How do Cl concentrations matter for simulating CH4, &delta;13C(CH4) and estimating CH4 budget through atmospheric inversions?

Thanwerdas, Joël; Saunois, Marielle; Pison, Isabelle; Hauglustaine, Didier; Berchet, Antoine; Baier, Bianca; Sweeney, Colm; Bousquet, Philippe

doi:https://doi.org/10.5194/acp-2021-950

Preprints

https://doi.org/10.5194/acp-2021-950

Preprints

02 Dec 2021

Status: this preprint is currently under review for the journal ACP.

How do Cl concentrations matter for simulating CH₄, δ¹³C(CH₄) and estimating CH₄ budget through atmospheric inversions?

Joël Thanwerdas¹, Marielle Saunois¹, Isabelle Pison¹, Didier Hauglustaine¹, Antoine Berchet¹, Bianca Baier^2,3, Colm Sweeney³, and Philippe Bousquet¹ Joël Thanwerdas et al.

Show author details

¹Laboratoire des Sciences du Climat et de l’Environnement, LSCE-IPSL (CEA-CNRS-UVSQ), Université Paris-Saclay 91191 Gif-sur-Yvette, France
²Cooperative Institute for Research in Environmental Sciences (CIRES), University of Colorado-Boulder, Boulder, CO, USA 80305
³NOAA Earth System Research Laboratory Global Monitoring Division, Boulder, CO, USA 80305

Received: 15 Nov 2021 – Discussion started: 02 Dec 2021

Abstract. Atmospheric methane (CH₄) concentrations have been rising since 2007, resulting from an imbalance between CH₄ sources and sinks. The CH₄ budget is generally estimated through top-down approaches using CH₄ observations as constraints. The atmospheric isotopic CH₄ signal, δ¹³C(CH₄), can also provide additional constraints and helps to discriminate between emission categories. The oxidation by chlorine (Cl) likely contributes less than 5 % to the total oxidation of atmospheric CH₄. However, the Cl sink is highly fractionating, and thus strongly influences δ¹³C(CH₄). As inversion studies do not prescribe the same Cl fields to constrain CH₄ budget, it can lead to discrepancies between estimates. To quantify the influence of the Cl concentrations on CH₄, δ¹³C(CH₄) and CH₄ budget estimates, we perform multiple sensitivity simulations using three Cl fields with concentrations that are realistic with regard to recent literature and one Cl field with concentrations that are very likely to be overestimated. We also test removing the tropospheric and the entire Cl sink in other sensitivity simulations. We find that the realistic Cl fields tested here are responsible for between 0.3 % and 1.8 % of the total chemical CH₄ sink in the troposphere and between 1.0 % and 1.2 % in the stratosphere. Prescribing these different Cl amounts in surface-based inversions can lead to differences in global CH₄ source adjustments of up to 12.3 TgCH₄.yr⁻¹. We also find that the globally-averaged isotopic signature of the CH₄ sources inferred by a surface-based inversion assimilating δ¹³C(CH₄) observations would decrease by 0.53 ‰ for each additional percent of contribution from the tropospheric Cl sink to the total sink. Finally, our study shows that CH₄ seasonal cycle amplitude is modified by less than 1–2 % but δ¹³(CH₄) seasonal cycle amplitude can be modified by up to 10–20 %, depending on the latitude.

Download & links

Preprint (PDF, 701 KB)

Supplement (855 KB)

Joël Thanwerdas et al.

Status: final response (author comments only)

RC1: 'Comment on acp-2021-950', Anonymous Referee #1, 22 Dec 2021

This paper investigates the role of pre-scribed Cl-fields on simulated CH₄ and the ¹³C(CH₄). The current inversion system does not yet include a tropospheric Cl sink, and since the group is currently extending their system to include ¹³C(CH₄) observations, this paper is a logical first step. The necessity of this paper apparently appeared when the authors performed their first inversion (presented in the Supplement). As a result, the paper is a bit unbalanced and thin in content. The title promises “atmospheric inversions”, but the main paper only presents forward simulations as a series of sensitivity simulations to map out the impact of various choices for the Cl field. I therefore suggest to remove “through atmospheric inversions from the title”.

For the rest, the paper is well written, and I will attach an annotated pdf with minor comments. The box model inversions are an elegant way to estimate the global impact of the Cl sink on emissions and their required signature. However, the comparison to the vertical profiles are only performed for CH₄ mixing ratios. The results indicate that the model does not perform very well, but that this likely a transport issue rather than an issue with the Cl sink. However, it remains totally unclear how well the model performs in the stratosphere concerning ¹³C(CH₄), while the action of Cl is critical here. I therefore suggest to include an analysis of the modeled ¹³C(CH₄) profiles and compare to the available observations (https://acp.copernicus.org/articles/11/13287/2011/acp-11-13287-2011.pdf)

Citation: https://doi.org/10.5194/acp-2021-950-RC1
RC2:
'Comment on acp-2021-950', Anonymous Referee #2, 02 Jan 2022

The manuscript of Thanwerdas et al. describes an attempt to quantify the effect of applying various assumed/model-generated Cl radical fields on the CH4 mixing and 13C/12C isotope ratios simulated within a CTM (i.e. using offline pre-calculated chemical kinetics and transport). Whilst I find the idea of such quantification useful (nothing innovative but another set of simulations will add to statistic and perhaps could thus help quantifying the uncertainties about other AC-GCM/CTM-specific terms, e.g. dynamics), no marked advance in Cl-CH4 interaction in the atmosphere is obtained, plus the analysis offered ruins the attempt. I foremost imply Section 3.2 (after which I could not continue with required scrutiny) which introduces very questionable “fit methodology” (see the general comment below). Even provided that this is repaired, my other major concern (seconding the Reviewer #1) is in that the study is largely based on results of another – to date not peer-review-completed – study by the same first and another four authors. It is necessary that the latter is finally reviewed in order to be certain that CH4 fluxes used in the simulations are adequate. After that, my suggestion is to consider resubmission of the current work to the GMD, as this journal appears to be more appropriate for the content presented. Compared to other manuscripts submitted to Copernicus journals by Thanwerdas et al. recently, the current one is somewhat better in terms of composition and information content but not sufficiently clearer in conveying the story and presenting methods and discussion (see the specific/presentation comments below). The authors still have to put a considerable effort in improving this.

line                General comments

235, 340         I am not fine with the averaging of absolute biases (in both surface- and column-wise comparisons) – their average may be spuriously reduced through the summation of negative and positive members. Thus obtained low global mean biases do not guarantee that local (per-station or per-altitude) biases are at their optimum. Also, an indication that Cl spatial distribution is wrong will be lost. As a remedy, use squares of biases (as conventionally used in, e.g., least-square fit); I also suggest not to use latitudinal averaging due to the same reasons.

Sect. 3.2         This section is full of confusing and contradictory statements, appended with an apparently erroneous “fitting” approach, see below:

244-246         What is meant by “temporal evolution of CH4 budget is not linear”? If you state that sink is proportional to CH4 abundance, how can both decrease/increase introduce both negative feedback? What feedback is meant here?

246-               You perform simulations with varying CH4 emissions and sinks (biases are derived for varying S and τ_i) yet you assume S ant τ_i constant over time in the analysis. How valid is this approach? How large are the errors introduced by this assumption?

Eqs. 4-7         Why inventing a cumbersome apparatus when you can simply diagnose changes in sink terms (hence ΔS) directly from the simulations? If you still like to use the “box-model” apparatus, why not writing solutions for Eq. 3 for each simulation and their differences (read biases) in analytical form? Ultimately, you confuse the Reader (and yourself) so much that in Eq. 7 you fit both A and B parameters. On which grounds? B represents τ_ref and should be the same for all simulations (it is from a reference simulation, isn’t it?) At t®¥ (steady state), Eq. 6 reduces to b(t)=ΔS´τ_ref. Using the biases from Table 5, this yields various τ_ref for different simulations (about 8 yrs for three of them and 12 yrs for the rest!), how do you explain that? My explanation is that by fitting A and B simultaneously you receive their whatever combination that minimises error-prone averaged absolute biases in the first two decades of simulations. What is the value of τ_ref in the reference simulation?

Sect. 3.3         Same argumentation as for Sect. 3.2 applies, plus you have to show how the fitting is done for isotope ratios, specifically how δ13C biases are obtained. In any case, regarding the errorneous fitting of total CH4, I suspect same or greater problems with 12CH4 and 13CH4.

line                Specific/presentation comments

                       Some of Cl fields are referred to in the manuscript as “realistic” – I strongly discourage that, as it creates impression that the regarded fields were (in)directly compared to Cl observations (they were not, although indirect estimates exist). If they were, would there be the need to test five different distributions?

3,32,41,etc.    just use “composition” instead of “signal” (see the definition of the latter in the dictionary)

5                     there is a lot of processes which may fractionate whatever elements in whatever phases, so you have to be specific here, e.g. use “sink kinetics is 13C/12C fractionating”

22-23             how large is “slight imbalance”?

34-35             this definition is wrong, δ notation always uses ATOMIC ratios, not molar ones – e.g. try to use your definition with isoprene (5 carbon atoms, most of isotopologues are singly substituted)

36-37             there exists the (V-)PDB belemnite-based 13C/12C standard isotope ratio, however there is no standard ratio of PDB known to me

40                   overlaps ® overlap

43                   “sinks are also fractionating” – in addition to which process? Emissions introduce molecules of with various isotope ratios, but this is not a fractionation process. See also comment to l. 5

71                   level of detail

83-84             you can’t claim that/reference the study that is not peer-reviewed yet

86                   do not use “modelling” in this context (modelling is an overall process of creating and applying models, what you refer “reproducing X and Y in the model” or similar)

88                   use “through the prism”, that’ll bear physical sense

95, 98             do not use “dedicate” (you can dedicate a poem to CH4 and Cl, however). Model levels are located in stratosphere and above

108-109         do prescribed species have a diurnal cycle in the model? It is relevant for T-dependent reactions and KIEs, e.g., average OH concentration may be times lower than that at midday, so most of the sink occurs at higher air temperatures in the low troposphere

149                 irrelevant statement (“… was not mandatory”)

Figure 1         please use the same colour scale for both upper and lower panel

161                 exhibits ® exhibit

169-170         vague statement – how may your wish influence the model so that it infers a good model-observation agreement? Most comprehensive studies by no means guarantee delivering most realistic results

208-209         the second sentence repeats the message of the first one, remove

215                 to what estimates “our own estimates” refers to?

Table 4          Cl concentrations are implied in columns 3 and 6?

267                 why not using a conversion factor derived directly from your model simulations? You are using a factor from a study employing a model actual for early 1990s using a very different OH field (Spivakovsky climatology) and meteorology. You may introduce an error in conversion larger than that of any other assumptions used…

Citation: https://doi.org/10.5194/acp-2021-950-RC2
- RC3: 'Corrections to RC2', Anonymous Referee #2, 02 Jan 2022
  
  The Reviewer regrets to have spotted problems with mathematical notation typesetting (on ACP side) after the comment was already submitted. Listed below are (hopefully) the corrected lines:
  
  line                General comments
  Eqs. 4-7         … At t→∞ (steady state), Eq. 6 reduces to b(t)=ΔS∙τ_ref. …
  
  line                Specific/presentation comments
  40                   overlaps → overlap
  161                 exhibits → exhibit
  
  Citation: https://doi.org/10.5194/acp-2021-950-RC3
RC4: 'Comment on acp-2021-950', Anonymous Referee #3, 03 Jan 2022

General comments

The budget of methane is a timely issue, the past changes in atmospheric methane have not been fully explained, and currently the atmospheric methane burden is rapidly increasing. The effect of the chlorine sink on the atmospheric methane burden is relatively small, but as the authors point out the chlorine sink is especially interesting regarding the C13 methane, and the atmospheric chlorine sink need to be included in top-down atmospheric inversions. Overall, the paper is an interesting study on the effect of the atmospheric chlorine sink on methane and C13 methane, and the results are in-line with earlier findings. The question posed in the title is interesting, but can it strictly speaking be answered based on this study, as the results are based on the forward simulations? There is an additional method of analysis used to derive the expected response to the sources if the changes in atmospheric chlorine sink would be made in an inversion system. However, there are some possible shortcomings in the analysis that should be addressed before the paper can be published.

Specific comments

The research topic/questions, the value of isotopic variations for explaining the methane budget in atmospheric inversions, are well justified in the introduction. The introduction presents a reasonable overview of the atmospheric isotopic signal, and especially the effect of chlorine, also for the stratosphere. However, it would be beneficial for the paper to briefly review the processes included in the models and possible differences between the models. Now the discussion is on a very general level, e.g., lines 61-64 “...have made important developments in tropospheric chemistry modeling...”.

For completeness, the photolysis of methane could be added to the reactions (Table 2) as the model extends to ca. 75 km, even though it would likely not impact any of the results in this study.

Lines 135-136. “More details on the modeling of this field are available in the supplement.” However, there are not really any further details given on the simulation, practically the same text is given on lines 120-128 as in the supplement lines 49-60, really the only addition is Table S3. Also, table S3 is almost a duplicate of Table 1 in the manuscript, only the sink column and the average KIE is added to table S3 compared to Table 1. Table S3 seems redundant.

Lines 135-140. The main missing reactions/processes could be mentioned.

Lines 169-176, Scaling the Cl-INCA field to match the tropospheric average of Cl in the Cl-Wang field may introduce some differences, at least visually it seems like the Cl fields would differ at high latitudes, even though the tropospheric average would be nearly the same.

Overall, it would be relevant to have an overview of the major differences between the simulations. The global average is interesting, but the latitudinal and vertical distributions are also important for understanding the impact. Here it would be beneficial to have an overview of the model differences, e.g., the Cl-INCA field seems to have a low (or missing) release of Cl from sea-salt aerosols (Figure 1). Elevated halogen concentrations are often observed in the spring at high latitudes, which could affect surface C13 methane concentrations observed at high latitude stations etc. Latitudinal differences in the Cl field would also cause different responses in the source estimates an inversion system.

Section 2.3. The setup, if I understood correctly, is based on an inversion using a Cl burden that is about half of the one used in the forward run, SimREF, that is then used as reference for the other forward runs using different Cl fields. The fluxes are therefore not optimized with the same Cl burden as in the SimREF, but nevertheless SimREF is used as reference for deriving the delta S, i.e., the source change required to adjust for the different loss rates in the different simulations. The total fluxes would not be affected much, but the distribution between the source categories could be affected. This should be elaborated.

Lines 234-236. Should the SimREF be validated more rigorously when it is used as reference for the other simulations? If the SimREF has biases compared to observations, it might affect the conclusions from the box model analysis (delta S).

Line 237. Is it justified to use mean bias in the comparisons? Positive and negative biases (time, latitude or vertical) will cancel out to some extent.

Lines 241-264. The reason for the introduction of the box model analysis is somewhat unclear. It seems that the driver data ended before the models reached steady state, therefore the steady state had to be estimated by the fitting procedure derived using the box model approach. A more straightforward alternative would be to repeat the simulated years until steady state is reached. The seasonal and interannual variability in the bias, seen in Fig 2, is relatively small compared to the bias. Therefore, it would be justified to repeat the same years to reach steady state. The steady state values could then be used in the analysis instead of fitted values. What is the information obtained in the fitting procedure from B in eq(6)? The values of B are not shown, but they should be almost identical for the different fits? Are they realistic?

Lines 245-247. What is meant by negative feedback from both decrease and increase? “CH4 decrease/increase induces a negative feedback on the magnitude of the sink, leading to a stabilization of the mass of CH4 after several decades if S and τi are constant over time.“

Lines 266-268. Is it justified to use the conversion factor of Lassey et al 2000? The distribution of sinks and the resulting methane distribution will affect the conversion factor between mixing ratio and emissions.

Lines 268-269. “For SimNoCl and SimSherwen, these estimations are very close (difference of less than 0.2 TgCH4.yr−1) to the tropospheric Cl sink discrepancies from Table 4.” Maybe the authors meant SimNoTropo? Then the discussion in the following lines is more understandable.

SimSherwen: 9.9-3.2=6.7 from Table 4 vs. 6.6 in Table 5

SimNoCl: 0-3.2=-3.2 from Table 4 vs. -5.7 in Table 5

SimNoTropo: 0-3.2=-3.2 from Table 4 vs. -3.2 in Table 5

Still this comparison to the tropospheric sink in Table 4 is not straight forward, the stratospheric sink also has an influence. You only need to adjust for the fraction of methane that does not return to the troposphere from the stratosphere, therefore the effect is significantly smaller than the sink itself. The Cl in the stratosphere is fairly similar in all simulations except for the SimNoCl and SimTaki. This could be discussed a bit more around line 271.

Table 5. Latitudinal dependency is reported as min/max, but it is unclear which latitude band is associated with which value.

Lines 286-288. How should B be interpreted when fitting eq(6) for delta13C methane?

Lines 290-291. How is it estimated? “We can estimate that each percent increase in how much CH4 is oxidized by Cl leads to an additional 0.53 ‰ increase in δ13C(CH4), ...” Linear fit to Total oxidation in Table 4 and Signature (Source adjustment) in Table 5?

Lines 294-297. The contribution from STE is estimated as 0.3 ‰, a small clarification could be made that the contribution is only from Cl, not the full contribution from stratospheric intrusions. “Intrusions of stratospheric air are therefore responsible of an enrichment at the surface stations of 0.30 ± 0.01 ‰ (depending on the latitude) after 21 years of simulation, larger than the value of Wang et al. (2002) inferred between 1970 and 1992 (0.23 ‰).” Some discussion could be added for the comparison to Wang et al result. Possible reasons for the discrepancy, different years etc. Also, the difference between SimNoCl and SimNoTropo is 0.36 (Table 5), but the value reported here 0.30 ‰, is from Fig. 2, which is not the steady state value, why is that used instead of a steady state value?

Text S2. P could be explained, first seen at line 85 (k*B)

Text S2. Lines 93-96.

Is it reasonable to assume delta_s equal to delta_a? Is it then also assumed that the isotopic fractionation due to the atmospheric sinks are negligible, even though the idea is to estimate the effect of chlorine on the mean atmospheric isotopic signal? Seems like this assumption going from eq(11) to eq(12) needs to be justified more thoroughly.

Lines 298-302. The value for SimREF, -52.6 from table 2, could be given here to aid the reader. Oscillate is not a good choice of word here, the value does not oscillate, it just depends on the simulation.

Lines 310-328. In the text related to Figure 3 in section 3.4 the reader could be reminded why the SimSherwen has an opposite bias compared to the others. It is also interesting that the SimNoTropo and SimINCA are so similar, this could also be discussed, the tropospheric Cl in SimINCA seems to have a small effect on the delta13C. While SimNoCl seems to have a significant effect on the delta13C amplitude in the northern hemisphere. It is easy to understand that the differences in the bias for methane is small, but to also see small differences in the delta13C is a bit more surprising.

The discussion in section 3.5 on vertical profiles.

The number of profiles in the SH is very low, three observations in tropical latitudes, and one on mid latitudes. The validation of the SimREF is therefore not very convincing, at least for SH. In the NH there are more observations but averaging all profiles from tropical to Arctic soundings into one for the whole hemisphere is probably not good. Already the tropopause height is quite different but also the stratospheric polar vortex might have influenced the Arctic soundings. It is, however, difficult to know the reason for the observed discrepancy without seeing the individual profiles. Now the simulations were only sampled when there was a sounding. To make a more thorough analysis of the differences between the simulations the full fields should also be used. Furthermore, some discussion should be added on the delta13C profile.

Line 343 “The mean bias relative to SimREF is given for all simulations and observations in Table 6.” Are there any values given for observations; I find only simulations?

Line 343-344: “A change in the Cl field (and keeping it realistic) induces a maximum mean bias of 51 ppb in the stratosphere (SimNoCl).” I don't understand the meaning of this statement. It is the difference between a realistic Cl distribution and no Cl.

Text S3.

It is unclear what the TCCON analysis adds to the paper, especially as nothing is mentioned in the paper about the results, results are only presented in the supplementary. It is not clear which bias is presented in Fig. S5 colored dots (surface observations compared to SimNoCl or SimREF). Also, for XCH4 it could be better to compare the simulations without filtering the data only for cases with observations. I understand that the point is to get an idea of differences in areas/times with data, in case the data would be assimilated, but without the assimilations it is not especially useful. It is somewhat interesting that the bias between SimNoCl and SimREF is latitude dependent. Probably a combination of Cl distribution and atmospheric transport but based on this analysis it remains unclear. It would perhaps make more sense to make a more rigorous analysis in another paper.

Lines 377-378 “...the change in the Cl field..” This conclusion should be reworded, unclear what is meant by “the change”.

Lines 384-385 This conclusion may need to be reworded once the previous comments have been considered, the result is not from an actual inversion. “In an inversion, this additional percent of contribution would reduce the inferred globally-averaged isotopic signature by 0.53 ‰.”

Lines 385-386. The authors probably mean only the contribution from Cl in the stratosphere, rather than the full impact of stratospheric intrusions. The given value is not from a steady state situation, how/why is this value chosen. The driver data ended?

Lines 388-390. This conclusion should be re-worded. This is true for the specific changes made in this study, not in general. Some other change in Cl could result in a change in delta13C methane outside the 10-20 range. “CH4 seasonal cycles are only slightly influenced by a modification of the Cl sink (1-2 % change in the seasonal cycle amplitude). Changing the Cl field can nevertheless modify the amplitudes of δ13C(CH4) seasonal cycle by up to 10-20 %, depending on the latitude”

The conclusions regarding the vertical profiles (Line 390-) may need to be revised once the discussion is updated. The comparison using hemispherical averages is likely not representative due to the large span of latitudes that are averaged.

technical corrections

Line 139. half lower than the mean tropospheric >> half of the mean

Table 2. The abbreviation VPDB is not explained

Lines 221-223 unclear which value is meant “The tropospheric value from Hossaini et al. (2016), used in recent studies (Saunois et al., 2020; McNorton et al., 2018), is also slightly above that of ClSherwen (Table 4: 1.4 times higher) but well above that of Cl-Wang and Cl-INCA (4 and 8.5 times higher).”

Table 4 third column Conc. Is not explained in the caption (average Cl conc. ?)

Line 369. Cl configuration >> Cl distribution

Citation: https://doi.org/10.5194/acp-2021-950-RC4

Joël Thanwerdas et al.

Supplement

https://doi.org/10.5194/acp-2021-950-supplement

Joël Thanwerdas et al.

Viewed

Total article views: 579 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
399	163	17	579	44	6	5

HTML: 399
PDF: 163
XML: 17
Total: 579
Supplement: 44
BibTeX: 6
EndNote: 5

Views and downloads (calculated since 02 Dec 2021)

Month	HTML	PDF	XML	Total
Dec 2021	262	96	3	361
Jan 2022	70	35	10	115
Feb 2022	26	13	3	42
Mar 2022	12	5	0	17
Apr 2022	24	13	0	37
May 2022	5	1	1	7

Cumulative views and downloads (calculated since 02 Dec 2021)

Month	HTML	PDF	XML	Total
Dec 2021	262	96	3	361
Jan 2022	70	35	10	115
Feb 2022	26	13	3	42
Mar 2022	12	5	0	17
Apr 2022	24	13	0	37
May 2022	5	1	1	7

Viewed (geographical distribution)

Total article views: 575 (including HTML, PDF, and XML) Thereof 575 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 23 May 2022

Short summary

Atmospheric methane (CH₄) concentrations have been rising since 2007, resulting from an imbalance between CH₄ sources and sinks. The CH₄ budget is generally estimated through top-down approaches using CH₄ and δ¹³C(CH₄) observations as constraints. The oxidation by chlorine (Cl) contributes little to the total oxidation of CH₄ but strongly influences δ¹³C(CH₄). Here, we compare multiple recent Cl fields and quantify the influence of Cl concentrations on CH₄, δ¹³C(CH₄) and CH₄ budget estimates.


Total:	0
HTML:	0
PDF:	0
XML:	0