Split N application and DMP based nitrification inhibitors mitigate N2O losses in a soil cropped with winter wheat

Nitrogen (N) fertilization to crops might lead to formation and release of reactive N—e.g. nitrate, ammonium, ammonia, nitrous oxide (N2O) —, contributing to eutrophication, atmospheric pollution, and climate change. Use of nitrification inhibitors and splitting of N fertilizer may reduce the N2O emission from arable soils cropped with winter wheat. We tested different N fertilizers treated with 3,4-dimethylpyrazol phosphate (DMPP) and 3,4-dimethylpyrazol succinic acid (DMPSA) by applying 180 kg N ha−1 in different N splitting strategies in a full annual field experiment on a loamy soil in Southwest Germany. A threefold split fertilization led to an emission of 2.3 kg N2O–N ha−1 a−1 (corresponding to a reduction of 19%) compared to a single application of ammonium sulphate nitrate (ASN) (p = 0.07). A single application rate of ASN with DMPP resulted in an emission of 1.9 kg N2O–N ha−1 a−1 and reduced N2O emissions from an ASN treatment without NI by 33%. Calcium ammonium nitrate (CAN) with DMPSA reduced N2O emissions during the vegetation period by 38% compared to CAN without a nitrification inhibitor, but this was offset by high emissions after harvest, which was driven by soil tillage with an annual reduction of 26% (CAN: 2.9 kg N2O–N ha−1 a−1; CAN + DMPSA: 2.1 kg N2O–N ha−1 a−1; p = 0.11). Among our tested treatments, a twofold split application of ASN with DMPP efficiently reduced N2O emissions and maintained grain yield when compared to the traditional system with threefold application without nitrification inhibitor. Despite resulting in lower protein contents in the twofold split application, this treatment should be further investigated as a potential compromise between wheat yield and quality optimization and climate protection.


Introduction
Nitrous oxide (N 2 O) is a climate-relevant trace gas which also contributes to the depletion of stratospheric ozone (Ravishankara et al. 2009;IPCC 2021). More than half of the anthropogenic N 2 O emissions are released from agricultural soils (Canadell et al. 2021). It is generally agreed that N 2 O production in agricultural soils is mainly controlled by the microbiological processes of nitrification and denitrification (Robertson and Groffman 2015), whereas the contribution of other processes such as autotrophic Abstract Nitrogen (N) fertilization to crops might lead to formation and release of reactive N-e.g. nitrate, ammonium, ammonia, nitrous oxide (N 2 O) -, contributing to eutrophication, atmospheric pollution, and climate change. Use of nitrification inhibitors and splitting of N fertilizer may reduce the N 2 O emission from arable soils cropped with winter wheat. We tested different N fertilizers treated with 3,4-dimethylpyrazol phosphate (DMPP) and 3,4-dimethylpyrazol succinic acid (DMPSA) by applying 180 kg N ha −1 in different N splitting strategies in a full annual field experiment on a loamy soil in Southwest Germany. A threefold split fertilization led to an emission of 2.3 kg N 2 O-N ha −1 a −1 (corresponding to a reduction of 19%) compared to a single application of ammonium sulphate nitrate (ASN) (p = 0.07). A single application rate of ASN with DMPP resulted in an emission of 1.9 kg N 2 O-N ha −1 a −1 and reduced N 2 O emissions from an ASN treatment without NI by 33%. Calcium ammonium nitrate (CAN) with DMPSA reduced N 2 O emissions during nitrifier-denitrification to the release of N 2 O is currently under discussion (Wrage-Mönnig et al. 2018).
The use of nitrification inhibitors (NIs) can raise fertilizer N use efficiency (NUE) in agriculture and thus reduce N surpluses. Commercially available NIs are substances capable of retarding the first step of nitrification-the oxidation of ammonia (NH 3 ) to hydroxylamine (NH 2 OH). Besides the delay of nitrification, this also results in a reduction of the availability for further nitrite (NO 2 − ) oxidation to nitrate (NO 3 − ), thus lowering substrate supply for denitrification (Ruser and Schulz 2015). The use of NIs with ammonium (NH 4 + )-based fertilizers is supposed to keep NH 4 + for a longer period in its reduced form; NH 4 + is then mainly adsorbed to negatively charged soil particles in upper soil layers for a longer period, reducing NO 3 − leaching losses. Another way of increasing NUE in wheat (Triticum aestivum L.) production is the splitting of N fertilization. Traditional N fertilization strategies for wheat comprise a splitting of fertilizer application in order to adapt N supply to the physiological needs of wheat plants. Wheat breeding schemes in the last decades have developed more flexible cultivars, which are less prone to yield losses through stress events and show a higher influence of N translocation during the grain-filling period than older cultivars (Makary et al. 2020). Due to these traits, wheat N uptake and N utilization of current cultivars is much higher when compared to older cultivars, and new fertilization strategies adapted to regional climatic and soil conditions must be developed. In a series of field experiments on medium to heavy textured soils in South Germany, Schulz et al. (2015) found no differences in grain yield or crude protein content when N fertilization was applied in one, two, or three application rates and soil mineral N (N min ) contents did not differ after harvest. Since splitting can lead to lower soil N min contents throughout the vegetation period (Arregui and Quemada 2006;Guardia et al. 2018)-and since soil N min often correlates with N 2 O release (Granli and Bøckman 1994)-lower N 2 O emissions can be expected when N fertilizer is split.
3,4-dimethylpyrazole phosphate (DMPP) is a widely used NI in Europe, which releases 3,4-dimethylpyrazole (DMP) as active compound. Because of its chemical characteristics, DMPP cannot be sprayed on calcium ammonium nitrate (CAN), the most frequently used N fertilizer in Western and Central Europe in 2018 (23% of the N fertilizer market; IFA 2021). 3,4-dimethylpyrazole succinic acid (DMPSA) is a novel NI that also provides DMP as active inhibitor after the microbial degradation of succinic acid (Pacholski et al. 2016). Because of its non-polar chemical nature, DMPSA can be sprayed on CAN, increasing the scope of applicability of DMP (Pacholski et al. 2016). As CAN contains a higher portion of the highly mobile NO 3 − -N than ASN, synchronization of an early high N demand of wheat and N availability right after fertilizer application can be better achieved when compared to ASN. Simultaneously, the NH 4 + -N is stabilized and thus prevented from leaching.
Since weather and soil conditions as well as the presence of fresh organic matter in soil can greatly influence soil redox potential and therefore potential nitrification and denitrification, it seems rather impossible to predict N 2 O emissions induced by different fertilizer types. For example, Lebender et al. (2014a) reported similar N 2 O losses after application of ammonium sulphate (AS) and CAN at two sites and higher losses in the AS treatment at a third site with different N 2 O emission patterns for the growth and post-harvest period.
In their meta-analysis on the effect of NIs on soil N 2 O emission, Ruser and Schulz (2015) calculated a 35% emission reduction as compared to a conventionally fertilized control treatment without NI. This was the mean N 2 O reduction over all NIs tested. For DMPP, they reported a 38% to 40% reduction. So far only few studies investigated the effect of DMPSA on N 2 O release. Under humid Mediterranean conditions, N 2 O emission from wheat fields was reduced by DMPSA (Huérfano et al. 2016). In an incubation experiment with DMPP and DMPSA, Torralbo et al. (2017) reported similar N 2 O reduction for both NIs.
Most of the field studies investigating DMPSA were conducted under Mediterranean conditions in irrigated systems with mild and rainy winters. In contrast, winter wheat production in South Germany is rainfed, although soils may dry very strongly in the summer months. Rewetting of dry soil in this period after heavy rainfall as well as thawing of frozen soil in winter were shown to induce N 2 O bursts highly relevant for annual balances Guzman-Bustamante et al. 2019). The latter can be significant in the context of NI application because some studies showed N 2 O reduction after DMPP usage a long time after harvest in winter (Pfab et al. 2012;Guzman-Bustamante et al. 2019).
We aimed to quantify the effect of different N splitting strategies and of the NIs DMPSA and DMPP on N 2 O emissions, yield, and N utilization in winter wheat under Southern Germany conditions. We hypothesized that (i) N 2 O emissions from a winter wheat field can be decreased in a conventional N fertilization system with three N application rates when compared to a single N application since temporally high mineral N availability which serves as substrate for N 2 O production is minimized. (ii) We also assumed lower N min concentrations that serve as substrate for N 2 O production in the split NI fertilization treatments. Consequently, annual N 2 O emissions can be even further mitigated than a single NI application without any decrease in crop yield or protein concentration; lastly, (iii) the NI DMPSA shows similar N 2 O reduction under field conditions compared to DMPP, due to the same active compound (DMP).

Field experiment
From a fully randomized block experiment (Guzman-Bustamante et al. 2019) located at the experimental station of the University of Hohenheim "Heidfeldhof", in Stuttgart, Germany (48 • 42′ 59″ N; 9 • 11′ 42″ E), 24 plots of 3 m x 5 m were selected. An overview of climatic conditions and soil properties is given in the supplementary online material. Plots were divided into a sampling and a harvest subplot (1.5 m x 5 m each). Gas and soil samples were taken from the sampling subplot, while the harvest subplot was used for grain and straw yield determination as well as for plant analysis (C and N). The winter wheat variety "Schamane" was sown 6 October 2011 after winter wheat as previous crop. Total fertilizer amount was 180 kg N ha −1 , calculated according to the German Fertilizer Ordinance ("good agricultural practice", DüV, 2006 Flessa et al. (2002). N 2 O fluxes were determined using the closed chamber method, for which circular PVC bases with an inner diameter of 0.3 m and a height of 11 mm were installed at a depth of ca. 70 mm in the middle of the sampling subplot. In order to account for the growth of the wheat plants inside of the soil rings, additional PVC extensions of 0.3 or 0.6 m height were used during gas sampling in the vegetation period. Closed chambers and extensions were fitted on the rings only when measurements took place. A closer description of the dark chambers fitted with a vent and identical in construction to our chambers was provided by .
During each gas sampling, four gas samples were taken out of the chamber's atmosphere with evacuated vials (22.4 mL) through a double cannula inserted into a sampling port with a septum in the chamber`s top at time intervals of 15 min. N 2 O gas concentration in the vials was measured with a gas chromatograph (5890 series II, Hewlett Packard) equipped with a 63 Ni electron capture detector (ECD) and an autosampler (HS40, Perkin Elmer). A linear regression (concentration enrichment over time) was used in order to calculate N 2 O fluxes as described by .
Air temperature and precipitation data was retrieved from weather station "Hohenheim" located 600 m south from the experimental site (Landwirtschaftliches Technologiezentrum Augustenberg 2015).

Soil measurements
Soil samples were taken weekly from a composite sample of eight soil cores (0.3 m depth and 14 mm diameter) per treatment in the sampling subplot outside of the chamber base ring. Soil samples were kept cold in the field and frozen after field sampling until extraction in the lab. 40 g of soil were extracted with 160 mL of 0.5 M K 2 SO 4 solution for one hour to determine N min concentration. NO 3 − and NH 4 + concentrations in the extracts were measured with a flow injection analyser (3 QuAAtro.AQ2.AACE, SEAL Analytical, UK). Soil moisture was calculated gravimetrically after drying the samples at 105 °C for at least 24 h. Water-filled pore space (WFPS) was calculated after Ruser et al. (1998) using the mean measured bulk density (1.25 Mg m −3 ) in the A p -horizon during our experimental period.
In order to assess the transport of N min in deeper soil layers, soil in three depths (0-0.3; 0.3-0.6 and 0.6-0.9 m) was sampled at three dates: before fertilization (6 March 2012), after harvest (8 August 2012), and at the end of the experiment (14 March 2013). At the first date, samples were taken as a composite for each treatment; at the second and third dates, samples were taken separately in each plot. For calculation of N min amounts, we assumed a bulk density of 1.5 Mg m −3 for the 0.3-0.6 m soil layer and 1.6 Mg m −3 for the 0.6-0.9 m layer.
The NH 4 -N/NO 3 -N ratio was calculated in order to follow inhibitory effect of treatments with NI.

Yield and plant analysis
All measurements on yield and yield components took place in the harvest subplot. Spike number per m 2 was calculated by counting the wheat spikes in a circular area of 0.6 m diameter. Wheat grain was harvested using a plot harvester. Straw and grain samples were taken for each subplot. Samples were dried for 48 h at 60 °C and ground using a cutting mill. C-and N-analyses were conducted with an elemental analyser (vario MAX CN, Elementar Analysensysteme, Hanau, Germany). Thousand grain mass (TGM) was determined gravimetrically after weighting 100 grain subsamples (n = 3) counted by a seed counter (Contador, Pfeuffer GmbH, Kitzingen, Germany). N surplus-the balance between N fertilizer input and N removal through N in the harvest-was calculated subtracting grain-N from fertilizer-N.

Seasonal and annual N 2 O emission and statistical analysis
Cumulative N 2 O emissions were calculated using a step function, i.e. the flux at a given date was assumed to be constant until the next sampling date . This was done for each "Season", which represents the experimental time interval vegetation period (6 March-9 August 2012), tillage (10 August-29 November 2012) and winter (30 November 2012-21 March 2013, and for the whole experimental year. Statistical analyses were performed with SAS (SAS Institute Inc., Cary, NC, USA). For N 2 O fluxes, a repeated measures model was implemented using PROC MIXED with block, season, weekly dates (nested in season) and treatments as fixed effects, weekly date as repeated term with plot as subject and season as grouping variable. A spatial power correlation matrix was used in order to avoid serial autocorrelation and to consider differing sampling dates. For a better distribution of residuals, N 2 O fluxes were transformed using the boxcox SAS Macro (Box and Cox 1964;Piepho 2017).
The effect of treatments and seasons on cumulative emissions was assessed using a repeated measures model with block, season and treatments as fixed effects, and season as repeated term, with plot as subject. An autoregressive correlation matrix was used. Effect of treatments on annual emissions and yield parameters were calculated with linear models. A logarithmic transformation was used when necessary to improve residual distribution.
The effect of different variables (soil NH 4 , soil NO 3 , soil temperature, WFPS and ΔWFPS) on N 2 O fluxes was calculated using PROC GLMSELECT and PROC GLM with the Akaike information criterion (AIC value) as selection parameter. To improve residual distribution, all variables were log transformed. The interaction between N min and use of NI was also assessed by including NI as a dummy variable. The relative importance of variables was calculated dividing the type I sum of squares of each variable by the sum of squares of the model. Using PROC MIXED the effect of depth, date and treatment on soil NO 3 − content was assessed using a repeated measures model, with depth and date as repeated terms and plot as subject. An autoregressive correlation matrix was used. Because the first soil sampling was done as composite, the model was used with the data of the second (after harvest) and third soil sampling (end of experiment, after winter).
Adjusted means were calculated using the LSMEANS and SLICE statements in PROC PLM, with letter display for pairwise comparisons at α = 5% using the Student-Newman-Keuls method for all linear models (Büchse and Zenk 2013). All graphs were done with the graphical R package ggplot2 (Wickham 2009).

Weather conditions
After sowing, precipitation summed up to 44 mm during October 2011. A dry November was followed by a mild, rainy winter with a median daily temperature of 3.9 °C and a total precipitation of 170 mm during December 2011 and January 2012. From the end of January, temperature dropped down (lowest mean daily temperature: -12.4 °C) without a snow cover for two weeks (S1, supplementary online material). Vegetation period started beginning of March with low precipitation and consequently low soil moisture (Fig. 1). To avoid drought stress due to the lack of rain, the field experiment was irrigated on 29 May 2012 with 17 mm. Precipitation was higher during June and July (172 mm), nevertheless, its clustered distribution led to dry soil conditions by the end of June (30% WFPS) and a rewetting event two days after (42% WFPS after 46 mm of rain). Before harvest, wheat plants showed signs for leaf rust infection-orange-red pustules on leaf surface.
After harvest ("tillage" season), precipitation continued while temperature dropped from 20 °C to 0 °C (end of November 2012), leading to higher soil Experimental time periods "vegetation period", "tillage" and "winter" are represented as coloured blocks. Irrigation took place once (29 May 2012, white bar) moisture (60% WFPS) at the beginning of the winter season (December 2012). The relatively harsh winter (104 days) was characterized by continuous precipitation with changing temperatures, nevertheless, in a small range (-7 to 11 °C) and with 21 ice days (daily maximum air temperature below 0 °C) (Fig. 1).

Temporal dynamics and drivers of N 2 O fluxes
Average N 2 O fluxes before the first N application were 13 (± 11) µg N 2 O-N m −2 h −1 (Fig. 2). After the first N application, only CAN treatment showed elevated fluxes (41 µg N 2 O-N m −2 h −1 ) one week after N application. Two peaks with flux rates higher than 100 µg N 2 O-N m −2 h −1 were registered in the CAN treatment in a period of rising temperatures and in conjunction with rainfall on 3 May 2012 and due to the irrigation, which had taken place two days before the second N 2 O flux measurements on 31 May 2012. The other fertilized treatments showed rather low fluxes during the vegetation period, ranging from 3 to 68 µg N 2 O-N m −2 h −1 and with a rise of fluxes in June and reduction at harvest.
After harvest, fluxes were high after each tillage event. With 98 µg N 2 O-N m −2 h −1 in the ASN and ASN + DMPP treatment, highest flux in this period was measured after seeding (13 October 2012). During winter fluxes were low, ranging between 0 and 37 µg N 2 O-N m −2 h −1 (Fig. 2).
Soil temperature was a main driver for N 2 O fluxes (Table 3). A comparison of soil temperature and fluxes shows a similar course, with higher fluxes during the warmer period between May and September 2012 ( Figs. 1 and 2). Soil NH 4 -N and NO 3 -N were the second and third main drivers for N 2 O fluxes, Fig. 2 Course of the median N 2 O emission (n = 4). Arrows represent nitrogen (N) fertilization (N = all treatments; N* = only three application rates treatment), harvest (H) and tillage (T). Experimental time periods "vegetation period", "tillage" and "winter" are represented as coloured blocks. The number of application rates is given in square brackets. ASN: ammonium sulphate nitrate; CAN: calcium ammonium nitrate; DMPP: 3,4-dimethylpyrazol phosphate; DMPSA: 3,4-dimethylpyrazol succinic acid followed by the weekly change of WFPS (ΔWFPS) and WFPS (Table 3).  (Table 1) and consequently also cumulative emission by 38% during the vegetation period (Table 2). Nevertheless, this effect did not have a repercussion on the cumulative annual emissions (p = 0.13).

Effect of N fertilization, splitting and N fertilizer type
Splitting of fertilization had a significant influence on the flux behaviour during seasons, with highest fluxes during the tillage period, followed by vegetation and winter period. In the single application treatments winter fluxes were significantly lower than during the rest of the seasons (Table 1).

Effect of nitrification inhibitor
Nitrous oxide fluxes were significantly reduced using DMPP and DMPSA, mostly during the vegetation period and, in the case of DMPP, also on an annual basis (  (Fig. 2).
This reduction of N 2 O fluxes induced a reduction of cumulative N 2 O emissions by 38% in treatments which used DMPSA and DMPP at the single Table 1 Type 3 tests of fixed effects and back transformed adjusted means of significant effect "treatment" (annual) and significant interaction "season x treatment" for N 2 O fluxes 1. Adjusted mean N 2 O fluxes followed by a common capital letter are not significantly different within treatments (Student-Newman-Keuls; α = 5%) 2. Adjusted mean N 2 O fluxes followed by a common small letter are not significantly different within seasons (Student-Newman-Keuls; α = 5%)   (Table 2).
Several logarithmized soil variables influenced N 2 O flux rates, with soil temperature and NH 4 -N and NO 3 -N content being the most influential ones (Table 3). Positively correlated variables were soil temperature, NO 3 -N, and ΔWFPS; NH 4 -N and WFPS were negatively correlated with N 2 O fluxes.

Soil N min
Highest NH 4 + amounts in the upper soil layer (0-0.3 m) were measured after fertilization in the single application treatments with DMPP or DMPSA (Fig. 3). Highest NO 3 − amounts were found in the single application treatments and after the third CAN application in the [3]ASN/CAN/CAN treatment on 31 May 2012.

Yield and yield components
Fertilization was a main driver for yield and yield components, with significant effects on grain and straw yield, spike number as well as on N related variables such as N concentrations in grain and straw (and N amount in these wheat fractions). Among fertilized treatments, the N amount in straw ranged between 25 and 33% of applied N fertilizer and N surplus varied only between 49 and 65 kg N ha −1 (Fig. 5). Among the fertilized treatments, a single application of ASN yielded 21.4% more grain and

Main drivers for N 2 O release
The positive correlation between temperature and N 2 O flux rates can be explained not only by a direct effect of temperature on enzymatic activity, but also by an increased soil anaerobiosis after stimulation of soil respiration (Butterbach-Bahl et al. 2013). We found a negative correlation between the N 2 O flux rates and the NH 4 + contents, which were mainly high after fertilizer application. This might be a hint on nitrification as the main N 2 O source in this period. However, we cannot exclude denitrification as another relevant N 2 O source, since NO 3 − , the end product of nitrification, serves as a substrate for denitrification. This was indicated by a positive correlation between N 2 O flux rates and NO 3 − contents. Using a stable isotope approach, Ruser et al. (2006) reported a contribution of denitrification of up to approx. 66%, at a low soil moisture (40% WFPS) in a soil similar in soil texture and humus content. They found this high ratio of denitrification to the total N 2 O flux especially after the rewetting of dry soil, and explained this phenomenon by inferring that increased oxygen consumption and microbial growth after rewetting was due to an enrichment of easily available carbon under dry soil conditions, which induced anaerobiosis even at low soil moisture. This would also explain the positive correlation between N 2 O fluxes and the change of soil moisture (ΔWFPS) between two sampling dates. Soil tillage also stimulated N 2 O flux rates. As summarized by Guzman-Bustamante et al. (2019), tillage increases C turnover in soil aggregates, nitrification and denitrification potential and enhances C and N availability of crop residues. Similarly, increased N 2 O fluxes after tillage have also been reported e.g., by Mutegi et al. (2010) after winter barley harvest and by Lebender et al. (2014b) after winter wheat harvest.
Use of NI diminished the slope of Log NO 3 − (Table 3) indicating that DMP based NIs were able to lower N 2 O fluxes by decreasing NO 3 − availability as a substrate for denitrification.
Although moisture plays a predominant role in triggering N 2 O production, by filling soil pores with water thus limiting oxygen diffusion and consequently stimulating denitrification  Roman-Perez and Hernandez-Ramirez 2021), we found a negative correlation between WFPS and N 2 O fluxes (Table 3). A negative correlation of these variables during the growing season was also found by Vitale et al. (2013), who hypothesized a limiting effect of high soil moisture on nitrification. Analog to the correlation between NH 4 + and N 2 O fluxes, WFPS in our study was higher during periods of time when other conditions were limiting, i.e., during winter, when WFPS reached ≈70%, but NO 3 − contents ranged only between 0.9 and 7.8 g N kg soil −1 . Additionally, Guzman-Bustamante et al. (2019) reported a temporal C limitation during the cropping season of winter wheat at the same study site and at the same time as the measurements presented here, overwriting moisture effects on N 2 O flux rates. As pointed out by Granli and Bøckman (1994), fertilizer application, as a seasonal operation, which takes place when temperature is high and-in the case of South Germany-in periods with most precipitation, can mask the effect of soil-physical variables on N 2 O fluxes and complicates the interpretation of field studies.

Effect of N fertilization and N splitting
Annual N 2 O emissions of [1]CAN and [1]ASN (2.85 and 2.82 kg N 2 O-N ha −1 a −1 ) were in accordance with the range of N 2 O emissions reported by Kaiser and Ruser (2000) who reported a mean N 2 O emission of 2.8 kg N 2 O-N ha −1 a −1 from 14 field experiments with wheat in Germany. Lebender et al. (2014b) found lower annual emissions at two sites in Germany cropped with winter wheat (1.7 and 1.8 kg N 2 O-N ha −1 a −1 ) although this experiment took place in the same year as our experiment (2012). In contrast to our study, soil texture at the study site of Lebender et al. (2014b) was characterized by higher sand and lower clay and silt contents.
Weather conditions before the vegetation period indirectly influenced N 2 O fluxes and annual emissions of N fertilized treatments in our study. Due to a mild winter with a short cold period without snow cover, wheat plants tended to create more stems and  (Ruser et al. 2001) and in winter oilseed rape (Kesenheimer et al. 2021). The [1]ASN treatment contained more NH 4 + -N than [1]CAN (CAN: 50% as NO 3 -N, 50% as NH 4 -N; ASN: 29% as NO 3 -N, 71% as NH 4 -N). Desorption of higher amounts of fertilized NH 4 + from clay minerals might have postponed N availability and the resulting substrate supply for N 2 O production into the post-harvest period in the [1]ASN treatment (Lebender et al. 2014a).
Increasing NO 3 − and low NH 4 + amounts during the tillage period indicate a rapid nitrification of mineralized N from N-rich straw. The turnover of easily degradable carbon fractions of the straw such as cellulose and hemicellulose might also have further contributed to O 2 consumption, thus increasing anaerobic conditions favouring denitrification and enhancing N 2 O fluxes after harvest.
A comparison with other field experiments shows that grain yield of our fertilized treatments was rather low, between − 16% and − 26% (Pasda et al. 2001;Schulz et al. 2015). The reasons for the low yields might be related to year, as Makary et al. (2020) also reported low grain yields in the same experimental year. They attributed this result to the exceedingly warm winter, which led to an unfavourable high tiller density in spring.
A comparison between the single application in [1]CAN and the traditional threefold application in [3]ASN/CAN/CAN showed significantly decreased fluxes and 38% less emissions during the vegetation period in the traditional split fertilization treatment. Although statistically not significant on an annual basis, the t-test comparison of the annual N 2 O emission between the treatments with single and three applications was very close to statistical significance (p = 0.056) and can at least be considered a substantial trend.
Possible reasons for lower N 2 O emissions with increasing number of N splitting compared to a single application rate were (i) the generally lower soil NO 3 − contents in the treatments with fertilizer splitting and especially during the time of the first N 2 O peak after rain, and (ii) the fact that fertilizer granules were only slowly dissolved due to relatively low soil moisture following the second and third application. As discussed by Knittel et al. (2007), the later fertilization occurs, the higher the probability that soil might be too dry for fertilizer granules to be dissolved.
The high soil NO 3 − amounts in the split treatments after the second and third application did not induce enhanced N 2 O fluxes; this might be the result of the low soil moisture and the corresponding good aeration in this period which limited denitrification. Mainly because of different precipitation patterns and the occurrence of heavy rainfall events after N applications, the success of splitting as a N 2 O reduction strategy can strongly vary as shown by Guardia et al. (2018) and others.
In our experiment, grain yield and quality were influenced by splitting of the N-fertilizer with higher grain yield in the treatment without splitting when compared to the traditional fertilization with three application rates and higher crude protein contents in treatments with split application. Neither Schulz et al. (2015) nor Makary et al. (2020) found differences in yield or N content for split N fertilization on similar study sites in Southwest Germany. Both recommended to consider one single CAN application in a late (shooting) stage when modern wheat varieties are grown on soils with low NO 3 − leaching during the growing season. In contrast, our results recorded under unusually dry conditions (19% lower rainfall from March to July when compared to the longterm annual mean) seem to be more similar to the ones reported under Mediterranean conditions: we observed slightly higher grain yields with one application rate (Guardia et al. 2020) and a higher N grain content when N fertilizer was split (Ercoli et al. 2013). Yield components such as spike number and TGM followed a similar trend as found by Pasda et al. (2001), with smaller spike numbers and higher TGM when N fertilizer was split. Since results from a previous experiment on the same field showed higher grain yields for a fertilization with three application rates (Guzman-Bustamante et al. 2012) and no difference between protein contents (data not shown), the comparatively milder winter and dryer vegetation period together with the high N amount might have driven spike numbers on [1]ASN and [1]CAN and so elevated competition among wheat plants and decreased grain yield (Maidl et al. 1998). In this sense, split fertilization was not able to contribute to yield formation, since the spike number was too high (Scharf and Alley 1993).

Effect of nitrification inhibitors
Both NIs in our study reduced the mean annual N 2 O emission (DMPSA: 26%; DMPP: 33%), with the reduction for DMPP being statistically significant. For [1]CAN + DMPSA the tillage operation after harvest might have masked N 2 O reduction during the vegetation period (Corrochano-Monsalve et al. 2020).
Similar reduction potentials for DMPP and DMPSA were reported for field studies by Ruser and Schulz (2015) and by Huérfano et al. (2016). The reduction of N 2 O emissions after the application of ammonium containing fertilizers with NIs was explained directly by lower N 2 O production during nitrification as well as indirectly by the lower substrate availability for denitrification (Ruser and Schulz 2015). Additionally, Torralbo et al. (2017) detected an increased N 2 O reduction during denitrification after NI application which also decreased net N 2 O release from soil.
A direct comparison between DMPP and DMPSA cannot be drawn with our dataset, as we used different N fertilizers for the two inhibitors. Differences between the two products (ASN + DMPP vs. CAN + DMPSA) might result either from different efficiencies of the inhibiting compounds or from the different share of NH 4 + and NO 3 − in CAN and ASN. The latter was reflected by the soil NH 4 -N/NO 3 -N ratio, which was higher in the ASN + DMPP treatment (vs. ASN) for approximately 3.5 months, whereas it did not differ that clearly for CAN + DMPSA (vs. CAN).
Twofold split application of ASN + DMPP treatment performed similarly to a single application of ASN + DMPP leading to 33% lower annual N 2 O emissions compared to a single application of ASN. One of the reasons for this reduction was the same as for the [3]ASN/CAN/CAN treatment: lower soil NO 3 − amounts were registered for split treatments during periods with conditions favourable for high N 2 O production.
Similarly to our results, splitting NI fertilizers did not further mitigate N 2 O emissions compared to a single N application under Mediterranean conditions (Huérfano et al. 2016;Corrochano-Monsalve et al. 2020), because soil conditions during the second fertilizer application were not favourable for N 2 O production (WFPS < 48%). Contrarily, if the second NI application occurs when denitrification conditions are optimal (high water content and high soil temperature), high N 2 O fluxes might raise emissions to the same level as soil fertilized without NI (Huérfano et al. 2015).
Despite In this regard, determination of the inhibiting compound and metabolites might be interesting, since it was shown that approx. 16% of DMPP were still present in a topsoil under winter wheat at the end of the vegetation period (Benckiser et al. 2013).
In terms of winter wheat yield and quality, our results agree with Pasda et al. (2001) and Huérfano et al. (2015), who did not find an effect of split NI on winter wheat grain yield, whereas protein content was increased in a twofold ASN + DMPP application compared to all single application treatments. Since our single application treatments with and without NI were all in a lower crude protein class (≈ 11.2%) compared to the treatments with split application (12.4-14%), splitting seems to be the main factor influencing crude protein in wheat grain as discussed before.
As enhanced-efficiency fertilizers are more expensive than regular mineral fertilizer, its use might not be profitable in a wheat system. From a climate protection point of view, farmers could waive its use when an appropriate N fertilization management is implemented (Li et al. 2018). However, due to expiration of patent protections, NIcontaining fertilizers became cheaper on the European market in the last years, and an economical re-evaluation of the use of NIs seems worthwhile.

Conclusion
Our first hypothesis-that a threefold split N application can decrease N 2 O emission compared to a single N application-can be partially corroborated (p < 0.1) as [3]ASN/CAN/CAN reduced annual N 2 O emission compared to one application of CAN and ASN. The second hypothesis-that split application of a NI fertilizer can further mitigate N 2 O emissions, compared to a sole NI application-must be rejected, as N 2 O emission levels of both split NI treatments ([2]ASN + DMPP and [2]ASN + DMPP/ CAN) showed the same emission levels as a single application of ASN + DMPP. Nevertheless [2] ASN + DMPP contributed to significantly higher grain protein content. Our third hypothesis-that DMPSA used with CAN shows a similar N 2 O reduction as ASN + DMPP-must be rejected as well, since a single application of CAN + DMPSA mitigated N 2 O emissions from CAN only during the vegetation period but not on an annual basis. Only DMPP was able to lower N 2 O fluxes during the vegetation period and winter, thus mitigating annual emissions. Our results support the splitting of N fertilizer in order to achieve high grain quality when appropriate wheat varieties are sown by simultaneously lowering N 2 O emissions. As a result of climate change, precipitation patterns (with more heavy rain events during the cropping season) will change more frequently in the future. Such strong rainfall events can trigger N 2 O production after N application, and thus the use of DMP-based nitrification inhibitors could be a powerful tool to mitigate N losses in these periods. Future studies should focus on the effects of DMPSA on N transformation in soils, especially after harvest. Determination of long-term effects on nitrification and probably also on denitrifiers may help to improve our understanding in this context.