Particulate Nitrate Photolysis as a Possible Driver of Rising Tropospheric Ozone

Tropospheric ozone is an air pollutant and a greenhouse gas whose anthropogenic production is limited principally by the supply of nitrogen oxides (NOx) from combustion. Tropospheric ozone in the northern hemisphere has been rising despite the flattening of NOx emissions in recent decades. Here we propose that this sustained increase could result from the photolysis of nitrate particles (pNO3−) to regenerate NOx. Including pNO3− photolysis in the GEOS‐Chem atmospheric chemistry model improves the consistency with ozone observations. Our simulations show that pNO3− concentrations have increased since the 1960s because of rising ammonia and falling SO2 emissions, augmenting the increase in ozone in the northern extratropics by about 50% to better match the observed ozone trend. pNO3− will likely continue to increase through 2050, which would drive a continued increase in ozone even as NOx emissions decrease. More work is needed to better understand the mechanism and rates of pNO3− photolysis.


Introduction
Tropospheric ozone is a short-lived climate forcer and a highly reactive gas that can damage human cells and tissue and reduce plant productivity.It forms from the oxidation of carbon monoxide (CO), methane, and volatile organic compounds in the presence of nitrogen oxides (NO x ≡ NO + NO 2 ).A smaller amount is transported from the stratosphere.Tropospheric ozone concentrations in the northern midlatitudes have risen by about 50% since the early 20th century because of rising anthropogenic emissions of ozone precursors (Tarasick et al., 2019).This trend has continued into the 21st century (Christiansen et al., 2022;Gaudel et al., 2020;Ziemke et al., 2019), although the rise in global precursor emissions has slowed down (Hoesly et al., 2018).Between the 1990s and 2017, ozone concentrations in the northern hemisphere increased by an average of 2 ppbv per decade in the free troposphere, the layer between ∼2 km and the tropopause (Christiansen et al., 2022;Gaudel et al., 2020).Ozone increases in the free troposphere have a larger effect on climate than increases near the surface (Lacis et al., 1990).Free tropospheric ozone also contributes to ozone pollution at the surface (Colombi et al., 2023;Lin et al., 2017).However, global atmospheric chemistry models cannot account for the observed ozone increase (Christiansen et al., 2022;Gaudel et al., 2020;Skeie et al., 2020;H. Wang et al., 2022), implying that key processes are missing.It is important to identify these processes to improve our ability to make accurate future projections of climate forcing and air pollution from tropospheric ozone.
Intercomparisons of global atmospheric chemistry models show major differences in their computed global production and loss of tropospheric ozone (Hu et al., 2017;Stevenson et al., 2006;Wu et al., 2007;Young et al., 2018).Only a few models include tropospheric halogen chemistry, an important sink for ozone and NO x (Saiz-Lopez et al., 2012;Sherwen et al., 2016;X. Wang et al., 2021).When included in the widely used GEOS-Chem model, halogen chemistry lowers the tropospheric ozone burden by 15% (X.Wang et al., 2021).At the same time, many models including GEOS-Chem underestimate ozone production over the tropical oceans because of an underestimate in NO x concentrations (Guo et al., 2023;Travis et al., 2020).The combination of halogen chemistry and low tropical NO x has led recent versions of GEOS-Chem to underestimate global tropospheric ozone (X.Wang et al., 2021).
The photolysis of nitrate particles (pNO 3 ) has been proposed as a major route for recycling NO x over the oceans (Andersen et al., 2023;Kasibhatla et al., 2018;Reed et al., 2017;Ye et al., 2016).pNO 3 is produced by the gasparticle partitioning of nitric acid (HNO 3 ), the dominant sink of NO x , at low temperature, high humidity, and low aerosol acidity, which is associated with high ammonia, low sulfate, and freshly emitted sea salt aerosols.Photolysis of nitrate ions in aqueous solutions produces nitrous acid (HONO) and NO 2 , which volatilize to the gas phase (Mack & Bolton, 1999): Using aircraft-based HONO observations over the oceans, studies have inferred a pNO 3 photolysis frequency on the order of 10 4 s 1 (Andersen et al., 2023;Ye et al., 2016), which is about two orders of magnitude faster than the photolysis frequency of gas-phase HNO 3 or nitrate in bulk solutions, likely because of enhancement of nitrate at the aerosol surface (Andersen et al., 2023;Ye et al., 2017).These fast rates are supported by laboratory studies on ambient pNO 3 (Bao et al., 2018;Gen et al., 2019;Ye et al., 2017).Shah et al. (2023) showed that including pNO 3 photolysis in GEOS-Chem corrects the NO x underestimate over the oceans and increases the production of tropospheric ozone in the model.However, some field and laboratory studies suggest that the reaction is too slow (<10 5 s 1 ) to be a significant path for recycling NO x (Romer et al., 2018;Shi et al., 2021;Y. Zhu et al., 2022).
Here we show that including the parameterization of pNO 3 photolysis from Shah et al. (2023) improves the ability of the GEOS-Chem model to simulate the observed tropospheric ozone distribution as well as the trends in the northern midlatitudes since the mid-1990s.Further simulations indicate that increasing pNO 3 concentrations has augmented the growth in the tropospheric ozone burden since the middle of the 20th century and could continue to do so until the middle of the 21st century.

Ozone Measurements
We use observations of ozone vertical profiles from the global ozonesonde network (Stauffer et al., 2022;Tarasick et al., 2021), and the In-service Aircraft for a Global Observing System (IAGOS) program (Petzold et al., 2015).We use ozonesonde data from the following archives: Tropospheric Ozone Assessment Report-II Harmonization and Evaluation of Ground-based Instruments for Free Tropospheric Ozone Measurements (HEGIFTOM v2) Working Group (2023), NOAA Earth System Research Laboratory-Global Monitoring Division (Sterling et al., 2018), the Southern Hemisphere ADditional OZonesondes (SHADOZ; Thompson et al., 2017;Witte et al., 2017Witte et al., , 2018)), and the World Ozone and Ultraviolet Data Center (WOUDC 2015).Ozonesondes measure ozone using an electrochemical cell with accuracy of better than ±10% (Tarasick et al., 2019;Thompson et al., 2019).The ozonesonde launch frequency varies among sites from once a month to thrice a week.We exclude stations with less than two profiles in a month and less than 8 months of observations, and aggregate the observations at each site to monthly means.The ozonesonde stations used in this work are listed in Supporting Information S1.
The IAGOS program provides ozone observations using commercial passenger aircraft (Petzold et al., 2015).In this work, we only use the vertical profile observations from the take-off and landing portions of the flights, excluding observations from the cruise portion.The IAGOS measurements are made using a dual-beam ultraviolet absorption photometer, with an accuracy of ±2 ppbv (Blot et al., 2021;Nédélec et al., 2015).We aggregate the profiles into eight areas to account for the irregular sampling frequency at any one airport, excluding areas with less than two profiles in a month and less than 8 months of observations (Supporting Information S1).
We further aggregate the observations in each area to monthly means.Comparisons of the two data sets show the IAGOS measurements to be low relative to the ozonesonde measurements by 5%-8% (Staufer et al., 2014;Tarasick et al., 2019).

GEOS-Chem
We simulate tropospheric ozone using the GEOS-Chem atmospheric chemistry model (version 14.2.0; https://doi.org/10.5281/zenodo.8411433)driven by meteorology from the NASA Global Modeling and Assimilation Office's (GMAO) Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2) reanalysis (Gelaro et al., 2017).GEOS-Chem includes a detailed representation of tropospheric and stratospheric gas and aerosol chemistry (Eastham et al., 2014;Gao et al., 2022;Sherwen et al., 2016;Travis et al., 2016), with recent updates to the tropospheric halogen chemistry (X.Wang et al., 2021).Sea salt aerosol debromination was disabled in recent applications of the model, despite evidence that it takes place (Sander et al., 2003), because it caused excessive ozone depletion in the marine boundary layer (Shah et al., 2023).This was partly due to an error in how the titration of sea salt aerosol alkalinity was represented in the model.This error is now fixed, and we include sea salt aerosol debromination in our simulation.We also reduce the loss of ozone by iodine radicals by including the uptake of HOI, IONO, and IONO 2 on alkaline sea salt aerosols as I using uptake coefficients from Sherwen et al. (2016), and by limiting the hydrolysis of IONO 2 to acidic aerosols assuming an uptake coefficient equal to that of BrONO 2 (Deiber et al., 2004), following X.Wang et al. (2021).Ozone dry deposition in the model was recently updated to explicitly consider the reaction of ozone with sea surface I (Pound et al., 2020), and to use a higher surface resistance over snow and ice (Barten et al., 2021).
pNO 3 photolysis in GEOS-Chem follows the original implementation of Kasibhatla et al. (2018) with modifications from Shah et al. (2023), and includes pNO 3 on fine and coarse mode particles.The photolysis frequency of pNO 3 is calculated by scaling the photolysis frequency of HNO 3 by an enhancement factor (EF), taken to be 100 for coarse mode pNO 3 and between 10 and 100 for fine mode pNO 3 depending on the fraction of pNO 3 in sea salt aerosols (Shah et al., 2023): Here, [pNO 3 ] and [SSA] are the molar concentrations in air of fine mode pNO 3 and sea salt aerosol.The molar concentration of sea salt is taken as [SSA] = 2.39 [Na + ] based on the fraction of Na + in seawater (Millero et al., 2008), and where Na + is the chemically inert sea salt aerosol species simulated by GEOS-Chem.The HONO:NO 2 yield is taken to be 2:1 (Kasibhatla et al., 2018).As described in Shah et al. (2023), Equation 1corrects previous GEOS-Chem low bias in simulating NO x concentrations over the remote oceans.It is consistent with the inverse relationship between EF and [pNO 3 ] derived by Andersen et al. (2023) based on a Langmuir isotherm model for the partitioning of pNO 3 to the aerosol surface.It also accounts for the enhancement of pNO 3 at the surface in the presence of halides in sea salt aerosols (Richards-Henderson et al., 2013;Wingen et al., 2008).EF values from Equation 1 are lower than those derived from HONO measurements over remote areas (Andersen et al., 2023;Ye et al., 2016), but consistent with those derived over polluted areas (Romer et al., 2018).The thermodynamic partitioning of HNO 3 to fine mode pNO 3 is computed with ISORROPIA II (Fountoukis & Nenes, 2007;Pye et al., 2009).Coarse mode pNO 3 forms by the uptake of HNO 3 on sea salt aerosols (X.Wang et al., 2019).Uptake of HNO 3 on dust is not included here, though it is an option in GEOS-Chem (Fairlie et al., 2010).Photolysis frequencies in the model are calculated using Fast-JX (Bian & Prather, 2002;Eastham et al., 2014).
Our main simulation is conducted for 2018 (with a spin-up period of 6 months) at a 4°latitude ⨯5°longitude resolution.For comparison with the ozonesonde and IAGOS measurements, we sample the model at the measurement location and within a 3-hr window of the measurement time.To evaluate the long-term changes in tropospheric ozone, we conduct additional simulations for the years 1960, 1980, 1995, and 2050 using year-specific anthropogenic emissions and methane concentrations but constant ( 2018) meteorology and natural emissions.We refer to the simulation with pNO 3 photolysis as the base simulation, and compare it to a parallel simulation without pNO 3 photolysis.Details about emission inventories and the tropospheric ozone budget in GEOS-Chem are in the Supporting Information S1.

Tropospheric Ozone Distribution in 2018
Figure 1 shows the annual mean ozone distribution in the middle troposphere (800-400 hPa or ∼2-7 km altitude) from the ozonesonde and IAGOS observations and GEOS-Chem simulations in 2018.Figure 2 shows the observed and simulated ozone vertical profiles between the surface and 200 hPa, and the seasonal variation of mid-tropospheric ozone concentrations over six regions.The global mean ozone concentration in the middle troposphere in the ensemble of ozonesonde and IAGOS observations is 47.4 ppbv.This is reproduced by GEOS-Chem with a mean bias of 2.8 ppbv, and within the observational uncertainty of about 8% implied by the bias between the ozonesonde and IAGOS measurements.The simulated ozone vertical profiles and seasonal variations also align with the observations in most regions (Figure 2).The model overestimates ozone observations in the tropics, which could be from an overestimate in lightning NO x emissions.Lightning NO x emissions in the model are calculated using a satellite-based climatology of lightning flash rates and NO x yields of 260 and 500 mol per flash in the tropics and the northern midlatitudes, respectively, amounting to a global lightning NO x emission of about 6 Tg N a 1 (Murray et al., 2012).However, there is large uncertainty in this source with estimates ranging from 2 to 8 Tg N a 1 (Schumann & Huntrieser, 2007).
The ozone concentrations from a simulation without pNO 3 photolysis are on average 2-6 ppbv lower than those in the base simulation, but they are still largely consistent with the observations within their uncertainty (Figure 2).An exception is the northern extratropics in spring, where the effect of pNO 3 photolysis is strongest, and excluding it introduces a negative bias of up to 10 ppbv compared to the observations over the Arctic, North America and Europe, and East Asia (bottom panel of Figure 2).pNO 3 concentrations are highest in spring because of efficient lifting to the free troposphere combined with relatively low temperatures and seasonally rising emissions of ammonia from agricultural sources.Actinic flux is also relatively high in spring to enable pNO 3 photolysis as well as ozone production.A low ozone bias in spring in the absence of pNO 3 photolysis had been previously reported in recent versions of GEOS-Chem and attributed to halogen chemistry (Christiansen et   pNO 3 photolysis in the model, as also found by Colombi et al. (2023) and Yang et al. (2023) in comparison with ozonesonde and aircraft observations over South Korea in May-June 2016.

Multi-Decadal Trends in Tropospheric Ozone
Ozone concentrations in the free troposphere have increased by 1-6 ppbv decade 1 in the northern hemisphere since the mid-1990s (Christiansen et al., 2022;Gaudel et al., 2020;H. Wang et al., 2022), but previous versions of GEOS-Chem could not capture that trend (Christiansen et al., 2022;H. Wang et al., 2022).pNO 3 concentrations in the northern midlatitudes have most likely increased over this period because of decreasing aerosol acidity due to decreasing sulfate and increasing ammonia (Bauer et al., 2020;Paulot et al., 2018).Here, we examine the effect of rising pNO 3 concentrations on ozone trends by conducting simulations for the year 1995.We switch only the anthropogenic emissions and methane levels to 1995 values, keeping the meteorology, and natural emissions constant.There is no long-term trend in global lightning NO x emissions (Kaplan & Lau, 2022).H. Wang et al. (2022) found that climatic factors contribute little to the global ozone trend between 1995 and 2018.
Figure 3 compares the 1995-2018 change in mid-tropospheric ozone concentrations in the base GEOS-Chem simulation and the simulation without pNO 3 photolysis to the observed trends in the ozonesonde and IAGOS data calculated by Christiansen et al. (2022) and H. Wang et al. (2022).Both simulations and the observations show the fastest increase in tropospheric ozone over Asia, reflecting the increase in ozone precursor emissions in the region over this period (Hoesly et al., 2018;Kurokawa & Ohara, 2020).However, the ozone increase in the base simulation is larger than that in the simulation without pNO 3 photolysis, particularly in the northern midlatitudes, and more consistent with the observed trends.Over East Asia, the ozonesonde and IAGOS observations show mid-tropospheric ozone trends of 2.1-3.3 ppbv decade 1 .In comparison, the base GEOS-Chem simulation shows an increase of 2.3-2.7 ppbv decade 1 , but the simulation without pNO 3 photolysis shows an  S1 and S2 in Supporting Information S1).The shaded areas denote ±1 standard deviation of the annual mean ozone profiles (top panel) and the monthly mid-tropospheric ozone concentrations (bottom panel) at the measurement sites in each region.
increase of 1.4-1.9ppbv decade 1 , suggesting that about a third of the increase in ozone over East Asia since the mid-1990s is driven by increasing pNO 3 .The base simulation also reproduces the trend in the IAGOS data over North America better than the simulation without pNO 3 photolysis, but it overestimates the IAGOS trend over Europe.The trends in the ozonesonde data over North America and Europe vary substantially, which suggests that they are strongly affected by meteorological variability (Christiansen et al., 2022).
pNO 3 concentrations in the northern midlatitudes in the model double between 1995 and 2018, mainly because of falling SO 2 emissions and rising ammonia emissions.According to the Community Emissions Data System (CEDS) inventory (Hoesly et al., 2018) used in our simulations, global anthropogenic SO 2 emissions over this period fell by 55% due to better emission controls in power plants, while anthropogenic ammonia emissions rose by 20% due to increased agricultural activity.pNO 3 formation in the northern midlatitude free troposphere is limited by high aerosol acidity (Nault et al., 2021).Aerosol acidity drops when there is less sulfate and more ammonia, allowing more HNO 3 to condense as pNO 3 (Guo et al., 2016).The increase in pNO 3 in the model happens mostly in the free troposphere, where NO x concentrations are sensitive to pNO 3 photolysis (Dang et al., 2023).At the surface, pNO 3 concentrations decrease over US and Europe, consistent with the observed trends (Ciarelli et al., 2019;Hand et al., 2020), but increase over Asia.Free tropospheric pNO 3 also increases due to the 50% increase in global aircraft NO x emissions between 1995 and 2018 (Simone et al., 2013).
Emissions of SO 2 , ammonia, and NO x have changed substantially since the middle of the 20 th century and further changes are expected in the future (Gidden et al., 2019;Hoesly et al., 2018).We examined the effect of these changes on long-term trends of pNO 3 and ozone by conducting additional simulations for the years 1960, 1980, and 2050.For 2050, we use the SSP2-4.5 emissions scenario from Phase 6 of the Coupled Model Intercomparison Project (CMIP6).SSP2-4.5 is a middle-of-the-road scenario in which future emissions largely follow the current trends (Fricko et al., 2017).Again, we only consider changes in anthropogenic emissions and methane concentrations.
Figure 4 shows the 1960-2050 change in the tropospheric ozone burden in the base GEOS-Chem simulation and the simulation without pNO 3 photolysis.It also shows the simulated tropospheric burdens of sulfate, HNO 3 , and pNO 3 , and global emissions of SO 2 , NO x , and ammonia for 1960-2050.The ozone burden increases between 1960 and 2050 in both simulations largely because of increase in methane, NO x and CO emissions.However, the burden in the base simulation increases by 86 Tg in this period, compared to an increase of 67 Tg in the simulation without pNO 3 photolysis, because of the increasing burden of pNO 3 .The burden of pNO 3 increases much faster than the burden of HNO 3 , reflecting the trends in SO 2 and ammonia emissions.Between 2018 and 2050, the base simulation projects an increase in the ozone burden by 12 Tg (under the SSP2-4.5 scenario), double the increase in the simulation without pNO 3 photolysis, because of a projected increase in pNO 3 , despite a decrease in the HNO 3 burden and NO x emissions.This result is not specific to the SSP2-4.5 scenario, as most SSP scenarios project a decrease in SO 2 and increase in ammonia emissions between now and 2050, making it likely that increasing pNO 3 will continue to amplify the trend in tropospheric ozone.

Conclusions
Photolysis of nitrate particles (pNO 3 ) is generally not included in global atmospheric chemistry models but we show here that it improves the tropospheric ozone simulation in the GEOS-Chem model and can help account for the observed multi-decadal trends in ozone.pNO 3 photolysis increases simulated free tropospheric ozone concentrations in the northern extratropics in the spring by up to 10 ppbv and counteracts the springtime ozone loss by halogen radicals that would otherwise cause a low bias in ozone in the model.We find that increasing pNO 3 concentrations due to falling SO 2 and rising ammonia emissions globally since the 1980s have amplified the increase in the tropospheric ozone burden and explain over a third of the observed growth in free tropospheric ozone over the northern midlatitudes between 1995 and 2018.Increasing pNO 3 will likely continue to raise the tropospheric ozone burden through 2050.
The significant effect of pNO 3 photolysis on tropospheric ozone calls for further work to characterize its mechanism and rates.Rate estimates for this reaction are highly variable, with values ranging from 1 to 1,000 times the HNO 3 photolysis frequency, but it is not fully clear what drives this variability (Andersen et al., 2023).pNO 3 photolysis also affects the HONO/NO 2 and NO x /HNO 3 ratios (Romer et al., 2018;Ye et al., 2023) and the diurnal cycles of HONO and NO 2 in the remote troposphere (Kasibhatla et al., 2018;Reed et al., 2017), and these metrics can provide constraints on pNO 3 photolysis rates.Simulation of pNO 3 also remains a challenge in models (Shah et al., 2018;Zhai et al., 2021) and is highly sensitive to the parameterization of wet deposition (Luo et al., 2020).Improvements in model representation of these processes would increase confidence in our assessment of the role of pNO 3 photolysis as a driver of tropospheric ozone trends.Emission-driven changes in tropospheric ozone burden, tropospheric burdens of sulfate, HNO 3 , and pNO 3 , and global emissions of SO 2 , NO x , and NH 3 .The burdens are from the base GEOS-Chem simulations using time-varying anthropogenic emissions and methane concentrations (for years 1960, 1980, 1995, 2018, and 2050), but constant meteorology (year 2018).Historical anthropogenic emissions are from the Community Emissions Data System (Hoesly et al., 2018), and emissions for 2050 are from the SSP2-4.5 scenario (Gidden et al., 2019) used in Phase 6 of the Coupled Model Intercomparison Project (CMIP6).The left panel also shows tropospheric ozone burdens from the GEOS-Chem simulation without pNO 3 photolysis.The right panel shows emissions from all sources in GEOS-Chem.

Figure 1 .
Figure 1.Annual mean ozone concentrations in the middle troposphere (800-400 hPa) in 2018.The left panel shows observed ozone concentrations from the ozonesonde (circles) and the IAGOS (squares) data sets.The right panel shows results from the GEOS-Chem simulation sampled at the measurement times and locations.The mean mid-tropospheric ozone concentrations from the observations and the model at the measurement sites are shown inset.

Figure 2 .
Figure2.Vertical profiles and seasonal variations of tropospheric ozone over six regions in 2018.The top panels show the annual mean ozone concentrations between the surface and 200 hPa from the ozonesonde and IAGOS observations, and GEOS-Chem simulations aggregated into six regions.The bottom panels show the observed and simulated monthly mean ozone concentrations in the middle troposphere (800-400 hPa) over the six regions.The figure shows results from the base GEOS-Chem simulation (solid line) and from a simulation without pNO 3 photolysis (dashed line).The number of sites (ozonesonde and IAGOS) in each region (n) is indicated in the top panels (TablesS1 and S2in Supporting Information S1).The shaded areas denote ±1 standard deviation of the annual mean ozone profiles (top panel) and the monthly mid-tropospheric ozone concentrations (bottom panel) at the measurement sites in each region.

Figure 3 .
Figure 3. 1995 to 2018 change in ozone concentrations in the middle troposphere.The top panels show the change of 800-400 hPa ozone concentrations in response to the change in anthropogenic emissions between 1995 and 2018 in the GEOS-Chem base simulation and a simulation without pNO 3 photolysis.Also shown are the trends in mid-tropospheric ozone from ozonesonde (circles) and IAGOS (squares) observations as reported by Christiansen et al. (2022) and H. Wang et al. (2022), respectively.These trends were reported for 25 ozonesonde sites for 1990-2017 (1980-2017 for nine sites), and in 11 IAGOS areas for 1995-2017.

Figure 4 .
Figure 4. Emission-driven changes in tropospheric ozone burden, tropospheric burdens of sulfate, HNO 3 , and pNO 3 , and global emissions of SO 2 , NO x , and NH 3 .The burdens are from the base GEOS-Chem simulations using time-varying anthropogenic emissions and methane concentrations(for years 1960, 1980, 1995, 2018, and  2050), but constant meteorology (year 2018).Historical anthropogenic emissions are from the Community Emissions Data System(Hoesly et al., 2018), and emissions for 2050 are from the SSP2-4.5 scenario(Gidden et al., 2019) used in Phase 6 of the Coupled Model Intercomparison Project (CMIP6).The left panel also shows tropospheric ozone burdens from the GEOS-Chem simulation without pNO 3 photolysis.The right panel shows emissions from all sources in GEOS-Chem.