This paper describes inflight measurements of atmospheric methane, which are particularly challenging, but can provide insights on the regional methane budget and on the main local methane sources. The application of the multi-Keeling model regression is of great interest and has been found useful to define the local background, given the difficulties in specifying a background in an area with such a multitude of sources.

One of the aims of this study is the attribution of new isotopic signatures to methane sources, as stated in few parts of the paper. However, I would strongly weaken this statement, as the little CH4 enhancements between samples lead to a very high uncertainty and therefore a large isotopic range. I would focus the study more on the identification of potential new sources that are not accounted in the inventories and on the quality of the measurement technique. I would also add a paragraph explaining how to better constrain the source isotopic signatures (e.g. collecting more samples to constrain better the keeling line? Is it possible to collect a smaller bag than 3L ? Perhaps explain better the reason why more samples could not be collected. I am not an expert of inflight measurements, I would need more clarification in the text).

Another issue that I think should be addressed more is the mismatch between the samples collected at different heights. It looks that in some cases there is a mismatch between the calculated footprint area and the observed area, some sources might have entered the domain and some other maybe not included. It is not the scope of this paper, but for few sources, forward modelling would help to see if some emission plumes would have been captured during the flight.

Overall, the method and results are thoroughly described, and given the importance of the findings included in this study, I would recommend this manuscript for publication after addressing the issues that I mentioned above and the following comments:

Abstract: it is too long. I am not sure there is a word limit but I think it could be heavily shortened.

Line 82: BU, I don’t think the acronym has been explained in the text above. Expand for readers who don’t know what you are referring to e.g emission factors x statistics.

Line 122: how can these challenges been tackled?

Line 197: “distributed sources”. These are explained later in the text, but I would move some details here as the reader might be confused by the term “distributed”.

209: refer to the Figure including also the symbol color to help the reader i.e. “The largest individual source in an open pit….red square in Fig 2a.

215: I was wondering how Fig 2b was created. Then you explained that later in the text. I would mention briefly about the isotopic signatures attribution here and then describe more in detailed in the following paragraph.

239: see my previous comment

245: why? Can you include a reference?

260: perhaps there are no studies on termite in this area, but I think there are some isotopic values in literature that you can refer to and you can include here (Monteil et al. 2011?).

Figure 3 b: what do the lines represent? The confidence bands? State that in the Figure caption.

349: instead of using only a visual identification of outliers, I would quantify them using a statistic approach, so that the identification is more solid. It is not clear to me just looking at Figure 3 how these outliers have been selected.

373: add “see appendix X”.

415: include here the Neininger background value.

533: again refer to the figure colors. “within the range listed in table A2 , grey in Fig 6”

Fig 6: include a figure title for each plot “Gracing Cattle; Feedlots…”

Line 555: the isotopic signature was…(blue line)

Line 562: maybe for the high altitude samples the footprint is different and you see different sources? See my previous comment

588: again include the line color

526: Fig 5 a?

740: also some atmospheric transport modelling would address this issue.

Review of “Atmospheric methane isotopes identify inventory knowledge gaps in the Surat Basin, Australia, coal seam gas and agricultural regions” by Kelly et al., for publication in Atmospheric Chemistry and Physics

The authors used measurements of CH4 and δ13CCH4 from samples collected during airborne surveys across the Surat Basin in Australia to study CH4 emissions from different types of sources in this region. This work is a follow up of two other studies: Lu et al. (2021) that estimated δ13CCH4 of single methane sources in the region using ground-based surveys and Neininger et al. (2021) that developed a bottom-up (BU) inventory of the CH4 emissions in the region. In this study, the authors identify in-flight atmospheric air samples with a predominant type of CH4 source based on a combination of HYSPLIT back trajectory footprints and the BU inventory developed in Neininger et al. (2021). They regroup samples with the same predominant CH4 source and sampling height into different sets. These sets are used in a multi-Keeling-model regression to estimate CH4 and δ13CCH4 of the background first, and then these background estimates are used to estimate δ13CCH4 of the different sets of samples.

This manuscript is very ambitious as it aims at assessing the quality of the BU inventory, extending the knowledge of δ13CCH4 from sources difficult to access with ground-based surveys, and identifying mitigation opportunities using these samples containing low signals from a mix of CH4 sources. This is a laudable goal and will be of great interest to the journal’s readership as the state of knowledge of coal seam gas (CSG) and cattle farming emissions and isotopic signatures is not very advanced. However, the manuscript currently requires multiple improvement before final publication. I recommend publishing it after addressing the comments listed below:

Major comments:

• The manuscript needs a bit of reorganization:
  • The end of the introduction needs to be reworked.
  • Some subsections from results could go to the “Material and methods’ section.
  • The ‘Results’ and ‘Discussion’ sections should be merged to avoid the repetitions of the results and the discussion teasing (all of the sections talking about the background could be together: 3.1, first part of 3.3 and 4.1.1 for example).
• Getting different isotopic signatures for a given source depending on the sampling height is very concerning regarding the ability of the method to correctly assess δ13CCH4. It is even more concerning since this problem occurs for the two type of sources. I would expect IPAA samples collected at lower heights to have higher methane signals since they are sampled closer to the sources and therefore making it a bit easier to distinguish methane sources with isotope techniques, not the other way around.
• Additional analysis/discussion should include:
  • The uncertainty on the estimated background values with the multi-Keeling-model regression (1.825 +/- 0.037 ppm and -47.3 +/- 0.3 ‰): Are these uncertainty include in the Keeling regression of the different sources? How does that affect the Keeling regressions for the different categories? If δ13CCH4 of the background signature was -47.0 ‰ or -47.6 ‰, this would impact all of the other sources signatures.   
  • Footprint calculation: is 2 hours enough? A footprint calculated from 250-350 mAGL will likely be larger than 100-200 mAGL due to the increase of wind speed with altitude. Maybe 2-hour BTF is not enough to properly capture sources that influence measurements at lower heights.
  • The IFAA samples contain a mix of methane coming from different types of sources, I am wondering if the 50% threshold to attribute a sample to a category is not too low (see below).
  • Could be interesting to try to see how different would be the results if 100-200 mAGL and 250-350 mAGL sets were merged for CSG and grazing cattle.
• The story about the outliers and the identification of potential mitigation opportunities is a bit wobbly… I am not sure that associating these 3 outliers because they seem to align is really convincing.

Detailed comments:

L35: I would add ‘based on a bottom-up inventory developed specifically for the region’ after ‘… could be attributed to a single source (CSG, grazing cattle, or feedlots)’ or something like that.

L60: Introduce the notation CH4 for methane here instead of L66.

L63: Introduce the notation CSG for coal seam gas here instead of L66.

L82: Introduce the BU abbreviation as bottom-up here instead of L99.

L148-150: “However, multiple IFAA samples were collected…” This sentence seems to imply that several IFAA samples were collected downwind of plumes coming from only one source category but looking at Fig. 4 I only see maybe 2 samples from grazing cattle with 100%. I would rather say with “… multiple IFAA samples were collected downwind of a predominant source category…”

L150-151: “One aim of this study…” I would remove this sentence, it does not add much.

L151-156: This is the introduction, there is no need to give too many details about the HYSPLIT footprints, it should rather be developed in the ‘Material and methods’ section. I would just say: “Predominant upwind sources were identified using a combination of the BU inventory presented in Neininger et al. (2021) and back-trajectory footprints (BTF) modeled with the Hybrid single particle Lagrangian integrated trajectory (HYSPLIT).”

L158-162: I would develop a bit on the multi-Keeling-model regression, this is the main part of your paper and it seems like a detail here. Maybe better link it to the sets of IFAA samples that you are mentioning L149-150 (we do not understand what these sets are for otherwise). The last sentence of this paragraph does not really belong to the introduction section, these type of conclusions are more for the abstract, the results/discussion or summary sections. 

Figure 1: There is a lot of information on these maps and their legend. I would try to simplify it a bit by removing the TD domain: it is a bit confusing at first because we do not really know what is this domain and it is just briefly mentioned in section 2.2. For the methane sources, I would already clearly separate point sources and distributed sources in different columns in the legend. I do not think that it is very useful to have the three types of grazing cattle (17a, 17b and 17c), grazing cattle should only be one category with one color. I would also remove the numbers in front of the methane sources, I did not see them anywhere else in the paper.

L198: I would start by explaining that there are two types of sources: point sources and distributed (or diffuse) sources, it is a bit difficult to understand the difference during the first read.  

 L203-206: As I mentioned earlier, I would remove any mention of the TD domain. It just got me confused and wondering how it was defined and why some IFAA samples are outside of it. I would just say that the area surveyed by the airborne platform has a much higher proportion of CSG and a lower proportion of grazing cattle. 

L233: Remove coma after ‘The’.

L237-240: I would remove this last part of the paragraph, it is already stated in the introduction that this study aims at extending the knowledge of isotopic signatures from various methane sources in the Surat Basin. The part about gaining access to a wider range of farms/CSG facilities vs. one sample for the ground-based measurements can go in the discussion section.

L258-261: I would move this part to the end of Section 2.3 or merge it with the last paragraph as they are both talking about source categories obtained from the literature.

L279: PU = polyurethane?

Section 2.5: Which meteorological archives were used with the HYSPLIT model? What is its resolution? Did the authors try different meteorological archives to see how it could affect the definition of the BTF polygons? I don’t know if it changes much at such scales but it is worth checking. HYSPLIT can also produce footprints (HYSPLIT dispersion model), did the authors try to simulate footprints from HYSPLIT and compare to their polygons?  It is not stated very clearly in the text that the back-trajectories start at the mid-point of the sample collection interval (it is only mentioned in the legend of Figure A1).

L294-298: The explanation of how the BTF inventory polygons are estimated is not easy to understand, figure A2 really helps!

L310: relative difference?

L311: 49 IFAA samples were useable out of the 92 collected?

Figure 3: I am not a big fan of Figure 3b, or rather the linear regression between the IFAA concentrations and the BTF BU inventory emissions. Our lives would be so much easier if the relationship between methane concentrations and emissions in the atmosphere was linear… But it depends on so many other parameters: meteorological conditions (these points were collected on different days with different conditions), distance from the source (the samples were collected at different altitudes: 100-200 mAGL vs. 250-350 mAGL) and so on… I don’t think this linear regression means much to be honest. Also, it would be nice to have the same markers than in the middle plot in the two other plots of this figure. IFAA sample 2111 is not mentioned in the legend like the other outliers are. 

L350: I am wondering if 50% is a good enough threshold to attribute a sample to a category of if it is not too low. Let say we have a sample with 55% of CSG (-54.5 ‰), 20% of grazing cattle (-59.7 ‰) and 25% of feedlot (-62.9 ‰), the resulting signature will likely be much lighter than the CSG typical signature and I do not see how this point could be useful in the Keeling inversion. It would be interesting to see how the results are changing with a threshold of 70-75% instead (based on Fig 4b, there should still be enough points for CSG and grazing cattle).

L406-408: “A subset of visually identified outliers with low δ13CCH4(a) values (1604, 1906, 2103) is analysed using the results of the multi-Keeling-model regression. Using the values for CH4(b) and δ13CCH4(b) derived from the multi-Keeling-model regression, the Keeling model (Eq. 1) is fitted to this outlier subset to determine its δ13CCH4(s).”

L427: Should be section 3.2 instead of 3.1.

L440: “IFAA samples 1604, 1817, and 1906 are also highlighted for later discussion.” I would remove “for later discussion” and just directly continue with what is in the next paragraph.

L449: Why do the authors show 3-hour back-trajectory if the footprint calculation is based on 2-hour trajectories?

L474: It would be even easier to see if the BTF polygons for each IFAA sample were also displayed on these figures.

Section 3.3: If I understand correctly, the system uses the different sets to estimate CH4(b) and δ13CCH4(b) first and then use these background values in the Keeling regression of each set. Is the uncertainty on CH4(b) and δ13CCH4(b) estimates taken into account for the Keeling regression of each set in the second step? Looking at Figure 5, there is no error bar for background point and it seems like all the regression lines are exactly passing by this point even if δ13CCH4(b) has a standard error of 0.1 ‰. I think this +/- 0.1 ‰ can have a big impact on the calculated δ13CCH4(s) for the different categories.

L500-501: “Below we also discuss…” I don’t understand this sentence, what is going to be compared? It is not clear to me what is the difference between the 2-hour upwind BU inventory estimates and the expected values based on the BU inventory.

L501-502: “For the IFAA samples discussed below details about the sample location, day and time of collection, and the upwind inventory are listed in Table A2.” There are many commas missing in the paper but sentences usually stay understandable, unfortunately it does not work here.

Figure 6: Add titles with the source category for each plot. L547: there is a missing “)” after “including derived source signatures”.

Section 4.1.3: Looking at Figure 6a, 100-200 mAGL points seem to better align than 250-350 mAGL points but somehow both sets end up with similar standard error. Also, it seems like 4 or 5 points (out of 9?) collected at 250-350 mAGL could fit the 100-200 regression line, this is a bit concerning… 

L595-598: I don’t think excluding sample 1808 because it was collected on a different day is a good reason! Different days have been used in all the previous categories and it was not a problem. Several points of CSG 250-350 mAGL falls into the CSG 100-200 mAGL δ13CCH4(s) signature regression line and have not been excluded! Excluding sample 1808 seems very random, how different is δ13CCH4(s) with this point?

L603-606: Not clear if the problem is that 2-hour back trajectories are not enough or if the BTF polygons have portions out of the BU inventory map. Maybe the border of the BU inventory map should be displayed on figures A3, A4, A5 (as well as the polygons). If the problem comes from the 2-hour back trajectories not being enough for this case, then why should it be enough for the other cases?

L626: Figure 5(a)

Section 4.2 and 4.3: These sections should be merged, most of what is said in section 4.2 is repeated in the beginning of section 4.3.

Section 4.3: Reading this section, it seems like it is difficult to draw any conclusions from these outliers… Outliers 1604 and 1906 are potentially sampling termite emissions and are associated with outlier 2103 whose signal is potentially coming from brine water ponds. Altogether, they end up having an isotopic signature of -80.5 ‰ but outliers 1817 and 2111 sampling these same brine water ponds don’t get the same isotopic signature…

L733-734: Sentence not clear.

L745: “For all three samples, termite emissions may have been sampled.” There is no mention of termite emissions at all in the paragraph discussing outlier 2103 (L671-677) in section 4.3.

Figure A1: I would only show the 2-hour back-trajectories rather than the 3-hour BT. Most of the time, it is difficult to see where is the 2-hour point on the red lines. This comment applies to all the other figures with back-trajectories.