Biochar’s effect on soil nitrous oxide emissions from a maize field with lime-adjusted pH treatment

Introduction Conclusions References Tables Figures

Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | time, is highly vulnerable to the consequences of a changing climate (IPCC, 2014). With its 300 fold warming potential compared to CO 2 , nitrous oxide (N 2 O) from soil is a downside of the large productivity increase in agriculture, due to synthetic nitrogen fertiliser application. Reducing agricultural N 2 O emissions would reduce the GHG induced radiative forcing (IPCC, 2014), improve the stability of the stratospheric ozone layer 5 (Ravishankara et al., 2009) and reduce agriculture's energy intensity when achieved with a lower nitrogen fertiliser use (IAASTD, 2009). Biochar is produced by thermal decomposition of organic material in a low-oxygen environment, called pyrolysis. This stable charcoal-like material has the potential to contribute to the mitigation of climate change by increasing soil carbon (C) (Lehmann, 10 2007; Woolf et al., 2010;Lal et al., 2011). In addition, biochar can increase crop yields (Jeffery et al., 2011;Biederman and Harpole, 2013;Crane-Droesch et al., 2013) and reduce water stress, which helps to adapt to climate change (Mulcahy et al., 2013). Its application to soils that have a small cation exchange capacity and low organic carbon content is associated with higher crop yields (Crane-Droesch et al., 2013) with 15 an overall mean response of 10 % (Jeffery et al., 2011).
Biochar also controls nitrogen (N) cycling (Clough et al., 2013). Biochar can reduce N leaching (Steiner et al., 2008;Güereña et al., 2013) and soil-borne N-containing GHG (van Zwieten et al., 2015). Especially nitrous oxide (N 2 O) emissions from soil are reduced on average by 54 % in lab studies and 28 % in field measurements (Cayuela 20 et al., 2015). In field situations, N 2 O reduction effects are typically difficult to verify because of less uniform conditions and a large spatial and temporal variability of fluxes (Felber et al., 2013;Schimmelpfennig et al., 2014). A few field experiments indicated an increase in N 2 O (e.g., Verhoeven and Six, 2014;Liu et al., 2014), many showed no significant effects (Angst et al., 2014;Karhu et al., 2011;Scheer et al., 2011;Suddick 25 and Six, 2013; Anderson et al., 2014) while other studies indicated decreasing N 2 O emissions (e.g., Felber et al., 2013;van Zwieten et al., 2010;Taghizadeh-Toosi et al., 2011;Zhang et al., 2010;Case et al., 2014). Only few studies with biochar have looked  (Verhoeven and Six, 2014), hence there is a large uncertainty about longer term effects of biochar addition. Biochars are often alkaline and therefore increase soil pH after application (Joseph et al., 2010). Denitrifying bacterial communities have the potential to increase their N 2 O-reducing activity with increasing pH, which may reduce N 2 O emissions from soils 5 (Cavigelli and Robertson, 2001;Simek and Cooper, 2002;Čuhel et al., 2010). Some authors suggest that the elevated soil pH is responsible for reduced N 2 O emissions following biochar application through increased activity of N 2 O reducing bacteria (van Zwieten et al., 2010;Zheng et al., 2012). In contrast, Yanai et al. (2007) argue that the suppression of N 2 O emissions by biochar is not through increased N 2 O reduction ac-10 tivity because biochar ash also increases soil pH but does not reduce N 2 O emissions. Cayuela et al. (2013) showed that biochar's acid buffer capacity was a more important factor in denitrification than the pH shift in soil. There are indications that biochar enhances nosZ expression, the gene responsible for the transcription of the N 2 O reductase in denitrifying microorganisms (Harter et al., 2014;Van Zwieten et al., 2014).

15
This could be a mechanistic link to the observed reduction in N 2 O emissions through biochar increasing soil pH and microbial activity. In contrast, under conditions favouring nitrification and not being as sensitive to pH as total denitrification, biochar addition increased N 2 O emissions in the lab (Sánchez-García et al., 2014) and possibly in the field (Verhoeven and Six, 2014). 20 In this study, we test (i) whether N 2 O emissions are reduced following the application of biochar to soil of a temperate maize cropping system and (ii) whether this possible reduction in N 2 O emissions is due to an increase in pH. The latter was tested by a treatment where limestone was added to increase soil pH to the same level as that from the addition of 20 t ha −1 biochar. N 2 O emissions and maize yield were quantified 25 during one growing season in the field. Group WRB, 2006) it is a Eutric Mollic Gleysol (Drainic). The untreated soil has a pH of 6.3 in water (1 : 2.5 w/v), total organic carbon content of 26.2 g kg −1 , total N of 0.29 g kg −1 and bulk density of 1.3 g cm −3 .

Biochar
Several biochars were screened in advance to pick one with a high liming capacity and 15 with properties in agreement to the guidelines for polycyclic aromatic hydrocarbons (PAHs), C-and N-content of the European Biochar Certificate (EBC, 2012). The chosen biochar was produced in a Pyreg reactor (Pyreg GmbH, Dörth, Germany) by Verora in Edlibach ZG, Switzerland in late 2013 (see chapter 30, case study 2 in Lehmann and Joseph, 2015). Pyreg reactors use slow pyrolysis in a continuous system with 20 an average residence time of circa 25 min and a peak temperature of approximately 650 • C. The feedstock was green waste mainly from tree pruning. The biochar has the following properties: 64.9 % total C; 62.1 % Corg, pH 9. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | the time of application was 12 %. Biochar was sieved < 3 mm shortly before it was spread on the field.

Experimental setup
Three different treatments were introduced; 20 t ha −1 biochar, control without additions and a limestone treatment to increase the soil pH to the same level as with biochar. The 5 field was split into 3 × 3 plots with a size of 2 by 3 m (6 m 2 per plot and 3 replicates for each treatment). One meter buffer zones were established between plots on all sides. The 3 different treatments were arranged in a randomized complete block design with the 3 × 3 grid accounting for spatial variability. The whole field, including the buffer zones, were planted with maize (zea mays). Initial pH values were not different among 10 treatment plots (see pH measurement in January on Fig. 2).

Field management
The field was ploughed in autumn 2013 after the maize harvest. In January 2014, 20 t ha −1 biochar and 2 t ha −1 limestone were spread on the wet, ploughed field surface.
Freshly applied biochar was gently mixed with the first 1-3 cm of soil by hand at the 15 same time. In mid-February 2014, the automated GHG chamber system was installed and in March the field was harrowed by a rototiller to a depth of circa 15 cm. The chamber frames were reset into the soil again and Decagon TE5 temperature and humidity sensors (Decagon Devices Inc., Pullman Wa, USA) were placed at a depth of 8 cm in the centre of each plot. 20 In May, potassium (K) and phosphorus (P) fertiliser was applied at a rate of 41.4 and 132 kg K ha was not in the same range as the biochar plots. Maize (Padrino from KWS SAAT AG, Einbeck, Germany) was sown on the 8 May with 0.14 m distance within rows that were 0.6 m apart from each other. For plant protection only one herbicide application was conducted on the 19 June with 1 L ha −1 Dasul (Syngenta, Basel, Switzerland) 1 L ha −1 Mikado (Bayer CropScience, Germany) and 1 kg ha −1 Andil (Omya AG, Switzerland).

5
Despite manual weeding and herbicides a considerable amount of weeds emerged. Plots were harvested on the 13th of October.

Nitrous oxide measurement
N 2 O and CO 2 emissions were measured with static chambers of a fully automated measurement system (Flechard et al., 2005;Felber et al., 2013) consisting of nine stainless steel chambers (30 × 30 × 25 cm). These chambers were placed on PVC frames inserted 3 cm deep into soil. Two frames were placed on each plot at a similar distance to the plot borders. These positions were moved three times during the growing season to obtain a better spatial representation of each plot. After maize had been sown, the chamber positions were between rows and no vegetation was grown 15 within the chamber frame. Each of the 9 chamber lids were automatically closed and opened sequentially (over a period of 3.5 h) allowing N 2 O and CO 2 to accumulate in the chamber headspace for 15 min. Chamber headspace air was circulated (1 L min −1 air flow) through an inlet and outlet line from each chamber through polyamide tubes (4 mm I.D.) to the analytical system and back to the chamber headspace continuously 20 after sample analysis. The analytical and chamber control instruments were installed in a nearby field cabin under temperature controlled air conditioning. N 2 O concentrations were continuously measured and stored every minute using a gas filter correlation technique (TEI Model 46C, Thermo Environmental Instruments Inc., Sunnyvale, CA, USA Hence a temperature correction factor was applied to the raw data from a regression of the device temperature with data during calibrations in May. N 2 O and CO 2 fluxes from soil were calculated from the continuous concentration measurement (resolution 1 per min) when chamber lids were closed. Data from the first 3 min of the total 15 min closure time were omitted from the flux calculation to re-5 move signal noise due to gas exchange from the system during chamber switching and closing (Felber et al., 2013). The same flux estimation procedure (R-script by R. Fuss on bitbucket.org, see Fuss, 2015) was used as in Leiber-Sauheitl et al. (2014). It is a modification of the HMR package (Pedersen et al., 2010) that chooses between exponential curvature for non-linear chamber behaviour (Hutchinson-Mosier regres-10 sion) and robust linear regression (Huber and Ronchetti, 1981). The exponential HMR scheme considers non-linear concentration increase in the chamber due to a possibly decreasing concentration gradient, chamber leakage and lateral gas transport. Robust linear regressions provide a more reliable flux estimate for low fluxes when there is a lot of variation due to limited measurement precision and outliers. The resulting flux 15 estimates from this procedure were then filtered for implausible large N 2 O uptake by soil. N 2 O fluxes smaller than −50 ng-N 2 O m −2 s −1 (Neftel et al., 2010) were removed as well as data associated with a likely invalid chamber functioning (i.e. frozen lids) when CO 2 flux < −0.5 µmol m −2 s −1 (Felber et al., 2013). In total 302 CO 2 and 351 N 2 O data points from the entire dataset (14 068 points) were rejected. 20

Yield
The yield was separated into grain (kernels) and plant material. Cobs were threshed and dried whereas the plants were weighed freshly on the field, chaffed and a subsample was then dried to measure water content and for further plant nutrient analysis. From both plant and grain, dry matter total N and P were measured (FAL, 1996).

Soil sampling and analysis
Soil samples for pH, ammonium (NH + 4 ) and nitrate (NO − 3 ) measurements were taken on the 31 January, 31 March, 26 May, 16 June and 4 September 2014. At each sampling, five randomly distributed soil cores per plot were taken (0-10 cm) and pooled. Soil pH was determined in moist soil samples using water at a ratio of 1 : 2.5 w/v and 5 measured with a PH100 ExStik pH meter (Extech Instruments Corp., Nashua, NH, USA). Soil bulk density was measured on the 27 June at a depth of 3-8 cm using 100 cm 3 steel cores, 3 per plot.
For soil NO − 3 and NH + 4 concentrations, 20 g of moist soil were mixed with 100 mL 0.01 M CaCl 2 solution. The suspension was shaken for 30 min, filtered and then anal-10 ysed by segmented flow injection analysis on a SKALAR SANplus analyser (Skalar Analytical B.V., Breda, the Netherlands).

Statistical analysis
The obtained fluxes from the automated chamber system were aggregated to 8 h means producing a regular, smoothed dataset. The system was able to measure each 15 chamber three times for every 11 h calibration cycle during regular operations, hence on average 2.2 measurements for each chamber were included in each a 8 h mean. Still missing values after this aggregation step were linearly interpolated for each chamber. Treatment averages and standard deviations were calculated from the 3 chambers on the replicated plots. 20 Statistical analyses were performed with R (version 3.0.1, The R Project, 2014). Significance level was chosen at p < 0.05 for all procedures, unless indicated otherwise. Significant treatment effects for cumulated fluxes were determined using ANOVA from rbase package (treatments: control, biochar and lime; n = 3). Bartlett test of homogeneity of variances showed conflicting ANOVA assumptions for the cumulative fluxes.

25
This could be solved by log transformation of the flux data. In addition, a generalized least squares model (GLS) was constructed with weekly cumulated N 2 O emissions as dependent variable, and weekly averages of soil volumetric water content (VWC) and the treatments (control, biochar, lime) as explanatory variables. A restricted maximum likelihood generalised linear model from nlme R package was used to calculate the GLS.

Meteorological data on the field
The year started with above average temperatures and low rainfall (Fig. 1). End of May to June was dry with high temperatures being on average for Switzerland 1.5 • C above the 1981-2010 norm (Meteoswiss, 2015). The soil's volumetric water content fell to 10 circa 20 %, inducing high water stress on the young maize seedlings. The lack of soil moisture presumably hampered the dilution of the first application of 40 kg ha −1 N in the soil solution. Along with the 2nd N fertilisation the field was therefore irrigated with 33 mm water (shown as green bar in the precipitation dataset). The summer months following (July and August) were rather cold and wet with daily mean air temperatures 15 below 20 • C (Meteoswiss, 2015).
The GLS model indicated a significant, treatment specific (p = 0.0202) effect of weekly mean soil VWC on weekly cumulated N 2 O fluxes (p = 0.0034). Biochar plots had significantly higher soil water content than lime and control plots (p < 0.001). However, there is no interaction between treatment and VWC on a weekly basis (p = 0.542).

Soil pH and nitrogen
Soil pH increased with limestone and biochar addition in medium terms by circa 0.4 pH units (Fig. 2). The initial soil pH was on average 6.3 and not different among treatments. Following biochar application soil pH increased to up to 7.4 whereas with addition of limestone soil pH increased to up to 6.9 (averages across replicates). The pH sharply Introduction Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | decreased after the initial peak, especially in those two liming plots, which were treated with another 1 t ha −1 in May. Soil pH of biochar and lime treatments were not significantly different at any sampling time, whereas soil pH of the control treatment was systematically below that of the amended soils. Mean soil bulk density was not statistically different between treatments (1.31 g cm −3 5 in the control, 1.29 g cm −3 in biochar and 1.36 g cm −3 in the liming treatment). Soil mineral N was not statistically different between treatments (Tables 1 and 2).

N 2 O fluxes
Emissions were characterized by peak events, particularly in summer, and by background emissions in spring and autumn (Fig. 3). Main emissions occurred after the second fertilisation event of 80 kg-N ha −1 around early August. Afterwards, there were only emissions from one of the lime plots but almost none until the end of October from all the other plots. This also corresponds to the low amounts of available soil N, indicating that the plants had taken up most of it. All treatments revealed similar temporal N 2 O emission dynamics but the height of the peaks differed. During peak events 15 emissions from the biochar treatment were often lower than those from the other treatments, especially compared to the control. This resulted in an increasing difference in cumulative fluxes (Fig. 4)  that were directly influenced by the N-fertiliser applied (between 26 May and 13 August = approx. 3 months) and subtract half of the cumulative emissions from the residual period measured (approx. 6 months). This resulted in IPCC emission factors of 0.58 % for biochar, 1.28 % for control and 1.25 % for the lime treatment. 10 Maize yields were not significantly different between treatments, for both grain and plant dry matter (Fig. 5). Nitrogen and P uptake did not differ among treatments (Figs. 6 and 7). same authors under field conditions (28 ± 16 %). In our temperate maize field, N 2 O emissions thus decreased with biochar addition as much as they have been shown to be reduced under controlled lab conditions. Our results show no a decrease in N 2 O emissions when limestone is used to increase the soil pH to the same level as that with biochar. This finding does not support 5 the hypothesis that biochar's N 2 O reduction effect is solely due to a geochemical manipulation of soil pH. However, it must be considered that the large variability among the three replicates hampers the power of this conclusion. The high variability solely in the liming treatment might be due to additional lime application to the field in May 2014 and the high spatial-temporal variability of that soil property in general. The two replicates that received additional limestone were the ones that emitted more N 2 O than the other plot. Hence, instead of reducing emissions by increasing the pH, the additional limestone application could have provoked local arbitrary disturbance to soil chemistry leading to emission hotspots. To determine the biochar effect on N 2 O emissions, we therefore also compared only the biochar and control treatments; the cumulative emis-15 sions in the biochar amended plots are significantly lower (by 53 %) than in the control treatment.

Maize yields and plant growth
The GLS model shows that not only treatment but also water content affects soil N 2 O emissions. However, the mechanism behind the overall negative feedback of VWC on N 2 O emissions (i.e. higher VWC leads to lower emissions) can not be derived from our 20 data. Biochar effects on soil physical properties have been shown to increase waterholding capacity, reduce bulk density and increase soil sub-nanopore surface together with a 92 % decrease in N 2 O emissions (Peake et al., 2014;Mukherjee et al., 2014). This suggests that increased soil aeration by biochar dominates the effect of increased water content and hence does not favour denitrification (van Zwieten et al., 2010).   -Toosi et al., 2011;Liu et al., 2012). A number of studies found no significant effect of biochar addition in the field (Schimmelpfennig et al., 2014;Angst et al., 2014;Scheer et al., 2011;Karhu et al., 2011;Anderson et al., 2014). Often the much higher variability in the field and the low number of replications make it difficult to reproduce reduction effects observed in laboratory studies. In particular, Angst et al. (2014) found 5 no significant difference but there was a tendency for lower emissions with biochar addition. However there are also studies that showed increased emissions from biochar application in the field (Verhoeven and Six, 2014;Shen et al., 2014). Sánchez-García et al. (2014) found that biochar increases soil N 2 O emissions produced by nitrification-mediated pathways. In our study, the water content ( Fig. 1) was 10 high during periods of high emissions and suggesting that during periods of high water content denitrification dominates the N 2 O production in soil. The high emissions were thus often triggered by large precipitation events. There are many indications from lab experiments that biochar can reduce N 2 O emissions in denitrifying conditions at high water content (Felber et al., 2013;Harter et al., 2014;Singh et al., 2010;Yanai et al., 15 2007). Under denitrification conditions, the pH exerts control over the N 2 O : N 2 ratio (Simek and Cooper, 2002). Various studies have suggested that an elevated soil pH is responsible for reduced N 2 O emissions following biochar application through increased activity of N 2 O reducing bacteria (van Zwieten et al., 2010;Zheng et al., 2012). In contrast, Yanai et al. (2007) argued that the suppression of N 2 O emissions by charcoal is 20 not due to increased N 2 O reduction activity because biochar ash increased pH to the same degree as biochar but did not reduce N 2 O emissions. Also Cayuela et al. (2013) found no N 2 O mitigation when soil pH was increased to the same level as biochar did but with CaCO 3 addition. They also showed that biochar's buffer capacity but not biochar pH was highly correlated with lower N 2 O emissions compared to pH-adjusted 25 biochars (Cayuela et al., 2013). In our case, we used a biochar with rather high liming capacity (17.2 % CaCO 3 ) and pH (9.8). We can confirm that with this kind of biochar N 2 O emissions can effectively be reduced also in real field conditions, although the high variability in the pH adjusted control does not allow us to reject the hypothesis of  Van Zwieten et al., 2014). Some authors relate this enhancement of N 2 O reducing bacteria to biochar's redox activity that facilitates electron shuttling for the sensitive process of N 2 O reduction (Kappler et al., 2014;Cayuela et al., 2013). This shuttling might be the connection between reduced N 2 O emissions and low H : Corg ratios (Cayuela et al., 2015) in biochar that 10 refers to condensed aromatic structures and its quinone/hydroquinone moieties being electro-active by allowing electron transfer across conjugated pi-electron systems (Klüpfel et al., 2014). Such high electro-catalytic activity has also been shown in Ndoped C nanotube arrays (Gong et al., 2009). Hence, in contrast to a promotion of microbial N 2 O reduction, there is also the possibility that biochar abiotically reduces 15 N 2 O through its electrocatalytic abilities represented by a high aromaticity with low H : Corg ratios. Indeed, this is one of the various abiotic mechanisms that reduce N 2 O emissions suggested by Van Zwieten et al. (2015).

Yield and nutrients
In our experiment, grain yield and plant biomass production were not increased by show clear effects within the first year of application yet. Our data is also in agreement with Jay et al. (2015) who showed that biochar had no effect on harvest yield of different crops after a single rotational application (20 and 50 t ha −1 ) in a sandy loam under intensive management. Nitrogen uptake was not changed by biochar or liming. Although there was no signif-5 icant difference in P uptake between the treatments, green plant material from biochartreated plots tended to have higher uptake then the control (+100 % increase). Vanek and Lehmann (2014) showed significant increase in P availability through enhanced interactions between biochar and arbuscular mycorrhizas.