Impact of a new condensed toluene mechanism on air quality model predictions in the US

G. Sarwar, K. W. Appel, A. G. Carlton, R. Mathur, K. Schere, R. Zhang, and M. A. Majeed Atmospheric Modeling and Analysis Division, National Exposure Research Laboratory, Office of Research and Development, US Environmental Protection Agency, RTP, NC 27711, USA Department of Mathematics, The Hong Kong University of Science & Technology, Clear Water Bay, Kowloon, Hong Kong, China Delaware Department of Natural Resources & Environmental Control, New Castle, DE, USA now at: Department of Environmental Sciences, Rutgers University, New Brunswick, NJ, USA


Introduction
Toluene is an important aromatic compound that can affect ozone (O 3 ) and secondary organic aerosol (SOA) in the atmosphere.However, there is currently a great deal of uncertainty related to toluene chemistry (Calvert et al., 2002).Different chemical mechanisms use different approximations for toluene reactions that can result in different Correspondence to: G. Sarwar (sarwar.golam@epa.gov)ozone predictions.For example, Faraji et al. (2008) used the Carbon Bond IV (CB-IV) (Gery et al., 1989) and the Statewide Air Pollution Research Center (SAPRC-99) chemical mechanisms (Carter, 2000) for Houston and reported that O 3 predictions differed by as much as 40 ppbv.The authors attributed the difference in predicted concentrations mostly due to the differences in aromatic chemistry in the two mechanisms.Yarwood et al. (2005) extended the CB-IV mechanism into an updated Carbon Bond (CB05) mechanism consisting of 156 chemical reactions involving 52 chemical species.Sarwar et al. (2008) evaluated the impact of the CB05 mechanism on model predictions in the US and reported that the use of CB05 increases predicted O 3 compared to those with the CB-IV.Toluene chemistry in CB05 was retained from the CB-IV chemical mechanism.Recently, Whitten et al. (2010) proposed a new condensed toluene mechanism for CB05 mechanism.They performed simulations using the existing CB05 mechanism as well as the new CB05 mechanism containing the new condensed toluene mechanism for 38 environmental chamber experiments involving different combinations of toluene and oxides of nitrogen (NO x ).They used four different performance metrics to compare model predictions with chamber data: maximum O 3 , maximum (NO-O 3 ), NO x crossover time, and cresol concentrations (NO = nitric oxide).The new toluene mechanism provided better results than the existing toluene chemistry in simulating chamber data.This study examines the impact of the new condensed toluene mechanism on air quality model predictions in the US.
G. Sarwar et al.: Impact of a new condensed toluene mechanism on model predictions 2 Methodology

Model description
The study uses the Community Multiscale Air Quality (CMAQ) modeling system (version 4.7) (Binkowski and Roselle, 2003;Byun and Schere, 2006) to simulate air quality.Evaluations for the CMAQ modeling system have recently been conducted by comparing model predictions to measured ambient pollutants (Eder and Yu, 2006;Appel et al., 2007;Foley et al., 2010).The CMAQ model has considerable skill in simulating O 3 mixing ratios in the atmosphere.For example, CMAQv4.7 predicts 8-h maximum O 3 with a normalized median bias of 6.9% and a normalized median error of 14.5% in August 2006 (Foley et al., 2010).Two modeling domains are used for the study.One domain focuses on the western US and consists of 213×192 horizontal grid-cells while the other domain focuses on the eastern US and consists of 213 × 188 horizontal grid-cells with a 12-km resolution.Each model contains 14 vertical layers of variable thickness between the surface and 100 mb with a surface layer thickness of approximately 36 m.Model simulations are performed for the eastern US for July 2001 and for the western US for July 2002.While the modeling time periods used for the western and eastern US domains are different, findings presented herein are not likely to change if same time period is used for the both domains.
The CMAQ chemical transport model is configured to use the mass continuity scheme to describe advection processes, the Asymmetric Convective Model Version 2 (ACM2) (Pleim, 2007) to describe vertical diffusion processes, the multiscale method to describe horizontal diffusion processes, and an adaptation of the ACM algorithm for convective cloud mixing.Aqueous chemistry, aerosol processes, and dry/wet deposition are included.The meteorological driver for the CMAQ modeling system is the PSU/NCAR MM5 system (version 3.5) (Grell et al., 1994).Initial and boundary conditions for this study are obtained from CMAQ model results of a larger modeling domain.
Each domain is first simulated using the CB05 chemical mechanism containing existing toluene chemistry (CB05-Base) (Yarwood et al., 2005) and then using the CB05 chemical mechanism containing the new toluene chemistry (CB05-TU) (Whitten et al., 2010).The difference in modeling results obtained with the two model simulations are attributed to the differences in toluene chemistry.
The CMAQ modeling system currently provides three different gas-phase chemistry solvers: the Sparse-Matrix Vectorized Gear Algorithm solver, the Rosenbrock solver, and the Euler Backward Iterative (EBI) solver.The EBI solver is dependent on chemical mechanism which necessitates the development of a new EBI solver for each new mechanism.The Rosenbrock and the Sparse-Matrix Vectorized Gear Algorithm solvers are generalized solvers that can usually be used for any chemical mechanisms without requiring any changes.The Sparse-Matrix Vectorized Gear Algorithm solver is the slowest but the most accurate among the three solvers.The Rosenbrock solver is faster than the Sparse-Matrix Vectorized Gear Algorithm solver and was used for this study.The use of CB05-TU increases computational time of the model by 3-6% compared to the CB05-Base.

Toluene chemistry
Toluene chemistry in CB05-Base contains 10 chemical reactions involving 5 chemical species.A simplified schematic diagram of the toluene chemistry in CB05-Base is presented in Fig. 1.Reaction of toluene (TOL) with hydroxyl radical (OH) proceeds via three different channels: 36% of the reaction produces cresol (CRES), 56% produces a bicyclic peroxy radical (TO2), and 8% produces benzaldehyde.Contribution of benzaldehyde to O 3 is negligible; thus, it is not further retained in CB05.TO2 reacts with NO to produce organic nitrate (NTR) and a ring-opening product (OPEN) or breaks down to produce CRES.OPEN reacts with OH to form acetyl peroxy radical (C 2 O 3 ), which subsequently produces peroxy acetyl nitrate (PAN) via reaction with nitrogen dioxide (NO 2 ).During daytime, CRES reacts with OH to produce methyl phenoxy radical (CRO), which produces NTR via reaction with NO 2 .Whitten et al. (2010) describe the detailed chemistry of CB05-TU containing 26 chemical reactions involving 13 species for toluene oxidation and provide a simplified schematic diagram for the chemistry (see Fig. 3 in the reference).Reaction of toluene with OH proceeds via four different channels: 18% of the reaction produces CRES, 65% produces TO2, 10% produces benzaldehyde, and 7% produces OH.TO2 reacts with NO to produce NTR, OPEN, and methyl glyoxal.OPEN reacts with OH to form a peroxyacyl radical (OPO3), which leads to an organic PAN (OPAN) via reaction with NO 2 .During daytime, CRES reacts with OH to generate CRO, which produces nitro-cresol (CRON) via reaction with NO 2 .CRON reacts with OH and leads to an alkoxy radical (CRNO) which further reacts with NO 2 yielding NTR.
The impact of the revised toluene mechanism on photochemistry and O 3 production can be best illustrated by the following reactions: where HO 2 = hydroperoxy radical, RO 2 = organic peroxy radical, and O( 3 P) = oxygen atom (triplet), O 2 = oxygen.
The reaction of NO with HO 2 converts NO into NO 2 and causes an increase in O 3 when NO 2 is photolyzed according to Reactions (1-3) (Finlayson-Pitts and Pitts, 2000).The reaction of NO with RO 2 also converts NO into NO 2 and causes an increase in O 3 when NO 2 is photolyzed (Reaction 4) (Finlayson-Pitts and Pitts, 2000).CB05-TU enhances HO 2 and RO 2 directly as well as indirectly through an increase in OH and its subsequent reactions with volatile organic compounds (VOC).If sufficient toluene is present, the new condensed toluene mechanism can enhance HO 2 and RO 2 , and subsequently O 3 .Since CB05-TU enhances OH, it can also affect other pollutants.

Emissions
Toluene is primarily emitted from anthropogenic sources, although some studies (Heiden et al., 1999 andWhite et al., 2009) suggest that biogenic sources can also potentially emit toluene.Anthropogenic toluene sources include industrial processes involving production of toluene, solvent usage, surface coating operations, printing and publishing industries, automotive exhaust emissions, gasoline storage and distribution facilities (USEPA, 1994).Heiden et al. (1999) conducted laboratory and field experiments, and reported the presence of toluene emissions from sunflowers and pine trees.They suggested that plants under stress can emit more toluene than plants without stress.White et al. (2009) recently reported that alfalfa and pine trees can emit toluene and suggested that biogenic sources in northern New England in the US can emit as much as 13% of the total anthropogenic toluene emissions.Toluene emissions from biogenic sources are generally low and not included in biogenic emissions models such as the Biogenic Emissions Inventory System (BEIS).In this study, we use the BEIS (version 3.13) for estimating biogenic emissions (Schwede et al., 2005); as such toluene emissions from biogenic sources are not included.
Anthropogenic emissions are derived from the 2002 National Emissions Inventory (NEI) for the western US and the  2001 NEI for the eastern US.Total toluene emissions in the western US are lower than those in the eastern US.Typical summertime daily toluene emissions in the western US are about 30% of those in the eastern US.Mobile source sector is the major contributor to toluene emissions burden.On-road and non-road mobile sources collectively contribute 44% and 32% of the total toluene emissions in the western and eastern US, respectively.Toluene emissions in urban areas are higher than those in rural areas.

Impact on daily maximum 8-h O 3
Monthly mean daily 8-h maximum O 3 with CB05-Base and increases in mean 8-h O 3 between CB05-TU and CB05-Base are shown in Fig. 5. Elevated levels of 8-h O 3 (> 55 ppbv) are predicted over only a small area in the western US while similar levels are predicted over a large area in the eastern US.CB05-TU increases monthly mean 8-h O 3 by a maximum of 2.8 ppbv in Los Angeles, 1.5 ppbv in Portland, 1.5 ppbv in Seattle, 2.0 ppbv in Chicago, 1.9 ppbv in Cleveland, 1.7 ppbv in northeastern US, and 1.3 ppbv in Detroit compared to those obtained with the CB05-Base.CB05-TU also increases mean 8-h O 3 by 0.5 ppbv or more in several other areas.
Day-to-day variation of the increases in daily 8-h maximum O 3 for Los Angeles, Portland, Seattle, Chicago, New York, and Detroit is presented in Fig. 6.For each area, changes in daily 8-h maximum O 3 between CB05-TU and CB05-Base vary from day to day.While increases are relatively high (5-10 ppbv) on some days, increases are modest on many other days.In Los Angeles, O 3 increases occur on most days while increases in O 3 in Portland and Seattle occur on fewer days.Increases in Chicago and New York are comparable, while increases in Detroit are lower than those in Chicago or New York.While the day-to-day variation in toluene emissions is relatively small, the variation in meteorology affects toluene concentrations and subsequent atmospheric chemistry.Generally, greater toluene levels produce greater increases in O 3 .

Comparison with observed data
Ambient monitoring data from the United States Environmental Protection Agency's Air Quality System are used to evaluate mean bias (MB) and mean normalized bias (MNB) -22 -

Impact on ozone production efficiency
Ozone production efficiency (OPE) is defined as the number of O 3 molecules formed from each molecule of NO x oxidized to NO z and can be calculated from the slope of a regression between O 3 and NO z .OPEs are estimated using O 3 and NO z values during daytime (10:00-05:00 p.m.) and when O 3 /NO x are greater than 46 (aged air mass) (Arnold et al., 2003).OPE obtained with CB05-TU is marginally lower than the value obtained with CB05-Base at Los Angeles (OPE = 5.9 with CB05-TU, OPE = 6.3 with CB05-Base, correlation coefficient = 0.90).OPE obtained with CB05-TU is similar to the value obtained with CB05-Base at Chicago (OPE = 4.1 with both CB05-TU and CB05-Base, correlation coefficient = 0.79).Increases in O 3 with the new mechanism are associated with increased NO z .Similar changes are obtained for other areas in the western and eastern US.Thus, CB05-TU increases O 3 by increasing NO z and without enhancing OPE.

Impact on ozone control strategy
While predicted O 3 mixing ratios are important for model evaluation, relative reduction factors (RRF) are valuable for developing emissions control strategies.reduction in NO x emissions using each mechanism.RRF are calculated using results obtained with normal and reduced NO x emissions for each mechanism.RRF calculated with CB05-Base are identical to values obtained with CB05-TU for most areas.Only minor changes (0.01-0.02) occur in RRF for some isolated areas.CB05-TU does not change RRF compared to those with CB05-Base; thus it is not expected to affect inferences on air pollution control strategies drawn from the model.

Sensitivity of predicted O 3 with toluene emissions
Developing a reliable emissions inventory is a resource intensive process.While tremendous improvements have been made in past years, current emissions inventories still contain large uncertainties (Placet et al., 2000;Sawyer et al., 2000;Werner et al., 2005).To evaluate the sensitivity of predicted O 3 to increased toluene emissions, two additional simulations were conducted by doubling toluene emissions (2 × toluene emissions obtained using NEI).One simulation was conducted using CB05-Base with enhanced  toluene emissions and the other simulation was conducted using CB05-TU with enhanced toluene emissions.Larger increases in O 3 occur between the two mechanisms with enhanced toluene emissions than those with normal toluene emissions.For example, CB05-TU increases daily maximum 8-h O 3 by 9 ppbv in Los Angeles with enhanced toluene emissions compared to an increase of 6 ppbv with normal toluene emissions on July 6.Similarly, CB05-TU increases daily maximum 8-h O 3 by 17 ppbv in Chicago with enhanced toluene emissions compared to an increase of 10 ppbv with normal toluene emissions on 8 July.Thus, CB05-TU can produce additional O 3 compared to those with CB05-Base if greater toluene emissions are present which suggests that the new mechanism can be important in areas with elevated toluene emissions.

Impact on selected aerosol species
Monthly mean anthropogenic SOA, biogenic SOA, and incloud SOA with CB05-Base and their percent increases between the two mechanisms are shown in   Carlton et al. (2010) describe SOA formation mechanisms in CMAQv4.7.In CMAQ, benzene, toluene, and xylene (precursors to anthropogenic SOA) react with OH to produce organic peroxy radicals.These organic peroxy radicals react with NO to produce semi-volatile organic compounds and HO 2 to produce non-volatile SOA.Semi-volatile organic compounds produced via the NO reaction pathway can partition to form SOA. Semi-volatile organic compounds can also form non-volatile oligomers through particle phasereactions.While the CB05-TU produces more organic peroxy radicals via reactions of VOCs with enhanced OH, increases in SOA via the NO reaction pathway are small since NO also decreases with CB05-TU.Since HO 2 increases with CB05-TU, SOA produced via the HO 2 reaction pathway becomes more important and consequently anthropogenic SOA increases.CB05-TU increases anthropogenic SOA in areas with high levels of both anthropogenic SOA precursors and toluene.
Monthly mean biogenic SOA concentrations exceeding 0.4 µg m −3 are predicted over the southeastern US while such values are predicted only over a small area covering northern California and southern Oregon.Biogenic SOA precursors are emitted mostly in the southeastern US; consequently SOA derived from these precursors are high in the southeastern US.CB05-TU increases biogenic SOA by more than 1.0% over a wide area in the southeastern US while increases in the western US are generally lower than 1.0% and occur only over small areas.In CMAQ, isoprene, monoterpene, and sesquiterpene are precursors to biogenic SOA.For SOA production from isoprene, only reaction with OH is considered.For SOA production from monoterpene, reactions with OH, O 3 , O( 3 P), and NO 3 are considered.For SOA production from sesquiterpene, reactions with OH, O 3 , and NO 3 are considered.These reactions produce semi-volatile organic compounds which partition to form SOA. The semivolatile organic compounds can also form oligomers through particle phase-reactions.Acid enhanced isoprene SOA is also accounted in the mechanism.Since oxidants increase with CB05-TU, biogenic SOA also increases.Although percent increases in biogenic SOA and anthropogenic SOA are similar, absolute increases of biogenic SOA are much larger.CB05-TU increases biogenic SOA in areas with high levels of both biogenic SOA precursors and toluene.
Monthly mean in-cloud SOA concentrations of greater than 0.2 µg m −3 are predicted over a large portion of the eastern US while in-cloud SOA concentrations in the western US are generally lower than 0.2 µg m −3 and are predicted only over small isolated areas.In-cloud SOA precursor emissions and cloudiness are more frequent in the eastern US, increases aerosol sulfate by 0.5-1.5% in small areas in the western US and 0.5-1.0% in small areas in the eastern US.CB05-TU increases aerosol sulfate in areas with high levels of both sulfur dioxide and toluene.
Predicted mean aerosol nitrate of greater than 0.8 µg m −3 are predicted over only isolated areas in the western US while similar values are predicted over a much larger area in the eastern US.CB05-TU increases aerosol nitrate by more than 2.0% in some isolated areas in the western US and larger areas in the eastern US.CMAQ produces aerosol nitrate from the partitioning of HNO 3 , which is produced via nighttime homogeneous and heterogeneous hydrolysis of dinitrogen pentoxide as well as daytime production via NO 2 + OH = HNO 3 .The primary reason for the increase in aerosol nitrate is the enhancement of the daytime production of HNO 3 .
Mean ammonium concentrations of greater than 1.2 µg m −3 are predicted over most of the eastern US while similar concentrations are predicted over only small areas in the western US.The spatial distribution of ammonium with CB05-Base follows to that of aerosol sulfate.CB05-TU increases ammonium over only small isolated areas in both the western and eastern US by 0.5-1.0%.Increases in ammonium concentrations with CB05-TU also follow the pattern of the increases in aerosol sulfate.CB05-Base predicts greater than 10.0 µg m −3 of PM 2.5 in most areas in the eastern US while concentrations are typically lower than 5.0 µg m −3 for most of the western US.CB05-TU increases PM 2.5 by 0.5-2.5% in some areas in the western and 0.5-1.0% in the northeastern US.However, increases over 1.0% occur in only a few isolated areas in the western US where predicted mean PM 2.5 concentrations are generally lower than 5.0 µg m −3 .
CB05-TU induced changes in predicted PM concentrations arise primarily from differences in atmospheric oxidant levels.Due to lower actinic flux and temperature in winter, atmospheric chemical reactions proceed at slower rates; consequently atmospheric oxidant levels are lower in winter than in summer.Thus, the impact of CB05-TU on secondary aerosols in winter would likely be less than described here.

Summary
CB05-TU enhances monthly mean daily 8-h maximum O 3 by a maximum of 2.8 ppbv in the western US and 2.0 ppbv in the eastern US.These changes are largely confined to the vicinity of major urban areas.CB05-TU decreases MB at higher observed O 3 concentrations, and increases MB at lower observed O 3 concentrations.CB05-TU enhances OH, HO 2 , RO 2 , and NO z levels compared to the CB05-Base.While it enhances O 3 , it does not increase OPE.The use of CB05-TU does not alter RRF; thus, relative to CB05-Base it is not expected to alter inferences on air pollution control strategy.Its impact on PM 2.5 is small.This study uses relatively coarse horizontal grid spacings since it focuses over a large geographical area.The impact of the new toluene chemistry on air quality modeling results could be more pronounced in areas with higher toluene emissions if finer horizontal grid spacings are used.

Fig. 2 .
Fig. 2. (a) Monthly mean O 3 with CB05-Base (b) percent increases in mean O 3 between CB05-TU and CB05-Base (c) toluene/VOC ratio Fig. 3. (a) Monthly mean OH with CB05-Base and percent increases in mean OH between CB05-TU and CB05-Base (b) mean HO 2 with CB05-Base and percent increases in mean HO 2 between CB05-TU and CB05-Base (c) mean RO 2 with CB05-Base and percent increases in mean RO 2 between CB05-TU and CB05-Base.In each row, the first two plots represent mean values (pptv) and the last two plots represent percent changes (%).

Fig. 4 .
Fig. 4. (a) Monthly mean (daytime) NO z with CB05-Base (b) percent increases in mean NO z between CB05-TU and CB05-Base.westernUS.CB05-TU increases NO z both in the western and eastern US in the same areas that it increases O 3 .The primary reason for the increase in NO z is daytime nitric acid (HNO 3 ) production via the reaction: NO 2 + OH = HNO 3 .CB05-TU enhances OH, thus produces more daytime HNO 3 and increases NO z .

Figure 6 :Fig. 6 .
Figure 6: Day-to-day variation of the increases in daily 8-hr maximum O3 in selected areas

Figure 7 :
Figure 7: The median and inter-quartile range of mean bias for the daily maximum 8-hr O 3 with CB05-TU and CB05-Base: (a) Los Angeles (b) Portland (c) Seattle (d) Chicago (e) New York/New Jersey (f) Detroit.Number beneath each paired evaluation represents the total sample number in each binned range of observed concentration.

Fig. 7 .
Fig. 7.The median and inter-quartile range of mean bias for the daily maximum 8-h O 3 with CB05-TU and CB05-Base: (a) Los Angeles (b) Portland (c) Seattle (d) Chicago (e) New York/New Jersey (f) Detroit.Number beneath each paired evaluation represents the total sample number in each binned range of observed concentration.

Figure 8 :
Figure 8: The median and inter-quartile range of mean normalized bias for the daily maximum 8hr O 3 with CB05-TU and CB05-Base: (a) Los Angeles (b) Portland (c) Seattle (d) Chicago (e) New York/New Jersey (f) Detroit.Number beneath each paired evaluation represents the total sample number in each binned range of observed concentration.

Fig. 8 .
Fig. 8.The median and inter-quartile range of mean normalized bias for the daily maximum 8-h O 3 with CB05-TU and CB05-Base: (a) Los Angeles (b) Portland (c) Seattle (d) Chicago (e) New York/New Jersey (f) Detroit.Number beneath each paired evaluation represents the total sample number in each binned range of observed concentration.
Fig. 9. Predicted anthropogenic SOA concentrations are greater in the eastern US than those in the western US.Concentrations over 0.05 µg m −3 are predicted over most areas in the eastern US while such levels are predicted only over smaller areas in G. Sarwar et al.: Impact of a new condensed toluene mechanism on model predictions -25 -anthropogenic SOA between CB05-TU and CB05-Base (b) mean biogenic SOA with CB05-Base and percent increases between CB05-TU and CB05-Base (c) mean in-cloud SOA with CB05-Base and percent increases in mean in-cloud SOA between CB05-TU and CB05-Base.In each row, the first two plots represent mean values (µg/m 3 ) and the last two plots represent percent changes (%).
Fig. 9. (a) Monthly mean anthropogenic SOA with CB05-Base and percent increases in mean anthropogenic SOA between CB05-TU and CB05-Base (b) mean biogenic SOA with CB05-Base and percent increases between CB05-TU and CB05-Base (c) mean in-cloud SOA with CB05-Base and percent increases in mean in-cloud SOA between CB05-TU and CB05-Base.In each row, the first two plots represent mean values (µg m −3 ) and the last two plots represent percent changes (%).