Trials were completed between August 14 and 29, 2018, in southeastern Michigan (fig. 1). Three test sites were chosen and included two golf course ponds and a single urban water retention pond with known RSC infestations. In addition to the three test ponds, two untreated control ponds in Livonia, Mich., were chosen. Due to heavy rainfall the week before testing, two of the test ponds were flooded rendering them untreatable. The remaining treatment site was a water retention pond in Novi, Mich. The treatment pond had a volume of ~2,500 cubic meters (m³) and a surface area of ~0.9 hectare (ha). Control pond 1 had a volume of ~400 m³ and a surface area of ~0.2 ha, and control pond 2 had a volume of ~2,000 m³ and a surface area of ~0.7 ha.

Ponds were prepared for testing by removing brush along the shorelines and installing silt fencing as a barrier approximately 1 meter (m) from the shoreline. A folded shade cloth was then placed between the shoreline and the silt fence barrier around each pond. The barrier was necessary to prevent RSC escapement and served as a boundary for perimeter surveys. Shade cloth between the barrier and shoreline provided a terrestrial habitat and a collection point for emergent crayfish (Abdelrahman and others, 2021). Minnow traps (Tackle Factory, Model G40M, 6.35-millimeter [mm] mesh) baited with dry dog food were also placed in control pond 1 (C1), control pond 2 (C2), and treatment pond (TP) throughout testing. Trap openings were increased slightly to approximately 3 centimeters (cm) to accommodate larger crayfish sizes following protocols detailed by previous studies (Capelli and Magnuson, 1983; Hein and others, 2006). Baited traps were placed in C1, C2, and TP according to a fixed distance sampling design (fig. 1). Traps were placed the day before treatment application and checked once daily to evaluate crayfish captures. Perimeter surveys were completed in the CO₂-treated pond approximately every 2–4 hours before, during, and after treatment application. Control ponds were also opportunistically surveyed, but timing was not standardized due to limited access to those sites at certain times throughout the day and night. Visual surveys were usually completed in two-person teams; one person waded in the water and the other walked on the shore near the silt fence. Locations where crayfish were collected during perimeter surveys were identified as (1) outside of the silt fence, (2) inside the fence, (3) under the shade cloth flap, or (4) in the water. Crayfish were collected by hand when they were on land and gigged when they were in the water. All crayfish collected from traps or perimeter surveys were removed from the pond and euthanized at the time of their collection.

Water quality was measured in C1, C2, and TP throughout testing. Baseline water quality was determined by collecting grab samples from each pond approximately 1.2 m from shore and 0.15–0.30 m under the water’s surface (fig. 1). Grab samples were collected approximately every 2 hours in the TP. Less frequent grab samples were collected from C1 and C2 due to limited golf course access at certain times of the day and night. Grab samples were collected from north, west, east, and south locations at the TP, and a single location at C1 and C2 (fig. 1). Intermittent water-quality samples were collected from the center of the TP. Samples were collected with a 1-liter (L) container and sealed. Each grab sample was analyzed for dissolved oxygen (HACH Model HQ40d, Hach Company, Loveland, Colorado), pH (HACH Model HQ40d), water temperature (Thermapen MK4, ThermoWorks, American Fork, Utah), CO₂ (HACH Method 8205), ammonia-N (HACH DR3900 photospectometer, TNT 830 test kits), alkalinity (HACH DR3900 photospectometer, TNT 870 test kits) (APHA, 2012), and hardness (HACH DR3900 photospectometer, TNT 869 test kits) (APHA, 2012). Autonomous water-quality sondes (YSI 600XLM, YSI, Yellow Springs, Ohio) were also deployed in each pond to collect time-series pH and water temperature every 30 minutes before, during, and after testing. Sondes were attached to a rope with a buoy and suspended approximately 0.15 m below the surface in C1 (water depth of 0.30 m), 0.20 m below the surface in C2 (water depth of 1.22 m), and 0.3 m below the surface in the TP (water depth of 1.22 m). Time-series pH data from sondes were converted to CO₂ concentrations using a standard curve relationship determined from grab sample analyses (fig. 2). Sondes remained in ponds until August 29, 2018, which is when the final water-quality samples were collected and analyzed. The water level had increased by greater than 0.30 m from rainfall and runoff by the end of the study.

Atmospheric CO₂ data were collected with a handheld meter (pSense Model AZ–0002, CO2meter.com, Ormond Beach, Florida) at the surface and 1 m above the surface of the ponds at water-quality grab sample locations. Atmospheric CO₂ was measured before, during, and after treatment. The purpose of atmospheric sampling was to ensure staff safety and quantify CO₂ exposure levels.

Carbon dioxide was injected into the TP to obtain a target concentration of 200 mg/L. This concentration was a benchmark established during pilot testing to elicit RSC emergence behaviors in outdoor ponds (Abdelrahman and others, 2021). Briefly, 24 porous air diffusers (Sweetwater Model: DYFP16, Pentair, Apopka, Florida) were attached to small bricks and submerged equidistantly throughout the pond. Each diffuser was attached to clear tubing (6.35 mm inner diameter [I.D.]) that connected to gas flow manifolds (oxygen flow meter manifold, Model: 1FMM222, Pentair, Apopka, Florida) on the shore. Manifolds were then connected with flexible tubing (9.525 mm I.D.) directly to heated gas regulators on CO₂ storage tanks (160-L liquid dewars, Airgas Inc.). Twenty-five CO₂ tanks were split between the north and east sides of the pond. Tanks were covered with canopies to reduce direct exposure to sunlight to minimize tank pressure, which could accelerate pressure relief of CO₂ into the atmosphere. Carbon dioxide treatments in TP began at approximately 1100 hours on August 21, 2018, and ended at approximately 0200 hours on August 24, 2018. Gaseous CO₂ was continuously injected at a flow rate of 15 liters per minute (maximum flow rate of gas flow manifolds) per tank until all tanks were empty.

The removal of two treatment ponds due to flooding reduced the number of treatment site replicates to one and confounded any possible inferential comparisons. Therefore, all water quality, air quality, trapping, and perimeter survey data were summarized for each pond using basic descriptive measures (for example, mean and variance estimates). Data summaries are intended to broadly describe outcomes from this study to inform future treatments and management actions. Inferences based on statistical significance are not included. All data collected for this report are available through a U.S. Geological Survey data release, https://doi.org/10.5066/P99OUHMV (Smerud and others, 2022).

Treatment applications met and exceeded target CO₂ concentrations. Maximum concentrations measured in TP were 282 mg/L of CO₂ (titrated from grab samples; table 1) and 352 mg/L of CO₂ (calculated from sonde pH measurements; fig. 3). The discrepancy in maximum concentrations between the two sampling methods was likely attributed to sample locations. Sondes were placed mid-depth in the center of the pond, whereas grab samples were collected near-shore. Target concentrations were achieved approximately 11 hours after treatment initiation and were sustained at >200 mg/L of CO₂ for approximately 68 hours. Maximum concentrations were reached at approximately 48 hours into treatment (fig. 3). Control ponds showed only minor fluctuations in ambient CO₂ concentrations with a range of 1.6–3.8 mg/L from grab samples and 0–14.5 mg/L from water-quality sondes (table 2).

Dissolved oxygen in the TP was 8.92 plus or minus (±) 0.85 (mean±standard deviation) mg/L and was 8.53±0.84 mg/L in the control ponds. Mean pH in the TP was 6.77±0.57 and was 8.49±0.85 in the control ponds. The pH suppression observed in the TP was expected because CO₂ mixed into water produces an equilibrium reaction with bicarbonate and carbonic acid. Minor pH variations in the untreated control ponds were likely attributed to diurnal photosynthetic activity. Water temperature was 22.5±1.90 degrees Celsius (°C) in the TP and 23.9±2.11 in the control ponds. Alkalinity (as milligrams per liter of calcium carbonate [CaCO₃]) was 347±135 in TP and 133±49.6 in the control ponds. Hardness (as milligrams per liter of CaCO₃) was 371±71.4 in TP and 131±43.2 in the control ponds. Ammonia (as milligrams per liter of unionized ammonia [NH₃-N]) was 0.017±0.002 in the TP and 0.027±0.016 in the control ponds. Complete summaries of water-quality data are documented in tables 1, 2, and 3.

Crayfish responses were monitored during injection using a combination of perimeter surveys and baited traps. Overall, catch rates for both sampling methods were relatively low throughout injection. Perimeter survey collections showed a rapid increase in crayfish captured in the water near shore on the second and third day of injection (fig. 4). However, catches on dry land at the other three sample locations saw a slight decline in captured crayfish. Results from baited traps showed a strong temporal decline because of catch rates in the treatment pond starting several weeks before injection (fig. 5). At the start of injection, control and treatment ponds had relatively similar catch rates. Control ponds had consistently low catch rates throughout the summer. Catch rates in the TP at the time of injection were low, which made discerning differences between treatment and control ponds not obvious (fig. 6). Further, there was no clear change in baited trap catches after injection. The low catch rates in all ponds continued throughout the end of trapping season. One possible explanation for the temporal decline in catch rates in the TP before injection was the corresponding decrease in water temperatures (fig. 7). Catch rates seemed to peak when water temperatures were at their highest and then declined with decreasing water temperature. However, this was not corroborated with control ponds where catch rates were consistently low.

Visual observations of crayfish and certain fish species were documented throughout testing. Approximately 12 crayfish were observed congregating during treatments on a concrete culvert on the west side of the pond. Crayfish were observed near the air-water interface, and most observations were made during the morning perimeter surveys. On the last day of treatment, four crayfish were observed displaying loss of equilibrium, presumed as narcotization from prolonged CO₂ exposure, in the northeast corner of the pond. After 12 hours into CO₂ application, large masses of fathead minnows (Pimephales promelas; Rafinesque, 1820) were observed porpoising at the water's surface. At 24 hours, a few small green sunfish (Lepomis cyanellus; Rafinesque, 1819) and four adult-sized channel catfish (Ictalurus punctatus; Rafinesque, 1818) exhibited anesthetic behaviors and loss of equilibrium. Mortalities included one catfish, three green sunfish, and several thousand fathead minnows by the conclusion of the treatment. No crayfish mortalities were documented.

Atmospheric data displayed minor fluctuations in the treatment pond when CO₂ was applied (table 4). The maximum atmospheric CO₂ concentration measured at the TP shoreline, 1 m above the water surface, was 1,415 parts per million (ppm). Control ponds reached a maximum of 400 ppm, which was consistent with ambient atmospheric CO₂. The highest single atmospheric CO₂ level, away from the TP, was 2,390 ppm and was recorded inside the enclosed canopy tent where tanks were stored. All atmospheric CO₂ readings were well below the 5,000 ppm 8-hour Time Weighted Average and 30,000 ppm acute 15-minute exposure limits for human safety (National Institute for Occupational Safety and Health [NIOSH], 2019).