LMDZ is an atmospheric general circulation model (GCM) developed by Sadourny and Laval (1984). The primitive equations of the GCM are solved with a time step of 3 min, a large-scale transport of tracers is carried out every 15 min, and the physical processes are calculated at a time interval of 30 min. The LMDZ model provides at each time step the physical quantities (temperature, pressure, water vapour, turbulence, boundary layer, etc.) needed to compute the transport and chemistry in INCA.

INCA is an atmospheric chemistry model developed at LSCE (Laboratoire des Sciences du Climat et de l'Environnement). The model takes into account the primary emissions from natural sources or by anthropogenic activities. It computes physical processes such as dry and wet deposition and photochemical reactions (Hauglustaine et al., 2004). The atmospheric concentration fields are integrated in time by solving differential equations when called by LMDZ, with a time step of 30 min. For this development, we used a version of INCA (version v_5.2.1) simulating interactively the gaseous photochemistry, whereas aerosol distributions obtained with a more complex version of INCA are prescribed through climatologies. This version represents in particular the chemistry of the inorganic species Ox, NOx and HOx, as well as CH4, CO, non-methane volatile organic compounds (NMVOCs) and their photo-oxidised products (Folberth et al., 2006).

Biogenic emissions like those of isoprene are provided by the global vegetation model ORCHIDEE (Organising Carbon and Hydrology In Dynamic Ecosystems) (Folberth et al., 2006). Anthropogenic emissions are based on the Community Emissions Data System (CEDS) v2016-07-2 emission inventory for the year 2010 (Hoesly et al., 2018).

The LMDZ-INCA model has been used and assessed in numerous previous model intercomparison exercises aimed at simulating the pre-industrial, present and future tropospheric composition (e.g. Szopa et al., 2013); the long range transport of air pollution in the HTAP exercise (Fiore et al., 2009); or the chemistry–climate interactions in ACCMIP (Naik et al., 2013).

The chemistry of halogenated compounds implemented in LMDZ-INCA is essentially based on the scheme developed for CAM-Chem (Ordóñez et al., 2012; Prados-Roman et al., 2015a; Saiz-Lopez et al., 2012a, 2014, 2015) and GEOS-Chem (Sherwen et al., 2016a), with a priority given to the reactions of importance for the ozone sinks (Fig. 1).

A total of 42 halogenated species have been added to the INCA chemical scheme (Table 1). The list of the 160 chemical reactions added to the INCA chemical scheme can be found in Tables 2, 3, 4 and 5. Photolysis reactions of halogens as well as relative quantum yields and absorption cross-sections are listed in Table 2. Uptake coefficients for heterogeneous reactions are based on experimental values and data selected for other models and are reported in Table 5. Recycling through heterogeneous chemistry on sea salt and sulfate aerosols is included, while the heterogeneous bromine reactions in water droplets and on ice crystals in clouds are not. Dry deposition is calculated with the Wesely scheme (Wesely, 1989), while washout of gases by precipitation is simulated by using Henry's law constants detailed in Table S1 in the Supplement. Some VSL halocarbons, CH2BrCl, CHBr2Cl and CHBrCl2, represent a few percent of the emissions of halocarbons (Ordóñez et al., 2012). Considering the high uncertainty in their emissions and their limited abundances, we chose not to consider their chemistry in this version.

Primary halogenated compounds can originate from both inorganic and organic sources. Inorganic species are reactive species emitted by different sources (e.g. oceans, sea salts) but also result from atmospheric reactions (Simpson et al., 2015). Organic species are only of primary origin and are photo-oxidised in the atmosphere to form halogen radicals (Simpson et al., 2015). Table 6 shows the emissions of halogenated compounds as considered in INCA. Organo-bromines (e.g. CHBr3, CH2Br2), organo-iodines (e.g. CH3I, CH2I2) and inter-halogen species (e.g. CH2IBr, CH2ICl) are those used by the CAM-Chem model (Ordóñez et al., 2012).

Sources of organic chlorine are diverse, mainly originating from biological activity on the surface of the oceans and from forest fires. The emissions of these chlorocarbons are detailed in Table S2. In INCA, inorganic chlorine, HCl – the major reservoir of chlorine in the atmosphere – is considered to be emitted by sea salts (90 000 Gg Cl yr-1) and by forest fires (6400 Gg Cl yr-1) with total based on Lobert et al. (1999) and Hossaini et al. (2016). Note that more comprehensive but hard to implement approaches relying on pH-dependent parameterisation of the dechlorination of sea salts have been recently implemented in GEOS-Chem. HCl emissions from industrial sources (McCulloch et al., 1999) are neglected as recommended by Wang et al. (2019) to avoid overestimating HCl in urban areas. ClNO2 is emitted by the heterogeneous reaction of N2O5 on sea salts and produces 2217 Gg Cl yr-1.

CH3Br emissions are from oceanic origin (40 %) and from agriculture (60 %) (Barker et al., 2016) and represent 91 Gg of Br per year (Schmidt et al., 2016). IBr and ICl originate from the heterogeneous reaction of HOI, INO2 and INO3 on sea salt aerosols (Table 5) as considered in other models (Saiz-Lopez et al., 2014; Sherwen et al., 2016a) but without accounting for pH dependence (McFiggans et al., 2000).

The inorganic iodine distribution of emissions (HOI and I2) are prescribed from a climatology computed by the CAM-Chem model considering sea–air fluxes as a function of the oxidation of aqueous iodide by atmospheric ozone on the ocean surface parameterised by Carpenter et al. (2013) and MacDonald et al. (2014) and which depends on the ozone deposition to the ocean surface, the sea surface temperature and the wind speed (Prados-Roman et al., 2015b).

The aim of this simulation is to explore how the sensitivity of ozone to anthropogenic emissions perturbations have changed since the pre-industrial period (a present-day climate is thus considered). Only emissions from natural sources are included for the simulation of the pre-industrial troposphere, following the approach used by other modelling teams to quantify this perturbation. Emissions of halogenated compounds from biomass burning are reduced to 10 % of their current values (Wang and Jacob, 1998). The methane concentration is set at 700 ppbv (IPCC, 2013). Emissions of biogenic volatile organic compounds are kept at their present-day level due to uncertainties in the effect of their drivers over the past century and the resulting high uncertainty regarding the quantification of change in their emissions (Szopa et al., 2021).

Organo-iodine emissions are almost unchanged between the PI and PD. For organo-bromines, pre-industrial emissions of CHBr3 and CH2Br2 are equivalent to the PD ones, following the methodology of Parrella et al. (2012). The concentration of CH3Br is set to ∼5 ppbv and is considered to be in agreement with the measurements from ice cores (Saltzman et al., 2004). Pre-industrial emissions of CH3Cl, CHCl3 and CH2Cl2 are from Carpenter et al. (2014), Hu (2012), Montzka et al. (2010), and Worton et al. (2006). HOI and I2 emissions, pre-computed by CAM-Chem, are ∼40 % lower than today because they are sensitive to atmospheric ozone content which was lower in the pre-industrial era (Prados-Roman et al., 2015a; Sherwen et al., 2017). Global emissions for each primary halogenated compound are reported in Table 6.

Iodine chemistry is the most impactful with regard to the ozone loss. The simplified iodine cycle is shown schematically in Fig. 4. The burden of inorganic iodine excluding IBr or ICl and iodine in aerosols is ∼22.0 Gg in the troposphere in LMDZ-INCA, comparable to the value of ∼27.9 Gg reported by Sherwen et al. (2016a, Fig. 7). HOI and OIO species, whose photolysis result in atomic iodine production, are the main intermediates for direct ozone loss. HOI is the most abundant iodine species, and its wet deposition controls the total atmospheric iodine burden. Its tropospheric burden of 12.8 Gg I in INCA is very close to the 13 Gg simulated by Sherwen et al. (2016b). The tropospheric burden of OIO in INCA, 0.43 Gg I, is comparable to the 0.55 Gg I also reported in Sherwen et al. (2016b). Overall, the global levels of HOI and OIO simulated by INCA are similar to their simulation by other models described in the literature.

The tropospheric burdens of HI, INO2 and I2 sum up to 1 Gg I, lower than the burdens reported for GEOS-Chem and totalling 2.48 Gg I (Sherwen et al., 2016b). The global burden of I2 in INCA is lower than that in GEOS-Chem (Sherwen et al., 2016b) because in Sherwen et al. (2016b) heterogeneous reactions transform INO3, INO2 and HOI into I2, whereas in INCA, these reactions produce IBr and ICl, similarly to the heterogeneous reactions adopted by Sherwen et al. (2016a). Other quantifications in the literature are lacking to explain the differences in INO2 and HI; however, Saiz-Lopez et al. (2014) indicate that minor iodine species (Imin= HI + OIO + INO2+ INO) represent less than 5 % of the mass in tropospheric inorganic iodine (2.3 % in our case). In INCA, the tropospheric Iy mean is 1.3 pptv (parts per trillion volume), whereas previous model studies found a range of ∼0.5 to 1 pptv (Saiz-Lopez et al., 2014; Sherwen et al., 2016a, b). A large set of observations has been assembled by Sherwen et al. (2016b) for IO. Measurements of IO during oceanic campaigns show a strong diurnal variation (IO being present only during daytime) with climatological concentrations (local observations averaged over 20∘ bins) comprised between 0.3 and 1 pptv. These IO concentrations show a weak latitudinal variability and a rapid decrease with altitude to reach values of about 0.2 pptv from 2 km to the tropopause (observations gathered and reported by Sherwen et al., 2016b). Surface values simulated by INCA agree within a factor of 2 with these observations. Over oceans, the annual mean surface concentrations of IO is approximately 0.2 pptv with concentrations of between 0 and 2 pptv. This is similar to the range simulated by Sherwen et al. (2016b) (0.25–1 pptv) and to the observations reported by Prados-Roman et al. (2015b). In the tropics (20∘ S–20∘ N), levels of surface IO in INCA (0.35 pptv) are in agreement with model results reported for CAM-Chem (∼0.4 pptv) (Saiz-Lopez et al., 2014). In the free troposphere, IO levels slightly increase with altitude to reach diurnal values of ∼0.5 pptv at 10 km in INCA. This overestimation of IO in the upper troposphere is common to all current model-simulated iodine and remains unclear (Sherwen et al., 2016b; Badia et al., 2019).

The IOx family influences the oxidative capacity of the troposphere through the catalytic cycles of ozone depletion (Brasseur, and Solomon, 2005), modifies the HOx (HO2/OH) and NOx (NO2/NO) ratios (Bloss et al., 2010), produces photolabile higher-order iodine oxides (IxOy) (Gómez Martín et al., 2020; Lewis et al., 2020), and generates different forms of inorganic iodine (Saiz-Lopez et al., 2012b). For that reason, a correct representation of the branching ratios and the generation and loss of IOx is crucial. Globally, IOx production in INCA is dominated by inorganic photolysis of HOI (80.5 %), OIO (10.2 %), INO3 (5.4 %) and I2Ox (3.9 %). This partition is comparable to the one reported in Sherwen et al. (2016a), who found a 76 % HOI photolysis contribution and a 11 % OIO photolysis contribution. The main IOx loss pathway is the production of HOI by IO reaction with HO2 (78.5 %), with additional loss pathways by self-reaction and reactions with NOx and BrO contributing 11.3 %, 5.7 % and 4.5 %, respectively – numbers that are very close to the branching ratios reported in Sherwen et al. (2016a) (77 %, 10 %, 7.7 % and 4.6 %, respectively).

The tropospheric bromine budget and simplified recycling cycle are illustrated in Fig. 5. The tropospheric inorganic bromine burden is ∼27 Gg in LMDZ-INCA, close to the 28 Gg reported for the most recent version of GEOS-Chem (Wang et al., 2021). On an annual average, INCA simulates daytime BrO values in the free troposphere of ∼0.3 pptv within the tropics, close to the results in CAM-Chem (∼0.2 pptv) (Fernandez et al., 2014). At the tropical tropopause layer, the modelled tropical BrO concentrations in INCA are ∼3 pptv, close to the observed concentrations of 2.0±1.5 pptv and of 3.2±1.6 pptv at two different altitudes performed by Dorf et al. (2008) and reported by Fernandez et al. (2014) for the evaluation of CAM-Chem. Overall, the global tropospheric mean of BrO simulated by INCA is 0.6 pptv (in agreement with model results from Parrella et al., 2012) but is in the lower end of the 0.5 to 2 pptv range inferred from observations (Dorf et al., 2008; Saiz-Lopez and von Glasow, 2012; Stone et al., 2012).

In this first implementation of halogen chemistry in a version of LMDZ-INCA, the focus is on gaseous chemistry. Aqueous chemistry in droplets or on ice crystals in clouds is not accounted for in the first version of the scheme because it is complex and uncertain and, more importantly, might not be a dominant mechanism for global ozone loss. Nonetheless, the scheme includes recycling through heterogeneous chemistry on sea salt and sulfate aerosols. A study by Badia et al. (2019) using WRF-Chem shows that heterogeneous halogen chemistry (on all types of aerosols) allows a better vertical Bry partitioning but contributes little (∼6 %) to ozone loss compared to the gas-phase halogen chemistry (25 %) on a global scale. Our choice to include, in this first step, reactions on sea salt and sulfate aerosols only is suited to reproduce bromine-mediated ozone loss in the marine boundary layer but probably tends to slightly underestimate the halogen-mediated ozone loss in the global troposphere.

Due to this lack of in-cloud halogen chemistry in cloud droplets and ice crystals, whose importance increases with altitude, the bromine chemistry in INCA is less dynamic and the recycling of bromine is less efficient in LMDZ-INCA compared to those reported for some versions of GEOS-Chem (Parrella et al., 2012; Schmidt et al., 2016; Sherwen et al., 2016a; Zhu et al., 2019). Note, however, that the recently updated HOBr heterogeneous chemistry implemented in the most recent GEOS-Chem version (Wang et al., 2021) results in much slower recycling of HOBr in clouds and aqueous aerosols than previously calculated. This results in fluxes from heterogeneous reactions (specifically HOBr + HBr → Br2+ H2O and HOBr + HCl → BrCl + H2O) that sum up to 32 500 Gg Br yr-1, which is smaller for example than our estimates of 86 000 Gg Br yr-1 but remains of the same order of magnitude. Finally, our branching ratios are found to be consistent with those in Schmidt et al. (2016) and Wang et al. (2021). This leads us to conclude that, although LMDZ-INCA does not integrate the same degree of details in tropospheric halogen heterogeneous chemistry compared to its peers, global fluxes in the bromine cycle of INCA appear to be more comparable in terms of the bromine cycle with those in the newest version of GEOS-Chem (see the detailed bromine budget and cycle in Fig. S1) and CAM-Chem (Iglesias-Suarez et al., 2020) than fluxes in the older versions of GEOS-Chem (Parrella et al., 2012; Schmidt et al., 2016; Sherwen et al., 2016a; Zhu et al., 2019). All these estimates remain of the same order of magnitude, but the range of fluxes illustrates the level of uncertainties about the strength of the halogen heterogeneous recycling.

The total tropospheric burden of Cly (inorganic chlorine) is 290 Gg of Cl, largely dominated by HCl. The total load of HCl is smaller than the one reported for GEOS-Chem (Wang et al., 2019) essentially because of the difference in the representation of the emissions. However, the total burden of reservoir species has a similar magnitude (10.4 Gg Cl). The halogen chemistry in INCA leads to a 25 % higher quantity of available reactive chlorine compounds: Cl and ClO* (ClO* being ClO + ClOO + OClO + 2Cl2O2).

Atomic chlorine has two main chemical sinks. The first and most important in terms of reaction flux is the transformation of atomic Cl into ClO*. This conversion takes place mainly by reaction with ozone (99.5 %), while the conversion of ClOx to Cl is dominated by reaction with NO (93.9 %). The second path is represented by the chemical loss of reactive chlorine Cl to inactive chlorine HCl (Fig. 6), which is due to the reactions of Cl with methane and several NMVOCs. The conversion of Cl to ClO* is in strong competition with the second route. By dividing the fluxes of the conversion Cl to ClO* by the flux of conversion of Cl to HCl, we find that the chain length is short and equal to 1.3 (4.8×104/3.6×104). This is close to the chain length of 1.6 calculated by Wang et al. (2019) and shows that the branching ratios are of the same order of magnitude in both models despite differences in NMVOC and CH4 concentrations between the two models. Overall, we find that the partition between Cl and ClO* is correct in the model as well as the balance between radical and non-radical species, although the latter is a little more shifted towards reactive species in INCA compared to GEOS-Chem and CAM-Chem.

The average of climatological observations (mean over several months and representative of large areas) over six oceanic and coastal remote locations shows a mean surface HCl mixing ratio of 196±114 pptv, whereas INCA simulates an average of 193±98 pptv, compared to the 246±90 pptv in GEOS-Chem (Wang et al., 2019) (see Fig. S2). However, for continental regions, LMDZ-INCA underestimates the HCl concentrations compared to observations mainly because it does not include industrial emissions. ClNO2 over continental areas does not exceed a few tens of pptvs in INCA, whereas observations at several continental locations, as gathered by Wang et al. (2019) (their Table 5), show concentrations of a few hundreds to thousands of pptvs over such locations for hourly maxima. This difference can arise from the coarse resolution of the model which does not resolve the spatial heterogeneity of these local observations (Sherwen et al., 2016a) and the absence of ClNO2 production in sulfate–nitrate–ammonium aerosols (Wang et al., 2019). The chlorine Cl* measurements to which we compare LMDZ-INCA results represent the sum of Cl2, HOCl, ClNO2, ClNO3 and other minor components as explained in Pszenny et al. (1993). LMDZ-INCA Cl* values are of the same order of magnitude as measurements (see Table S3).