Sensitivities of sulfate aerosol formation and oxidation pathways on the chemical mechanism employed in simulations

The processes of aerosol sulfate formation are vital components in the scientific understanding of perturbations of earth’s radiative balance via aerosol direct and indirect effects. In this work, an analysis of the influence of changes in oxidant levels and sulfur dioxide oxidation pathways was performed to study the underlying pathways for sulfate formation. Sensitivities of this constituent were calculated from a series of photochemical model simulations with varying rates of NOx and VOC emissions to produce variations in oxidant abundances using a photochemical model (CMAQ) that covers the eastern US for part of the ICARTT 2004 campaign. Three different chemical mechanisms (CBIV, CB05, and SAPRC99) were used to test model responses to changes in NO x and VOC concentrations. Comparison of modeled results and measurements demonstrates that the simulations with all three chemical mechanisms capture the levels of sulfate reasonably well. However, the three mechanisms are shown to have significantly different responses in sulfate formation when the emissions of NO x and/or VOC are altered, reflecting different photochemical regimes under which the formation of sulfate occurs. Also, an analysis of the oxidation pathways that contribute to sulfur dioxide conversion to sulfate reveals substantial differences in the importance of the various pathways among the three chemical mechanisms. These findings suggest that estimations of the influence that future changes in primary emissions or other changes which perturb SO 2 oxidants have on sulfate abundances, and on its direct and indirect radiative forcing effects, may be dependent on the chemical mechanism employed in the model analysis.


Introduction
Since the 1970s, studies of particle sulfate (SO 2− 4 ) formation and fate have played a key role in advancing the scientific understanding of diverse phenomena such as acid precipitation, tropospheric particle matter composition, and, more recently, in the role SO 2− 4 particles play in the direct and indirect forcings of the earth's radiative budget.The study of the formation of SO 2− 4 aerosols involves a complex coupling among gas-and aqueous-phase photochemical reactions and meteorological processes.Formation of SO 2− 4 is chemically linked to primary emissions of sulfur dioxide (SO 2 ) and to the abundance of atmospheric oxidants such as hydroxyl radical (OH), hydrogen peroxide (H 2 O 2 ), ozone (O 3 ), methylhydroperoxide (MHP), and peroxyacetic acid (PAA) (Seinfeld and Pandis, 1998).All of these oxidant species are formed via photochemical reactions which originate from emissions of nitrogen oxides (NO x ) and volatile organic compounds (VOC).Therefore, it is expected that variations in primary emissions of NO x and VOCs may have an effect on the amount and distribution of sulfate (Stein and Lamb, 2002).
The aim of this work was to investigate the dependency of sulfate formation on oxidant levels and on the choice of the chemical mechanism employed to describe the processes by which these oxidants are produced.Towards that end, we present simulated SO 2− 4 concentrations calculated using a three-dimensional regional air quality model with three different photochemical mechanisms and compare results with airborne in situ measurements.Changes in the amount of primary emitted NO x and VOC were then introduced for each mechanism to understand the responses of sulfate formation processes to variations in SO 2 oxidant concentrations.In addition, an oxidation partitioning analysis was performed to delineate the main oxidation pathways participating in the conversion of gas-phase SO 2 to particle SO 2− 4 for each chemical scheme.

Model and measurements
The Community Multiscale Air Quality (CMAQ) modeling system (Byun and Schere, 2006) Version 4.6 was used to simulate the study period that extended from 14-23 July 2004.This modeling period encompasses three flights that took place on 18, 20, and 22 July 2004, corresponding to a portion of the International Consortium for Atmospheric Research on Transport and Transformation (ICARTT) field campaign (Fehsenfeld et al., 2006).The flight patterns corresponding to the study period have been described in Singh et al. (2006).The model configuration included 3-D advection, cloud processes, gas and aqueous chemistry, cloud scavenging and wet and dry deposition.Three different chemical mechanisms were used to test model responses to changes in NO x and VOC emission rates, namely CBIV (Gery et al., 1989), CB05 (Yarwood et al., 2005), and SAPRC99 (Carter, 2000).The aerosol calculation is based on the ISORROPIA v1.7 model (Nenes et al., 1998) and the aqueous phase simulation includes uptake of gases into the cloud droplets, aqueous oxidation of SO 2 by H 2 O 2 , O 3 , MHP, PAA, and catalytic oxidation by iron and manganese (Binkowski and Roselle, 2003) and wet scavenging.The model domain covered the East-ern US with a horizontal resolution of 12 km and 22 vertical layers of variable thickness.The meteorological data for this study was based on Eta model (Black, 1994) outputs linked to the CMAQ modeling system using the methodology described in Otte et al. (2005).Emissions include point, mobile, area, and biogenic sources (Mathur, 2008).The area emissions are based on the 2001 National Emissions Inventory.Emissions of SO 2 and NO x from point sources were projected from 2001 to 2004 based on the Department of Energy's 2004 Annual Energy Outlook.Mobile sources were estimated based on the MOBILE6 model (Pouliot and Pierce, 2003) and biogenic emissions were calculated with the BEIS3.12model (Pierce et al., 2002).
The measurements used for this study from the ICARTT field campaign were obtained from http://www-air.larc.nasa.gov/cgi-bin/arcstat. Results for 18, 20, and 22 July 2004, along the path of the DC8 flight have been utilized in this analysis.Since the aim of this study is focused on sulfate formation, only species relevant to its formation have been included in the comparison.A more complete chemical comparison of CMAQ model results with ICARTT measurements can be found elsewhere (Yu et al., 2010).
Sulfate-NO x -VOC sensitivities were calculated from a series of photochemical model simulations with varying NO x and VOC emission rates to produce responses in oxidant abundances.For each chemical mechanism three simulations were performed: a base case simulation, a simulation with 35 % reduction in anthropogenic NO x emissions, and a simulation with 35 % reduction in anthropogenic VOC emissions.

Discussion
Comparison of base case model results and measurements demonstrates that all three chemical mechanisms capture the levels of boundary layer H 2 O 2 , SO 2− 4 and total nitrate (nitric acid/nitrate) (HNO 3 /NO − 3 ) within 50 % for 20 July 2004 (Fig. 1).(All results presented in this work for July 20 are representative of similar results obtained for July 18 and 22, which are not shown for brevity.)These results are consistent with the model comparisons performed by Yu et al. (2010) for the same dataset.In particular, it is important to notice that all three chemical mechanisms show a very similar performance with regard to fine particle sulfate, with the largest differences occurring in the boundary layer where concentrations are highest.At their largest, the differences in sulfate between the mechanisms are less than 20 %.The similar sulfate concentrations obtained with the three different chemical mechanisms suggest that oxidant-limited conditions (i.e. the SO 2 oxidation is limited by the oxidant availability) prevailed during the study period.Therefore, a sensitivity analysis and an SO 2 oxidant partitioning study were undertaken to investigate whether or not the three mechanisms form these similar sulfate concentrations via similar underlying processes.In order to analyze the response of sulfate levels to changes in primary emissions of NO x and VOCs, Stein and Lamb (2002) proposed the use of a combination of afternoon concentrations of HNO 3 , H 2 O 2 , and SO 2− 4 as "indicator species" of ambient "potential" SO 2− 4 -VOC-NO x sensitivities, where potential sulfate is defined here, after Stockwell (1994) . The concept of potential sulfate was first proposed by Stockwell et al. (1988) and can be interpreted as the maximum concentration of SO 2− 4 that could be produced in the gas and aqueous phases.This idea provides an integrated tool to deal with this complex chemical system.The link between the indicator ratio (defined as } and hereafter denoted as I SO 4 ) and NO x -VOC chemistry can be understood in terms of the dominant sinks and sources for odd-hydrogen.A NO x -sensitive regime, defined as the chemical conditions conducive to a stronger decrease in potential sulfate as a result of a NO x reduction when compared with the same reduction in VOC (see Sillman, 1995 andStein andLamb, 2002), favors the formation of potential sulfate over the production of HNO 3 .Under this condition, the production of H 2 O 2 constitutes the main loss of HO x and the gas-phase SO 2 oxidation is favored over the formation of HNO 3 .On the other hand, the VOC-sensitive regime, featuring a reduction in potential sulfate when reducing VOC and an increase in potential sulfate as NO x is lowered, is characterized by a higher production rate of HNO 3 that overwhelms the formation of potential sulfate.Therefore, larger values of I SO 4 are associated with NO x -sensitive conditions while smaller values for the indicator are associated with VOC sensitive regimes.
Figure 2 Although each particular chemical mechanism shows a distinct response to NO x and VOC changes, the shape of the sensitivity curves depicted in Fig. 2 share some common general features arising from the sensitivity of HO x to changes in NO x levels for all the chemical mechanisms.Each curve shows that for relative low values of I SO 4 (i.e. less than approximately 4, 1, and 3 for CBIV, SAPRC-99, and CB05, respectively) the potential sulfate concentrations tend to decrease when VOC levels are reduced and tend to increase when a reduction in NO x levels occurs.This is, by definition, a VOC sensitive regime.Under these conditions, lowering VOC in the photochemical system will reduce HO x producing less H 2 O 2 .Furthermore, a reduction in NO x increases the availability of HO x , that otherwise will form HNO 3 , to form peroxides instead, therefore increasing the potential sulfate.
Even though VOC sensitive conditions exist for the three mechanisms, when comparing Fig. 2a, b, and c it can be observed that the range indicator values for which VOC conditions prevail differ among the mechanisms with I SO 4 values lower than about 4, 1, and 3 for the CBIV, SAPRC-99, and CB05 chemical mechanisms, respectively.Differences in these values cannot be attributed to a single component of the chemical mechanisms but to a complex interrelation among the radical cycles depicted by each mechanism.For instance, the increasingly higher formation rate of H 2 O 2 from the SAPRC-99, CB05, and CBIV mechanisms (Luecken et al., 2008) explains the higher simulated H 2 O 2 .Since in the CBIV mechanism the formation of H 2 O 2 is the only sink of HO 2 a decrease in NO x will free up more HO x that will directly form even more H 2 O 2 .In CB05 and SAPRC-99, on the other hand, the formation of organic peroxides diverts HO 2 from forming H 2 O 2 , leading to less potential sulfate formation.The three chemical mechanisms give a more convergent picture when modeling HNO 3 , with the CBIV mechanism producing slightly more than the other two.SAPRC-99 and CB05 produce similar HNO 3 levels (Fig. 1b) because in SAPRC-99 the reaction rate for the formation of HNO 3 is lower but the OH levels are higher than in CB05 (Luecken et al., 2008).Therefore, when combining the increasingly higher H 2 O 2 peroxide formation from CBIV, CB05 and SAPRC-99 with a small variation in the HNO 3 among the three mechanisms we end up with increasingly higher values of I SO 4 .Indeed, the linear regression coefficient of the I SO 4 calculated with the three chemical mechanisms against measurements shows values of 1.15, 0.79, and 0.89 for CBIV, SAPRC-99 and CB05, respectively.Conversely, for I SO 4 values higher than 7, 2, and 5 for CBIV, SAPRC-99 and CB05, respectively, the curves in Fig. 2 show features corresponding to NOx sensitive conditions, i.e. potential sulfate is reduced more when NO x levels are reduced than when VOC are.Differences in the VOC speciation and more reactivity of the aromatic fragments (Faraji et al., 2008) can help explain why the SAPRC-99 mechanism tends to show more NO x -sensitivity than the CB05 and CBIV.In SAPRC-99, the relative higher abundance of reactive VOC species with respect to NO x leads to VOC saturated conditions.Under these conditions, a reduction in NO x slows down the cycling between HO 2 and OH lowering the formation of potential sulfate (Stein and Lamb, 2002).
The three chemical mechanisms exhibit markedly different responses in potential sulfate formation when emissions of NO x and VOC are altered; these differing responses reflect different photochemical regimes under which the formation of sulfate occurs with the three chemical mechanisms.Therefore, for a given change in oxidation conditions, (i.e.changes in solar radiation, seasons, primary emissions, etc.) the three chemical mechanisms will produce substantially different responses in sulfate formation.In particular, the SAPRC99 chemical mechanism shows that reductions in NO x are more effective in reducing H 2 O 2 +SO 2− 4 than are reductions in VOCs.On the other hand, results using CBIV show that H 2 O 2 +SO 2− 4 will be reduced more effectively if VOCs are reduced.Finally, the CB05 chemical mechanism exhibits an intermediate behavior somewhere between the other two mechanisms but with closer similarities to CBIV given their common formulation.Theoretical, as well as model-derived, estimates of the threshold value of I SO 4 for the transition from VOC to NO x sensitivity in the formation of potential sulfate were derived by Stein and Lamb (2002) to range from 1.4 to 2.2.In that work, the SAPRC99 chemical mechanism was used and the period of study was from 14-16 July 1995, covering the eastern US. Figure 2b shows that when SAPRC99 is used, the indicator transition seems to be in agreement with the estimated values from Stein and Lamb (2002).On the other hand, Fig. 2a and c shows that the transition has approximate values of 7 and 5 for the CBIV and the CB05 chemical mechanisms, respectively.
The main chemical processes that drive behavior of the SO 2− 4 -VOC-NO x sensitivities for the three chemical mechanisms are expected to be the gas-phase oxidation by OH and the aqueous-phase oxidation mediated by H 2 O 2 while other oxidants are expected to play a minor role (Seinfeld and Pandis, 1998).To determine the importance of the different formation pathways, the sulfate tracking diagnostic model configuration included in CMAQ (Mathur et al., 2008) was employed.Significant differences in sulfate production chemical pathways are observed to be present between the CBIV, CB05, and SAPRC99 mechanisms (Fig. 3).In general, for all three chemical mechanisms, oxidation pathways for SO 2 are dominated by the gas-phase OH and the aqueousphase H 2 O 2 reactions.However, for the CBIV mechanism, up to 30 % of sulfate formation occurs through aqueous phase oxidation by methylhydroperoxide (MHP) from 15:30 to 16:00 p.m. (Figs.3a and 4).Comparison with measured MHP concentrations shows that the model (using CBIV) tends to overestimate levels of MHP by as much as an order of magnitude (not shown).Furthermore, for the CB05 and the SAPRC99 mechanisms, the aqueous SO 2 oxidation pathway mediated by peroxyacetic acid (PAA) accounts for up to 20 % of the sulfate formation from 15:30 to 16:00 p.m. (Figs.3b, c and 4).Comparison of PAA concentrations with observed values indicates that the model (using CB05 or SAPRC-99) tends to overestimate the levels of PAA by as much as 4 times the observed values (Fig. 5).This can be understood in terms of the stoichiometric parameters used to describe the formation of PAA though the reaction of peroxyacetyl radical (PA) and HO 2 .In CB05 and SAPRC99, 0.8 and 0.75 PAA molecules are formed for each PA+HO 2 , respectively, while for the CBIV mechanism only 0.21 PAA are formed.When interpreting these results it has to be taken into account that CBIV, CB05, and SAPRC-99 are condensed mechanisms that have originally been developed to model ozone concentrations and its precursors (mostly focused on hydrocarbons and NO x ) by replicating smog chamber experiments.The use of lumping techniques to condense the complex atmospheric processes into these chemical mechanisms produces reasonably good representations of the ozone formation (Carter, 2010).However, the description of the formation of other species, like PAA, does not appear to be as accurate.

Conclusions
This study illustrates that, despite the close agreement between the observed and simulated sulfate concentrations, each of the three chemical mechanisms presents a different sensitivity response of sulfate formation to changes in oxidant levels.The SAPRC99 mechanism generally presents sulfate levels that are NO x sensitive along the path of the flights for the cases analyzed.On the other hand, CBIV and CB05 show VOC sensitivity to be dominant for the formation of sulfate.In addition, beyond the dominant pathways for SO 2 to sulfate conversion (aqueous H 2 O 2 and gaseous OH oxidation), different secondary sulfate formation pathways are simulated among the three chemical mechanisms.For CBIV, the aqueous phase oxidation via MHP constitutes the third dominant reaction while for SAPRC99 and CB05 aqueous oxidation via PAA dominates.
These findings imply that good agreement between measured and modeled concentrations is a necessary but not a sufficient condition for an accurate depiction of sulfate chemical formation processes.Indeed, recent studies (e.g.Goto et al., 2011) highlight the importance of an accurate description of sulfate formation chemical pathways, in particular in global aerosol models.The more detailed formulation of sulfate chemical processes presented in Goto et al. (2011) showed an improvement in simulation performance when compared with observations and revealed a difference of approximately 50 % in the estimated radiative forcing with respect to the very simplified approaches generally utilized in the global simulations included in the IPCC-AR4 report.However, as shown in the present work, simulations of sulfate may agree with the measured values despite the use of completely different chemical mechanisms and consequent differing formation pathways.Our analysis of the influence of changes in oxidant levels and SO 2 oxidation pathways strongly suggests that the choice of chemical mechanism may produce noticeable differences in sulfate distributions given changes in primary emissions of NO x or VOCs, or given some other change which affects the amount or relative concentrations of SO 2 oxidants.This implies that estimations of the influence of future changes in emissions on sulfate levels, and therefore on its direct or indirect radiative forcing effects, may be highly dependent on the chemical mechanism employed in the model analysis.Giving the limited time and space scales examined in the present study, further research is needed to establish the extent of influence of these findings in a global-scale modeling framework and to determine what impacts of these sensitivities may be potentially propagated to long-term climate scenario analyses.
An additional conclusion that can be drawn from this work is one which is well known among three-dimensional atmospheric chemistry modelers in the context of tropospheric ozone simulation, namely that the right answer can be obtained for the wrong reasons.In other words, compensating errors within the modeling framework can lead to results which compare well with measurements, but the model errors can remain hidden unless the right measurement-model comparisons are undertaken.In this work, we have identified a potential source of compensating errors in the simulation of fine particle sulfate.Further research is necessary to understand the full implications of this finding.
(b) SAPRC99, and (c) CB05 along the path of the flight of the DC8 on July 20, 2004.In 391 the legend: ICBC stands for initial conditions and boundary conditions.
ICBC stands for initial conditions and boundary conditions.

Fig. 3 .
Fig. 3. Percent contribution of sulfate formation pathways to total sulfate for (a) CBIV, (b) SAPRC99, and (c) CB05 along the path of the flight of the DC8 on 20 July 2004.In the legend: ICBC stands for initial conditions and boundary conditions.
presents the normalized percentage reduction in {[H 2 O 2 ]+[SO 2− 4 ]} (i.e., potential sulfate) concentration as a consequence of either a NO x or a VOC emission reduction for each chemical mechanism along the path of the DC8 flights for 18, 20, and 22 July 2004.The difference in potential sulfate levels between the base case and the NO x and VOC reduction scenarios is plotted as a function of the simultaneously modeled I SO 4 , 26

Figure 4 .Fig. 4 .
Figure 4. Pie chart showing the percent contribution of sulfate formation pathways to