This is a technical paper, describing a “relaxed eddy accumulation” (REA) system to collect air for ¹⁴C analysis, in conjunction with CO₂ eddy covariance systems.  Adding ¹⁴C allows the partitioning of CO₂ into fossil and non-fossil components.  REA is designed to separate and capture air from updrafts and downdrafts, to separate (in this case) ¹⁴C values for up and down drafts, and therefore calculate fluxes. The paper gives very detailed information on the REA system design, as well as carefully evaluating the sources of uncertainty in the system. 

This is, I believe, the first time an REA system has been built for ¹⁴C sample collection, and while REA is conceptually straightforward, it is quite technically challenging to implement.  Thus the level of detail in system design and evaluation is appropriate.  The detail provided would allow others to mimic the system – although my suspicion is that this system is complex enough that there will be few takers. 

The paper is well written, and I have only a few minor comments.  I recommend acceptance with the minor revisions I have noted.

Page 2 line 36.  I agree that EC is the only method to directly measure fluxes.  Given the discussion in the previous paragraph of inventory methods, it would be worthwhile to mention that there are multiple other atmospheric methods available besides EC.  This is particularly relevant to this paper because while this paper presents the first time ¹⁴C has been used in REA, ¹⁴C measurements are widely used in other urban atmospheric methods, such as tracer ratios.

Page 2 line 44.  Human respiration is more typically ~5% of the total annual CO₂ flux, and it depends not only on population density, but on emissions density.

Page 2 lines 49-51.  Suggest rephrasing this sentence, as it implies that in situ ¹⁴CO₂ measurements are available, just not at fast-response.  In fact, ¹⁴CO₂ measurements can not yet be made in situ at all, except a few novel laser-based measurements that have not yet acheived the precision or method development to allow them to be used for atmospheric applications such as this. 

Page 13.  Line 286.  Suggest renumbering the 3 points as 5.1, 5.2, 5.3 to match the following section labels.

Page 14.  Line 308.  Here it says that the difference between two buffer fillings is 0 ± 52 ppm.  In the next sentence, the additional uncertainty is estimated at 0.15 ppm.  I don’t quite follow how this is calculated, and wonder whether the ± 52 ppm is a typo?

Page 21.  Line 470.  Consider adding Turnbull et al 2015, Miller et al 2020 as additional references.

Turnbull JC, Sweeney C, Karion A, Newberger T, Lehman SJ, Tans PP, Davis KJ, Lauvaux T, Miles NL, Richardson SJ et al. 2015. Toward quantification and source sector identification of fossil fuel CO2 emissions from an urban area: Results from the INFLUX experiment. Journal of Geophysical Research: Atmospheres. 120.

Miller JB, Lehman SJ, Verhulst KR, Miller CE, Duren RM, Yadav V, Newman S, Sloop CD. 2020. Large and seasonally varying biospheric CO2 fluxes in the Los Angeles megacity revealed by atmospheric radiocarbon. Proceedings of the National Academy of Sciences of the United States of America.

Page 21-22 section 6.  ∆CO₂ partitioning.  The authors note that the ∆¹⁴C differences between up and downdrafts are small relative to the measurement uncertainty, resulting in many of the differences being indistinguishable from zero.  There is some discussion in this section and also in appendix D about this, but the paper would benefit from expanding this discussion. 

First, the actual ∆¹⁴C measurement precision acheived for these samples is not given, instead the uncertainty in calculated ∆ffCO₂ is reported.  It would be helpful to indicate what the ∆¹⁴C uncertainties are for these measurements.  My estimate from the reported  ∆ffCO₂ uncertainties is that the ∆¹⁴C measurement uncertainties are around 2‰.  Reducing these uncertainties would go a long way to improving the utility of the method.  Several other labs are now reporting around 1.5‰ uncertainty on ¹⁴C measurements, and this modest improvement would make a significant difference to the fraction of usable measurements. 

Since the REA method doesn’t appear to be sample size limited, one could also consider measuring multiple graphite targets to reduce the overall uncertainties.  This would of course come at considerable cost in money and instrument time, but might be worth considering in the future.

Another option would be to make these REA measurements at EC sites that are closer to the surface and/or to sources, so that the ∆¹⁴C differences are larger.  This would certainly be useful in demonstrating the technique, and could be a necessary constraint for the foreseeable future if ∆¹⁴C uncertainties cannot be beaten down further.

I don’t suggest that the authors try to implement these things in this paper, but some discussion of these points would be helpful.

Page 23.  Lines 508-518.  The same comments as above apply – indeed the signal-to-noise seems to be the main challenge.

Page 23 line 522.  Including CO and/or other species in these analyses would be very interesting.  I wonder if incorporating CO measurements could also help resolve the signal-to-noise issues?

Review on „A relaxed eddy accumulation flask sampling system for 14C-based partitioning of fossil and non-fossil CO2 fluxes“ by A.-K. Kunz et al submitted to AMT

The manuscript describes an innovative set-up that would allow, presumably for the first time, to use the relaxed eddy accumulation technique and laboratory analysis to measure 14CO2 fluxes and ultimately infer fossil vs. non-fossil CO2 fluxes between a mixed-source land surface like a city and the atmosphere. The amount of consideration and effort that went into the set-up, processing chain and various additional quality control tests is impressive. Finally, the language and figures are of high standard.

The only issues I found concern clarity, and are briefly summarized here (no need to reply since they all re-occur in the detailed comments):

Intelligibility: The length and content of the methods section suggest that easy access for readers approaching the interdisciplinary method from various backgrounds was aimed at, however this was not always successful.

Aims and scope: Large parts of the manuscript, maybe unconsciously, raise expectations to go further towards partitioned fluxes than is actually done.

Systematization of QA/QC: The way the large amount of quality checks and the reasons for discarding measurements are described, can cause some confusion (as to what was done why and what wasn’t, and with respect to the “final” number of 103 measurements mentioned in the abstract).

Detailed comments with line number:

L13: At this point the mentioning of quality control flask pairs is quite cryptic to the reader. Depending on length limitations of the abstract, consider to either briefly explain what they are about, or reword like “112 flask pairs in total (103 for real-world fluxes and 9 for quality control purposes)”.

L93: “with a dynamic deadband” sounds a bit arbitrary, maybe briefly refer to the standard deviation of w again (L72), or clearly define there that “dynamic deadband” in the rest of this manuscript will always refer to

L118: Given that the description is otherwise very detailed and inclusive for readers from outside the isotope community, wouldn’t it be consistent to also briefly clarify the normalization process, e.g. with another equation (I guess it would be the first one in sect. 2 of the cited source)?

L126: abbreviating Delta^14 C by ^14 Delta looks a bit arbitrary and inconsistent. If it is rooted in common nomenclature of the isotope community, briefly mention it, if not consider a more consistent solution, like: Did I need the delta^14 C above in the first place? If yes, do I really need the abbreviation? If yes, can I at least keep the order between 14 and Delta consistent?

L142: “14Δphoto is best approximated by the current atmospheric signature 14Δmeas”: Again, more explanation needed, it does not get clear why the current measured mixed signal should best approximate the pure signal of one of its components.

L146: “and set to 10”: This statement only makes sense once the reader knows that measurement uncertainty is much smaller than that, maybe clarify by adding something, like at the end “which is well above the typical measurement uncertainty as will be shown in ….”.

L172 ff. and Fig. 1: The order of introducing the buffer tanks before the loop system in the text, depicting the latter in Fig. 1 without any mentioning of a pump and then only clarifying much later in Fig. 2 is confusing. Fig. 1 as it is shown now would not work – readers must assume that the only pump in the system is the one mentioned in Fig. 1, which would however leave unclear how air can leave the outflow (rather than being sucked in) and why all valves before the pump are closed at the moment of sampling.

L185: The photo reference is confusing, readers would first expect a reference to Fig. 3 which shows the inlets. Fig. 3 suggests that the 2 inlets were at the same height and horizontally displaced, while Fig. 1 suggests they are vertically displaced. Even if the figure is difficult to fix in that respect, it would be good to state the arrangement in the text (e.g. near L170).

L208: Out of the 3 valve state abbreviations NO, NC and CO mentioned in the Figure and caption, only one (NC) is referred to in the text. Establishing a better connection between figure, caption and text could help readers understand the system.

L221: tr is introduced as “an air parcel needs the time” above the equation and then re-defined as rinse time two lines later. If it is the same, change the wording here, like e.g. “we refer to tr as rinse time in the following because it is exactly time the sampling needs to be artificially delayed to avoid sampling air from before the event”.

L238: maybe for discussion somewhere (not necessarily here): For which of the other gases would the REA system also be of interest due to a lack of fast analyzers? L240 “remaining”: Does this only refer to the standard ICOS procedure or also to this manuscript, i.e. were the Zürich REA samples also analyzed for the other gases before extracting the CO2 for 14-C-analysis?

L274: What does the lower flow rate at the MFC during rinse time imply? In hindsight, shouldn’t then this be the flow to use in calculating rinse time? Or was the rest assumedly going through the outflow?

L290: What is excluded by the words “only” and “focus”, and why?

L312: Reference to appendix B3 missing?

L322-323: It remains unclear how (and why?) only using the MGA here relates motivation-wise to comparing to both the IRGASON and MGA later as described near L390.

L398: Clarify “1/t flow rate”

L432: How does the number 102 relate to the 103 from the abstract? If the difference results from the discarded measurements mentioned in the next sentence, why are they mentioned in plural?

L453: What “additional CO2 density output”? The text before suggests that raw CO2 density was available, and so were probably sonic temperature and H2O density, which is all that is needed to compute fast-response air temperature offline. An example for dry molar fraction is given in the appendix, but it can be done for any other measure of CO2 “concentration”, and with even less slow-response data (basically only p), using H2O density from the IRGASON instead of q, only solving the equations gets a bit more complicated then.

L460: 103 *selected* REA flask pairs seems to refer to the earlier mentioned fact that due to the costs of the 14CO2 analysis more flasks were taken but then carefully selected for the analysis. Then it is not understandable why subtracting so many more afterwards, except maybe the (unforeseeable) loss during graphitization. The 8 not analyzed could (as it looks now) as well be included in the selecting process from the start (thus lowering the number of 103), while for the 4 mentioned last it is unclear whether they were analyzed for 14CO2 and only discarded from data analysis afterwards. The way it is described now leaves unclear why the number 103 is important, as well as how objective the criteria for omitting the last 4 samples were.

L484 and L520: The paper stops somewhat abrupt, with everything on the table (at least according to how it is described in the methods section) to compute estimated fossil fuel fluxes but not presenting any. It is understandable that given the huge effort behind these measurements and the already long paper, the authors want to publish two separate papers about the methodological groundwork and the actual results. But then the title, abstract, parts of the introduction and methods section, frequent use of the word “partitioning” and presence of appendix D in this manuscript should be carefully reconsidered. For example, everything about estimating the deltas of other components seems to be needed here just for uncertainty estimations, and could more logically go into a later paper presenting the fluxes (together with their uncertainties).

L469-470 and 506-507: If data points to the right/below of the 1:1 line indicate respiratory and other non-fossil fuel signals and data points to the left/above photosynthesis, this means that any data point near the line could also result from simultaneous photosynthesis sinks and respiration/biofuel sources instead of from fossil-based CO2? This question is out of curiosity, and probably more relevant for a follow-up paper on the fluxes than for this one.

L533-534: “water trap operated in reversed order” = (custom) dewpoint generator? Were the 12°C a setting or an uncontrollable result of the device’s setup? How does the humidification avoid surface effects?