Anthropogenic Impacts on Tropospheric Reactive Chlorine Since the Preindustrial

Tropospheric reactive gaseous chlorine (Cly) impacts the atmosphere's oxidation capacity with implications for chemically reduced gases such as methane. Here we use Greenland ice‐core records of chlorine, sodium, and acidity, and global model simulations to show how tropospheric Cly has been impacted by anthropogenic emissions since the 1940s. We show that anthropogenic contribution of nonsea‐salt chlorine significantly influenced total chlorine and its trends after the 1940s. The modeled regional 170% Cly increase from preindustrial to the 1970s was driven by acid displacement from sea‐salt‐aerosol, direct emission of hydrochloric acid (HCl) from combustion, and chemical reactions driven by anthropogenic nitrogen oxide (NOx) emissions. Since the 1970s, the modeled 6% Cly decrease was caused mainly by reduced anthropogenic HCl emissions from air pollution mitigation policies. Our findings suggest that anthropogenic emissions of acidic gases and their emission control strategies have substantial impacts on Cly with implications for tropospheric oxidants, methane, and mercury.

Anthropogenic emissions and acid displacement of HCl can lead to enrichment or depletion of Cl relative to sodium (Na) compared to their ratio in sea water, denoted as Cl exc (Equation 2 in Methods). Legrand et al. (2002) calculated ice-core HCl after removing sea-salt and continental chloride from the measured total chloride, and attributed the increases in alpine ice cores to enhanced coal combustion and waste incineration in western Europe during 1925-1970. Observations show decreasing trends of non-SSA Cl − deposition over the past 20-30 years in the US (Haskins et al., 2020) and UK (Evans et al., 2011), suggesting that the post-1970s air pollution mitigation policies targeting SO 2 and NO x emissions have reduced emissions of HCl. At Summit (central Greenland), Legrand et al. (2002) found that Cl exc originates mainly from acid displacement of HCl from SSA, which increased by a factor of 2-3 over the twentieth century due to enhanced aerosol acidity resulting from growing anthropogenic NO x and SO 2 emissions. Greenland ice-core records of sulfate and nitrate, the main sinks for NO x and SO 2 , show increases beginning in the 1900s, peaking in the 1970s, followed by a rapid decline in sulfate and a more gradual decline in nitrate (Geng et al., 2014), consistent with trends of anthropogenic SO 2 and NO x emissions from combustion (Smith et al., 2011).

Ice-Core Records
We present ice-core chlorine, sodium, and acidity records from the six Greenland ice cores ( Figure S1). Precise locations and other information are summarized in Table S1, and details on core extraction and dating are described in previous publications (Geng et al., 2014;Iizuka et al., 2018;McConnell et al., 2019;Opel et al., 2013;Spolaor et al., 2016). Measurements of ice-core Na and Cl were using either a continuous flow analysis with an online ion chromatography system (CFA-IC) with an accuracy of 5% at annual resolution (for Summit07) (Geng et al., 2014;Iizuka et al., 2018), or Inductively Coupled Plasma Mass Spectrometry (ICP-MS) Spolaor et al., 2016) with an uncertainty of ±10% (for NEEM, NGT_B19, Tunu, ACT_11d and Summit10). For NEEM, NGT_B19, Tunu, ACT_11d and Summit10 cores, acidity (H + ) was measured directly using a flow-through bubbling chamber method described in Pasteris et al. (2012), with an error less than 5%. For Summit07 ice core, acidity was calculated based on the ion balance, according to Equation 1: with concentrations in units of μeq L −1 (Geng et al., 2014). Note that this calculation may underestimate snow acidity because it does not consider organic-acid anions (e.g., formate and acetate), which were measured to be 0.3 ± 0.1 μM at Summit during 1767-1945 (Legrand & Mayewski, 1997).
To separate the contribution of SSA relative to more reactive forms (e.g., HCl) of Cl, we calculated the chlorine excess (Cl exc ) relative to what would be expected from SSA alone, which is defined with a sea water Cl/ Na mass ratio (([Cl]/[Na]) sea water ) of 1.796 (Riley & Tongudai, 1967) ICP-MS measurements may lead to an underestimate of Cl exc because it measures both the soluble Na and the insoluble Na fraction which may originate from nonsea-salt aerosol (e.g., dust), whereas IC measures the soluble Na and a small fraction of leachable Na from dust.
To analyze the relationship between measured species in the ice cores, we adopt the Passing-Bablok (PB) regression model (Passing & Bablok, 1983). In contrast to the traditional Ordinary Least Squares regression (OLS) which only considers measurement errors on the y-axis variable, PB regression assumes both x-axis and y-axis variables contain measurement errors and is insensitive to outliers (e.g., due to volcanic eruptions). We use the Pearson's correlation coefficient (r) to show the relationships between species, and r is not affected by the choice of the regression model.

GEOS-Chem Simulations
To estimate impacts of anthropogenic emissions on tropospheric HCl and reactive chlorine abundances, we use a global 3-D chemical transport model GEOS-Chem (version 11-02d, Text S1) described in Bey et al. (2001) with updates described in the supporting information. The model is driven by MERRA-2 assimilated meteorological observations from the Goddard Earth Observing System (GEOS) (Gelaro et al., 2017). The model simulates detailed HO x -NO x -VOC-ozone-halogen-aerosol tropospheric chemistry, which includes SSA (Jaeglé et al., 2011) and tropospheric gas-phase, liquid-phase, and heterogeneous-phase reactive chlorine chemistry (X. Wang et al., 2019), and fully coupled stratospheric chemistry (Eastham et al., 2014).
Model simulations are performed using three emission scenarios: preindustrial (PI, year 1750), peak atmospheric acidity (PA, year 1975), and present day (PD, year 2007), as summarized in Text S1 and Table S2. We run each simulation for 5 years to equilibrate stratosphere-troposphere exchange, and use only the fifth year for analysis. All simulations are conducted at 4° × 5° horizontal resolution and 72 vertical levels up to 0.01 hPa. We use MERRA-2 meteorological fields of the same year (2007) for all three simulations to isolate changes induced by anthropogenic emissions. This configuration will also keep emissions that are dependent on meteorological parameters, such as wind-blown dust, lightning and soil NO x , biogenic VOCs, and SSA from the open ocean and sea-ice, constant. Only anthropogenic and biomass-burning emissions are allowed to vary between simulations in order to isolate their impacts on tropospheric chlorine.

Backward Trajectory Analysis
To determine the source regions of Cl y at the six Greenland ice-core sites, we run backward trajectory analysis using the HYSPLIT model (Hybrid Single-Particle Lagrangian Integrated Trajectory) (Stein et al., 2015). We calculate the cumulative air mass probability for the 5-day backward trajectories, considering the modeled lifetime of acidic gases (e.g., SO 2 and NO x of about 1 day), accumulation-mode aerosol (up to 6 days) (Alexander et al., 2005), and gas phase HCl (2.5 days), as well as possible seasonal variations in transport.
To retrieve the source regions across the ice-core covered time periods, we conduct the backward trajectory analysis for 1959-2010, and present the averaged results. Initial altitudes of air masses are at 10, 500, 1,000 and 1,500 m above ground level (a.g.l.), and the calculation was constrained within 1,500 m a.g.l., which was assumed to be the depth of the mixing layer. Daily precipitation from the reanalysis data sets (ERA-40 and ERA-Interim (Dee et al., 2011;Uppala et al., 2005)) was used for weighting the probability of air masses. Considering the proximity of the two Summit cores, and Tunu and NGT_B19, we only conduct the analysis for four locations: NEEM, Tunu, ACT_11d, and Summit, and chose a region that covers most of the backward trajectory probabilities as the backward trajectory region (TRJ) ( Figure S2).

Ice-Core Observations
Figure 1 shows annual concentrations of Na, Cl, Cl exc , and acidity from six Greenland ice cores ( Figure S1). Positive mean Cl exc values (1.0-5.9 ng g −1 ) over the reported time periods (1750 or 1776-end of the records) in all records are consistent with negligible loss of HCl from the snowpack after atmospheric deposition where snow accumulation rates are greater than 40 kg m −2 yr −1 (Röthlisberger, 2003) (Table S1). Cl exc represents a higher fraction of total Cl in lower latitude (medians of 9%-49% over the full records) compared to higher latitude ice cores (medians of 8%-17% over the full records) ( Figure S3) likely due to closer proximity to North American (NA) and Western European (WE) anthropogenic source regions, as determined by back trajectory analysis ( Figure S2). For all ice cores, Cl exc records show no trends before 1940, followed by a twofold to sevenfold increase until ∼1975. After 1975, Cl exc either declined (Summit07, ACT_11d) or leveled off (Summit10, Tunu, NEEM, and NGT_B19). ice-core acidity is similar to Cl exc , with no long-term trends before the 1900s (lower latitude cores) or 1940s (higher latitude cores), an increase from 1940 to 1975, followed by a leveling off or decrease. Acidity trends are consistent with previous Greenland ice-core sulfate records (Geng et al., 2014) and trends in anthropogenic sulfur emissions in NA and WE (Smith et al., 2011). Table S3 show relationships between annual ice-core Na and Cl, and between ice-core acidity and Cl exc . Na and Cl were strongly correlated in all ice cores, with a stronger relationship in higher latitude (r = 0.86-0.94) compared to lower latitude (r = 0.49-0.80) cores. Continued strong correlation after the 1940s (Table S3) suggests that SSA was the dominant source of chlorine throughout the records.
ZHAI ET AL.
10.1029/2021GL093808 4 of 12  Figure 3 shows the 30°-90°N regional distribution of modeled annual-mean surface Cl y in PI, PA, and PD, where the model considers past changes in anthropogenic emissions (Methods, and Text S1). Surface Cl y is shown because most tropospheric Cl y is confined to lower altitudes (2 km) due to the dominance of direct surface emissions and near-surface chemistry sources ( Figure S4). The highest surface Cl y concentrations in PA and PD are distributed in continental outflow regions where anthropogenic acids encounter SSA Cl − , leading to acid displacement of HCl. The tropospheric burden of Cl y in the 30°-90°N region increased 132% from PI to PA and 7% from PA to PD. While surface Cl y increases everywhere in the region from PI to PA and PI to PD, the trend from PA to PD shows spatial variability that is, consistent with regional trends in anthropogenic emissions of SO 2 and NO x .

Model-Observation Comparisons
The simulated trends in Cl y in the calculated 5-day back trajectory region (TRJ, green dashed region in Figure 3) are qualitatively consistent with and within the ranges of the observed trends in Greenland ice-core Cl exc (Figure 4a). On average, the modeled Cl y burden in TRJ increased by 170% from PI to PA, and decreased by 6% from PA to PD. From PI to PA, ice-core Cl exc showed increases ranging from 105% to 631%, with an average increase of 335%. Although modeled average trends from PI to PA lie below the 25th percentile of observations, the modeled increase in Cl y in continental outflow regions of NA (276%) and WE (203%) lie within the interquartile range (IQR) of the observations, suggesting that trends in chlorine deposition in ZHAI ET AL.
10.1029/2021GL093808 5 of 12 d inland Greenland may be more influenced by these source regions. From PI to PD, the simulated average TRJ Cl y increased 153%, similar to the median increase (163%) in the observations, and was within the IQR range of the ice-core trends. The average modeled trend (+153%) fell on the lower end of the observed range, but again the simulated trends in NA and WE continental outflow regions (223% and 185%, respectively) showed a more robust comparison with the average change in ice-core Cl exc (253%). From PA to PD, most inland Greenland ice cores showed a decreasing trend in Cl exc , with an average decrease of 20%, and a median decrease of 41%. The magnitude of the average modeled Cl y trends from PA to PD in the TRJ region (−6%) was smaller than the average of ice-core observations, but the modeled changes in the NA and WE continental source regions (−13%) were more similar to the observations. The range of PA to PD changes in the model (−46% to 19%) fell within the range of the ice-core observations (−114% to 102%). The modeled PI-to-PA (210%) and PA-to-PD (−25%) change in Cl y at the location of the Col du Dome ice core in the French Alps also was qualitatively consistent with ice-core Cl exc changes (383% from PI to PA, −71% from PA to PD) from Legrand et al. (2002). Model underestimation of the trends may be due in part to uncertainties in anthropogenic HCl emissions (Methods).

Anthropogenic Impacts on Reactive Gaseous Chlorine
Simulated trends in Cl y reflect trends in HCl, since >94% of the burden and 99% of deposition is of the form HCl ( Figure S5). The dominant source of HCl in all three-time periods is acid displacement of SSA Cl − , contributing 73%, 47%, and 61% to the total source in PI, PA, and PD, respectively (Figure 4b). Chemical reactions that convert Cl* (= Cl y − HCl) to HCl is the second largest source (20%, 27%, and 26% in PI, PA, and PD, respectively). In PA, the Cl* source is closely followed by direct anthropogenic HCl emissions, which contributes 21% of the total source. Other sources are minor (<10%). The increase in HCl from PI to PA in the TRJ region (mean of 238%) is driven by increases in direct anthropogenic emissions of HCl (35%), acid displacement (29%), and heterogeneous reactions involving Cl* (29%). The 12% decrease in HCl from PA to PD is driven by decreases in the direct anthropogenic source of HCl (71%) and in conversion of Cl* to HCl (16%), and is partly compensated by a continued increase (15%) in acid displacement.
ZHAI ET AL.
10.1029/2021GL093808 6 of 12 HCl acid displacement is controlled by thermodynamic equilibrium between gas phase (HCl) and aerosol phase ( SSA Cl − ). Lower aerosol pH and aerosol water content (AWC) both favor acid displacement of HCl , but the relationship is nonlinear. At higher pH in PI, the equilibrium is more sensitive to pH than to AWC. At lower pH in PA and PD, AWC becomes more important . From PI to PA, the mean aerosol pH in TRJ decreased 1.7 pH units, resulting in a large increase in HCl displacement despite the increase in AWC (223%) (Figure 4c). From PA to PD, continued increase in HCl displacement is driven by lower AWC in the PD relative to the PA. Although accumulation-mode aerosol pH increases slightly from PA to PD (0.3 pH units), the equilibrium is less sensitive to aerosol pH at the lower pH values during the PA and PD   (Figure 4c).
The HCl source from Cl* chemistry originates from reactions between Cl• with hydrocarbons and the incloud reaction between dissolved SO 2 and HOCl ( Figure S5). Trends in the Cl* source of HCl reflect trends in Cl* abundance. The Cl* burden increased by 252% from PI to PA, and decreased by −9% from PA to PD in TRJ (Figure 4d). Enhanced formation of ClNO 2 (395-fold) from heterogeneous reaction of N 2 O 5 with particulate chloride, driven by elevated NO x emissions ( Figure S6), dominates the increase in Cl* from PI to PA. The decrease in Cl* from PA to PD is caused by the decrease of ClNO 2 (−14%) and Cl 2 (−32%) from PA to PD due to less N 2 O 5 in continental outflow regions ( Figure S7) driven by a decrease in NO x emissions ( Figure S6), consistent with satellite observations (Kim et al., 2006;Konovalov et al., 2010).
Changes in ClNO 2 production from PI to PA and PA to PD drive changes in total Cl* abundance and in all individual Cl* species except Cl• ( Figure S5). Opposite to the Cl* trends, Cl• abundance decreased 27% from PI to PA and increased 20% from PA to PD in the TRJ region. These changes are driven by their reactions with alkanes producing HCl ( Figure S5 and Table S4). Enhanced emissions of alkanes from transportation ZHAI ET AL.
10.1029/2021GL093808 7 of 12 and energy extraction (Hoesly et al., 2018) and increasing methane from PI to PA (Murray, 2016) increased conversion rate of Cl• to HCl, driving the Cl• decrease. From PA to PD, although methane levels continue to increase, anthropogenic emissions of alkanes in NA and WE decrease (Hoesly et al., 2018), resulting in an increase in Cl• from PA to PD. Changes in sink reactions of Cl• are driving the conversion of Cl* to HCl, which shows an increase from PI to PA and a decrease from PA to PD (Figure 4d).

Conclusions and Implications
This study investigates total and nonsea-salt chlorine (Cl exc ) trends since preindustrial time using six Greenland ice cores and examines the contribution of anthropogenic emissions to these trends using the GEOS-Chem model. Observed trends in inland Greenland ice-core Cl exc are captured by historical model simulations that isolate the impact of changes in anthropogenic emissions while holding meteorology constant. Model results indicate that from PI to PA, the increases in acid displacement of HCl from SSA, direct anthropogenic HCl emissions and enhanced Cl* production were responsible for the increasing trend in Cl y . From PA to PD, acid displacement continued to increase, but was overcompensated by reduced direct anthropogenic HCl emissions and chemical conversion of Cl* to HCl. Although direct anthropogenic emissions of HCl represent <21% of the total HCl source, it is required to explain the decreasing trends in Cl exc observed in ice cores since PA.
Cycling of Cl y species can destroy O 3 directly through catalytic cycles, and indirectly through reducing NO x abundance (X. Wang et al., 2019). Consequently, increases in Cl y lead to decreases in OH due to reduction in ozone. The implications of Cl y for ozone, OH and NO x have been demonstrated previously (X. Wang et al., 2019). This study shows that anthropogenic emissions of HCl, SO 2 and NO x have had significant impacts on tropospheric Cl y abundance (up to +170%), which should be considered in the estimation of anthropogenic impacts on changes in tropospheric oxidation capacity.