|Flux event sampling|
I realize logistics, manpower, communications, were a major issue, however, some event driven sampling days would have been beneficial (eg adopting a hybrid sampling strategy [standard weekly plus targeted sampling post manure/fertilizer; even from 1-2 sites per land class), particularly when we know that annual management events (N inputs, tillage, etc) are major drivers of GHG emissions, and that these systems had not been well investigated. I agree that low N inputs would likely not have resulted in major fluxes, however this approach may have given greater power to identify land class differences, without undue bias for capturing increased flux in the more intensively samples sites.
Consider not including ‘soil respiration’ CO2 data from these systems, particularly the grazed sites. It is more appropriate from the bare soils e.g., between crop rows, but even here, the relatively long chamber closure times are less than ideal. Non-event based sampling also argues against its inclusion (see above). You use CO2 as a ‘proxy’ for determining vial integrity for the other gases; I suggest this should be its only use in these chamber methods, unless you plan to use the CO2 data for e.g., building up a C budget or broader GWP for these systems?
Please be consistent with flux unit order (m-2 h-1, not h-1 m-2 is more typical).
Consider being bolder, i.e., more geographically broad. Given that you argue (and I believe appropriately) that the 100km2 area in western Kenya is representative of the highlands in east Africa, maybe this should be reflected in title.
Reconsider “limited”; restricted, impeded…. maybe just “low”?
32-34: This appears to be just tacked on at the end of the abstract. You don’t mention GHG mitigation or food security as issues previously. If emissions are very low now, would increasing inputs mitigate them further? I can see that increasing e.g., N inputs could lead to a lowering of N2O intensity (ie emissions remain low up to a point below crop N uptake requirements, but yield increases disproportionately). Is this what you are implying; if so, could you make clearer, and maybe bring in the concepts of mitigation and food security earlier in the abstract, as opposed to the focus on measuring for improved coverage and potentially more accurate inventories.
38: “the N2O”?
55-56: What is “proper”? Consider ‘this shortage of baseline data makes it difficult….”
57: IPCC Tier 2 methods can be broader than just country specific, e.g., regional, or crop specific.
63: Check out Shcherbak, I., Millar, N., Robertson, G.P. 2014. Proc Nat Acad Sci. 111,9199–9204, here and elsewhere for improved estimate of N2O emissions at low N rates using a non-linear approach.
66: S and B, 2006 is a meta-analysis. Would seem inappropriate for citing this single factor; replace?
74-77: Can NDVI be considered a ‘top-down’ approach? I understand these methods to involve using global budgets and atmospheric concentrations to help determine emissions from various processes and sectors (eg Crutzen et al. 2008. Atmos. Chem. Phys 8, 389–395, Davidson, E. A. Nature Geosci. 2, 659–662 (2009). Please check, and clarify/revise as needed.
78-80: I do see how farmer surveys could inform N management and therefore act a as a proxy for N2O emissions, however, I understood ‘bottom-up’ to refer to field based (eg chambers) measurements that are used to develop emissions factor that are then extrapolated over time and space to estimate larger scale emissions. Eg, see Reay et al. 2012 Nature CCDOI: 10.1038/NCLIMATE1458. Again, please check and clarify/revise as needed.
85: Quite the contrary, I’d suggest. Kenya seems ‘over-represented’ in the dearth of ‘African’ GHG literature at least wrt Table 1 (3 publications). I’d agree that smallholder farms are indeed poorly represented, as indeed are many landscapes in SSA. NB consider adding another Kenyan study, Millar et al 2004 Global Biogeochemical Cycles 18:GB1032. doi:1010.1029/2003GB002114 to refs and Table 1.
M and M
106: I’ll admit to not reading Sijmons et al. 2013, but could you justify in a sentence or two how East African tropical highlands data can be extrapolated to other regions worldwide?
117: Can you estimate the range of N inputs associated with 100 kg manure. Although you note that EF generation is not a primary reason for the study, you do mention inventories and IPCC Tier 2 in the text, and of course N inputs related to N2O emissions is the basis for EF determination. I’d encourage you to reconsider estimating EFs and assigning one or more for the highlands region; at least it could be seen as a useful start; without them, I think the impact of all this work is reduced? Also, later in manuscript (L 584) you mention 100 kg C input with manure application; what is correct?
127-186: Phew, quite an undertaking. I appreciate the explanation, and I understand the rationale and the approach (to some extent!), but where there no ‘simpler single’ ways of identifying land type/parcels (eg soil class/type, rainfall, temperature, or a combination of these; understanding that they are likely inherent/incorporated in your method). Given that you didn’t find too much in the way of differences associated with your divisions, is there a more straightforward approach to ‘retroactively’ define a better relationship with emissions (and by that I mean one that can identify differences!)? Or indeed, as you note (and is well accepted), are N2O emissions primarily driven by management (N inputs). Maybe surveys looking at manure inputs/types/analysis would be useful (do they exist?). Please comment.
187-213: I’m not convinced that the incubation adds sufficiently to the paper to be included. I think I understand that you were looking here for GHG emissions potential based on manipulating WHC? Given that you do record soil moisture in the field (was this used to estimate WFPS and correlated with N2O?) is its apparent use as an assay for field hot spot sufficient, or indeed justified? Could you justify more please? Also check out Bouwman 1998 NATURE|VOL 392 | 30 APRIL 1998 for overview on WFPS and N2O in trop ag.
209-213: What was the time interval between successive flushings?
How was the rate N addition chosen? Were other nutrients replaced?
226-229: See CO2 comments.
231: Use of fans in ‘small’ chambers typically not recommended (pressure gradients induced/altered. How were the fans used (e.g., timing, duration of operation)? Could be a major issue. Please comment on how your fans did not cause these problems.
233-238: Could you expand on how this bias was calculated? Were chamber temperatures recorded during deployment? Maybe better in flux calculation section?
239-255: I’m aware of the Arias-Navarro paper, and accept/agree with your rationale.
256-273: Please include a flux calculation equation for clarity.
261: Please clarify why a 10% decline was chosen as a measure of detecting vial leakage?
261: What was the decline unit? (concentration [at consecutive time points], slope [between consecutive time points])?
262-263: Unclear of process here. Irrespective of position of point (2 ,3, or 4) were they thrown out? (all 3 gases?) if they failed the 10% test?
264: What is the instrument precision? How was it calculated? How does this differ from flux detection limit? See Parkin et al. J. Environ. Qual. 41:705–715 (2012). Please include a flux detection limit for your GC (see below).
264: I think a better approach is to report the flux detection limits (± x ug N2O-N m-2 hr-1) and include all fluxes in the analyses that fall within these limits, rather than zeroing. This zeroing may bias the results if many more of one sign (+ or -) exist. Please clarify/confirm that zeroing has no effect.
269: Were other non-linear fits tested? Please rationalize your choice of 2nd order polynomial? eg see Venterea 2013. Soil Sci. Soc. Am. J. 77:709–720 for other options.
269-270: How were the threshold R2s chosen? R2s not always useful when fluxes (slopes) are low. Consider using F/t tests to determine slope differences from zero.
310: Presume labor restrictions/accessibility limited harvesting ability? Could you clarify why these 9 chosen
316-317: Note equipment vendor as before.
317-321: How was growing season determined/defined; presumably different duration for the crops/systems investigated? Please clarify.
Were emissions in the ‘off-season’ (after harvest and before next crop planted) included?
Does above ground N uptake = grain N (i.e., harvested N)? Maybe emissions intensity (yield-scaled) calculations/text better in flux calculation section?
366-383: The pH, C, N, and CN values don’t appear to match with Table 2. I understand that table 2 is an average of the multiple sites (could you note the number of sites per land class in the table), however the errors (presume SEM; again, please note in Table 2) seem too small for the ranges you note in this section?
384-451: Other results data introduced here. Inorganic N concentrations, C:N data, crop yields, soil core GHGs. Could you please re-visit and better parse out this information into discrete sections.
389: “two weeks later” or (17 March 2014)
396: Very low mean value (1.6). Is this flux ‘detectable’ (i.e, above the detection limit) on your GC? Irrespective, as noted earlier I believe all flux values should be included in linear interpolations and not zeroed. Please include a flux detection limit in your manuscript.
443-444: mung beans and green grams synonymous, but probably should be consistent.
509-510: Clarify how you determined/calculated this. Were you able to assign a manure/fertilizer N input for each site, or even some of them, and therefore tie N input directly with N2O emissions If so, this should be included in manuscript.
510-511: Possible, even likely, but you didn’t sample frequently enough (as you note in 514-518) to conclude this – rephrase.
516-520: This works both ways. You may have missed even lower positive (or negative) fluxes with your weekly sampling (cold moments!) I think better to use N2O peak or pulse event rather than “hot moment”. Agreed that one off large pulses cause consternation when interpolating to seasonal and annual emissions estimates, but I don’t think “incorrect” is the word. Its why when you’re uncertain of the emissions you expect from your systems that a ‘hybrid’ sampling regime is the most judicious, so that you can capture the highs and the lows so to speak, as well as the ‘background’ emissions.
523-526: Maybe not so much that you did not miss these events rather, even if you did miss them your approach is not bias, and your systems can be compared. Over an even longer term study (many years) of these systems your weekly sampling regime would effectively capture the management and event based pulses, just by sheer number of times sampled. Note, if I interpret correctly, Hickman may not have shown differences between N rate treatments, but he did show peaks/pules of N2O closely following N fertilization vents even at the lowest N rate treatments (Fig 1 d,e).
527-533: Is total C and N content of the topsoil a useful parameter here? Could you include more discussion on other factors both from your sites (eg do you have SOC and SON values?) and from the literature. Grazed land, therefore presumably animal manures added, patchily and in large quantities historically; ties in with hotspot theory.
537-539: Are the Brazilian and Chinese studies on smallholder farms with similar low input management and productivity? Are there similar studies not in SSA?
547-556: Useful information; I think some of this could go in Methodology though. Try not to fall on your sword too hard!
557-564: See also M and M (187-213). I think using the soil core incubation results as a proxy for predicting field hotspots is a bit of a stretch; I’d remove that claim. You appear to be conflating annual emissions with hot spots; I’m uncertain if land-use type and GHG emissions has been tied to prevalence of hot spots, which appears to be what you are saying? I would agree that they could be useful (and have been used) as an assay for getting at comparisons between land-use types, that are not well known or quantified.
598-601: Agreed. See Shcherbak et al 2015 (above). N2O emissions are low at low N rates until crop N demand quenched (~150 kg N ha-1).
605-608: See Vitousek et al. SCIENCE VOL 324 19 JUNE 2009 for global N (nutrient) imbalances.
Can’t read/examine them easily - spaces, indents?
Table 1. Poorly presented; maybe use lines, shading or other for column/row boundary clarity. Column headings don’t align well with column data/text in some cases.
Fig 1: Given the efforts that went into land stratification, I was hoping to see map of the 100 km2 landscape divided into the 5 land classes along with maybe cover class, field type and site location. Is this not possible; shape files? Please include. Google earth view is not particularly appealing; try to exclude their insignia?
Fig 2: This may not be your fault, but Fig 2 (N2O; 3rd panel on right), cut off on my pdf version. I don’t see the land-classes noted. Please include here and in the legend.
Fig 3: I don’t see flux units for the three gases on the y axis (please include). Also increasing the vertical height of the GHG panels would be helpful (more than the SWC and IN); they’re very difficult to examine. Please include (shortened) land class names in panel key. Also label panels 3 a, b, c…etc.
Fig. 4. My version is of poor resolution. The 59 sites appear to be split by land class, field and crop type – please revise legend. For clarity, please label in graph and in legend what the land class (1-5) and field type (1-3) are. What do the error bars represent? What defines the outliers?