Estimating ηred and δ0 values
With respect to robust estimation of N2O reduction, a first question
arises: to what extent δ0 values and η values were variable
or constant during incubations. When assuming constant δ0
values during the experiment, calculated η values were highly variable.
The large ranges obtained are clearly in strong disagreement with previous
knowledge on possible η values (Jinuntuya-Nortman et al., 2008;
Lewicka-Szczebak et al., 2014; Ostrom et al., 2007; Well and Flessa, 2009a).
In the further interpretation of data we therefore suppose that δ0
values were variable and η values constant. While we cannot rule out
that η values varied to some extent, it is not possible to verify that
using the current data set.
Another question is whether the assumption of isotopic fractionation pattern
of closed systems holds. Logarithmic fits provided best correlations with the
measured data, whereas linear correlations that would be indicative for open
system dynamics (Decock and Six, 2013) yielded worse fits (data not shown).
This indicates that the N2O reduction follows the pattern of a closed
system according to Rayleigh distillation equation (Eq. 13), as suggested
previously (Köster et al., 2013; Lewicka-Szczebak et al., 2015;
Lewicka-Szczebak et al., 2014).
To what extent are the observed ηred and δ0 values in
agreement with previous data and how could differences be explained? For Min
soil we can compare the ηred and δ0 values obtained
here to the previous experiment, carried out with the same soil (Exp 1E, 1F, Lewicka-Szczebak et al., 2014) but using the acetylene inhibition
technique. The actual ηred15Nsp values from -8.6
to -6.7 ‰ (Fig. 4a) are quite close to that previous result of
-6.0 ‰, whereas δ015Nsp values from 4.0 to
4.5 ‰ are significantly higher than the previously determined value
of -2.7 ‰. While that previous value was within the δ015Nsp range of bacterial denitrification (-7.5 to
-1.3 ‰, Toyoda et al., 2005), the clearly higher actual values
indicate that the previous method must have strongly influenced the microbial
denitrifying communities, most probably favouring bacterial over fungal
denitrification. Much wider ranges of ηred values were found for
ηred18O (from -22.7 to -9.9 ‰) and
ηredNbulk (from -6.6 to -2.0 ‰, Table 2),
which is also consistent with the previous findings, indicating that these
values depend on enzymatic and diffusive isotope effects and as result can
vary in quite a wide range (Lewicka-Szczebak et al., 2014). The
ηred determined in Exp 1 are similar to the previous results
(-18 ‰ for ηred18O and -7 ‰ for
ηred15Nbulk, Lewicka-Szczebak et al., 2014), whereas in
Exp 2 the absolute values are much smaller, suggesting a different
fractionation pattern there. Most probably this difference is an effect of a
different range of rN2O in both experiments (Table 2). In Exp 2
we partially deal with extremely low rN2O values, which results
in smaller overall isotope effects, as also shown before (Lewicka-Szczebak et
al., 2015). But δ015Nbulk values are very robust
since the actual δ015Nbulk (-45 ‰,
Table 2) corresponds very well to the one previously determined
(-46 ‰) using the acetylene method. Conversely, δ018O
is much higher (+36 ‰, Table 2) compared to the value of
19 ‰ obtained previously (Lewicka-Szczebak et al., 2014). This may
indicate a significant admixture of fungal denitrification characterized by
higher δ018O but similar δ015Nbulk
values (Lewicka-Szczebak et al., 2016; Rohe et al., 2014).
For Org soil, much higher absolute values of ηred were found
(Table 2), being in contrast to all previous studies (Jinuntuya-Nortman et
al., 2008; Lewicka-Szczebak et al., 2015; Well and Flessa, 2009a). Hence, it
has to be questioned whether this observation is not an experimental artefact.
Actually, the Org soil anoxic treatment was the only case where 15N-pool-derived N2O was dominant (Fig. S3.1b), hence the isotopic signatures
should not be altered due to different N2O-producing pathways but mostly
governed by the rN2O. But for Org soil, based on the NA
treatment, we observe a constant and very significant increase in the
contribution of N2O from fungal denitrification during the incubation
(Fig. 5). Future studies should clarify whether such a rapid microbial
shift is possible. Fungal denitrification adds N2O characterized by
higher δ15Nsp values and presumably also higher
δ18O values (Lewicka-Szczebak et al., 2016; Rohe et al., 2014). As
a result the ηred values determined from correlation slopes are
biased because the production of 18O- and15Nα-enriched
N2O increased in time parallel to a decrease in rN2O. In
15N treatments this increase in N2O added from fungal
denitrification cannot be distinguished from bacterial denitrification
because both originate from the same 15N nitrate pool.
The Org soil data thus demonstrate that a high and variable in-time
contribution of fungal denitrification complicates the application of the
N2O isotopic fractionation approach for quantification of N2O
reduction. This is because a highly variable contribution implies that
changes in the measured δ15Nsp values can either result
from variations in δ015Nsp or rN2O. Only when the
contribution of fungal denitrification is stable, robust rN2O values can
be derived from δ15Nsp data. Although the Min soil
exhibited a smaller range in fF, the contribution of fungal
denitrification was apparently also not constant. Simultaneous application
of the other isotopic signatures, i.e. δ15Nbulk and/or
δ18O, as discussed further in Sect. 4.2.3, may help solving
this problem.
Calibration and validation of rN2O quantification
The successful calibration shows that δ015Nsp and
ηred values were stable enough within Min soil incubation
experiments for calculating rN2O using the isotope fractionation
approach.
The results of the calibration were very similar if we treated the oxic and
anoxic conditions separately and if we used a mean ηred and
δ015Nsp value of the oxic and anoxic phase of Exp 1
to all the results (Fig. 6). This indicates that the fractionation factors
determined experimentally under anoxic conditions may also be applied for isotopic
modelling for oxic conditions, e.g. for parallel field studies in
regard to denitrification processes. But importantly, our experiments were
performed under high soil moisture and the majority of cumulative N2O
flux also in oxic treatments originated from denitrification (Sect. 3.3),
which explains the similar δ015Nsp values obtained
for oxic and anoxic conditions. For lower soil moisture, differences in
δ015Nsp values should be expected due to the possible
significant admixture of nitrification processes under oxic conditions.
The results of validation show very different agreement between measured and
calculated rN2O values depending on the experimental approach
used for determination of ηred and δ015Nsp values (Fig. 7). When the experiments performed in this study were used
(Val2) the agreement was quite good. These experiments are characterized by
simultaneous N2O production and reduction and a longer duration of the
experiment of 5 to 9 days. However, when we used values found in a previous
experiment using the acetylene inhibition technique (Val1), the agreement is
much worse. Estimation of ηred and δ015Nsp using the acetylene inhibition technique included
several experimental limitations that might have affected results. Specifically,
this approach was based on separate parallel experiments with and without
N2O reduction, acetylene amendment required an anoxic atmosphere, and the
duration of incubation had to be shorter than 48 h. These limitations most
probably influence the microbial denitrifying community and do not provide
the true δ015Nsp values.
Whereas finding the true δ015Nsp values is rather
challenging, fewer problems seem to be related to the ηred15Nsp values. For them similar values were found in
all the experiments, where He incubations, 15N gas flux or acetylene
inhibition methods were applied. The determined values were also similar to
the mean literature ηred15Nsp value of
-5 ‰ (Lewicka-Szczebak et al., 2014). Therefore, applying this
common literature value for the calculations (Val3) also provided a very good
agreement between measured and calculated rN2O values. Hence,
this reinforces the previous conclusion that the ηred15Nsp value of -5 ‰ can be commonly
applied for rN2O calculation (Lewicka-Szczebak et al., 2014),
but major caution should be paid to the proper determination of δ015Nsp values, which may cause much larger bias of the
calculated rN2O.
Mapping approach to distinguish mixing and fractionation
processes
The emitted N2O is analysed for three isotopocule signatures and the
relations between them (δ15Nsp / δ18O,
δ15Nsp / δ15Nbulk, δ18O / δ15Nbulk) can be informative. Namely, the
observed correlation may result from the mixing of two different sources or
from characteristic fractionation during N2O reduction, or from the
combination of both processes. If the slopes of the regression lines for
these both cases were different, mixing and fractionation processes could be
distinguished. Such slopes were often used for interpretations of field data
(Opdyke et al., 2009; Ostrom et al., 2010; Park et al., 2011; Toyoda et al.,
2011; Wolf et al., 2015) but recently this approach was questioned because of
very variable isotopic fractionation noted during reduction for O and N
isotopes (Lewicka-Szczebak et al., 2014; Wolf et al., 2015). A recent study
showed that for moderate rN2O (> 0.1) the
δ15Nsp / δ18O slopes characteristic of
N2O reduction are quite consistent with previous findings
(Lewicka-Szczebak et al., 2015), i.e. they vary from ca. 0.2 to ca. 0.4
(Jinuntuya-Nortman et al., 2008; Well and Flessa, 2009a). Hence, in such
cases, the reduction slopes may significantly differ from the slopes
resulting from mixing of bacterial and fungal denitrification, characterized
by higher values of about 0.63 and up to 0.85 (Lewicka-Szczebak et al.,
2016).
In theory, the slopes for calculated δ0 values are not influenced
by N2O reduction and hence should be mostly caused by the variability of
mixing processes, whereas the slopes of the measured δ values reflect
both mixing and fractionation due to N2O reduction. For Min soil, there
is no correlation between calculated values of δ015Nsp
and δ018O (Table 3), which indicates that the correlation
observed for measured δ values was a result of fractionation
processes during N2O reduction. In contrast, for Org soil all the
correlations for calculated δ0 values are still very strong and
show similar slopes as the correlations for measured δ values
(Table 3). This indicates a very significant impact of the mixing of various
N2O-producing pathways.
The δ15Nsp / δ18O slopes for Org soil are
generally higher (from 0.65 to 0.76) than for Min soil (from 0.30 to 0.64)
(Table 3). This supports the hypothesis from the previous Sect. 4.2.1 about a
higher contribution of fungal N2O in Org soil. But we can also notice
that the slopes in Exp 1 are lower than in Exp 2. Most probably less stable
microbial activity is present under the longer incubation in Exp 2 (9 days)
compared to short phases analysed in Exp 1 (3 days). As observed from the
calculated δ0 values (Fig. 5) the estimated contribution of fungal
N2O most probably increases with incubation time. Hence, the higher
slopes for Exp 2 probably result from the admixture of fungal denitrification
and the lower slopes for Exp 1 better represent the typical bacterial reduction
slopes. The δ15Nsp / δ18O slopes may thus
be helpful in indicating the admixture of various N2O sources.
Interestingly, there is no correlation between isotopic values in oxic Exp 2
for Min soil. A single process or the combination of several processes, which
cause large variations in δ15Nsp but not in δ18O, seems to be present there. This might be due to admixture of N2O
from different microbial pathways and possibly also due to O exchange with
water. In this treatment we also observe the lowest N2O fluxes and also
the lowest fP_N2O values, which suggest the largest input
from nitrification. The δ15Nsp values for hydroxylamine
oxidation during nitrification are much larger (ca. 33 ‰) than for
bacterial denitrification or nitrifier denitrification (ca. -5 ‰)
(Sutka et al., 2006), whereas δ18O may be in the same range for
both processes (Snider et al., 2013; Snider et al., 2011). This could be an
explanation for the missing correlation between δ15Nsp
and δ18O (Table 3).
The graphical interpretations including δ15Nbulk values
are more difficult since the isotopic signature of the N precursor must be
known, but can be also informative and were often used (Kato et al., 2013;
Snider et al., 2015; Toyoda et al., 2011, 2015; Wolf et al., 2015; Zou et
al., 2014). The slopes between δ18O and
δ15Nbulk observed in our study range mostly from 1.94 to
3.25 (Table 3), which corresponds quite well to the previously reported
results from N2O reduction experiments where values in the range from
1.9 to 2.6 were reported (Jinuntuya-Nortman et al., 2008; Well and Flessa,
2009a). Only for Org soil in anoxic conditions (in both Exp 1 and 2) is this
slope largely lower and it ranges from 0.61 to 0.84. These values are more
similar to δ18O / δ15Nbulk slopes for the
calculated δ0 values (0.56 for Min soil and 1.04 for Org soil
(Table 3)) and are significantly lower than typical reduction slopes. Thus,
most probably, they are instead due to the mixing of various N2O sources.
However, the calculated δ0 values cannot be explained with mixing
of bacterial and fungal denitrification only (Fig. S4.3b).
For the relation of
δ15Nsp / δ15Nbulk (Fig. S4.2) the
reduction and mixing slopes cannot be separated so clearly. The calculated
δ0 values are not all situated between the mixing end-member of
bacterial and fungal denitrification. This observation is similar to that for
δ18O / δ15Nbulk and is due to some data
points showing very low δ015N(N2O/NO3-)bulk values down to
ca. -70 ‰. This value exceeds the known range of the 15N
fractionation factors due to the NO3- / N2O steps of
denitrification, i.e. based on pure culture studies, from -37 to
-10 ‰ for bacterial and from -46 to -31 ‰ for fungal
denitrification (Toyoda et al., 2015) (as displayed on graphs in Fig. S4)
and, based on controlled soil studies, from -55 to -24 ‰
(Lewicka-Szczebak et al., 2014; Well and Flessa, 2009b). This additional
N2O input may originate from nitrifier denitrification, as already
suggested based on the 15N treatments results (Sect. 3.3). Frame and
Casciotti (2010) determined that fractionation factors for nitrifier
denitrification are ε15NNH4/N2Obulk= 56.9 ‰, ε18ON2O/O2= -8.4 ‰, and ε15Nsp= -10.7 ‰. When recalculated for values presented in our study,
δ018ON2O/H2O will range from 22 to 25 ‰
(taking the variations in δ18OH2O into account).
Unfortunately, the δ015Nbulk value for this process
could not be assessed in our study, since the δ15NNH4
was not measured. In case the δ15NNH4 is lower than
0 ‰, the very low δ015N(N2O/NO3-)bulk values may be well explained
with nitrifier denitrification.
Although the interpretation of the relations between particular isotopic
signatures is not completely clear yet, it seems to have potential to
differentiate between mixing and fractionation processes. Note that by using
the literature ranges of isotopic end-member values, they must be
recalculated according to respective substrate isotopic signatures for the
particular study; hence δ15NNH4,
δ15NNO3, and δ18OH2O should be
known. Only the δ015Nsp can be directly adopted.
Progress in interpretations could be made if all three isotopic signatures
would be evaluated jointly in a modelling approach. In order to produce
robust results, precise information on δ0 values for all possible
N2O source processes must be available for the particular soil.
Unfortunately, the complete modelling is not possible for the data presented
here as information on the NH4+ isotopic signature and the δ015Nbulk value for possible nitrification processes is
lacking.
The mapping approach had been used before based on δ15Nsp
and δ15Nbulk to estimate the fraction of bacterial
N2O (Zou et al, 2014). Because N2 fluxes were not measured in that
study, scenarios with different assumptions for N2O reduction were
applied to show the possible range of the bacterial fraction. Here, we
evaluated the mapping approach for the first time using independent estimates
of N2O reduction. Most informative are the relations between
δ15Nsp and δ18O, because δ015Nbulk was poorly known, whereas the estimation of
δ018O is quite robust due to the large O exchange with water and
constant fractionation during O exchange, as shown previously
(Lewicka-Szczebak et al., 2016). Therefore we proposed here a method based on
δ15Nsp and δ18O values to
simultaneously calculate the N2O residual fraction (rN2O) and the
contribution of the mixing end-members as described in Sect. 2.7.3. From Fig. 8 we
can assume that the method works quite well in the case of a significant
admixture of fungal N2O and allows the quantification of its fraction
(fF). For the three treatments where a good agreement between
measured and calculated rN2O is observed, we deal with a
significant contribution of fungal N2O (Sect. 4.2.1). The fF
values calculated here from the mapping approach are very consistent with the
values found based on estimated δ015Nsp only (Fig. 5),
i.e. without considering δ18O values. In the oxic Min soil
treatment we probably deal with a significant contribution of N2O
originating from nitrification or nitrifier denitrification, as supposed
previously from the 15N treatment (Sect. 4.1) and from the isotopic
relations discussed above. The oxic Min soil treatment thus results in rather
poor agreement of the mapping approach results. The combination of these
processes seems to be too complex to precisely quantify their contribution in
N2O production based on three isotopocule signatures only.
Importantly, for Org soil where fF values are very high and
variable with time (see also Sect. 4.2.1), the mapping approach was the only
method to get any estimation of both fF and rN2O.
The other approach, presented in Sect. 2.7.2 and successfully applied for Min
soil, failed for Org soil due to the inability to assess a stable δ015Nsp. Hence, for the case of varying contribution of
fungal N2O, the mapping approach presented here may be the only way of
assessing the range of possible fF and rN2O values.
However, the precision of the results obtained from the mapping approach is a
complex issue depending on the size of end-member areas and variability of η values. We did not aim to determine the resulting uncertainty in the
present paper. The following paper will address the precision problem in
detail (Buchen et al., 2017).