We used GEOS-Chem version 14.4.0, a state-of-the-art global chemical transport model that includes detailed oxidant-aerosol chemistry in the troposphere and stratosphere (Bey et al., 2001). Aerosol thermodynamic calculations are performed using the HETerogeneous vectorized or Parallel (HETP) module for estimating NH3-NH4+, HNO3-NO3-, and HCl-Cl-, along with aerosol properties such as pH and liquid water content (Miller et al., 2024). HETP does not calculate the thermodynamic partitioning of sulfate as it does for semi-volatile species. Sulfate is formed kinetically via chemical oxidation reactions and is assumed to reside entirely in the aerosol phase.

Global anthropogenic emissions are from the Community Emissions Data System (CEDS v2) with aircraft emissions from the Aircraft Emissions Inventory Code (AEIC) 2019 (Simone et al., 2013). Shipping emissions of NOx (NO + NO2) are calculated in the PARANOx module (Holmes et al., 2014; Vinken et al., 2011). Marine emissions of dimethyl sulfide (DMS) are from Breider et al. (2017) based on Lana et al. (2011) (Breider et al., 2017; Lana et al., 2011). Wet and dry deposition (including gravitational settling) of aerosols and gases are from Liu et al. (2001), Emerson et al. (2020), and Li et al. (2023) (Emerson et al., 2020; Li et al., 2023; Liu et al., 2001). Photolysis rates are computed in Cloud-J (Prather, 2015). GEOS-Chem Classic simulations in this study were conducted at 4°×5° resolution with 72 vertical levels driven by MERRA-2 meteorology. Model runs were conducted for the year 2022 with 1 year spin-up (see Table 1 for configuration details).

Online sea salt aerosol emissions from the sea surface and blowing snow are from Jaeglé at al. (2011) and Huang and Jaeglé (2017). The current halogen chemistry in GEOS-Chem already includes sea salt debromination, anthropogenic HCl and aerosol chloride emissions, Iy uptake on alkaline sea salt aerosol, and stratospheric halogen chemistry (Eastham et al., 2014; Sherwen et al., 2016a, b; Wang et al., 2019, 2021; Zhang et al., 2022). Modeled sea salt aerosols are assumed to have an initial pH of 8 upon emission, after which their alkalinity begins to be titrated by uptake of gas-phase acidic species (HNO3, SO2, HCl). Fine mode SSA pH is calculated in HETP while coarse mode sea salt aerosols are assumed to have a pH of 5 after all of the initial alkalinity is depleted.

The continental chlorine emission inventory is described in Zhang et al. (2022). HCl emission factors from open fires are from Andreae (2019) utilizing Global Fire Emissions Database version 4 (GFED4) (Andreae, 2019). Organic halogen gases, bromocarbons, and iodocarbons are from Meinhausen et al. (2017), Bell et al. (2002), Liang et al. (2010), and Ordóñez et al. (2012) (Bell et al., 2002; Liang et al., 2010; Meinshausen et al., 2017; Ordóñez et al., 2012). These gases are photolyzed in the model to form reactive inorganic Iy. Surface emissions of inorganic iodine in the model are driven by O3 deposition on the sea surface, which reacts with iodide in seawater to produce I2 and HOI. Sea surface iodide concentrations in version 14.4.0 are from MacDonald et al. (2014), though this scheme has a known low bias to underestimate sea surface iodide by more than a factor of two globally with especially poor performance in polar regions (Sherwen et al., 2019). More recently, Chance et al. (2014) and Sherwen et al. (2019) used sea surface iodide parameterizations that show better agreement with observations of sea surface iodide concentrations than the scheme of MacDonald et al. (2014) (Chance et al., 2014; Pound et al., 2024; Sherwen et al., 2019). However, these new schemes have not been included in the GEOS-Chem base model yet and are not considered in this work.

Here we expand upon the heterogeneous and gas-phase iodine chemistry in GEOS-Chem. Additions include size-resolved, speciated iodine aerosol and heterogeneous dehalogenation of aerosol iodide to the gas-phase by reaction with hypohalous acids (HOCl, HOBr, and HOI, referred to collectively as HOX) and halogen nitrates (ClNO3, BrNO3, and INO3, referred to as XNO3). Additions to the model chemical mechanism are described in detail below.

Table 1 details the assumptions for the three model simulations reported in this work. The base model is out-of-the-box GEOS-Chem 14.4.0 under the GEOS-Chem Classic configuration. The new iodine chemistry simulation incorporates speciated aerosol iodine, iodide dehalogenation chemistry, aerosol uptake of HIO3, and HIO3 new particle formation. New iodine chemistry (no HIO3 NPF) is the same as new iodine chemistry without NPF (Eqs. 4 and 5).

Global observations of speciated iodine aerosol are compiled from Gómez Martín (2022b), who synthesized all prior literature, including surface measurements from both site-based and shipborne campaigns. Surface HIO3 observations are from He et al. (2021), another synthesis of all prior measurements (Beck et al., 2021; Finkenzeller et al., 2023; He et al., 2021a, b; Jokinen et al., 2018; Sipilä et al., 2016; Thakur et al., 2022; Zhang et al., 2024). We compiled surface IO observations including site-based and cruise measurements (Allan et al., 2000; Butz et al., 2009; Carpenter et al., 2001; Gómez Martín et al., 2013; Grilli et al., 2012, 2013; Großmann et al., 2013; Huang et al., 2010; Inamdar et al., 2020; Mahajan et al., 2010b, a, 2012, 2021; Oetjen, 2009; Peters et al., 2005; Prados-Roman et al., 2015a; Read et al., 2008; Saiz-Lopez et al., 2008; Saiz-Lopez and Plane, 2004; Stutz et al., 2007). Vertical profiles of non-sea-salt iodine aerosol are from the NASA Atmospheric Tomography Mission (ATom) (Schill et al., 2025). Vertical profiles of IO are from the Tropical Ocean Troposphere Exchange of Reactive Halogen Species and Oxygenated VOC (TORERO) and Convective Transport of Active Species in the Tropics (CONTRAST) campaigns (Koenig et al., 2020; Pan et al., 2017; Volkamer et al., 2015, 2020; Volkamer and Dix, 2017). We use over two decades of iodine observations to ensure adequate spatial coverage for model and measurement comparison, even though the model was run for the year 2022. This may introduce uncertainties in model and measurement comparison in regions with high variability in iodine emissions (e.g., with strong interannual variability in surface ozone concentrations or marine biogenic production). The spatial coverage of the observations used for model comparison may be viewed in Fig. 2. See the data availability section for links to access these datasets.

Figure 2 compares modeled annual-average, surface bulk aerosol iodine (a–d), HIO3 (e), and IO (f) from the new iodine chemistry model simulation with surface observations (Gómez Martín et al., 2022b; Großmann et al., 2013; He et al., 2021a; Prados-Roman et al., 2015b). Modeled and observed total aerosol iodine concentrations peak in the tropics, with the highest concentrations in the equatorial Northern Hemisphere. Modeled SOI (Fig. 2b), iodate (Fig. 2c), and iodide (Fig. 2d) aerosols exhibit different spatial patterns despite similarities in their zonal distribution. SOI concentrations tend to be higher in biogenically productive marine environments (e.g. the equatorial Pacific and regions with coastal upwelling) and where iodide is also abundant. Iodate concentrations are enhanced in the Mediterranean and off the West coasts of the United States and Africa due to higher HIO3 abundance in these regions (Fig. 2c and e). Because iodide abundances result from gas-phase diffusion of a myriad of Iy species onto aerosol as well as the decomposition of SOI and iodate, it has the most diffuse spatial pattern (Fig. 2d). The model predicts that all three aerosol iodine species have the highest concentrations over the North Indian Ocean, a region that currently does not have speciated or bulk aerosol iodine observations.

Figures 2e, f and 3a, b show modeled annual-mean HIO3 (e) and IO (f) compared to surface observations. The model predicts surface HIO3 between 0.001–1.1 ppt. The normalized mean bias for modeled HIO3 in the new iodine chemistry simulation is -61.8 % for mean observed HIO3 concentrations in 3a and -17.7 % for median observed HIO3. Modeled HIO3 compares well with observations where available, though it's worth noting that the regions with the highest surface HIO3 in the model do not have observations available for model evaluation (Figs. 2e and 3a) (Beck et al., 2021; Finkenzeller et al., 2023; He et al., 2021a, b; Jokinen et al., 2018; Sipilä et al., 2016; Thakur et al., 2022; Zhang et al., 2024). HIO3 mixing ratios are enhanced near the West Coast of the US, the Mediterranean, and the Indian Ocean (Fig. 2e). The spatial pattern of HIO3 concentrations follow IO in the model since O3 deposition enhances the emission of Iy from the sea surface in these regions.

Figure 2f shows that GEOS-Chem performs fairly well in reproducing surface IO over the open ocean (Allan et al., 2000; Butz et al., 2009; Carpenter et al., 2001; Gómez Martín et al., 2013; Grilli et al., 2012, 2013; Großmann et al., 2013; Huang et al., 2010; Inamdar et al., 2020; Mahajan et al., 2010b, a, 2012, 2021; Oetjen, 2009; Peters et al., 2005; Prados-Roman et al., 2015a; Read et al., 2008; Saiz-Lopez et al., 2008; Saiz-Lopez and Plane, 2004; Stutz et al., 2007). Figure 3b compares surface IO with the zonal-mean IO in the base and new iodine chemistry simulations, where, on average, the new iodine chemistry version underestimates surface IO by -0.5 ± 1.6 ppt. The normalized mean bias for IO in the new iodine chemistry simulation is -62.7 %, a slight improvement from the base model which had a normalized mean bias of -68.9 %. The largest differences in measured and modeled IO are at Mace Head in Ireland, the Isles of Shoals in Maine, Roscoff, France, and Halley Station, Antarctica, demonstrating that the coarse model resolution is not able to capture concentrated emissions from coastal iodine hot spots (Fig. 2f) (Alicke et al., 1999; Furneaux et al., 2010; Huang et al., 2010; Mahajan et al., 2009; Saiz-Lopez et al., 2007; Saiz-Lopez and Plane, 2004; Thurlow et al., 2014; Wada et al., 2007; Whalley et al., 2007). It is also possible that uncertainties in current chemical mechanisms and emissions of iodine contribute to the model biases in these coastal hot spots.

The overall bias for modeled HIO3 and IO may be improved further by changing the sea surface iodine emission scheme from MacDonald et al. (2014) to Sherwen et al. (2019), which had 80 % higher sea surface iodide concentrations on average with the largest increases in emissions in polar regions (Sherwen et al., 2019). When the Sherwen et al. (2019) scheme was incorporated into GEOS-Chem v14.1.1 by Pound et al. (2024), the relative mean bias of IO shifted from -0.43 ppt with the MacDonald et al. (2014) scheme to +0.43 ppt, with the largest increases occurring in the polar regions. However, v14.1.1 had sea-salt debromination deactivated and did not include speciated aerosol iodine or iodide dehalogenation, making it unclear how the updated surface emissions would affect Iy under the new iodine chemistry. Sea surface iodide concentrations from Sherwen et al. (2019), which were used in GEOS-Chem in Pound et al. (2024), have not been implemented in the base iodine emission scheme of GEOS-Chem and are therefore not considered in this work.

Figure 4 shows that adding speciated aerosol iodine and iodide dehalogenation to GEOS-Chem significantly improved agreement between modeled and measured bulk soluble aerosol iodine. While the 14.4 base model overestimated aerosol iodine, the new iodine parameterization brings the model much closer to the 1:1 ratio line (Fig. 4a and b). The model also shows a stronger correlation with observed total aerosol iodine for latitudes less than 30° N, with an r2 values of 0.76 compared to 0.34 for the new model and base model, respectively. The normalized mean bias of total aerosol iodine for latitudes below 30° N is +7.3 %, showing a substantial improvement over the base model, which had a normalized mean bias of +285.9 %. For latitudes >30° N, the model underestimates SOI, iodide, and iodate, suggesting missing iodine sources at mid- to high-latitudes in the Northern Hemisphere (Figs. 4 and C1–C3). Zonal comparisons of modeled and measured SOI, iodide, and iodate (including size-resolved observations) may be found in Figs. C1 and C2.

In general, the model underestimates mean bulk SOI and iodide by -0.25 ± 1.0 and -0.16 ± 0.2 ppt, respectively (Fig. 4c and d). On the other hand, iodate is overestimated for latitudes <30° N, which compensates for the underestimate in SOI and iodide (Fig. 4e). The normalized mean biases for SOI, iodide, iodate, and total aerosol iodine for all latitudes are -83.3 %, -72.3 %, +21.0 %, and -53.4 %, respectively. The normalized mean bias for total aerosol iodine in the base model for all latitudes is +22.1 %. The lower bias overall is due to the overestimate for aerosol iodine at latitudes less than +30° N, which compensates for the underestimate at latitudes greater than +30° N.

Aerosol iodine interconversion rates could theoretically alter the individual concentrations of the aerosol iodine species to increase the abundances of SOI and iodide relative to iodate. However, by increasing the conversion rate of iodate to iodide, the reactions between HOX and XNO3 with iodide are too efficient to preserve aerosol iodide before it is recycled to the gas-phase in the model. Tuning the aerosol iodine interconversion rates in the model is somewhat arbitrary, since changes in aerosol surface area and acidity in the model can alter the acid-catalysed reaction rates for HOX + iodide. Additionally, it's likely that we are missing sources of SOI that are not derived from primary marine organic aerosol, such as secondary marine organic aerosol and continental organic sources. Global observations from Gómez Martín et al. (2022b) suggest that 20 %–35 % of total aerosol iodine is non-soluble and likely derived from combustion and biomass burning, which is not currently considered in this model. Adding additional SOI sources would also help address the low bias in this study. Speciated aerosol iodine observations are strongly correlated with the amount of observed total soluble iodine at all latitudes, with r2 of 0.76, 0.92, and 0.66 for SOI, iodide, and iodate, respectively (Fig. C3). As this study is the first attempt of modeling speciated aerosol iodine at the global scale, the main goal was to reproduce total soluble iodine observations. The low bias for SOI and iodide and high bias for iodate all suggest that the rates used for aerosol iodine interconversion in this study still need to be refined once quantitative experimental results are available.

The poor model agreement with observed total aerosol iodine shown in Fig. 4 at latitudes higher than 30° N arises from severe underestimates of aerosol iodine in the North Atlantic (mainly at Mace Head, Ireland) and off the coast of Northern Alaska, which is also evident in Fig. 2a–d. The inability of GEOS-Chem to reproduce iodine observations at Mace Head is not surprising. The highly productive algae beds make this site a volcano of reactive iodine emissions to the atmosphere, with a median observed total aerosol iodine of 5.2 ppt and observations up to 37 ppt during the Marine Aerosol Production (MAP) (2006) campaign (Gilfedder et al., 2008). Nearly all of the aerosol iodine at Mace Head is SOI (96 % ± 4 %), with 58 % and 42 % of SOI in the coarse and fine mode, respectively, suggesting that a large portion is primary SOI (Fig. C1) (Gilfedder et al., 2008). Other regions rich in kelp, including The Russian Far East, the West Coast of South America, and the West Coast of Australia (Eger et al., 2023), do not exhibit the same underestimate in aerosol iodine, suggesting that this is not a systematic issue in the model for the lower latitudes (Fig. 2b) The underestimate of SOI at latitudes >30° N, however, does point to the need for more SOI sources in the model (Figs. 4c and C1, C2).

Another known missing source in GEOS-Chem is iodine emissions from snow and ice, which have also previously been demonstrated to be an important bromine source contributing to Arctic O3 depletion events (Raso et al., 2017; Thompson et al., 2015). The snowpack iodine in Brown et al. (2025) appeared to originate from the deposition of aerosol iodine onto the snowpack and subsequent re-emissions through aerosol recycling back to Iy. Since observations suggest that both iodine and bromine are enriched in snowpack relative to seawater, this iodine source should be considered in future work (Brown et al., 2025; Celli et al., 2023; Raso et al., 2017).

Figure 5a–i shows modeled vertical profiles of annual-mean aerosol iodine concentrations and their speciated fractional contributions to total aerosol iodine for the tropics, mid-latitudes, and polar regions. Non-sea-salt aerosol iodine (nssI) observations from the ATom campaign (plotted in silver in Fig. 5a–c) provide a lower limit for fine-mode aerosol iodine in the atmosphere (Schill et al., 2025). During ATom, aerosol iodine was detected in several types of non-sea-salt organic particles including sulfate-rich organic aerosol, biomass burning particles, and metal-rich particles associated with ship emissions (Schill et al., 2025). While the concentration of iodine in non-sea salt particles is reported, the partitioning between SOI, iodide, and iodate in the observations is unknown. Schill et al. (2025) showed that nssI is ubiquitous in the upper atmosphere, despite the distance from the ocean surface, which is the main source of atmospheric iodine. The vertical profile of nssI closely resembles the profile of non-sea salt organic particles (Schill et al., 2025) (Fig. 5a–c). While the speciation of the ATom nssI observations is unknown, one explanation for the correlation between organics and nssI could be the formation or transport of SOI to the upper troposphere.

Figure 5a–c shows that fine-mode iodine aerosol is likely underestimated in the model compared to the nssI ATom observations. Fine aerosol iodine in the base model (plotted as a dashed black line) is <0.01 ppt in polar regions, underestimating fine aerosol iodine by more than an order of magnitude (Fig. 5a). Fine aerosol iodine in the base model shows an even steeper decrease in mass with height, resulting in worse agreement with observations in the upper troposphere. Part of the increase in modeled upper-troposphere aerosol iodine compared to the base model is from HIO3 new particle formation in the upper atmosphere. Because SOI and IO3- are not directly recycled to Iy in the model, their presence helps preserve the aerosol iodine to allow it to be transported away from the surface. Despite these additions, the amount of fine iodine in the upper troposphere is still underestimated compared to observations. The model's underestimation in the upper troposphere could be due to the underestimated surface emissions with the MacDonald scheme in addition to issues with aerosol iodine transport, deposition, or a combination of these factors. Due to the high abundance of HOX and XNO3 in the lower troposphere, the mean lifetime of fine mode iodide against dehalogenation is only 12 min, resulting in low iodide abundance throughout the troposphere (Figs. 4a–c and C1). Because HOX dehalogenation reactions are acid-catalysed, fine-mode aerosol in the model may be too acidic, making the recycling rate to form Iy too fast. Alternatively, surfactants on the aerosol surface may slow down dehalogenation, which is not currently considered in the model.

Figure 5d–f includes coarse mode aerosol concentrations in addition to the fine mode concentrations to represent total aerosol iodine in the model. Total aerosol iodine peaks at the surface and decreases with height through the mid-troposphere. In the upper troposphere, this trend reverses and aerosol iodine begins to increase with height. Schill et al. (2025) also observed an increase of nssI with height in the upper troposphere during ATom. Figure 5g–i shows that the modeled fraction of SOI increases with height in the tropics and mid-latitudes. This is also consistent with the hypothesis that organics have a stabilizing effect on aerosol iodine in the troposphere by binding with iodide to form adducts, effectively slowing down reactive iodine chemistry by not allowing iodide to be recycled back to the gas phase. While we lack laboratory studies that explicitly investigate reaction rates between iodide and organic aerosol to form adducts, observations from Yu et al. (2019) found that iodide-organic adducts constituted 64 % ± 8 % of aerosol iodine at their inland site 200 km from the coast, which was higher than the coastal site contribution at 31 % ± 16 % of total aerosol iodine. In the aged particles from their study, SOI contributed 76 % ± 7 % of total aerosol iodine, which further supports the hypothesis of organics having a stabilizing effect on aerosol iodine.

Modeled fine mode iodate is the dominant form of aerosol iodine above the boundary layer in the troposphere (Fig. 5g–i). Koenig et al. (2020) found an increase in the fraction of iodate moving upward from the upper-troposphere to the lower stratosphere based on I+:HI+:I2+ ratios in aerosol mass spectrometry (AMS) observations. They report a mean iodate fraction of 56 % compared to total aerosol iodine in the upper troposphere, however, only one organic iodine aerosol species (5-iodo-2-furfural) was considered, so the contribution of iodate to total aerosol iodine is not easily comparable to this study (Koenig et al., 2020). Obtaining more speciated aerosol iodine observations in the upper atmosphere would shed light on the relative stabilities of SOI, iodate, and iodide as they age in the atmosphere and provide further guidance for model improvement. Additionally, the relative distribution of aerosol iodine in non-sea salt vs. sea salt particles is unknown in the upper atmosphere, though sea salt abundances decrease by around a factor of 10 for every 2 km in altitude, with very low sea salt abundance in the upper troposphere (Murphy et al., 2019). This likely means that nssI aerosol becomes a relatively larger contributor to total aerosol iodine in the upper troposphere, as sea salt particles are more efficiently removed by deposition due to their high solubility and larger size. Quantifying the abundance and speciation of iodine in different particle types as a function of altitude would further aid in model improvement.

Figure 6 shows vertical-mean profiles and zonal-mean surface Iy composition for the base model (a and b), the new iodine chemistry simulation (c and d), and the new iodine chemistry (no HIO3 NPF) simulation (see Table 1 for model configuration information). HIO3 is a large contributor to Iy in the troposphere in the model. The HIO3 concentration increases with height and peaks in the mid-troposphere (around 600 hPa). In the sensitivity study where HIO3 NPF is not included (“no HIO3 NPF”), a large bubble of HIO3 forms in the upper troposphere, with mean concentrations around 1 ppt (Fig. 6e). HIO3 accumulates as a function of altitude without NPF due to the lack of existing aerosol surface area for uptake of HIO3 to form iodate. This makes HIO3 the globally dominant iodine species in the upper atmosphere, contributing almost all of total Iy upward from 800 hPa (Fig. 6e). The difference in the tropospheric burden of HIO3 in the “no HIO3 NPF” and “new iodine chemistry” simulations in Fig. 6 is large, contributing 6.6 and 2.4 GgI of HIO3, respectively. In the “new iodine chemistry” simulation, HIO3 NPF contributes 24 % of the HIO3 loss rate, with fine aerosol uptake and coarse aerosol uptake contributing 46 % and 25 %, respectively, and deposition contributing the remainder of HIO3 loss rate at 5 %. These results suggest that HIO3 is a critical terminal product of iodine in the gas phase and that the NPF mechanism is an important sink for HIO3 in the upper troposphere in the model, allowing for iodate formation and recycling back to other forms of Iy. This further indicates that global-scale simulations should include HIO3 when assessing iodine's impact on atmospheric oxidizing capacity and particle formation.

With the added speciated aerosol iodine and HIO3 chemistry, the abundance and composition of Iy changed compared to the base model. In the “new iodine chemistry” simulation, non-HIO3 Iy increased by 75 % while total Iy (including HIO3) increased from 11.5 to 23.1 Gg in the troposphere, more than doubling the total Iy abundance (Tables 2 and D1). The largest increases in Iy were due to iodide dehalogenation to form ICl, IBr, and I2. The total dihalogen iodine concentration increased by a factor of 11, with ICl, IBr, and I2 concentrations increasing by factors of 7, 22, and 47, respectively (Table D1). This corresponded to a 74 % decrease in total aerosol iodine (from 6.9 Gg total aerosol I in the base model to 1.8 Gg in the new iodine chemistry simulation) (Tables 2 and D1).

The updated iodine chemistry in GEOS-Chem increases IO concentrations in the upper-troposphere in the model, which we compare with observations from two aircraft campaigns, TORERO and CONTRAST (Fig. 7a–d) (Koenig et al., 2020; Pan et al., 2017; Volkamer et al., 2015, 2020; Volkamer and Dix, 2017). IO was measured during both campaigns by the CU AMAX-DOAS instrument. The Tropical Ocean Troposphere Exchange of Reactive Halogen Species and Oxygenated VOC (TORERO) campaign in the eastern Pacific measured IO profiles between 1000–200 hPa, observing mean concentrations of 0.19 ± 0.17 ppt (Fig. 7a) (Koenig et al., 2020; Volkamer et al., 2015; Volkamer and Dix, 2017). Selecting gridboxes along the flight path between 12:00–15:00 LT to represent daytime concentrations in the model, the new iodine simulation performs better at reproducing mean IO concentrations during TORERO (0.12 ± 0.12 ppt) compared to the base model (0.06 ± 0.11 ppt) (Fig. 7a). This reduces the magnitude of normalized mean bias of IO in the new iodine chemistry simulation (-39.3 %) compared to the base model (-65.1 %).

Similarly, the Convective Transport of Active Species in the Tropics (CONTRAST) campaign provided aircraft-based IO measurements in the Western Pacific upper troposphere (260–180 hPa), reporting mean IO concentrations of 0.12 ± 0.06 ppt. Along the flight paths between 12:00–15:00 LT, January-mean GEOS-Chem IO concentrations were 0.14 ± 0.04 ppt with the new iodine scheme, an improvement over the base model's 0.03 ± 0.02 ppt (Koenig et al., 2020; Pan et al., 2017; Volkamer et al., 2020) (Fig. 7b and d). The normalized mean bias in the new iodine chemistry simulation compared to CONTRAST observations is +29.2 %, which is a slight improvement over the base model -67.4 %, though it also represents a substantial increase in IO in the upper troposphere and leads the model to slightly overestimate compared to observations. The increase in gas-phase iodine in the upper troposphere with the new iodine chemistry is also consistent with Dix et al. (2013), who found that a substantial portion of the IO column resided above the marine boundary layer and hypothesized that its source originated from heterogeneous recycling of iodine from aerosols (Dix et al., 2013).

Above 300 hPa, the new iodine chemistry simulation yields IO levels that are closer to the TORERO and CONTRAST observations than the base model, indicating improved model performance in the upper troposphere (Fig. 7c and d). During TORERO the mean difference in IO concentration above 300 hPa between the model and observations was -0.04 ppt (compared to -0.12 ppt in the base model) and during CONTRAST the mean difference above 300 hPa was +0.02 ppt (compared to -0.08 ppt in the base model). GEOS-Chem IO in version 12.9 compared well with TORERO observations, however, the emission of iodine in the model is extremely sensitive to O3 concentrations in the marine boundary layer (Wang et al., 2021). This means that iodine abundances can vary substantially across GEOS-Chem versions along with changes in VOC, NOx, and oxidant chemistry. Even so, the improvement in modeled IO in the upper atmosphere in this work compared to observations suggests better representation of iodine in the upper troposphere with the new scheme.

Table 2 and Fig. 8 show the changes in the global iodine sources and sinks after implementing the new chemistry. The total emission of inorganic iodine in the updated model decreases by -0.12 TgIyr-1 (-5.4 %) due to a decrease in the tropospheric burden of O3 resulting in decreased sea surface reaction between iodide and O3.

The tropospheric ozone burden decreases by -2 % in the new iodine chemistry simulation compared to the base model (see Table E1). On average, surface O3 concentrations decrease by -3 %; however, the spatial pattern of O3 changes are not uniform across the two hemispheres (Fig. E1). Surface O3 increases for latitudes less than -30° S by up to +11 % over the Southern Ocean and Antarctica. The increase in surface O3 over the Southern Ocean coincides with decreases in BrO by -0.6 ppt on average. Lower BrO concentrations reduce the rate of the BrO+HO2→HOBr+O2 reaction, which is a major contributor to the ozone loss rate (Bates and Jacob, 2020). At latitudes greater than -30° S, surface O3 decreases by up to -18 %, with the largest decreases over the tropical Pacific and Indian Oceans and in the Arctic (Fig. E1). We discuss the impacts of the new iodine chemistry on oxidants further in a follow-up publication.

The net decrease in surface O3 is responsible for the -0.12 Tg decrease in inorganic iodine emissions in the new iodine chemistry simulation. Figure E2 compares annual-mean surface ozone concentrations in the base model and new iodine chemistry simulations with Tropospheric Ozone Assessment Report II (TOAR II) ship and buoy observations (Kanaya et al., 2025a, b). Overall, both the base model and new iodine chemistry simulations reasonably simulate surface ozone concentrations, with r2 values of 0.84 and 0.90 for the base model and new iodine chemistry simulations, respectively. The decrease in surface ozone for latitudes >-30° S leads to a slightly more negative normalized mean bias in the new iodine chemistry simulation (-16.0 %) compared to the base model (-10.7 %).

Iodide dehalogenation reactions produce 6.4 TgIyr-1, which is more than double the total source of Iy in the model from ocean emissions and the photolysis of organic iodine gases, which produce 2.8 TgIyr-1 (Table 2 and Fig. 8). This suggests that even though the aerosol is not the initial Iy source, the overall distribution and abundance of reactive iodine is strongly mediated by heterogeneous chemistry of aerosol iodine.

As a result of the increased Iy recycling rate, the aerosol uptake rate of Iy substantially increases from 1.0 to 7.2 TgIyr-1. The majority (62 %) of Iy to aerosol iodine conversion is from HIO3 uptake onto aerosol and new particle formation to form iodate (Table 2). The lifetime of fine mode aerosol iodate in the model is 1 h before it is converted to iodide, which can then be liberated to the gas phase via heterogeneous reactions (Reactions R3 and R4). The short modeled lifetime of fine iodate is consistent with rapid reduction of iodate to yield Iy in Reza et al. (2024) and Li et al. (2024), who observed iodate recycling even in the absence of light (Li et al., 2024; Reza et al., 2024). The overestimate of iodate compared to surface observations suggests that the conversion of iodate to either iodide or Iy is likely underestimated. The uncertainty of HIO3 uptake and subsequent recycling is compounded by the fact that the reactive uptake coefficients (γHIO3) for HIO3 in the updated iodine simulation of 0.03 and 0.10 for the fine and coarse modes, which is on the lower end of the estimates from other studies. Previous modeling studies have set γHIO3 from 0.2 to unity (Pechtl et al., 2007; Zhao et al., 2024). In a follow up paper, we evaluate the sensitivity of the model to prescribing γHIO3 to unity and iodate to iodide conversion rates on the order minutes instead of hours.

The effective lifetime of iodine in the atmosphere can be calculated by dividing the tropospheric burdens of gas and aerosol iodine by their total loss rate from wet and dry deposition to the surface (Eq. 6). 6τiodine=Total iodine burden⋅Total iodine deposition rate-1 The effective lifetime of iodine in the atmosphere increases by 42 % in the updated model from 2.4 to 3.4 d (Table 2 and Fig. 8). This coincides with the tropospheric burden of Iy increasing from 11.5 to 23.1 Gg (Table 2 and Fig. 8). The increase in the effective iodine lifetime from the speciated iodine chemistry and iodide dehalogenation is responsible for the increase in the Iy burden, where less time is spent in the aerosol phase compared to the gas phase, increasing the amount of time the iodine can persist in the atmosphere before deposition. With the addition of aerosol recycling, the contributions of Iy deposition increases compared to aerosol I deposition; with a larger contribution from Iy in the new iodine chemistry simulation (71 % of total I deposition) compared to the base model (66 % of total I deposition).

The average number of cycles that Iy is converted to aerosol I and back to Iy before deposition can be calculated by using the ratio between the iodide dehalogenation rate and the aerosol iodine deposition rate. On average, this cycle occurs 8 times in the troposphere with the current model configuration. We note that the interconversion rates between SOI → iodide, iodide → SOI, and iodate → iodide effectively dictate the overall iodide dehalogenation rates, where higher SOI and iodate concentrations relative to iodide slows down reactive iodine chemistry and increases the depositional loss from aerosol I relative to Iy. We explore the sensitivity of atmospheric oxidants to iodine parameterizations in GEOS-Chem including the rates of conversion between SOI, iodide, and iodate and model sensitivity to HIO3 NPF and reactive uptake coefficients in a follow-up paper.

We have implemented speciated aerosol iodine and heterogeneous chemistry of iodide aerosol in GEOS-Chem, largely relying on the existing network of surface observations presented in Gómez Martín et al. (2022b). We would like to emphasize that observations of the relative abundances of aerosol iodine species are uncertain. This is because speciated aerosol iodine is measured through offline filter extraction into solution, separation via ion chromatography (IC), and then measurement of iodine as a function of retention time (IC-ICPMS) (Gómez Martín et al., 2022b). Given the sensitivity of iodine speciation to pH, the speciation in the sample may be altered the moment the filter is extracted by simply diluting the solution, inducing interconversion between SOI, iodide, and potentially iodate. Several studies have also shown that sonication during sample extraction causes interconversion between SOI and iodide. Sonication can either increase the amount of iodide by causing SOI to dissociate (observed in Yodel and Baker, 2019 ), or increase the amount of SOI as iodide reacts with organics in bulk solution (observed in Baker et al., 2000). These uncertainties in measuring the speciation of aerosol iodine propagate into global modeling uncertainties. Therefore, we call for future endeavours to carry out in situ online measurement of these components to constrain the uncertainties.

We do not represent non-soluble iodine in the model, though it has been observed in the atmosphere and contributes 20 %–35 % of total aerosol iodine (Gómez Martín et al., 2022b). It is hypothesized that non-soluble iodine (NSI) aerosol is formed in hydrophobic aerosol like soot and has a continental origin (Gómez Martín et al., 2022b). Schill et al. (2025) also identified biomass burning as a source of primary aerosol iodine, though neither of these formation processes has been included in the model. NSI may be an important sink for Iy in biomass-burning plumes since they provide a surface for aerosol uptake. Shi et al. (2021) observed large increases in both primary and secondary organic iodine aerosol during the heating season in Beijing, which seemed to come from coal combustion, suggesting that there are likely also anthropogenic sources of organic iodine aerosol (Shi et al., 2021).

There are likely other missing iodine sources in the model. GEOS-Chem performs poorly in reproducing total aerosol iodine observations north of 50° N. This may be improved by transitioning from the MacDonald Iy emission scheme to the Sherwen et al. (2019) scheme, which has substantially higher Arctic emissions. Adding a snow source for iodine may also improve model performance in the Arctic.

The model is also missing continental sources of iodine. Biomass burning and anthropogenic emissions are potentially important sources of iodine (Schill et al., 2025; Shi et al., 2021). Additionally, aircraft observations of IO have indicated that dust is a potential driver of gas-phase iodine, where IO concentrations were enhanced by up to a factor of 10 within lofted dust layers off the coast of Chile, coinciding with lower ozone abundances (Koenig et al., 2021). This is in line with observations in the Canary Islands, where dust events were associated with higher gas-phase iodine abundances, with CH3I concentrations increasing by factors of 2 to 14 compared to background conditions (Puentedura et al., 2012; Williams et al., 2007). At a mountain site in Colorado, Lee et al. (2024) found that GEOS-Chem underestimated IO by about a factor of 3. This bias could be due to overestimated IO sinks, likely in part because of the previous lack of aerosol recycling and shorter effective iodine lifetime in GEOS-Chem (Lee et al., 2024). However, their findings could also indicate missing continental iodine sources. More observations of iodine in inland sites will further aid in identifying the relative importance of non-marine sources in the global iodine budget.

Because almost all speciated iodine observations are from the marine boundary layer, relatively little is known about the processes governing the relative abundance of these species as they age and are transported upwards in the atmosphere. Interconversion rates between aerosol iodine species are uncertain and are currently parameterized by first-order rate constants tuned to global surface observations. Iodate is the dominant aerosol iodine species throughout most of the troposphere in the model, though as discussed before, it is overestimated compared to surface observations (Figs. 4 and 5). It is possible, however, that iodate is the dominant species in the upper troposphere due to the efficiency of HIO3 NPF at low temperatures and its relative stability against dehalogenation compared to iodide (He, 2023; He et al., 2021a).

We do not account for ionic strength and pH in modeling the uptake of Iy onto aerosol because we use a kinetic approach, where the aerosol formation rates from HI and HIO3 are estimated based on the estimated reaction probability and aerosol surface area. In reality, this is a thermodynamic process influenced by temperature, the pH of aerosol liquid water, and the ionic strength of the solution. To better represent and understand gas-to-aerosol iodine partitioning, inclusion of HI-I- and HIO3-IO3- partitioning using a thermodynamic approach may be better.

Given the sensitivity of dehalogenation chemistry to pH, further evaluation of this model parameter is warranted. Cloud water pH in GEOS-Chem was evaluated in Shah et al. (2020), where they found good agreement between global observed and modeled cloud water pH, with mean values of 5.2 ± 0.9 and 5.0 ± 0.8, respectively (Shah et al., 2020). On the other hand, fresh sea salt aerosol has been found to rapidly acidify in the remote marine atmosphere, with aerosol pH ranging between 1.5 and 2.6 within minutes of emission, compared to the seawater pH around 8 (Angle et al., 2021). Observations of aerosol pH in the marine boundary layer are spatially and temporally sparse, making evaluating model performance in simulating aerosol pH difficult. In general, aerosol pH used to calculate dehalogenation rates ranges between -0.3–5.3 and 5–8 for the fine mode and coarse mode, respectively, which is within the ranges estimated in the literature at -1.1 to 5.3 for fine mode and 1.2–8.0 for coarse mode (Angle et al., 2021; Pye et al., 2020). More work is needed in the atmospheric chemistry community to evaluate model performance in simulating the pH of fine and coarse mode aerosol in the remote marine atmosphere.

We have added aerosol iodine dehalogenation to the model using the same rate constants for debromination. Additional BrCl, ICl, and IBr observations will help constrain aerosol dehalogenation rates, since these dihalogen species have only been measured in a handful of studies (Finley and Saltzman, 2008; Tham et al., 2021). Finally, more synchronous gas and aerosol phase observations would help us better understand the role of heterogeneous chemistry in regulating the abundance of Iy and oxidants in the atmosphere.

While there are still many uncertainties in understanding the formation and interconversion between aerosol iodine species, this study serves as a framework for future work. Adding the new SOI and iodate tracers to GEOS-Chem allows for explicit representation of the different types of aerosol iodine and examination of their potential importance in the atmosphere. Additionally, the first-order rate constants used to represent the interconversion rates between aerosol are easily tuneable until a more detailed parameterization informed by new laboratory studies is developed. Previous versions of GEOS-Chem have showed better performance in reproducing IO observations than the 14.4 base model presented in this study. In the absence of aerosol iodide recycling and speciated aerosol iodine chemistry in the base model, the abundance of Iy is largely tied to the surface abundances of ozone, which dictate the release of I2 and HOI from the sea surface. Because of this, iodine chemistry in GEOS-Chem is sensitive to the model version as NOx, VOC, and oxidant abundances are altered. It is possible that performance will vary with future model developments, as we've already observed in comparing the results of previous iodine modeling work in GEOS-Chem, including v10 in Sherwen et al. (2016b), v12.9 in Wang et al. (2021) and Schill et al. (2025), v13.2 in Lee et al. (2024), and v14.4, which served as the base model in this study.

Despite the modeled sensitivity of iodine chemistry to surface ozone concentrations, the large increase in the Iy burden and the substantial contribution of iodide dehalogenation to Iy production demonstrated in this work are expected to persist across model versions. Additionally, the treatment of aerosol iodine here differs fundamentally from previous GEOS-Chem implementations: rather than serving primarily as a depositional sink for Iy, aerosols are shown to be a significant source. This new framework for modeling aerosol iodine will ultimately make for a more robust representation in GEOS-Chem in future model releases because we present additional model parameters that can mediate the abundances and recycling of atmospheric iodine.