Limiting downstream dispersal of invasive carp egg surrogates using a laboratory-scale oblique bubble screen

Abstract

We conceptualize and test a non-intrusive barrier, comprised of an oblique bubble screen (OBS) oriented at an angle to the mean flow, to prevent the downstream dispersal of invasive carp egg surrogates. Three surrogates of different densities and diameters were tested. Secondary flows created by the OBS were tuned to redirect surrogate eggs to facilitate their capture. Surface particle image velocimetry and acoustic Doppler velocimetry were used to characterize secondary flows. We assessed the influence of airflow rate, OBS angle, mean flow velocity, and surrogate density on particle redirection. In general, redirection efficiency improves by increasing the OBS angle with respect to the cross-section. At a mean flow velocity of 0.75 metres per second (m/s), the OBS system redirected up to 60% (%) of positively buoyant particles (specific gravity SG = 0.9, and diameter d = 7.09 millimetres [mm]) and 40% of semi-buoyant particles (SG = 1.001, d = 3.1 mm). Negatively buoyant particles (SG = 1.04, and d = 5.90 mm) were redirected by the physical structure of the diffuser rather than by OBS-induced flow. The study shows that an OBS system can be used to effectively redirect carp-egg surrogates over a wide range of particle sizes and densities, allowing for selective targeting of undesired particles in streams.

Keywords:

• Invasive carp
• bubble screen
• non-physical barrier
• particle redirection

1. Introduction

Invasion of non-native biological species can have harmful effects on native species, local ecosystems, and the economy of the regions linked to that ecosystem (Pimentel et al. 2005; Lodge et al. 2006). Globally, fish species have been introduced in high numbers into non-native regions, and, as a result, freshwater aquatic ecosystems are among the biomes most affected by non-native invaders (Gozlan 2008; Bernery et al. 2022). In North America, nearly 40% of fish species are categorized as imperilled (Jelks et al. 2008) and non-native species are the second highest contributor to this imperilment (Wilcove et al. 1998). Many of the introductions of invasive fishes have occurred as a by-product of interventions in aquaculture and agriculture, primarily driven by economic interests (Courtenay and Williams 1992; Milardi et al. 2020). For instance, four fish species native to Asia were introduced in the United States in the 1960s and 1970s, mainly for controlling primary producers or mollusks in aquaculture ponds or sewage treatment (Mitchell and Kelly 2006; Kolar et al. 2007; Kelly 2011; Kelly et al. 2011). These species, namely grass carp (Ctenopharyngodon idella), silver carp (Hypophthalmichthys molitrix), bighead carp (Hypophthalmichthys nobilis), and black carp (Mylopharyngodon piceus), make up the four major aquaculture carps, known in the United States as invasive carp. These are important food fishes, both economically and culturally, in Asia and in many other places (Kočovský et al. 2018). However, in some areas where they have been introduced, these carp have been identified as highly destructive invasive species (Kolar et al. 2007). The abundance of invasive carp has resulted in the reduction of zooplankton, along with a decline in the condition of native planktivorous fishes in the upper Illinois River (Irons et al. 2007; Sass et al. 2014; Pendleton et al. 2017). Also, with the presence of invasive carp, a reduction in populations of native sport fish has been observed in the upper Mississippi River due to competition for zooplankton between early life stages and invasive carp (Chick et al. 2020).

Deterrents that prevent movement have been proposed to control invasive carp populations and prevent dispersal to new aquatic ecosystems, as removing invasive fishes following their establishment is almost impossible (Leung et al. 2002). Preventing additional spread and movement is particularly important for invasive carp in the upper Mississippi River Basin because of the linkage to the Great Lakes ecosystem through the Chicago Area Waterway System. As such, controlling populations of invasive carp is among the five focus areas identified in the Great Lakes Restoration Initiative (Leung et al. 2002; USFWS 2010).

Different technologies have been developed and implemented to control the movement of invasive carp with the ultimate objective of stopping them from invading new ecosystems (Cupp et al. 2021). Progression of deterrent technologies generally begins in the laboratory and, if successful and feasible, transitions to field-scale trials before application. Examples of deterrent technologies tested in the laboratory and/or field include electrical barriers (laboratory—Holliman et al. 2015; field—Moy et al. 2011), bubble barriers (laboratory—Zielinski and Sorensen 2016; field—Zielinski and Sorensen 2015), acoustic barriers (laboratory—Vetter et al. 2015, 2017; Murchy et al. 2017; field—Riesgraf et al. 2022; Brey et al. 2023), strobe light barriers (laboratory—Dennis and Sorensen 2020), carbon dioxide (CO₂) (laboratory—Kates et al. 2012; Cupp et al. 2017; field—Zolper et al. 2022; Raboin et al. 2023), and ensonified bubble barriers (laboratory—Dennis et al. 2019; Feely and Sorensen 2023; field—Fritts et al. 2023). Special attention has been given to non-physical deterrents (Noatch and Suski 2012) that can stop or deter fish passage while still safely permitting navigation and other activities in a waterbody. An electric barrier, tuned to immobilize fish when they attempt to swim upstream through the electric field, has been deployed in the Chicago Area Waterway System to prevent the upstream movement of aquatic invasive species between the Mississippi Basin and the Great Lakes Basin (Moy et al. 2011). However, Parker et al. (2016) found that fish were more abundant immediately downstream of the electrical barrier, even at the peak voltage, and were observed to be persistently challenging the barrier, posing a risk of a breach in case of any discontinuity in the power supply. Some additional drawbacks of electric barriers include a reduced efficacy at deterring small fish, high energy consumption, safety risks for humans, unintended threats to native species, and potential passage of fish during disruption of the electrical field by the movement of metal barges (Dettmers et al. 2005; Davis et al. 2016, 2017; LeRoy et al. 2019). Barriers involving high-intensity light have shown success in stopping juvenile bighead carp in dark and dimly lit environments in the laboratory (Dennis and Sorensen 2020). However, the efficacy of light-based barriers depends on the species and ambient conditions, and light barriers can be an effective deterrent where depth and turbidity allow for light to penetrate the entire water column. Barriers based on acoustic stimuli (underwater acoustic deterrent system [uADS]) and their combination with bubble curtains showed high efficacy in deterring invasive carp in a laboratory flume (Vetter et al. 2015; Murchy et al. 2017; Vetter et al. 2017; Dennis et al. 2019). However, broadband sound alone was not effective in deterring common carp when tested in the field at a navigation lock on the Mississippi River, although the study was limited, and the authors identified a range of factors that may have affected the efficacy of the study (Riesgraf et al. 2022). In a pair of ongoing field studies, the interim results from multi-year deployments of a uADS system at Lock and Dam 19 on the Mississippi River and a multimodal deterrent system at Barkley Lock and Dam on the Cumberland River have both shown a reduction in the upstream passage of the invasive carp by about one-half (Brey et al. 2023; Fritts et al. 2023). A system of sound projected through a bubble curtain was capable of blocking more than 90% of bighead carp, silver carp, and grass carp in a laboratory flume and showed no signs of habituation (Feely and Sorensen 2023). In addition to the stated limitations, all the barrier technologies mentioned above focus on behavioural responses of juvenile or adult fish and rely on fish choosing to actively avoid specific deterrent stimuli (Moy et al. 2011). However, these stimuli fail to target early life stages, such as eggs and larvae, when organisms have no or a reduced ability to choose whether to avoid or follow a specific feature, and have movements directed by flow patterns. Thus, additional barrier types, with the ability to overcome weaknesses in existing barriers, would be beneficial to deter the movement of invasive carp and prevent further spread.

Harvesting of adult carps is an alternate approach used to suppress populations and prevent additional expansion of the current population (Altenritter et al. 2022). This intense effort using commercial harvesters is expensive for a large aquatic ecosystem and needs sustained efforts, often through contracted harvesting or harvest incentives (ACRCC 2018). In addition, an effective population control strategy harvests invasive carp at multiple life stages, including eggs and larvae, to effectively suppress population growth (Tsehaye et al., 2013). Control technologies that target early life stages of fish (e.g. non-motile eggs and larvae) have not been well developed and therefore represent a key gap in population-control capabilities. Together, existing strategies for deterring the spread of invasive fishes not only have inherent weaknesses that leave them vulnerable to fish passage but also fail to target early life stages that are critical for population control. The development of new barrier technologies capable of controlling the movement and abundance of eggs and larvae, could therefore contribute to effective control of populations of invasive fish and prevent additional spread.

We present a novel approach to reducing the survival of early life stages of invasive carp that relies on secondary flows generated by a series of oblique bubble screens (OBS). While the visual and audible presence of the bubble screens can deter upstream passage of adult carp (Zielinski and Sorenson 2016), we examine the ability of secondary flows generated by an OBS to control the downstream drift of surrogate eggs. An OBS can be considered a special case of multiphase plumes, consisting of air and water (Tekeli and Maxwell 1978). The inclination of a linear diffuser with respect to the cross-section of the flume can increase complexity. Air bubble screens have been used for different applications, including, but not limited to, aeration of sewer overflows, enhancing mixing in lakes, preventing shock waves, and restricting oil spills (Ditmars and Cederwall 1974; Buscaglia et al. 2002; García and García 2006). Many of these studies developed predictors for velocities and width of the zone of influence of the bubble screens (Kobus 1968; Ditmars and Cederwall 1974; Brevik 1977; Tekeli and Maxwell 1978; Socolofsky 2001). Recent applications used bubble screens at an oblique angle to the flow for removing floating trash at laboratory- and field-scales (Spaargaren 2018; Zhang et al. 2022). Presence of crossflow in the form of mean flow in the flume, and the oblique alignment of the linear diffuser with the flow, makes the OBS more complicated compared to typical two-phase plumes. The interaction of the OBS-induced currents with the crossflow distorts the bubble curtain and tilts it in the flow direction. Spaargaren (2018) studied a laboratory-scale bubble barrier system aimed at redirecting floating plastic trash of different shapes and densities. They tested the redirection of drifting plastic objects by varying the angles of diffusers, crossflow velocities, and airflow rates, using empirical relations (Bulson 1968) and momentum balance equations to predict the bubble screen-induced lateral velocities. The first field-scale prototype of the bubble barrier was installed in Amsterdam, Netherlands, in 2019 to collect the floating plastic trash in a canal (TGBB 2019). The installation can effectively redirect the trash to a floating passive collection bin for removal. In a recent study, Zhang et al. (2022) used a bubble screen to study redirection of positively buoyant plastics (with diameters ranging in size from 2–4 millimetres [mm] and 80–450 micrometres [μm]). Their results showed a high variability in trapping efficiency based on flow conditions for two different screen configurations (linear and chevron-shaped diffusers).

Physical properties of live invasive carp eggs are of particular interest in this study, which allows us to replicate flow-particle interactions using surrogates. Although water-hardened invasive carp eggs are slightly denser than water, they can remain suspended in turbulent river flows and survive until hatching (Yih and Liang 1964; Garcia et al. 2015; George et al. 2018). Newly hatched larvae can swim vertically (Chapman and George 2011) but lack horizontal swimming capabilities until the gas bladder inflation stage. Therefore, pre-gas bladder inflation invasive carp are subject to the hydrodynamics of streams, making hydraulic-based control efforts attractive.

The aim of the current study is to quantify the potential of an OBS for redirecting particles of different densities and diameters representative of invasive carp eggs to develop a novel control technology able to target invasive carp at early life stages. The novelty in this study lies in the application of secondary flows induced by the bubble screen as the only stimuli for redirection of the invasive carp eggs to aid with population suppression. We hypothesize that if an OBS can be tuned to redirect the positively buoyant, negatively buoyant, and the semi-buoyant particles used in this study, it may be capable of redirecting the semi-buoyant, water-hardened invasive carp eggs that fall within the range of densities tested.

2. Methods

2.1. Facilities

Experiments were performed in an 11-metre (m) long recirculating flume (Figure 1(a,b)) located at the Ecohydraulics and Ecomorphodynamics Laboratory (EEL) at the University of Illinois Urbana-Champaign. A rotating paddle wheel drives flow, making it suitable for testing live eggs, larvae, or egg surrogates without damage that could be incurred from the use of a centrifugal water pump. The flume width (B) is 0.77 m and water depth (H) was kept at 0.4 m for all tests. Flow straighteners were removed from the flume to avoid any potential hindrance to particle transport during experiments. The OBS diffuser was made of flexible 2.54-centimetre (cm) (1-inch [in]) outer diameter aeration tubing (Pentair Aquatic Eco-Systems, Inc., Apopka, Florida, https://pentairaes.com/). The OBS diffuser tube was attached to a steel pipe of 1.27-cm (0.5-in) nominal diameter that keeps the diffuser rigid and fixed at the flume bed. A transition made of 1.9-cm (0.75-in) nominal diameter split-PVC pipe was attached to the upstream side of the diffuser to prevent particle entrapment between the diffuser and the bed (Figure 1(a)). A reciprocating air compressor (1471-Watt [2-Horsepower] motor with 114-liter [30-gallon] storage tank) was attached to the diffuser, with a gate valve controlling total air discharge. Total airflow rate (Q_a) through the diffuser was measured with an inline air flowmeter (King Instrument Company, Inc., Garden Grove, California, https://kinginstrumentco.com/). A constant pressure of 2.07 × 10⁵ pascals (Pa) (30 pounds per square inch [psi]) was maintained at the outlet of the compressor in all the tests. The OBS was tested at three different inclination angles (θ = 30 degrees [°], 45°, and 60° as seen in Figure 1(a)) to define the effect of angle on particle redirection. The inclination angle is measured with respect to the y-axis (perpendicular to the flow) as the reference (θ = 0°). Total length of the diffuser was adjusted depending on the angle to cover the entire width of the flume. Total airflow in the diffuser(s) was varied to keep the airflow rate per unit length (q_a) constant for all three inclination angles. For each given airflow and inclination angle, diffusers were tested at three nominal mean channel velocities (U₀ = 0.23, 0.45, and 0.75 metres per second [m/s]). This velocity range is generally consistent with observed mean cross-sectional velocities in the Sandusky River downstream from known grass carp spawning areas at spawning flows (Murphy and Jackson 2013; Embke et al. 2019; Jackson et al., 2019; Heer et al. 2020). Velocity profiles for the undisturbed incoming flow are available in the Supplementary Material.

Figure 1. (a) Sketch of the setup in the racetrack flume. Cross section of the diffuser system (A-A’ in Figure 1(a)). Aeration tubing is attached to a steel pipe and a ramp made of split polyvinyl chloride (PVC) pipe (all dimensions in mm). (b) Picture for the overall setup focusing on the area around the diffuser seen from the downstream end of the oblique bubble screen (OBS) (photo credit: V. Prasad, UIUC); notice the surfacing bubble screen, mounted acoustic doppler velocimeter (ADV), and high-speed camera used for surface particle image velocimetry (PIV; red rectangle).

2.2. Hydrodynamics

2.2.1. Acoustic Doppler velocimetry

Acoustic Doppler velocimeters (ADV) (Nortek Vectrino, Nortek USA, Boston, Massachusetts, https://www.nortekgroup.com/) were used to collect point measurements of velocity. Table 1 shows velocity data for the incoming flows at three different paddle frequencies. The mean incoming velocity (U₀) is vertical-average of streamwise velocity profile, and all the near bed quantities are evaluated at a height above the bed [z] of 1.5 cm at three velocities (U₀ = 0.23, 0.45, and 0.75 m/s), all within a fully turbulent regime. Detailed three-dimensional hydrodynamics were acquired for two different diffuser configurations (data acquisition locations shown in Figure 2): (a) OBS perpendicular to the mean flow (θ = 0°) without crossflow (U₀ = 0), and air flow rate Q_a = 5.10 cubic metres per hour (m³/h); 3 standard cubic feet per minute [scfm]); and (b) OBS with θ = 60°, U₀ = 0.23 m/s, and Q_a = 5.10 m³/h (3 scfm).

Figure 2. A schematic of the data acquisition using acoustic doppler velocimeter (ADV) while the diffuser is in operation. (a) ADV measurements with no crossflow (U₀ = 0). Each point denotes the location of measurement on the Central plane. (b) ADV measurements with mean flow. Each asterisk denotes the location of a vertical profile of eight points in the vertical. Surface particle image velocimetry (PIV) field of view shown in a green rectangle. The solid black line represents the diffuser, and the dotted grey line shows the approximate location of the bubbles when they approach the water surface [cm, centimetres].

Table 1. Summary of bulk flow parameters corresponding to the three nominal flow velocities (U₀), including near-bed turbulent kinetic energy (TKE_bed) and turbulence intensity percent (TI_bed × 100), Reynolds number (Re), Froude number (Fr), and shear velocity (u_*) [m, metres; cm, centimetres; s, seconds; Hz, Hertz].

No.	U0 (m/s)	Paddle frequency (Hz)	TKEbed (cm^2/s^2)	TIbed (%)	Re (×10^5)	Fr	Shear velocity at the bed (cm/s)
1	0.23	7	7.8	12.50	0.92	0.116	1.22
2	0.45	14	24.4	11.79	1.8	0.227	2.15
3	0.75	28	55.6	10.73	3.0	0.379	3.25

To characterize OBS-induced hydrodynamics, velocity data were collected over a grid with spacing of 20, 10, and 5 cm in the longitudinal (x), transverse (y), and vertical (z) directions, respectively (Figure 2). The topmost location in each profile had to be shifted 3 cm below the planned location of measurement due to the space constraints and water surface fluctuations. Therefore, the profiles are measured at z = 5, 10, 15, 20, 25, 30 and 32 cm. The origin, x = 0, corresponds to the upstream end of the bubble screen as it appeared on the surface. Measurements covered a length of 100 cm in the longitudinal direction at six cross-sections along the entire width. Data were acquired for 40 s (s) at 120 Hertz (Hz) at each point. A convergence analysis showed that streamwise velocity converged within a ± 5% error in just 20 s, even at locations with the highest concentration of bubbles. A similar convergence is achieved for root mean square values of velocity fluctuation (u’_rms) in 17 s. ADV data were filtered using a kernel density based on a despiking algorithm (Islam and Zhu 2013) to remove spikes introduced by the bubbles.

Velocity fluctuations (u’) are obtained by Reynolds decomposition and are used to calculate the u’_rms values. (1A) $$u^{\prime }=u-U$$(1A)

Where u is streamwise velocity time series and U is time-averaged streamwise velocity. Similarly, v’ and w’ were calculated for the transverse and vertical components of velocities, respectively. (1B) $${u^{\prime }}_{\mathit{\text{rms}}}=\sqrt{\overline{({u^{\prime })}^2}}$$(1B)

Near-bed turbulence intensity (TI_bed), measured at z = 1.5 cm, is calculated as: (2A) $${\mathit{\text{TI}}}_{\mathit{\text{bed}}}=\frac{u_{\mathit{\text{rms}}}^{\prime }}{U_0}$$(2A)

Turbulent kinetic energy near the bed (${\mathit{\text{TKE}}}_{\mathit{\text{bed}}}$), measured at z = 1.5 cm, is calculated as: (2B) $${\mathrm{T}\mathit{\text{KE}}}_{\mathit{\text{bed}}}=0.5\bullet \left(\overline{{\mathit{u}}^{\prime 2}}+\overline{{\mathit{v}}^{\prime 2}}+\overline{{\mathit{w}}^{\prime 2}}\right)$$(2B)

Bulk Reynolds number, Re, is calculated as: (3) $$\mathit{\text{Re}}=\frac{U_{0}D}{\nu }$$(3) where D is hydraulic depth (same as the flow depth H for a rectangular channel), ν is kinematic viscosity of water at 20 degrees Celsius (^oC), and U₀ is calculated using a velocity profile upstream of the diffuser when the diffuser is off.

All the flows are subcritical, Froude number (Fr) <1, as defined by: (4) $$\mathit{\text{Fr}}=\frac{U_0}{\sqrt{\mathit{\text{gD}}}}$$(4) where g is acceleration due to gravity.

Shear velocity $u_{*}$ is calculated as: (5) $$u_{\mathrm{*}}=\sqrt{\frac{{\tau }_b}{\rho }}$$(5) where ρ is density of water at 20 °C (998.2 kilograms per cubic metre [kg/m³]) and bed shear stress (τ_b) is calculated as: (6) $${\tau }_b=C_1\rho \left[0.5\left(\overline{u^{\prime 2}}+\overline{v^{\prime 2}}+\overline{w^{\prime 2}}\right)\right]$$(6) where C₁ is a constant (0.19) from (Thompson et al. 2003). All relevant flow parameters for all tests are summarized in Table 1. The OBS system is intended to be used in the field for subcritical flows only, therefore, we have tested the OBS system only for subcritical flows in the laboratory (Fr < 1 for all three velocities).

2.2.2. Surface particle image velocimetry

Secondary flows generated by the OBS resulted in a flow field with high spatial variability. We used surface particle image velocimetry (PIV) to measure the flow field on the water surface using bubbles on the surface as tracers. The optical properties of the bubbles differ from the rest of the water surface momentarily when air trapped in the bubbles is about to be released at the surface. Once bubbles are on the surface, they travel along with water parcels as tracers. A high-speed camera (Edgertronic SC1, Sanstreak Corp., Campbell, California, https://edgertronic.com/) was set up 2 m above the water surface, looking downwards. Images were acquired at a sampling frequency of 120 Hz with a time between images of 8.3 milliseconds (ms). The field of view for PIV captured a longitudinal section of ∼0.8 m with the whole width of the flume. Images were pre-processed to remove background noise and analysed using the open source PIVlab software (Thielicke and Sonntag 2021).

2.3. Surrogate egg particles

We used three types of particles as surrogate eggs: (1) positively buoyant (PB) plastic spherical particles, d = 7.09 mm, specific gravity (SG) = 0.9 (Cospheric LLC, Somis, California, https://www.cospheric.com/); (2) negatively buoyant (NB) plastic spherical particles, d = 5.90, SG = 1.04 (Cospheric LLC); and (3) semi-buoyant (SB) hemispherical particles made of soy wax, mean d = 3.50 mm, SG ≈ 1.001 (DPS Wax, Inc., Kansas City, Kansas, https://www.dpswax.com/).

We characterized the particles by measuring their settling speed. Three randomly selected particles were dropped in a settling column filled with water at 23 °C and recorded with a high-speed camera. A Matlab (The MathWorks, Inc., Natick, Massachusetts, https://www.mathworks.com/) subroutine was written to calculate the instantaneous settling speed of the particles using consecutive frames from the video, obtaining terminal settling speed once the instantaneous settling speed converges to a constant value (e.g. Prada et al. 2018). Particle diameters, measured with Vernier callipers, and settling speeds are shown in Table 2. The SG range of the tested particles is generally representative of invasive carp eggs, as summarized in Table 3 (George et al. 2017). The settling speed of the hemispherical SB particles was 0.5 cm/s. Because hemispherical particles of a given diameter and density settle at a lower settling speed compared to spherical particles, we estimated the settling speed of a perfectly spherical particle of the same diameter and density using the models given by Haider and Levenspiel (1989). The settling speed of a sphere of the same material and diameter was estimated to be 0.7 cm/s compared to 0.5 cm/s for hemispherical particle (Haider and Levenspiel 1989). Settling speeds of SB particles of both the shapes compare well with the settling speed range of the live eggs of invasive carp as reported by George et al. (2017) (also reported in Table 3). Of all the particles tested, the SB soy-wax particles best represented the size, specific gravity, and settling speed of invasive carp eggs.

Table 2. Characteristics of the particles used as surrogates [mm, millimetres; cm/s, centimetres per second].

No.	Particle type	Mean diameter (mm)	Mean settling speed (cm/s)	Specific gravity	Shape, colour
1	Positively buoyant	7.10	Not applicable	0.9	Spherical, white
2	Negatively buoyant	5.90	7.2	1.04	Spherical, fluorescent green under UV light
3	Semi-buoyant	3.50	0.5	∼1.001	Hemispherical, white

Table 3. Physical properties of water-hardened invasive carp eggs determined by George et al. (2017).

No.	Species	Diameter (mm)	Terminal settling speed (cm/s)	Specific gravity
1	Grass carp (Ctenopharyngodon idella)	4.5 ± 0.6	0.74 ± 0.16	1.00068 ± 0.00052
2	Silver carp (Hypophthalmichthys molitrix)	4.7 ± 0.4	0.69 ± 0.07	1.00044 ± 0.00008
3	Bighead carp (Hypophthalmichthys nobilis)	5.8 ± 0.5	0.78 ± 0.07	1.00043 ± 0.00019

Note: Specific gravity was calculated for different temperatures using the data and methodology presented in George et al. (2017) [mm, millimetres; cm/s, centimetres per second].

2.4. Particle redirection and capture

We conducted three replicates for each configuration, with 50 particles released in each trial. Particles were released en masse at a fixed location in the parallel branch of the racetrack flume (Figure 1(a)). The presence of bubbles made it difficult to visualize the drifting particles, preventing the use of particle tracking analysis on the surrogate particles. To assess their trajectories, a custom-made 1.6-mm mesh-size net (Pentair Aquatic Eco-Systems, Inc., Apopka, Florida, https://pentairaes.com/) composed of eight cells (Figure 3(c,d)) was placed downstream of the OBS to capture the particles and determine their location after passing through the OBS (Figures 1(a) and 3). Two equal rows covered a total depth of 50 cm of the flume (25 cm each row). Because the depth was kept constant at 40 cm, only 15 cm of the top row was submerged. The numbers of particles in each cell were counted after each trial to calculate the percentage of particles captured in each cell. An average percentage from the three replicates was used as a metric of capture efficiency. Three airflow rates were tested for each water velocity: zero, low, and high airflow rate (Table 4). In zero-airflow replicates, flow is modified only by the physical structure of the diffuser. Airflow per unit length of the diffusers were maintained the same to ensure a valid comparison among different lengths (summarized in Table 4).

Figure 3. (a) Schematic for proof-of-concept experiments using positively buoyant particles including the four general trajectories of particles based on their final location (A, B, C, D). (b) Photo of semi buoyant (SB) particles being redirected as they pass through the oblique bubble screen (OBS). (c) Picture of the eight-cell net used for particle capture. (d) Sketch showing dimensions of net cells. The blue dashed line in (c) and (d) denotes the water surface level. [θ, diffuser angle; B, channel width; cm, centimetres] (photo credits: V. Prasad, UIUC).

Table 4. Airflow rates for diffusers of different angles (θ) and lengths, where Q_a is the total airflow rate and q_a is the airflow rate per unit length of diffuser [m, metres; h, hours].

No.	θ (degrees)	Diffuser length (m)	Qa (m^3/h) for [high, low] airflows	qa (m^3/h/m) for [high, low] airflows
1	30	0.82	4.9, 1.5	6.0, 1.8
2	45	1.00	6.0, 1.8	6.0, 1.8
3	60	1.42	8.5, 2.6	6.0, 1.8

We defined effective redirection in terms of the percentage of particles captured in the cells on the right bank (outer bank, at the downstream end of the OBS), namely cells A4 and B4. Effective redirection is used as the primary metric to evaluate the efficacy of the different configurations of the OBS in this study.

3. Results

3.1. Hydrodynamics of a bubble screen without a crossflow

The simplest setup was a bubble screen covering the entire width of the flume, placed perpendicular to the flume (θ = 0°) and without crossflow (U₀ = 0). This case was used to characterize the bubble screen in terms of the surface velocity, bubble slip velocity, and extent of secondary flows in absence of crossflow for comparison to previous studies (Figure 2(a)). Data demonstrated nearly symmetrical secondary flows on both sides of the diffuser with maximum horizontal velocities observed near the surface (Figure 4). Bubble-curtain models predict a zone of horizontal velocity away from the diffuser (surface current) in the top one-fourth of the water depth (Ditmars and Cederwall 1974; Buscaglia et al. 2002; Spaargaren 2018). We compared the maximum surface velocity, $v_m,$ and depth of the surface current, $h$ (Equations (7)(7) $$\frac{v_m}{{\left({\mathrm{\text{gq}}}_{\mathrm{a}}\right)}^{\frac{1}{3}}}=1.46{\left(1+\frac{\mathrm{H}}{{\mathrm{h}}_0}\right)}^{-\frac{1}{3}}$$(7) and (8)(8) $$\frac{\mathrm{h}}{\mathrm{H}}=0.32\frac{{\mathrm{h}}_{\mathrm{o}}}{\mathrm{H}}\mathrm{\text{ln}}\hspace{0.17em}\left(1+\frac{\mathrm{H}}{{\mathrm{h}}_0}\right)$$(8) ), with the models developed by Bulson (1961): (7) $$\frac{v_m}{{\left({\mathrm{\text{gq}}}_{\mathrm{a}}\right)}^{\frac{1}{3}}}=1.46{\left(1+\frac{\mathrm{H}}{{\mathrm{h}}_0}\right)}^{-\frac{1}{3}}$$(7) (8) $$\frac{\mathrm{h}}{\mathrm{H}}=0.32\frac{{\mathrm{h}}_{\mathrm{o}}}{\mathrm{H}}\mathrm{\text{ln}}\hspace{0.17em}\left(1+\frac{\mathrm{H}}{{\mathrm{h}}_0}\right)$$(8) where H = water depth, h₀ = atmospheric pressure as head of water, and q_a is the airflow rate per unit length of the diffuser.

Figure 4. Mean velocity vectors generated by the bubble screen in the absence of crossflow with the diffuser aligned perpendicular to the flume (θ = 0°). the colour scales show mean velocity magnitude in centimetres per second (cm/s) in the x-z plane (u_xz).

A variable equivalent to v_m (in Bulson 1968) is defined as the velocity magnitude in x-z plane in our analysis. (9) $$U_{\mathit{\text{xz}}}=\sqrt{U^2+W^2}$$(9)

U_xz in absence of crossflow (θ = 0°, U₀ = 0) is shown in Figure 4, where maximum U_xz is seen near the water surface. A more quantitative comparison is done by using Equation (7)(7) $$\frac{v_m}{{\left({\mathrm{\text{gq}}}_{\mathrm{a}}\right)}^{\frac{1}{3}}}=1.46{\left(1+\frac{\mathrm{H}}{{\mathrm{h}}_0}\right)}^{-\frac{1}{3}}$$(7) (Bulson 1968) that gives a value of v_m = 0.386 m/s, whereas our data show U_{xz =} 0.358 m/s. The depth of surface current from Equation (8)(8) $$\frac{\mathrm{h}}{\mathrm{H}}=0.32\frac{{\mathrm{h}}_{\mathrm{o}}}{\mathrm{H}}\mathrm{\text{ln}}\hspace{0.17em}\left(1+\frac{\mathrm{H}}{{\mathrm{h}}_0}\right)$$(8) (Bulson 1968) is h/H = 0.31, compared to h/H between 0.25 and 0.37 in this study. Because the models are developed for no-crossflow conditions, OBS cases with a non-zero current cannot be compared using these predictors.

3.2. Hydrodynamics of the OBS in a crossflow

3.2.1. Flow field around the diffuser

We used ADV measurements to characterize flow around a diffuser for θ = 60° and U₀ = 0.23 m/s. In the presence of a crossflow, recirculation is generally absent in the first three cross-sections, x = 0, 20, and 40 cm (hereafter referred to as the undeveloped region), but was clearly present at x = 60, 80, and 100 cm (hereafter referred to as the developed region as shown in Figure 5). The mean flow yields a bubble screen tilted away from the vertical in the direction of the flow, with the degree of tilting increasing with flow velocity. Bubbles get advected in the streamwise direction as they rise vertically, resulting in a tilted bubble screen. Two independent factors enhance the degree of tilting – (1) the flow velocity is lower near the bed because of boundary layer effects, and (2) bubble vertical velocity decreases as the bubbles rise in the water column (Kobus 1968; Tekeli and Maxwell 1978). The secondary flows are not developed under the surface for almost the first half of the OBS as seen at the surface. The secondary flows result in an increased longitudinal velocity in the downstream wake of the OBS and a decreased longitudinal velocity on the upstream side of the OBS.

In the developed region, between x = 60 cm and x = 100 cm, a clear recirculation pattern is seen in the vector plots. Recirculation on both sides of the diffuser is asymmetric, unlike the perpendicular diffuser without crossflow (Figure 4), due to the oblique orientation of the diffuser and the limited space on the sides constrained by the flume walls. A vertical upward mean velocity component is observed at the core and in the developed part of the bubble screen, and a downward velocity component was observed at the wall. The downward velocity near the walls is enforced by the physical boundary and is a common feature for all cases with θ ≠ 0°.

A region of higher TKE was observed in the proximity of the bubbles (Figure 6). The values are substantially higher compared to the flow in the absence of the bubble screens (given in Table 1). Bubbles create a zone of high turbulence within the area of influence as seen in Figure 6. The zones of higher TKE are consistent with the secondary flows created by the mean flows shown in Figure 5. A substantial difference in the values of TKE was observed near the bubble screen indicating that the region away from the bubble screen is not affected by the turbulence generated by the OBS.

Figure 5. Mean velocities around the oblique bubble screen (OBS) in the y-z plane are shown using vectors. The colourmap shows streamwise mean velocity (U) as seen from the upstream (flow is into the page). The OBS system generated recirculation towards the downstream end of the diffuser [cm, centimetre].

Figure 6. Turbulent kinetic energy (TKE) calculated using acoustic doppler velocimeter (ADV) measurements for six cross-sections seen from upstream (flow is into the page). A zone of higher turbulence can be seen near the core of the bubble screen [cm, centimetre; s, second].

3.2.2. Flow field on the surface

The lateral velocity component at the surface is a key parameter controlling particle redirection. A ratio of the magnitudes of the lateral (V) and longitudinal (U) components, |V|/|U|, was used to quantify the lateral push a particle experiences as it drifts with the flow (Figure 7). The lateral push generated by the secondary flow is highest at the lowest velocity, U₀ = 0.23 m/s, and is concentrated on the upstream side of the diffuser. At the intermediate case, U₀ = 0.45 m/s, the high lateral push region was slightly weaker and centred at the top of the bubble screen, with resultant velocity vectors corresponding to the highest |V|/|U| aligned almost parallel to the bubble screen. At the highest velocity, U₀ = 0.75 m/s, the zone of highest |V|/|U| was still close to the location of the bubble screen on the surface. However, the magnitude of |V|/|U| was much lower compared to lower U₀. For all three cases, the airflow rate per unit length was kept constant (values given in Table 4). Achieving a lateral push (or |V|/|U|) near unity at higher mean flow velocities would require an increase in the airflow rate. The hydrodynamics on the surface can thus be used to gain insight into particle redirection, particularly for positively buoyant particles.

Figure 7. Mean velocity ratios |V|/|U| for three different mean flow velocities (U₀ = 0.23, 0.45, 0.75 metres per second [m/s]) at an airflow rate, q_a = 6 standard cubic metre per hour per metre length for the 60-degree diffuser [cm, centimetres].

3.3. Particle redirection

We evaluated the efficacy of the OBS based on the number of surrogate eggs redirected to a desired location in the flume (the right, outer wall in this study). Preliminary experiments identified four general trajectories of particles based on their final location: (A) zone of recirculation, (B) full redirection along the bubble screen, (C) partial redirection at the further downstream end of the flume after particles pass through the diffuser, and (D) particle trajectories unimpeded as they pass through the OBS (Figure 3(a)). These preliminary experiments served as a proof-of-concept to determine the viability of the diffusers, showing different redirection patterns over different flow conditions, with higher lateral push resulting in higher redirection efficiency (Prasad et al. 2022). The efficacy of OBS to redirect particles to the desired capture location was quantified using the average percentage of particles captured in each net cell for the three replicates conducted for each configuration for PB, NB, and SB particles, as shown in Figures 8–10, respectively.

Figure 8. Mean capture percentages for positively buoyant (PB) particles for different diffuser angles, mean flow velocities, and airflow rates (q_a, in cubic metres per hour per metre of diffuser [m³/h/m]): zero air flow (a), low air flow (b–d) and high air flow rates (e–g). A schematic of the net on top-left shows location of the cells with the flow going into the page. Error bars show the standard deviation in the three replicates for each configuration. Only one case was tried for the trial with zero airflow rate because all the particles stayed on the surface for all three angles of the diffusers [m/s, metres per second].

Figure 9. Capture percentages for negatively buoyant (NB) particles for different diffuser angles, mean flow velocities, and airflow rates (q_a, in cubic metres per hour per metre of diffuser [m³/h/m]): zero air flow (a–c), low air flow (d–f) and high air flow rates (g–i). Error bars show the standard deviation in the three replicates for each configuration [m/s, metres per second].

Figure 10. Mean capture percentages for semi-buoyant (SB) particles for different diffuser angles, mean flow velocities, and airflow rates (q_a, in cubic metres per hour per metre of diffuser [m³/h/m]): zero air flow (a), low air flow (b–d) and high air flow rates (e–g). Error bars are the standard deviation in the three replicates for each configuration [m/s, metres per second].

3.3.1. Positively buoyant (PB) particles

With no airflow, a higher percentage of particles in cells A1 and A2 for U₀ = 0.23 and 0.45 m/s was observed for the PB particles. However, we observed a different pattern for the high-velocity treatment, in which the largest percentage was captured in cell A3. A negligible percentage appears in the bottom row of the net (B1–B4) (Figure 8).

At U₀ = 0.23 m/s, all three inclination angles show nearly 100% of particles captured on the right bank (cells A4 and B4), for both high and low airflow rates. Although PB particles remained on the surface, in general, a high percentage of particles were pushed to the bottom row (cell B4) for U₀ = 0.23 m/s when they reached the downstream end of the OBS. This downward movement of PB particles shows that the OBS creates a recirculation zone downstream of the OBS strong enough to sink a majority of PB particles during low-velocity flow. High values of the standard deviation corresponding to cells A4 and B4 in Figure 8(a) show that this variation occurs within the same column (A4 and B4) and is arising only due to variation in the vertical recirculation, not because of uncertainty in lateral distribution. Trials with higher inclination angles (θ = 45° and 60°), however, showed consistent redirection with low standard deviations. For the medium mean flow velocity, U₀ = 0.45 m/s, a steady increase in particle redirection can be seen with increasing angle of the diffuser. However, for the highest velocity cases, U₀ = 0.75 m/s, we observed the highest redirection for θ = 45° (57%).

High airflow demonstrated almost 100% redirection to the right bank (A4 and B4 cells) for the slowest mean flow velocity at all three angles. More than half of the particles were captured in the second row (cell B4) for θ = 30° and 45° and U₀ = 0.23 m/s (Figure 8(e,f), respectively). Redirection increased for U₀ = 0.45 and 0.75 m/s as the inclination angle of the diffuser increased. We observed the highest percentage of particles on the right bank for high airflow and high inclination cases (Figure 8(g)) among all the cases for PB particles.

3.3.2. Negatively buoyant (NB) particles

NB particles resulted in a different distribution pattern compared to the PB surrogates. NB particles (SG = 1.04) sank after their release, and then rolled and saltated along the bed. The turbulence created by the diffuser, even at the high airflow rate, was not sufficient to resuspend the NB particles. Most particles were captured by the cell B4 for all three control trials (Figure 9(a–c)).

The redirection of NB particles for the control replicates was solely due to their interaction with the physical structure of the diffuser at the flume bed, and not due to recirculation caused by airflow injected into the flume. Due to its oblique placement along the flume bed, the diffuser structure guided the rolling NB particles towards the right bank. At the low airflow rate (Figure 9(d–f)), a reduction in the percentage of particles in cell B4 was observed. Even at the high airflow rate, the percentage of particles captured in B4 decreased for the 30° and 60° angles at all flow velocities. The only exception was the high flow velocity (U₀ = 0.75) at θ = 45°, where the percentage of particles trapped in cell B4 remained around 80%.

3.3.3. Semi buoyant (SB) particles

Due to their physical properties, SB particles were expected to behave the closest to the invasive carp eggs out of the three surrogates tested. The control replicate (no airflow) was tested for only θ = 60° (Figure 10(a)). Because the SB particles did not saltate or roll along the bed, repeating the control cases with the remaining two angles was not necessary. Most SB particles were captured by the cells in the top row of the net, towards the left bank (A1–A3) in the control test (Figure 10(a)). A higher redirection to cells A4 and B4 was observed for the low airflow, but only for the low flow velocity replicates (Figure 10(b–d)). The percentage of redirected particles increased with the inclination angle (θ). We also observed a sizeable portion of the SB particles pushed to the bottom row for low velocity for θ = 60° (Figure 10(d)). However, we did not observe any effect of low airflow for the 30° or 45° diffusers at higher velocities (Figure 10(b–c)). An improved particle redirection is observed for θ = 60° diffuser at 0.45 m/s (Figure 10(d)).

A higher airflow did not increase particle redirection at faster velocities for θ = 30° (Figure 10(e)). The θ = 45° diffuser demonstrated some increase, and θ = 60° resulted in a much higher particle redirection towards the desired target bank (cell A4) with higher airflow rates (Figure 10(f,g)). However, the increased airflow at low flow velocity has an opposite effect and reduces the particle redirection for θ = 45° and θ = 60° (Figure 10(f,g)).

3.3.4. Comparison of redirection efficiencies

A summary of particle redirection efficiencies for all configurations of the OBS is given in Figure 11. For the PB, redirection was observed to increase with increasing the airflow rate and higher inclination angle of diffuser. The pattern was substantially different for the NB particles. Higher redirection efficiencies were observed at low airflow rates compared to high airflow rates for the NB particles. The high airflow replicates with θ = 60° demonstrated the least redirection of the NB particles for all three velocities. However, θ = 60° diffuser demonstrated much more efficient redirection at low airflow compared to the high airflow replicates. A similar pattern was observed for θ = 30° and 45° when airflow rates are compared. The airflow created regions of high turbulence close to the diffuser that induced mixing and disrupted the streamlined movement of NB particles along the diffuser ramp. Redirection of NB particles is more sensitive to the airflow rate compared to the inclination angle of the diffuser.

Figure 11. Effective redirection as a function of diffuser angle, mean flow velocity, and airflow rate for all three types of particles for all the configurations of the oblique bubble screen (OBS). (a), (b) and (c) show the plots for positively buoyant (PB), negatively buoyant (NB), and semi buoyant (SB) particles, respectively. Effective redirection is defined as the average percentage of particles captured in the rightmost column of the net (A4 and B4 in Figure 3) [deg, degrees; m/s, metres per second].

Redirection of the SB particles did not show any straightforward ubiquitous patterns. Instead, the redirection efficiency was strongly dependent on the mean flow velocity and the inclination angle. The lowest mean flow velocity (U_o = 0.23 m/s) with low airflow rate for θ = 45° and θ = 60° resulted in the highest redirection towards the right bank of the flume. But for θ = 30°, the highest redirection was seen at the high airflow rate. At relatively higher mean flow velocities (U_o = 0.45 and 0.75 m/s), redirection efficiencies increased with the inclination angle and airflow. The pattern at higher velocities was found to be similar to the SB particles with high airflow rate and θ = 60° demonstrating the highest redirection efficiency.

3.3.5. Correlation between hydrodynamics and redirection

Lateral redirection of the particles at the water surface, generated by the OBS, is quantified in terms of the mean velocity ratio (|V|/|U|) and shown in Figure 7. The maximum values of |V|/|U| are found to occur close to the location of the OBS on the surface. A new parameter that is representative of maximum values of |V|/|U| is defined to establish a correlation between the hydrodynamics and particle redirection close to the surface. We defined this new hydrodynamic parameter, the bulk mean velocity ratio ($〈\left|\mathrm{V}\right|/\left|\mathrm{U}\right|〉$), calculated from the surface PIV results shown in Figure 7. Bulk mean velocity ratio is defined in Equation (10)(10) $$〈\left|V\right|/\left|U\right|〉=\mathit{\text{mean along}}\mathrm{ }x\mathrm{ }\mathit{\text{axis}}\left[90\mathit{\text{th percentile along}}\mathrm{ }y\mathrm{ }\mathit{\text{axis}}\mathrm{ }\left(\frac{\left|V\right|}{\left|U\right|}\right)\right]$$(10) as: (10) $$〈\left|V\right|/\left|U\right|〉=\mathit{\text{mean along}}\mathrm{ }x\mathrm{ }\mathit{\text{axis}}\left[90\mathit{\text{th percentile along}}\mathrm{ }y\mathrm{ }\mathit{\text{axis}}\mathrm{ }\left(\frac{\left|V\right|}{\left|U\right|}\right)\right]$$(10)

We have chosen 90^th-percentile instead of the maximum value of |V|/|U| to avoid any biases because of the errors in surface PIV results. We could see a correlation between ($〈\left|\mathrm{V}\right|/\left|\mathrm{U}\right|〉$) and redirection efficiencies of PB and SB particles (Figure 12). A higher value of $〈\left|\mathrm{V}\right|/\left|\mathrm{U}\right|〉$ resulted in better redirection efficiencies for all three angles. The mean velocity ratio was found to decrease with an increase in the flow velocity in the flume at the same airflow rate. Similarly, particle redirection was also found to decrease with higher downstream flow velocities if airflow was unchanged (also shown in Figure 11). Because the $〈\left|\mathrm{V}\right|/\left|\mathrm{U}\right|〉$ is calculated for the surface hydrodynamics and NB particles always stayed close to the bed, we did not expect the mean velocity ratio at the surface (|V|/|U|) to have any effect on the redirection of NB particles.

Figure 12. Correlation between the bulk velocity ratio (<|V|/|U|>; bars) and effective redirection for positively buoyant (PB) and semi buoyant (SB) particles (markers and trendlines) at high airflow rate. Error bars show standard deviation in three replicates for the selected configurations [deg, degrees; m/s, metres per second].

4. Discussion

The development of an OBS for redirecting the drifting early life stages of invasive carp could be a potentially transformative technique that can have important benefits for population control, mainly because it is a control technology that is independent of behavioural responses (e.g. larval light trapping) and could be used to target early life stages currently not well represented in population suppression efforts. Overall, our results demonstrate that specific characteristics of an OBS can directly change the trajectory of particles (surrogates for invasive carp eggs) of different densities and, therefore, provide guidance on the targeted redirection of particles or drifting early life stages in lotic ecosystems needed to assist in controlling invasive carp populations. More specifically, in the current study, the target location for the collection of eggs redirected by an OBS is at the trailing end where the bubble screen hits the surface, on the bank furthest downstream. The lateral push created by the OBS changed with velocity for a given airflow rate. High water velocity advects the bubble screen to a much further extent downstream of the diffuser’s physical location. The advection does not happen uniformly across water depth as the top of the bubble screen advects to a longer distance, tilting the bubble screen. Because of this, the surface currents generated by the bubble screen get weaker, resulting in a smaller lateral component at high flow velocities. Therefore, a higher airflow is required, in general, to create a comparable lateral push at a higher streamflow velocity. The location of a capturing mechanism would thus depend on mean flow velocity, assuming that diffusers are fixed at the bed. In laboratory tests, we had to change the location of the net when flow velocity changed, which could require adjustability of a collection system in a field deployment.

Particle density was critical to determining trajectory as particles interact with the OBS, with even slight differences in density resulting in contrasting results. Redirection patterns were noticeably different within the narrow range of SG of the particles (0.9–1.04) tested in this study. Any particles with SG lower than 0.9 would likely exhibit a behaviour similar to the PB particles. Spherical NB particles with SG higher than 1.04 are expected to be redirected along the physical structure of the diffuser, assuming the flow has enough energy to sustain rolling or saltation. A similar result would be expected for the eggs of any organism with densities in such a range. This can have important repercussions in the field application of an OBS where early life stages of non-target species, or other objects with similar physical properties, will exhibit a similar redirection. However, we note that most invasive carp spawning events occur later in the year (Kolar et al. 2007) than those of most fish species native to North America, especially those few with drifting eggs, and thus the timing of application of an OBS barrier could thus reduce non-target effects.

Airflow rate and secondary flows correlated positively, ultimately affecting particle redirection, but the effect of secondary flows on particle trajectory was not always straightforward. For the PB particles that always tend to remain on the surface, we see a higher redirection of particles towards our target collection point at higher airflows because of the stronger surface currents that resulted from enhanced recirculation. The redirection results showed a pattern similar to the findings of Spaargaren (2018) where the focus was only on the floating plastic objects. In contrast, increased airflow reduced particle redirection for NB particles by mixing them across the diffuser and thus preventing their rolling along the diffuser’s axis. For the SB particles, this trend was much more complex, and the optimal airflow rate to redirect SB particles towards the collection point was dependent on incoming water velocities such that increasing the airflow rate did not necessarily increase SB particle redirection. This result implies additional tuning to optimize an OBS for live eggs because of their lower settling speed relative to the particles tested currently is warranted (George et al. 2017). Because of secondary flows, even slight changes in the flow field might affect the trajectory of SB invasive carp eggs. The optimal airflow rate to maximize particle capture would depend on the settling speed for the SB surrogates. This is of particular interest for determining the optimal airflow rates needed to collect live invasive carp eggs that are also treated as semi-buoyant. An additional factor here is the hemispherical shape of the SB particles. A spherical particle of the same diameter and density will have slightly higher settling speed due to the lesser drag experienced by a perfect spherical shape. The slight difference in the settling speed can potentially have substantial difference in the redirection pattern in the presence of secondary flows.

NB particles were redirected solely by the physical structure of the diffusers along the bottom of the flume, and not due to recirculation of water due to the bubble screen. The likelihood of particle resuspension can be predicted by comparing the shear velocity at the bed and particle settling speed. If bed shear velocity exceeds settling speed, the turbulence generated at the bed can counter the settling and particles can remain in suspension (Rijn 1984; Prada et al. 2018). Shear velocity at the bed was lower than the settling speed of NB for all flow cases, indicating insufficient turbulence to resuspend the NB particles (Tables 1 and 2). While a strong recirculation was observed in the wake of the diffuser, which could keep NB particles temporarily suspended, eventually, all particles leave the recirculation zone and travel to the net.

An increased angle of the OBS resulted in increased redirection for PB and SB particles, mainly due to strengthened lateral velocity component. However, the cost of installation and operation also increases with the higher inclination angle due to additional materials and airflow required, resulting in a trade-off between the high inclination angle and the cost of installation and operation. For instance, the redirection efficiency of SB surrogates for U₀ = 0.75 m/s, at the highest airflow, shows a 10% higher redirection for θ = 60° compared to θ = 45°. But the length of the diffuser for θ = 60° was 1.4 times longer. Longer diffusers also require a higher amount of total airflow rate (Q_a) to maintain the same value per unit length airflow rate (q_a). A higher inclination angle will thus result into a higher cost of installation and operation. However, findings from Spaargaren (2018) show that, with a higher diffuser inclination, the total airflow needed for the same capture percentage of PB plastic was lower despite a larger diffuser. So, keeping airflow rate per unit length might not be the most optimal strategy to compare the effect of angles on particle redirection because of higher energy consumption. Future experiments could consider this possibility and include an inclination angle higher than θ = 60° at a lower airflow.

Configuration of an OBS is another aspect that warrants further research. We have conceptualized an OBS using a single diagonal diffuser and target location at the downstream end of the bubble screen. The target location would also change with the layout of the OBS. For instance, an inverted V-shaped or chevron configuration of diffusers would push drifting particles towards both banks or to the centre of the channel depending on whether the diffuser “V” was pointing upstream or downstream.

All the experiments in this study were conducted in a laboratory setup using a flume with a rectangular cross-section and mean velocity not exceeding 0.75 m/s. These conditions are different from those in the field whereby stream bottoms are irregular and banks often slope towards the surface. The fundamental approach for the field deployment of an OBS would stay the same even relative to the laboratory study described here. Non-uniform flows, irregular cross-sections, and bed forms in a natural channel impose additional challenges for field implementation of an OBS. Maintaining a uniform secondary flow across the entire width of the stream is challenging and will require sophisticated local controls to keep the hydrodynamics consistent. Moreover, an array of diffusers, rather than the single diffuser used here, might be needed in the field to create a redirection zone wide enough to keep semi-buoyant eggs in suspension for longer durations to facilitate redirection and collection. Testing the interaction of multiple diffusers placed in parallel is also important to see their combined effect on particle redirection.

In addition to the cross-section of the flume, water temperature for all the experiments was controlled in a temperature range of 21 ± 2 °C, but temperature fluctuations and gradients can be substantially higher in the field. Viscosity and density of water change with temperature, resulting in substantial changes in particle settling characteristics (Ahrens and Member 2000). Therefore, efforts related to field implementation would benefit from considering the effects of temperature differences on particle behaviour.

Initial studies on other applications of bubble screens relied on experiments and analytical models to predict velocities and width of the bubble screen (Kobus 1968; Ditmars and Cederwall 1974; Brevik 1977), while more recent studies could characterize the turbulence structures of different scales induced by the bubble screens (Bombardelli et al. 2003; García and García 2006; Bombardelli et al. 2010). Computational models capable of incorporating mass transfer and liquid chemistry for a large-scale two-dimensional domain were developed in recent years (Buscaglia et al. 2002; Oldeman et al. 2020). Because most field prototypes of the bubble screen have occurred at a much larger scale compared to the laboratory-scale models, scaling analysis of hydrodynamics becomes important (Tekeli and Maxwell 1978; Bombardelli et al. 2010), resulting in various relevant non-dimensional parameters suggested for scaling of bubble screens (Milgram 1983; Schladow 1992; Asaeda and Imberger 1993; Bombardelli et al. 2010). Scaling analysis is also important considering the expensive implementation of large field-scale prototypes. For instance, Bombardelli et al. (2010) showed a scaling analysis using results from a prototype, and a model with a geometrical scaling of 10:1 to predict a quasi-universal behaviour of bubble plumes. Following a similar approach, an intermediate scaled prototype of an OBS could be used as a nearly-field scale prototype for scaling analysis in future work using a model of appropriate geometrical scaling.

Another challenge of field implementation may arise due to presence of a mobile bed. Secondary flows created by bubble screens can change the morphology of a mobile sediment bed. Dugué et al. (2015) used a sediment bed (mean particle diameter d_m = 2 mm) and noticed that the changed morphology allowed for growth of secondary flows to more than double their extent on a rigid flat bed. The extent of their secondary flow, however, was found to be independent of the scale of the flow, for flow depth in the range 0.1 m to 5 m. This can be of particular importance in the field implementation of OBS, where the stream bed will be susceptible to erosion due to the secondary flows. OBS setup will require modifications to mitigate the unintended effects on sediment transport dynamics in natural streams.

An extension of this study investigating the efficacy of an OBS on redirection of live grass carp eggs using a similar laboratory setup has recently been completed and will be published in a separate article. While an OBS might also be useful for redirecting larvae, further research would be beneficial before any generalizations can be made due to their mobility and unknown behavioural response to the OBS. Grass carp larvae try to avoid regions of high turbulence, particularly areas with turbulent eddies of length scales similar to their size (Prada et al. 2021). These additional insights about their behaviour might be helpful in modifying the approach to study the effects of OBS on live larvae of invasive carp in future studies.

5. Conclusions

An oblique bubble screen (OBS) in flowing water produces a much more complex flow distribution than a bubble screen in a stagnant environment. The OBS can redirect nearly 100% of positively buoyant particles to a target collection point at low (0.23 m/s) and intermediate (0.45 m/s) mean channel velocities, and up to 60% at the highest flow velocity evaluated (0.75 m/s). The OBS redirection efficiency was lower for semi buoyant particles—up to 80% at intermediate flow velocity and up to 40% at high flow velocity—indicating a dependence on particle density and settling speeds. Redirection efficiency of semi buoyant particles had an optimal airflow rate corresponding to each flow velocity and angle of diffuser, showing the relevance of both mean flow, air flow rate, and diffuser configuration on redirection efficiency, which can be captured with velocity ratios such as the proposed |V|/|U| ratio at the surface. The airflow did not affect the redirection of negatively buoyant particles, as they are only guided by the physical structure of the diffuser along the bed. The observed redirection patterns of three classes of surrogate eggs within a narrow range of densities demonstrates the potential of OBS as a tool for controlling the dispersal of particulate matter in streams. While most effective for positively buoyant particles, this technology has potential for redirection of semi buoyant eggs of invasive fish, potentially providing a novel technology to target early life stages of invasive fishes and supplement existing control efforts.

Acknowledgments

Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. We thank Andrew Waratuke, lab manager at Hydrosystems Laboratory, UIUC and CEE Machining shop staff for their help with assembling the initial setup of the oblique bubble diffusers.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article and its Supplementary Material.

Additional information

Funding

This work was supported by the Great Lakes Restoration Initiative and the U.S. Geological Survey under Cooperative Ecosystem Studies Units (CESU) Grant G21AC10372. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.