics, Hydrology and Glaciology of ETH Zurich. The orographic left channel wall and the channel bed were made of concrete, and the right channel wall was made of glass to allow for visual observations. The glass wall was covered with a perforated foil to mimic the concrete wall and to avoid reflections (Figure 1). Two hard foam floaters and a honeycomb flow straightener were installed at the channel inlet to reduce surface waves and to achieve symmetrical flow conditions, respectively. The closed water circuit was equipped with a cooling system, such that all experiments were carried out at water temperatures in the range of 12 °C $\le T\le$ 14 °C with a water conductivity of 230 μS/cm. The approach flow depth was $h_o$ = 0.90 m and the approach flow discharge was $Q_o$ = 0.675 m³/s, leading to an average approach flow velocity of $U_o$ = $Q_o$/($h_o{w}_o$) = 0.50 m/s. Fish movements were recorded with five submerged cameras (Basler acA2040-35gmNIR) with 185° fish eye lenses (Fujifilm FE185C086HA-1) and waterproof domes (autoVimation IP67 Orca S; Figure 1). The cameras were all synchronized in a GigE Vision 2.0 network with a Precision Time Protocol (PTP) IEE1588. Frame rates were kept constant at 20 fps [28,29]. White sheets were installed above the channel and illuminated with eight 1000 Watt halogen spot lights to ensure constant indirect light conditions. The 25 cm wide full depth open channel bypass with its inlet at the downstream rack end was separated from the main channel with a 10 cm wide concrete wall (Figure 1b). The HBR consisted of aluminum bars with a foil-shaped cross-section (dimensions in [12]), which were connected with threaded bars (tie-bars) covered by cylindrical aluminum spacers.

For the e-HBR tests, a stainless steel metal mesh with 5 by 5 cm openings was installed 18 cm downstream of the rack (clear distance, measured orthogonally, Figure 1) and used as a cathode, and the rack itself acted as an anode. The electric field was independent of the mesh size, and the mesh did not influence the velocity distribution upstream. Fish were always either protected by the rack (stayed upstream of the rack) or passed through the rack and the cathode. The cathode did, therefore, not prevent any rack passages. The HBR and the metal mesh were connected by cables and multiple crocodile clips to the control unit of the electrification system (NEPTUN from the Procom company [16]). To reduce the electric field in front of the bypass, the most downstream 20 cm of the e-HBR were isolated with insulating tape.

The HBR was electrified with low-voltage gpDC (38 V $\le U_e\le$ 80 V). The following pulse parameters could be varied independently: number of pulses within a pulse group $N_{pulse}$ (–), pulse length $L_{pulse}$ (s), gap length $L_{gap}$ (s) and repetition time $t_{rep}$ (s) (Figure 2). The duty cycle D is defined as the percentage of time with an electric current in the water and is therefore a measure for the amount of electricity acting on the fish (Equation (1)). The electric power consumption of the e-HBR in the laboratory was very low, at ≈0.08 W.

$$D=\frac{N_{pulse}{L}_{pulse}}{t_{rep}}$$

(1)

2.2. Experimental Procedure

2.2.1. Electric Field Measurements

An oscilloscope was used to verify that the pulse pattern in the water, defined by $N_{pulse}$, $L_{pulse}$, $L_{gap}$ and $t_{rep}$, matched with the settings of the NEPTUN system. The peak voltage $U_e$ varied across the channel and was measured with a custom-built voltmeter with a measurement range of 1–62.5 V and an accuracy of ±0.1 V. The voltmeter was validated with a condenser with a known voltage potential. It measured the voltage difference between the cathode (metal mesh) and measurement points in the channel at the peak voltage. Different flow velocities and pulse pattern did not affect $U_e$. All electric field measurements were conducted for output voltages of the control unit of $U_e$ = 38 V. The voltmeter was mounted on a 3D traverse system to measure $U_e$ at systematic locations upstream of the e-HBR and in the bypass with a positioning accuracy of ±5 mm. The density of the measurement grid increased towards the e-HBR (anode) to a minimal distance between measurement points of 3 cm in a cross-section perpendicular to the HBR.

2.2.2. Live Fish Tests

The HBR live fish tests were carried out between 16 October 2018 and 23 November 2018. The e-HBR live fish tests were carried out between 10 October 2019 and 8 November 2019. All experiments were conducted during daytime, between 8 a.m. and 6 p.m. Wild fish were caught with mild DC electrofishing (voltages of 220 V $\le U_e\le$ 250 V, certified devices) in the Swiss cantons of Zurich (Himmelbach) and Thurgau (River Thur and a side channel of the Murg). In total, 75 spirlin and 58 eels were transported to the laboratory in aerated and temperature-stable tanks, and subsequently slowly acclimatized to the water temperature in the fish holding tanks ($\Delta T$ ≈ 1 °C/h). They were allowed to adapt to the holding tanks for at least one day before the first experiment was conducted. Spirlin and European eels were kept for up to four and seven days in the laboratory, respectively, and they were not fed during this time. The condition of the fish (physical appearance, natural behavior) and the holding tanks, including the water quality (oxygen concentration and pH), water temperature and turbidity, were monitored daily. No fighting, illness or mortality was observed during the time fish spent in the laboratory.

Previous experiments indicate that fish are more active and show a more natural behavior when tested in groups of three than when tested individually [7]. The number of tested fish per experiment was therefore set to three specimens, which still allowed us to distinguish individual fish during visual observations in a test. In accordance with Swiss animal welfare law, which calls for reusing animals to reduce the total number of animals tested, fish were used for up to three (spirlin) or four (eels) experiments. Eel were tested more often than spirlin because of the limited number of individuals which were available for the experiments. Each fish was tested only once a day and always subjected to different configurations (e.g., different pulse patterns) to reduce learning effects. Injured fish were not tested again. At the end of each experimental week, all fish were released in the same river reach where they were originally caught.

At the start of each experiment, three individuals were caught randomly with a dip net from the holding tanks and placed into a cuvette, where pictures with a reference scale were taken to determine the total fish length $TL$. The fish were then placed into the acclimatization compartment, of which the outlet was located 10.5 m upstream of the bypass inlet on the left channel side (opposite of the bypass inlet), where they could acclimatize to the flow conditions for 15 min. The acclimatization compartment was then opened, such that the fish could swim freely in the entire channel. The experiment lasted for 30 min or until all fish either passed through the rack or entered the bypass.

2.3. Parameter Range and Test Program

All live fish tests were conducted at $U_o$ = 0.5 m/s and $\alpha$ = 30°. The average flow velocity at the bypass inlet was $U_{by,in}$ = 0.6 m/s for all experiments with spirlin, and the bypass discharge corresponded to 13.7% of the approach flow discharge (details in [9]). The experiments with eels at the e-HBR were also conducted with $U_{by,in}$ = 0.6 m/s. For the HBR, the average flow velocity at the bypass inlet was in the range of 0.6 m/s $\le U_{by,in}\le$ 0.9 m/s.

The pulse parameters were chosen based on recommendations of the companies Procom from Wrocław, Poland, and IUS Weibel and Ness GmbH from Kandel, Germany, who have experience with electric deterrent systems for downstream passage and were involved in the experiments of Berger (2018) [21] and Tutzer et al. (2019) [22] (cf. Section 1.1). The electrified experiments (Test F1 to F4,

Table 1. Test program of live fish tests with eel (Anguilla anguilla) and spirlin (Alburnoides bipunctatus) with a horizontal bar rack (E1 and E2) and an electrified horizontal bar rack (F1–F5). The characteristics of each configuration are defined by the bar spacing $s_b$, applied voltage $U_e$, pulse length and gap length $L_{pulse}$ and $L_{gap}$, number of pulses $N_{pulse}$, repetition time $t_{rep}$, duty cycle D, fish total length $TL$ (minimum, maximum and average), standard deviation ${\sigma }_{TL}$, total number of individuals tested per configuration N and the total number of active fish n (fish that swam into the observation area).

Test	${\mathit{s}}_{\mathit{b}}$	${\mathit{U}}_{\mathit{e}}$	${\mathit{L}}_{\mathit{pulse}}$	${\mathit{L}}_{\mathit{gap}}$	${\mathit{N}}_{\mathit{pulse}}$	${\mathit{t}}_{\mathit{rep}}$	D	Fish Species	${\mathit{TL}}_{\mathit{min}}-{\mathit{TL}}_{\mathit{max}}$	N	n	$\mathit{n}/\mathit{N}$
	(mm)	(V)	(ms)	(ms)	(−)	(ms)	(%)		$(\overline{\mathit{TL}},{\mathit{\sigma }}_{\mathit{TL}}$) (cm)	(−)	(−)	(%)
E1	20	-	-	-	-	-	-	Spirlin	8.1–10.7 (9.6, 0.7)	33	33	100
E2	20	-	-	-	-	-	-	Eel	41.2–82.8 (65.2, 10.2)	26	23	88
F1	20	38	0.3	7	5	200	0.75	Spirlin	8.9–13.8 (10.9, 1.0)	36	29	81
F2	20	38	0.3	7	5	200	0.75	Eel	45.4–84.2 (65.1, 9.2)	15	12	80
F3	51	38	0.3	7	5	200	0.75	Spirlin	7.4–11.4 (8.9, 0.9)	51	27	53
F4	51	38	0.3	7	5	200	0.75	Eel	42.9–87.4 (68.1, 8.6)	72	59	82

) were conducted with the following pulse parameters: $L_{pulse}$ = 0.3 ms, $L_{gap}$ = 7 ms, $N_{pulse}$ = 5, $t_{rep}$ = 200 ms (Table 1). The duty cycle of this configuration was D = 0.75% (Equation (1)).

According to $TL$ in Table 1, all tested spirlin could have physically passed through the rack at a 20 mm bar spacing [9,10]. Additionally, all eels with $TL$ < 70 cm could have physically passed through the rack at 20 mm [10]. The main goal of the study was to test whether e-HBRs with a large clear bar spacing ($s_b$ = 51 mm) are an alternative to HBRs with a small clear bar spacing ($s_b$ = 20 mm) without electrification. Experiments with $s_b$ = 20 mm were carried out with and without electrification to check for adverse effects on the guidance efficiency of an electrified HBR.

An HBR with 51 mm bar spacing was not tested because the German [10] and Swiss guidelines [11] for downstream migration require a bar spacing of $s_b$ < 20 mm for physical barriers like HBRs, which implies that HBRs with $s_b$ > 20 mm are unsuitable for fish protection. This statement is support by multiple studies, as discussed in Section 4.4.

2.4. Data Analysis

2.4.1. Fish Guidance and Protection Efficiency

Three different kinds of fish behaviors were defined, namely, bypass passage, rack passage and refusals. The numbers of bypass passages $N_{by}$, rack passages $N_{rack}$ and refusals $N_{ref}$ were used to calculate the fish guidance efficiency (FGE) and the fish protection efficiency (FPE) with Equations (2) and (3), respectively.

$$\mathrm{Fish}\mathrm{guidance}\mathrm{efficiency}\mathrm{FGE}=\frac{N_{by}}{N_{by}+N_{rack}+N_{ref}}$$

(2)

$$\mathrm{Fish}\mathrm{protection}\mathrm{efficiency}\mathrm{FPE}=\frac{N_{by}+N_{ref}}{N_{by}+N_{rack}+N_{ref}}$$

(3)

Bypass passages, rack passages and refusals were counted manually through visual observation and validated through manual analysis of the video files and the fish tracks generated with a MATLAB based fish-tracking software (see Section 2.4.2). At the HBR, bypass passages were only counted if the fish had a rack interaction before passage, which was defined as a maximum distance of 15 cm between the fish’s center and the HBR (Figure 3, Sector 5). Fish that directly passed into the bypass without a rack interaction were not used to calculate FGE and FPE, as these fish were likely not guided by the rack. Only the first behavior was counted in case of a rack and bypass passage. If a fish swam back out of the bypass, this was therefore still counted as a bypass passage and further interactions were not counted. Fish that neither swam into the bypass nor passed the rack within the maximal experimental duration of 30 min, but swam into sector 4, 5 or 7 were counted as refusals. Fish were classified as inactive and thereby excluded from the calculation of FGE and FPE, when they did not swim into any of sectors 4, 5 and 7.

The distinction between guided fish (bypass with rack contact) and fish that swam into the bypass "by chance" led to a conservative estimation of FGE and FPE compared to other studies where this distinction was not made (e.g., [7,22,30]). To allow for comparison, both direct bypass passages and bypass passages with rack contact are reported in Section 3.

For the e-HBR tests, the electric field extended across the entire channel (details in Section 3.1). Therefore, it was not possible to distinguish whether a fish was affected by the electrified rack or not, and all fish that entered the bypass were counted as bypass passages independently of their swimming paths. Rack passages and refusals were counted in the same way as for the HBR.

2.4.2. Sector Analysis

The video recordings of all experiments were analyzed with a MATLAB based fish-tracking software described in [28,29,31]. The code is freely available under [32]. The tracking code allows one to track the fish swimming paths in each experiment, whereby each fish is represented by the coordinates in x and y-directions with the corresponding timestamps. In some experiments with spirlin, individuals could not be distinguished from each other by the tracking code, because they were swimming too close together and often swam in front of one another, as seen from the camera. For these experiments, the position of three individuals was assumed as the centroid of the detected group of fish.

To analyze multiple tracks and compare different test configurations with each other, the channel was divided into sectors (Figure 3). Sectors 1, 2 and 7 covered the 15 cm wide boundary areas along the channel walls. Sector 6 represents the bypass, and sector 5 encompasses the 15 cm wide area upstream of and parallel to the HBR, which was used to determine if fish were interacting with the HBR. The streamwise, transversal and vertical coordinates are denoted as x, y and z, respectively, and the point of origin was set to the channel bottom at the downstream rack end (Figure 3).

To measure the shares of time individual fish spent in the different sectors, while accounting for the sector size, the residence coefficient $R_{c,i}$, defined by Equation (4), was calculated. The normalized residence coefficient $R_{c,i,norm}$ (Equation (5)) was limited to values between zero and one, where $R_{c,i,norm}$ = 0 means that no fish entered sector i for this specific configuration. In contrast, $R_{c,i,norm}$ = 1 specifies that all fish swam within sector i only.

$$R_{c,i}=\frac{1}{N}\sum _{j=1}^N\frac{t_{i,j}{A}_{tot}}{t_{j,tot}{A}_i}$$

(4)

$$R_{c,i,norm}=\frac{R_{c,i}}{{\displaystyle \sum _{i=1}^7}{R}_{c,i}}$$

(5)

where i = sector number (–), j = fish number (–), N = total number of individuals tested per configuration (–), $t_{i,j}$ = time the fish j spent in sector i (s), total time the fish j spent in any of the sectors 1–7 (s), $A_i$ = area of sector i (m²) and total area of all sectors (m²).

2.4.3. Statistical Analysis and Data Limitations

The experimental limitations explained in Section 2.2.2 led to some dependencies in the data which rendered a statistical analysis difficult. Testing fish up to four times led to a certain data dependency. Additionally, testing fish in groups of three led to pseudoreplication. However, eels are known to be solitary [33] and did not show any schooling behavior in the present study. In most experiments eels, swam downstream independently with breaks of several minutes between individuals. Spirlin, on the other hand, showed strong schooling behavior. Spirlin were still not tested individually, as fish are known to be more active and show more natural behavior when tested in groups [7,34]. An additional limitation of the experiments was that individual marking was not possible during these tests.

To assess if differences in FGE and FPE were statistically significant, despite the mentioned limitations, two-sided ${\chi }^2$-tests were applied, which assume all data points to be independent. However, due to this simplification, the significance values only give an indication of whether differences in FGE and FPE were statistically significant. The two-sided ${\chi }^2$-tests were applied with a significance level of ${\alpha }_{sig}$ = 0.05 for each possible outcome from the live fish tests, i.e., refusal, bypass and rack passage. The null hypothesis $H_0$ states that there was no statistically significant difference between the two tested configurations, whereas the alternative hypothesis $H_1$ applies for significant differences.

3. Results

3.1. Measurements of the Electric Field

Figure 4 shows the interpolated contour map of the electric field strength. Measurements at different flow depths revealed that the variation of the field strength in the vertical direction was fairly within the measurement accuracy (±0.02 V/cm). Therefore, only the measurements collected close to the bottom are shown in Figure 4. Likewise, changing the clear bar spacing $s_b$ from 20 to 100 mm did not affect the electric field strength distribution. The measurements shown in Figure 4 ($U_e$ = 38 V; $s_b$ = 100 mm) can therefore be used to explain the fish behavior for all tested configurations with the e-HBR listed in Table 1. For a voltage output of $U_e$ = 38 V from the NEPTUN system, the largest voltage measured at the anode was 31 V, indicating a ≈20% reduction compared to the output voltage of the control unit. That voltage reduction stemmed from losses in the cables and a thin layer of calcium depositing on the electrodes during the course of the experiments.

The electric field resembled an electric field generated by a plate condensator, with the rack and downstream grid acting as the plate electrodes. The field strength was the greatest at the rack and gradually decreased with the distance from the rack. Voltage gradients expected to trigger fish reactions (0.05 V/cm $\le E\le$ 0.10 V/cm) occurred within an area of 30–40 cm upstream of the rack (Figure 4). The electric field strength peaked around E = 0.35 V/cm at the downstream rack end (Figure 4). An increase of up to 0.25 V/cm was also visible at the upstream rack end, though less pronounced than at the bypass entrance. The difference is due to the geometry at and past the bypass entrance, the potential between the electrodes is discharged through the water, whereas at the upstream end, the rack touches the concrete wall, leading to a voltage discharge through the concrete wall instead of through the water.

3.2. Fish Behavior, Protection and Guidance

3.2.1. Spirlin

The characteristic behavior of spirlin is described in the following paragraphs, where the percentage of spirlin showing this specific behavior is given in parentheses. In the experiments with the non-electrified HBR, spirlin typically swam downstream in zigzag movements between the channel wall and the HBR and were thereby guided towards the bypass [9,13] (78%). Similarly, most spirlin approached the e-HBR in zigzag movements with positive rheotaxis (82%), i.e., facing the oncoming flow. In contrast to the non-electrified HBR, spirlin typically reacted to the electric field around 10–20 cm upstream of the e-HBR with sudden but controlled upstream burst swimming movements (82%). Instead of fleeing towards the headwater, spirlin often swam only a couple of centimeters upstream, before they re-approached the e-HBR (88%). After several attempts, they often continued their downstream zigzag movements between the e-HBR and the channel wall and were thereby guided towards the bypass inlet (93%). Despite several attempts, spirlin often refused to enter the bypass until the end of the experiment (73%). This was only observed in e-HBR experiments where a voltage gradient of up to E = 0.35 V/cm was present at the downstream rack end (cf. Figure 4). For the HBR, no bypass refusals were observed. Some spirlin swam very close to the e-HBR and almost touched it (50%), and other specimens reacted to the electric field more than 50 cm upstream of the e-HBR (15%). Compared to the HBR without electrification, spirlin did not only swim very close to the channel bottom, but they frequently used the whole water column along the e-HBR and the bypass inlet and seemed to actively search for an appropriate downstream passage corridor (32%). Figure 5 shows the typical behavior of spirlin at the e-HBR, where three individuals approached the rack with positive rheotaxis. Similarly to the non-electrified HBR, they were mostly almost rack-parallel when swimming towards the e-HBR (Figure 5a), swam close to the rack (Figure 5b) before they changed their swimming direction and moved towards the glass wall (Figure 5c). After another zigzag movement (Figure 5d–f), they tried to swim into the bypass (Figure 5g) but refused to enter it and swam back upstream (Figure 5h). In this experiment, two spirlin entered the bypass ≈14.5 min after they first approached the e-HBR after several attempts. The sudden upstream swimming movements can hardly be visualized with screenshots, but can be best seen on video, which is provided as supplementary material.

In Figure 6, the fish protection and guidance efficiencies at the e-HBR with $s_b$ = 20 mm and $s_b$ = 51 mm are compared to those of the HBR with $s_b$ = 20 mm for the same hydraulic conditions ($U_o$ = 0.5 m/s, $U_{by,in}$ = 0.6 m/s). The white numbers in Figure 6 represent the absolute numbers of fish for all the reactions. The green bar represents the fish guidance efficiency FGE (cf. Equation (2)), and the sum of the green and yellow bars represents the fish protection efficiency FPE (cf. Equation (3)). For configuration E1, six spirlin entered the bypass without rack interaction and were therefore not considered for calculation of the FGE and FPE, as explained in Section 2.4.1.

For the HBR and $s_b$ = 20 mm, all spirlin entered the bypass (E1, Figure 6). The e-HBR with the same bar spacing (F1), led to significantly fewer bypass passages (34% compared to 100%, p < 0.001, ${\chi }^2$ = 23.931) and more refusals (59% compared to 0%, p < 0.001, ${\chi }^2$ = 20.039). The number of rack passages increased as well (0% to 7%), but this difference was not significant (p = 0.503, ${\chi }^2$ = 0.448). Among the electrified racks (F1 and F3), no statistically significant differences were observed between $s_b$ = 20 mm and $s_b$ = 51 mm for bypass passages (p = 0.919, ${\chi }^2$ = 0.010), refusals (p = 0.730, ${\chi }^2$ = 0.119) and rack passages (p = 1.000, ${\chi }^2$ = 0.000).

The e-HBR with a bar spacing of 51 mm therefore provided similar protection as the HBR with a bar spacing of 20 mm (FPE > 96%). Independent of the bar spacing, the bypass acceptance and guidance was reduced for the e-HBR compared to the HBR.

Two rack passages were observed for the e-HBR with $s_b$ = 20 mm and one with $s_b$ = 51 mm. One of these spirlin passed through the e-HBR with $s_b$ = 20 mm at the downstream rack end in the upper water column 13 min after the start of the experiment. The swimming capabilities of the spirlin seemed impaired, and it was swept downstream through the HBR. During rack passage, the spirlin was capable of rotating to the side and pass through the rack with minimal contact.

The second spirlin which passed through the e-HBR with $s_b$ = 20 mm approached the rack with positive rheotaxis and actively swam towards the rack, such that it was oriented almost rack-parallel during rack passage. This orientation parallel to the equipotential lines led to a very low body voltage. After approximately one second, the spirlin swam upstream through the e-HBR again and entered the bypass actively with negative rheotaxis (i.e., head-first with the flow). The rack passage of the spirlin at the e-HBR with $s_b$ = 51 mm is described in Section 3.3, as this fish was severely injured during rack passage.

Figure 7 shows the normalized residence coefficients of the e-HBR with different $s_b$ and the HBR without electrification for reference. The electrification did not affect the time fish spent near the left channel wall (sector 1) and the channel center (sectors 3 and 4) with $R_{\mathrm{c},\mathrm{i},\mathrm{norm}}$ < 0.1 for E1, F1 and F3. At the e-HBR (F1 and F3), spirlin avoided the rack (sector 5) and bypass area (sector 6) where $R_{\mathrm{c},\mathrm{i},\mathrm{norm}}$ decreased from 0.24 and 0.37 to less than 0.12, respectively, compared to the HBR. With the electrification, spirlin swam more frequently close to the right channel wall upstream of the bypass inlet (sector 2, $R_{\mathrm{c},\mathrm{i},\mathrm{norm}}$ = 0.08 for the HBR, 0.32 and 0.53 for the e-HBR), which indicates that they were guided in the direction of the bypass but hesitated to enter it. The measurements of the electric field showed that the voltage gradients were highest at the bypass inlet (cf. Figure 4).

3.2.2. European Eel

Most eels approached the e-HBR actively with negative rheotaxis (79%), but some eels swam slowly downstream with positive rheotaxis (19%) or let themselves drift passively (2%). The reactions of eels to the electric field varied strongly between individuals. A few specimens changed their rheotaxis several meters upstream of the e-HBR (7%), where the voltage gradient was in the range of $E\approx 0.02$ V/cm (cf. Figure 4), and others reacted only after rack contact (67%). In general, eels which swam actively downstream with negative rheotaxis reacted closer to the rack than eels which approached the e-HBR slowly or passively. When approaching the rack, eels typically aligned rack-parallel and swam towards the bypass (36%) or escaped upstream (64%), either directly against the main flow direction (16%) or in the direction perpendicular to the rack (84%). While some eels swam upstream to the acclimatization compartment after rack contact (24%), most individuals swam only a couple of meters upstream, before they re-approached the rack (76%), like the eel shown in Figure 8, which represents a frequently observed reaction to the e-HBR. This eel approached the e-HBR with negative rheotaxis (Figure 8a) and aligned itself almost perpendicularly with the equipotential lines, such that the body voltage was $\Delta U_e\approx$ 5 V and the maximum voltage gradient was E = 0.25 V/cm (Figure 8c). The eel changed its rheotaxis (Figure 8b,c) and fled upstream, such that it was again aligned almost perpendicular to the equipotential lines, where it was exposed to $\Delta U_e\approx$ 6 V close to the rack (Figure 8d) and $\Delta U_e\approx$ 4 V further upstream (Figure 8e). The eel then drifted downstream along the glass wall where the electric field strength was rather small (Figure 8f,g), before it changed back to negative rheotaxis to enter the bypass (Figure 8h). Almost all eels flinched during bypass passage at the bypass inlet where the field strength was 0.10 V/cm $\le E\le$ 0.15 V/cm and up to $E\approx$ 0.35 V/cm very close to the rack (cf. Figure 4).

Figure 9 shows the fish guidance (green bar) and protection efficiencies (green and yellow bar). The white numbers represent the absolute numbers of active fish. For E2, three eels entered the bypass directly without rack interaction and are therefore not included in Figure 9, as explained in Section 2.4.1. For the HBR and $s_b$ = 20 mm (E2), the majority of eels was protected and subsequently entered the bypass (FGE = 90%). Counting also bypass passages without rack interaction, the FGE would be 91%. No rack passages were observed at the e-HBR with $s_b$ = 20 mm (F2), such that FGE and FPE were 92% and 100%, respectively. At the e-HBR with $s_b$ = 51 mm (F4), the fish guidance and protection efficiencies were similar with FGE = 85% and FPE = 86%. The differences between the tested configurations were not statistically significant with a two-sided ${\chi }^2$ test and ${\alpha }_{sig}$ = 0.05 (assumptions explained in Section 2.4.3): between the HBR with $s_b$ = 20 mm (E2) and the e-HBR with $s_b$ = 51 mm (F4) p = 0.831 and ${\chi }^2$ = 0.045 for bypass passages; p = 1.000 and ${\chi }^2$ = 0.000 for refusals; and p = 0.526 and ${\chi }^2$ = 0.402 for rack passages. The e-HBR with a bar spacing of 51 mm therefore provided similar, very good protection (FPE > 85%) as the HBR with a bar spacing of 20 mm. Unlike for spirlin, guidance was barely reduced for the e-HBR compared to the HBR.

Without electrification, eels showed thigmotactic positive behavior, meaning that they sought contact with structures and were swimming along the HBR with continuous rack contact. This behavior was quantified with the sector analysis in Figure 10. Without electrification (E2 in Figure 10), eels spent 37% ($R_{\mathrm{c},\mathrm{i},\mathrm{norm}}=0.37$) of the time in the experiment in front of the HBR (sector 5) and only < 5% ($R_{\mathrm{c},\mathrm{i},\mathrm{norm}}<0.05$) in front of the bypass (sector 7). The behavior was completely different with the electrification, where eels avoided the rack area ($R_{\mathrm{c},\mathrm{i},\mathrm{norm}}<0.05$, sector 5) and spent more time near the wall upstream of the bypass inlet ($R_{\mathrm{c},\mathrm{i},\mathrm{norm}}=0.22-0.3$, sector 7; F2 and F4 in Figure 10). The variations between the e-HBR configurations with different $s_b$ (F2 and F4) were small, which was expected as $s_b$ does not affect the electric field (cf. Section 3.1).

The acclimatization compartment was installed close to the left channel wall. Without electrification, many eels swam out of the acclimatization compartment with negative rheotaxis, such that they used the left channel wall more frequently than the right channel wall (sectors 1 vs. 2 in Figure 10). In contrast, with the electrification, eels slightly favored the right channel wall.

Although many eels were guided towards the bypass by the e-HBR with $s_b$ = 51 mm, eight individuals passed through the rack (Figure 9). One rack passage is illustrated in Figure 11. This eel approached the e-HBR with negative rheotaxis at the upstream rack end, resulting in an almost perpendicular orientation to the equipotential lines with a body voltage of $\Delta U_e\approx$ 3 V (Figure 11a). With backwards swimming movements, the eel changed its orientation at the upstream rack end (Figure 11b,c), and fled with an angle of ≈45° to the equipotential lines, such that the tail touched the rack, which provoked active swimming movements (Figure 11d). As the eel swam along the rack, this angle reduced to ≈35° (Figure 11e) and further to ≈15°, resulting in $\Delta U_e\approx$ 1 V (Figure 11f). Finally, the eel was aligned almost rack-parallel along the equipotential lines, such that it was exposed to a very low body voltage when passing through the e-HBR (Figure 11g,h). Five out of the eight eels, which passed through the e-HBR with $s_b$ = 51 mm (cf. Figure 9), passed through it similarly to the eel in Figure 11; that is, they passed through the e-HBR with an almost rack-parallel orientation. Two eels directly approached the e-HBR almost rack-parallel with negative rheotaxes and passed through it, whereas one eel swam from the glass wall towards the e-HBR with a positive rheotaxis and also passed through it rack-parallel. Hence, all rack passages occurred with the eel being oriented rack parallel.

A typical behavior for eels that were guided towards the bypass was similar to the eel in Figure 11a–e, but instead of aligning rack-parallel, they touched the rack multiple times with their tails and kept an angle of ≈30°–45°, between their body axis and the e-HBR (similar to Figure 11d,e) until they reached the bypass.

3.3. Fish Injuries

Without electrification, neither fish impingement at the HBR nor any other injuries were observed. In recent studies with electrified racks, where gpDC with $U_e\le$ 80 V was used (e.g., [21,22]), no injuries were reported. Therefore, it was not expected that fish would get injured in the present experiments with the lower applied voltage (38 V). However, after the e-HBR experiments, local dark coloration of the skin was observed for several spirlin. This coloration was particularly pronounced for two spirlin that passed the e-HBR with $s_b$ = 51 mm. They were anesthetized and subsequently euthanized with MS222 (Tricaine-S) and checked for internal injuries through autopsy at the Centre for Fish and Wildlife Health in Bern, Switzerland. One of the spirlin was tested with the e-HBR with $s_b$ = 51 mm, 38 V and the pulse pattern described in Table 1. The other one was tested with $s_b$ = 51 mm, 38 V, but a different pulse pattern ($L_{pulse}$ = 0.2 ms, $L_{gap}$ = 0 ms, $N_{pulse}$=1, $t_{rep}$ = 100 ms). Due to the severe injury, this pulse pattern was not used in further tests and hence is not listed in Table 1. However, since the field strength was equivalent, the autopsy results of both spirlin are discussed here. The full autopsy reports were published in German in the appendix of Meister (2020) [9].

The spirlin, which passed through the e-HBR with the standard pulse pattern (cf. Figure 6, F3), actively approached the rack with negative rheotaxis directly after the acclimatization compartment was opened. Figure 12 shows this spirlin before the experiment and after the rack passage, which resulted in internal bleeding in the eye and dark colorations of the skin (Figure 12b). The autopsy did not detect any internal bleeding in the muscles or along the vertebrae, and although likely, it could not be verified whether the dark colorations of the skin resulted from contact with the electrified rack. It is possible that the injuries resulted from impingement at the fine net (opening 1 cm) installed at the downstream end of the channel. This net was installed to prevent fish from entering the water circuit and getting injured in the pump. This scenario was, however, unlikely if the fish arrived at the downstream net unharmed by the electric field. The flow velocity at the downstream net was $U_o$ = 0.5 m/s, which was similar to the average approach flow velocity in the experiments with and without electrification. According to the proposed equation by Ebel (2016) [8], the swimming speed of rheophilic fish species such as the spirlin should allow the spirlin to swim for the full duration of the experiment of 30 min.

The second spirlin first passed the rack; reemerged shortly afterwards, passing the rack again; and then passed into the bypass. It had dark colorations of the skin behind the dorsal fin (Figure 13), and its swimming behavior was impaired. The video recordings indicate that the spirlin touched a vertical tie-bar at the location of the dark coloration. The diagnosis of autopsy report stated that this spirlin had a spinal fracture in the area of the spinal canal with associated bleeding in the spinal canal and the surrounding muscles, which was likely caused by an electric shock.

Although small external injuries of the tested eel could not be certainly attributed to the e-HBR, it was observed on the video recordings several times that eels hemorrhaged a dark fluid after they swam very close to the rack or touched the rack. Figure 14 shows an eel which slightly touched the e-HBR at the upstream rack end and subsequently escaped upstream. The dark fluid, which is encircled in Figure 14c, can be seen best on the video recordings provided as supplementary material. The dark fluid was not only observed when eels had contact with the anode, but also when they swam out of the bypass through the area with voltage gradients up to E = 0.35 V/cm (cf. Figure 4). Although multiple videos were shown to several biologists within the present study, it could not be clarified what kind of liquid it was and whether it was harmful for the eels. Possible explanations for the observed fluid are either blood that was excreted from the gills or some liquid from the intestine excreted through the pharynx.

No fish injuries were expected prior to the present, study and thus they were not systematically analyzed, resulting in no systematic quantification of fish injuries at e-HBRs from the present investigation. However, this study showed that fish can get injured at electric barriers even with low output voltages of $U_e$ = 38 V, especially if they directly touch the electrodes. The electric field map in Figure 4 shows that $E\le$ 0.35 V/cm in front of the e-HBR for $U_e$ = 38 V is in the range considered suitable for electrofishing (cf. Section 1.2). However, if a fish passed through the e-HBR, like the spirlin shown in Figure 13, it was exposed to a larger voltage gradient between the e-HBR and the metal mesh. Due to experimental restrictions, it was not possible to measure the exact voltage gradient at this location. Assuming a linear voltage gradient analogous to a plate condenser between the e-HBR (anode) and the metal mesh (cathode), which was installed 18 cm downstream of the e-HBR, the voltage gradient was $E\approx$ 38 V/18 cm = 2.1 V/cm.

4. Discussion

4.1. Fish Protection and Guidance Efficiency

The fish protection and guidance efficiencies are crucial parameters used to evaluate and compare the effectiveness of fish guidance structures. Silver eels in particular need to pass multiple HPPs on their migration to the sea; therefore, a high protection efficiency is crucial to ensuring that as many eels as possible reach their spawning grounds. High guidance efficiency is also of prime importance, as this is an indicator of the delay faced by a fish until it can pass the HPP. Different methodologies have been used to evaluate the FGE and FPE, depending on the experimental setup. Absolute numbers should be carefully compared, as they depend highly on the counting method and definition of bypass passages and refusals and the number of tested fish. The counting methods available depend on the experimental setup (PIT-tags, sonar, video observation or direct visual observation).

Counting methods that have been used included:

• Only fish that passed through the rack or swam into the bypass were counted. All other fish placed in a given experiment were counted as refusals. This approach was used with PIT tag antennas in the bypass and downstream of the rack by [22] and with direct observation by [7,21,34]. The FPE was overestimated, as it did not account for fish that were inactive and never approached the rack. The FGE is difficult to evaluate, as this method does not account for inactive fish, nor for fish swimming directly into the bypass without rack contact.
• A distinction between active fish (entering the area upstream of the rack) and inactive fish was made. All fish that entered the bypass were counted as bypass passages, all active fish that neither passed through the rack nor the bypass were counted as refusals. This approach was used here for e-HBRs. It is also generally used in monitoring campaigns in the field. In a laboratory setting, the FGE may be overestimated, since the bypass is large compared to the length of the rack and certain individuals may directly swim into the bypass along the channel walls without interacting with the rack.
• A distinction was made between active (entering the area upstream of the rack) and inactive fish. Only fish that interact with the rack before entering the bypass were counted as bypass passages; all active fish that neither passed through the rack nor the bypass were counted as refusals. This approach leads to the most conservative estimate of FGE and FPE. It was used here for the HBR, and by [9,35].

We used counting method 3 for the HBR choosing a conservative approach. For the e-HBR, counting method 2 was used, assuming that the electric field extended across the entire channel and all fish that entered the bypass interacted with the electric field (c.f. Section 2.4.1). Therefore, it was not necessary to distinguish bypass passages with and without contact as in counting method 2. The assumption that all fish entering the bypass were affected by the electric field was confirmed by the time fish spent in different sectors of the channel during the HBR and e-HBR experiments (c.f. Figure 7 and Figure 10). Spirlin used sectors along the right and left channel walls equally at the HBR, but with the e-HBR, the sector upstream of the bypass and away from the electric field was used much more. For eels, a similar observation was made. At the HBR, the left wall upstream of the rack was used more than the right channel wall. At the e-HBR no preference of the channel wall was observed. Both observations indicate a guidance effect of the electric field towards the right channel wall and confirm that fish are affected by the electric field several meters upstream of a rack. Therefore a distinction between bypass passages with and without contact was not feasible, as it could not be defined precisely at which electric field strength the electric field was perceived, leading to a switching of the channel side preference.

To asses the effect of the different counting methods of the bypass passages on the results reported in Section 3.2, all data analyses were repeated by counting all bypass passages for the HBR and e-HBR, regardless of whether fish interacted with the rack or not. All statistically significant differences reported in Section 3.2 were also significant according to this simplified data analysis. However, distinguishing between bypass passages with and without rack interaction with the HBR is more appropriate and conservative, which is why these results were reported in Section 3.2.

4.2. Limitations of the Statistical Analysis

As described in Section 2.4.3, fish reactions were assumed to be independent for the ${\chi }^2$-tests. This is a simplification, as fish may have been affected by the behavior of other fish of the same group. No schooling behavior was observed in any of the eel experiments, which is typical for eels [33]. In most eel experiments, several minutes passed between the downstream passage of different individuals. It is therefore unlikely that the behavior of one eel affected the behavior of the other eels, which means that the assumption of independent behavior is valid and did likely not affect the results. In contrast, spirlin showed strong schooling behavior, which means that the behavior of one spirlin likely affected the behavior of the other spirlins of the same group. It is therefore important to assess to which degree this simplification affected the results. The only significant differences detected for the spirlin experiments were that the e-HBR with $s_b$ = 20 mm and $s_b$ = 51 mm led to significantly less bypass passages and more refusals than the HBR with $s_b$ = 20 mm (cf. Section 3.2.1). The effect of the schooling behavior was assessed by repeating the ${\chi }^2$-tests, but instead of comparing the behavior of each individual, the most common behavior of each group was used as one data point. This analysis supports the findings described in Section 3.2.1. The e-HBR led to significantly less bypass passages (p = 0.007, ${\chi }^2$ = 7.194) and more refusals (p = 0.007, ${\chi }^2$ = 7.194) than the HBR with $s_b$ = 20 mm. This congruence between the two types of analyses suggests that the individual behavior of spirlins may have been influenced by that of congeners. To adequately address this in future experiments, the identity of individuals should be accounted for in the data analyses.

Within the present study, individual marking was not possible. Individual variations could therefore not be accounted for in a statistical model. Although unlikely, it can therefore not be ruled out that individual variations affected the results. In order to account for individual variations in a statistical model (generalized linear model), at least five replications per fish and individual marking would be necessary [36], which was not possible with the animal experiment permit given for the present experiments. To asses effects of individual variations, the authors recommend individual marking, at least five replications per fish and a data analysis including generalized linear models in future studies.

4.3. Electrification Setup and Layout

In the present study, an e-HBR was tested where the bar rack acted as an anode and a downstream grid as a cathode. This electrode arrangement generates an electric field with equipotential lines mostly parallel to the rack (Section 3.1). The electric field strength was largest between the rack and the downstream electrode and decreased upstream as distance to the rack increased. As highlighted by Beaumont (2016) for a horizontal voltage gradient, the fish’s reaction is dependent on the fish’s swimming orientation. This effect was also observed in the present study. The alignment of the equipotential lines seemed to favor guidance towards the bypass, as fish experienced the least electric stimulus when swimming parallel to the rack. However, all rack passages occurred when individuals, especially eels, aligned parallel to the rack and thereby must have barely felt any electric impulses.

The electrification setup with $U_e$ = 38 V led to relatively large voltage gradients of up to E = 0.35 V/cm at the bypass inlet (cf. Figure 4), triggering distinct avoidance reactions and many refusals of spirlin, which would likely delay or impede downstream passage at prototype application and might increase predation risk. Eels seemed to be less sensitive to the gradient at the bypass entrance, leading to similar FGE at the HBR and e-HBR. This indicates that (i) eels reacted less sensitively to the electric field, despite their length, which exposes them to larger body voltages when aligned perpendicular to the equipotential lines, and/or (ii) eels were more strongly motivated to move downstream despite the electric stimuli.

To reduce the variability of the body voltage with the orientation of the fish and reduce the electric field strength at the bypass entrance, a vertical instead of a horizontal voltage gradient could be generated. One possibility is to use each bar alternately as an anode and cathode, instead of using the e-HBR as the anode and a metal mesh as the cathode like in the present study. This would lead to a voltage gradient varying mainly in vertical direction and extending only a few cm upstream of the rack, similarly to the “fish protector”, where horizontal steel cables are alternately electrified [22]. Fish would then perceive the voltage gradient over their body height, independent of their orientation in x- and y-direction. It is recommended that such an electrification setup be investigated in future studies to further improve the technology of e-HBRs, before they are implemented at pilot HPPs.

In further studies, the electric field should be simulated numerically and validated with measurements to detect potential areas with critical voltage gradients, for example, in front of the bypass.

4.4. Fish Protection at Large Bar Spacings

Within the present study, an HBR with 20 mm bar spacing and an e-HBR with 20 and 51 mm bar spacing were investigated. Although not specifically tested, it can be assumed that HBRs without electrification with $s_b$ = 51 mm offer hardly any protection for eels and spirlin. This assumption is supported by the results of Beck et al. (2020) [35], whose experiments were conducted in the same laboratory flume. The majority of eels passed through a vertical bar rack (curved bar rack, CBR) with $s_b$ = 50 mm (FGE = FPE < 50%, n = 12). Furthermore, various studies indicate that even a 20 mm screen is not sufficient to protect an eel with a body length of 70 cm ([37,38,39]), as cited in [10]. Meister (2020) [9] proposed an equation which can be used to estimate FPE at HBRs for spirlin, trout, eel, barbel, trout and nase based on $s_b$ and fish dimensions. For the largest spirlin tested in this study ($TL$ = 13.8 cm), his equation predicts FPE = 0% for $s_b$ = 51 mm.

Despite the large bar spacing of 51 mm, FPE > 86% for spirlin and eels was observed at the e-HBR, which is a significant improvement compared to the expected large number of rack passages for an HBR with 51 mm bar spacing. Berger (2018) [21] studied the behavior of eels at an e-HBR, which was electrified in a similar way to the present study. She observed a FPE of 99% at the e-HBR with $s_b$ = 30 mm compared to 86% in the present study for $s_b$ = 51 mm. These variations were likely caused by differences in the experimental setup, such as the inclusion of a bottom overlay by Berger (2018) [21], different bar shapes, pulse patterns or the size range of the tested fish. The geometry of the electrodes in the setup of Berger (2018) [21] was similar to the present study, but the distance between the rack and the anode was larger (30 cm). Therefore, the shape of the electric field was likely similar but with lower field strength, as the distance between the rack and anode was larger. Berger (2018) [21] limited her tests to eels, and only a few rack passages were observed, which may explain why no fish injuries were reported. Most eels were protected well with 18 mm $\le s_b\le$ 20 mm with and without electrification in both studies, confirming the efficiency of HBRs with a bar spacing < 20 mm. For e-HBRs with 30 mm $\le s_b\le$ 50 mm the protection efficiency was similar to HBRs with a bar spacing < 20 mm.

Tutzer et al. (2021) [22] investigated the “fish protector”, where horizontally arranged steel cables were alternately used as anode and cathode and electrified with the same pulse pattern used here (Table 1). They reported FPE of 97.5–99.5% with the electrified “fish protector” with 30 mm $\le s_b\le$ 60 mm for trout, chub and grayling. Similarly, a FPE of 93–96% was achieved for spirlin in the present study with the e-HBR with 20 mm $\le s_b\le$ 51 mm. Nevertheless, the FPE cannot directly be compared, since Tutzer et al. (2021) [22] did not distinguish fish that approached the rack from inactive fish, which stayed in the upstream area throughout the experiment, leading to an overestimation of the FPE (Counting method 1, Section 4.1). Furthermore, with their counting approach, the FPE was already very high (83.4% for $s_b$ = 60 mm and 92.9% for $s_b$ = 30 mm) without electrification, despite the large cable spacing. They did not report the percentage of bypass passages but state that they were not reduced by electrification.

Both the electrified "fish protector" and e-HBRs are promising fish guidance structures, as they provide high fish protection efficiencies despite large bar/cable spacings. Nevertheless, electrification still cannot provide full protection and guidance may be poor. To prevent injuries and improve guidance, the electrode setup, electric field strength and pulse pattern need to be studied in more detail and reported in future studies, which was not the case in Berger (2018) [21].

4.5. Implications for Practical Application

The clear bar spacing $s_b$ of HBR-BSs is typically chosen such that the target fish species and size class are physically protected, which means that $s_b$ has to be smaller than the fish width [8,9]. The German and Swiss Guidelines currently demand $s_b\le$ 20 mm [10,11]. Up to now, HBR-BSs were primarily installed at small to medium-sized HPPs with design discharges of $Q_d\le$ 120 m³/s, where HBRs with such small bar spacing are feasible [9]. Challenges arising from the installation of HBR-BSs at larger HPPs include increased hydraulic load, high investment costs, larger damage in case of fatigue, increased clogging by floating debris and potentially larger $U_o$ which may cause fish impingement [9]. All these aspects can likely be mitigated with larger $s_b$. E-HBR-BSs with $s_b=$ 51 mm or even larger could be installed at HPPs where it is not feasible or economical to install HBRs with $s_b\le$ 20 mm. If the technology of e-HBRs can be further improved, which might be achievable with another electrification setup as described in Section 4.3, it could contribute to the implementation of the European Water Framework Directive, which demands undisturbed fish passage. Nevertheless, the authors do not recommend the installation of an e-HBR-BS with a similar setup as investigated in the present study at a pilot HPP because of the observed fish injuries and the increased number of refusals.

Despite the large clear bar spacing of $s_b$ = 51 mm, a FGE up to 35% and a FPE up to 96% for spirlin and a FGE = FPE up to 86% for eels were observed in the present study with an e-HBR. Eel and spirlin differ significantly in terms of size but also morphology, so variations in behavior and in FGE and FPE were expected. Eels can align in various ways with respect to an electric field (i.e., S-shape or straight), whereas spirlin are more restricted by their body shape. Furthermore, there are large differences in size and muscle distribution between the two species. Since the electric field stimulates the muscles, fish reactions and the risks of injury are most likely highly related to muscle distribution However, despite the large differences, good protection rates could be achieved for both species.

With all current technologies of electrified racks, fish are mainly protected and less guided towards the bypass compared to non-electrified physical barriers or mechanical behavioral barriers. Although the consequences of passage delays are poorly understood, they are a critical component of fish population management [40], as fish may not reach their spawning habitats on time due to the cumulative delays at multiple HPPs.

As observed in the present study, fish injuries may occur even with pulsed direct current and low voltages, depending on the electrode arrangement. Prior to the installation of e-HBRs or other electric barriers at HPPs, suitable pulse patterns and field strengths need to be determined for different fish species, life history stages and thus size classes.

While the present study focused on fish protection with e-HBRs, there are several other aspects to be considered for a prototype application, which are briefly mentioned in the following. Permanent clogging of HBRs is typically prevented with automated rack cleaning machines [8]; to avoid a short circuit, these need to be linked to the electric system such that power can be shut off during cleaning. The electrification setup of the present study requires a cathode downstream of the e-HBR, which needs to be designed with sufficiently large spacing to prevent clogging and trapping of fish between the electrodes. Within the present experiments, the metal mesh (cathode) did not affect the velocity field upstream of the rack and it did not physically block fish. Increasing the spacing of the cathode in prototype situations is therefore not expected to reduce FPE and FGE. To avoid the clogging problem at the cathode, an e-HBR with individual bars electrified alternately as anode and cathode may be used. It would also be necessary to secure the area around the barrier with warning signs and fences to protect pedestrians.

The present study was carried out in a laboratory flume which allowed for studying the fish behavior in a controlled environment. However, the results have to be carefully interpreted and the findings of this study cannot be directly transferred from a laboratory setting to an upscaled prototype setting due to abiotic and biotic factors, which could not be mimicked with the current flume setup. This includes variations in environmental conditions, differences in the fish biology and geometric restrictions. Environmental conditions encompass parameters such as turbidity, floating debris, changes in light conditions, discharge, conductivity or water temperature, which are often related to seasonal patterns and were not varied within the present live fish tests. Metal elements close to the bypass or the rack may also influence the electric field, leading to a deterrence effect in front of the bypass or gaps in the electric field in front of the rack. The fish biology includes the natural behavior of different fish species and their life stages, which can differ between laboratory experiments due to stress from experimental handling, unnatural environment, small number of individuals affecting the schooling behavior or the absence of predators. The geometric restrictions include the dimensions of the flume, the HBR and the bypass. If an e-HBR is installed at a pilot HPP, an extensive monitoring campaign is indispensable to verify functionality.

5. Conclusions and Outlook

The key findings of this work include:

• The HBR with $s_b$ = 20 mm protected and guided all tested spirlin (FPE = FGE = 100%). For the e-HBR with $s_b$ = 51 mm a FPE of 96% was achieved, which is comparable to the HBR with $s_b$ = 20 mm. The FGE was significantly reduced for the e-HBR (20 mm and 51 mm) compared to the HBR (20 mm). This is due to the electric field in front of the rack and an unfavorable electric field in front of the bypass.
• Eels were well protected with the HBR with $s_b$ = 20 mm (FGE = 90%, FPE = 95%). The e-HBR with $s_b$ = 51 mm had a high FPE and FGE of 86% comparable to the HBR with $s_b$ = 20 mm. For a bar spacing of $s_b$ = 20 mm, no significant differences in FGE and FPE could be observed between the HBR and e-HBR. Eels may therefore have reacted less sensitively to the electric field in front of the bypass or were more motivated to move downstream than spirlin.
• The reactions of fish to electric fields with a horizontal voltage gradient as used in this study depends on their swimming orientations. To avoid this, a setup creating voltage gradients in vertical direction may be considered in the development of further electrification schemes for fish guidance racks.
• The electrode setup studied here leads to an electric field strength up to 0.35 V/cm at the transition from the barrier to the bypass, which compromises bypass acceptance. The setup therefore needs to be improved in further studies.
• The e-HBR caused fish injuries even for a low applied voltage of $U_e$ = 38 V due to the electrode arrangement, which led to maximal voltage gradients of $E\approx$ 2.1 V/cm downstream of the rack. Future studies should therefore report not only the applied voltage but also the voltage gradient and applied pulse pattern in order to make different studies comparable. In addition, the electric field should preferably be quantified by measurements or simulations.

These findings show the potential to combine a horizontal bar rack with a large bar spacing and a low voltage electric field while still maintaining good fish protection efficiency. Although high protection efficiencies were demonstrated, the presented electrification setup is not recommended for installation at a pilot HPP.