2.1 Experimental set-up

Silt loam soil Albic Luvisol from arable cropland of Merklingsen experimental station (Germany) was used (silt content approx. 87 %, 11 % clay, 2 % sand). Three treatments were applied: (1) soil was sieved with 4 mm mesh size, the tracer solution was added evenly, soil was homogenized and packed into the incubation column (treatment H + M: homogenized + mixed); (2) intact soil cores were directly collected in the incubation columns and the tracer solution was added through the injection needles to 12 homogeneously distributed injection points at 6 depths (in total 72 injection points per column) (treatment I + I: intact + injected); (3) soil was sieved with 4 mm mesh size (like in treatment H + M), packed into the incubation column, and the tracer solution was added through the injection needles (like in treatment I + I) (treatment H + I: homogenized + injected). For each treatment the soil columns were 0.3 m high with a diameter of 0.15 m. A mesh size of 4 mm was used because this enabled us to sieve the necessary amount of soil (56 kg) within an adequate time. The soil density of intact cores was 1.3 g cm⁻³, and the packed columns were compacted to the same density. Each column contained 6.89 kg soil. For each soil column, 216 mL of 319 mgN L⁻¹ NaNO₃ solution with 73 atom % ¹⁵N was added. This resulted in the following initial experimental settings: 75 % water-filled pores space (WFPS), 37 mg N kg⁻¹ ${\mathrm{NO}}_{\mathrm{3}}^{-}$, 42.5 atom % ¹⁵N measured in the subsamples of the homogenized soil immediately after tracer addition and mixing. The incubation lasted 8 d. The columns were continuously flushed with a gas mixture with reduced N₂ content to increase the measurement sensitivity (2 % N₂ and 21 % O₂ in He; Lewicka-Szczebak et al., 2017) with a flow of 10 mL min⁻¹. The gas samples were collected daily in the first 4 d and every second day in the last 4 d in two 12 mL septum-capped Exetainers^® (Labco Limited, Ceredigion, UK) connected to the vents of the incubation columns.

2.2 Gas analyses

The gas samples were analysed with a modified GasBench II preparation system coupled with a MAT 253 isotope ratio mass spectrometer (Thermo Scientific, Bremen, Germany) according to Lewicka-Szczebak et al. (2013). In this set-up, N₂O is converted to N₂ prior to analysis, which allows the simultaneous measurement of stable isotope ratios ²⁹R (²⁹N₂∕²⁸N₂) and ³⁰R (³⁰N₂∕²⁸N₂), of N₂, of the sum of denitrification products (N₂+N₂O) and of N₂O. Based on these measurements the following values were calculated according to the respective equations (after Spott et al., 2006).

The ¹⁵N abundance of ¹⁵N-labelled pool (a_P), from which N₂ ($a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}})$ or N₂O ($a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}\mathrm{O}})$ originates, is calculated as follows:

The calculation of a_P is based on the non-random distribution of N₂ and N₂O isotopologues (Spott et al., 2006), where ³⁰x_M is the fraction of ³⁰N₂ in the total gas mixture:

a_M is ¹⁵N abundance in total gas mixture:

a_bgd is ¹⁵N abundance of non-labelled pool (atmospheric background or experimental matrix).

The fraction originating from the ¹⁵N-labelled pool (f_P) for N₂ ($f_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}}$), N₂+N₂O ($f_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}+{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$), and N₂O ($f_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$) within the sample is calculated as follows:

N₂O residual fraction ($r_{{\mathrm{N}}_{\mathrm{2}}\mathrm{O}})$ represents the unreduced N₂O mole fraction of pool-derived gross N₂O production (Lewicka-Szczebak et al., 2017):

where y represents the mole fractions.

2.3 Soil analyses

At the end of incubation, soil samples were collected from each column using a Goettinger boring rod with a diameter of 18 mm (Nietfeld GmbH, Quakenbrück, Germany). Three cores were taken from each column and separated into a top (0 to 15 cm) and bottom (15 to 30 cm) layer. For injected treatments ((H + I) and (M + I)) these sample cores were taken between injection points, and additional cores were collected from the injection points. All soil samples were homogenized and analysed for water content (by weight loss after 24 h drying in 110 ^∘C), nitrate content (by extraction in 2 M KCl 1:4), and ¹⁵N enrichment in nitrate (with the bacterial denitrification method of Sigman et al., 2001).

2.4 Statistics

For testing the statistical significance of the differences between treatments, ANOVA and Tukey HSD post hoc test were applied using R 3.4.2 (R Core Team, 2013).

In Table 3, for the comparison of particular $a_{{\mathrm{NO}}_{\mathrm{3}}}$ and a_P values, we applied the following calculated parameters:

• cumulative relative difference (cumulated diff) calculated as the sum of differences in ¹⁵N enrichment of different pools for all 24 samples: cumulated diff $={\sum }_{i=\mathrm{1}}^n(a_{\mathrm{1}}-a_{\mathrm{2}}{)}_i$;

• absolute mean difference (mean abs diff) calculated as the mean of modulus of differences in ¹⁵N enrichment of different pools: mean abs diff $=\left({\sum }_{i=\mathrm{1}}^n\left|{\left(a_{\mathrm{1}}-a_{\mathrm{2}}\right)}_i\right|\right)/n$.

In the above equations a₁ and a₂ represent the ¹⁵N enrichment of two compared pools ($a_{{\mathrm{NO}}_{\mathrm{3}}}$ or $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}}$ or $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}\mathrm{O}})$.

Figure 1Comparison of the temporal changes in N₂O concentration (a), fraction of ¹⁵N-pool-derived N₂ (b), fraction of ¹⁵N-pool-derived denitrification products (N₂+N₂O) (c), and N₂O residual fraction (d) in three treatments: homogenized soil mixed with fertilizer (black dots), intact soil cores with fertilizer added through needle injection (red triangles), and homogenized soil with fertilizer added through needle injection (green squares). Error bars represent the standard deviation of four replicates within one treatment.

Download

3.1 Gas fluxes and denitrification product ratio

In order to compare the treatments, the time course of the results must be taken into account as the gas production differed largely between the sampling dates (Fig. 1). Therefore, we checked for statistically significant differences between the treatments individually for each sampling date. The results show comparable trends and no statistically significant differences between treatments (Fig. 1). Notably, $r_{{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$ shows very good agreement at the beginning of the experiment, when the large gas concentrations were measured and starts to differentiate when the fluxes drop from the third day (Fig. 1d), but these differences are not statistically significant. However, if the experiment is evaluated for the cumulative values, significant differences between treatments appear (Table 1). The cumulated gas fluxes of N₂O and N₂ are significantly different between the treatments I + I and H + I, whereas the H + M treatment does not differ significantly from the others. However, comparison of the entire denitrification gas flux (joint N₂+N₂O flux) reveals no statistically significant difference between treatments (Table 1). Product ratios are compared as cumulated $r_{{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$ (calculated with the cumulated fluxes) and mean $r_{{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$ (average value of all sampling points). Cumulated $r_{{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$ shows an identical pattern of significant differences as the cumulated N₂ and N₂O fluxes. For mean $r_{{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$ values, H + M and H + I treatment are significantly different, whereas the I + I treatment does not differ significantly from the others.

These results show that the different tracer application strategies tested had no impact on the total denitrification (N₂+N₂O), but the product ratio may be slightly shifted, which results in differences by comparing N₂ or N₂O flux separately. This presumably results from the differences in distribution of moisture and nitrate between treatments (see Sect. 3.2). All determined $r_{{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$ values, although partially different, indicate a pronounced dominance of N₂ over N₂O emission. Importantly, no significant differences were noted between the H + M and I + I treatment; only the H + I treatment shows higher N₂O flux, lower N₂ flux and higher $r_{{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$ (Table 1). In this treatment we probably observe joint artefacts associated with soil homogenization and needle injection technique.

The homogenized treatments show better comparability between the repetitions – they show lower standard deviations for gas emissions and for $r_{{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$ (Table 1) and smaller error bars for the daily measurements (Fig. 1). The H + I treatment shows the lowest standard deviations for the cumulative gas emission measurements (Table 1). This indicates that the observed heterogeneity for I + I treatment is not due to needle injection procedure but rather due to the intact structure of soil cores, which naturally represents the typical soil heterogeneity.

3.2 Soil parameters

In this study the high dose of added N resulted in more than doubled ${\mathrm{NO}}_{\mathrm{3}}^{-}$ content. This was much above the common recommendations of tracer addition of 10 %–25 % of native soil N (Davidson et al., 1991). These recommendations are motivated by the need to minimize the fertilization effect and to trace the naturally occurring N transformation processes. But in this study we only aimed to compare tracer addition strategies and did not intend to draw conclusions for this particular study site. Establishing a high ¹⁵N enrichment of the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ by high addition of ¹⁵N-labelled ${\mathrm{NO}}_{\mathrm{3}}^{-}$ enhanced the sensitivity of N₂ flux detection, which is a prerequisite for reliably identifying potential experimental artefacts, which we aimed to evaluate in this study.

A good insight into heterogeneity within columns is also provided by the soil analyses performed at the end of experiment, by collecting samples from various areas of each soil core (Table 2). Clearly, I + I treatment shows the largest standard deviations between repetitions. Also, the most pronounced differences between top and bottom soil layer can be noted for this treatment, but only soil moisture is significantly lower for the bottom layer. Since this is not the case for H + I treatment, it reflects the natural heterogeneity of intact cores rather than a result of label injection procedure. The values from injection points are never significantly different from samples between injection points (within one treatment), which indicates a good distribution of the tracer solution (Table 2).

Table 1Comparison of cumulated fluxes, cumulated product ratio (cumulated $r_{{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$) and mean product ratios (mean $r_{{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$) in three treatments: homogenized and mixed (H + M), intact and injected (I + I), homogenized and injected (H + I). Statistically significant differences are indicated with superscript letters (^* p<0.05, ${}^{**}$ p<0.01, ${}^{***}$ p<0.001).

Download Print Version | Download XLSX

Table 2Soil analyses at the end of the experiment: mixed samples, and separately from the top and bottom layer and for injected columns also from injection points (including both top and bottom layer). Statistically significant differences are indicated with superscript letters (${}^{**}$p<0.01, ${}^{***}$p<0.001). For individual values the differences within treatment were tested, and for mean values the differences between treatments were tested.

Download Print Version | Download XLSX

Significant differences in soil parameters between treatments (Table 2) were observed. The I + I treatment shows significantly lower nitrate content compared to homogenized treatments (Table 2). This must be due to initial soil nitrate content. The soil was stored for 2 weeks before the experiment. Storing of mixed soil or sieving and homogenization procedures probably intensified N mineralization and the formation of additional nitrate through intensified nitrification, which has also been observed in previous studies (Kaur et al., 2010). Moreover, the H + M treatment shows significantly higher ¹⁵N enrichment of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ ($a{{}^{\mathrm{15}}\mathrm{N}}_{{\mathrm{NO}}_{\mathrm{3}}}$) than injected treatments. This may be due to injection procedure where the needles might get partially clogged with soil causing the addition of tracer solution to be lower than planned. The assumption that the injected volume was lower than the target and thus also lower than the addition of tracer solution to H + M treatment can also be supported by the slightly lower soil moisture and nitrate content of the injected treatments.

Table 3Differences between the measured ¹⁵N abundance in soil nitrate ($a_{{\mathrm{NO}}_{\mathrm{3}}}$) and determined ¹⁵N abundance of ¹⁵N-pool-derived N₂ ($a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}}$) and N₂O ($a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$) expressed as the cumulative relative difference for all samples (n=24), mean absolute difference (see Sect. 2.4 for calculation procedure). In the above equations a₁ and a₂ represent the ¹⁵N enrichment of two compared pools ($a_{{\mathrm{NO}}_{\mathrm{3}}}$ or $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}}$ or $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$).

Download Print Version | Download XLSX

3.3 ¹⁵N abundance in soil active pools

Despite the pronounced difference in ¹⁵N content between treatments, the results can still be compared because the ¹⁵N abundance of actively denitrifying pool (a_P value) for each sample is individually calculated based on the distribution of N₂ and N₂O isotopologues. We checked how well these calculated a_P values for N₂ and N₂O correspond with the respective ¹⁵N enrichment measured in soil nitrate ($a_{{\mathrm{NO}}_{\mathrm{3}}}$) and between each other (Table 3). This comparison gives additional information about the distribution of the ¹⁵N label. The cumulative relative difference represents the overall deviation between the analysed pools. Very high cumulative difference was noted between the a_P values of both gases and $a_{{\mathrm{NO}}_{\mathrm{3}}}$ in H + M treatment. This is mostly due to the first two sampling days, where a_P values were significantly lower than $a_{{\mathrm{NO}}_{\mathrm{3}}}$ (mean difference of ca. 15 atom % ¹⁵N, Fig. 2), whereas for the next samplings they corresponded very well (mean difference of ca. 1 atom % ¹⁵N, Fig. 2). This shows that initially the gases were produced in soil microsites depleted in ¹⁵N compared to the mean soil value. This is the case for all three treatments; however, the largest difference is observed for H + M treatment due to highest $a_{{\mathrm{NO}}_{\mathrm{3}}}$ values. The absolute mean difference represents the average variation range of the compared values. For the comparison of mean absolute difference between $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}}$ and $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$, we obtained quite a good agreement, much better than for the comparisons with $a_{{\mathrm{NO}}_{\mathrm{3}}}$ (Table 3). This shows that both gases originate mostly from the same soil pool. Importantly, even in the H + M treatment where large mean difference between $a_{{\mathrm{NO}}_{\mathrm{3}}}$ and a_P values was noted, the mean difference between $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}}$ and $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$ is very low. The fact that $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$ shows much closer agreement with $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}}$ than $a_{{\mathrm{NO}}_{\mathrm{3}}}$ suggests that, when missing data on $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}}$, which is often the case due to high N₂ detection limit of the gas-flux method, the $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$ should be used rather than $a_{{\mathrm{NO}}_{\mathrm{3}}}$ or a theoretical value on ¹⁵N abundance, as has also been proposed in previous studies (Bergsma et al., 2001; Stevens and Laughlin, 2001).

Figure 2Comparison of ¹⁵N abundance in total initial and final soil nitrate ($a^{\mathrm{15}}{N}_{{\mathrm{NO}}_{\mathrm{3}}})$ and in active soil pool emitting N₂ ($a_{\mathrm{P}}^{\mathrm{15}}$N${}_{{\mathrm{N}}_{\mathrm{2}}})$ and N₂O ($a_{\mathrm{P}}^{\mathrm{15}}$N${}_{{\mathrm{N}}_{\mathrm{2}}\mathrm{O}})$ in three treatments: homogenized soil and mixed fertilizer (H + M, black points)), intact soil core and injected fertilizer (I + I, red points), homogenized soil and injected fertilizer (H + I, green points).

Download

Interestingly, for the I + I treatment lower differences between $a_{{\mathrm{NO}}_{\mathrm{3}}}$ and $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$ or $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}}$values were obtained, but there was a larger difference between $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}}$ and $a_{\mathrm{P}\mathrm{\_}{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}$ compared to homogenized treatments (Table 3, Fig. 2). This shows that the multiple-injection technique reduced the formation of isolated soil microsites characterized by distinct ¹⁵N enrichment when compared to the bulk $a_{{\mathrm{NO}}_{\mathrm{3}}}$ value measured. However, the slightly higher difference between a_P values for N₂ and N₂O suggests non-identical origins for both gases, i.e. probable slight admixture of hybrid N₂ (Spott et al., 2011) since the ¹⁵N enrichment of N₂ shows lower values than N₂O. This could explain the higher cumulated N₂ flux for I + I treatment (Table 1).

3.4 Homogeneity of ¹⁵N tracer distribution and accuracy of results

Surprisingly, the inconsistency in ¹⁵N abundance in total and actively denitrifying nitrate soil pools (Fig. 2) indicates the largest inhomogeneity at the beginning of the incubation for the homogenized soil, which is then equilibrated after 2 d of incubation. This resulted most probably from the imperfect mixing of the relatively wet (gravimetric water content of 29.3 %) silt loam soil and could be due to delayed equilibration of added ¹⁵N solution into the centre of soil aggregates where denitrification rates are probably highest (Sextone et al., 1985). But, importantly, these first two days are also the ones with the highest gas production and close agreement of results between all three treatments (see Fig. 1). This suggests that even non-homogeneous distribution of ¹⁵N label and thus heterogeneity in content and ¹⁵N enrichment of nitrate in soil does not lead to severe bias in determining denitrification and its product ratio.

This study allows only for the comparison of these different treatments but not for checking the true emission values, since we have not used any independent method for flux determination. However, we can conclude that, despite pronounced differences in a¹⁵N values of different treatments and different pools, the calculated results for gas fluxes and product ratios were mostly not significantly different between the treatments. This supports the assumption that in a real soil situation even imperfect label distribution allows for obtaining accurate results (Arah, 1997; Davidson et al., 1991; Deppe et al., 2017). But, importantly, this is possible only if we measure and use a_P values representing the ¹⁵N values of the pools actively producing N₂ and N₂O. The fluxes would be significantly underestimated if the $a_{{\mathrm{NO}}_{\mathrm{3}}}$ value was applied for calculations. For example, for the first sampling point this would result in about a 20 % underestimation of the N₂ flux when the measured final $a_{{\mathrm{NO}}_{\mathrm{3}}}$ value was applied and an about 30 % underestimation when the initial $a_{{\mathrm{NO}}_{\mathrm{3}}}$ value was applied. Significant differences in ¹⁵N enrichment of total and active nitrate pool have also been found in our previous laboratory and field studies (Buchen et al., 2016; Deppe et al., 2017). It was shown that in such cases the ¹⁵N enrichment of N pool undergoing denitrification is well represented by a_P values but not by $a_{{\mathrm{NO}}_{\mathrm{3}}}$ values.

The homogeneity of ¹⁵N label distribution depends not only on the tracer addition technique but even more on the soil type, water content, and initial nitrate and ammonium content. In our previous laboratory experiments quite a good agreement between $a_{{\mathrm{NO}}_{\mathrm{3}}}$ values and a_P values was achieved, indicating a homogenous denitrifying pool (Lewicka-Szczebak et al., 2017). In that study similar soil texture was used (silt loam), but the initial amount of nitrate and ammonium was very low, and soil samples were prepared at soil moisture of 70 % WFPS with rest water added on top, and soil was incubated in high moisture conditions. But, notably, the anoxic conditions showed perfect agreement in $a_{{\mathrm{NO}}_{\mathrm{3}}}$ and a_P values, whereas for oxic conditions slight differences have also been noted (Lewicka-Szczebak et al., 2017). Oxic conditions can be expected to yield greater disagreement between $a_{{\mathrm{NO}}_{\mathrm{3}}}$ and a_P due to dilution of the bulk $a_{{\mathrm{NO}}_{\mathrm{3}}}$ by soil-derived non-labelled N sources in contrast to anoxic soil microsites (Deppe et al., 2017). In the H + M treatment of the actual experiment, inhomogeneity was probably the result of soil moisture during soil homogenization being too high (75 % WFPS), causing the formation of larger aggregates. But this problem can be overcome if the ¹⁵N label is incorporated at low soil moisture and target moisture is established by adding water afterwards (Lewicka-Szczebak et al., 2017; Well et al., 2019).