4.3.1. The δ15Nbulk and δ18O
 To interpret the observed isotopic signatures it is necessary to consider the control of isotope effects that cause specific signatures in the sequential reaction members of the denitrification pathway, i.e., NO3−, NO2−, NO, N2O, N2. For NO3− it is known that isotope fractionation during movement into or out of cells is small or negligible [Bryan et al., 1983] and the magnitude of fractionation during denitrification largely depends on the relative rates of uptake and reduction within the cell [Ostrom et al., 2002]; the relative importance of these rates depends on the relative concentrations of electron acceptors (NO3−) within and external to the cell as well as on the factors affecting the reduction rate, i.e., concentrations of electron donors (organic C) and enzymes (NO3− reductase). If NO3− is non-limited in relation to the reduction capacity, then NO3− reduction is incomplete, resulting in isotopic enrichment of NO3− leaving the cells. In contrast, when the supply of NO3− is less than its reduction rate, then little or no NO3− escapes the cells to express an isotope effect. Thus a substantial isotope effect results if NO3− supply is high in relation to reduction capacity of the system, whereas the effect is low or negligible if the opposite is the case. Principally, the same fractionation control applies to the other N species subject to reduction during the further progress of denitrification, i.e., NO2−, NO, and N2O [Barford et al., 1999]. For these species, however, the situation is even more complex because the concentration of electron acceptors of each reaction step of denitrification depends on the rate of the previous step. Furthermore, some microbes are lacking enzymes of some of the reduction steps [Stein and Yung, 2003] which implies that transport within denitrifying species is a precondition of further reduction. The isotopic signature of N2O as an intermediate is thus resulting from two processes, i.e., production during NO reduction and consumption during N2O reduction to N2. N2O that is instantaneously produced is depleted in 18O and 15Nbulk in relation to its precursor. The δ18O is also affected by oxygen exchange with water, where the exchange ratio varies among microbial species [Casciotti et al., 2002]. Reduction to N2 causes increasing δ15Nbulk in the residual N2O [Barford et al., 1999].
 To explain the pattern of observed isotopomer signatures, the temporal and spatial dynamics of the investigated system also need to be considered. It is assumed that each groundwater parcel represents a nearly closed system; that is, diffusion during groundwater flow is small compared to production and consumption of N-gases. Within a denitrifying layer, the concentration of NO3− decreases with residence time or with depth below the groundwater surface (see equation (2)), where the completeness of NO3− reduction can be described by the reaction progress (RP = [denitrified NO3−-N]/[NO3−-N of recharge water]; [Böhlke, 2002]). Close to the groundwater surface, δ15NO3− can be assumed to be close to the NO3− signature of the unsaturated zone, because residence time and thus RP are low. The δ15Nbulk of instantaneously produced N2O (δ15Nibulk) is depleted with respect to the NO3− signature. With ongoing RP, both δ15NO3− and δ15Nibulk must increase. δ15Nbulk of residual N2O is further increased by N2O reduction to N2. The δ15NO3− of seepage water can be assumed to be close to zero or, due to denitrification in the unsaturated zone, slightly positive [Kendall and Aravena, 1999], i.e., not more negative than the NO3− source. Thus δ15Nbulk should be initially negative and then continuously increase during the reaction progress. This pattern is roughly reflected by the groundwater data (Table 2), which exhibited lowest δ15Nbulk for the samples with lowest residence time (December samples at 115 and 135 cm depth: δ15Nbulk = −38 to −42‰) and a large range of δ15Nbulk at a higher level (−21 to +86‰) for the samples of longer residence time. The δ18O exhibit a similar pattern, i.e., lowest values for the 115 and 135 cm samples (23 to 37‰), and higher values for the other samples (48 to 77‰). The correlation between residence time and δ18O was significant (R2 = 0.62, P > 0.95). One sample (July, 160 cm, ti = 148 days) strongly deviates from the temporal trend of δ15Nbulk and δ18O since both values are highest among the data set (86 and 89‰, respectively) while ti is only 148 days, i.e., less than half of the maximum ti (324 days). This anomaly could be explained by a “hot spot“ of elevated denitrification activity within the flowpath preceding this sample: Isotopic signatures increase with reaction progress (RP, see above) which is the product of ti and denitrification rate. Hence, for a given ti, RP is directly related to denitrification rate. When this deviating sample is excluded, closer correlations between ti and δ15Nbulk (R2 = 0.81, P > 0.95) and between ti and δ18O (R2 = 0.89, P > 0.99) are obtained.
4.3.2. Site-Specific N Fractionation
 Owing to the asymmetry of the N2O molecule, fractionation at the central (15Nα) and peripheral N-position (15Nβ) is not equal. It is generally accepted that the stability of the Nα-O bond governs isotope fractionation during N2O reduction resulting in increasing site preference (SP) in the residual N2O [Toyoda et al., 2002; Stein and Yung, 2003; Schmidt et al., 2004]. This has recently been confirmed in laboratory experiments with surface soils [Ostrom et al., 2004]. The N2O production step is more complex: The formation of the N-N-bond during NO-reduction to N2O can proceed via sequential or parallel binding of two NO-molecules to the NO reductase, depending on the type of this enzyme. There is disagreement on the extent of associated site-specific N fractionation, which is reflected by SPiP (SP of N2O that is instantaneously produced, i.e., that is not affected by partial reduction). Stein and Yung  propose positive and low SPiP during sequential and parallel binding mechanisms, respectively, whereas others assume positive SPiP for the parallel binding [Toyoda et al., 2002; Schmidt et al., 2004], and Schmidt et al.  proposed variable SPiP for the sequential mechanism. The variability of SPiP among denitrifiers was recently confirmed experimentally, giving SPiP close to zero for P. chlororaphis and P. aureofacines [Sutka et al., 2004b] and SPiP = 23.3 and 5.1‰ for P. fluorescens and P. denitrificans, respectively [Toyoda et al., 2005].
 SP was positive in all of the laboratory and groundwater samples. Since production and reduction of N2O have occurred during the reaction progress (see section 4.1), SP in the residual N2O results from fractionation during both partial processes. Site-specific fractionation factors of the partial processes given by the difference in δ15N between substrate and instantaneous product [Barford et al., 1999] cannot be determined from our data because both processes proceeded in parallel. Given the relatively large values of SP, it could be concluded that both processes probably contributed to positive SP of residual N2O. However, the possible effect of negative SPiP could theoretically also be masked by fractionation during N2O reduction.
 Similar to δ18O and δ15Nbulk, SP increased with ti (R2 = 0.63, P > 0.95) and the regression between ti and SP improved if the supposed “hot spot” sample (July, 160 cm, see above) is excluded (R2 = 0.96, P > 0.999). This clearly demonstrates that all components of the isotopic fingerprint, i.e., δ15Nbulk, δ18O and SP, reflect reaction progress. Furthermore, anomalies of the temporal pattern might be used to identify variation of process rates within the investigated system.
 The control mechanisms of isotope effects associated with denitrification given above (section 4.3.1) can be used to discuss possible explanation for the discrepancy between observations of high SP of residual N2O in our groundwater samples and relatively low SP of soil emitted N2O. High N2O concentration and/or low N2O reduction rate would be theoretical explanations, since isotope effects can increase with electron acceptor concentration-to-reduction rate ratio. However, the observed groundwater concentrations (0.005 to 2.7 mg N L−1) were within the range of reported pore space concentration (up to 8 mg N L−1 [Heincke and Kaupenjohann, 1999]). Furthermore, the low N2O concentration of samples with high SP (Table 2) demonstrates that reduction capacity was not rate limiting. This suggests that these parameters were not the main drivers. A fundamental difference between soil and groundwater lies in the fate of produced N2O: Owing to high diffusivity in the unsaturated zone a large fraction bypasses further reduction to N2 and is emitted to the atmosphere. In contrast, gases are more or less trapped in the groundwater. Thus there is a relatively high probability that N2O with positive SP leaving denitrifying cells is taken up by N2O reducing bacteria, where partial reduction may result in further increase SP of residual N2O. Consequently, multiple partial reduction cycles might be responsible for the observations of high SP in groundwater N2O. However, further evidence is needed to prove this hypothesis.
 In contrast to the observed field trends, SP in the laboratory experiment did not increase continuously but fluctuated with time. This could indicate temporal variation of site specific fractionation which might be explained by microbial adaptation during the initial phase of the experiments. It is possible that reduction was initially inhibited by a limited density of denitrifying enzymes resulting in relatively large isotope fractionation during N2O production as well as reduction (see section 4.3.1). Establishing anaerobic conditions and adding NO3− probably induced production of NO and N2O reductases which may in turn have resulted in decreasing site specific fractionation during the initial phase. After 30 days, SP of the laboratory incubation follows a positive trend (Figure 3a) similar to that observed in the field which could indicate the end of microbial adaptation. The change in fractionation pattern at this time is also reflected in increasing δ18O.
 The final SP and δ15Nbulk of the laboratory experiment was approximately half the highest field values. The following characteristics of the laboratory conditions are possible explanations for the observed differences: (1) N2O concentrations were lower. (2) Between the sampling events, the headspace gas containing most of the N2O was separated from the denitrifying environment, i.e., the submersed aquifer material, by a diffusion barrier consisting of the slurry (approximately 7 mm in height) and the supernatant (approximately 17 mm in height). This is unlike the in situ conditions, where maximum diffusion length is governed by the pore size distribution and is thus several orders of magnitude lower compared to the laboratory system. Furthermore, (3) the liquid-to-solid ratio was higher than in the field, (4) periodical shaking may have increased availability of organic substrates, and (5) the microbial community may have been altered because samples were not collected aseptically. The first and fourth enumerated impacts both imply a decrease in the electron acceptor–to-reduction rate ratio which potentially lessens isotope fractionation and could thus explain the observed differences.
 The positive relationship between SP and δ18O (Table 2, Figure 2b) demonstrates that the stability of the Nα-O bond was responsible for position-specific N-fractionation and O-fractionation at this site, which is in agreement with earlier assumptions [Yoshida and Toyoda, 2000; Popp et al., 2002; Toyoda et al., 2002; Schmidt et al., 2004]. The large growth of δ18O shows that either oxygen exchange with water played a minor role or the fractionation factors for oxygen are large compared to SPiP and SPiR.