General comments
The authors addressed all comments and solved most of them. The additional data presented are very valuable to better show the potentials and limitations of the new approach. But the data and some of the author’s responses also revealed even more the uncertainty and maybe even bias of their approach. Nevertheless the data-set is valuable despite those uncertainties in view of the paucity of in situ denitrification data and due to the fact that all in situ reports so far are subject to uncertainty or even bias to some extent (e.g. Felber et al., 2013). But it is necessary to better highlight and discuss these limitations also in view of the recent paper showing the detailed fluxes from this study (Sgouridis & Ulla, 2015). Moreover I encourage the authors to add some conclusions how the approach could be improved in future studies. A major problem still associated with the presented data is how the authors dealt with the question of linearity in 15N fluxes during 20 h of chamber enclosure. This was one of my main concerns in the first review. The reported R2 is not an appropriate measure for linearity in view of the uneven sampling intervals. So there is needed for some more data analysis to address this issue. Based on that and due to the further points addressed below, the limitations of the approach must be fully reported and discussed and also addressed in the abstract.
Non-linearity of fluxes (N2+N2O and N2O):
I strongly disagree with the view of the authors here. First, their basic assumption that 15N fluxes are not affected by decreasing concentration gradients due to the high N2 concentration in the atmosphere is not correct, because due to diffusion physics each of the N2 isotopologues (ie 28N2, 29N2 and 30N2) acts like an individual gas species. Hence, for the 15N gas flux method subsoil and lateral diffusion is even more relevant compared to total N2o fluxes since production of enriched N2 is limited to the labelled soil volume, whereas N2O can be produced anywhere in soil.
Second, testing linearity of fluxes over time using linear regression functions assumes equal sampling intervals which was not the case, since the three intervals were 1 h, 1h and 18 h. Hence any event when 1 and 2h values are similar but lower than the 20 h value must lead to R2 close to 1, irrespective of linearity. Assuming the analytical uncertainty was negligible, linearity could be shown by comparing flux rates of the 3 intervals, i.e doubling 15N2 concentrations between 1h and 2 h and increasing tenfold between 2 and 20h. From the individual data (Table S4) it is evident that fluxes were close to linearity only in few cases. In most cases, 2h values were less than twice 1h values or were even lower than 1h values. 20h values were sometimes much higher than tenfold 2h values, in other cases much lower or even the lowest values of all samplings (eg C-MW3). This probably reflects the combination of several antagonistic artefacts of the approach (see below). But analytical errors must also be considered here, especially when 1 and 2 h values were close to detection. In consequence, the results and discussion sections on (non-)linearity must be completely rewritten and must address also individual site observations, not only trends of land use averages.
Non-homogeneity of labeling:
Ten injections for an area of 0.05 m2 (ie 50 cm2 per injection!) is clearly suboptimal, but it is clear that the ideal number of injections suggested by Wu et al was not possible due to the large number of sites and measurements. The 15XN data clearly reveal that distribution of the label was far from homogeneity, since 15XN was always close to the 15N enrichment of the added tracer (98at%) while the expected average 15N enrichment of the soil N2O was about ¼ of 15XN (around 20-30 at%). Hence only about ¼ of the soil volume contributed to the dilution of the added 15N tracer by soil NO3 and consequently, only this small volume was 15N labelled and thus contributed to the fluxes. This must clearly lead to a substantial underestimation of N2 fluxes.
Moisture effect:
First , there is need for clarification on the reported data. In the methods, the authors mention that the increase in moisture was equivalent to less than 2 mm precipitation, but they report up to 5 % increase in volumetric soil water content, which is equivalent to 5 mm given the soil depth of 10 cm. Moreover, the reported maximum soil water volume is > 5000 cm3, but this is more than the total soil volume of 5000 mL (0.05 m2 * 10 cm). This needs clarification. But anyway, the injection of up to 200 mL in 10 injections over 10 cm depth would initially saturate a small cylindrical volume and there would be a distance of several cm between these saturated cylinders of the 10 injections. A thumb rule of soil hydrology is that up to two days are necessary to reach field capacity after full saturation. Hence, depending on texture and structure, it would take at least several hours until the initially saturated volumes equilbrate toward homogeneous water content in the amended soil. The authors started chamber closure immediately after injection (“Following the 15N tracer application the collars were covered with the acrylic chamber fitted with a rubber septum for gas sampling”), ie did they did not wait until water might have equilibrated before starting their measurements: I suspect their idea was to avoid substantial plant take up of the small amount of added NO3 (would be good to explain that). Hence there might have been initially favorable denitrification conditions in the saturated injection cylinders diminishing over time. These moisture-induced denitrification events would be associated with synthesis of denitrification enzymes, were the N2O reductase is known to be typically slower compared to the enzymes of the preceding steps. Hence there might be overlap of two opposing effects: decreasing saturation (=>lowering total denitrification but increasing its product ratio due to faster gas exchange) and increasing enzyme density (=>increasing total denitrification but lowering its product ratio). Depending on texture, structure and abundance of substrates, the extent and balance of these effects must be variable. Moreover, these moisture effects interact with the impact from extended enclosure (20h), which must lead to decreasing fluxes over time due to decreasing gradients and increasing subsoil diffusion. This in combination might explain why the temporal trends of N2 fluxes and of the product ratio were not consistent among sites as described above.
Repeated measurements after 15N labeling might reveal the extent of the abovementioned moisture artefacts. So in case the authors did repeated measurement (e.g several closure periods during one week) in some cases it would be valuable to show results.
Despite these limitations the data are valuable and worth to be published. As mentioned before, there are no unbiased denitrification field flux data until now. There are hundreds of papers published with the invalid AIT and the approach proposed here is certainly far more reliable compared to the AIT data which is also shown by the comparison of this study. Today we have no better method for measuring denitrification in the field and in the presence of plants than the 15N gas flux method. It’s accuracy and precision certainly increases with the effort take to label homogenously without augmenting nitrate too much and by measuring as often as possible. But clearly there are limitations if several sites are investigated repeatedly as in the present study. Therefore the presented data-set must be seen as an attempt for a best compromise when balancing data quality and quantity. But a very valuable contribution to progressing denitrification methodology can only be obtained if the authors evaluated and discussed the different source of uncertainty and derived some suggestions for future improvements.
But how could the uncertainty be addressed? I suggest to calculate the rates for each sampling interval and compare them. Based on that discuss the factors mentioned above and come up with possible suggestions for future improvement (more injections, allow water equilibration, shorter and equal time intervals). The discussion should address the different patterns. E.g. highest rates during last interval might suggest delay of enzymes, highest initial rates dominance of moisture effects. You might check if you find patterns e.g. showing that moisture effects were more relevant in loamy soils, enzyme effects more in soils with low denitrification rates etc. Same for the product ratio: decreasing product ratio might reflect a combination of increasing N2O reductase and N2O accumulation during closure. Increasing ratio could be due to decreasing moisture and/or initiation of a denitrification event with delay of N2O reductase. I think your data-set shows clear lag-phase in many cases, indicating that fluxes started only when enhanced moisture lead to O2 depletion and/or when sufficient enzymes were produced. I would mostly trust rates and ratios of the first sampling interval in cases when conditions were favorable for denitrification before injection due to coincidence of sufficient moisture, nitrate and reductants, because then the injection would not change denitrification too much. In cases when only moisture limited denitrification the last interval might be more reliable because injected soil water was better equilibrated. Under limitation by nitrate only, the tracer injection would have complex effects. It might be useful to check whether the time pattern of rates and product ratio shows some similarities among soil and land use types.
In summary, before the paper can be accepted the authors must show and analyse the temporal change of N2 fluxes and product ratios and discuss their method based on that, addressing the abovementioned (and maybe further) aspects. This must be addressed in a separate section of the discussion and also summarized in the abstract. I also recommend to mention that this detailed uncertainty analysis complements the data presented in the recent paper (Sgouridis and Ulla, 2015) since that paper did not address these aspects.
Details:
L 267: moisture effect < 2 mm equivalent is incorrect in view of 5 % vol water content change : 5 % of 100 mm = 5 mm
L 271 there was immediate enclosure and sampling after labeling (see general comments).
L 238 -250 only 10 injections for 0.05 m2 not enough(see general comments)
Table 2: the fact that 15XN by far exceeded expected enrichment of total soil NO3 demonstrates huge non-homogeneity of labeling. The small number of injections apparently caused denitrifying hot spots in the injection area with 15XN (0.8 to 0.9 on average) close to the enrichment of the tracer solution (0.98) but far from the NO3 target enrichment (0.13 to 0.25). Note that due to imperfect distribution of tracer solution the local increase in water content was far more than the average of 5% (which is still quite a lot) (see also general comments). So the non-homogenity of the label is an indication that the moisture effect on 15N fluxes was much larger than expected from the increase in average water content in the entire soil.
Table S1: Soil water numbers are questionable (up to 5 L) since the volume of labelled soil was 5 L only. Please check.
Table S6: the fact that there were no clear time trends for the product ratio probably shows the overlap of several processes (see general comments)
L 652-657 the conclusion with respect to hybrid N2 or N2o is incorrect (see Spott & Satnge 2007 and Spott et al., 2011): hybrid N2 and/or N2o would be proven by 15XN was lower than 15N atom fraction of NO3 but not from the deviation between 15XN of N2 and N2O. In fact the fraction of hybrid gas could be different in N2 and N2O fluxes which could lead to different values in 15XN. But this could not be determined due to missing 15NO3 analysis and the large non-homogeneity in labeling.
L 535 to 538 this statement is not well justified. Your precision for R29 and R30 is in the same order compared to previous studies including as early as Siegel et al., 1982 (see comparison of precision in Well ea 1998). So please formulate more cautious or give exact numbers in identical units (eg. Standard dev for R29 and R30) to show to which extent your analysis was better.
L 563 to 567 it is not well clarified what this means. Suggest: “the soil cores or slurries were incubated in fully enclosed systems and were thus not affected by potential bias from diffusion of evolved N2 and N2O to the subsoil (Clough et al. 2005). But please check if the reference still fits to this modification.
L 570 -572 this is indeed by no means the case (see first general comment). So you have to keep the possibility that increasing subsoil diffusion during extended chamber closure was a potential source of bias.
L 681-684 this would not only result from subsoil diffusion of N2O but also from enhanced reduction in the topsoil due to increasing N2O concentration during extended cover periods.
L 734 please cite also Bollman & Conrad 1996, who were the first to show the artefacts by catalytic NO decomposition and to clarify that this artefact is known since long.
In the entire manuscript: use consistently the correct spelling of the product ratio: N2O/(N2+N2O), one or both brackets were often missing
Details to the response file with marked changes of the text.
L 682: suggest: “ to maintain natural drainage and root growth during the measurements” since natural drainage is also needed if the ground water table is far below
L 699 delete “equal” since 4*6 is not equally spaced. A pattern with triangles of equal side length would be optimal. So the distance between your injections varied between 4, 6 and about 7.5 cm, isn’t it?
L 881-895 these statements are not justified, see general comments
L 913-914 but this statement only applies for landuse average, whereas individual sites could have any pattern. Please be more detailed here and explain that there was no consistent pattern for all sites.
L 1070 to 1073 sentence not clear to me. Do you want to highlight that you could detect fluxes in view of low enrichment? But in fact your active pool was close to the enrichment of the tracer solution since 15XN was around 90 at%. So you can’t state that your method worked at low enrichment.
Conclusions must be partly rewritten:
L 1257 to 1260 not clear to me why this is related to smaller sample size. In fact you improved analytical (IRMS) precision somewhat , but not greatly. Also your fluxes came from highly enriched pools
Pleas add some conclusions on the aspects raised in the general comments |
- Printer-friendly version
- Supplement
