A Miscanthus plantation can be carbon neutral without increasing soil carbon stocks

National governments and international organizations perceive bioenergy, from crops such as Miscanthus, to have an important role in mitigating greenhouse gas (GHG) emissions and combating climate change. In this research, we address three objectives aimed at reducing uncertainty regarding the climate change mitigation potential of commercial Miscanthus plantations in the United Kingdom: (i) to examine soil temperature and moisture as potential drivers of soil GHG emissions through four years of parallel measurements, (ii) to quantify carbon (C) dynamics associated with soil sequestration using regular measurements of topsoil (0–30 cm) C and the surface litter layer and (iii) to calculate a life cycle GHG budget using site‐specific measurements, enabling the GHG intensity of Miscanthus used for electricity generation to be compared against coal and natural gas. Our results show that methane (CH4) and nitrous oxide (N2O) emissions contributed little to the overall GHG budget of Miscanthus, while soil respiration offset 30% of the crop's net aboveground C uptake. Temperature sensitivity of soil respiration was highest during crop growth and lowest during winter months. We observed no significant change in topsoil C or nitrogen stocks following 7 years of Miscanthus cultivation. The depth of litter did, however, increase significantly, stabilizing at approximately 7 tonnes dry biomass per hectare after 6 years. The cradle‐to‐farm gate GHG budget of this crop indicated a net removal of 24.5 t CO2‐eq ha−1 yr−1 from the atmosphere despite no detectable C sequestration in soils. When scaled up to consider the full life cycle, Miscanthus fared very well in comparison with coal and natural gas, suggesting considerable CO2 offsetting per kWh generated. Although the comparison does not account for the land area requirements of the energy generated, Miscanthus used for electricity generation can make a significant contribution to climate change mitigation even when combusted in conventional steam turbine power plants.


Introduction
Climate change is unlikely to be solved with a shortterm solution, but alternative renewable fuel sources, like bioenergy, can be a part of the long-term solution. Therefore, it is essential to ensure these bioenergy crops are helping to turn atmospheric carbon dioxide (CO 2 ) into stable long-lived carbon (C) forms, rather than the reverse. As alternative energy sources, bioenergy crops and lignocellulosic feedstocks often fare well against conventional fuels in both socio-economic (Paine et al., 1996;Domac et al., 2005;Remedio & Domac, 2003) and environmental (Cherubini et al., 2009;Smeets et al., 2009;) comparisons. The bioenergy crop, Miscanthus x giganteus Greef et Deu (Hodkinson & Renvoize, 2001) (herein Miscanthus), has attracted attention in North America and Europe due to high yields (Christian et al., 2008;Heaton et al., 2008), low management requirements (Miguez et al., 2008;Gopalakrishnan et al., 2011;McCalmont et al., 2015) and the potential for improved soil C stocks (Hansen et al., 2004;Schneckenberger & Kuzyakov, 2007;Poeplau & Don, 2014). These characteristics make Miscanthus a particularly attractive crop in the light of climate change mitigation options McBride et al., 2011).
A key area of uncertainty when assessing the sustainability of bioenergy crops surrounds their potential to sequester more C in crop residues and soils than is emitted through production, transport and end-use processes of the harvested biomass. Quantifying the complete life cycle C budget of bioenergy plantations is therefore essential to accurately determine any potential GHG savings. This GHG mitigation potential is an important part of formal life cycle assessments (LCAs) for bioenergy crops that evaluate their environmental impact from cradle to grave (e.g. Adler et al., 2007;Rowe et al., 2011). To date, empirical measurements of the GHG balance of Miscanthus cultivation have produced inconsistent outcomes (Toma et al., 2011;Drewer et al., 2012;Zimmermann et al., 2012;Poeplau & Don, 2014). As a consequence, GHG emissions data included in Miscanthus LCAs are often modelled (e.g. Hamelin et al., 2012) or use IPCC default emission factors (e.g. Brandão et al., 2011). To address this area of uncertainty, we focused on cultivation of Miscanthus from the cradle-to-farm gate to quantify the C sequestration potential of Miscanthus. For this, we measured four years of soil GHG emissions and net ecosystem exchange (NEE) from a 3-to 7-year-old commercial Miscanthus plantation in the United Kingdom, also measuring soil C stocks and accumulated plant litter.
Assessing the GHG budget of Miscanthus requires more than estimates of C assimilation through photosynthesis as soil C sequestration can offset a large proportion of GHG emissions from the field (Lal, 2004). Temperature (Kirschbaum, 1995) and water availability (Orchard & Cook, 1983;Wood et al., 2013) are both major drivers of the microbial processes that incorporate C into soils. Further, the 'quality' of plant litter (quantified by C : N ratios or lignin : N ratios) can influence how quickly that C is decomposed (Taylor et al., 1989;Donnelly et al., 1990;Bonanomi et al., 2013). Consequently, it is important to consider these factors when evaluating soil C sequestration. Senesced Miscanthus biomass is typically very low in N due to nutrient translocation. This results in low litter quality (Amougou et al., 2011) which has a significant impact on the rate of C turnover from the litter layer into the topsoil . Root decomposition also contributes to soil C sequestration, but Miscanthus-specific data are limited to a few studies (Rasse et al., 2005;Agostini et al., 2015). The majority (>50%) of belowground biomass is found in the top 30 cm (Neukirchen et al., 1999;Amougou et al., 2011), with C inputs from roots and rhizomes estimated to be as high as 0.86 tC ha À1 yr À1 and 2.66 tC ha À1 yr À1 , respectively . However, a recent study suggests that rhizosphere activity under Miscanthus may stimulate priming, causing a loss of native soil C and offsetting fresh C inputs (Zatta et al., 2014). Long-term studies are therefore required to assess litter accumulation, belowground biomass and soil C stock changes in Miscanthus plantations, in order to quantify its benefits for climate change mitigation (Poeplau & Don, 2014;Robertson et al., 2015).
While C stocks in litter, standing biomass and soils are important 'pools' to quantify, their changes over time are relatively slow compared to the 'fluxes' of the system that include photosynthesis and respiration (Kuzyakov, 2011). These processes continually respond to environmental conditions and often follow diurnal patterns strongly influenced by crop physiology (Linn & Doran, 1984;Rochette et al., 1999;Cheng et al., 2003). At the ecosystem scale, the balance between C uptake and CO 2 efflux is described as the NEE, and within the C cycle, this is the largest flux between atmosphere and a bioenergy plantation. NEE is typically calculated using eddy covariance to continuously monitor changes in CO 2 concentration above the plantation canopy (Baldocchi, 2003). Although the C stored in aboveground biomass is often quantified for bioenergy crops when they are harvested, measurements of the NEE are required to ensure that the amount stored in pools is in excess of the amount emitted through fluxes.
In many agricultural systems, CO 2 is not the only GHG of importance with nitrous oxide (N 2 O) emissions often contributing more to a crop's overall GHG balance than the NEE (Flessa et al., 2002). Despite established measurement techniques, relatively few studies have measured soil GHG emissions from Miscanthus plantations. The limited data available show that emissions of both N 2 O and methane (CH 4 ) from soils are low and CO 2 efflux dominates soil GHG emissions (Toma et al., 2011;Drewer et al., 2012;Gauder et al., 2012). To accurately quantify an average annual efflux of these GHGs, data are required throughout the year and ideally over several years. In this study we measured GHG emissions and NEE in a Miscanthus plantation in Lincolnshire, UK, from 2009 to 2013 (growth years 3 to 7). We then used parallel measurements of climatic variables to explore the environmental controls on soil respiration (CO 2 ), CH 4 and N 2 O emissions, including the temperature sensitivity of respiration at different stages in the crops growth cycle. The aims of the study were to quantify the relative contributions of each GHG towards the net GHG balance of the site, and to better understand their relationship to temperature and soil moisture as environmental drivers. CO 2 was expected to dominate site GHG fluxes, with warmer and wetter periods driving the greatest soil respiration rates. In addition, changes in soil C stocks and the litter layer were quantified over time, with the expectation that the dynamics of these C pools are largely responsible for sequestration rates reported for Miscanthus (e.g. Dondini et al., 2009). These data were then used to calculate a life cycle GHG balance of Miscanthus cultivation in order to compare Miscanthus as a source of electricity to coal and gas.

Study site
The field experiment was conducted in an 11.5-ha commercial Miscanthus plantation near Lincoln, Lincolnshire, UK. The soil type is a compacted loam that behaved like a heavy clay, with approximately 15 %, 36 % and 49 % of clay, silt and sand, respectively, in the top 30 cm of soil. The top 30 cm of soil had a mean total C and N concentration of 1.86 % and 0.18 %, respectively, with a soil pH ranging from 6.8 to 7.3. The bulk density of the soil was 1.46 AE 0.03 g cm À3 for the 0to 15-cm layer and 1.53 AE 0.02 g cm À3 for the 15-to 30-cm soil layer. Root biomass (live and dead) was estimated at the end of the 7th growth year: 2.61 t dry mass ha À1 for 0-15 cm and 1.85 t dry mass ha À1 for 15-30 cm. Additional soil characteristics sampled monthly for two years within this study can be found in Table S1. The deeper soil profile showed an increasing bulk density (1.59 AE 0.20 g cm À3 , 30-50 cm; 1.62 AE 0.10 g cm À3 , 50-100 cm) and a clear B-horizon at the plough depth (30 cm). There was little evidence of root biomass propagation below 70 cm when trenches were dug in early 2009. The site had a mean annual precipitation of 605 mm and a mean annual temperature of 9.9°C (30-year average 1980-2009). The Miscanthus was established in 2006 at a density of 10 000 rhizomes ha À1 . The crop was harvested annually in the spring, beginning in March 2008, but biomass was only removed from 2009 onwards; bale yields (20% moisture content) were recorded as 6.95, 10.28, 6.24, 7.58 and 6.87 dry t ha À1 for 2009 to 2013, inclusive. The only addition of fertilizer was in April 2010, when a phosphorus-potassium fertilizer was applied at a rate of 125 kg ha À1 . The land management prior to conversion to Miscanthus was a crop rotation of wheat and oilseed rape, with three years of wheat directly before conversion. Further site details can be found in Robertson et al., 2016. Sampling strategy and eddy covariance In early May 2008 a meteorological tower was established in the north east corner of the Miscanthus plantation, along with a flux mast positioned to maximize CO 2 measurements given prevailing winds over the cropped area. The tower and mast were equipped with a number of devices to continuously (every 30 min) monitor a range of environmental conditions (Table S3), including an ultrasonic anemometer and infrared gas analyser (IRGA) to employ an eddy covariance (EC) system to examine NEE (more details can be found in S.1). Measurements were taken from 7 May 2008 until 10 March 2013 with some exceptions around the harvesting times where instrumentation was removed. NEE data were cumulated for each growth year (March to February) and an average taken over the four full years of measurements (March 2009to February 2013, reported in g CO 2 -C m 2 .

Soil-atmosphere gas fluxes
Measurements of soil GHGs (CO 2 , CH 4 and N 2 O) were taken from October 2008 until March 2013 using the static chamber method described by Livingston & Hutchinson (1995), adapted to include the use of a pressure 'vent'. Five chambers made from PVC (40 cm diameter and 20 cm height) were inserted approximately 3 cm into the soil surface (exact volumes noted). This avoided severing many of the fine roots that were found very close to the soil surface (similar strategies have been recommended in different land uses by Heinemeyer et al., 2011 andMills et al., 2011). All chambers remained in the soil except at harvest times. Chambers were replaced in the same approximate location after each harvest, with proximity to plants taken into consideration, aiming to represent the average spacing throughout the plantation. At the exact time of GHG sampling, and near the location of GHG sampling, volumetric soil moisture (0-6 cm depth) was measured using a ML29 Theta Probe and Meter HH2 (Delta T Devices, UK) as well as soil (0-7 cm depth) and air temperature measurements using a Tiny Tag temperature logger with integral stab probe (Gemini Data Loggers, UK). Measurements were not taken between December 2010 and April 2011 or in April 2012 due to funding constraints and harvest activities, respectively.
At times of sampling, chambers were closed with a reflective aluminium lid, which had a rubber seal around the edge to prevent leakage. Chambers were enclosed for 30 min with one 10-ml sample taken every 10 min for a total of four time points collected for each plot. At the time of sampling, gas samples were transferred from the chamber headspace into a 3-ml gastight exetainer (Labco Ltd, Lampeter, UK) via a needle and syringe inserted into the self-sealing septa in the chamber lid. The majority (>85%) of GHG measurements were taken between the hours of 10:30 and 14:30 with some exceptions due to field logistics. Exetainer gas samples were analysed on a Perkin-Elmer Autosystem XL Gas Chromatograph (GC) fitted with a flame ionization detector (FID) for CO 2 and CH 4 and an electron capture detector (ECD) for N 2 O. All results were calibrated against certified gas standards (BOC, UK) (Case et al., 2014) and converted to a total flux reported as mg CO 2 -C m À2 h À1 , lg CH 4 -C m À2 h À1 or lg N 2 O-N m À2 h À1 in accordance with methods detailed in Holland et al. (1999).

Carbon and nitrogen in soil, vegetation and litter
In parallel with monthly GHG measurements, soil samples were collected using PVC pipes (5 cm internal diameter) hammered into the topsoil (0-15 cm) from five locations, one each within a 10 m radius of the static chambers. These cores were taken in March 2009 and March 2010 and then at monthly intervals from May 2011. Further, in October 2011, May 2012, October 2012 and March 2013 additional 30-cm-depth cores (split into 0-to 15-cm and 15-to 30-cm layers) were taken using a 2.5-cm-diameter gouge auger (Van Walt, Haslemere, UK). All soil collected was for destructive sampling and used for C and N determination. The routine monthly 0-15 cm cores were homogenized and freeze-dried (Alpha 1-4 LD, Martin Christ, Osterode am Harz, Germany) before being gently ground by hand to pass through a 2 mm sieve. The 0-30 cm cores were air-dried to constant weight at room temperature before being homogenized, ground and sieved. No differences in C or N concentration were seen between the freeze-dried and air-dried samples. All visible plant matter remains (e.g. roots and leaf litter) were removed before grinding. Small subsamples of the ground soil were taken for analysis of C and N concentration through combustion in an elemental analyser (Costech ECS 4010; Milan, Italy). C and N stocks were estimated by relation to fixed site bulk densities (1.46 for 0-15 cm and 1.53 for 15-30 cm) and the depth layer (Guo & Gifford, 2002). These bulk densities were taken from 15 replicates using a 4.8-cm-diameter, 40-cm-deep split-tube sampler (Eijkelkamp Agrisearch Equipment BV, Giesbeek, the Netherlands). Care was taken to avoid compaction during coring and, where necessary, bulk density was corrected for compression based on the depth of the hole. To ensure consistency when calculating C and N stocks, the resulting bulk density for 0-15 cm was checked against the PVC cores taken monthly.
In October 2011, an adjacent field was sampled to provide an estimate of soil conditions before the Miscanthus was planted (a paired-site approach). This allowed a comparison to be made where samples from the adjacent field represent timezero reference values of soil C and N stocks. This field had followed the same land use as the Miscanthus field prior to planting in 2006, was seeded with oil seed rape in 2006 and 2010, and winter wheat all other years. Before sampling in 2011, it had recently been harvested for winter wheat before being ploughed and cultivated again. Three replicates at five random locations were cored using the same split-tube sampler (Eijkelkamp Agrisearch Equipment BV, Giesbeek, the Netherlands) and split into 0-15 cm and 15-30 cm (n = 15). The soil was then freeze-dried, sieved to 2 mm and analysed for C and N. The same procedure to remove plant matter remains from the soil samples was applied. Further, these cores were analysed for bulk density and corrected for compression through coring (0-15 cm, 1.13 AE 0.17 g cm À3 ; 15-30 cm, 1.41 AE 0.15 g cm À3 ). C and N stocks were calculated using the field-specific bulk density values. No carbonates were detected at either depth from either field.
Between October 2008 and March 2013 senesced aboveground biomass was collected using twenty five 2-m 2 litter traps. Traps were placed on top of the litter layer throughout the plantation, with senesced biomass collected and weighed on a monthly basis and values extrapolated to an average rate per hectare. Subsamples of the senesced biomass were weighed and returned to the laboratory for moisture content determination (oven-dried at 105°C for 24 h). The resulting dried subsample was then ground by freeze-milling (6770 Freezer/Mill, SPEX SamplePrep, Stanmore, UK) before C and N concentrations were determined. The amount of biomass added to the litter layer after harvesting, termed harvesting inefficiency, was also quantified by measuring the size of the litter layer before and after harvest. This varied between years but was proportional to the aboveground yield. Using an average of the measurements taken, a standard value of 5% of the year's harvest was used in future calculations (this value was similar to that reported by Sanderson et al., 1997).
After harvesting (in May 2011, March 2012and March 2013, the size (t ha À1 ) of the litter layer was quantified by collecting all of the O-horizon (lightly raked from the soil surface) from 1.6 m 2 circles at 25 random locations throughout the plantation before extrapolating to a per area average for the site (after moisture content was determined by drying in an oven at 105°C until constant weight,~24 h). Additionally, the litter layer was quantified at 15 locations at six time points between March 2012 and March 2013 (May, June, August, September, October and January). Subsamples of the litter layer were dried, milled and analysed for C and N concentration. The decomposition rate of this litter layer was assessed assuming first-order decay rates as per Olson (1963), deriving a constant (k) to match a line of best fit through measured litter layer points. This constant was compared to two other studies for Miscanthus litter, Amougou et al. (2012) and Yamane & Sato (1975) who reported k values of 0.776 and 0.511, respectively.
Finally, standing biomass was partially harvested in October 2012 and March 2013 to assess C and N concentrations at the beginning and end of crop senescence. Nine stems were selected at random from different plants. Stems and leaves were separated, weighed and dried at 105°C until constant weight (~24 h) to calculate moisture content. Dry biomass was then freeze-milled and analysed for C and N concentrations. All C and N concentrations were determined using the same elemental combustion analyser (Costech ECS 4010).

Site-specific life cycle GHG balance
To assess the contribution of site GHG emissions and changes in C stocks to the life cycle GHG balance of Miscanthus, an annual budget was calculated taking into account soil GHG fluxes, NEE and topsoil C stocks (0-30 cm). The mean annual NEE was used for net CO 2 emissions and cumulative annual CH 4 and N 2 O emissions were derived from chamber fluxes using monthly data from the four years. CO 2 chamber data refer to soil emissions only and were not used in life cycle estimates. CH 4 and N 2 O cumulative annual emissions were transformed using 100-year global warming potentials (GWPs), calculated as CO 2 equivalents (CO 2 -eq) according to Myhre et al. (2013) (CH 4 = 34; N 2 O = 298). The cradle-to-farm gate GHG balance was presented as an annual GHG balance per unit area extrapolated to the end of the plantation lifetime. This assumed an 18 year lifecycle of the plantation (DEFRA, 2007) and followed conventional cultivation routines (Table S4) including ploughing before planting as well as at the end of the plantation lifetime to prepare the site for the next crop (Styles & Jones, 2007;Thornley et al., 2009). Direct and indirect emissions associated with other site operations were considered according to Miscanthus-specific estimates of diesel requirements reported by Lewandowski et al. (1995), Smeets et al. (2009) and Thornley et al. (2009).
Applying an assumed 20% moisture content of Miscanthus biomass when harvested and combusted (Lewandowski et al., 2000) a realized calorific value (lower heating value (LHV)) of 14 MJ kg À1 (ECN, 2015) was used to estimate GHG intensity. Additionally, a lifetime harvested yield from the plantation was estimated to be 129.2 tonnes dry biomass ha À1 (Table S5).
In accordance with the common observation that productivity declines as a Miscanthus stand ages (Clifton-Brown et al., 2007;Angelini et al., 2009;Arundale et al., 2014), the findings of Lesur et al. (2013) were applied to decrease yields proportional to stand age. Lesur et al. (2013) observed a maximum yield of 16.8 dry t ha À1 in year 8 and a decrease of 0.647 dry t ha À1 in each subsequent year. This reported maximum yield seems unrealistic at our site; therefore, the highest observed yield (10.28 dry t ha À1 in 2010) was assumed to be the site-specific maximum. Consequently, this is approximately 49% of that reported by Lesur et al. (2013) and so the rate of yield decline is scaled accordingly (0.396 dry t ha À1 yr À1 ). The resulting lifetime plantation yield (129.2 dry t ha À1 ) compares well with the alternative approach (121.3 dry t ha À1 ) to average measured yields of the first seven years and assume that average is stable over the plantation's lifetime (Table S5). It is important to note that in other areas of the world the harvested biomass may have a lower moisture content (Heaton, 2006), thereby incurring an increased LHV.
The final cradle-to-grave GHG balance was estimated for the Miscanthus plantation and reported using the standard notation of emissions per unit of energy generated (GHG intensity; g CO 2-eq kWh À1 ). This calculation was divided into three procedures (combustion, transportation and production) and was based on a number of informed assumptions. The Miscanthus biomass was assumed to be cofired for electricity generation in conventional steam turbine power stations where conversion efficiency of this solid biomass was 30% (1 MJ biomass = 0.30 MJ electricity) (Howes et al., 2002). Although the conversion rate efficiency of biomass to energy can be considerably higher in combined heat and power (CHP) plants (~70%; Cannell, 2003), conventional electricity generation was employed to estimate the most realistic current scenario when comparing with traditional fossil fuels. This resulted in a GHG intensity associated with combustion as defined by Eqn (1).
where GHG com is the GHG intensity of Miscanthus combustion for electricity generation in g CO 2-eq kWh À1 ; Y is the harvested yield in g biomass ha À1 at an assumed 20% moisture content (i.e. 129200000 over this plantation's lifetime); C conc is the carbon concentration of the harvested biomass as a fraction (0 to 1); CO 2 mol is the molecular mass of CO 2 ; C mol is the molecular mass of carbon; Cal is the calorific content of Miscanthus in MJ g biomass À1 (i.e. a LHV of 0.014 given an assumed 20% moisture content); Eff is the conversation rate efficiency in power stations as a fraction (i.e. 0.30); and E conv is the energy conversion from MJ to kWh (i.e. 0.278 as 1 kWh = 3.6 MJ; Thompson & Taylor, 2008). The GHG intensity associated with transporting the biomass to a power plant assumed a 160-km round trip (based on the location of local power plants) using a vehicle averaging 2.44 km per litre of diesel while carrying the equivalent of 25 tonnes biomass (NAP, 2010). Total GHG emissions of using 1 l of diesel to transport over land were assumed to be 3644 g CO 2 -eq (Smeets et al., 2009). This resulted in Eqn (2).
where GHG trans is the GHG intensity of Miscanthus biomass being transported between the plantation and a power station in g CO 2-eq kWh À1 ; PP dist is the round trip distance to the power station (i.e. 160 km); F eff is the fuel efficiency of the truck used in transportation (i.e. 0.41 l km À1 ); F emi is the truck emissions associated with 1 l of fuel used during transportation (i.e. 3644 g CO 2-eq l À1 ); and L is the truck load of biomass (i.e. 250 00 000 g). Finally, the GHG intensity of cradle-to-farm gate production was calculated using Eqn (3).
where GHG prod is the GHG intensity in g CO 2-eq kWh À1 of Miscanthus biomass being grown and harvested including; GHG site is the GHG balance in g CO 2-eq ha À1 yr À1 of all direct and indirect emissions, using NEE to estimate CO 2 exchange as well as CH 4 and N 2 O measurements at the soil surface; and P life is the plantation lifetime in years (i.e. 18). Ultimately, the sum of these three procedures were compared to full life cycle GHG budgets for coal and natural gas when used for electricity generation, as derived from MacKay & Stone (2013).

Statistical analysis
Outliers of GHG measurements were excluded when outside 29 standard deviation, as per Altman & Bland (1995), assuming normal distribution between all measurements of each gas at each time point, thereby retaining 95% of the data. All statistical analyses were performed with R version 3.0.2 (R Core Team, 2014). A global model was formed to define relationships between GHG fluxes and environmental variables (soil temperature, soil moisture, crop phase and a soil temperature * soil moisture interaction). User-defined growth phases of the crops were used to specify whether the crop was dormant (D), emerging (E) or growing (G). These each referred to four months of the year (November to February, March to June and July to October, respectively); the phases were found to be a significantly better predictor of CO 2 efflux than the traditional spring-summer-autumn-winter divisions. Regression analysis was used to quantify the variance in GHG emissions explained by each of the environmental variables through use of the lme function as part of the NLME package (Pinheiro et al., 2013) and the r.squaredGLMM function, part of the MUMIN package (Barto n, 2012). To meet the assumptions of linear mixed effects (LME) models, log transformations to the flux data were required for soil CO 2 emissions and residuals were transformed using the varPower function (in NLME) for CH 4 and N 2 O fluxes. Each chamber was used as the random effect to account for repeated sampling from the same location. This allowed estimates of how much variation in the measurements was explained by the different environmental factors.
Relationships of soil GHG emissions with soil temperature and soil moisture were explored in detail. The temperature sensitivity of CO 2 fluxes was determined as per Raich & Potter (1995) and Luo et al. (2001) to estimate a Q 10 value associated with the relationship, defined as the relative change in CO 2 flux given a 10°C rise in temperature. This followed a nonlinear (exponential) relationship and applied the nls function as part of the base stats package within R, reporting an associated P value to describe the closeness of the defined relationship and data points. Further, because the goodnessof-fit r 2 metric is not as statistically robust for nonlinear relationships (Spiess & Neumeyer, 2010), these are not reported and instead a Q 10 value was calculated for each chamber individually, and therefore, a standard error could be applied to the average. These relationships were defined for both monthly averages and the full data set. This was done for two reasons: (i) to reduce bias where more measurements were taken in some certain crop phases and (ii) to assess how a few measurements at extreme temperatures influenced Q 10 values. To test which relationship (monthly vs. all data) best described the temperature sensitivity a generalized additive model (GAM) approach was applied using the gam formula in the MGCV package (Wood, 2011). The resulting nls relationships were compared using the ANOVA function as part of the base stats package within R.
To compare the difference in chamber GHG measurements between temporal groups (days, months, phases or years), repeated-measures analysis of variance (ANOVA) was used applying the aov function as part of the base stats package in R. Where the assumptions of ANOVAs could not be met, residuals were transformed using either the varPower or varExp function as described earlier. The transformed (modelled) data were then analysed using the lme function with chamber as the random effect. This provided significance levels (i.e. P-values) to the tests performed.

Climatic conditions and net ecosystem exchange
Continuous half-hourly measurements of air and soil temperature showed clear seasonal trends with annual means (9.60 and 9.55°C, respectively) in line with 30year averages (Fig. S1). While precipitation was distributed relatively evenly over the whole measurement period, on average March had the least rainfall (16.68 mm; 0.54 mm day À1 ) and November had the most (70.60 mm; 2.35 mm day À1 ). Both soil temperature and precipitation saw notable interannual variation with highs and lows in growth years 6 (9.86°C) and 5 (8.91°C) and in years 7 (818 mm) and 6 (405 mm), respectively (Table S2; Fig. S2). Mean NEE over the four full growing seasons was À678.08 AE 110.70 g CO 2 -C m À2 yr À1 with more days between frosts in 2010 leading to the greatest uptake during this year. The large standard deviation reflects the notable interannual variation.

Soil GHG emissions and environmental controls on soil respiration
Soil fluxes of CH 4 and N 2 O were largely negligible, with no discernible temporal trends and no clear relationships to environmental variables (Fig. 1). Using linear integration to cumulate average monthly fluxes to annual totals, CH 4 and N 2 O emissions were found to be the same weight, totalling 0.38 kg CH 4 -C ha À1 yr À1 and 0.38 kg N 2 O-N ha À1 yr À1 , respectively. In the case of N 2 O emissions, only the fluxes in June 2010 were significantly different from zero and therefore contributed largely to the cumulative annual average. Soil CO 2 emissions were significantly higher than those of CH 4 and N 2 O, contributing 3.00 AE 0.22 t CO 2 -C ha À1 yr À1 . Emissions throughout the year followed a clear seasonal trend with highest emissions during the crops growth phase when soil temperatures were warmer; the lowest emissions were seen during the dormant crop phase when temperatures were cooler ( Table 1). The climatic variables of temperature and precipitation explained the differences between years, with particularly warm and dry periods during measurements taken in June and September 2009 responsible for high cumulative totals in growth year 4. The highest single measurement (283 mg CO 2 -C m À2 h À1 ) was observed in September 2009 and the lowest (0.83 mg CO 2 -C m À2 h À1 ) in January 2013 (Fig. 2).
Using either all available data points or monthly averages, soil respiration correlated well with both soil temperature and soil moisture (GAM results for all correlations P < 0.01) (Fig. 3). Using nonlinear regressions for each block of chambers, mean Q 10 values and standard errors were derived using both monthly average data and the full data set (Table 2). In all cases soil respiration was most sensitive to temperature during the crop growth phase and least sensitive during the dormant crop phase, when average temperatures were highest and lowest, respectively. ANOVA results showed the uncertainty of these Q 10 values was lower (P = 0.009) when monthly averages were used in place of the full data sets.
Less than 5% of the variance observed for CH 4 or N 2 O fluxes was explained by any of the environmental variables studied (Table 3). However, the same variables explained far more variation in chamber CO 2 fluxes; soil temperature explained more than half of the variance seen in soil respiration throughout the 4-year measurement period.

Carbon and nitrogen stocks
The paired-site proxy used as a 'time-zero' indicated that there was no temporal difference (P > 0.05) in soil C or N stocks between 0-to 15-cm and 15-to 30-cm layers (Fig. 4). Soil C stocks were estimated to be 81. Annual inputs to the litter layer through crop senescence (not including harvesting inefficiency) decreased over time from 2.59 t dry biomass ha À1 in growth year 3 to 1.75 t dry biomass ha À1 in growth year 7. After heavily stunted growth during the first two years, all standing biomass was cut and left on the site in April 2008, estimated to be 3 t biomass ha À1 . From this point, litter inputs comprised both senesced leaves (green bars; Fig. 5) and residues from harvesting inefficiency (grey bars; Fig. 5). Considerable litter accumulation was observed between 2009 and 2013 (blue points; Fig. 5), suggesting a decomposition rate (k) slower than the rate of inputs. Using our measurements of the litter layer, we estimated a decomposition rate between those reported by Amougou et al. (2012) and Yamane & Sato (1975): k~0.63.
Both senesced and living Miscanthus biomass had similar C concentrations (Table 4). In contrast, N concentration in standing biomass almost halved between October (when senescence and nutrient translocation began) and March, and was reduced by a further 40 % in the litter layer (Table 4). Relatively little difference was seen in C concentration between stems and leaves, whereas N concentration was significantly different, resulting in C : N ratios of 206 and 56 for stems and leaves, respectively, in harvested biomass (Table 4). The mean oven-dried (0% moisture content) harvested yield was 6.07 t ha À1 yr À1 over the 5-year measurement period, equating to 2.85 t C ha À1 yr À1 (assuming 47% C concentration; Table 4); litter inputs were estimated as 2.69 t ha À1 yr À1 on average, equivalent to 1.24 t C ha À1 yr À1 (assuming 47% C concentration).

Life cycle GHG balance of Miscanthus vs. fossil fuels
When calculated over the predicted crop life cycle of 18 years, the total GHG balance from cradle to farm gate was a net removal of 441 t CO 2 -eq ha À1 (Table 5). Soil C stocks were assumed to remain constant (as this creates the most cautious scenario and no empirical data at the site suggest otherwise) and the litter layer unchanged for the remainder of the crop's lifetime following the measurement period. Both CH 4 and N 2 O emissions contributed very little to offsetting the net sequestration observed through NEE measurements. Cutting and baling the harvested biomass contributed the most to direct emissions but these were orders of magnitude lower than NEE measurements.  Fig. 3 Relationships between soil respiration and soil temperature (a and c) and soil moisture (b and d) using all available data points (a and b) and monthly average data (c and d) from measurements beneath a Miscanthus plantation in Lincolnshire, UK. Colours refer to crop phase: dormant (green), emergent (orange) and growth (purple). Regression analysis was used to fit an exponential relationship for soil temperature, reporting the associated P-values of how well the suggested relationship fit the measured data. Dashed vertical lines indicate 0 on plots where negative values were measured. Dotted horizontal lines are applied to aid comparison between top and bottom panels given that the scales differ.  Compared to the life cycles of coal and natural gas, Miscanthus had a substantially lower GHG intensity (Table 6). Further, the life cycle estimate of À1401 g CO 2-eq kWh À1 suggests noteworthy sequestration beyond offsetting the known emissions. Any GHG intensity associated with cradle-to-farm gate 'production' below À1525.03 g CO 2-eq kWh À1 would completely offset the emissions from transportation and combustion when using conventional power plants with conversion efficiency of 30% (Table 6). However, an important consideration in using GHG intensity as a comparison metric is that it does not account for the land area required to generate each unit of energy (kWh ha À1 ). Consequently, a higher yield at this site, or an improved conversion efficiency (e.g. 70% achieved by CHP generators), would lead to lower emissions per kWh but would not necessarily increase net sequestration per kWh (Table 6). For reference, using 1 t of Miscanthus biomass (at 20% moisture content; LHV = 14 MJ kg À1 ) for electricity generation produces 1167 kWh at 30% efficiency and 2722 kWh at 70% efficiency, while both emit 1722 kg CO 2 -eq through combustion (assuming 47% C concentration) (Eqn 3).

Discussion
This study addressed three main objectives: i) to quantify GHG emissions from a Miscanthus plantation and examine the influence of soil temperature and moisture on these emissions, ii) to examine the dynamics of litter and soil C stocks that define long-term sequestration and iii) to estimate the life cycle GHG intensity of electricity generation using Miscanthus harvested from this site, ultimately comparing this with conventional fossil fuels.

Net ecosystem exchange and soil GHG emissions
The annual net CO 2 flux, reported as NEE, was on average À24.85 t CO 2 ha À1 yr À1 (Table 5), despite low yields compared to other studies in similar climatic regions (Lewandowski et al., 2000;Christian et al., 2008). A trial in Illinois, USA, comparing Miscanthus with switchgrass (Panicum virgatum) and prairie grasslands reported a GHG balance of À20.31 t CO 2 ha À1 yr À1 for Miscanthus in its third year after establishment (Zeri et al., 2011), 14% lower than switchgrass (À17.78), 88% lower than prairie (À10.82) and 18% higher than our reported NEE. This Illinois Miscanthus plantation produced approximately 16 t dry biomass ha À1 in October of the third growth season, more than double the spring yield at our Lincolnshire site. Both studies emphasize the large sequestration potential of Miscanthus, despite annual harvests removing all aboveground biomass. While the negative NEE at our site implied considerable sequestration, soil respiration (10.99 t CO 2 ha À1 yr À1 ) offset a large portion and dominated the GHG flux at the soil surface. This value was within the same range as other Miscanthus plantations (Wanga et al., 2005;Behnke et al., 2012;Case et al., 2014), as well as other bioenergy crops: switchgrass (Panicum virgatum) (Frank et al., 2004;Lee et al., 2012), maize (Zea mays) (Rochette et al., 1999;Ding et al., 2007) and short rotation coppice (SRC) poplar (Populus spp.) (Verlinden et al., 2013).
In contrast to the CO 2 fluxes, both CH 4 and N 2 O made a negligible contribution to the GHG budget of the plantation over 4 years. That said, in June 2010 N 2 O emissions were an order of magnitude larger than all other months (Fig. 1). Soil N 2 O efflux is often very sporadic (Parkin, 1987;Dalal et al., 2003) and most commonly associated with rainfall events and rapid changes in water filled pore space (Dobbie et al., 1999). Consequently, rainfall events that occurred prior to measuring are likely to have influenced the high flux measured in June 2010, although this is unlikely to be the sole cause. To elucidate the drivers of this lone peak, more regular flux measurements are required to gauge the influence of explanatory variables. If these events are short bursts and occur more often than detected by our measurement schedule, the contribution of N 2 O to the overall GHG budget would be much larger due to the high GWP of N 2 O.
The Miscanthus plantation was shown to be a small source of CH 4 contradicting two previous studies at other sites (Toma et al., 2011;Gauder et al., 2012); however, spatial heterogeneity in soils is likely to cause variation between sites (Smith et al., 2000). While there are a number of factors which influence the processes that govern CH 4 and N 2 O efflux (e.g. disturbance, H€ utsch (2001)  the low trace GHG emissions seen in this study and reported elsewhere (e.g. Toma et al., 2011;Drewer et al., 2012;Gauder et al., 2012). It is worth noting that land use change to intensive management practices after Miscanthus propagation may stimulate rapid mineralization of labile nutrients (particularly C and N) that accumulated during the plantation's lifetime. Bold values refer to summed totals -both annually and over the full plantation lifetime.

Environmental drivers of soil respiration
Due to very low CH 4 and N 2 O fluxes, it is not possible to draw conclusions regarding the weak relationships observed between climatic variables and emissions. In contrast, soil respiration did vary significantly with season, closely following changes in soil temperature and crop phenology (Table 1; Fig. 2). This confirms results from other studies where largest CO 2 emissions were observed when temperatures and photosynthetic activity were greatest (Yazaki et al., 2004;Wanga et al., 2005;Gauder et al., 2012) and follows conventional understanding of both heterotrophic and autotrophic soil respiration (Ryan & Law, 2005;Tang et al., 2005). Soil respiration also varied interannually (4.22 to 2.24 t C ha À1 for growth years 4 and 7, respectively; Table 1) despite similar climatic conditions between years (Table S6). Yazaki et al. (2004) took similar measurements from a Miscanthus sinensis plantation in Japan, estimating much more consistent emissions between two years. While the average aboveground biomass was similar, annual soil respiration from the Japanese plantation was more than three times higher than ours (~14 t C ha À1 ). Additionally, in the same study the temperature sensitivity (Q 10 ) of total soil respiration varied between 2.7 and 3.1. This agrees well with the average Q 10 values calculated for our site (Table 2), despite the Japanese site having higher soil temperatures and not including Q 10 estimates between December and April (when they are likely to be lowest). The relatively low soil temperatures at our site, and their impact on soil respiration, may explain why the low productivity still creates a lower NEE than that of the higher yielding site in Illinois (Zeri et al., 2011); while C assimilation through photosynthesis in Illinois is considerably higher than in Lincolnshire, so is the annual mean air temperature (11.1 vs. 9.6°C) and, in particular, temperatures during the growing season. Consequently, soil respiration is likely to greatly offset the increased C sequestration through photosynthesis; while biomass production in Illinois is larger than that in Lincolnshire, the overall GHG balance of the Miscanthus plantation may be more favourable in the cooler climate.

Carbon and nitrogen stocks
Soil C and N stocks did not change over 4 years and when compared with a proxy for before Miscanthus was planted, stocks were still unchanged (Fig. 4). While this is consistent with some studies of Miscanthus (Zatta et al., 2014;Rowe et al., 2015), many others report increases in topsoil (0-30 cm) C stocks of more than 1 t C ha À1 yr À1 with prior land use and management practices playing a key role in the direction of change (Kahle et al., 2001;Dondini et al., 2009;Zimmermann et al., 2012;Poeplau & Don, 2014). There is a reasonable chance that topsoil C stocks were negatively impacted through disturbance of ploughing and planting, but were also enhanced by the addition of rhizomes and rapid fine root turnover as the plantation established itself. Indeed, Amougou et al. (2011) reported combined rhizome and root C input rates of 2.91 t ha À1 for the top 30 cm over the first three years after planting. These input rates are then expected to decline as the plantation ages; Richter et al. (2015) noted a combined C input rate of 1.43 t ha À1 for the top 100 cm over the first 14 years after planting (see Agostini et al. 2015 for a review of existing data on this topic). Aside from the lower yields noted at this Lincolnshire site, and therefore likely smaller belowground biomass pools, there is no clear reason why soil C stocks are not increasing over time. We hypothesize that at this site fresh C inputs may be stimulating (priming) the decomposition of existing soil C, therefore negating any C sequestration (Zatta et al., 2014). Testing this hypothesis would require the use of stable isotopes to trace the fate of native soil C and fresh C inputs in these crops.
N deficiency in the soil may also explain low C sequestration rates through limitation of decomposition and microbial activity (Hu et al., 2001;Craine et al., 2007). The C : N ratio of senesced Miscanthus biomass was between 70 and 120, and soil C : N was around 10 (Table 4; Fig. 4). These are high values for an arable crop, and therefore, a lack of N fertilizer may be a limiting factor in microbial decomposition (Anderson & Domsch, 1989). That said, these C : N ratios are within a normal range for Miscanthus plantations where soil C sequestration has been noted (Dondini et al., 2009;Amougou et al., 2011) and therefore cannot alone explain the lack of sequestration at this site. Additionally, other studies have observed similar accumulation rates of senesced biomass (2 t ha À1 yr À1 ; k~0.63) while also reporting increased soil C stocks (Yamane & Sato, 1975;Amougou et al., 2011Amougou et al., , 2012. In the absence of soil C sequestration at this site, the measured NEE of À6.78 t C ha À1 yr À1 is very low and requires an explanation for where C is being sequestered. Following biomass removal at harvest, C pools may remain in live belowground biomass, an increased O-horizon and in the soil organic matter (SOM) that was removed before calculating soil C stocks. When these additional pools are considered, À6.78 t C ha À1 yr À1 is not unrealistic: 2.85 t C ha À1 yr À1 was present in harvested biomass and 1.24 t C ha À1 yr À1 was added to the O-horizon through senescence and harvesting inefficiency (Fig. 5). This leaves 2.69 t C ha À1 yr À1 to be allocated to live belowground biomass, to soils below the measured topsoil (30 cm) and to SOM fractions, a realistic possibility given the recalcitrant nature of Miscanthus biomass (Amougou et al., 2011) and its characteristic deep-rooting (Neukirchen et al., 1999). Indeed, live and dead root biomass was estimated to be 4.46 t dry mass ha À1 in the top 30 cm of soils at this site and annual C inputs under Miscanthus can be substantial . It is also important to note that dissolved organic carbon and carbon lost through root exudation may contribute to this unquantified sink of soil carbon (Hromadko et al., 2010).

Comparative life cycle GHG budgets of Miscanthus
Miscanthus was calculated to remove 441 t CO 2 -eq ha À1 (over 18 years) from the atmosphere using a 'cradle-tofarm gate' analysis (Table 5). This compares well against a SRC willow plantation, grown for 23 years, removing 496 t CO 2 -eq ha À1 without consideration of soil GHG emissions (Heller et al., 2003). It is worth noting that while our method of linear integration to cumulate soil CO 2 emissions is robust, it may be less appropriate for N 2 O. Soil N 2 O emissions are spatially and temporally heterogeneous and as a result chamber measurements may not capture the true site-scale emission rates (Williams et al., 1992;Bouwman et al., 2002;Stehfest & Bouwman, 2006). This may have contributed towards the favourable cradle-to-farm gate GHG balance in comparison with other studies, where soil GHG emissions were modelled rather than measured (Brandão et al., 2011;Hamelin et al., 2012). While we acknowledge that the low temporal resolution of measurements may limit our ability to accurately quantify the contribution of N 2 O to the life cycle GHG budget, both this study and those previously published report low N 2 O emissions under Miscanthus (Toma et al., 2011;Drewer et al., 2012;Gauder et al., 2012). Higher resolution (both temporally and spatially) N 2 O measurements would reduce uncertainty and are needed to underpin the refinement of emission factors for use in LCAs. With respect to NEE, limiting gaps in NEE measurements would also improve the accuracy of field GHG emissions data for LCAs. The measurement gaps reported here were assumed to cause limited error because they occurred in winter when photosynthesis and GHG fluxes were low. Further, average annual values were derived from a full 48-month period. Ultimately, gaps during winter months are likely to have far smaller impact on annual NEE estimates than other factors such as interannual climatic variation (Massman & Lee, 2002;Baldocchi, 2014).
The life cycle GHG intensity of electricity generation using Miscanthus from this site is very low compared to that of electricity generated from coal or natural gas. While both fossil fuels are a net source of GHGs, the Miscanthus plantation was a noteworthy GHG sink, offsetting between 0.6 and 1.4 kg CO 2 -eq per kWh (Table 6). This range is very low compared to a similar study of Miscanthus grown in Canada (Sanscartier et al., 2014) where between 0.02 and 0.19 kg CO 2 -eq was offset per kWh, including soil C sequestration. However, GHG intensity (emissions per unit energy generated) does not account for the land area required to generate each kWha major concern when determining the sustainability of bioenergy crops (Dornburg et al., 2003;Rowe et al., 2009). At this site, each hectare is capable of producing 8372 kWh of electricity, assuming a combustion efficiency of 30% and an average annual yield of 7.18 ha À1 (20% moisture content). A higher yielding site with similar environmental characteristics may increase C sequestration through NEE but not necessarily enough to improve the GHG balance per kWh produced, especially if these higher yields come at a cost of increased emissions during production and growth through intensive management (e.g. fertilizer application or precision planting). A recent study comparing Miscanthus with maize and switchgrass in North America  drew similar conclusions to those described here: Miscanthus has the potential to produce energy at low, or even C-negative, GHG intensities. It is also important to recall that soil C sequestration can offset a significant portion of the emissions derived from generating electricity. Given a 30% combustion efficiency and 129.2 t ha À1 yield (18 years at Lincolnshire), an increase of 1 t C ha À1 yr À1 in soils would offset 438 g CO 2 -eq kWh À1 on a life cycle basis (Eqn 3, GHG site fixed at À3.66 t CO 2 -eq ha À1 yr À1 ). An increase of 1 t C ha À1 yr À1 in the top 30 cm is not unrealistic; at this site, Miscanthus inputs were previously shown to add 0.86 t C ha À1 yr À1 to the top 30 cm (Robertson et al., 2016) and Poeplau & Don (2014) saw an average increase of 1.68 AE 0.7 t C ha À1 yr À1 from a range of Miscanthus crops across Europe. The unchanged topsoil C stocks reported here, therefore, have important consequences for whether it is deemed a preferable alternative to conventional fossil fuels.
Due to minimal land management and fertilizer requirements , Miscanthus is often seen as an attractive option when land is unsuitable for conventional arable crops. However, policymakers still require more data to reliably assess its sustainability when used for bioenergy by combustion. As hypothesized, this study found CO 2 to dominate site GHG fluxes but noted substantially more sequestered than emitted over each year. Furthermore, despite relatively low yields and a lack of soil C sequestration, the crop studied here had a considerably lower GHG intensity than coal or natural gas when used for electricity generation. Additional research is required to elucidate why soil C stocks are not changing under this plantation (Zatta et al., 2014;Robertson et al., 2016) and future bioenergy sustainability studies should prioritize land use efficiency over GHG intensity comparisons. Nevertheless, this study demonstrates that even when yields are lower than many other sites due to climate or establishment issues, GHG benefits can still outweigh costs and contribute to climate change mitigation through the provision of low C renewable energy.

Supporting Information
Additional Supporting Information may be found online in the supporting information tab for this article: Figure S1 Daily soil and air temperature (°C) measured at the Miscanthus plantation in Lincolnshire, UK, between growth years 3 and 7. Half-hourly data were averaged to give daily points between 1 May 2008 and 10 March 2013. Figure S2 Daily soil temperature (°C) and precipitation (mm) measured at the Miscanthus plantation in Lincolnshire, UK, between growth years 3 and 7. Half-hourly data were averaged or summed for temperature and precipitation, respectively, between 20 August 2008 and 10 March 2013. Table S1 Soil characteristics at the Lincolnshire Miscanthus plantation estimated using 5 reps taken from each month between February 2009 and November 2010 (inclusive).  (2007-2013) and predicted (2014-2024) Miscanthus yields (assumed 20% moisture content throughout) at the Lincolnshire site estimated using one stable and one declining approach. Table S6 Average temperature (AE 1 SE) and cumulative precipitation and radiation measurements from continuous (half-hourly) data collected between growth years (March-February) 3 and 7 of a Miscanthus plantation in Lincolnshire, UK.