Manure Application Timing and Incorporation Effects on Ammonia and Greenhouse Gas Emissions in Corn

Abstract

Manure application influences ammonia (NH₃) and greenhouse gas emissions; however, few studies have quantified the effects of manure application methods and timing on NH₃, nitrous oxide (N₂O), carbon dioxide (CO₂), and methane (CH₄) fluxes simultaneously. We evaluated surface-applied liquid manure application with disk incorporation versus injection on NH₃, N₂O, CO₂, and CH₄ fluxes in central Wisconsin corn silage (Zea mays L.) plots during pre-plant (PP) and sidedress (SD) application windows from 2009 to 2011. Manure treatments were PP injection (PP-Inject) and injection at sidedress time (SD-Inject) to growing corn, along with two incorporation times for PP surface application (within 24 h—PP-1-hr; within 3 days—PP-3-day). Mean NH₃ emissions were 95% lower for injected treatments compared to surface application in both years, with larger losses for PP-3-day and SD-Surf. While N₂O fluxes were generally low, larger increases after manure application were associated with injection and triggered by soil moisture/temperature changes. Mean CO₂ and CH₄ were unaffected by manure treatments and influenced more by weather. Overall, injection conserved more available soil N while contributing to modest N₂O emission, suggesting manure injection may offer greater agri-environmental benefits on the whole over surface application.

1. Introduction

Ammonia (NH₃) losses from surface-applied manure can be large, reducing plant-available N and the economic value of manure as a N source. Ammonia emission into the atmosphere is an environmental concern because NH₃ can combine with sulfur and nitrogen oxides to form fine particulates that can contribute to human health problems [1]. It may also contribute to the eutrophication of surface waters (especially marine and estuarine) via atmospheric deposition [2]. Volatilization of N as NH₃ and deposition downwind additionally can serve as a source of indirect N₂O emissions [3]. The decreased amount of available N in manure reduces the N:P ratio, potentially leading to a more rapid build-up of soil P per unit of applied manure N, increasing the potential for P loss in runoff.

A common approach for controlling NH₃ volatilization from manure is incorporation into the soil with tillage or injection equipment, typically reducing NH₃ losses by 50 to >90% compared to broadcast [4,5,6,7,8,9]. Manure application timing also affects N loss potential and availability to crops. Injecting liquid dairy manure into a growing corn crop at or near the early season N application window (i.e., sidedress application) might be used as a viable substitute for commercial fertilizer to meet corn N demands [10,11]; however, studies indicate that manure incorporation increases N₂O fluxes compared to broadcast/surface application due to greater NH₃/NH₄₊ conservation [8,9,12,13,14,15,16,17]. Ammonia is considered a secondary contributor of N₂O. The IPCC [3] assumes that 1% of the N volatilized as NH₃ could be released as N₂O after redeposition on the land. In a systems agri-environmental context, the larger N₂O fluxes for manure injection/incorporation might be offset by more efficient NH₃-N capture and lower secondary N₂O emission relative to surface application [18].

Compared to N₂O, manure application effects on CH₄ and CO₂ fluxes are less consistent and appear to be more tied to weather, especially soil moisture and temperature [13,16,19,20,21,22,23,24]. Rhizosphere dynamics in maize systems have important effects on carbon cycling via microbe–root–soil interactions, with root respiration and rhizodeposition both being important processes affecting net CO₂ fluxes [19]. In general, CH₄ pulses from manure application typically occur within the first few days after injection; however, annual CH₄ fluxes vary among soils, with some acting as net CH₄ sinks or sources depending on landscape attributes, drainage, and weather conditions [14,16,25,26,27]. Recent studies suggest that manure could be a larger source of greenhouse gases (GHGs; CO₂, N₂O, and CH₄) compared to organic amendments and that biochar addition can substantially reduce GHGs compared to manure application [20,21]. Overall, relatively few studies have quantified the effects of dairy manure incorporation method and timing on both NH₃ and GHG fluxes in humid, temperate corn silage systems. The objective of our study was to evaluate the effect of liquid dairy manure application method and timing (shallow injection or disk incorporation at two times) on emissions of NH₃ and GHGs in a central Wisconsin corn silage production system.

2. Materials and Methods

This field research was conducted at the University of Wisconsin/USDA Agricultural Research Station in Marshfield, Wisconsin (WI), from 2009 to 2011. To reduce the potential for residual manure N effects, a new site was selected each year that had a previous corn crop. Soil on all sites was predominantly Withee silt loam (fine-loamy, mixed, superactive, frigid Aquic Glossudalfs), a somewhat poorly drained soil with 0 to 2% slope. Soil pH and organic matter content averaged 6.7 and 30 g kg⁻¹, respectively. A 92-day relative maturity corn silage hybrid was planted on 19 May 2009, 18 May 2010, and 3 June 2011 in 0.76-meter rows at 86,500 seeds ha⁻¹ with 112 kg ha⁻¹ of 9-11-30-6S-1Zn starter fertilizer in a 2 × 2 configuration (50 mm to the side of seed row, 50 mm deep).

Liquid dairy manure was applied either during the pre-plant (PP) window (mid- to late May) or in-season (SD) (5–6-leaf stage; late June to early July) (

Table 1. Manure application dates, nutrient measures, and application rates.

		Manure Nutrient Measures			Application Rate
Date	Time	Solids	TN	NH4-N	TN	NH4-N
		%	g L−1		kg ha−1
15-May-09	Pre-Plant	16.6	3.1	1.2	185	74
23-June-09	Sidedress	11.9	5.4	2.4	330	143
10-May-10	Pre-Plant	23.9	3	1.1	184	65
30-June-10	Sidedress	22.5	2.9	1.1	176	70
26-May-11	Pre-Plant	13.0	1.8	0.7	111	42
06-July-11	Sidedress	15.9	1.7	0.6	104	39

). Pre-plant treatments were either injected (PP-Inject) or incorporated with a tandem disk immediately after manure application (<1 h) (PP-1-hr) or 3 days later (PP-3-day). Injection was performed with an S-tine (Kongsgilde Vibro-flex) injector (68-mm width) with 0.38 m spacing at a 0.10- to 0.15-meter depth (Figure 1). All plots were chisel plowed 3 to 5 days after manure application. Sidedress (SD) manure applications were either injected with an S-tine injector (0.76-meter spacing) equipped with shields (SD-Inject) or surface-applied (SD-Surf) with the same means with the injectors raised above the soil surface (2010 and 2011) (Figure 1). Manure (average solids content = 14%) was applied at a target rate of 62,000 L ha⁻¹. The manure supplied an average of 177 kg total N and 69 kg NH₄-N ha⁻¹, but rates varied somewhat across years and application times (Table 1). Plots (4.5 by 15 m) were replicated four times per treatment in a randomized complete block design. Ammonia emission was measured from 2009 to 2011 and GHG emission was measured in 2010 and 2011 (only three replicates were sampled in 2011).

Ammonia emission was measured using the dynamic chamber/equilibrium concentration technique [28,29] that is well suited to small, replicated plots and has been used successfully by others [8,9,29,30,31,32]. We placed two 31 by 38 by 20 cm ventilated chambers and an open collector, or ambient meter, in each plot. Duplicate passive diffusion samplers of two types were placed in each chamber and in each ambient meter: one with an acidified filter paper disk directly exposed to the air and the other with the filter paper disk 10 mm below a semipermeable Teflon membrane, requiring NH₃ to diffuse along a 10-millimeter path to the acidified trap. Ammonia flux was calculated based on the micrometeorological law of resistance, using NH₃ concentrations from diffusion samplers to estimate the required parameters. Details on chamber design and further calculations are found elsewhere [28,30,32]. Measurements started immediately after manure application and continued for six separate periods through day 3. Day 1 measurements started immediately (Time 0), with successive periods starting approximately 1, 3, and 8 h (overnight) after application, followed by 10-hour measurements during days 2 and 3 (not overnight). Overnight emission between day 2 and day 3 was estimated from linear interpolation adjusted for measured temperature and wind conditions [29,33]. Ammonia measurement ended just before disking of the 3-day incorporation treatment, so the 3-day treatment represents surface-applied manure for NH₃ measurement.

Nitrous oxide, CO₂, and CH₄ were measured using the static vented chamber technique following the GRACEnet protocol [33]. Chambers consisted of galvanized steel utility pans 38 cm in diameter and 10 cm deep. A sample port was placed in the center of the utility pan bottom and a vent tube (3 mm ID and 40 cm long) was installed horizontally on the side and coiled inside the pan. Weather stripping was attached along the pan lip to serve as a gasket, and the entire chamber was covered with reflective insulation to minimize temperature changes. Bases were made by cutting out the bottom section of the utility pan. The base was then pressed into the soil, leaving 1–2 cm exposed above the soil surface. During sampling, the chamber top was placed on the base and secured with four binder clips. Chamber construction was based on a design from R. Venterea (http://www.ars.usda.gov/pandp/docs.htm?docid=19008 accessed on 15 April 2010). Bases were left in the field for the full season and removed and replaced only for field operations. Gas samples were collected by inserting a 10-milliliter syringe into the port, removing a sample, and immediately transferring the sample to a 5.9-milliliter capped, non-evacuated vial containing ambient air. Sample concentrations were later adjusted for the dilution by ambient air. Gas samples were collected four times for each measurement (0, 20, 40, and 60 min) from 13 May to 8 Jul 2010 and three times (0, 30, and 60 min) from 16 Jul to 14 Oct 2010 and 28 May to 25 Oct 2011. Gas samples were collected over a 2- to 3-hour period, typically between 0900 h and 1200 h, to approximate the mean daily temperature.

Gas fluxes were calculated from the rate of change in concentration over the sampling period using linear regression and were adjusted for theoretical flux underestimation from deployment of the chamber as described by [34]. Measurements for PP treatments began two days after manure application and continued approximately weekly (more frequently after manure or rain and less frequently late in the season) into October. Measurement on SD treatments began 20 days before manure application in 2010 and 5 days before application in 2011. Analysis of gas samples was performed by gas chromatography using an infrared gas analyzer (IRGA, LiCor 820, Lincoln, NE, USA) for CO₂, an electron capture detector (micro-ECD) for N₂O, and a flame ionization detector (FID) for CH₄ (Agilent 7890A GC System, Santa Clara, CA, USA). Annual cumulative gas emission was estimated by linear interpolation between sampling times.

Volumetric soil moisture (5 cm depth; Delta-T Devices Theta Probe) and soil temperature (5 cm depth; digital soil thermometer) were measured in all plots during each gas sampling period. Soil bulk density was measured (two 48 mm diam. × 0.1 m deep cores per plot) 3 to 6 times per year at the beginning of each sampling year and after tillage or other activities that would be expected to affect bulk density. Bulk density was used in calculating theoretical flux underestimation [31] for adjusting N₂O and CO₂ flux values. Monthly precipitation and air temperature were obtained from a standard weather station at the University of Wisconsin Agricultural Research Station about 1 km from the field sites. Weather data during the NH₃ emission measurement period were collected from a portable weather station at each site set up at a lower height than usual (0.3 m from the soil surface for temperature and 0.6 m for wind) to better capture near-soil surface conditions, where NH₃ is volatilized.

Computation and Statistics

Annual cumulative GHG fluxes were calculated using trapezoidal integration of flux versus time, assuming linear changes in daily fluxes. Main effects of manure application by treatment and timing on cumulative NH₃ and GHG emissions were assessed using the general linear modeling procedure (proc glm) in SAS [35]. When necessary, data were logarithmically (log10) transformed to achieve normality/equality of variance. Treatment means were compared using Fisher’s protected LSD. We used p ≤ 0.10 to declare statistical significance due to high background uncertainty for GHG measures. Since there was a significant treatment*year interaction for NH₃ cumulative emission over the study period, data are presented by year.

To more fully evaluate the simultaneous effects of manure application method, weather and soil temperature/moisture on N₂O fluxes, a generalized linear mixed model (proc glimmix) was developed to estimate treatment effects and differences between treatments across a defined range of days since manure application [35]. Average N₂O flux was the dependent variable (expressed as the natural logarithm). Non-detects (n = 43) were randomly assigned values between zero and the minimum observed N₂O-N flux value (0.000048) according to the uniform distribution. The model included random intercepts for year, block, plot, and chamber. The error term in the model was modified to explicitly accommodate correlation of samples within chambers. A spatial power correlation structure was employed with the correlation relating to the time between sample collection dates. Fixed model effects included treatment, days since manure application, treatment * [days since manure application], soil temperature, water content, and [soil temperature] * [water content]. Days since manure application, soil temperature, and water content were represented as natural cubic splines. Least square means and differences between treatments were plotted across the range of days since manure application with corresponding 95% pointwise confidence bands for the latter. Statistical significance was inferred when either the lower confidence limit was >0 or the upper confidence limit was <0. A Bonferroni correction was applied to account multiple statistical tests performed at each value of days since manure application. Days since manure application ranged from 2 to 157; days 112 to 157 were used for PP treatments. Graphics depicting the estimated treatment effects and treatment differences were restricted to days 2 to 111 since there were minimal differences noted after day 111.

3. Results and Discussion

3.1. Weather

Precipitation for the May to October growing season was close to the 30-year mean of 607 mm in 2009 and 2011, although a few individual months departed substantially from the long-term mean (

Table 2. Monthly mean precipitation and temperatures. April 2009 precipitation and temperature were both unavailable due to a logger malfunction.

	Precipitation				Temperature
Month	2009	2010	2011	30-Year Mean	2009	2010	2011	30-Year Mean
		mm				°C
April	-	26	75	76	-	10.1	5.2	7.2
May	99	90	81	102	13.7	14.6	12.8	13.7
June	93	172	105	118	18.2	18.6	18.3	18.9
July	63	281	207	104	18.2	22.2	23.1	21.2
August	185	112	61	109	18.9	21.8	20.9	20.1
September	9	228	92	99	16.8	13.8	14.1	15.4
October	157	61	59	75	5.4	10.4	10.1	8.7
May–October	605	944	605	607	15.2	16.9	16.5	16.3

). Precipitation in 2010 was 50% greater than the 30-year mean and more than twice the 30-year means for July and September. Mean temperatures for the May to October period were similar to 30-year means for 2010 and 2011, while 2009 was slightly cooler.

3.2. Ammonia Emission

Ammonia emissions from surface broadcast manure applied either PP or SD followed a similar pattern in all three years, with the greatest emission occurring immediately after application, ranging from 5 to 12 kg ha⁻¹ h⁻¹, and decreasing dramatically after the first few hours (Figure 2). This resulted in over 75% of the total loss occurring in the first 8 h in 2009 and 2011, and somewhat less occurred in 2010. This pattern of NH₃ loss emphasizes the importance of prompt incorporation to reduce losses and conserve N for crop use. While there were significant treatment differences at each sampling time, NH₃ fluxes at days 2 and 3 were very low (<1 kg NH₃-N ha⁻¹), similar to other studies for both annual crops and grassland [4,5,6,8,9,26,27,32,36].

Cumulative NH₃ emission was greatly reduced by injection at either PP or SD, with typically >90% less NH₃ lost than with broadcast application (Figure 2). Most of the reduction occurred in the first several hours after manure application, but emissions from the two injection treatments were consistently the lowest throughout the measurement period, resulting in values close to zero in most cases (Figure 2). Other researchers have reported reductions in NH₃ emission by injection compared to surface application of about two-thirds [5,8,24,36,37,38], >90% [4,9,32], or close to 99% [26,29]. In our study, the cumulative emission from immediate disk incorporation was intermediate between surface broadcast and injection, with reductions of between 40 and 60% compared to the 87% decrease for PP-3-day in the first two years (not statistically different from PP-Inject in 2011). Ammonia loss reductions of 50 to 80% by chisel plow incorporation were measured in a Pennsylvania study [5]. Disk harrowing reduced NH₃ loss by an average of 75% in the Netherlands [39], while disking or chisel plowing decreased NH₃ loss by 84% in a Maryland study [4].

Total NH₃ emission from surface applications ranged from 16 to 24% of the total N and 44 to 63% of the NH₄-N in applied manure. These conversion values for NH₄-N are similar to the 40 to 60% reported by several other studies [5,31,40] and slightly lower than the 55 to 70% reported by Dell et al. [5] for dairy and swine slurry. Huijsmans et al. [36] reported that 68% of NH₄-N was converted to NH₃-N when averaged over many experiments for swine slurry. For liquid dairy manure, Powell et al. [7] found a wide range of NH₄-N to NH₃-N conversion values for liquid dairy manure over four seasons, with two years nearing 50% while the other two years averaged just 26% and 13%, attributed to a combination of lower slurry dry matter content and lack of rainfall during the two seasons.

We found no consistent differences in cumulative NH₃ emissions between PP and SD treatments for 2010 or 2011 despite the quite different weather conditions (

Table 3. Average air temperature, wind speed, and total precipitation during ammonia (NH₃) sampling periods.

		Average Temperature		Average Wind Speed		Total Precipitation
Date	Time	3 Days	First Day	3 Days	First Day	3 Days	First Day
		°C		m s−1		mm
15-May-09	Pre-Plant	10.5	16.0	3.00	3.58	3.6	3.05
23-June-09	Sidedress	26.0	30.7	0.39	0.56	10.9	0.25
10-May-10	Pre-Plant	7.20	12.2	3.97	4.06	7.9	0.25
30-June-10	Sidedress	21.4	22.2	0.03	0.31	0.0	0.00
26-May-11	Pre-Plant	12.4	13.2	1.53	1.00	8.1	0.00
6-July-11	Sidedress	23.1	25.6	0.06	0.18	0.3	0.00

). Temperatures were much higher at the SD time in late June/July than at the PP time in May, as would be expected; however, wind speeds were much lower at the SD time, probably a function of the crop. The surface application treatments were not significantly different for the two application times. Perhaps the greater wind speed at the PP time and the higher temperatures at the SD time (both of which would tend to increase NH₃ volatilization) compensated for the lower wind velocity. Ammonia emissions from the injected manure treatments were low in all cases, with no significant differences in 2010 or 2011. The substantial differences in 2009 were perhaps related to the higher NH₄-N content of the sidedressed manure that year (Table 1). The total N and NH₄-N contents of PP and SD manure were similar in 2010 and 2011.

3.3. Nitrous Oxide Emission

Nitrous oxide flux was relatively low for most manure treatments during much of the May to October period in both years (Figure 3). However, there were pronounced N₂O peaks after injection for PP (2010) and SD (2011). The time from manure application to peak fluxes was approximately 15 days for PP-Inject (2010) and 6 days for SD-Inject (2011), which is likely related to the lower soil temperatures around the PP time (10 °C) compared to the SD time (26 °C) and associated rates of microbial activity. Times between manure application and peak N₂O fluxes for PP surface applications were shorter (typically 7 to 10 days). Despite considerable differences between site-specific conditions, our results are not that dissimilar to those of Flessa and Beese [14] and Sistani et al. [16], who reported peak N₂O fluxes at 13 to 18 days after application for injection and 3 to 6 days after application for surface application versus the somewhat broader N₂O emission peaks reported by Rodhe et al. [15] at 4 to 12 days after manure injection.

Changes in precipitation and soil temperature/moisture provide some explanation for N₂O patterns. Manure was applied on 10 May 2010 with low soil temperature (10 °C); however, the temperature increased markedly over the next 2 weeks to 22 °C on 25 May, the same date as the peak N₂O emission for PP-Inject (Figure 3). The decreasing soil moisture in the period following application may have limited the N₂O fluxes along with lower soil temperatures, contributing to some of the smaller fluxes after PP (Figure 3). Previous research has demonstrated the important roles of weather, soil temperature, and soil moisture changes in N₂O fluxes associated with manure application [15,25,27,40]. SD manure application on 6 July 2011 was followed by 30 mm of rain over 10–11 July and another 113 mm from 16 to 19 July, increasing the soil water content from 22 to 37%. The soil temperature fluctuated between 21 and 27 °C during the same time and, together with elevated soil water, may have facilitated the increased denitrification potential and pronounced N₂O peak on 12 July. Mean N₂O increases after application were much lower in 2010, with only a slight N₂O flux increase for SD-Inject. In fact, N₂O fluxes increased for some treatments before manure application, presumably because of favorable denitrification conditions (25 °C temperature and 34 to 39% water content).

There were no significant differences in annual cumulative N₂O emission in 2010; however, PP-Inject (3.5 kg ha⁻¹) had a mean value 45% greater than the mean of the four other treatments (2.4 kg ha⁻¹) (data not shown). The mean N₂O emission from SD-Inject (4.2 kg ha⁻¹) was significantly (two to four times) greater than that of all other treatments (0.9 to 1.6 kg ha⁻¹) in 2011. Additionally, injected treatments had 60% greater N₂O fluxes than the other treatments when averaged over the two years of the study (

Table 4. Cumulative and seasonal cumulative nitrous oxide (N₂O) emissions.

	Average	2010		2011		Average
	Cumulative	Post-Plant	Pre-Sidedress	Post-Sidedress	Post-Sidedress	Summer/Fall
			kg N ha−1
PP-Inject	2.94	1.20 a †	0.65	0.22	0.32 b	0.39
PP-1-hr	1.85	0.69 b	0.62	0.13	0.41 b	0.38
PP-3-day	1.78	0.42 b	0.67	0.17	0.22 b	0.45
SD-Inject	3.15	.	.	0.31	3.7 a	0.96
SD-Surf	1.91	.	.	0.45	0.7 b	0.69
p-value	NS	0.01	NS	NS	0.01	NS
PP	2.19	.	.	0.17 b	0.32 b	0.40
SD	2.53	.	.	0.38 a	2.20 a	0.83
p-value	NS	.	.	0.04	0.003	NS
Treatment × Year	NS	NS	NS	<0.0001		NS

^† Means with a different lowercase letter differ at p ≤ 0.10.

) and in a range of other trials with full growing season measurements using liquid swine manure that found cumulative N₂O fluxes of 0.3 to 0.8 [15], 1.1 [25], and 4.6 to 7.1 kg ha⁻¹ [11].

Cumulative growing season N₂O fluxes for most treatments ranged from 0.8 to 1.9% of the total N applied in manure, nearly five times greater than the IPCC’s default factor of 0.4% for the high end of the flux range. Mean fluxes for SD-Inject were much greater in 2011, representing 4% of the applied N. Other research with a range of application methods for liquid swine manure or cattle slurry reported a range of 0.2 to 5% of total N applied lost as N₂O [12,14,16,25,26]. Chantigny et al. [12] reported that 3.1 to 5% of total N applied with liquid swine manure on a clay soil was lost as N₂O, noting that the IPCC default values may also considerably underestimate N₂O emissions on fine textured soils. On the other hand, Velthof and Mosquera [41] reported an average emission factor of 0.9% over a range of N sources (pig slurry, cattle slurry, and fertilizer N), application methods (injection and surface), crop types (grassland and corn), and soils (sandy and clay), with greater rates for injection vs. surface application; injected pig slurry for corn averaged 3.6% vs. 0.9% for surface-applied pig slurry and 0.4% for surface-applied cattle slurry. These emission factors are broadly similar to the ranges found in our study.

We further examined cumulative N₂O emissions in four segments of the season (Table 4). Despite the lack of significant treatment effects for the full sampling seasons, there were some significant differences in the post-plant period. For PP application, PP-Inject was greater than PP-3-day, with PP-1-hr being intermediate. Due to a significant treatment by year interaction for the post-SD time period, comparisons were made individually by year. The mean N₂O emission after 2010 SD application was relatively low, with numerically larger emissions from SD-Surf. There were significant treatment differences in 2011 for the post-SD period, with much larger fluxes for SD-Inject. While there were no significant differences in the late summer–fall period, there was a trend of greater emission from SD, especially SD-Inject.

The greater N₂O emission from manure injection compared to surface application has been attributed to conducive denitrification conditions in the injection zone (abundant inorganic N and labile C) coupled with anoxic conditions [14,42]. The lack of consistent effects with injection may be related to the codependence of N₂O emission on soil moisture content and/or injection depth [14]. The lower N₂O flux from slurry injected deeper into the soil has been attributed to a longer diffusion path to the soil surface, increasing the probability that denitrification will proceed to N₂ [43] or that rainfall would not reach the injection zone to increase soil moisture [21]. In our study, all PP plots were chisel plowed 3 to 5 days after manure application to prepare a consistent seedbed for corn, which may have limited some injection zone effects on N₂O. A recent study in Quebec, Canada, showed that splitting N fertilizer application reduced N₂O fluxes in corn by >50% compared to applying N all at once earlier in the season when corn plants had less demand for N [23].

The linear mixed model developed for continuous N₂O flux estimate curves (ex-pressed as ln (N₂O-N flux)) as a function of the number of days since manure application showed clear patterns by treatment (Figure 4a) and for select differences between treatments (Figure 4b). Contrasts based on the linear mixed model showed similar results, with no significant differences over the year, beginning with manure application between PP-Inject and versus PP-Disk, and SD-Inject versus SD-Surf, or PP-3-day and SD-Surf.

The model predicted greater N₂O fluxes for PP-Inject+PP-1-hr compared to PP-3-day, in addition to greater fluxes for PP-1-hr over PP-3-day in the first 30 days after incorporation. Additionally, greater N₂O fluxes were predicted for SD-Inject compared to PP-Inject within 20 days of application. Interaction terms for time (treatment * [days since manure application] and soil water/temperature [soil temperature] * [soil water content]) were both significant (p < 0.0001). Ranges of statistical significance for days since manure application (i.e., days where the 95% confidence band does not overlap zero) were identified for four of the six treatment comparisons examined: PP-1-hr vs. PP-3-day, days 3–5 (positive); PP-3-day vs. SD-Surf, days 2–7 (negative) and days 12–35 (positive); PP-Incorp (Inject+1-hr) vs. PP-3-day, days 3–5 (positive); PP-Inject vs. SD-Inject, days 12–32 (positive). Positive values indicate that estimates for the first treatment listed in the above comparison were greater than the corresponding second treatment value.

3.4. Carbon Dioxide and Methane Emissions

Changes in CO₂ fluxes showed strong seasonal trends such as temperature changes, with maximum fluxes occurring near mid-July to mid-August (Figure 5) and lower fluxes during early spring and fall (Figure 5). The consistent decrease in CO₂ emissions during mid- to late July (2010 and 2011) paralleled the soil temperature declines (Figure 3), and soil temperature and CO₂ fluxes were significantly correlated (Pearson r = 0.43; p < 0.0001) over the two cropping seasons. Overall, the manure treatments had minor effects on CO₂ fluxes, with some limited seasonal differences. SD-Surf had the greatest emission in the 2010 post-SD period (p = 0.04), and SD-Inject had the highest flux in the late summer/fall of 2011 (p = 0.10; data not shown). These CO₂ increases from SD treatments may be related to the recent addition of labile C from SD manure at a time when soil temperatures were relatively high. It is unclear what factors led to the greater CO₂ fluxes for SD-Surf (2010) and SD-Inject (2011).

Flessa and Beese [14] measured their highest CO₂ flux immediately after dairy slurry application, an effect attributed to high labile C availability in the slurry, and reported no difference between injected vs. surface application. Sistani et al. [16] also reported no application effect for swine effluent on cumulative CO₂ emission. In contrast to our results, they observed elevated CO₂ flux initially and then a decline as soil moisture decreased during a period when soil temperature was increasing (emissions increased again later in the season following a period of rain). Their study was conducted on a moderately well-drained Kentucky soil with temperatures of 15 to >20 °C in the first week and mostly ≥25 °C for the rest of the season, with low soil moisture status (12 to 15% by mid- to late June). Our study was conducted on a somewhat poorly drained loess soil where soil moisture was not limiting (volumetric soil moisture ranged between 30 and 40% for most of the season, with occasional periods of 20 to 25%), likely contributing to the different results for the two trials. Using 11 soil- and climate-related measures over 542 days of study (2012 to 2015), Abasi et al. [22] used six machine learning algorithms to predict CO₂ fluxes in a maize–soy system in southern Quebec, Canada. They reported that the Random Forest algorithm was the most efficient and accurate for predicting CO₂ fluxes and that soil temperature and moisture were the most sensitive input parameters.

In a review of isotope methods for tracking ecosystem carbon and CO₂ fluxes, Pang et al. [19] suggested six fates for carbon additions based on results from 13 CO₂ labeling method studies: (i) lost as CO₂ to the atmosphere, (ii) transient storage in roots, (iii) carbon release via rhizodeposition, (iv) stored in microbial biomass/necromass, (v) lost via microbial/faunal respiration, or (vi) stored as soil organic matter. Changes in labile organic carbon influenced by injection/disk incorporation likely influenced below-ground biomass and root respiration, contributing to some of the CO₂ variability in our study. While labile soil carbon or dissolved organic carbon forms were not measured in our study, we suspect labile carbon measures and other suitable soil biomarkers (i.e., amino sugars, sugars, phospholipid fatty acids, and DNA/RNA) at different soil depths could help better track potential CO₂ differences for manure–tillage combination studies.

Average CH₄ fluxes were low and sometimes negative, indicating net CH₄ consumption (Figure 6). However, there was a sharp increase in CH₄ flux from both PP and SD injection in the first sampling 2 to 3 days after manure application in both years (second sampling 5 days after sampling in PP period in 2011).

The exception to this pattern was the sharp spike in CH₄ from the PP-3-day treatment in the first sampling of 2010. Net annual CH₄ emissions ranged from 180 to −109 g ha⁻¹ over the two years. However, there were no significant differences among treatments on an annual or seasonally segmented basis (data not shown). Other researchers have reported similar results, with transient CH₄ flux increases immediately after manure application and larger emissions from injection [14,16,26,27]. Net emissions were low for the remainder of the measuring period, resulting in no overall treatment effects.

4. Conclusions

The results from our study indicate that injection was the most effective means for NH₃ conservation of the methods investigated, with more modest reductions from tillage incorporation. Manure application method effects on CO₂ and CH₄ emissions were minimal. While effects on N₂O flux varied across the years and application times, larger N₂O fluxes were associated with injection. While this suggests a trade-off between the beneficial effects of injection for the control of NH₃ emission and the negative effects on greenhouse gas due to increased N₂O flux, it is important to include the secondary effects of volatilized NH₃ on N₂O emission. The IPCC [3] estimates that 0.4% of N applied to land will be emitted as N₂O; thus, injection may result in a net greenhouse gas benefit compared to surface application. Moreover, some researchers have measured substantially larger N₂O emissions than those in our experiment [12], concluding that the IPCC default values may underestimate N₂O emission potential. Another point to consider is that higher manure application rates are needed for broadcast application in order to meet crop N demands compared to injected manure because of large NH₃ losses from broadcast, potentially exacerbating secondary N₂O emissions.