Significant Decrease in Wet Deposition of Anthropogenic Chloride Across the Eastern United States, 1998–2018

Using deposition observations from precipitation samples collected by the National Atmospheric Deposition Program at 125 sites across the United States, we show that the mean wet deposition flux of non‐sea‐salt chloride (NSS Cl−) has decreased by 83% throughout the eastern United States between 1998 and 2018. We find that 30% of the sites switch from having excess Cl−  to being depleted in Cl−. We attribute the observed decreases in NSS Cl−  deposition to a 95% decrease in U.S. anthropogenic HCl emissions since 1998. We propose that industry emission controls that remove HCl as a cobenefit of NOx and SO2 have caused significant decreases in NSS Cl−  deposition throughout the eastern United States, in addition to shifts from coal to natural gas and to coal with lower Cl−  content. Our analysis implies that the lower tropospheric reactive inorganic chlorine burden was larger over the United States in the past than it is today.


Introduction
The multiphase photochemical cycling of chlorine (Cl) has a widespread influence on the chemical composition of the troposphere, affecting climate, the oxidant budget, acid deposition to the biosphere, and the chemistry of primary and secondary pollutants such as nitrogen oxides (NO x ≡ NO + NO 2 ), nitrate aerosols, and ozone (O 3 ) (Evans et al., 2011;Finlayson-Pitts, 2003;Massucci et al., 1999;Osthoff et al., 2008;Pszenny et al., 1993;Saiz-Lopez & von Glasow, 2012;Sherwen et al., 2016;Thornton et al., 2010;Young et al., 2012). The broad influence of chlorine chemistry stems from the high reactivity of Cl atoms with many gas phase species, including methane (Atkinson, 1997), many volatile organic compounds (VOCs) (Simpson et al., 2015), dimethyl sulfide (Chen et al., 2017;Hoffmann et al., 2016), and mercury . Cl atoms can even be the dominant early-morning oxidative agent in the winter, serving to increase OH production by enhancing formaldehyde through Cl atom-initiated VOC oxidation and enhancing O 3 production from NO x recycling (Haskins et al., 2019).
The main source of tropospheric gas phase inorganic chlorine (Cl y = HCl + ClNO 2 + ClNO 3 + HOCl + 2Cl 2 + Cl) is mobilization of Cl − from sea-salt aerosol (SSA) (Graedel & Keene, 1995;Keene et al., 1999). Wang et al. (2019) estimate that 3.6% of the global flux of SSA is mobilized as Cl y through acid displacement and heterogeneous reactions, resulting in a source of 64 Tg Cl year −1 . They find that nearly 80% of the mobilized SSA Cl y is released as hydrochloric acid (HCl) (51 Tg Cl year −1 ). Over continents, HCl can be directly emitted into the atmosphere through coal combustion, industrial processes, waste incineration, road salt application, fugitive dust, and biomass burning (Andreae, 2019;Keene et al., 1999;Kolesar et al., 2018;Lobert et al., 1999;McCulloch et al., 1999;Sarwar et al., 2012;World Meteorological Organization [WMO], 2014). Although these non-sea-salt (NSS) emissions of Cl y are much smaller in magnitude (together <6.6 Tg Cl year −1 , Keene et al., 1999) relative to the source from SSA, they are potentially important as inland sources of HCl and thus reactive Cl y (Haskins et al., 2018;Thornton et al., 2010).
Since the 1990 U.S. Clean Air Act Amendments, anthropogenic emissions from power generation, industry, transportation, and waste incineration have decreased significantly, thereby affecting the direct anthropogenic sources of Cl y . Additionally, decreases in anthropogenic SO 2 and NO x emissions have led to reduced concentrations of HNO 3 , H 2 SO 4 , and N 2 O 5 , which may have weakened the mobilization of Cl − from SSA via acid displacement and heterogeneous reactions. Between 1998 and 2018 in the United States, anthropogenic HCl emissions have decreased by 95%, SO 2 emissions have decreased by more than 88%, and NO x emissions have decreased by more than 59% (U.S. EPA, 2018a). These anthropogenic emission changes have led to significant and well-documented decreases in SO 4 2− deposition to the environment, increases in rainwater pH, and the recovery of previously devastated ecosystems (Butler et al., 2001;Giang & Selin, 2016;Fedkin et al., 2019;Lajtha & Jones, 2013;Lehmann et al., 2007;Paulot et al., 2017).
These emission reductions stem from the widespread installation of emission control measures. Other contributing factors include the shift from coal to natural gas in electricity generation after 2005(de Gouw et al., 2014U.S. EPA, 2020) and the increasing use of coal sourced from west of the Mississippi River, which has lower chlorine and sulfur content than that sourced in the eastern United States (U.S. Energy Information Administration, 2020; U.S. EPA, 2014). Emission control technologies, such as flue-gas desulphurization (FGD) and selective catalytic reactors (SCRs), were designed to remove SO 2 , NO x , and mercury as part of the 1990 Acid Rain Program, the 2005 Clean Air Interstate Rule (CAIR), the 2011 Cross State Air Pollution Rule, and the 2012 Mercury and Air Toxics Standards (MATS) rule. As a cobenefit of NO x , SO 2 , and mercury targeted reductions, emission control technologies have been shown to be effective at removing gas phase HCl, which can be emitted from burning chloride-containing items (e.g., coal, waste, and biomass) (Lee et al., 2004;McCulloch et al., 1999). FGD uses an alkaline sorbent to remove SO 2 , which also reacts readily with other acid gases such as HCl, and is used to control SO 2 and HCl emissions at municipal and industrial facilities with a removal efficiency estimated to be greater than 95% . Emission controls through SCR, originally designed to remove NO x before emission, can be coupled with FGD in order to convert insoluble elemental mercury (Hg 0 ) to oxidized mercury (Hg 2+ ). This conversion is achieved via the heterogeneous reaction of Hg 0 with HCl on various metal oxide surfaces to form HgCl 2 , which is then captured, thereby removing not only mercury but also HCl within flue gases (Lee et al., 2004).
In this work, we examine whether these anthropogenic HCl emission decreases are reflected in observations of wet deposition fluxes of total chlorine (HCl and particulate Cl − ) in the United States over the last two decades. We assess how changes in direct anthropogenic HCl emissions as well as in the secondary mobilization of Cl y from SSA (from reduced NO x and SO 2 concentrations) have affected observed trends in chlorine wet deposition.

Methodology
The National Atmospheric Deposition Program (NADP)/National Trends Network (NTN) has measured the concentrations of acids, nutrients, and base cations in weekly precipitation samples since 1978 to characterize the chemical composition of rain in the United States and its temporal and spatial trends (Lamb & Bowersox, 2000; NADP, 2020). As of 1 January 2020 there are 380 NADP sites. In this work, we use NADP reported weekly measurements of total Cl − ( T Cl − ), sodium (Na + ), total sulfate ( T SO 4 2− ), and precipitation (NADP, 2020). In order to separate natural and anthropogenic contributions to T Cl − deposition, we calculate the NSS contribution using NADP observations of Na + . The mass ratio of Cl − to Na + ratio in fresh sea spray has a value of 1.80, corresponding to bulk seawater composition (Keene et al., 1986). NSS Cl − is thus calculated as follows: Similarly, NSS SO 4 2− can be calculated using the SO 4 2− to Na + mass ratio in seawater (0.507; Keene et al., 1986): Using observed T SO 4 2− , T Cl − , Na + , and precipitation data from the NADP/NTN, we calculate weekly NSS Cl − and NSS SO 4 2− wet deposition fluxes for 1998-2018. Only samples flagged as valid by the NADP were used in our analysis. Valid samples are defined as those collected following standard procedures: not flagged as being contaminated, those not exposed to excess dry deposition, those with a sampling interval less than 8 days, and those samples with a reported rain gauge depth or sample volume (NADP, 2016). We require data to meet three completeness criteria as defined by the NADP for each week throughout the year, throughout the period of analysis (NADP, 2016): (1) that there are valid samples for at least 75% of the summary period, (2) that there must be precipitation amounts for at least 90% of the summary period, either from the rain gage or from the sample volume, and (3) that there must be valid samples for at least 75% of the total precipitation amount reported for the month. The first two criteria ensure that measurements on valid wet deposition samples and of precipitation amounts were reported for a minimum acceptable fraction of the summary period (i.e., functioning equipment), while the third ensures that there are valid precipitation chemistry data to represent a majority of the precipitation that was estimated to have occurred.
We performed a seasonal Kendall test (SKT) using Sen's method on the weekly data at individual sites to detect trends underlying seasonal deposition patterns from 1998-2018 (Burkey, 2020;Marchetto et al., 2013). The SKT trends were applied to NADP sites that fulfilled the three criteria described above, and in addition, we excluded sites that had more than a 3-month gap in meeting the completeness criteria. For the 1998-2018 period considered here, 125 sites met these criteria, and we only show analysis and summary statistics for these 125 sites throughout this work. Seasons in the SKT test are based on the annual year, so seasonality trends are assessed from January-December. The null hypotheses were that the trend was 0 (Kendall's tau-b statistic was 0), and the statistical significance level was set at 90%. The magnitude of the trend was determined by taking the Sen's slope of the weekly deposition data.
We are particularly interested in analyzing trends at sites where NSS Cl − is a significant portion of the T Cl − wet deposition flux. However, SS can often be a very large component of the T Cl − wet deposition flux. In order to distinguish sites of interest from those where sea-salt (SS) trends dominate T Cl − wet deposition, we will refer to sites where the NSS Cl − deposition flux exceeds 5% of T Cl − wet deposition flux as sites with nonnegligible NSS Cl − deposition flux. These sites tend to be located near industrial centers where anthropogenic Cl − sources might be expected to contribute to T Cl − deposition or over inland regions where the SS Cl − deposition is so small that even a small amount of deposited NSS Cl − would be relevant. Sites with less than 5% contributions from NSS Cl − to T Cl − are referred to as sites with negligible NSS Cl − and tend to be located in coastal regions where direct SS deposition is large or where there is little anthropogenic Cl y contribution to deposition.
The deposition flux of NSS Cl − can be positive or negative because the concentration of NSS Cl − is calculated as a difference (Equation 1). Positive values of NSS Cl − deposition flux show that excess Cl − was measured relative to what would be expected from SSA, thus indicating the influence of anthropogenic sources of Cl y . Negative values of NSS Cl − deposition flux show that the measured rainwater sample was depleted in 10.1029/2020GL090195

Geophysical Research Letters
Cl − relative to what would be expected in SS aerosols. Therefore, negative deposition fluxes of NSS Cl − suggest that SS Cl − was displaced into the gas phase as Cl y either through acid displacement by HNO 3 or H 2 SO 4 or through heterogeneous reactions with N 2 O 5 .
Inherent in our calculation of NSS Cl − (Equation 1) is the assumption that all measured Na + is from SS. Other potential sources of Na + include road salt, soil, and refuse incineration (Ooki et al., 2002). By using correlations between Na + and Mg 2+ (Text S1 in the supporting information), we determined that only 11 NADP sites showed evidence of a persistent NSS Na + source (for less than 25% of the 1998-2018 time period). At these 11 sites, the calculated NSS Cl − concentration could thus be an underestimate during some weeks. Results excluding these sites were consistent with those including them (see supporting information Text S1); therefore, we have chosen to keep these sites in our analysis.

Geophysical Research Letters
NSS Cl − deposition fluxes are those where Cl − from SSA was displaced into the gas phase as Cl y via heterogeneous chemistry involving HNO 3 , H 2 SO 4 , or N 2 O 5 . The sites with the largest depletion in Cl − relative to Na + are located in the south and southeast, particularly in Missouri, Arkansas, Mississippi, and Texas. By 2016-2018, these sites are generally less depleted in Cl − than in 1998-2000, likely reflecting decreases in NO x and SO 2 reducing heterogeneous mobilization of SS Cl − , which will be discussed further below.
Additionally, in this two-decade span, the number of sites with nonnegligible NSS Cl − (NSS Cl − / T Cl − > 0.05) decreased from 91 to 68 indicating that fewer sites now experience either (1) a significant contribution from anthropogenic sources of Cl y or (2) significant heterogeneous chemistry mobilizing natural SS Cl − . This reflects that the number of sites experiencing a significant anthropogenic influence on the observed Cl − wet deposition flux has significantly decreased over time.  (Fedkin et al., 2019;Lehmann et al., 2007).
The largest statistically significant trends in NSS Cl − deposition are also seen in the midwestern and northeastern United States (Figure 2b), again mirroring spatial patterns of NSS SO 4 2− . The mean change at all sites with statistically significant trends in NSS Cl − is −6.6 g Cl ha −1 year −1 , with decreases at 85% of these sites. Considering only sites with statistically significant decreases and nonnegligible NSS Cl − , the mean change is −10.8 g Cl ha −1 year −1 , with these sites largely located in Maryland, New Jersey, New York, North Carolina, Ohio, and Pennsylvania.
In contrast to the eastern United States, several sites in the southeastern United States had heavily depleted NSS Cl − wet deposition fluxes (negative NSS Cl − ) in the late 1990s (Figure 1b). Thirteen of these sites displayed statistically significant increases (+3.8 g Cl ha −1 year −1 ) in NSS Cl − wet deposition flux. Of these 13 sites, only 6 had nonnegligible NSS Cl − between 1998 and 2000 (blue circles in Figure 2b). For the remaining seven sites in the Pacific Northwest and coastal Florida (blue diamonds), the amount of NSS Cl − deposited is negligible relative to the SS signal, such that the increases are occurring for a very small amount of deposited NSS Cl − relative to T Cl − deposited. The average trend in NSS Cl − at the six sites with nonnegligible NSS Cl − wet deposition flux was +7.8 g Cl ha −1 year −1 , increasing from a mean deposition of −174 g Cl ha −1 in 1998-2000 to −104 g Cl ha −1 in 2016-2018 (Figures 1b and 1c). Despite these increases in NSS Cl − wet deposition, by 2018 these sites in the southeastern United States were still depleted in NSS Cl − , just to a lesser degree than in 1998. We attribute these positive trends in NSS Cl − deposition in the southeast to stricter emission controls having decreased both SO 2 and NO x , thus weakening displacement of Cl − from SSA via anthropogenically driven secondary chemistry.  Here, we only consider deposition data from NADP sites with nonnegligible NSS Cl − for the 1998-2000 period and statistically significant negative Sen's slopes. The resulting percentage change in deposition compares well to the trends calculated from the Sen's slopes (blue "X" in Figure 3; see supporting information Text S2). In total, wet deposition of NSS Cl − has decreased by 83% on average since 1998 across the eastern United States (Regions 1-5) at sites with significant trends.  Figure 3 illustrates that the observed NSS Cl − wet deposition flux declines at a rate similar to the 3-year moving mean decrease in HCl emissions throughout the comparison period. For these five EPA regions, anthropogenic HCl emissions decrease by 95%-100%. Between 1998 and 2018, NSS Cl − wet deposition decreases by 72%-95%. The time series of HCl emissions and NSS Cl − wet deposition decrease are well correlated temporally, with r 2 ranging from 0.76 to 0.96 depending on the region. Temporal correlations between HCl emissions and NSS Cl − deposition are largest in Region 3 (r 2 = 0.96, n = 8). Over this region, anthropogenic HCl emissions were 61 Gg year −1 in 1998, or 22% of the national anthropogenic HCl emissions that year, and NSS Cl − wet deposition flux decreased from +0.59 g Cl ha −1 (1998)(1999)(2000) to −0.14 g Cl ha −1 (2016-2018) at sites with nonnegligible NSS Cl − . This is the largest absolute value change in NSS Cl − wet deposition flux seen across all EPA regions.

Geophysical Research Letters
The impact of upwind HCl emissions can be seen at NADP sites in New York and New Jersey in Region 2, which has minimal direct anthropogenic HCl emissions from industry but experiences some of the largest decreases in the wet deposition flux of NSS Cl − both in absolute value and percentage (Figure 3b). Regions 1 and 5 both show a slight difference between their trends in HCl emissions and NSS Cl − deposition fluxes (Figures 3a and 3e). This difference could be due to the fact that not all regional anthropogenic HCl emissions are necessarily deposited within that region and that transport of upwind HCl emissions can influence the local deposition flux. Overall, the very similar temporal evolution in NSS Cl − and HCl emissions in EPA Region 1-5, together with the spatial correlation between SO 4 2− and NSS Cl − deposition decreases, suggests that decreases in anthropogenic HCl emissions are driving the observed NSS Cl − wet deposition decreases in near source regions.
To determine the extent to which anthropogenic HCl emissions can explain the observed trends in NSS Cl − wet deposition, we estimated the decrease in NSS Cl − wet deposition by multiplying the weekly observed NSS SO 4 2− wet deposition by the mean HCl/SO 2 anthropogenic emission ratio (ER) in each region, considering only electric utility sources and then taking the Sen's slope of the predicted NSS Cl − to estimate the decrease from HCl emission changes only, as described in supporting information Text S2 ( Figures S4 and  S5). The best agreement between observed and predicted change in NSS Cl − deposition flux was found in EPA Region 3, with a mean 16% (+0.935 g Cl ha −1 year −1 ) overestimation from the predictions ( Figure  S5), consistent with our finding that Region 3 has the strongest temporal correlation between decreases in anthropogenic HCl emissions and NSS Cl − wet deposition flux (Figure 3c). Overall, anthropogenic HCl emission decreases were enough to explain observed NSS Cl − wet deposition decreases particularly in large source regions (see supporting information Text S2 for discussion) but were unable to capture increasing trends in the southeast influenced more strongly by Cl − mobilization through acid displacement and heterogeneous reactions.
Together, the 1998-2018 decrease in excess NSS Cl − deposited across the Midwestern and northeastern United Stated and the increase in depleted NSS Cl − deposited in the southeastern United States represent a return to Cl − deposition patterns less impacted by anthropogenic activities. Therefore, regardless of the exact mechanism for the decrease in the absolute value of NSS Cl − wet deposition flux across the United States (reductions in primary emissions of anthropogenic HCl or reductions in emissions of SO 2 and NO x promoting less Cl − mobilization from SS aerosols), we find that decreases in anthropogenic emissions have reduced the anthropogenic impact on the wet deposition of NSS Cl − across the eastern United States. A similar decrease in anthropogenic HCl emissions from coal burning (−95%) in the United Kingdom has been shown to have resulted in a 66% decrease in deposition of NSS Cl − between 1990 and 2007 (Evans et al., 2011).

Conclusions
Between 1998  Observed decreases in NSS Cl − wet deposition were spatially well correlated with SO 4 2− wet deposition decreases across the United States (r 2 = 0.70) and were temporally well correlated with observed HCl emission decreases (r 2 ≥ 0.76). We propose that the 95% decrease in anthropogenic HCl emissions across the United States has driven an observable decrease in NSS Cl − wet deposition in the Midwest and eastern United States. The negative trends in NSS Cl − deposition reflect a reduced anthropogenic impact on Cl − deposition, with more Cl − deposition explained by a SS source alone in 2016-2018 than in 1998-2000 and with sites switching from having excess NSS Cl − deposition in the late 1990s to having either depleted or near-zero NSS Cl − patterns by 2018. A small number of sites in the southeastern United States were depleted in NSS Cl − in 1998 and saw statistically significant increases in NSS Cl − deposition but remained depleted in Cl − by 2018. We attribute this positive trend to a decrease in displacement of NSS Cl − as stricter emission controls decreased both SO 2 and NO x , which in turn decreased the amount of Cl − from SSA that was displaced into the gas phase as Cl y .
Overall, our results suggest that the tropospheric Cl y burden was distinctly larger over the United States in the past than it is today, both from higher direct HCl emissions and larger mobilization of SS Cl − . Global model simulations by Wang et al. (2019) recently showed that displacement of Cl − from SS aerosol provided a reactive Cl y (Cl y -HCl) source of 12 Tg Cl year −1 via heterogeneous reactions with HNO 3 , H 2 SO 4 , and N 2 O 5 in the year 2015, higher than previous estimates of 5.6 Tg Cl year −1 (Hossaini et al., 2016) and 6.1 Tg Cl year −1 . Our results suggest that this number may have been even higher in the past 20 years in the United States, when HNO 3 and H 2 SO 4 production was larger and when anthropogenic HCl concentrations were a more significant inland source of Cl y . The past impacts of Cl y on the oxidant budget, downwind O 3 formation, methane and VOC oxidation, dimethyl sulfide, and mercury concentrations remain to be quantified. Given that observations have shown Cl atoms can be the dominant early morning oxidant in winter in select polluted coastal regions as recently as 2015 (Haskins et al., 2018), our results suggest that the impact of tropospheric halogens may have played an even larger role on air quality and climate prior to large anthropogenic emission reductions.

Data Availability Statement
Deposition data used in this work from the National Atmospheric Deposition Program are available to the public at the website (http://nadp.slh.wisc.edu/data/ntn/ntnAllsites.aspx). HCl emission data used in this work from the EPA's Toxic Release Inventory are available to the public at the website (https://bit.ly/ 3ocV9fg). Facility level SO 2 emission data used in this work from the EPA's National Emissions Inventory are available to the public at the website (https://www.epa.gov/air-emissions-inventories/2017-nationalemissions-inventory-nei-data).