Net soil – atmosphere fluxes mask patterns in gross production and consumption of nitrous oxide and methane in a managed ecosystem

Nitrous oxide (N2O) and methane (CH4) are potent greenhouse gases that are both produced and consumed in soil. Production and consumption of these gases are driven by different processes, making it difficult to infer their controls when measuring only net fluxes. We used the trace gas pool dilution technique to simultaneously measure gross fluxes of N2O and CH4 throughout the growing season in a cornfield in northern California, USA. Net N2O fluxes ranged 0–4.5 mg N m d with the N2O yield averaging 0.68± 0.02. Gross N2O production was best predicted by net nitrogen (N) mineralization, soil moisture, and soil temperature (R = 0.60, n= 39, p< 0.001). Gross N2O reduction was correlated with the combination of gross N2O production rates, net N mineralization rates, and CO2 emissions (R = 0.74, n= 39, p< 0.001). Overall, net CH4 fluxes averaged−0.03± 0.02 mg C m d. The methanogenic fraction of carbon mineralization ranged from 0 to 0.27 % and explained 40 % of the variability in gross CH4 production rates (n= 37, p< 0.001). Gross CH4 oxidation exhibited a strong positive relationship with gross CH4 production rates (R = 0.67, n= 37, p< 0.001), which reached as high as 5.4 mg C m d. Our study is the first to demonstrate the simultaneous in situ measurement of gross N2O and CH4 fluxes, and results highlight that net soil–atmosphere fluxes can mask significant gross production and consumption of these trace gases.


Introduction
Greenhouse gas emissions from soils are major contributors to climate change (Ciais et al., 2013).While carbon dioxide (CO 2 ) is the most abundant greenhouse gas in the atmosphere, both nitrous oxide (N 2 O) and methane (CH 4 ) are more potent with 298 and 34 times the global warming potential of CO 2 on a 100-year timescale, respectively (Myhre et al., 2013).Both N 2 O and CH 4 are produced and consumed in soils by microbially mediated redox-sensitive processes.However, most studies only measure net soil-atmosphere exchange of N 2 O and CH 4 .This approach cannot differentiate between production and consumption of these trace gases and thus limits our ability to infer controls on these processes and to diagnose model inaccuracies in predicting net N 2 O and CH 4 fluxes.This hinders predictions of how soil-atmosphere N 2 O and CH 4 fluxes will respond to future changes in land use practices or climate change.
Agricultural soils account for nearly two-thirds of global soil emissions of N 2 O, which is produced from nitrification and denitrification of fertilizer nitrogen (N) that supports agroecosystem productivity (Ciais et al., 2013).Managing soil N 2 O emissions from agroecosystems can go beyond direct reductions in N 2 O production from decreased fertilizer inputs because denitrifying bacteria can consume N 2 O to dinitrogen gas (N 2 ), completing the N cycling.Nitrous oxide consumption is not generally considered to be an important process in upland soils because it is an anaerobic process.Rates of N 2 O reduction to N 2 decrease as O 2 and NO − 3 availability increases (Weier et al., 1993;Firestone et W. H. Yang and W. L. Silver: Net soil-atmosphere fluxes mask patterns in gross fluxes al., 1980).Theoretically, this results in a high N 2 O yield (N 2 O / (N 2 O + N 2 )) in unsaturated soil where diffusive resupply of O 2 and the production of NO − 3 from nitrification would inhibit N 2 O reduction.Thermodynamics also predict that high soil NO − 3 from fertilizer N inputs in agricultural soils would lead to high N 2 O yields.However, N 2 O yields average 0.375 ± 0.035 in agricultural soil and span the entire range from 0 to 1 in oxic, upland soils (Schlesinger, 2009;Stevens and Laughlin, 1998).This high variability, in part, reflects the difficulty in measuring rates of N 2 O reduction to N 2 , particularly under field conditions (Groffman et al., 2006).It also reflects other important factors that influence the N 2 O yield, such as soil type (Woli et al., 2010), labile C (Weier et al., 1993), and pH (Stevens et al., 1998).The lowerthan-expected average N 2 O yield in agricultural soils and large range in N 2 O yields in upland soil in general also suggests that N 2 O reduction to N 2 could play an important role in mitigating soil N 2 O emissions to the atmosphere in agroecosystems.
Upland soils globally consume atmospheric CH 4 at a rate similar to the accumulation of CH 4 in the atmosphere (Ciais et al., 2013), and thus changes in the CH 4 sink strength of soils could influence atmospheric CH 4 concentrations.The inhibition of CH 4 oxidation associated with fertilizer application of NO − 3 (Aronson and Helliker, 2010), urea (Mosier et al., 1991), and NH + 4 (Bedard and Knowles, 1989) is thought to cause lower net rates of CH 4 uptake in agricultural systems compared to natural ecosystems (Nesbit and Breitenbeck, 1992;Bender and Conrad, 1994;Koschorreck and Conrad, 1993;Dutaur and Verchot, 2007;Mosier et al., 1991).Inhibition by NH + 4 has been attributed to enzymatic substrate competition due to the similarities between the CH 4 monooxygenase and NH + 4 monooxygenase enzymes (Gulledge and Schimel, 1998) as well as toxicity effects from nitrite produced during NH + 4 oxidation (King and Schnell, 1994).However, the effect of N on CH 4 oxidation varies by soil (Gulledge et al., 1997), and at least some of this effect is due to inhibition by salts included in the fertilizer applications (Adamsen and King, 1993;Dunfield et al., 1993;Gulledge and Schimel, 1998;Nesbit and Breitenbeck, 1992).In addition, the response of CH 4 oxidation to NH + 4 and NO − 3 may depend on the methanotrophic community; for example the high affinity type II methane-oxidizing bacteria that dominate under low (< 1000 ppm) CH 4 conditions (Bender and Conrad, 1992) may be less sensitive to mineral N availability (Jang et al., 2011;Reay and Nedwell, 2004;Wang and Ineson, 2003).Thus, there remains uncertainty surrounding N inhibition of CH 4 oxidation as the mechanism leading to low net rates of CH 4 uptake in agricultural soils.
A major confounding factor in studies assessing controls on CH 4 oxidation is the simultaneous occurrence of methanogenesis and CH 4 oxidation.Net changes in CH 4 concentrations under oxic soil conditions are assumed to reflect only CH 4 oxidation (e.g., Nesbit and Breitenbeck, 1992) because methanogenesis occurs only under highly reducing conditions (Conrad, 1996). However, von Fischer andHedin (2002) demonstrated that CH 4 production occurred in a wide range of dry, oxic soils with water-filled pore space as low as 20 %.Similarly, Teh et al. (2005) documented the occurrence of methanogenesis under well-aerated conditions in an upland tropical forest soil.Macroaggregates can support net CH 4 efflux in unsaturated soil (Jackel et al., 2001;Sey et al., 2008), likely because O 2 consumption in the centers of the aggregates exceeds diffusive re-supply of O 2 to create reducing conditions (Sexstone et al., 1985).Microsites of methanogenesis could also occur in the rhizosphere where high rates of O 2 consumption from rhizosphere priming could create reducing conditions (Cheng et al., 2003).Because the controls on methanogenesis and CH 4 oxidation are likely very different, the co-occurrence of these processes means that we must measure gross rates of both processes simultaneously to elucidate the mechanisms driving patterns in net soil-atmosphere CH 4 fluxes.
We used the stable isotope trace gas pool dilution technique to measure gross N 2 O and CH 4 fluxes in cornfield soils throughout the growing season in order to improve our understanding of trace gas dynamics in upland soils of agroecosystems.Fertilized agroecosystems are typically large net N 2 O sources and small net CH 4 sinks (Haile-Mariam et al., 2008;Kessavalou et al., 1998;Gelfand et al., 2013;Nangia et al., 2013;Robertson et al., 2000).However, little is known about the rates of gross production and consumption of these gases in upland soils, or their controlling factors.Different controls on production and consumption processes may result in complex responses of net soil-atmosphere gas fluxes to climate or land management.Thus, the objectives of this study were to quantify field rates of gross N 2 O and CH 4 production and consumption, and explore environmental and plant-mediated controls on these rates.

Study site
The study site was a cornfield planted on a drained peatland located on Twitchell Island (38.11 • N, 121.65 • W) in the Sacramento-San Joaquin River delta region of northern California.The region is very productive agriculturally, producing USD 500 million in crops in 1993 (Ingebritsen and Ikehara, 1999).The climate is Mediterranean with a winter wet season and summer dry season.The mean annual temperature is 15.1 • C, and mean annual precipitation is 335 mm (Hatala et al., 2012).The soils consist of mucky clay over buried peat and are classified as fine, mixed, superactive, thermic Cumulic Endoaquolls (Drexler et al., 2009).The field was fertilized once, at seeding, at a rate of 118 kg N ha −1 with UAN 32, which consists of 45 % ammonium nitrate, 35 % urea, and 20 % water.The water table was maintained around 50 cm soil depth throughout the growing season via subsurface irrigation.

Study design
We measured gross and net fluxes of CO 2 , CH 4 , and N 2 O at five time points during the growing season from May to November 2012 on the following days after seeding (DAS): 11 (germination stage), 24 (seedling stage), 59 (peak growth stage), 94 (flowering stage), and 171 (senesced stage).The corn began senescing around DAS 104 and was harvested on DAS 178.We performed measurements in row and inter-row locations with the assumption that plant effects, if any, would be greater in the rows where the corn was growing (Cai et al., 2012;Haile-Mariam et al., 2008;Kessavalou et al., 1998).We established three parallel transects spaced 50 m apart.We measured gross production and consumption of CH 4 and N 2 O as well as net fluxes of CO 2 , CH 4 , and N 2 O along the northernmost transect and measured only net fluxes in the other two transects.In each transect, we used paired measurements in the bed (in between corn rows) and furrow (in row) with replicate pairs spaced 10 m apart (n = 4 pairs per transect).After each gas flux measurement was completed, we measured air, chamber headspace, and soil temperature at the surface flux chamber location.We also used an auger to sample the soil from the chamber footprint in 10 cm increments to 50 cm depth for the gross flux transect and only 0-10 cm depth in the net flux transects.The soils were processed the next day for determination of gravimetric soil moisture and net rates of nitrogen (N) mineralization and nitrification as described below.

Laboratory assays
We determined net rates of N mineralization and nitrification from 6-day laboratory incubations.We mixed each soil core by hand and subsampled 15 g for extraction in 75 mL of 2 M KCl, 10 g for determination of gravimetric soil moisture, and 50 g for incubation in Mason jars kept in the dark at ambient temperature.The jars were covered in perforated plastic wrap to minimize evaporation during the incubation.After 6 days, the soils in the jars were mixed and 15 g of soil was subsampled for KCl extractions.The KCl extracts were analyzed colorimetrically for NH + 4 and NO − 3 concentrations on a Lachat QuickChem flow injection auto-analyzer (Lachat Instruments, Milwaukee, WI, USA).We calculated net N mineralization rates from the change in NH + 4 plus NO − 3 concentrations over the incubation period and net nitrification rates from the change in NO − 3 concentrations over the incubation period.
The remaining soil (not utilized in the net rates incubation) was air dried for archival.Air-dried samples from the May sampling date were ground in a Spex Mill (Metuchen, NJ, USA) for total C and N analyses on a vario MICRO cube elemental analyzer (Elementar, Hanau, Germany).

Gas flux measurements
We used the stable isotope trace gas pool dilution technique to measure field rates of gross N 2 O and CH 4 production and consumption (von Fischer and Hedin, 2002;Yang et al., 2011).We injected 10 mL of isotopically enriched spiking gas into the headspace of a 28 L surface flux chamber inserted 6 cm into the soil surface.The spiking gas consisted of 70 ppm N 2 O at 98 atom % 15 N enrichment, 280 ppm CH 4 at 99 atom % 13 C enrichment, and 28 ppm SF 6 to achieve a 15 N-N 2 O enrichment of 5.42 atom % and 13 C-CH 4 enrichment of 5.61 atom %.This spiking gas injection increased the chamber headspace gas composition by 25 N 2 O, 100 CH 4 , and 10 ppb SF 6 .We sampled the chamber headspace at 5, 15, 30, 45, and 60 min after spiking gas injection.We analyzed samples on a Shimadzu GC-14A gas chromatograph (Columbia, MD, USA) equipped with a thermal conductivity detector, flame ionization detector, and electron capture detector for determination of CO 2 , CH 4 , N 2 O, and SF 6 concentrations.We analyzed separate samples for 15 N-N 2 O and 13 C-CH 4 on an IsoPrime 100 continuous flow isotope ratio mass spectrometer interfaced with a trace gas preconcentration unit (Isoprime Ltd, Cheadle Hulme, UK) and Gilson GX271 autosampler (Middleton, WI).The trace gas analyzer was equipped with a combustion furnace using palladium to catalyze the conversion of CH 4 to CO 2 for isotopic analysis after CO and CO 2 were scrubbed from the sample (Fisher et al., 2006).One out of the 40 gross N 2 O flux measurements and three out of the 40 gross CH 4 flux measurements were lost due to autosampler needle clogs that occurred during isotopic analysis.
Gross N 2 O and CH 4 production and consumption rates were estimated using the pool dilution model as described by Yang et al. (2011) andvon Fischer andHedin (2002).The iterative model solves for gross production rates based on the isotopic dilution of the isotopically enriched chamber headspace pool of N 2 O or CH 4 by natural abundance N 2 O or CH 4 emitted by the soil.Gross consumption rates were estimated from the empirical loss of the 15 N 2 O or 13 CH 4 tracer, using the loss of the SF 6 tracer to account for physical losses such as diffusion.We note that gross N 2 O consumption rates are not equivalent to N 2 production via denitrification because heterogeneous 15 N 2 O distribution in the soil and complete denitrification intracellularly could lead to underestimates of N 2 production (Well and Butterbach-Bahl, 2013;Yang et al., 2013).We assumed that the isotopic composition of produced N 2 O was 0.3431 atom % 15 N and the fractionation factor associated with N 2 O reduction to N 2 was 0.9924.The justification for these assumptions is discussed by Yang et al. (2011).We assumed that the isotopic composition of produced CH 4 was 1.0473 atom %, based on measurements of the 13 C isotopic composition of soil CH 4 in a nearby study site (Y.Teh, personal communication, 2011).We assumed that the fractionation factor associated with CH 4 oxidation was 0.98 as justified by von Fischer and Hedin (2002).Sen-sitivity analyses performed by both Yang et al. (2011) andvon Fischer andHedin (2002) showed that the pool dilution model output is not sensitive to these assumed values at the high isotopic enrichments used.Net fluxes of CO 2 , N 2 O, and CH 4 were determined from the change in concentration over time using an iterative model that fits an exponential curve to the data (Matthias et al., 1978).Fluxes were considered to be zero when the relationship between trace gas concentration and time was not significant at p = 0.05.The methanogenic fraction of C mineralization was calculated as the gross CH 4 production rate divided by the sum of the gross CH 4 production rate and CO 2 production rates.

Statistical analyses
We used SYSTAT version 13 (SPSS Inc., Evanston, IL, USA) to perform statistical analyses and Microsoft Excel 2007 (Microsoft Corporation, Redmond, WA, USA) to run the iterative pool dilution model.We log transformed the data to meet the normality assumptions of ANOVAs; soil moisture, soil temperature, soil C and N concentrations, and soil C : N ratios did not require transformation.We analyzed net and gross fluxes of CO 2 , N 2 O, and CH 4 using sampling date as the within-subjects factor and location (i.e., bed versus furrow) as the between-subject factor in repeated measures ANOVAs.We also analyzed net N mineralization and nitrification rates using sampling date as the within-subjects factor, and soil depth and location as the between-subjects factors in repeated measures ANOVAs.We explored relationships between trace gas fluxes and potential drivers (soil moisture, air and soil temperatures, soil NH + 4 and NO − 3 concentrations, net N mineralization and nitrification rates, soil C and N concentrations, etc.) using linear regressions.We determined the model that best fit observed trace gas flux data using backwards stepwise multiple linear regressions starting with all potential explanatory variables; the best model fit was determined by minimizing the Akaike information criterion.Statistical significance was determined at p values < 0.05.

Soil characteristics and N cycling
Air and soil temperature differed significantly among sampling dates (p< 0.05, Table 1).Mean air temperature spanned a small range from a low of 24.5 ± 0.7 • C on DAS 171 to a high of 28.2 ± 0.7 • C on DAS 94.Soil temperature was more variable, with the lowest mean soil temperature on DAS 171 at 14.8 ± 0.1 • C and the highest mean soil temperature on DAS 59 at 24.2 ± 0.3 • C.
In surface soils (0-10 cm depth), gravimetric soil moisture ranged from 0.24 ± 0.01 g H 2 O g −1 soil on DAS 94 to 0.38 ± 0.02 g H 2 O g −1 soil on DAS 11 (Table 1).Soil moisture decreased as the growing season progressed until DAS 171, when soil moisture increased to a value intermediate of that on DAS 59 and 94 (Table 1).Soil moisture was significantly higher in the row than in the inter-row on DAS 11 and 24 only (Table 3).Mean soil moisture increased significantly with depth (Table 2), although differences were not statistically significant for all dates (Table 3).
Across the entire data set (n = 216), net N mineralization rates averaged 3.3 ± 0.5 and net nitrification rates averaged 2.7 ± 0.6 µg N g −1 d −1 .Net N mineralization and nitrification rates did not differ significantly among soil depths, sampling locations, or sampling dates (Table 3), although rates trended higher at 0-10 cm depth across all sampling dates and locations (Table 2).Across all sampling dates and soil depths, 96 % of the variability in net nitrification rates was explained by net N mineralization rates (p< 0.001, n = 215, Table 4).
Total C and N concentrations for soils sampled on DAS 11 differed between row and inter-row sampling locations (soil C, F 1,30 = 5.295, p = 0.03; soil N, F 1,30 = 4.546, p = 0.04) but not among soil depths (Table 2).Both soil C and N concentrations were higher in rows than in inter-rows, averaging 16.1 ± 0.8 % C and 0.99 ± 0.03 % N in rows and 13.7 ± 0.5 % C and 0.89 ± 0.02 % N in inter-rows.Soil C : N ratios averaged 15.8 ± 0.2 overall (n = 40), and did not differ significantly between sampling locations or among soil depths.

Gross and net CH 4 fluxes
Net CH 4 fluxes ranged from −1.3 to 0.44 mg C m −2 d −1 but net fluxes were not detectable for 94 out of 112 measurements.Overall net CH 4 fluxes averaged −0.03 ± 0.02 mg C m −2 d −1 .Using the trace gas pool dilution technique, we detected gross CH 4 production in 36 out of 37 measurements.Gross CH 4 production reached as high as 5.4 mg C m −2 d −1 with rates trending higher throughout the growing season (Fig. 2b).However, rates were only significantly different between DAS 11 and 94 (F 4,12 = 4.1, p = 0.03).Gross CH 4 production rates were marginally significantly higher in rows than in inter-rows (F 1,3 = 5.8, p = 0.10).Overall, gross CH 4 production rates were weakly correlated to soil CO 2 emissions (R 2 = 0.17, Table 4) but exhibited a stronger positive correlation with the methanogenic fraction of C mineralization (R 2 = 0.40, n = 37, p< 0.001, Fig. 4a), which ranged from 0 to 0.27 % and averaged 0.06 ± 0.01 %.The strength of the relationship increased to R 2 = 0.60 (n = 23, p< 0.001) when considering only dates when the corn was not actively growing (Fig. 4a).When only peak growth sampling dates were con- sidered (DAS 59 and 94), 57 % of the variability in gross CH 4 production rates was predicted by the combination of CO 2 emissions, net N mineralization, and net nitrification (n = 14, p = 0.03).

CO 2 emissions
Carbon dioxide emissions ranged 0.6-10.5 g C m −2 d −1 across the entire data set.Emissions trended higher in the rows than in the inter-rows after the corn germinated, but repeated measures ANOVA showed that CO 2 emissions differed significantly among sampling dates (F 4,56 = 80.1, p< 0.001) but not between row and inter-row locations (Fig. 1b).The highest CO 2 emissions occurred on DAS 59 and 94, at the height of the growing season, averag- ing 6.7 ± 0.2 g C m −2 d −1 ; the lowest emissions occurred on DAS 11 and 24 at the beginning of the growing season, averaging 2.6 ± 0.2 g C m −2 d −1 .The variability in CO 2 emissions was poorly explained by environmental and soil variables with soil moisture and soil temperature together as the best, yet weak, predictors (R 2 = 0.15, Table 4).

N 2 O dynamics
Net N 2 O fluxes at our study site were comparable to those reported for other fertilized crop fields (Gelfand et al., 2013;Smith et al., 2011;Stevens and Laughlin, 1998;Nangia et al., 2013;Robertson et al., 2000), averaging 1.5 ± 0.2 mg N m −2 d −1 across the growing season.Prior field estimates of N 2 O yield using 15 NH 4 or 15 NO 3 addition at application rates of 200-300 kg N ha −1 span a wide range from 0.06 to 0.7 (Mosier et al., 1986;Rolston et al., 1976Rolston et al., , 1978Rolston et al., , 1982)).In contrast, the N 2 O yield varied little throughout the growing season at our site, averaging 0.68 ± 0.02, despite significant differences in both net and gross N 2 O fluxes among sampling dates.This is similar to a field estimate of the N 2 O yield for a nearby pasture on the same soil type (0.70 ± 0.04; Yang et al., 2011).Soil NO − 3 concentrations in surface soils (0-10 cm depth) were 1-2 orders of magnitude greater in the cornfield than in the pasture, so it is surprising that the N 2 O yields were similar.Soil NO − 3 concentration was the strongest predictor of N 2 O yield in a US Midwest cornfield soil incubated in the laboratory (Woli et al., 2010).
Other factors such as soil pH, labile C availability, or soil aggregation may have played a more important role in controlling the N 2 O yield in our cornfield (Sey et al., 2008).
The best predictors of gross N 2 O production and consumption changed over the growing season, likely reflecting the influence of plant-microbial competition for N on N 2 O dynamics.This is a novel finding because, to our knowledge, this is the first study that has made repeated measurements of gross N 2 O dynamics over the growing season in the presence of active plant-microbial competition for N. When the corn was actively growing, 89 % of the variability in gross N 2 O production was explained by soil moisture, soil temperature, net N mineralization, and CO 2 emissions together.In contrast, when the corn was not actively growing, both gross N 2 O production and reduction were best predicted by soil CO 2 emissions alone.This may reflect the role of CO 2 emissions as proxy for the availability of labile C as an electron donor for denitrification; during the growing season, the contribution of autotrophic respiration to soil CO 2 emissions obscured this role.Net N mineralization was an explanatory variable for gross N 2 O production only during the growing season when plant uptake of N could have limited N 2 O production.
Overall, gross N 2 O reduction rates were strongly correlated to gross N 2 O production rates.This relationship was also observed in a managed grassland with high soil mineral N concentrations and net soil N 2 O emissions (Yang et al., 2011), but not in a salt marsh with low mineral N availability where net N 2 O uptake by soil occurred (Yang and Silver, 2016).The strong relationship between N 2 O production and reduction may have driven the well-constrained N 2 O yields in both this study and the managed grassland study because N 2 O reduction increased proportionally to N 2 O production rates.Additional studies using the trace gas pool dilution technique in the field could elucidate whether or not this relationship holds only in soils with high mineral N concentrations to drive high rates of N 2 O production.

CH 4 dynamics
The small and zero net CH 4 fluxes we observed, which are typical of cornfields (Mosier et al., 2006), masked gross CH 4 fluxes which were 2 orders of magnitude greater.Net CH 4 fluxes were generally undetectable because CH 4 oxidation was tightly coupled to methanogenesis, especially at high gross CH 4 production rates.The ability of methanotrophs to adjust activity to but not exceed, rates of methanogenesis could reflect oxidation of soil-derived CH 4 at high concentrations near methanogenic microsites but not atmospheric CH 4 at low concentrations in the bulk soil.There are a few mechanisms that could drive a stimulatory effect of high CH 4 concentrations on CH 4 oxidation without increasing oxidation rates at atmospheric concentrations (Benstead and King, 1997).First, high microsite CH 4 concentrations can increase the number of methanotrophs as well as shift the methanotrophic community composition from high affinity type II methanotrophs, who consume CH 4 at low concentrations, to low affinity type I methanotrophs, who consume CH 4 only at high concentrations, in or near the methanogenic microsites (Bender andConrad, 1992, 1995).Second, the enzyme affinity of type II methanotrophs can change from high affinity in the presence of atmospheric CH 4 concentrations to low affinity at high CH 4 concentrations, thereby reducing their capability to oxidize CH 4 at low concentrations (Dunfield et al., 1999).Third, high CH 4 availability may be needed to stimulate enzyme synthesis (Bender andConrad, 1992, 1995;Nesbit and Breitenbeck), and thus methanotrophic activity may be induced only near methanogenic microsites and not in the bulk soil.Additional studies investigating gross CH 4 dynamics in soil aggregates or through the soil profile could provide insight into the mechanisms coupling CH 4 production and consumption.Regardless of the mechanisms, our observations suggest that using in situ methods that preserve spatial variability in soil CH 4 concentrations and allow for the occurrence of both CH 4 production and oxidation, such as the trace gas pool dilution technique, is important for accurately characterizing CH 4 dynamics in soil.
Gross CH 4 production rates were strongly positively correlated with the methanogenic fraction of C mineralization, an index of anaerobic soil microsites where electron acceptors are depleted relative to C supply (von Fischer and Hedin, 2007).Von Fischer et al. (2007) found that the methanogenic fraction was constrained below 0.04 % and gross CH 4 production rates below 1 mg C m −2 d −1 in tropical and temperate forest soils with less than 60 % water-filled pore space.Though the slope of the relationship between gross CH 4 production rates and the methanogenic fraction observed here was similar to that reported by von Fischer et al. (2007), the maximum methanogenic fraction observed here was nearly 7 times greater.The maximum gross CH 4 production rate was also an order of magnitude greater than the maximum rate of 0.5 mg C m −2 d −1 reported by von Fischer and Hedin (2002) for a range of unsaturated upland soils in which net CH 4 fluxes were near zero (−0.2 to 0.2 mg C m −2 d −1 ).This suggests a higher potential for the development of methanogenic microsites in these drained peatland soils, which are rich in C.
The near-zero net CH 4 fluxes measured in our cornfield are consistent with other studies in agricultural systems, but the relatively high gross CH 4 oxidation rates we documented challenge the paradigm that agricultural soils have low potential for CH 4 oxidation compared to unsaturated soils in natural ecosystems (Bender and Conrad, 1994;Koschorreck and Conrad, 1993;Mosier et al., 1991;Nesbit and Breitenbeck, 1992;Zhuang et al., 2013).Our soils had high NH + lated soils incubated in the laboratory under conditions to isolate CH 4 oxidation from CH 4 production, and vice versa.Application of the trace gas pool dilution technique to other agricultural fields could reveal whether or not the tight coupling of CH 4 production and consumption rather than low rates of CH 4 production and oxidation could be responsible for the general observation of small and near-zero net CH 4 fluxes in agricultural ecosystems.A greater understanding of limitations on gross CH 4 oxidation under field conditions is needed to accurately predict how land use change will alter soil-atmosphere CH 4 exchange and to better manage agricultural soils to be atmospheric CH 4 sinks.
Our data provide circumstantial evidence that plants could mediate gross CH 4 dynamics in upland soil both directly and indirectly.An increase in plant C inputs to the soil over the growing season may have directly driven a steady, though not statistically significant, increase in rates of methanogenesis by providing more C substrate to support methanogenesis.Both gross CH 4 production and oxidation rates were approximately 2.5 times greater at DAS 171 compared to DAS 11.This trend in gross CH 4 fluxes cannot be explained by changes in environmental variables such as soil temperature, which peaked in the middle of the growing season, and soil moisture, which decreased over the growing season.However, von Fischer and Hedin (2007) showed that methanogenesis was not limited by C supply in a wide range of upland soils, but rather, it was limited by the number of anaerobic microsites that could support methanogenesis in the soils.Our data also support the latter mechanism controlling methanogenesis: we observed a strong relationship between gross CH 4 production and the methanogenic fraction of C mineralization (an index of the abundance of anaerobic soil microsites) on DAS 11, 24, and 171, when root respiration likely did not contribute significantly to CO 2 effluxes.We also observed higher gross CH 4 production and oxidation rates in rows than in inter-rows, suggesting that plants could indirectly control methanogenesis through rhizosphere priming, fueling biological O 2 demand for C mineralization (Zhu et al., 2014) that creates a greater number of anaerobic soil microsites supporting methanogenesis.

Conclusions
Our study demonstrates that the anaerobic processes of N 2 O reduction to N 2 and methanogenesis can play important roles in mediating soil-atmosphere greenhouse gas fluxes in upland crop field soils where these processes have previously been discounted.Moreover, despite high soil NO − 3 and NH + 4 concentrations that theoretically inhibit N 2 O reduction to N 2 as well as CH 4 oxidation, gross N 2 O reduction rates were approximately one-third of gross N 2 O production rates and CH 4 oxidation kept pace with methanogenesis that reached relatively high rates for unsaturated soil.Our field measurements of gross N 2 O and CH 4 fluxes thus challenge our cur-rent understanding of the controls on the production and consumption of N 2 O and CH 4 in upland soils.The strong correlations that gross N 2 O and CH 4 fluxes exhibited with soil characteristics and soil N cycling process rates can help guide controlled studies to investigate the controls on the processes that lead to the production and consumption of N 2 O and CH 4 .A better understanding of the controls on these processes can help refine modeling efforts to characterize the effects of anoxic microsites in unsaturated soil on greenhouse gas emissions (Riley et al., 2011) and also inform land management decisions to mitigate soil greenhouse gas emissions from crop fields.

Figure 1 .
Figure 1.Mean (a) net N 2 O flux and (b) CO 2 efflux for all three transects (n = 24 per sampling date except n = 16 on DAS 94) in inter-rows (black bars) and rows (grey bars).Error bars represent standard errors, and different letters indicate statistically significant differences among sampling dates.

Figure 2 .
Figure 2. Mean (a) gross N 2 O production rates (black bars) and reduction rates (grey bars) and (b) gross CH 4 production rates (black bars) and oxidation rates (grey bars).Error bars represent standard errors (n = 8 per sampling date).

Table 1 .
Environmental and soil (0-10 cm depth) variables by sampling date (mean ± SE).Degrees of freedom are shown in subscripts, and statistically significant F statistics at P < 0.05 are indicated by bold text.Letters indicate statistically significant differences among sampling dates.* One transect was excluded from the repeated measures ANOVA because data are missing for one sampling date.

Table 2 .
Soil characteristics and N cycling rates across all sampling dates by soil depth in the gross flux transect (mean ± SE).
Letters indicate statistically significant differences among soil depths.* Data from DAS 11 only.

Table 3 .
Results from repeated measures ANOVAs with sampling date, the interaction of sampling date and soil depth, and the interaction of sampling date and sampling location as the within-subjects, and soil depth and sampling location as the between-subjects factors.

Table 4 .
Coefficients for multiple linear regressions predicting trace gas fluxes using soil variables.