IntroductionInvasive carps, including silver carp (Hypophthalmichthys molitrix) and bighead carp (H. nobilis), have migrated throughout the Mississippi River Basin since the 1970s and are now expanding into other large rivers basins in the United States. Recent expansion toward the Great Lakes Basin has raised conservation and economic concerns related to how invasive carps could adversely affect the $7-billion Great Lakes fishing industry (Conover et al. 2007). Lock and dam structures positioned throughout these interconnected waterways have been identified as key management locations to limit or block the upstream migration of invasive carps toward the Great Lakes. Permanent lock closure was one possible solution to block fish passage, but this approach was not considered a viable option to many stakeholders due to the negative economic consequences of limiting waterway access (Schwieterman 2010, 2015). Therefore, researchers and resource managers have focused on the development of other fish deterrents that could be implemented at lock structures to reduce upstream migration of invasive carps while maintaining a fully operational waterway.Several technologies have been explored to prevent fish passage through locks (Noatch and Suski 2012; Zolper et al. 2019). For example, invasive carp deterrent systems such as underwater acoustics and electric barriers have shown promise under various scenarios (Cooper et al. 2021; Jones et al. 2021; Noatch and Suski 2012). Underwater speakers can be used to broadcast acoustics in certain areas to elicit strong behavioural responses from invasive carps (Bzonek et al. 2021; Vetter et al. 2015, 2017). Similarly, electric barriers create an electrified field that deters and immobilizes fish as they attempt to move upstream (Egly et al. 2021). Each deterrent method has benefits and challenges related to target species responsiveness, acclimation to the deterrent stimulus, variable efficacy with fish size, human safety concerns, and operational costs. Consequently, supplementing existing control methods and further developing additional methods would be beneficial in deterring upstream movements of invasive carps.Chemosensory stimuli, such as carbon dioxide (CO2) infused into water, offer an alternative approach to deter carp movements. The concept is to infuse CO2 into water at specific areas (e.g., lock structures) to repel or immobilize invasive carps. Research has already demonstrated that invasive carps respond strongly to localized CO2 plumes (Cupp et al. 2017a, 2021; Donaldson et al. 2016; Tix et al. 2018), and infusion into water could be accomplished at lock structures with minimal expected effects to navigation or lock operation (Zolper et al. 2019). Aspects of CO2 deterrents may also address limitations with other deterrent technologies due to its nonselectivity across species and taxa, efficacy across most life stages, limited evidence of acclimation to the chemical stimulus, low cost, wide availability, and low risk to human health (Suski 2020). Pesticide registration with the US Environmental Protection Agency in 2019 has further facilitated continued development and created a legal pathway to apply CO2 as an approved pesticide for invasive carp control in the United States (USEPA 2019). However, much of the previous research has been limited to relatively small scales and controlled environments where CO2 applications were accomplished using crude gas injection methods. The next important step is to determine the feasibility of CO2 applications at a management scale in navigation lock structures using more purposefully engineered gas injection systems.The transition of CO2 testing from controlled settings (e.g., laboratory tanks, outdoor ponds) into full-scale environments (e.g., navigation lock structures) requires additional engineering considerations that were not addressed during previous biological research. Several gas-to-liquid infusion methods, including mechanical and fluid dynamic mixers, have been developed for chemical, mechanical, and environmental engineering applications that could potentially be adapted for this new application (Bumrungthaichaichan et al. 2016; Chen et al. 2017; Fossett 1951; Fossett and Prosser 1949; Fox and Gex 1956; Lane and Rice 1982). Jet mixing stands out as a highly effective nonphysical method of mixing. It generally uses a pump to draw liquid from a chamber, pressurizes it, and reinjects it through a nozzle as a high velocity jet. The momentum of a fluid jet entrains surrounding fluid and induces efficient mixing within the chamber. Jet driven liquid-gas mixing also produces multiphase flow that can alter fluid properties, resulting in cavitation, efficiency losses, and other phenomena (Freudigmann et al. 2017; Liu et al. 2017; Loeb 2017; Tian and Van de Ven 2017).The performance of machinery used in mixing applications is often gauged by the uniformity, concentration, and efficiency of the processes. Mixing performance can be quantified in two ways: (1) “mixing homogeneity” describes the uniformity of CO2 concentrations within the lock using the variance in measured CO2 at all monitoring points; and (2) “injection efficiency” is the ratio of dissolved CO2 to injected CO2, an indicator of gas-to-liquid infusion ability. Both mixing homogeneity and injection efficiency are affected by many factors, including the chamber volume, injection velocity, fluids being mixed, jet injection angles, nozzle geometry, manifold configurations, operating pressure, and Reynolds number (Bumrungthaichaichan et al. 2016; Dinsmore et al. 2017; Fox and Gex 1956; Grenville and Tilton 2011; Lane and Rice 1982; Maruyama 1986; Patwardhan and Gaikwad 2003).Nozzle geometry also directly affects mixing time, operating pressure, and flow rate, and a variety of nozzle designs have been evaluated with circular apertures (Cui et al. 2015; Yang et al. 2013), noncircular apertures (Majamaki et al. 2003; Nikitopoulos et al. 2003; Quinn 2005; Smith et al. 1997; Yu et al. 2004; Zaman et al. 1994; Zhdanov and Hassel 2012), and vortex-inducing nozzle covers (Bradbury and Khadem 1975; Foss and Zaman 1999; Reeder and Samimy 1996; Samimy et al. 1993; Zaman et al. 1994; Zhdanov and Hassel 2012). Fox and Gex (1956) demonstrated that good mixing performance can be attained by maximizing a nozzle’s “momentum flux,” the product of velocity and mass flow rate through an aperture (Bumrungthaichaichan et al. 2016; Fox and Gex 1956). Manifolds are used to divide a single fluid stream into multiple outflow streams, and detailed studies of diverse manifold designs, including customized designs, are available in the scientific literature (Bajura and Jones 1976; Bajura 1971; Gandhi et al. 2012; Hassan et al. 2014; Majumdar 1980; Pathapati et al. 2016; Subaschandar and Sakthivel 2016; Tong et al. 2009).The goal of this research is to evaluate the performance of a CO2 injection system at a navigation lock structure on the Fox River near Kaukauna, Wisconsin, USA. System performance was evaluated based on several metrics that characterized the efficiency and timing of this engineered system to reach target concentrations. A target concentration range of 100–150  mg/LCO2 was established based on previous fish behavior testing and discharge allowances on state and federal permits (USEPA 2019). Two manifold distribution types were tested that were designed as possible options to treat the lock without obstructing lock operation. Previous testing in outdoor ponds detailed the mixing performance of singular floor-based and wall-based CO2-to-water injection manifolds in a closed simulated lock chamber (Zolper et al. 2019). The present research uses a system of manifolds, activated in different combinations, to generate a uniform vertical and lateral CO2 field with the intent to prevent fish from entering the lock chamber or force resident fish downstream. This research introduces several new factors to previous research, including full-scale commercial injection equipment, multiple combinations of manifolds, and operation in a management applicable setting in a lock structure.Experimental Setup and MethodsThe CO2 infusion experiments were designed to simulate, as closely as possible but without vessels, the treatment of a lock chamber with CO2 during an upstream lock operation. In a typical upstream passage of a vessel through a navigational lock, the lock must be at tailwater level (downstream water level) with the downstream miter gates open to allow the vessel to enter. Once the vessel is moored in the lock, the downstream gates are closed and the chamber is filled to the head-water level. Upon attaining the head-water level, the upstream miter gates are opened and the vessel can continue upstream. To prevent upstream migration of aquatic invasive species during this process, the lock water can be infused with CO2 before the vessel enters the lock to (1) drive fish out of the chamber and into the downstream pool; and (2) to deter any fish from entering the lock with the vessel. In these experiments, we simulate the pretreatment of the lock chamber with CO2 without the complication of vessel passage.The objective of these experiments was to evaluate the ability of the CO2-to-water infusion system to attain a uniform target concentration of 100  mg/L throughout a lock chamber. Multiple combinations of operating parameters using wall-based and floor-based distribution manifolds were tested to generate diverse flow fields. This resulted in a greater number of operating parameters in comparison to the earlier research (Zolper et al. 2019) and substantially increased overall complexity. Certain parameters were constrained by the greater scope of activities and regulations of the participating agencies, including the US Army Corps of Engineers (USACE), US Geological Survey (USGS), and Wisconsin Department of Natural Resources (WDNR). The maximum allowable CO2 concentration of 150  mg/L in a lock chamber was established by the WDNR under an approved National Pollutant Discharge Elimination System (NPDES) permit based on state requirements in compliance with Federal Clean Water Act standards.A temporary CO2-infusion system with water distribution manifolds was constructed and tested at Kaukauna Lock #2 on the Fox River located in Kaukauna, Wisconsin, as shown in Fig. 1. The test site was entirely on property that was owned and operated by the Fox River Navigational System Authority (FRNSA). Site access approval was gained from FRNSA, and all necessary local, state, federal and historical permits were obtained prior to construction and testing. Construction was completed by a private firm that specialized in water distribution systems. Experiments were performed from August 5, 2019, through September 6, 2019.The lock chamber was approximately 51.82 m (170 ft) long and 11.12 m (36.5 ft) wide at the top, tapering down to 10.72 m (35.2 ft) wide at the bottom (Fig. 2). The upstream (west) miter gates were closed during the experiments, and each gate was 5.7 m (18.7 ft) long and met in the middle. The downstream (east) miter gates were open during the experiments: the line between their hinges divided the lock chamber from the downstream reservoir. The Kaukauna Lock #2 measurements (Fig. 2) were accurate to ± 50.8 millimeters (mm) (±2  in.). Thus, the volume enclosed by the chamber and upstream miter gate was about 2,920,000±47,000  L (771,000±12,400  gal.) at head-water level (5.08 m or 16.7 ft) and 1,548,000±37,000  L (409,000±9,700  gal.) at tail-water level (2.64 m or 8.7 ft). For reference, a typical 183-m (600-ft)-long commercial navigation lock has a tailwater volume of approximately 20 times that of Kaukauna Lock #2.An open-loop CO2 injection system was designed for these experiments (Fig. 2). A diesel-powered pump drew near-bottom water from the northwest corner of the chamber at the upstream end of the lock, lifted it approximately 6.1 m (20 ft), passed it through a filter and into parallel mixing chambers where CO2 was added to carrier water under a prescribed pressure differential, and then distributed the CO2-enriched carrier water in the lock chamber using a series of manifolds that were adjusted for each experiment. The pump intake was positioned approximately 203.2 mm (8 in.) above the floor of the lock, the same height as the discharge of the wall and floor manifolds, resulting in a net pumping height of 0 mm. The piping networks, pressurized solution feed (PSF), manifolds, and nozzles were the primary causes of the system head losses. All treatments occurred with the chamber at tail-water level and downstream (east) miter gates open. After each experiment, the downstream miter gates were closed and the lock was filled to dilute excessive CO2 concentrations before discharging the treated water downstream in preparation for the next experiment.The pump produced water flow rates (V˙H2O) from 6,810 to 12,110  L/min (1,800 to 3,200  gal./min) in accordance with the requirements for the experiment. A TOMCO2 Systems (Loganville, Georgia) PSF was used to infuse CO2 into the carrier water from the pump at mass flow rates of m˙CO2=1,134 to 1,588  kg/h (2,500 to 3,500  lb/h). The CO2-infused-water was mixed into the lock chamber using either a series of wall-based or floor-based manifolds, shown in Fig. 2.High-density polyethylene (HDPE) was used for all piping except for the floor manifold, which used polyvinyl chloride (PVC). The pump intake and outflow pipes were made of 304.8-mm (12-in.) HDPE. Downstream from the PSF, 406.2-mm (16-in.) HDPE pipe was used to divide the flow into two streams (Fig. 2). The downstream (east) pipes provided flow to four wall manifolds. The upstream (west) pipes provided flow to four wall manifolds and included a valve to provide flow to the floor manifold. The floor manifold was made of 203.2-mm (8-in) schedule 40 PVC and was affixed to the bottom using distributed weights. The wall manifold branches used 406.4-mm to 203.2-mm (16-in. to 8-in.) T-fittings and elbows to connect to the eight wall manifold dropdowns (Figs. 2 and 3). Several gate valves were used to control the water flow to the full floor manifold or a maximum of four wall manifolds at a time. The pump intake and discharge pipes were placed on one side of the chamber to minimize obstructions to lock personnel.The wall manifolds were designed to use the kinetic energy and turbulence of a series of jets to induce mixing and diffusion of CO2 into the water. The design is also adaptable to flush-mounted scenarios that would not physically interfere with lock operation or vessel passage. Each wall manifold was constructed from 203.2-mm (8-in.) HDPE and designed to produce six equal-discharge jets from TOMCO2 Systems (Loganville, Georgia) using cast 304 stainless steel nozzles spaced 762 mm apart, as shown in Figs. 3 and 4.Converging nozzles were selected to maximize momentum flux, which has been shown to enhance long-range mixing (Fox and Gex 1956). The nozzles had 76.2-mm (3-in.) inlets and 23.4-mm (59/64-in.) outlets to allow the necessary back-pressure for proper CO2 dissolution and water flow rate. All nozzles discharged horizontally toward the south wall and were located approximately 203.2 mm (8 in.) above the bottom of the lock chamber. In total, eight wall manifolds (designated A–G) were evenly spaced at 5.33 m (17.5 ft) on center along the north wall of the lock chamber (Fig. 5). Combinations of two, three, and four manifolds were variously activated for individual experiments during the test period.The floor manifold was composed of a 406.4-mm (16-in.) HDPE main branch that equally distributed the CO2-infused water into four 203.2-mm (8-in.) PVC side branches with center-to-center spacing of 2.67 m (105 in.), as shown in Fig. 5. Each 48.77-m (160-ft)-long side branch ran parallel to the length of the lock, and branches were connected to one another at the downstream end of the lock by a 203.2-mm (8-in.) PVC pipe. Seventy-two 6.35-mm (1/4-in.) holes (dH2O) were drilled into the tops of the four side branches at a spacing of 508 mm (20 in.) each (288 total). The main branch also had twelve 6.35-mm (1/4-in.) holes spaced at 609.6 mm (24 in.) over its 8-m (26.25-ft) length. Thus, a total of 300 orifices in the floor manifold distributed CO2-enriched water throughout the lock chamber approximately 203.2 mm above the floor of the lock.Flow meters, pressure gauges, and control valves allowed the system to be configured and operated for a variety of experimental conditions. The pump was fitted with a vacuum pressure gauge at the intake port and an operating pressure gauge at the outlet port (Fig. 2), both with accuracies of ±6.9  kPa (±1  psi). The pressure gauge readings at the pump intake port Pin and pump outlet port Pout were used to determine the operating pressure of the system Poper [Eq. (1)] (1) A Fluid Components International (San Marcos, California) CO2 mass flowmeter (accuracy ±1% of reading [lb/h]) was used to measure CO2 flow rate into the PSF every minute during experiments (Fig. 2). A Hach sc200 (Loveland, Colorado) pH and temperature sensor measured the acidity and temperature, respectively, of the incoming and outgoing water of the PSF to ±0.1% of full scale at 25°C. Two Bourdon pressure gauges measured the PSF of incoming and outgoing water pressure with an accuracy of ±6.9  kPa (±1  psi). The volume flow rate of water was measured at three points in the system (Fig. 2) using Greyline Instruments (Largo, Florida) Doppler Flow Meters (DFM 6.1) with an accuracy of ±75.7  L/min (±20 gal./min). The primary flowmeter, located on the 406.4-mm (16-in.) HDPE pipe between the pump and the PSF, measured the total flow through the piping system. Two additional flowmeters were also placed on the 406.4-mm (16-in.) lines feeding each bank of manifolds (Fig. 2). A portable Greyline Doppler Flow Meter (DFM 6.1) calibrated to 203.2-mm (8-in.) pipes was used to assess the flow rate of the dropdowns to individual manifolds. Flow rates were logged every 10 s on all meters.The distribution of dissolved CO2 in the lock chamber was measured during each experiment using an array of pH sensors on multiparameter sondes with real-time telemetry. Ten YSI (Yellow Springs, Ohio) multiparameter sondes (eight YSI Model 600XLM and two YSI/Xylem EXO2) were divided between five Ocean Science (North Falmouth, Massachusetts) tethered boats positioned on taglines throughout the lock chamber (Fig. 6). Each tethered boat was equipped with two sondes positioned at depths of 0.914 m (3 ft) and 1.828 m (6 ft) below the free surface, a Campbell Scientific (Logan, Utah) CR6 data logger, a RF407 high-speed 900-MHz spread-spectrum radio, a 12-volt battery, and a 10-W solar panel and regulator. The radios allowed wireless transmission of real-time data at 5-s intervals from all instruments to a single laptop computer with custom display for real-time feedback during the experiments. A pulley driven hand line attached to each boat allowed the boat to be positioned at any point along the static tagline by a single person. During testing, the boats were simultaneously repositioned from south (S), center (C), and north (N) stations at carefully timed intervals to assess the variation of the CO2 field and other basic water-quality properties (temperature, specific conductance, and dissolved oxygen). Tethered boat #3, located near the center of the lock chamber (Fig. 6), was equipped with the EXO2 sondes, which included additional sensors not found on the 600XLM sondes (optical dissolved oxygen, turbidity, total algae, and phycocyanin sensors).Dissolved CO2 concentration (mg/L) was computed from pH measurements using a CO2-pH regression equation [Eq. (2)] and water samples collected from the Fox River Lock #2 in Kaukauna, Wisconsin, at the beginning and end of the experiments (July 26, 2019, and September 5, 2019, respectively). Details of the development of Eq. (2) are given in the appendix (2) CCO2=3.361·108·e−2.323·pHThe multiparameter sondes were three-point calibrated using pH 4, 7, and 10 standards at the beginning of every week, and all sensors (except temperature) were calibrated. The thermistors on the sondes do not require calibration but are checked annually in a thermal bath. During calibration, sonde clocks were synchronized with the master clock for the experiments.Design of Experiments—Test PlanPrevious research showed that water flow rate V˙H2O and CO2 mass flow rate m˙CO2 have the greatest influence on the performance of the floor and wall manifolds (Zolper et al. 2019). A design of experiments (DOE) full-factorial approach was used to extricate their respective influences on the performance of the manifold systems (Box et al. 2005). Sixteen experiments were identified using four manifold configurations at four flow rates each, as shown in Table 1. The pump provided stable water flow rates at pump speeds of ω=1,200 and 1,800 RPM, which corresponded to different water flow rates whether using floor manifolds or combinations of two, three, or four wall manifolds. The PSF was set to steady CO2 mass flow rates of m˙CO2=18.9 and 26.5  kg/min (2,500 to 3,500  lb/h) with negligible effects on water flow rates. It was found that operating four wall manifolds at low pump speed produced insufficient operating pressure to allow the PSF to efficiently infuse CO2; therefore, those experiments were not pursued.Table 1. Design of experiments factorial design settings showing water volume flow rates (V˙H2O) and CO2 mass flow rates (m˙CO2) of the floor manifold and three configurations of wall manifoldsTable 1. Design of experiments factorial design settings showing water volume flow rates (V˙H2O) and CO2 mass flow rates (m˙CO2) of the floor manifold and three configurations of wall manifoldsManifoldOperating parametersUnits1234FloorVolume flow rate V˙H2O(L/min)8,3308,33011,36011,360(gal./min)2,2002,2003,0003,000Mass flow rate m˙CO2(kg/min)18.926.518.926.5(lb/h)2,5003,5002,5003,500Wall 2 manifoldsVolume flow rate V˙H2O(L/min)6,8106,81010,60010,600(gal./min)1,8001,8002,8002,800Mass flow rate m˙CO2(kg/min)18.926.518.926.5(lb/h)2,5003,5002,5003,500Wall 3 manifoldsVolume flow rate V˙H2O(L/min)8,3308,33011,36011,360(gal./min)2,2002,2003,0003,000Mass flow rate m˙CO2(kg/min)18.926.518.926.5(lb/h)2,5003,5002,5003,500Wall 4 manifoldsVolume flow rate V˙H2O(L/min)9,080a9,080a12,11012,110(gal./min)2,400a2,400a3,2003,200Mass flow rate m˙CO2(kg/min)18.9a26.5a18.926.5(lb/h)2,500a3,500a2,5003,500Test configurations are identified by the following nomenclature: manifold type [floor (F) or wall (W)], wall–manifold combinations (e.g., A-B-C-D; for W manifold type only), water volumetric flow rate (V˙H2O), and CO2 mass flow rate (m˙CO2) in US customary units. Therefore, a test configuration using wall manifolds A, C, E, and G at V˙H2O=3,200  gal./min, and m˙CO2=2,500  lb/h is designated “W A-C-E-G 3200-2500”. Meanwhile, a test configuration of the floor manifold at V˙H2O=3,000  gal./min, and m˙CO2=2,500  lb/h is designated “F 3000-2500”.General Experimental ProceduresThe full scope of the research also includes fish behavioral studies, which largely dictated the duration for each trial. Therefore, nearly 2 h were allocated for each experiment to perform baseline pH measurements and track tagged fish using acoustic fish telemetry. Baseline measurements of CO2 concentration were collected prior to injection (t<0  min). At about t=−10  min, the pump was set to the nominal water flow rate to allow enough time to develop a stable flow field. At the beginning of CO2 injection (t=t0=0  min), the PSF was set to the nominal CO2 mass flow rate. The CO2 injection was stopped when the measured CO2 reached a threshold concentration of 150  mg/L at any sensor on the tethered boats. This was a requirement to comply with issued permits from WDNR for conducting this research project. In practical administration, injection would continue until all locations within the lock reached the desired concentrations and not just at any one sensor. Measurements continued for up to 90 min after the CO2 injection began (t>0  min). The five tethered boats with pH sensors simultaneously measured for 3 min at each station in the lock (north, center, south) throughout the trial period. The following sequence of activities were undertaken to simulate upstream lock changes: 1.At t=−40  min: Baseline testing: Tethered boats start at center stations of lock, then move to north stations (t=−37  min), back to center stations (t=−34  min), to south stations (t=−31  min), and return to center stations (t=−28  min).2.At t=−10  min: Pump starts and attains desired flow rate.3.At t=−1  min: Lower miter gates open.4.At t=t0=0  min: CO2 injection begins.5.Over the course of the test, the tethered boats are moved to the following positions: •(0150  mg/L), thus foreshortening CO2 injection times and dissolved CO2 masses. Three-wall manifolds (Fig. 10) produced good flushing fields at the start of tests and nearly uniform CO2 concentrations at the end of tests. Concentrated wall manifolds (Figs. 9 and 10) produced stronger flushing fields than four distributed manifolds (Fig. 11). However, distributed manifolds (Fig. 11) were less likely to produce localized CO2 concentrations that may exceed WDNR regulatory limits (>150  mg/L), thus allowing longer injection times and greater dissolved CO2 masses.Treatment time and fish exposure periods are important considerations relative to fish biology and effective fish control treatments. Commercial shipping endeavors to minimize locking time but longer treatment duration ensure that localized CO2 concentrations can be distributed in the lock to more uniform levels. Additionally, fish must encounter the CO2 plume for a certain length of time to make favorable cognitive and behavioral responses. It is likely that prolonged injection times across larger treatment volumes would result in a more effective outcome. Regardless, the testing showed that floor manifolds and several (3–4) wall manifolds can effectively reach target concentrations and could be used to conduct lock treatments.Performance at 15 MinutesThe previously described CO2 concentration versus time plots (Figs. 8–11) show the variation of dissolved CO2 for the duration of the trials. Details of the comparative performance of the manifolds at a specific time (e.g., 15 min) can be obtained by interpolation of the data from the 10 sondes at each position [south (S), center (C), and north (N)]. Fig. 12 shows the four-wall manifolds A-B-C-D operating at c=12,110  L/min (3,200  gal./min) and m˙CO2=18.9  kg/min (2,500  lb/h). The data for the center of the lock are taken as the mean of five measurements at time 14:30 to 14:50 min. The data for the north wall of the lock are taken as the mean of five measurements at time 15:10 to 15:30 min. The data for the south wall of the lock use interpolation between five measurements at times 11:30 to 11:50 and five measurements at 21:10 to 21:30 min.Data from each of the tests were used to generate CO2 concentration arrays corresponding to each of the sonde locations shown in Fig. 5 for the upper (top) and lower (bottom) sondes on each boat. The data were imported into MATLAB version 9.10 R2021a, and a grid was generated with 40 subdivisions in the x-direction (west–east), y-direction (south–north), and z-direction (bottom–top). Linear interpolation was used to assign values to grid points from the measured data points (sondes).Fig. 13 shows the interpolated dissolved CO2 field from Fig. 12 as contour plots divided into arrays of lower sondes (bottom) to focus on the performance concerns discussed earlier. The field is contained within the rectangular chamber created by the pH sensors as depicted in Fig. 5. The positions of the connecting pipes to the active wall manifolds (A-B-C-D) are depicted by black circles in the upper region of relevant figures. The figures graphically and numerically illustrate the expected CO2 concentration gradient from the upstream (west) to downstream (southeast) portions of the chamber. Based on fish response to CO2 in previous studies, such a gradient is expected to drive fish from the lock through the downstream (east) gates and out of the lock.Fig. 14 shows the dissolved CO2 concentrations at time = 15 min for the floor manifolds depicted in Fig. 8. Each figure shows the bottom sensor arrays corresponding to trials F 3000–2500 [Fig. 14(a)] and F 3000–3500 [Fig. 14(b)]. In general, all trials show a decreasing concentration gradient from north to south. This is attributable to the south-to-north flow field generated by the pump intake in the northwest corner. The lower CO2 concentrations at the upstream end of the lock in the southeast corner (lower left) are believed to be caused by leaks in the upstream miter gates combined with the water circulation pattern induced by south branch pipe position and the pump intake. Aside from minimizing leakage, this lateral asymmetry may be able to be somewhat alleviated by placing the manifold branches closer to the lock walls and placing the pump intake centrally within the chamber.Fig. 15 shows the dissolved CO2 concentrations at the lower sondes at time = 15 min for the wall manifolds A-B depicted in Fig. 9 under the test names W A-B 2800-2500 [Fig. 15(a)] and W A-B 2800-3500 [Fig. 15(b)]. In general, all trials showed a decreasing concentration gradient from upstream to downstream (west to east). This was caused by the activation of the wall manifolds only on the upstream (west) section of the lock as shown by the black circles. Although the position of the manifolds generated the desired deterrence field gradient, it also caused CO2 spikes (Fig. 9) that foreshortened the injection time and thereby reduced the available CO2 for dissolution. Similar conditions occurred for the W A-B-C manifolds (Fig. 10) where an early spike in CO2 caused shorten injection times and less overall CO2 injection. Thus, the CO2 concentrations attained by two-wall and three-wall manifolds were not able to meet the CO2 concentration requirements due to the short injection times to meet WDNR limits on dissolved CO2 (<150  mg/L).Fig. 16 shows the dissolved CO2 concentrations at the lower sondes at time = 15 min for the distributed wall manifolds A-C-E-G depicted in Fig. 11. Each figure shows the bottom sensor arrays corresponding to trials (1) W A-C-E-G 3200-2500; and (2) W A-C-E-G 3200-3500. The uniform distribution of manifolds, shown by the black circles, gives a uniform CO2 distribution that nearly attained target concentrations in both cases. Additionally, the absence of high CO2 regions allowed CO2 injection to continue for 10–12 min and approach the desired concentration. These trials also exhibit a more subtle form of the upstream–downstream (west–east) concentration gradient that was sought in the W A-B, W A-B-C, and W A-B-C-D trials.In general, the floor manifolds and the combinations of four-wall manifolds (W A-B-C-D and W A-C-E-G) most nearly approached the target CO2 concentrations and generated rather uniform fields. This is due to the greater distribution of water injection points and the need to carry less CO2 per unit volume of water. Meanwhile, the high momentum flux of combinations of two-wall and three-wall manifolds (W A-B and W A-B-C) produced good mixing and strong upstream-to-downstream (west-to-east) CO2 gradients, but the resulting high CO2 concentration quickly approached the WDNR limits (150  mg/L) and foreshortened the injection times, thereby reducing the total volume of water treated. These results indicate that multiple concentrated manifolds can meet performance requirements for unidirectional lockages, whereas distributed manifolds are better suited for bidirectional lockages.Manifold Performance: Concentration and Injection EfficiencyThe injection efficiency is a composite of the PSF’s ability to infuse CO2 into water (gas–liquid mixing) and the manifolds’ ability to distribute the treated water uniformly through the lock chamber. Previous research in a concrete test pond at USGS Upper Midwest Environmental Sciences Center allowed the authors to study floor and wall manifolds under highly controlled conditions of injection times (Δt=40  min), flow rates, and treatment volume (Zolper et al. 2019). This ensured that accurate calculations of injection efficiency were feasible. The present studies, representative of actual field installations subject to regulatory conditions, pose greater challenges to consistent calculations of injection efficiency due to greater variation in injection times (3:01<Δt<13:37). These challenges are further compounded by the effect of water flow rate on mixing homogeneity and off-gassing rates.In an effort to reconcile the challenges presented previously with the need to quantify the performance of each manifold configuration, the authors took an approach similar to previous work (Zolper et al. 2019) with slight modifications. The mass of CO2 dissolved in the lock water at 15 min after CO2 injection began was approximated by the product of the average CO2 concentration CCO2 in the water and the water volume of the lock (VH2O=1,459,700±18,000  L), as shown in Eq. (3). The average CO2 concentration CCO2 at 15 min was obtained from the sum of individual concentrations CCO2  i, divided by the total number of sensors (30) shown in Eq. (3). Thus, average concentration used the same 30 data points that were interpolated to generate the CO2 concentration fields in Figs. 13–16(3) mCO2D=CCO2·VH20=∑i=130CCO2  i30·VH20Table 2 lists the average CO2 concentration at 15 min and the calculated dissolved CO2 mass for each of the tests listed in Table 1, using the naming convention introduced after Table 1. It also lists the standard deviation of the CO2 concentration CCO2 at 15 min, an indicator of mixing homogeneity, as well as the resulting uncertainty in dissolved mass. Only four experimental trials reached the target mean concentration of 100  mg/L due to the requirement that tests be terminated when any sensor reached a CO2 concentration of 150  mg/L. Nonetheless, the relative magnitude of the standard deviation with respect to concentration is an indicator of mixing homogeneity and the magnitude of any concentration gradients that formed in the chamber. For example, the minimum number of wall manifolds (W A-B) had relatively high standard deviations, whereas the maximum distributed wall manifolds (W A-C-E-G) had relatively small standard deviations due to the more uniform concentrations they produced.Table 2. Average CO2 concentration at 15 min, standard deviation, and total dissolved CO2 mass with corresponding uncertaintiesTable 2. Average CO2 concentration at 15 min, standard deviation, and total dissolved CO2 mass with corresponding uncertaintiesTest name (gal./min-lb/h)Avg. CO2 concentration (mg/L)Standard deviation (mg/L)Dissolved CO2 mass (lb)Dissolved CO2 mass (kg)F 2200-2500118.254.1382.1±178.9173.3±81.2F 2200-350058.534.3189.0±113.385.8±51.4F 3000-250071.231.6230.1±104.6104.4±47.5F 3000-3500120.243.0388.4±142.4176.2±64.6W A-B 1800-250063.435.5204.9±117.592.9±53.3W A-B 1800-350032.115.2103.7±50.347.1±22.8W A-B 2800-250035.112.9113.4±42.851.5±19.4W A-B 2800-350035.315.0114.1±49.551.7±22.4W A-B-C 2200-250043.619.8140.9±65.463.9±29.6W A-B-C 2200-350052.323.6169.0±78.176.7±35.4W A-B-C 3000-250069.018.6223.0±61.6101.1±28.0W A-B-C 3000-350076.528.4247.2±94.0112.1±42.6W A-B-C-D 3200-250076.723.3247.9±77.2112.4±35.0W A-B-C-D 3200-3500W A-C-E-G 3200-2500106.231.1343.2±103.0155.7±46.7W A-C-E-G 3200-3500118.326.9382.3±89.1173.4±40.4The mass of CO2 injected into the lock through the PSF was calculated as the product of the CO2 mass flow rate and the injection time, as shown in Eq. (4). The CO2 mass flow rate was measured at 60-s intervals and remained stable for the duration of the injection period, leading to relatively low uncertainty values. Injection time was limited by the WDNR criteria that the CO2 concentration must not exceed 150  mg/L(4) The carbon dioxide injection efficiency ε was calculated from the quotient of the masses of CO2 dissolved in the water (mCO2D) and CO2 injected through the PSF (mCO2I), as shown in Eq. (5). The carbon dioxide injection efficiency serves as an indicator of the expected operational costs of each injection manifold. This method gives the composite performance of the PSF, injection manifolds, and off-gassing rates in addition to the limits imposed on CO2 injection time by WDNR CO2 limits. This is representative of operational and regulatory constraints of an actual installation (5) ε=mCO2DmCO2I=CCO2·VH20m˙CO2·tinjTable 3 lists the injection times, total injected masses of CO2, and calculated injection efficiencies corresponding to each of the trials in Table 2. The CO2 flowmeter was an accurate instrument, so the uncertainty was most affected by injection time. The efficiency represents the effects of many factors, including the level of CO2 saturation of water, turbulence of the water, injection time, operating pressure, Reynolds number, and manifold type (W or F) and combination (A-B, A-B-C, etc.). Most of the uncertainty in injection efficiency was propagated through dissolved CO2 mass from the standard deviation listed in Table 2.Table 3. Injection times, total masses of CO2 injected, and injection efficiencies with corresponding uncertaintiesTable 3. Injection times, total masses of CO2 injected, and injection efficiencies with corresponding uncertaintiesTest name (gal./min-lb/h)Injection time (m:s)Injection time (s)Injected CO2 massInjected CO2 massInjection efficiency(±1  s)(±1  s)(lb)(kg)(%)F 2200-250013:37817588±23267±1065±31F 2200-350006:13373372±10169±551±30F 3000-250009:36576409±16185±756±26F 3000-350011:41701698±19317±956±20W A-B 1800-250006:55415296±12134±569±40W A-B 1800-350003:01181183±583±257±28W A-B 2800-250004:33273193±888±359±22W A-B 2800-350003:01181182±583±263±27W A-B-C 2200-250005:26326235±9107±460±28W A-B-C 2200-350004:22266257±7117±366±30W A-B-C 3000-250006:31391278±11126±580±22W A-B-C 3000-350006:25385380±11172±565±25W A-B-C-D 3200-250008:28508361±14164±669±22W A-B-C-D 3200-3500W A-C-E-G 3200-250012:02722508±20231±968±20W A-C-E-G 3200-350010:33633617±18280±862±15Design of Experiments: Injection EfficiencyInjection efficiency heavily depends on both CO2 mass flow rate and water volume flow rate. If the CO2 mass flow rate is too high for the water flow rate, there is substantial off-gassing when it is ejected from the manifolds. If CO2 mass flow rate is too low for the water flow rate, the carrier water does not maximize its carrying potential. The design of experiments outlined in Table 1 was used to assess the optimum operating parameters of CO2 mass flow rate and water volume flow rate. Efficiency optimization was addressed using four DOE trials for each of three manifold combinations: floor (F), wall manifolds A-B (W A-B), and wall manifolds A-B-C (W A-B-C). The resulting efficiencies at 15 min (Table 3) were subject to regression analysis using Minitab 18. Fig. 17 shows the optimum water and CO2 flow rates to maximize efficiency of three concentrated manifolds (W A-B-C) under the present constraints. The axes are labeled according to the convention established in Table 1. Fig. 17 shows that the CO2 flow rate was too high to maximize efficiency. In fact, there was visible off-gassing from the lock for some of the high CO2 flow rate trials. Fig. 17 also shows that configuration could accommodate even greater water flow rates.For a given manifold configuration, high water flow rates result in higher turbulence and operating pressure. High turbulence generally helps entrain and dissolve gaseous CO2, resulting in high injection efficiency (Gebhart 1993; Rolle 2016). Furthermore, the mass diffusivity for CO2 into water near the experiment temperature (δ=0.2×10−8  m2/s) and the target concentration (100  mg/L) renders high injection efficiency feasible (Rolle 2016). Although high operating pressure raises the amount of CO2 that can be dissolved into the water, when the pressure decreases after exiting the manifolds, the dissolved gases can come out of solution. These results can be used to further improve the design configurations and operating conditions of CO2 injection manifolds (e.g., floor versus wall) and related technologies (Shane 1996).Operating Power and CostsOperating power is an influential factor when selecting a large-scale CO2-to-water injection system. Water flow rate V˙H2O, operating pressure Poper, and pump efficiency all contribute to the operating costs of equipment. The operating power W˙oper of the pump-manifold system was calculated as the product of the operating pressure [Eq. (1)] and the water flow rate V˙H2O, as shown in Eq. (6) (6) Table 4 provides the water volume flow rates, operating pressures, calculated operating power, and associated uncertainties for each of the manifold systems in this study. For each manifold system, the operating pressure increased with water volume flow rate. The compound effect can double the operating power, such as in the case of the trials of manifolds W A-B 1800-3500 in comparison to W A-B 2800-3500. Among the following systems, the two-wall manifold (W A-B) had the highest operating pressure for a given flow rate. Meanwhile, the floor (F) and three-wall manifolds (W A-B-C) had similar operating pressures and power at low flow rates, but the floor manifold operating power increased more substantially at higher flow rates. To treat larger volumes and keep operating power low, better results were obtained using a greater number of manifolds operating at low pressure than using a few manifolds operating at high pressure.Table 4. Water flow rate, operating pressure, and operating power of manifoldsTable 4. Water flow rate, operating pressure, and operating power of manifoldsTest name (gal./min-lb/h)Operating power(±0.00126  m3/s)(±20  GPM)(±6.9  kPa)(±1  psi)(kW)F 2200-25000.1402,218337.849.047.3±1.1F 2200-35000.1402,224379.255.053.2±1.1F 3000-25000.1842,924668.897.0123.4±1.5F 3000-35000.1892,992675.798.0127.6±1.6W A-B 1800-25000.1151,828379.255.043.7±0.9W A-B 1800-35000.1161,845386.156.045.0±0.9W A-B 2800-25000.1722,723772.2112.0132.7±1.5W A-B 2800-35000.1752,770779.1113.0136.1±1.6W A-B-C 2200-25000.1502,375365.453.054.7±1.1W A-B-C 2200-35000.1472,322358.552.052.5±1.1W A-B-C 3000-25000.1903,014503.373.095.7±1.5W A-B-C 3000-35000.1913,021537.878.0102.5±1.5W A-B-C-D 3200-25000.2003,169441.364.088.5±1.5W A-B-C-D 3200-3500W A-C-E-G 3200-25000.1963,106406.859.079.7±1.4W A-C-E-G 3200-35000.1983,135427.562.084.6±1.5Operating costs for the present manifold systems are a function of operating power W˙oper and the processes needed to obtain and infuse the CO2 to the required specifications. Under steady-state conditions, the energy needed to operate the pump Eoper was the product of the operating power and pump operating time, as shown in Eq. (7). The pump operating time was about 40 min (0.67 h) for each test. An estimated pump efficiency εpump=80% was selected as representative of a typical pump. The energy required to operate the TOMCO2 PSF was negligible because the ambient temperatures in August–September 2019 were ample to vaporize the liquid CO2 in the storage tank and provide the necessary CO2 flow rates (7) Table 5 lists the pump energy, pump operating costs, and material (injected CO2) costs using the data from Tables 3 and 4. The pump used diesel fuel, but these calculations project the case of an electric pump located at a permanent installation of similar size. The industrial electrical rates in Kaukauna, Wisconsin, were about $0.10/kWh at the time of the experiments. The cost of CO2 delivered to Kaukauna, Wisconsin, was $0.24/kg ($0.11/lb).Table 5. Pump energy, electricity costs, CO2 costs, and total costs for each trial in Kaukauna, WisconsinTable 5. Pump energy, electricity costs, CO2 costs, and total costs for each trial in Kaukauna, WisconsinTest name (gal./min-lb/h)t=40  minElectricity costs ($)CO2 costs ($)Total cost ($)Pump energy (kWh)$0.10/kWh$0.24/kg$0.11/lbF 2200-250039.43.9464.6668.60F 2200-350044.34.4340.9245.36F 3000-2500102.810.2844.9455.22F 3000-3500106.310.6376.7887.40W A-B 1800-250036.43.6432.5236.16W A-B 1800-350037.53.7520.1323.88W A-B 2800-2500110.611.0621.2632.31W A-B 2800-3500113.511.3520.0131.36W A-B-C 2200-250045.64.5625.8930.45W A-B-C 2200-350043.84.3828.2932.67W A-B-C 3000-250079.87.9830.6038.58W A-B-C 3000-350085.48.5441.8250.36W A-B-C-D 3200-250073.57.3539.7347.08W A-B-C-D 3200-3500W A-C-E-G 3200-250066.46.6455.9262.57W A-C-E-G 3200-350070.57.0567.8874.92Operating pressure and the resulting operating power clearly affect the electricity cost for each configuration of manifolds. However, the greater cost driver was the CO2 consumption required to treat the lock. This highlights the benefit of setting the water and CO2 flow rates at levels that optimize the injection efficiency and minimize off-gassing, as shown in Fig. 17. Ultimately, the combination of operating conditions and manifold configurations that meets the injection requirements (100  mg/L

Source link