The study was conducted at the Bio-Environmental Engineering Center (BEEC) at Dalhousie University's Agricultural Campus in Truro, Nova Scotia, Canada (45°45' N, 62°50' W). The research site contained 6 in-ground, cement, meso-scale manure tanks (6.6 m² and 1.8 m deep). This site has been previously described by Wood et al. (2012) and Le Riche et al. (2016). Each tank was filled with 10.6 m³ (160 cm depth) of liquid dairy manure, consisting of 20% inoculum (2.1 m³) and 80% fresh manure (FM; 8.5 m³). Manure was obtained from a local diary operation which housed 95 lactating cows in a free stall barn. The manure was gathered from an in-ground manure tank adjacent to the dairy barn and was a mixture of feces, urine, and sand bedding.

Two types of inoculum were used in this study: (i) 1-year-old untreated manure, and (ii) 1-year-old manure that was previously acidified (Table 1). This manure inoculum was obtained from the same farm in spring 2017 (12 months prior to the start of this trial). The previously acidified (PA) manure was acidified using sulphuric acid (70% H₂SO₄; 2.4 L m⁻³ manure) to pH 6. Both the PA manure and untreated manure inoculum remained in storage for 1-yr (Sokolov et al., 2019). Additional information about storage conditions of the inoculating manure prior to this study can be found in Sokolov et al. (2019).

The six manure tanks were assigned within two blocks, each containing two treatments and a control. Inoculum was prepared on May 15–16, 2018 by pumping out of old storages and distributing 2.1 m³ to new storages using a pumping truck. The newly acidified (NA) inoculum treatment received 2.1 m³ of 1-yr-old untreated inoculum and was acidified on May 17, 2018 with 1.1 L m⁻³ 70% H₂SO₄ (i.e., 12 L per 10.6 m³ tank). The PA inoculum treatment received 2.1 m³ of 1-yr-old inoculum which had been acidified the previous year (spring 2017) at 2.4 L m⁻³ (i.e., 5.04 L added to 2.1 m³; Table 1). Lastly, the control received 2.1 m³ of 1-year-old untreated manure inoculum. Fresh manure was added to each tank on May 28 and 29, 2018 using a pumping truck to transport manure from the farm to the research site. The NA inoculum treatment received 1.1 L m⁻³ H₂SO₄, which is half as much as a previous meso-scale study (Sokolov et al., 2019) but considerably more than in the laboratory study where rates were only 0.16 L H₂SO₄ m⁻³ of total manure (i.e., 0.03 mL of 98% H₂SO₄ in 180 mL of stored manure; Sokolov et al., 2020).

Sulfuric acid (industrial grade) was obtained from Bebbington Industries (Dartmouth, NS) and was pumped into the inoculum manure using acid resistant tubing and a peristaltic pump. The tubing was attached to an aluminum pole which was moved around the inoculum as the acid was being pumped.

Emissions of CH₄ and N₂O were monitored continuously from Jun 8 to Nov 10, 2018 (155 days). Each manure storage tank was covered by a flow-through, steady-state chamber (~13 m³ headspace) consisting of an aluminum frame and 0.15 mm greenhouse plastic. Air was pulled through the chamber through intake slits at the front of the chamber and out through an exhaust fan and outflow exhaust duct and the opposite end. The rate of air flow within each chamber was approximately two full air exchanges per minute (~0.5 m³ s⁻¹). The airspeed was measured within each exhaust duct using cup anemometers recorded by a CR1000 data logger (Campbell Scientific, Edmonton, AB). The air temperature within each chamber was measured using copper-constantan thermocouples at 10 cm above the manure surface and along with manure temperature at 80 cm depth and 150 cm depth, recorded by the same CR1000 data logger (Campbell Scientific, Edmonton, AB). Ambient air temperature was obtained from the nearest Environment Canada climate station (Debert, NS, 45.42 N, 63.42 W; Climate ID: 8201380).

Air samples were continuously pumped (RC0021, Busch Vacuum Pumps and Systems, Boisbriand, QC) from the exhaust duct of each tank and two ambient inflow locations, and carried through polyethylene tubing (3.2 mm i.d.; Rubberline Products Ltd., Kitchener, ON) to a 8 × 2 manifold (Campbell Scientific In., Logan, UT) containing 12 V DC valves (The Lee Co., Essex, CT). The valves directed two samples every 30 s through high-flow air dryers (Perma Pure LLC.; Toms River, NJ) and into one of two tunable diode trace gas analyzers (TDLTGA, Campbell Scientific, Logan, UT). Sample CH₄ and N₂O concentrations were continuously recorded by a CR5000 data logger (Campbell Scientific Inc., Logan, UT) and an adjacent PC computer monitored the analyzer performance by running the TDLTGA software (Campbell Scientific, Logan, UT).

Concentrations were averaged hourly and used to calculate flux rates using to the following:

where F is the hourly flux (mg m⁻² h⁻¹), Q is the flowrate of air out of the chamber [m³ h⁻¹; calculated using average hourly windspeed × cross-sectional area of the exhaust duct (0.0645 m²)], A is the surface area of the manure surface (6.63 m²), and C is the concentration of gas (mg m⁻³) in the ambient inflow air (C_i) and sample outlet air (C_o).

Due to technical issues, block one had missing flux data Aug 21-Sep 2, and block two had missing data Jul 18-Sep 2. This resulted in missing the peak fluxes in block two tanks. Linear interpolation was used to estimate the missing data, although the values were likely underestimated. All values are presented as treatment average.

Ammonia concentrations were determined using 125 mL 0.005 M H₃PO₄ acid traps. Three times per week, air was pumped (Model 2107CA20B; Thomas Pumps and Compressors, Sheboygan, WI) from the exhaust of each tank and two ambient inflow locations and bubbled through acid traps (dispersion tubes id = 35 mm) at 1.5 L min⁻¹. Air was continually pumped through the traps for 24 h at each deployment. Airflow for each sample was measured using inline flow meters (Gallus 2000; Actaris Metering Systems, Greenwood, SC). Following deployment, evaporated liquid was replaced to 125 mL and a sample frozen until analysis. Samples were shipped to Agriculture and Agri-Food Canada (Ottawa, ON) where they were analyzed for NH₃-N using the QuikChem® Method 12-107-06-2-A modified for 0.005 mol L⁻¹ H₃PO₄ matrix using a Lachat QuikChem FIA+ Q8500 Series. Daily gas concentrations were calculated using the following:

where C_{NH3 air} is the daily NH₃-N concentration (mg m⁻³), C_{NH3 aq} is the NH₃-N concentration in sample liquid (mg L⁻¹), V_aq is the volume of liquid in the acid trap (L), and V_air is the volume of air pumped through the acid (m³) (Hofer, 2003).

Ammonia emissions on days that were not sampled were estimated using linear interpolation and daily total NH₃-N losses were added together to find the entire monitoring period.

Six FM composite samples were taken during tank filling (May 29). Manure in each tank was sampled monthly throughout the study with one sample per tank made from a composite of 12 subsamples. Subsamples were taken from each tank in a grid at two depths and six locations. All samples were kept frozen until analyzed at the Nova Scotia Department of Agriculture's Provincial Soils Lab (Bible Hill, NS). Samples were analyzed for total solids (TS) and volatile solids (VS) (American Public Health Association method 2540 B), total nitrogen (TN) (combustion method AOAC 990.03-2002), ammonium-N (TAN) (American Public Health Association method 4500-NH₃ B), and pH using an electrode (American Public Health Association method 4500-H⁺) (Clesceri et al., 1998). To verify pH, a FieldScout pH 400 meter (Spectrum Technologies, Aurora, IL, USA) was used to measure pH in the manure at 10, 50, 100, and 150 cm across 6 locations in each tank (24 pH points) on May 26, Jul 1, and Jul 31, 2018. These are not reported in the paper but verify the results of lab analysis.

For microbial analysis, duplicate composite samples were taken during storage tank filling (May 29, 2018) of FM, untreated inoculum, and previously acidified inoculum. Throughout the study, monthly composite manure samples were collected in duplicate and kept frozen until nucleic acid extraction. For each sampling, ~2 g of manure sample was stored in 5 mL of LifeGuard soil preservation solution (MoBio Laboratories Inc., Carlsbad, CA).

Based on the typical CH₄ emission curve, three sampling dates and starting FM and the two inoculums were chosen for analysis. The DNA or RNA PowerSoil total DNA/RNA isolation kit with RNA/DNA elution accessories (MoBio Laboratories, Inc., Carlsbad, CA) were used for DNA or RNA extraction. In triplicate, 8 μL of each extracted RNA sample was reverse transcribed to cDNA using Maxima First Strand cDNA Synthesis Kit (Thermo Scientific, Waltham, MA) following manufacturer's protocols.

Real-time qPCR was performed using an Applied Biosystems StepOnePlus real-time PCR system using clear 96-well PCR plates (Bio-Rad Laboratories, Inc., Hercules, CA). The total and active fraction of the methanogen populations were quantified by targeting methyl coenzyme A reductase (mcrA) genes and transcripts, respectively, using mlas-mod F and mcrA-rev-mod R primers (Habtewold et al., 2018). The total and active fraction of the methanogen populations were quantified by targeting methyl coenzyme A reductase (mcrA) genes and transcripts, respectively, using mlas-mod F and mcrA-rev-modR primers (Habtewold et al., 2018). Methyl Coenzyme M Reductase A gene is a fragment of DNA commonly found in methanogens that encodes the α-subunit of methyl-coenzyme M reductase enzyme which catalyzes the final step in methanogenic pathway (i.e., releases CH₄) (Evans et al., 2019). Although methanogens may involve one or more of the methanogenic pathways (i.e., hydrogenotrophic, acetoclastic, and methylotrophic), all pathways share the final step which requires methyl-coenzyme M reductase enzyme. Thus, the functional mcrA gene has been used extensively to effectively target all methanogens (Evans et al., 2019). Each reaction well-contained 10 μL of Ssofast EvaGreen supermix (Bio-Rad Laboratories, Inc.), 1 μL (10 pM) of each primer, 2 μL of DNA or cDNA, and 6 μL of PCR-grade water. Plasmid standard curves were prepared for mcrA from Methanosarcina mazei (ATCC 43340), and for 16S rRNA genes, plasmid with 16S rRNA gene insert from soil bacterium Clostridium thermocellum was used. Although primer sets used to target mcrA or 16S rRNA genes are specific to the respective gene fragments, target specificity of the primers were also confirmed by assessing the presence of a single district peak for melting curves (fluorescence vs. temperature) of each target gene. The mcrA gene standard curve had an efficiency of 101.6%, r² of 0.99, and slope of −3.29. The highest diluted standard had a cycle of quantification of 30.2 and no-template controls of 31.4. The 16S rRNA gene standard curve had an efficiency of 100.1%, r² of 0.998, and slope of −3.32. The highest diluted standard had a cycle of quantification of 27.0 and no-template controls of 28.9. StepOne software v2.3 (Bio-Rad Laboratories, Inc., Hercules, CA) was used to calculate sample copy numbers.

To compare treatments based on their global warming potential, GHG emissions were converted to 100-yr CO₂-equivalent (CO₂-eq) values and summed. Conversion values for the global warming potentials of CH₄ and N₂O were 34 and 298, respectively (IPCC, 2014). The contribution of indirect N₂O emissions from NH₃ volatilization were calculated using the IPCC emission factor of 0.01 (Dong et al., 2006).

Given that PA inoculum could reduce the need for acidification following every other emptying event, to compare use of PA inoculum and NA inoculum it is necessary to compare estimated total emissions over two storage periods. The total PA inoculum over two storage periods was calculated using the following:

where Total CH₄ is the total production over two storage periods, Acidified manure CH₄ is the production from one storage period where all manure was acidified (reported by Sokolov et al., 2019), and PA inoculum CH₄ is the total CH₄ production from the PA inoculum treatment.

The NA inoculum for two storage periods was calculated by doubling the total CH₄ production from the NA inoculum treatment. Lastly, the control for two storage periods was calculated by doubling the total CH₄ production from the controls. Note that this assumes both storage periods to have the same temperatures. Therefore, the two storage periods do not represent spring/summer and fall/winter, as emissions would be dramatically different during cold weather storage.

For each treatment the methane conversion factors (MCF) was calculated following IPCC methods (Dong et al., 2006). The calculation used the average VS of FM (disregarding VS of the inoculum) and maximum potential CH₄ production (B_o) of 0.24 m³ CH₄ kg⁻¹ VS. Cumulative N₂O and NH₃-N emissions for each tank were scaled by TN and TAN in FM and then averaged for each treatment.

Treatment effects were assessed using repeated measures, mixed linear model analysis using PROC Mixed in SAS software (SAS Institute Inc., Cary, NC) using the Kenward-Roger fixed effects method on total biweekly CH₄, N₂O, and NH₃ emissions. The CH₄ data was skewed and therefore log transformed to conform to normality. The spatial Gaussian covariance structure was chosen based on best fit statistics. Significance was considered when p < 0.05. Treatment effects on mcrA and 16S rRNA gene and transcript copy numbers over all dates were assessed using a general linear model using PROC GLM in SAS software, which uses ordinary least squares with Sidak adjustment to control familywise error. Effect size was calculated using partial eta² (η_p²). Significant results were followed up with a post hoc Sidak groupings comparison using a significance of p < 0.05.

Average ambient air temperature during the study (Jun 1–Oct 31, 2018; 160 d) was 15.1°C (with an average relative humidity of 77%) as recorded by the closest Environment Canada climate station. The 30-years normal for this location, Jun–Oct, is 14.7 ± 0.2°C (Mean ± SD). The temperature inside the tank chambers (10 cm above the manure surface) was 17.6 ± 0.8°C, which was, on average, 2.6°C warmer than the ambient air (Figure 1). The average manure temperature in the tanks was 13.7 ± 0.1°C at 150 cm depth and 17.6 ± 0.1°C at 80 cm depth. The manure temperature peaked at week 12 (d 68, Aug 14) at 80 cm (20.9°C) and week 15 (d 91, Sep 6) at 150 cm (15.6°C). The CH₄ production followed a similar pattern, peaking a week earlier (d 61, Aug 7). Following the peak, the 80 cm temperature quickly fell. By the end of the study the temperature at 80 and 150 cm were both on average 14.3°C (~122 days, Oct 7).

The manure pH had no clear treatment differences until days 85 into the study (Figure 1). The control was expected to have the highest pH, but was on average the lowest on Jun 15, 2 weeks after the start of the study. The pH dropped throughout storage until Sept when it increased. By Sept 8, the control tanks had the highest pH and the NA inoculum tanks had the lowest. This trend continued in Oct as well (Figure 1).

The TS, VS, and TN were all highest in the FM and fell markedly by Jun 15 (Table 2). This is most likely due to settling of solids which occurs rapidly following storage tank filling, and issues with unrepresentative sampling of the manure depth (Sokolov et al., 2019). The control had consistently the least VS, TS, and TN (Table 2). This may be due to faster degradation of organic matter and loss of TN to the atmosphere, although given the small differences, it may also be due to natural variability in manure (Sokolov et al., 2019).

On a CO₂-eq basis, the total GHGs were 94–97% comprised of CH₄ emissions, due to the anaerobic conditions within the manure storages (Table 4). Clear treatment difference was observed, where PA inoculum reduced total GHGs by 38% and NA inoculum reduced total GHGs by 77%, compared to control. All sources of GHG were reduced due to PA and NA inoculums, although CH₄ was the most important in reducing total GHGs.

The results of the 2-way ANOVA on copies of mcrA, mcrA transcript, 16S rRNA, and 16S rRNA transcript are shown in Supplementary Table 1. There were significant treatment effects on mcrA (p <0.0001) and 16S rRNA (p <0.0126) but not in mcrA or 16S rRNA transcript.

The FM had higher copies per gram of dry manure of mcrA transcript and bacterial 16S rRNA genes and transcript than untreated inoculum (30–45%; percentages are calculated on values prior to log transformations) sampled prior to the start of the study (Table 5). An exception was mcrA gene in untreated, control inoculum which had 64% more copies per gram of dry manure of than FM. These results differed from Habtewold et al. (2018) who reported more (11–458%) copies of genes and transcripts of both mcrA and bacterial 16S in inoculum compared to FM. At the start of the trial, previously acidified inoculum had lower mcrA copies of genes and transcript (88%) and lower 16S rRNA genes and transcripts (90%) compared to the untreated inoculum. Given that both inoculums were stored for 1-year under the same conditions, the difference in abundance are likely due to acidification with H₂SO₄ 1-year prior.

Averaged over the entire study period, the control had significantly more mcrA gene copies compared to PA inoculum (39%) and NA inoculum tanks (65%, p < 0.05; Table 5). The PA inoculum tanks have significantly more mcrA gene copies than NA inoculum tanks (43%, p < 0.05).

The mcrA transcript copies were variable over time, although the most marked difference between treatments was Jul 27 (d 42) when the NA inoculum and PA inoculum were 95 and 85% less than the control copies, respectively. This corresponds with the initial increase in CH₄ emissions. The average CH₄ emissions during the sampling week were 43.0, 7.70, and 4.12 g m⁻² d⁻¹ from control, NA inoculum, and PA inoculum treatments, respectively.

On Sept 8 (85 days) the NA inoculum treatment had the highest copies mcrA transcript, with the control and PA inoculum having 97 and 86% fewer copies, respectively (Table 5). This corresponds to CH₄ emissions during the sampling week of 43.4, 55.5, and 21.4 g m⁻² d⁻¹ from control, NA inoculum, and PA inoculum, respectively.

Lastly, on Oct 31 (108 days) the PA inoculum had the highest copies of mcrA transcript, with control and NA Inoculum having 81 and 86% fewer copies, respectively (Table 5). This corresponds to CH₄ emissions during the sampling week of 30.0, 34.2, and 11.0 g m⁻² d⁻¹ from control, NA inoculum, and PA inoculum, respectively.

The 16S rRNA gene copies varied less over time and between treatments (Table 5). The 16S rRNA transcript copies in the control treatment increased and decreased following the same pattern as the mcrA transcript copies. This pattern was not observed in the NA and PA inoculum, suggesting that the methanogen and bacterial communities had differing influences.