How do Cl concentrations matter for simulating CH 4 , δ 13 C(CH 4 ) and estimating CH 4 budget through atmospheric inversions ?

. Atmospheric methane (CH 4 ) concentrations have been rising since 2007 , resulting from (cid:58)(cid:58)(cid:58) due (cid:58)(cid:58) to (cid:58) an imbalance between CH 4 sources and sinks. The CH 4 budget is generally estimated through top-down approaches using (cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58) chemistry-transport (cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58) models (cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58) (CTMs) (cid:58)(cid:58)(cid:58) and (cid:58) CH 4 observations as constraints. The atmospheric isotopic CH 4 signal (cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58) et 2020) between sources that release CH 4 into the atmosphere and sinks that remove it. Sinks are mostly due to oxidation reactions in the atmosphere . Three radicals react with (cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58) between CH 4 in the atmosphere (cid:58)(cid:58)(cid:58) and (cid:58)(cid:58)(cid:58)(cid:58) three (cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58) radicals : hydroxyl (OH), atomic oxygen (O 1 D), and chlorine (Cl). These chemical reactions account for about 93 % of the total CH 4 sink, with 40 the remainder being removed by methanotrophic bacteria in the soil (Saunois et al., 2020). On the other hand, CH 4 sources are varied and result from radically different processes (biogenic, thermogenic and pyrogenic). the of Cl the estimated CH 4 ﬂuxes. the troposphere and the stratosphere) for the of an inversion assimilating can to radically different source mixtures and/or source

for the year 2020. The accumulation of CH 4 (∼ 8 ppb.yr −1 on average since 2007) in the atmosphere is the result of a slight imbalance between sources that release CH 4 into the atmosphere and sinks that remove it. Sinks are mostly due to oxidation reactions in the atmosphere. Three radicals react with CH 4 in the atmosphere: hydroxyl (OH), atomic oxygen (O 1 D), and chlorine (Cl). These chemical reactions account for about 93 % of the total CH 4 sink, with the remainder being removed by 25 methanotrophic bacteria in the soil (Saunois et al., 2020). On the other hand, CH 4 sources are varied and result from radically different processes (biogenic, thermogenic and pyrogenic).
Top-down atmospheric inversions are known to be efficient approaches to estimate CH 4 sources at different scales and have become increasingly relevant over the years as observational networks have developed (Houweling et al., 2017, and references therein). However, inversions that assimilate only total CH 4 observations can only rely on variations in seasonal cycles to dif-30 ferentiate co-located emissions. To better separate these sources, assimilating observations of the 13 C: 12 C atmospheric isotope signal of CH 4 , denoted δ 13 C(CH 4 ), can be relevant. This value is based on the ratio between the isotopologue 12 CH 4 , which represents about 99 % of the CH 4 in the atmosphere (Stolper et al., 2014) and its counterpart 13 CH 4 . δ 13 C(CH 4 ) is commonly defined using a deviation of the sample mole isotopic ratio relative to a specific standard ratio : at the surface. In particular, Wang et al. (2002) estimated that stratospheric Cl was responsible for a δ 13 C(CH 4 ) enhancement of 0.23 ‰ at the surface between 1970 and 1992 due to stratosphere-troposphere exchanges (STE).
In the troposphere, the Cl sink likely accounts for less than 5 % of CH 4 oxidation (Wang et al., 2019(Wang et al., , 2021Hossaini et al., 2016;Sherwen et al., 2016;Gromov et al., 2018;Allan et al., 2007). Several studies have estimated Cl concentrations in the troposphere and in the Marine Boundary Layer (MBL) and discussed the Cl sink. Allan et al. (2007) estimated the 60 Cl sink in the troposphere to be 25 TgCH 4 .yr −1 , representing about 5 % of the total CH 4 chemical sink. More recently, Hossaini et al. (2016), Sherwen et al. (2016), Wang et al. (2019) and Wang et al. (2021) have made important developments in tropospheric chemistry modeling and obtained oxidation contributions of 2.6 %, 2 %, 1 % and 0.8 % respectively with mean tropospheric Cl concentrations between 620 and 1300 molec.cm −3 . However, Gromov et al. (2018) concluded that variations in Cl concentrations above 900 molec.cm −3 in the extratropical part of the Southern Hemisphere are very unlikely ; thus 65 suggesting that the high estimates from Allan et al. (2007) and Hossaini et al. (2016) are likely overestimated. These variations in oxidation contributions may appear small but Strode et al. (2020) recently showed a high sensitivity of the tropospheric δ 13 C(CH 4 ) distribution to variation in Cl fields by testing, among others, those of Allan et al. (2007), Sherwen et al. (2016) and Hossaini et al. (2016), indicating that each percent increase in how much CH 4 is oxidized by Cl leads to a 0.5 ‰ increase in δ 13 C(CH 4 ), therefore larger than the global downward shift observed since 2007 (Nisbet et al., 2019). 70 Forward and inverse 3-D modeling studies focusing on CH 4 and δ 13 C(CH 4 ) consider the Cl sink at different level of details.
Most studies consider only the Cl sink in the stratosphere (e.g., Fujita et al., 2020;Rigby et al., 2012;Monteil et al., 2011;Fletcher et al., 2004), and a very few account for tropospheric Cl only (e.g., Thompson et al., 2018). In single-box models, sinks are combined and an overall fractionation coefficient is used (e.g., Schaefer et al., 2016;Schwietzke et al., 2016). In recent studies, Cl is often prescribed in both the troposphere and stratosphere (e.g., McNorton et al., 2018;Rice et al., 2016;Warwick et al., 2016;Neef et al., 2010), although most studies use the Cl configuration suggested by Allan et al. (2007), which is likely to be overestimated as mentioned above.
In the atmospheric inversions performed with the LMDz-SACS chemistry-transport model (Locatelli et al., 2015;Pison et al., 2009), Cl sink was omitted so far, even in the stratosphere (Saunois et al., 2020;Locatelli et al., 2015;Pison et al., 2009;Bousquet et al., 2006). For these studies assimilating only total CH 4 observations, the impact of the Cl sink on the estimated 80 CH 4 emissions was considered negligible. However, the number and quality of isotopic observations have considerably increased since the 2000s, and developments of the CIF-LMDz-SACS inversion system to use the isotopic constraint have been made (Thanwerdas et al., 2021). Joint assimilation (CH 4 and δ 13 C(CH 4 )) is proving to be relevant and necessary in order to reconcile the estimated CH 4 budgets with the atmospheric isotope signal. Considering the large impact of the Cl sink on δ 13 C(CH 4 ), it is necessary to include and evaluate the Cl sink and its impact on the CH 4 modeling process.

85
Here, we detail the influence of tropospheric and stratospheric Cl on the modeling of CH 4 and δ 13 C(CH 4 ) in LMDz-SACS by using several Cl fields. Ultimate goal being to use the isotopic signal observations to perform multi-constraint inversions with the LMDz-SACS model, results are analyzed throughout the study under the prism of atmospheric inversion. In the first part, we present the characteristics of the available Cl fields, model inputs and observations used for evaluation. Then, we analyze the influence of the different Cl fields on CH 4 and δ 13 C(CH 4 ) at the surface and on the CH 4 vertical profile.  (Hourdin et al., 2006). The version of LMDz used here is an "offline" version dedicated to the inversion framework created by Chevallier et al. (2005): the pre-  Vertical diffusion is parameterised by a local approach of Louis (1979), and deep convection processes are parameterised by the scheme of Tiedtke (1989). The offline model LMDz, coupled with the Simplified Atmospheric Chemistry System (SACS) module (Pison et al., 2009), was previously used to simulate atmospheric mole fractions of trace gases such as CH 4 , carbon monoxide (CO), methyl chloroform (MCF), formaldehyde (CH 2 O) or hydrogen (H 2 ). This system has been recently converted into a chemistry parsing system (Thanwerdas et al., 2021). It follows the principle of the chemical parsing system of the 105 regional model CHIMERE (Mailler et al., 2017;Menut et al., 2013) and allows the user to prescribe the set of chemical and HFC-134a, and their associated photochemical reactions, were included in the INCA chemical scheme to produce Cl rad-125 icals (Terrenoire et al., 2020). In the LMDz-INCA simulations, surface concentrations of these long-lived Cl source species were prescribed based on historical data sets prepared by Meinshausen et al. (2017). The model was run for the 1850-2018 period (Hauglustaine et al., 2021). we adopted the value of Saueressig et al. (2001) as they indicate that this data is of considerably higher experimental precision and reproducibility than previous studies, in particular Cantrell et al. (1990), which suggested a value of 1.0054.

Description of Cl fields
Four fields of Cl are used in this study. The first field was simulated by the LMDz-INCA model, as mentioned above. More , v10 and v12.9, respectively). These fields were generously provided by the respective authors of the two studies. They will be referred to as the Cl-Sherwen and Cl-Wang fields. Differences between the two fields are detailed below.

145
The last field was simulated by version 5.7b of the CCSR/NIES/FRCGC (Center for Climate Sytem Research/National Institute for Environmental Studies/Frontier Research Center for Global Chance) atmospheric GCM (Takigawa et al., 1999).
This was provided by the GCP-GMB (Global Carbon Project -Global Methane Budget) team to run the inversions used in Saunois et al. (2020), although it was not mandatory. It is referred to as the Cl-Taki field.  In this study, we do not test the Cl fields from Hossaini et al. (2016) and Allan et al. (2007) because we want to carry 150 out the sensitivity analysis while keeping a realistic and up-to-date range of concentrations. Their concentrations are indeed very likely to be overestimated (Gromov et al., 2018). Although Cl concentrations in the Cl-Taki field are also very large, to our knowledge, CH 4 oxidation resulting from the prescription of this field has not been studied before. We therefore choose to include it here in order to quantify the associated sink and to illustrate the influence of such concentrations on CH 4 and δ 13 C(CH 4 ).

155
The four fields are shown in Fig. 1. We use the lapse rate (2 K/km) definition from the World Meteorological Organization Our SimREF reference simulation uses the Cl-Wang field as it is the most recent field and is taken from the most comprehensive study to date. In addition, we want our reference simulation to infer realistic CH 4 and δ 13 C(CH 4 ) distributions with a good is obtained on global averages, and is considered sufficient to validate the conclusions of this study. More information about the inversion is given in the supplement (Text S1). Emissions and source isotopic signatures are given in Table 2. Both vary over time and space and are prescribed as monthly fields at the horizontal resolution of the model. thus, we tested this with the SimNoTropo simulation where the Cl-Wang is used but with no Cl in the troposphere. Moreover, as LMDz-SACS completely omitted the Cl sink in previous studies, we estimate the errors generated by this omission running the SimNoCl simulation, which has no Cl sink. A summary of the simulations and their characteristics is provided in Table 3.  The station locations and additional information can be found in the supplementary Figure S3, Tables S5 and S6.
Finally, an analysis of the impact of Cl on CH 4 vertical profiles is conducted using a set of 115 AirCore profiles recovered from 11 different sites over the 2012-2018 period. A total of 80 profiles are provided by the NOAA GML aircraft programme (Baier et al., 2021;Karion et al., 2010) and 35 others by the French AirCore programme (Membrive et al., 2017). The balloon-

200
borne AirCore technique (Karion et al., 2010) allows air samples to be taken from the stratosphere (up to approximately 30 km) to the ground, upon a parachute-based descent. Figure S4 and Table S4, in the supplement, provide information about the provider, location and number of profiles collected. Reported uncertainties generally increase with altitude due to endmember mixing within the AirCore samples. They are below 2 ppb in the troposphere and can reach 10 ppb in the lower stratosphere.

Quantification of the Cl sink
The simulated chemical sink of CH 4 due to Cl oxidation varies depending on the prescribed Cl field. We therefore obtain different sink intensities depending on the simulation. Table 4  From our simulations, contributions from the tropospheric Cl sink with Cl-Wang (0.6 %) and Cl-Sherwen (1.8 %) are slightly lower than those given in the associated papers (i.e., 0.8 % and 2 %, respectively). This discrepancy is likely due to a slight difference in the definition of the tropopause level or/and in the prescribed OH sink that is used to calculate the total chemical sink.

220
The tropospheric sink provided by Allan et al. (2007) is well above the other recent values. 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 Cl-Sherwen (Table 4: 1.4 times higher) but well above that of Cl-Wang and Cl-INCA (4 and 8.5 times higher). In the troposphere, the sink induced by Cl-Taki is much larger than the other sinks (up to 28 times larger) and therefore even larger than the value suggested by Allan et al. (2007) which is very likely to be overestimated (Gromov et al., 2018). In the stratosphere, the sink is 225 also slightly larger (1.3 times that of Cl-Sherwen).
Apart from the Cl-Taki field, we selected here only the fields that provided a realistic range of concentrations when applying the conclusions of Gromov et al. (2018). We therefore consider only the Cl fields that give a CH 4 tropospheric oxidation below 2 % as realistic. These fields happen to be the most recent and up-to-date estimations.
In the stratosphere, all tested fields are considered realistic because they provide an oxidation between 1.1 and 1.6 %, therefore in agreement with Saunois et al. (2020)  Cl field should be rigorously analyzed (concentration, oxidation) before prescribing it in a forward or inverse simulation.

CH 4 surface concentrations
The impact of the Cl sink on CH 4 mole fractions is analyzed by comparing the simulations against SimREF at MBL station locations providing δ 13 C(CH 4 ) data. Since the Cl fields vary mainly as a function of latitude, the comparisons are made by 235 averaging values over bands of latitude. Here, the bias b is defined as : where b X,i,l is the bias for a specific quantity X (i.e., CH 4 or δ 13 C(CH 4 )), a specific simulation i, and a specific band of latitude l. X i,s denotes the CH 4 or δ 13 C(CH 4 ) values simulated by a simulation i and at a station s. The (.) l symbol indicates the mean over all the stations whose location is inside the band of latitude l.

240
In a box model, the temporal evolution of the CH 4 budget for a simulation i is described by the equation below : where B i is the mass of CH 4 in the atmosphere in TgCH 4 , S is the source in TgCH 4 .yr −1 and τ i is the chemical lifetime of CH 4 in the atmosphere in yr. Note that the same sources are prescribed for all simulations. The temporal evolution of the CH 4 budget is not linear, because the sink is proportional to CH 4 mole fractions. CH 4 decrease/increase induces a negative feedback 245 on the magnitude of the sink, leading to a stabilization of the mass of CH 4 after several decades if S and τ i are constant over time. Here, the bias between two simulations is caused by a change in τ because we modify the Cl field. The evolution of the bias b can therefore be described by the equation : However, in a surface-based inversion (i.e., an inversion assimilating observations from surface stations) without sink optimiza-250 tion, the bias is compensated by a correction of the CH 4 global surface flux S + ∆S. The inversion system therefore answers the question : "What is the value of ∆S that will offset the bias caused by a change in the prescribed sink ?". The temporal evolution of the bias between a simulation and the reference simulation can therefore be described by the equation : τ re f denotes the chemical lifetime in the reference simulation. We consider that ∆S is constant over time as the inter-annual 255 variability of the Cl sink is below 0.4 TgCH 4 .yr −1 . In that case, the solution of this equation is : The value of ∆S can be obtained by analyzing the temporal evolution of the bias and, in particular, by looking at the value of the bias when it is stabilized (steady state). Here, after 21 years of simulation, the stabilization is not reached yet (see Fig. 2, top row). Therefore, we choose to extend our results by applying a curve fitting function to our simulated values : A X,i,l and B X,i,l are two constants that the curve fitting algorithm returns in order to maximize the agreement between the simulated values and the curve fitting function. Using this function, the results are extended until 2070 to reach a clear stabilization of simulated biases (see Fig. 2, top row).
At steady state, the bias of CH 4 at the surface varies between -20.0 ppb for SimSherwen and 24.5 ppb for SimNoCl (Table   265 5, second column). An estimation of ∆S is given by the coefficient A X,i,l . It provides a result in ppb.yr −1 . To convert this value in TgCH 4 .yr −1 , we use a conversion factor of 2.767 Tg.ppb −1 (Lassey et al., 2000) and show the final estimates in Table 5, fourth column. For SimNoCl and SimSherwen, these estimations are very close (difference of less than 0.2 TgCH 4 .yr −1 ) to the tropospheric Cl sink discrepancies from Table 4. Indeed, as the stratospheric Cl sinks in SimREF, SimINCA, SimNoTropo and SimSherwen are almost identical, the biases induced by tropospheric Cl sink discrepancies will be logically compensated 270 by a source adjustment of the same intensity as the sink discrepancy. For SimNoCl, the biases at the surface are also influenced by large stratospheric sink discrepancies. Therefore, the inferred adjustment values cannot be so simply related to the sink discrepancies. Also, latitude has a very low influence on biases and adjustment values, causing a variation of less than 5 % around the mean value (see Fig. 2).
We conclude that a source adjustment of 12.3 TgCH 4 .yr −1 would be necessary between a surface-based inversion without Cl  3.3 δ 13 C(CH 4 ) signal at the surface In contrast with the CH 4 biases, the δ 13 C(CH 4 ) biases between the simulations are much larger than recent δ 13 C(CH 4 ) observed downward shifts (∼ 0.3 ‰ since 2007). We use the same curve fitting method as before to propagate the time-series until 2070 in order to reach a steady state (see Fig. 2, bottom row).
SimNoTropo, SimINCA, SimREF and SimSherwen have very similar stratospheric Cl sinks (Table 4). Therefore, biases for SimNoCl, SimNoTropo, and SimSherwen are mostly generated by discrepancies in tropospheric Cl sink intensity. We can 290 estimate that each percent increase in how much CH 4 is oxidized by Cl leads to an additional 0.53 ‰ increase in δ 13 C(CH 4 ), To reduce these biases to zero, an inversion system would adjust the globally-averaged isotopic signature of the CH 4 sources, denoted by δ 13 C(CH 4 ) source . This adjustment factor would be roughly equal to the opposite of the bias at steady state (see demonstration in the supplementary Text S2). It would therefore oscillate between −0.60 ‰ (SimSherwen) and +0.66 ‰ 300 (SimNoCl) around the mean isotopic signature of the global CH 4 source prescribed in SimREF. We would therefore obtain, after the inversion process, a mean global signature between −53.20 ‰ (SimSherwen) and −51.94 ‰ (SimNoCl).
The system would modify δ 13 C(CH 4 ) source by changing the source mixture and/or the isotopic signatures of the multiple emission categories, with a weight depending on uncertainties associated to both. For instance, an adjustment of −0.60 ‰ could be made by increasing the wetlands share from 32 % to 43 % or by shifting the mean isotopic signature of wetlands from 305 −56.6 to −58.5 ‰, more in agreement with recent estimates (Ganesan et al., 2018;Sherwood et al., 2017) than our inverted value (see Table 2). However, the system would likely change not only wetlands but all emission categories, possibly limiting an unlikely large change in wetlands emissions only. Nevertheless, the configuration used to represent the Cl sink could largely influence the result of an inversion assimilating both CH 4 and δ 13 C(CH 4 ). The impact is more important for the δ 13 C(CH 4 ) seasonal cycle and is dependent on latitude. In the Southern Hemisphere, the variation in amplitude between SimREF and SimSherwen is about 0.02 ‰, which represents 20 % of the total seasonal cycle amplitude. In the Northern Hemisphere, the variation can exceed 0.03 ‰, which represents 10 % of the seasonal cycle amplitude.

320
SimTaki (not shown on Fig. 3 for clarity reason) causes a much larger variation in seasonal cycle for both CH 4 and δ 13 C(CH 4 ). For CH 4 , variations reach 5 % in the Southern Hemisphere and 10 % in the Northern Hemisphere. As for δ 13 C(CH 4 ), variations go up to 99 % in the Southern Hemisphere and 58 % in the Northern Hemisphere.
The influence of Cl on δ 13 C(CH 4 ) seasonal cycle must be considered as it will impact the results of an inversion with δ 13 C(CH 4 ) data assimilation. A misrepresentation of the seasonal cycle forces the system to adjust the intensity of sources that 325 actively participate in the seasonal cycle, such as wetlands or biomass burning emissions. This influence is negligible for CH 4 and noticeable for δ 13 C(CH 4 ) when keeping realistic Cl concentrations but becomes very large when using other Cl fields, such as the Cl-Taki field.

CH 4 vertical profiles
Vertical profile measurements of CH 4 are too scarce to be considered as a stand-alone constraint in inversion systems, and 330 so are rather used as evaluation data. Nevertheless, as their accuracy, spatial coverage and number increase, their assimilation will become increasingly relevant. It is, however, necessary to increase the model-observation agreement, especially in the stratosphere, before considering their assimilation. We analyze here the influence of the Cl configuration on these profiles. We also compare the simulated vertical profiles to observations to investigate whether modifying the Cl configuration can help to reduce the model-observation discrepancies.

335
Simulated vertical profiles are sampled at the same locations and time as the observations available. The bias b X,y,p,d 1 ,d 2 between two vertical profiles d 1 and d 2 (simulated or observed) for a specific profile p, a specific quantity X (i.e., CH 4 ) and a specific layer y (troposphere, stratosphere or total) is given by: The (.) y symbol indicates the mean over all the vertical levels in the layer y. We also define the mean bias as the bias averaged 340 over all available vertical profiles : b X,y,d 1 ,d 2 = b X,y,p,d 1 ,d 2 p (9) The mean bias relative to SimREF is given for all simulations and observations in Table 6. A change in the Cl field (and keeping it realistic) induces a maximum mean bias of 51 ppb in the stratosphere (SimNoCl). For all simulations besides SimNoCl, the bias is roughly constant over the entire column (see Fig. 4  Although modifying the prescribed Cl field can induce local differences in stratospheric mole fractions of the same order of magnitude as the model errors, none of the tested Cl sink really improves our model-observation agreement in the stratosphere as the inflections of mole fractions are not properly represented. Patra et al. (2011) already mentioned that strong vertical gradients of CH 4 around the tropopause may be caused by a too slow Brewer-Dobson circulation so these discrepancies are possibly due to transport errors rather than errors in removal rates. Further investigating the discrepancy in the stratosphere is 360 however beyond the scope of this study. Ostler et al. (2016) showed that model errors in simulating stratospheric CH 4 contribute to model biases when compared to observed column-averaged CH 4 dry-air mole fractions (XCH 4 ) from the Total Carbon Column Observing Network (TCCON).
XCH 4 obtained by remote sensing techniques are now massively assimilated in inversions because satellite observations offer a much larger spatial coverage than in situ measurements. Rigorously estimating the influence of Cl concentrations on a satellite-365 based inversion would require more than an one-box model approximation. We therefore include only a simple analysis using data from the GOSAT satellite in the supplementary Text S3.

Conclusions
In this study, we tested multiple Cl fields suggested by recent studies to investigate the influence of the Cl configuration on CH 4 and δ 13 C(CH 4 ), and to estimate its potential impact on the estimation of CH 4 sources and isotopic signatures with top-down 370 approaches.
We tested a realistic range of Cl concentrations, i.e., resulting in Cl tropospheric and stratospheric oxidations that are in agreement with recently published studies. We also included a Cl field suggested by the GCP 2018 protocol to be prescribed in inverse simulations in order to investigate its influence on CH 4 and δ 13 C(CH 4 ) values in comparison with more realistic and recent Cl fields. The realistic Cl fields tested here are responsible for between 0.3 % and 1.8 % of the total CH 4 sink in the 375 troposphere and between 1.0 % and 1.2 % in the stratosphere.
At the surface, the change in the Cl field and thus in the associated CH 4 sink results in a bias in CH 4 mole fractions that reaches a maximum value of 44.5 ppb at steady state. An inversion system would adjust the CH 4 surface fluxes by a value of 12.3 TgCH 4 .yr −1 to compensate for these biases. This adjustment remains small in comparison to the uncertainties inferred by Saunois et al. (2020). However, the use of perhaps more unrealistic Cl fields (as suggested by recent literature) can generate 380 much larger biases.
δ 13 C(CH 4 ) values at the surface are also shifted by a change in the prescribed Cl field. In particular, we find an increase in the δ 13 C(CH 4 ) global mean at the surface of 0.53 ‰ at the surface for each additional percent of contribution from the tropospheric Cl sink to the total CH 4 sink. In an inversion, this additional percent of contribution would reduce the inferred globally-averaged isotopic signature by 0.53 ‰. Also, we find that intrusions of stratospheric air are responsible for an enrichment of δ 13 C(CH 4 ) 385 by 0.30 ‰ at the surface between 1998 and 2018. Neglecting the influence of stratospheric Cl on δ 13 C(CH 4 ) surface values could therefore increase the global mean isotopic signature estimated by an inversion by 0.30 ‰.
CH 4 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 δ 13 C(CH 4 ) seasonal cycle by up to 10-20 %, depending on the latitude. 390 We also investigate the influence of Cl concentrations on the modeling of vertical profiles. We find that statospheric modelobservation discrepancies in LMDz-SACS are unlikely to be caused by a misrepresentation of the Cl sink, although a change in Cl concentrations can shift CH 4 mole fractions at 25 km by up to 130 ppb. Also, a change in the tropospheric Cl sink influences tropospheric and stratospheric CH 4 mole fractions to the same magnitude.
It is difficult to conclude which Cl field provides the most realistic representation of the Cl sink among those tested here.
Recent developments and efforts have nevertheless narrowed the range of uncertainties regarding the Cl concentrations (less than 1.1 × 10 3 molec.cm −3 in the troposphere and 1.4-1.6 × 10 5 molec.cm −3 in the stratosphere). Our study shows that the