Agricultural soils: A sink or source of methane across the British Isles?

This study summarizes a large diverse dataset of methane (CH4) fluxes measured from agricultural sites across the British Isles. A total of 53,976 manual static chamber measurements from 27 different sites were investigated to determine the magnitude of CH4 fluxes from a variety of agricultural fields across the UK and Ireland. Our study shows that contrary to some studies, agricultural soils (both arable and grassland) are small net emitters of CH4 rather than sinks. Mean fluxes measured from arable and grassland sites (excluding fertiliser and tillage events) were 0.11 ± 0.06 and 0.19 ± 0.09 nmol m−2 s−1, respectively, and were not found to be significantly different (Welch t‐test, p = 0.17). Using the values reported in this study, we estimate that an annual emission of 0.16 and 0.09 Mt of CO2‐eq is expected from arable and grassland agricultural soils in the UK and Ireland (comparable to 0.3 and 0.7% of the current annual CH4 emission inventories, respectively). Where CH4 uptake occurs in soils, it is negligible compared to expected emissions of the application of animal manures and tillage events, which were both found to significantly increase CH4 emissions in the immediate few days to months after events. Our study highlights that there are significant differences in CH4 uptake and emissions between sites, and that these differences are partially the result of the moisture content of the soil (i.e., the aerobic status of the soil). We expect uptake of CH4 to be more prevalent in drier soils where volumetric water content does not exceed 35% and emissions to be exponentially greater where agricultural fields become waterlogged.


| INTRODUCTION
Due to anthropogenic activities, concentrations of the powerful greenhouse gas (GHG) methane (CH 4 ) in the atmosphere have risen by over a factor of 2.5, from 772 (pre-industrial era) to 1,866 ppb (Dlugokencky, 2020) presently. Atmospheric CH 4 continues to rise at a rate of approximately 22 Tg every year, as global sources are larger than sinks (also known as uptake, or negative flux) (Ehhalt et al., 2001). The majority of naturally occurring emissions of CH 4 can be attributed to the biogenic processes of methanogens (methane-emitting microorganisms), which are predominant in anaerobic areas rich in organic carbon such as wetlands (Conrad, 2009;Segers, 1998). However, human activities have altered the magnitude of many natural processes over the past two centuries, as well as creating new sources of significant magnitude, which have grown exponentially as the human population has increased. Current estimates show that anthropogenic sources of CH 4 now contribute more to emissions than natural sources, and are responsible for approximately 60% of global emission inventories (Karakurt et al., 2012;Saunois et al., 2016). The largest sources of anthropogenic CH 4 emissions include energy production (i.e., fossil fuel extraction and gas flaring), rice agriculture, biomass burning and ruminant livestock.
Reduction of atmospheric concentrations of CH 4 is required if we are to meet the Paris Agreement target of keeping global warming below 1.5 C (Nisbet et al., 2020). Due to the relatively short lifetime of CH 4 in the atmosphere (approximately 9 years), reductions in CH 4 could have a significant impact on short-term trends in global warming (Dlugokencky et al., 2011;Collins et al., 2018). The primary removal (sink) of CH 4 from the atmosphere occurs through the abiotic reaction of CH 4 with hydroxyl radicals in the troposphere (Ehhalt, 1974). Globally, this process removes approximately 500 to 600 Tg of CH 4 from the atmosphere annually (Saunois et al., 2016). Another smaller and less well-quantified CH 4 sink is that of terrestrial soils, which contributes to the removal of approximately 30 to 60 Tg of CH 4 annually (Cicerone and Oremland, 1988;Ridgwell et al., 1999). Uptake of CH 4 into soils occurs primarily in aerobic conditions, when CH 4 is oxidized by methanotrophic microorganisms (Angle et al., 2017;Holmes et al., 1999).
Methanogenic bacteria produce CH 4 as a product of anaerobic metabolism during the degradation of organic materials in the soil (Stams, 1994). Conditions best suited to these bacteria are anaerobic soils with a rich supply of organic carbon, such as natural wetlands in the arctic and tropics, and rice paddies (Ehhalt et al., 2001). As methanogenic and methanotrophic (CH 4 consuming) microorganisms are both present in soils, both CH 4 production and oxidation occur simultaneously. Emissions of biogenic CH 4 from soils would be significantly larger if not for the methanotrophic microorganisms that oxidize CH 4 before it escapes the soil profile, reducing CH 4 released by up to 99% of the primary production (Reeburgh, 2007). The heterogeneous nature of soils allows for both anaerobic and aerobic microsites to exist in close proximity, thus the net flux from a particular location is dependent upon a variety of conditions within the soil profile, such as aeration, temperature, pH and available nutrients (Savage et al., 1997;Smith et al., 2000).
Methane currently accounts for 11% of all GHG emissions in the UK (BEIS, 2018) and 23% of all GHG emissions in Ireland (EPA, 2020) (CO 2 eq, CO 2 equivalent). Of the estimated 51.5 Mt CO 2 eq of CH 4 emitted across the UK and 14.0 Mt CO 2 eq of CH 4 emitted across Ireland, 49 and 93% are attributed to agricultural activities, respectively. These emissions are predominantly the result of ruminant livestock production (CH 4 released via enteric fermentation) and manure storage (BEIS, 2018;EPA, 2020). Currently, there is global pressure to reduce GHG emissions in supply chains to reduce the impacts of global warming, and as a result, policymakers are seeking opportunities to reduce CH 4 emissions from intensive agricultural activities. One topic of contention in GHG accounting at the farm level is whether intensively managed arable and grassland soils should be included as an offset of CH 4 emissions due to the oxidation uptake that is expected to occur (Chianese et al., 2009;Oertel et al., 2016;Jackson et al., 2020). Although wet soils high in organic matter are considered a net source of CH 4 emissions in the UK (Levy et al., 2012), mineral soils commonly used for intensive agriculture are often recognized as a sink (Dutaur and Verchot, 2007;Castaldi et al., 2007). However, the effects of intensive tillage/mechanical agitation of soils, irrigation, compaction (machinery and livestock trampling) and chemical applications (mineral/organic fertiliser, lime, pesticides, etc.) and resultant disturbed soil conditions across managed agricultural land on CH 4 emissions are currently poorly understood. Due to the small magnitude of CH 4 emissions and uptake typically observed from mineral soils, it can be difficult to identify drivers and quantify fluxes from different soil types and management practices (Kim et al., 2010;Segers, 1998). A large number of studies have investigated the net exchange of CH 4 for agricultural soils (e.g., Dengel et al., 2011;Le Mer and Roger, 2001;Levy et al., 2012;Smith et al., 2000) with varying conclusions and some suggesting that uptake of CH 4 by agricultural soils is substantially less than in unmanaged natural soils (Nesbit and Breitenbeck, 1992;Ridgwell et al., 1999).
Agriculture currently accounts for over 70% of land use in the UK, 65% of land use in Ireland and approximately half of the land use in EU countries. This study aims to better understand the reality of CH 4 production and uptake from agricultural soils in temperate northern hemisphere soils such as those found across the British Isles. Here we analyse a large dataset of CH 4 measurements carried out over the past 15 years by the authors. The dataset includes 53,976 manual static chamber measurements from across 27 separate fields, covering a variety of common crop types (arable, plantation and grassland) and agricultural practices such as tillage and fertiliser application. With this dataset, we aim to investigate CH 4 fluxes from intensively managed agricultural soils and establish whether these soils should be classed as a net sink or source of CH 4 for the purposes of GHG accounting across the British Isles.

| Flux data
Data used in this study were collected from numerous research groups within the UK and Ireland that specialize in the measurements of GHG from agricultural sources. The majority of the data was originally generated during various research projects with the aim of quantifying emissions of gases from soils, but primarily with a focus on nitrous oxide (N 2 O). The method is the same for measuring fluxes of both CH 4 and N 2 O and both gases are typically measured in tandem; thus a largely unused and unreported dataset was generated for CH 4 fluxes.
A total of 53,976 manual static chamber measurements from 27 different sites across the British Isles were collated (Table 1). These measurements were all carried out using the principle of flux chamber methodology, whereby an enclosed chamber (of 20 to 40 L volume) was inserted into the soil to form an air-tight seal. During closure of the chamber, gases were allowed to accumulate within the sealed volume, and gas samples were manually extracted from the chamber via a syringe and tap. Samples were analysed on different gas chromatography (GC) instruments based across the UK and Ireland, fitted with flame-ionization detector (FID) and electron capture detector (ECD) devices, which allow concentrations of carbon dioxide (CO 2 ), CH 4 and N 2 O to be quantified at the same time. The rate of exchange of a particular gas between the soil and the atmosphere was then determined using Equation (1): where F is the gas flux from the soil (nmol m −2 s −1 ), dC/ dt is the rate of change in the concentration in time in nmol mol −1 s −1 estimated by linear regression, ρ is the density of air in mol m −3 , V is the volume of the chamber in m 3 and A is the surface area enclosed by the chamber in m 2 . Although based on a similar principle, the flux chamber measurement methodology used for the different datasets varied between research groups (see Table 1 publications list for further details of methodology). Chamber size and sampling procedures differed between studies, but fluxes remain consistent as a representation of emissions from a given surface area of the soil over a given period of time (as described in Equation (1)). More efforts have been focused on improving the detection limit of the measurement methodology in recent years by increasing the number of samples taken per chamber from a minimum of one to a standard procedure which requires at least four. A larger number of samples reduces the uncertainty in the regression fit of concentration change over a set period of time (dC/dt), which is recognized as the largest source of error in chamber measurements . The least precise method is that of taking T A B L E 1 A summary of the origins and measurement methodology of the CH 4 flux data presented in this study (n is the number of measurements recorded at each site) several samples of background air as an 'ambient' to act as a t = 0 sample for all chambers, and taking one sample per chamber after a fixed period of time (typically 40-60 min). The uncertainty in a two-point regression fitting cannot be calculated on an individual basis, limiting quality control of data collected using this method. Here, we determine the range of uncertainty for CH 4 measurements to be in the order of ±0.52 nmol m −2 s −1 , although in reality this value will vary depending upon instrument capability and maintenance, user experience and execution of the chamber measurements themselves (i.e., leaking, disturbance of soil, timing of measurement). Where four (or more) samples are taken, this reduces the uncertainty in dC/dt significantly; thus the detection limit is estimated to be 0.17 nmol m −2 s −1 for this study, based on unpublished laboratory work carried out by the authors investigating quality control of flux chamber measurements. However, the same caveats exist in that this value is a generalization and in reality, detection limits of a particular method are dependent on many factors that will vary across research groups and experimental methodology. In this study, these approximate detection limits (0.17 to 0.52 nmol m −2 s −1 ) were used to determine the significance of the individually observed fluxes.

| Site classification
The field sites from which the collated data originate are classified in this study as arable, bioenergy (BE) plantations or grasslands (see Table 1). Fields in which cereal crops were grown and ground that was left unsown for a season (fallow) are classed as arable. These fields had a history of intensive tillage and were typically tilled once or twice per year depending on crop rotation. BE plantations represent fields in which crops (trees) were planted on soils that had until recently been classified as arable. As such, these fields did not have soils similar to those associated with traditional commercial forests. The crops planted in the BE plantations were short rotation coppice willow, short rotation forestry scots pine and Miscanthus perennial grass, all commonly used for bioenergy purposes. Where grassland was grazed or grown as a silage , the data are classed as grassland. These sites were less frequently tilled (usually more than 5 years between tillage events), and some grassland sites in upland or boggy locations were only managed extensively (e.g., frequent livestock grazing).

| Site data
Soil data were collected for each of the different field sites used in this study where available. At the site level, soil pH, bulk density, annual rainfall and temperature data were available for most sites (Table 2). Data were also T A B L E 2 Site data and soil properties of field sites from which CH 4 flux data were collected in this study. Annual rainfall and soil temperature are based on averages of 10 years of measurement data (where available) collated where measurements of soil properties had been taken in close proximity to flux chambers during or shortly after flux measurements. The three variables most often recorded during gas measurements were volumetric soil moisture, soil pH and soil temperature. The volumetric soil moisture was the most commonly collected parameter across the experiments for the purpose of relating CH 4 fluxes to soil moisture contents (water-filled pore space was not commonly available).
In situ measurements were taken using hand-held moisture probes, which were inserted to 5-10-cm depth of the soil (typically using an ML3 ThetaProbe Soil Moisture Sensor, Delta-T Devices, Cambridge, UK, or Campbell Hydrosense II, Campbell Scientific, Inc., Logan, UT, USA). A variety of hand-held temperature probes were used at the different sites to measure soil temperature in situ. Soil temperature probes would be left for approximately 2 to 10 min in the soil to reach equilibrium temperature before taking a reading. Soil pH was measured by taking a soil sample (0-10 cm deep) from the field for laboratory analysis (see Table 1 publications list for further details of the methodology). Where pH measurements were made in water rather than calcium chloride (CaCl 2 ), a conversion factor was applied to normalize the methods. Due to the unavailability of electrical conductivity (EC) measurements, the linear conversion of Ahern et al. (1995) was used, which is representative of soils ranging from 4.8 to 8.5 (Equation (2)):

| Statistical analysis
Data in this study were analysed using the statistical software 'R' (R Core Team, 2017) and presented using the 'ggplot2' package (Wickham, 2016). Where 95% confidence intervals (95% CIs) are presented, these were estimated by taking the standard error of a sample population and multiplying by 1.96, assuming a ''normal'' statistical distribution of data. The ''leaps'' package for R was used to perform stepwise regression to find the best-fitting model, based on the Akaike information criterion (AIC) (Lumley, 2015). The AIC is a measure of model goodness-of-fit derived from information theory, widely used in model selection (Burnham and Anderson, 2004). For a set of candidate models, the model with the lowest AIC value represents the best choice, given the trade-off between model likelihood and complexity.

| Flux data
A total of 53,976 manual flux measurements collected using standard static chamber methods are reported in this analysis (Figure 1a), 93% of which are from grassland soils (Figure 1d). The maximum CH 4 emission recorded from an individual measurement was 303.7 nmol m −2 s −1 , whereas the largest negative flux (uptake) was −19.4 nmol m −2 s −1 . However, observations of this scale are rare in this dataset and the vast majority of measurements were near zero, with a 95% quantile range of −1.44 to 2.29 nmol m −2 s −1 (Figure 1). The mean of all fluxes measured from the grassland sites was 0.62 ± 0.21 nmol m −2 s −1 , whereas the means of all fluxes measured from the arable and BE plantation sites were 0.11 ± 0.06 and − 0.05 ± 0.02 nmol m −2 s −1 , respectively. The large difference between grassland emissions and other management types is partly due to the inclusion of high flux data from one grassland site after a tillage event and contributions from sites after organic fertiliser applications (discussed below), and with their exclusion, the mean flux measured from grassland sites is 0.19 ± 0.09 nmol m −2 s −1 . With the exclusion of fluxes measured after the tillage event and application of organic fertilisers, differences between fluxes measured from grassland and arable sites were not found to be statistically significant (Welch t-test, p = 0.17); however, emissions measured from the plantation fields were found to be significantly different to both grassland and arable fields (p < 0.001 for both).
The mean value of all CH 4 fluxes measured is 0.58 nmol m −2 s −1 ; however, the majority of observations were negative, with 62% of all fluxes reporting uptake of CH 4 into the soil. The positive mean values are the result of a skewed dataset, with a log-normal distribution tending toward higher fluxes (Figure 1a). If the measurements after manure application (up to 30 days) and the impacts of tillage and subsequent large flux observations  are removed from the dataset, then the mean value of the reported CH 4 flux falls to 0.16 ± 0.08 nmol m −2 s −1 (heavily dominated by grassland sites). Splitting the data by field site reveals that there are significant differences between different fields under the same management type. Specific trends also appear, with uptake most likely to be observed in BE plantation fields and uptake more likely at arable than grassland sites (Table 3). Statistically significant uptake of CH 4 was observed at three of the four plantation sites, three of the eight arable sites and three of the 19 grassland sites. Assuming a global warming potential (GWP) of 28 over 100 years (IPCC 2014), we present the estimated annual flux in units of g-CO 2 eq m −2 . Although CH 4 emission or uptake was shown to be statistically significantly different from zero in some cases, estimated annual fluxes of CH 4 remained relatively small in terms of GWP, with the largest annual emission of 31.9 g CO 2 eq m −2 (Kirkton) and largest annual uptake of −3.49 g CO 2 eq m −2 (House OʼMuir) (Table 3).

| Impact of soil conditions
Direct linear correlations between volumetric soil moisture content, pH and soil temperature measured in tandem with flux measurements (excluding tillage and fertiliser application experiments) were each found to individually correlate with CH 4 flux in a statistically significant manner (p < 0.005); however, R 2 values when using individual or multiple linear regression remained small (<0.1). As the majority of measurements fell below the detection limit of the chamber method, comparisons were highly scattered and the meaningful use of more complex statistical comparisons was limited. Trends in the data were clear for both soil moisture content and temperature ( Figure 2). Where soils were very dry, uptake was more likely, with emissions typically increasing as soil moisture increases. Fluxes appeared to rise sharply as volumetric soil moisture reached 85% (Figure 2a). There also appeared to be a trend of uptake increasing as soils got warmer, although soil moisture T A B L E 3 A summary of mean fluxes and 95% confidence intervals (CIs) reported for each of the field management types at the different field sites. Annual global warming potential (GWP) estimates are presented as a multiplication of the mean flux, assuming a GWP of 28 over 100 years. Where uptake is larger than the 95% CI the site is designated as 'uptake significant'

Site
Crop type Mean CH 4 flux (nmol m −2 s −1 ) 95% CIs (±) Annual GWP (g CO 2 eq m −2 ) Uptake significant and soil temperature were strongly correlated (R 2 = 0.3). When using multiple linear regression, with all soil properties as variables, soil moisture retained statistical significance, whereas soil temperature and pH both became insignificant as predictors (p > 0.1). This suggests that the pattern observed between CH 4 flux and soil temperature was almost entirely due to the change in soil moisture content, and not a direct reaction to the temperature change.
At the site level, all measured variables were averaged for comparisons (excluding tillage and fertiliser application experiments). Using multiple linear regression, and rated by AIC value, we investigated the model that best explained the variability in CH 4 fluxes (Figure 3a). Univariate linear regression between CH 4 flux and annual rainfall accounted for 31% of the variance between the field sites; however, this could be slightly improved upon by including the other known soil properties, which then accounted for 38% of the variance between sites (Figure 3b). Site annual rainfall was the only statistically significant predictor of CH 4 flux at the site scale when using multiple linear regression and no clear patterns emerged between other site variables and CH 4 emissions when comparing differences between the sites.

| Impacts of tillage
Although tillage would have occurred annually at most of the arable sites, and less frequently (every 5 years) F I G U R E 2 Fluxes of methane (CH 4 ) grouped into statistical bins based on (a) volumetric soil moisture, (b) soil pH and (c) soil temperature, measured in tandem with chamber measurements at each chamber location. Plots show the median values for each data bin, with hinges on the 25% and 75% quantiles [Color figure can be viewed at wileyonlinelibrary.com] on intensively managed grasslands (or not at all on upland hill fields), such data were not available for most of these studies. Chambers are often removed from sites during any tillage operations, and as most of the experiments aimed to study the impact of nitrogen fertiliser application, tillage was deliberately avoided. The exception to this was one study that specifically aimed to investigate the impacts of tillage . This study reveals that after tillage events, relatively large fluxes of CH 4 can be released from an otherwise low CH 4 system (Figure 4). The results from this study showed that during these large emission events, the data shifts from a normal distribution of flux data close to zero, to a log-normal distribution, but retains many values close to zero and negative measurements were still observed even when exponentially high fluxes occur in the same field (Figures 4 and  5). Although rainfall may have played some part in the increase in emissions after tillage, CH 4 fluxes are exponentially higher from the tilled field than from the adjacent untilled field in this study (see .

| Impacts of fertiliser application
Mineral fertiliser application appears to have minimal impacts on CH 4 emissions. Over the 30 days immediately after fertiliser application in all of the studies, only a small number of high flux points were recorded with no obvious temporal pattern ( Figure 6). Differences between the fertiliser types were largely experiment/site specific and there was no clear pattern as to whether a particular mineral fertiliser type had any impact on CH 4 (large emissions are generally from grazed fields where animal dung is also present, as is the case for ammonium nitrate (AN) measurements). Application of some organic fertilisers did appear to have a larger impact on CH 4 emissions (Figure 7). In the experiments where dung was applied, a clear disruption of CH 4 emissions was observed for the first week after application. After this, CH 4 emissions dropped back down to background levels rapidly, with a handful of measurements still reporting elevated fluxes. In the experiment carried out by Sánchez-Rodríguez et al. (2018), industrially digested food waste (digestate) was applied in both its acidified and normal state. These peaks were not observed for the acidified digestates for which fluxes were generally negative in the 30 days after application. The application of cow urine appeared to have no effect on emissions, although a small number of high uptake measurements were recorded from these plots (Jones et al., 2005;Maire et al., 2020;Krol et al., 2016).
In the absence of a meaningful model able to gap-fill CH 4 between measurements, any cumulative estimates were open to wide interpretation. Integration using linear interpolation between points is highly sensitive to the number of measurements taken, the presence of outliers and the timing between samples. In this study, therefore, we made the simplistic assumption that the mean flux over 30 days after an application event is proportional to the magnitude of the cumulative flux expected over this period (Table 4). Assuming a GWP of 28 over 100 years (IPCC 2014), we present the estimated fluxes in units of g-CO 2 eq m −2 . Although the impact of individual (or grouped) events was shown to be significantly different from zero in some cases, fluxes of CH 4 during these periods were low in terms of GWP. The exception to this F I G U R E 5 Measurements of CH 4 fluxes taken before (a) and after (b) the tillage event reported in  show that data shift from a normal to log-normal distribution after disturbance was dung application, with an estimated emission of up to 21 g CO 2 eq m −2 in the 30 days after application.

| DISCUSSION
This study highlights numerous discussion points regarding the emission and uptake of CH 4 from agricultural soils. The first point to address is that our data show that the vast majority of CH 4 fluxes measured from all agricultural soils are close to zero and fall below the limit of detection of the flux chamber methodologies used to quantify fluxes. For this reason, these CH 4 fluxes are often disregarded by researchers and not reported in studies that use flux chamber methodologies as they are considered negligible or insignificantly different from zero. In many cases where the number of measurements for each particular ''treatment'' is small in short-term experiments, it is unlikely that statistically significant results will be possible. The issue of detection limits and statistical significance of ''noisy'' data does explain in part the contradictory studies from previous decades of CH 4 research where the number of gas samples taken per chamber were low (<4) and GC instruments were less precise than modern equivalents. Care should be taken when assessing small CH 4 flux datasets from soils, F I G U R E 6 Measurements of CH 4 fluxes taken up to 30 days after mineral and organic fertiliser application events (excluding treated digestate and dung) at all sites. Plots show the median values for each data bin, with hinges on the 25% and 75% quantiles. Plots are curtailed visually but not statistically to the −5 to 5 nmol m −2 s −1 range to exclude outliers. AN, ammonium nitrate; CAN, calcium ammonium nitrate especially when detection limits are poor. Ways to improve CH 4 measurements in future studies would be to use modern laser instruments with higher precision than GC instruments, capable of measuring high frequency measurements (1 Hz) from chambers (Hensen et al., 2006).
This study shows that with sufficient data, statistically significant differences can be observed when investigating CH 4 from different agricultural soils during experiments. The data gathered using static chamber methodologies allow for differentiation between fluxes measured from different sites and management practices. Mean fluxes from the 27 different sites reported in this study range from −0.33 ± 0.04 to 3.01 ± 0.76 nmol m −2 s −1 , showing that although small in magnitude, uptake and emissions of CH 4 are both viable conclusions to reach regarding any particular field site. Our data also suggest that bioenergy crop plantations are significantly more likely to be a small sink of CH 4 , with a mean flux of −0.05 ± 0.02 nmol m −2 s −1 calculated from all plantation site flux measurements. This is perhaps due to no-tillage conditions and limited disturbance by management operations, and agrees well with previous studies that suggested that forest and natural woodland soils are more likely to act as CH 4 sinks than soils under arable and grassland management (Nesbit and Breitenbeck, 1992;Smith et al., 2000). Plantation soils are also likely to be less compacted (i.e., lower bulk densities) than agricultural soils (Nesbit and Breitenbeck, 1992;Smith et al., 2000). Mean fluxes measured from arable and grassland sites (excluding all fertiliser and tillage events) were 0.11 ± 0.06 and 0.19 ± 0.09 nmol m −2 s −1 , respectively, and were not found to be significantly different (Welch t test, p = 0.17).
The strongest predictor of individual flux measurements in the available data was soil volumetric water content (VWC). Although noisy (R 2 < 0.1), partly due to the limit of detection of the chamber measurements, the number of measurement points was sufficient to identify a statistically significant relationship between CH 4 flux and VWC. This relationship predicts that in drier soils (<20% VWC), CH 4 uptake is more likely, whereas in very wet (waterlogged, >80% VWC) soils, CH 4 flux will F I G U R E 7 Measurements of CH 4 fluxes taken up to 30 days after (top) waste food digestate and (bottom) dung fertiliser application at all sites. Plots show the median values for each data bin, with hinges on the 25% and 75% quantiles. Plots are curtailed visually but not statistically to exclude outliers increase substantially. These observations agree to some extent with work carried out by van den Pol-van Dasselaar et al., (1998), which identified soil VWC of approximately 20 to 35% as the optimal range of CH 4 uptake in grasslands in the Netherlands. Comparisons between CH 4 flux and VWC reported in this study were somewhat limited by the depth at which measurements were taken, representing only the top 0-10 cm of the soil profiles. Emissions of CH 4 from agricultural soils were also likely to be influenced by CH 4 produced deeper in the soil profiles, dependent upon water table depth, soil aerobicity, soil compaction and available carbon. Where soils are dry and porous on the surface, but very wet and warm with a large carbon content deeper in the profile (anoxic and organic carbon rich), higher emissions of CH 4 might also be expected (Morel et al., 2019).
Fluxes of CH 4 at the site level correlated somewhat significantly with annual rainfall (R 2 = 0.31); however, adding further soil properties to the model only slightly improved the prediction. The correlation between sitelevel CH 4 fluxes and annual rainfall may indicate that CH 4 emissions from a particular area are more dependent on long-term conditions than short-term changes at a site. If this is the case, wetter soils may favour populations of methanogenic bacteria. At the individual measurement and site level, we find poor correlation between fluxes and soil properties that are believed to influence CH 4 emissions (pH, soil temperature, bulk density). Segers (1998) highlighted the difficulty in predicting CH 4 fluxes from soil properties, but other studies have identified relationships between pH and temperature (e.g., MacDonald et al., 1996;Smith et al., 2000). However, the strongest correlations between soil properties and CH 4 fluxes are often found in forest and wetland soils, with emissions from agricultural soils remaining close to zero in most cases (MacDonald et al., 1996;Smith et al., 2000).
The largest impact on CH 4 fluxes of any of the management practices studied here was that of tillage. In , adjacent intensively managed grazed grasslands were tilled in spring, resulting in a large increase in CH 4 emissions (Figures 4 and 5). This was likely to be a result of the incorporation and subsequent decomposition of grass into the lower soil layers, providing the necessary anaerobic conditions and carbon source for methanogenesis (Badagliacca et al., 2020). Taking simplistic means of the fluxes before tillage and the mean of fluxes after tillage (up to 150 days), fluxes increased from −0.03 ± 0.07 nmol m −2 s −1 to 4.03 ± 1.29 nmol m −2 s −1 . Based on these values, it can be estimated that tillage resulted in the net loss of approximately 0.63 g CH 4 -C m −2 during a 150-day measurement period. Assuming a GWP multiplier of 28, this means that emissions from the field over a 150-day T A B L E 4 The impact of mineral and organic N fertiliser types on CH 4 fluxes. Data are mean and cumulative fluxes of CH 4 estimated for a 30-day period, post fertiliser application, presented as nmol m −2 s −1 and g CO 2 eq m −2 (over 30 days) using a global warming potential multiplier of 28. In the absence of a model able to predict CH 4 between measurements, cumulative flux estimates are based on the mean of fluxes recorded from 0 to 30 days after fertiliser application period resulted in a net loss of 17.7 g CO 2 eq m −2 . To put that value into context, it would take approximately 55.4 years for this field to absorb the same mass of CH 4 as was released after this tillage event, assuming uptake occurred at pre-tillage rates (−0.03 nmol m −2 s −1 ) for the duration. Large emissions of CH 4 have also been observed from other grassland sites, including Merbold et al. (2014), where CH 4 emissions of 2.65 g CH 4 -C m 2 were reported as the annual flux during the year of a tillage event, measured using the eddy covariance method. The change in statistical distribution from normal to lognormal highlights that, potentially, the statistics used to describe this type of activity should take log-normal maths into account. Methods such as a Bayesian approach should be used to determine appropriate means and uncertainties when handling this type of data in future studies (Cowan et al., 2017). The immediate impact of nitrogen fertilisers (over 30 days) was typically below statistical significance for each of the treatments and dependent on the site at which experiments took place. Individual data points where high fluxes were measured skewed the data. These fluxes were real, but their representation of a particular nitrogen fertiliser application is questionable. In the case of fertiliser application to grazed fields (most of the experiments reported in this study), occasionally animal waste is present in the chambers (although not recorded), which can affect CH 4 flux. Interestingly, all mineral fertiliser applications resulted in uptake, with the exception of AN. As calcium ammonium nitrate (CAN) is expected to slightly increase the pH of soils when applied, an observation of uptake after application is counterintuitive to previous studies, which suggest increasing pH can increase CH 4 emissions from soils (Wang et al., 1993). Our observations mirror those of other studies in which the application of mineral fertilisers is not always well explained by the known soil and microbial chemistry (Hütsch, 1998;Le Mer and Roger, 2001). This is likely to be due to a mixture of the impacts of unmeasured variables and the limit of detection of the chamber methodologies applied. This study cannot determine the long-term impacts of mineral fertilisation on the different field sites, but this factor may affect the extent to which soils can act as a CH 4 sink (Hütsch et al., 1994).
The application of organic fertilisers has a more immediate impact than the application of mineral fertilisers. In the case of dung and digestate additions, an immediate peak and decline in CH 4 emissions were observed, which lasted for approximately 1 to 2 weeks in the data presented. These emissions remained relatively small with a maximum flux of 21 g CO 2 eq m −2 over a 30-day period after application. This study cannot state a definitive emission factor for CH 4 expected after organic fertiliser application, as this would take many experiments and a functional gap-filling model. The complex properties of livestock dung and urine depend on factors such as animal diet, age and health, which are beyond the scope of this study. What we emphasize in this study is a realistic magnitude of emissions expected after organic manure application. These are documented well in previous studies where similar flux ranges were observed, with emission peaks which typically lasted for 2 to 4 weeks after application (Chadwick et al., 2000;Saggar et al., 2004).
To put the impact of CH 4 emissions and uptake from agricultural soils into context with the other GHGs, we present the data from the Easter Bush grassland (Midlothian, Scotland), where a variety of GHG measurements have occurred (carried out by the authors of this study, see Table 5). CO 2 eq emissions are calculated using values provided from the IPCC 2014, Fifth Assessment Report (AR5). In Jones et al. (2017), the CO 2 emissions measured at the site as net ecosystem exchange ranged from −605 to 72 g C m −2 , with an annual mean uptake of approximately 217.9 g C m −2 . Typically, the Easter Bush field receives approximately 210 kg N per year in the form of AN fertiliser. In Cowan et al. (2020a), emission factors of N 2 O measured via the eddy covariance method varied per fertiliser event, ranging from 0.2 to 2.78% of the nitrogen applied as AN, with a mean flux of approximately 0.3 nmol N 2 O m −2 s −1 during periods outside of fertiliser events. Using this estimate as a background flux, an annual estimate of 0.27 g N 2 O-N m −2 is released from the field when no fertiliser is applied. Using the more refined EF from , which estimates an EF of 1.34% of the applied N to be emitted as N 2 O, the typical annual estimate of N 2 O released from the field is approximately 2.75 g N 2 O-N m −2 . In , annual soil carbon monoxide (CO) emissions of 0.37 g C m −2 at the site were reported. Carbon monoxide is classed as a secondary GHG due to its interactions with OH radicals in the atmosphere, thus extending the lifetime of CH 4 and having a positive GWP as a result. Excluding the study in which the tillage event occurred, the mean CH 4 flux for the Easter Bush field site was 0.1 nmol m −2 s −1 , which equated to 0.04 g C m −2 y −1 , and is close to the mean of the CH 4 measured from all sites presented in this study (i.e., 0.19 ± 0.09 nmol m −2 s −1 for undisturbed grassland).
Using these estimates, the relative contribution of CH 4 to emissions from the soil is negligible when compared to uncertainties that would be expected in CO 2 and N 2 O emission estimates. The contributions to the overall CO 2 eq emissions from the site account for only 0.4% of the total budget, only fractionally more than those estimated for CO, which contributes an estimated 0.3% of the total budget. At sites where uptake is expected, a similar comparison could be expected, in which uptake of CH 4 would be statistically insignificant when compared to the uptake and emission of other GHGs. Interestingly, emissions of CO in terms of mass of carbon are higher at the Easter Bush site than CH 4 , although CH 4 still contributes more to the total GWP of the site due to the higher GWP of CH 4 molecules. If CH 4 uptake was considered in GHG accounting for agricultural soils, the argument also exists that it would be prudent to include emissions of CO, which is likely to reduce the overall effect of any uptake to even less significance. This budget does not include the effect of tillage, organic fertiliser application or livestock respiration or excreta returns. If these were included, the contribution to the annual GWP of CH 4 is likely to increase (BEIS, 2018;EPA, 2020).
Previous studies have suggested that soils across Europe act as a methane sink, including grassland and upland soils similar to those reported in this study (e.g., Kravchenko 2017;Hörtnagl et al., 2018). It is uncertain from the data presented in this study if there are geographical reasons for soils becoming a CH 4 source across the British Isles (such as high soil moisture or organic matter content), or if comparisons across studies are too sparse or methodologically different to draw a valid comparison. In particular, given the highly skewed nature of the data, the mean is a very uncertain quantity, and the uncertainty bounds on this are rarely quantified correctly, and this affects whether authors conclude that their sites are net sinks or sources. According to estimated global inventories, soils are expected to act as a sink for an estimated 30 to 60 Tg of CH 4 annually (Cicerone and Oremland, 1988;Ridgwell et al., 1999). On average, soil CH 4 uptake at a global level is estimated to be in the order of −0.40 to −0.79 nmol m −2 s −1 , with an annual uptake of approximately 4.2 to 8.4 g CO 2 eq m −2 . The UK and Ireland account for 244,654 and 70,273 km 2 of land cover, respectively. If generic global soil uptake values were applied to UK and Irish agricultural soils, an annual CH 4 uptake of 0.94 to 1.88 Mt of CO 2 eq would be expected to occur across both countries. In the UK, 56,506 km 2 of land is classed as arable and 96,949 km 2 is classed as grasslands (improved, neutral, calcareous and acid) (Land Cover statistics derived from LCM2015; Rowland et al., 2017). In Ireland, 6,676 km 2 of land is classed as arable and 42,867 km 2 of land is classed as grasslands (CSO, 2018). Using the CH 4 fluxes estimated for these land classes reported in this study, these areas become a source, and an emission of 0.16 and 0.09 Mt of CO 2 eq is expected to be emitted from the UK and Ireland, respectively. Compared to the 51.5 and 14.0 Mt CO 2 eq of CH 4 emitted across the UK and Ireland from other sources, such as energy production and distribution, landfills, ruminants, manure management and wetlands (BEIS, 2018;EPA, 2020), emissions of CH 4 from UK and Irish agricultural soils are insignificant, representing only 0.3% and 0.7%, respectively, of the total budget.

| CONCLUSION
Our study concludes that release and uptake of CH 4 by agricultural soils are both possible and can be statistically significant at any given field site. The data presented also suggest that fluxes of CH 4 from plantation fields are more likely to be negative than those of arable or grassland sites. However, overall, both arable and grassland fields were on average small emitters of CH 4 due to the presence of hotspots (small areas [valid individual measurements] of high fluxes) that skewed data from uptake into a positive emission. Findings suggest that uptake of CH 4 should not be considered for GHG inventories of specific farms, or at a national level. Alternatively, we predict CH 4 emissions from intensively managed agricultural soils to be approximately 0.16 Mt of CO 2 eq for UK land and 0.09 Mt of CO 2 eq for Irish land. Furthermore, we predict agricultural management practices, such as the application of organic fertilisers and the tillage of fields, are likely to counteract most CH 4 consumption by methanotrophic bacteria in agricultural soils. In comparison to other GHGs, uptake or T A B L E 5 The annual greenhouse gas (GHG) budget for the intensively managed grazed grassland for Easter Bush (Midlothian, Scotland) is presented, using global warming potential (GWP) values provided from the IPCC 2014, Fifth Assessment Report (AR5). Annual estimates are made based on the years that measurements took place in Jones et al. (2017) for CO 2 ,  for CO and Cowan et al. (2020a) for N 2 O GHG Annual emission (g C m −2 )/(g N m −2 ) Typical emission (g C m −2 )/(g N m −2 ) AR5 GWP ratio Annual flux (g CO 2 eq m −2 ) Annual GWP (%) release of CH 4 from soils alone can be expected to be in the same order of magnitude in terms of GWP as CO emissions (secondary GHG), further reducing the significance of GHG mitigation by CH 4 uptake in soils if the CO factor is also considered. Our overall recommendation is that where no long-term measurements from a particular (or representative) field site are available, soil CH 4 emissions can be assumed to be negligible in any localized GHG budget within the British Isles. This is due to a lack of evidence to classify soils from a particular area as being either a significant CH 4 source or sink without a dedicated measurement campaign. As such, uptake of CH 4 in agricultural soils should not be used as a sink to offset emissions from livestock in the UK and Ireland.