Real‐time monitoring of greenhouse gas emissions with tall chambers reveals diurnal N2O variation and increased emissions of CO2 and N2O from Miscanthus following compost addition

Miscanthus x giganteus's efficacy as an energy crop relies on maintaining low greenhouse gas (GHG) emissions. As demand for Miscanthus is expected to rise to meet bioenergy targets, fertilizers and composts may be employed to increase yields, but will also increase GHG emissions. Manipulation experiments are vital to investigate the consequences of any fertilizer additions, but there is currently no way to measure whole‐plant GHG fluxes from crops taller than 2.5 m, such as Miscanthus, at the experimental plot scale. We employed a unique combination of eddy covariance (EC), soil chambers and an entirely new automated chamber system, SkyBeam, to measure high frequency (ca. hourly) fluxes of carbon dioxide (CO2), methane (CH4) and nitrous oxide (N2O) from a Miscanthus crop amended with green compost. Untreated controls were also monitored in a fully replicated experimental design. Net ecosystem exchange (NEE) of CO2 was partitioned into soil respiration (Rs), gross primary productivity (GPP) and ecosystem respiration, and the crop was harvested to determine the effect of compost on crop productivity. Compost increased NEE emissions by 100% (p < .05), which was the result of a 20% increase of Rs (p < .06) and a 32% reduction in GPP (p < .05) and biomass of 37% (p < .06). Methane fluxes were small and unaffected by compost addition. N2O emissions increased 34% under compost during an emission event; otherwise, fluxes were low and often negative, even under dry conditions. Diurnal variation in N2O fluxes, with uptake during the day and emission at night was observed. These fluxes displayed a negative relationship with soil temperature and a hitherto undescribed diurnal temperature hysteresis. We conclude that compost addition negatively affected the productivity and environmental effects of Miscanthus cultivation during the first year following application.


| INTRODUCTION
To combat climate change, it may be necessary to use biomass to produce as much as a third of our future energy requirements (IPCC, 2014). The perennial C 4 grass Miscanthus (Miscanthus x giganteus) has been the subject of great interest (Clifton-Brown, Stampfl, & Jones, 2004;van der Weijde et al., 2013), due to its high productivity and relatively low nutrient requirements, particularly its low demand for nitrogen (N) fertilizer (St Clair, Hillier, & Smith, 2008), which it recycles during senescence over winter (Strullu, Cadoux, Preudhomme, Jeuffroy, & Beaudoin, 2011). It is expected that the uptake of Miscanthus cultivation across Europe will increase greatly over the coming years to meet energy needs, particularly if economic barriers are alleviated . One way to improve economic viability is to increase productivity, and various studies have investigated the effect of fertilizer on Miscanthus yield, with mixed results. While some authors have reported no response (see Maughan et al., 2012;Teat, Neufeld, Gehl, & Gonzales, 2015), recently the balance of evidence suggests that fertilizer increases yields (Chen et al., 2019;Wang, Smyth, Crozier, Gehl, & Heitman, 2018;Xu, Gauder, Gruber, & Claupein, 2017). Emissions of the powerful greenhouse gas (GHG) nitrous oxide (N 2 O), derived from N fertilizer, are the biggest source of GHGs from agriculture after livestock methane (CH 4 ;FAO, 2015), and the use of fertilizer and composts can also lead to increased soil respiration (R s ; Garcia-Delgado et al., 2018;Ozlu & Kumar, 2018) and CH 4 emissions (Fernandez-Luqueno et al., 2010;Thangarajan, Bolan, Tian, Naidu, & Kunhikrishnan, 2013), two other important components of a net GHG balance. Previous work has shown that GHG emissions, and especially N 2 O fluxes, from Miscanthus are much lower than conventional crop rotations (Drewer, Finch, Lloyd, Baggs, & Skiba, 2012), and this is key to its viability as a bioenergy crop . Therefore, if it becomes common practice to apply fertilizer to energy crops, it may fundamentally change the GHG balance of energy crop production which is crucial to their purpose. It is long recognized that any potential increase in GHG emissions is factored into the overall GHG balance for cultivation of bioenergy feedstocks, and this demands robust measurements of GHG fluxes during the life cycle of the crop's cultivation.
The eddy covariance (EC) approach measures GHG fluxes at ecosystem scale, accounting for all the sources and sinks within its footprint and thus providing an estimate of the exchange of the gases of interest. Due to this, and the continuous data the technology yields, EC has become increasingly popular for quantifying GHG fluxes of carbon dioxide (CO 2 ) and CH 4 in particular. The EC approach has limitations, however, including topographical restrictions and difficulties measuring during stable atmospheric conditions at night, but perhaps the biggest drawback is its inability to resolve to the plot scale. Plot scale GHG measurements are vital for manipulative experiments, where the response of a system to a treatment may be followed in replication. Whilst wholetree chambers have been built for flux measurements before (Mordacq, Ghashghaie, & Saugier, 1991), here we deployed a novel automated system, SkyBeam, which we believe is the first fully automated mobile chamber capable of measuring from tall (>2.5 m) vegetation. The clear chamber allowed photosynthesis to continue, with short chamber closure time ensuring that the crop was exposed to ambient conditions for as much of the study period as possible. Circulating the headspace gas through multiple analysers allowed the quantification of fluxes of CO 2 , N 2 O and CH 4 from a single chamber closure. This new approach allowed monitoring of N 2 O, CO 2 and CH 4 from Miscanthus in near real-time, to explore the extent to which episodic emissions of GHG occurred.
Using a combination of SkyBeam, automated soil flux chambers and EC, we were able to conduct a fully replicated field experiment to investigate the effect of adding a green compost to Miscanthus x giganteus on: (a) the frequency and magnitude of N 2 O emissions events; (b) net exchange of CO 2 , N 2 O and CH 4 ; (c) R s ; and (d) crop productivity in the first year after application. Furthermore, the combination of techniques applied allowed investigation into the partitioning of carbon fluxes from NEE data to assess which elements of the C cycle would influence changes to the GHG balance following compost addition.

| Site description and compost experiment
All work was undertaken at a commercial farm in the east of the United Kingdom on a soil type of Beccles 1 association, fine silt over clay (described in detail by Drewer et al., 2012). The experiment was undertaken in an 11.5 ha field. The Miscanthus was a mature 7 year old stand which had not received any fertilizer for at least 2 years prior to this experimental work and the biomass was grown to supply nearby Drax power station in North Yorkshire, United Kingdom. Typically, Miscanthus emerges during May-June and grows rapidly until October, after which the biomass is left to senesce over winter and is harvested the following spring. However, following a disappointing harvest in 2012, the field was harrowed (ploughed to ca. 10 cm depth) in spring 2013 in an attempt to redistribute the rhizomes more evenly in order to improve yield, followed in July 2013 by an application of green compost (46.6% C, 1.2% N, C:N = 38) which consisted of a range of decomposed woody material smaller than 5 cm to fine sawdust-like particles.

| Experimental design
The experimental area was selected following harvesting and subsequent harrowing of the Miscanthus crop in spring 2013. Emerging shoots across the field were surveyed using quadrats and, within a representative area, six plots (ca. 1.5 m × 2.5 m) were demarcated with each containing one landing base for the SkyBeam automated flux system and a 20 cm diameter collar for use with a Licor automated flux chamber (LI-8100, Licor; Figure 1). In July 2013, coinciding with the contractor's compost addition to the Miscanthus field, a subsample of the same compost was taken, well mixed and applied to the experimental plots by hand at the equivalent rate (4 T/ha). Plots were paired and one of each pair chosen at random to receive compost (+COMP), or to be maintained as a control (−COMP).

| SkyBeam automated chamber system
A full description of the SkyBeam (University of York) system can be found in Keane (2015). Briefly, it is an automated chamber system capable of measuring GHG fluxes from vegetation greater than 3 m tall ( Figure 1). A single chamber is suspended from a 12 V powered trolley mounted on 10 m long aluminium gantry (height 6 m; Figure 1). Suspending the chamber allowed repeated measurements to be taken from preselected points along a transect directly underneath the beam. Each measurement consisted of lowering the chamber onto a landing base at one of the designated positions, before raising the chamber and moving to the next position. Automation of the system was achieved using a Licor infrared gas analyser (IRGA; LI-8100, Licor) to control the opening and closing and the length of closure of the large SkyBeam chamber in place of Licor-built automated soil flux chambers. The system was a dynamic closed-chamber system, where the headspace gas continually circulated via a 10 m length of polyethylene tubing (Bev-A-Line IV; Cole-Parmer) through the gas analysers, with the flow rate maintained at 2 L/min by the LI-8100.
The chamber (1 m internal diameter and 1.5 m in height) was a cage-like structure, with a circular clear Perspex roof sitting on top of a framework of vertical aluminium rods. A flat circular acrylic flange formed the chamber bottom, to which three concentric rings of closed cell rubber were fixed to ensure a gas tight seal with the landing based during chamber closure. The walls of the chamber were formed by stretching clear 720 gauge (180 µm) polythene (cat no. PM0026; First Tunnels), around the framework, sealed using fibreglass tape. Pressure inside the chamber was equalized with ambient pressure through the inclusion of a vent, after Xu et al. (2006). Landing bases consisted of flat circular flanges on which the bottom of the chamber sat during a measurement. Bases were positioned on the soil surface and packed with fine building sand to form a seal.

| GHG flux measurements
The SkyBeam chamber was deployed from June to December 2013. Measurement length was programmed as 10 min, with a delay of 2 min separating each measurement to allow the gas lines to purge. CO 2 fluxes from the SkyBeam system were calculated using the internal Licor software (Healy, Striegl, Russell, Hutchinson, & Livingston, 1996), with a dead band of 30 s to allow for mixing. The flux was calculated as a linear regression over 2 min, which was found to best describe the instantaneous flux at the time of closure. CH 4 and N 2 O fluxes were measured for two discrete campaigns of approximately 4 weeks each during the study: campaign 1 from 16 July to 12 August; campaign 2 from 6 September to 3 October. During these periods, two cavity ring down laser (CRD) analysers-a fast GHG analyser for CH 4 and an N 2 O analyser (Los Gatos Research)-were incorporated into the SkyBeam assembly, drawing the headspace gas from the exhaust of the IRGA before returning it to the chamber. Both CRD analysers measured at 1 Hz, and fluxes were calculated as the linear regression of the change in concentration over time for a 240 s window following chamber closure. Fluxes were adjusted for chamber volume, area and temperature, which were measured using a thermistor in the chamber headspace. Further adjustment was made to the CO 2 fluxes during daylight hours based on the light response curve to account for attenuation of light by the chamber material, after Heinemeyer, Gornall, Baxter, Huntley, and Ineson (2013). A detection limit was calculated for N 2 O fluxes (Cowan et al., 2014) and is discussed further in Supporting Information. Fluxes of CO 2 measured using the SkyBeam chamber were discarded when r 2 < .9; using this as an indication of a successful chamber closure, N 2 O and CH 4 fluxes which were significant (p < .05) were retained, and non-significant fluxes were considered to be zero.
Soil respiration measurements were made using opaque Licor automated chambers (Li-8100A; Licor) with a multiplexer (Department of Biology Workshops, University of York, York, UK). Chambers were placed over 20 cm diameter collars in the soil, and the collars were kept free of aboveground vegetation. Chambers were programmed to close for 3 min, with a 30 s dead band allowed for mixing, and fluxes were calculated as the linear regression of headspace CO 2 over the remaining closure period using Licor software.

| Environmental variables and harvest
Soil moisture and temperature were measured at 5 cm depth within each of the six experimental plots using SM200 moisture probes and ST1 temperature sensors, and logged as hourly averages on GP1 and DL2 dataloggers (Delta-T). Meteorological data (air temperature, solar radiation and relative humidity) were recorded as hourly averages using an on-site weather station (WP1; Delta-T). Solar radiation (W/m 2 ) was approximated to photon flux density of photosynthetically active radiation (µmol m −2 s −1 ) using the equation (PPFD = 0.47 * solar radiation) from Pankaew, Milton, and Dash (2013). Rainfall data were retrieved from the Met Office MIDAS weather station ca. two miles from the site. Biomass was harvested by hand to coincide with the commercial harvest when the vegetation from each plot was cut at height analogous to mechanical harvesting in spring 2014 and oven-dried at 70°C until at constant weight.

| Data processing
All data analyses and manipulations were conducted in SAS 9.4 (SAS Institute). Cumulative fluxes for NEE were calculated for the period of continuous SkyBeam operation between July and October, and compared to EC fluxes over the same period. Cumulative N 2 O and CH 4 fluxes were calculated for the two discrete measurement campaigns. The cumulative calculation was made using trapezoidal integration; repeated measures analysis of variance for treatment effects on daily mean fluxes was conducted using mixed effects models (SAS proc mixed), with chamber as a random effect. Similar to EC, SkyBeam NEE data were partitioned into ecosystem respiration (R eco ) and GPP following the method of Reichstein et al. (2005).

| CO 2 fluxes
When the first measurements of NEE were collected using SkyBeam in mid-June, all fluxes were positive, indicating net emission of CO 2 to the atmosphere, but by the beginning of July, the crop had developed sufficiently to be a net sink of CO 2 during daylight hours (Figure 2). The magnitude of uptake increased through the summer, peaking at the beginning of September, where values measured were ca. −17 µmol m −2 s −1 (−COMP), −8 µmol m −2 s −1 (+COMP) and −22 µmol m −2 s −1 (Eddy covariance). Maximum release of CO 2 occurred during the night in late July, where values measured were ca. 10 µmol m −2 s −1 ( − COMP and +COMP) and 11 µmol m −2 s −1 (EC). R s was low (<5 µmol m −2 s −1 ) at the start of July, but increased through the month (Figure 3). Fluxes declined through August and September except for a brief period around 14 September when maximum rates of R s (ca. 7 and 10 µmol m −2 s −1 for −COMP and +COMP respectively) were seen which coincided with rainfall, and this spike in R s was reflected in a period of increased (more positive) NEE (Figure 3).
Daily mean NEE from the +COMP treatment (1.8 µmol m −2 s −1 ) was more than 100% greater (more positive) than the −COMP (0.85 µmol m −2 s −1 , F = 220.86, p < .001) and this was reflected with a corresponding 20% increase in daily mean R s from the +COMP plots which approached significance (F = 3.34, p < .07; Figure 4). Although the magnitude of NEE measured with the EC system tended to be larger than the fluxes measured using SkyBeam, there was good agreement in the sign of the flux, and night-time values largely agreed (Figure 4). Daily mean flux fell within one standard error of the mean daily flux from the control plots ( Figure 4). As the crop senesced through November into December, uptake reduced and night-time emissions also fell back to under 5 µmol m −2 s −1 . Over the study period, F I G U R E 2 Net ecosystem exchange (NEE) of CO 2 measured from Miscanthus x giganteus using SkyBeam and eddy covariance (EC).
EC data include measurements at field scale, which was treated with a green compost, SkyBeam was used to measure plots amended with the same compost and an untreated control. EC data are half hourly integrated measurements, SkyBeam are mean (n = 3, ±1 SE) values of each measurement cycle (ca. hourly) GPP was 32% greater in −COMP plots (3.9 µmol m −2 s −1 ) than from +COMP (2.6 µmol m −2 s −1 ) plots (F = 46.90, p < .001; Figure 4), and the reduction in GPP and NEE under the compost treatment was reflected in 37% reduction in the

| N 2 O and CH 4 fluxes
The CH 4 fluxes were predominantly negative, indicating oxidation occurred in the soil ( Figure 5). However, the magnitude of fluxes was small, so that CH 4 made a negligible contribution to the total GHG balance. N 2 O fluxes were also small for the majority of the study, with 95% of fluxes in the range from 0.2 to −0.5 nmol m −2 s −1 ( Figure  6), and over the entire study, there was no significant effect of compost addition on either the N 2 O or CH 4 fluxes. However, there was a clear peak in N 2 O emissions over 5 days, beginning on 22 July, where a maximum flux of 12 nmol m −2 s −1 was seen in the +COMP plots and 8 nmol m −2 s −1 from the −COMP plots. During this time, an average of 109.11 ± 2.11% and 94.70 ± 12.25% of the total N 2 O flux from the first campaign was emitted from the −COMP and +COMP plots, respectively, with a significant (F = 3.97, p < .05) 34% increase in N 2 O emissions from the +COMP treatment. This peak occurred following significant rainfall (8 mm) which was the first precipitation in over 4 weeks.

| CO 2 fluxes
The expected diurnal pattern of NEE was seen, with uptake during the day and net release at night. The relationship between NEE and PAR was well described by Michaelis-Menten response curves for fluxes measured using both EC and SkyBeam ( Figure S1). For all but chambers 4 and 6, rates of CO 2 uptake were lower in SkyBeam plots than the entire EC footprint and the Pmax (maximum rate of GPP) values of the response curves were reflected in the biomass harvested from the chambers. The diurnal pattern in R s was not a simple sinusoidal form, but peaked twice during the day, once in the afternoon and once in the evening (Figure 7). This was reflected in the low correlation between R s and soil temperature in both treatments (r 2 = .13 and .11, p < .001, −COMP and +COMP, respectively) across the whole period, though no other environmental variable explained more variation in fluxes.

| N 2 O fluxes
During the emission event (22-25 July), N 2 O fluxes were most closely related to the cumulative rainfall over the previous 24 hr for both the control and the compost treatments (r 2 = .52 and .45, p < .001 respectively). Between July and October, there was a significant increase in soil moisture at 5 cm depth under the +COMP treatment (F = 7.65, p < .05) compared to the control, but there was no effect on soil temperature (Figure 8). Outside of this event, the most important predictor of N 2 O flux during the first campaign was soil temperature, which displayed a clear negative relationship with N 2 O flux (r 2 = .26 and .16, p < .001); furthermore, on a diurnal scale, the hourly N 2 O fluxes showed a clear, anticlockwise hysteresis with soil temperature (Figure 9). The hysteresis is such that at similar temperatures, there was a switch in the direction of fluxes from positive to negative, with the greatest N 2 O uptake coinciding with the first daily R s peak (ca. 15.00).

| DISCUSSION
Using real-time monitoring of GHG emissions with tall chambers over Miscanthus, we have shown the importance of short-lived emissions' events with regard to the total GHG F I G U R E 9 Mean hourly N 2 O flux, measured using SkyBeam from Miscanthus from untreated controls (a) and with compost addition (b), against mean hourly soil temperature at 5 cm. Whilst there is a negative relationship between soil temperature and N 2 O flux (as indicated by the linear regression and statistics included on the panel), there was also a clear hysteresis, which displayed an anticlockwise pattern. Each flux measurement is labelled with the hour | 1465 KEANE Et Al. flux and revealed a diurnal variation in N 2 O emissions. We were also able to investigate the extent to which compost addition altered GHG emissions from this bioenergy crop.

| Effect of compost on N 2 O emissions
The effect of compost addition was shown to be detrimental to several aspects of commercial Miscanthus production, most notably a reduction in the biomass produced. Whilst the productivity from both treatments was low, they were typical of year one harvest values for UK climatic conditions (McCalmont, McNamara, Donnison, Farrar, & Clifton-Brown, 2017), which perhaps reflects the disturbance caused by the harrowing of the field. Not only did the addition of compost to Miscanthus fail to improve productivity in our study but also it negatively affected the carbon balance of Miscanthus production by reducing GPP, whilst increasing NEE and R s . Furthermore, compost addition also increased emissions of N 2 O by more than a third compared to the noncomposted control. Whilst the total emitted N 2 O was small in terms of the GHG balance, any N 2 O production during Miscanthus cultivation will reduce its efficacy as a CO 2 mitigation strategy. The increase in N 2 O emissions measured under compost addition in our study was ca. 28 kg CO 2 -eq/ ha, and under annual application might lead to a further 513 kg CO 2 -eq/ha (Table S1)

| High N 2 O emission event
Although the observed high flux event lasted only a few days following compost addition, its duration was slightly under 10% of the total time N 2 O measurements were made during our study. If further high flux periods occurred at the same regularity outside of measurement campaigns, the annual budget would be greater than reported here. This flux event was characterized by a rapid increase in N 2 O emissions following heavy rain (8 mm) marking the end of four dry weeks. Similar N 2 O emissions events have been previously reported after rainfall (e.g. Breuer, Papen, & Butterbach-Bahl, 2000;Mummey, Smith, & Bolton, 1997;Saha et al., 2017), and we suggest that the driver of the event reported in this paper was that nitrate (NO 3 ) built up in the soil during the preceding dry weeks, which was rapidly denitrified as the soil became anaerobic following the rainfall, a process experimentally described elsewhere (Krichels, DeLucia, Sanford, Chee-Sanford, & Yang, 2019). This event reinforces the necessity to measure GHG fluxes at a temporal resolution appropriate to detect such emission events. The peak in N 2 O emissions seen here was equivalent to ca. 2000 µg m −2 hr −1 , which is more than ten times the magnitude of fluxes reported from Miscanthus in a long-term study at this site (Drewer et al., 2012) and approximately half the maximum rate seen at this farm following the addition of 120 kg N/ha ammonium nitrate fertilizer to oilseed rape (OSR, Brassica napus L.; Keane et al., 2018). While fluxes of a similar magnitude have been reported from unfertilized Miscanthus elsewhere (Guzman, Ussiri, & Lal, 2017), N 2 O emissions from established Miscanthus plantations are widely thought to be much lower than those from conventional cropping .

| Uptake of N 2 O and the diurnal variation of fluxes
A striking characteristic of the N 2 O fluxes presented here was the diurnal pattern, particularly for the first N 2 O campaign. Whilst diurnal patterns of N 2 O emission have been seen several times before (e.g. Alves et al., 2012;Ryden, Lund, & Focht, 1978;Shurpali et al., 2016;Yamulki et al., 2001), and were very recently observed in OSR at the same farm (Keane et al., 2018), the diurnal variation reported by these authors was defined by a peak in emissions during the day and lower emissions during the night. The sole exception was reported by Shurpali et al. (2016), who saw higher emissions at night under N-limiting conditions. Of these studies, all but Keane et al. (2018) attribute the pattern to soil temperature, and to our knowledge, there are no reports of a transition between N 2 O uptake during the day and production during the night.
The majority of N 2 O uptake in soils is thought to be the result of complete denitrification of N 2 O to dinitrogen (N 2 ) gas, which is a process that occurs in anoxic conditions, and therefore at high soil moisture content (Conen & Neftel, 2007). The negative fluxes presented here occurred at very low soil moisture content (<0.2 m 3 /m 3 ). It has been suggested that soil microbes will reduce N 2 O even in relatively dry (20% water-filled pore space) soils (Warneke, Macdonald, Macdonald, Sanderman, & Farrell, 2015), with negative fluxes even reported at 2% soil water content (Wu et al., 2013). Negative fluxes of N 2 O have also been reported in dry and oxic vegetated soils under grasslands (Flechard, Neftel, Jocher, Ammann, & Fuhrer, 2005) and forestry (Goldberg & Gebauer, 2009). There is strong evidence that uptake in soils is a biological process (Warneke et al., 2015;Wu et al., 2013) and it has been demonstrated recently that, contrary to previous understanding, some obligate aerobic bacteria can reduce N 2 O to N 2 (Park, Kim, & Yoon, 2017). It has largely been accepted that the consumption of N 2 O by microbes is through its use as an electron acceptor, particularly when NO 3 in the soil is scarce (Flechard et al., 2005); in such conditions, denitrification by heterotrophic nitrifiers might drive greater N 2 O reduction than production, leading to a net negative flux. However, the specific aerobic microbial N 2 O consumption pathway has not yet been confirmed, despite recent identification of novel N 2 O-reducing organisms (Hallin, Philippot, Loffler, Sanford, & Jones, 2018). But laboratory work on several marine bacteria has shown that microbes possessing the nifH gene responsible for N 2 fixation can also use N 2 O to assimilate N directly (Farias et al., 2013) in the presence of O 2 , a process which is twice as energetically efficient as N 2 fixation and therefore may be preferable to such microbes. Stable isotope probing, using 15 N 2 O to incubate a variety of aerated low-N soils has revealed the presence of 15 N in microbial organic compounds (J.E. Stockdale, C.B. Whitby, B. McKew, & P. Ineson, in preparation), suggesting that similar N 2 O fixation may also occur in soils. It has been shown that Miscanthus rhizomes host N 2 fixers (Davis et al., 2010;Liu & Ludewig, 2019), which have been suggested as an explanation for its high N use efficiency and this could be a contributing factor to N 2 O uptake seen here. Diurnal variation in nifH expression and N fixation has been demonstrated to be controlled by CO 2 concentrations (Stockel, Elvitigala, Liberton, & Pakrasi, 2013) and given that the highest daily rates of N 2 O uptake presented here coincided with highest daily rates of R s , it would suggest that N 2 O consumption peaked when CO 2 concentration in the soil was greatest. Alternatively, if N 2 O uptake was driven by heterotrophic nitrifier denitrification, this may have been stimulated by the arrival of C in the rhizosphere from photosynthate, which would also drive the first daily peak in R s at around 15.00.
Given the presence of vegetation in the SkyBeam chamber, it should not be discounted that N 2 O uptake was a plant-mediated process, or the interaction of the plant and soil microbes. It has previously been shown that maize (Zea mays L), a C 4 grass like Miscanthus, can absorb N 2 O through stomata which it both metabolizes and stores (Grundmann, Lensi, & Chalamet, 1993). As diurnal maximum N 2 O uptake coincided with peak GPP, it is possible that this was the cause of the pattern seen in our study. Positive plant-mediated fluxes of N 2 O have been posited previously, where the N 2 O is produced in the soil and transported to the atmosphere through the transpiration stream (Ferch & Römheld, 2001); if N 2 O is being consumed at depth in the soil, it is perhaps possible that the reverse is true and that atmospheric N 2 O is transported along a concentration gradient through the Miscanthus.

| Temperature hysteresis of N 2 O flux
The negative relationship of N 2 O fluxes with soil temperature in this study suggests that N 2 O uptake increased with rising temperature. Mills, Dewhirst, Sowerby, Emmett, and Jones (2013) found a similar negative relationship between N 2 O flux and soil temperature as reported here, even to the extent that fluxes switched from positive to negative above 20°C, but that work was conducted on a podzol at field capacity, a contrast to the drier conditions at our study site. The situation is complicated as both uptake and emission occurred in the same location, with the balance altering throughout the day. The net flux is the sum of both of these processes, which may have different drivers. This is perhaps reflected in the diurnal hysteresis in the relationship between flux and soil temperature. Hystereses have been demonstrated several times between R s and soil temperature (e.g. Phillips, Nickerson, Risk, & Bond, 2011;Riveros-Iregui et al., 2007;Zhang et al., 2015) and even under Miscanthus at the same study site (Keane & Ineson, 2017), but we believe that this is the first time temperature hysteresis on a diurnal scale has been reported for N 2 O fluxes. Since it appears that increasing temperature stimulated greater uptake, the anticlockwise hysteresis presented here is analogous to a clockwise hysteresis in R s fluxes, which has been suggested to be a consequence of thermal effects on gas diffusivity in soil (Zhang et al., 2015). The same authors did not consider that soil moisture fluctuations in the rooting zone were sufficiently large to be an important driver of hysteresis on a diurnal scale, and moisture levels in the bulk soil certainly did not vary so greatly over the course of the day to attribute N 2 O uptake to this process here. As with the diurnal pattern of N 2 O flux, we suggest that the temperature hysteresis may be driven by plant carbon; increasing soil temperature between 09.00 and 15.00 drives N 2 O uptake in the soil, which declines with temperature until 17.00. The arrival of photosynthate to the rhizosphere could then supply C which stimulates heterotrophic denitrification and therefore N 2 O production between ca. 20.00 and 09.00. Pulse labelling with 13 C in Miscanthus at this site has shown that C is available to soil microbes four hours after assimilation (Elias et al., 2017), and with highest rates of NEE seen from 11.00 to 15.00, photosynthate should start to arrive at the rhizosphere around 15.00. This coincided with the increase in N 2 O fluxes, and the arrival of plant C to the soil late in the day might also explain why R s remained above the daily average throughout the night time, despite coinciding with the lowest soil temperatures.

| CONCLUSION
The SkyBeam system performed well, producing CO 2 data which compared favourably with EC observations. The chamber system facilitated GHG flux measurements at the plot scale, enabling replicated manipulation experiments impossible with EC alone. Whilst large chambers have been built previously to measure gas exchange from over large vegetation (Mordacq et al., 1991), we have not found another automated flux system working at this scale for such | 1467 KEANE Et Al. a continuous period. Automation of the system was vital to identify not only the diurnal characteristics of GHG fluxes but also to detect the short-lived burst of N 2 O emissions between 22 and 26 July. Had a monthly or even weekly measurement schedule of manual measurements been conducted, it is likely that this event would have been missed.
Our study into the effects of compost addition on Miscanthus cultivation after harrowing identified negative effects on the GHG balance and crop productivity. This supports the notion that all farm interventions, across the life cycle of a crop, need to be considered for understanding the GHG balances of bioenergy crop cultivation. In this particular study, we have presented strong evidence that both cyclical and episodic events of N 2 O can occur that impact on the net GHG balance of bioenergy systems. Our measured N 2 O emissions in the context of reversion (McCalmont et al., 2018) and conversion (Holder et al., 2019) of bioenergy crops are relatively small, however, with other fertilizers and in other situations, such emissions may become tangible factors in GHG life cycle emissions. Fundamentally, there is a need for further investigation of the processes underlying N 2 O uptake and emission in plants and soil, which is relevant to both fertilized and unfertilized bioenergy and food crops, to reduce uncertainties surrounding environmental benefits .