Photochemical Modeling in California with Two Chemical Mechanisms: Model Intercomparison and Response to Emission Reductions

ABSTRACT

An updated version of the Statewide Air Pollution Research Center (SAPRC) chemical mechanism (SAPRC07C) was implemented into the Community Multiscale Air Quality (CMAQ) version 4.6. CMAQ simulations using SAPRC07C and the previously released version, SAPRC99, were performed and compared for an episode during July– August, 2000. Ozone (O₃) predictions of the SAPRC07C simulation are generally lower than those of the SAPRC99 simulation in the key areas of central and southern California, especially in areas where modeled concentrations are greater than the federal 8-hr O₃ standard of 75 parts per billion (ppb) and/or when the volatile organic compound (VOC)/nitrogen oxides (NO_x) ratio is less than 13. The relative changes of ozone production efficiency (OPE) against the VOC/NO_x ratio at 46 sites indicate that the OPE is reduced in SAPRC07C compared with SAPRC99 at most sites by as much as approximately 22%. The SAPRC99 and SAPRC07C mechanisms respond similarly to 20% reductions in anthropogenic VOC emissions. The response of the mechanisms to 20% NO_x emissions reductions can be grouped into three cases. In case 1, in which both mechanisms show a decrease in daily maximum 8-hr O₃ concentration with decreasing NO_x emissions, the O₃ decrease in SAPRC07C is smaller. In case 2, in which both mechanisms show an increase in O₃ with decreasing NO_x emissions, the O₃ increase is larger in SAPRC07C. In case 3, SAPRC07C simulates an increase in O₃ in response to reduced NO_x emissions whereas SAPRC99 simulates a decrease in O₃ for the same region. As a result, the areas where NO_x controls would be disbeneficial are spatially expanded in SAPRC07C. Although the results presented here are valuable for understanding differences in predictions and model response for SAPRC99 and SAPRC07C, the study did not evaluate the impact of mechanism differences in the context of the U.S. Environmental Protection Agency's guidance for using numerical models in demonstrating air quality attainment. Therefore, additional study is required to evaluate the full regulatory implications of upgrading air quality models to SAPRC07.

IMPLICATIONS

CMAQ simulations of gas-phase pollutants over California were conducted using the newly released SAPRC07C gas-phase chemical mechanism and the previous released version, SAPRC99. SAPRC07C predicted slightly lower concentrations of O₃ and important radical species than did SAPRC99 in key polluted regions of California. In certain regions, SAPRC07C simulates an increase in O₃ in response to 20% reductions in NO_x emissions, whereas SAPRC99 simulates a decrease in O₃. A consequence of this difference is that areas where NO_x controls would be disbeneficial are spatially expanded for SAPRC07C compared with SAPRC99.

INTRODUCTION

Regional chemical transport models (CTMs) are widely used in air quality applications for forecasting ozone (O₃) and particulate matter (PM) pollution events1,2 and for developing emission control strategies for reducing pollution to protect human health.3 CTM simulations are the only reliable method for estimating the response of pollutant concentrations to changes in emissions and meteorology over regional domains. Therefore air quality management agencies use CTM-derived estimates of concentration responses to emission reductions in developing State Implementation Plans for reducing O₃ and PM pollution to comply with the National Ambient Air Quality Standards (NAAQS).4,5 The U.S. Environmental Protection Agency's (EPA) Community Multiscale Air Quality (CMAQ) model6 is routinely used for regulatory applications in California.

A major component of a CTM is the gas-phase chemical mechanism, which is a computational representation of atmospheric chemistry. Atmospheric chemistry is highly nonlinear and involves numerous chemical species and reactions. A CTM's chemical mechanism must adequately represent the rates and relationships of the key gas-phase reactions because many pollutants targeted for reduction form in the atmosphere by chemical reactions (i.e., secondary pollutants). O₃ and a considerable fraction of PM are secondary pollutants, and predictions for these species by CMAQ and other CTMs could be affected by changes to the chemical mechanism. Therefore, the impact of chemical mechanism updates must be understood and evaluated before widespread adoption of a new mechanism is warranted.

The 1999 version of the Statewide Air Pollution Research Center (SAPRC99) mechanism7 is currently used for regulatory applications in California, although several other chemical mechanisms are also routinely used in regional air quality applications elsewhere.8–10 The use of SAPRC99 in California is based on the October 8, 1999 recommendation of the Reactivity Scientific Advisory Committee (RSAC; http://www.arb.ca.gov/research/reactivity/rsac.htm) to the California Air Resources Board (CARB). The SAPRC99 mechanism provides a relatively detailed representation of known atmospheric chemistry while maintaining the computational efficiency necessary for conducting regional air quality simulations over long time periods (e.g., seasonal and annual simulations). Models that incorporate SAPRC chemical mechanisms have been evaluated against chamber11,12 and ambient measurements13,14 as well as predictions based on other mechanisms.15–17

Recently, the SAPRC mechanism was updated to reflect an improved understanding of chemical reactions and kinetics. Detailed science reviews of the updated mechanism (SAPRC07) have been conducted and presented to RSAC (for details, see “March 25, 2009 Meeting” at http://www.arb.ca.gov/research/reactivity/rsac.htm). In contrast to those reviews, this paper focuses on how the updates from SAPRC99 to SAPRC07 affect predictions of the CMAQ air quality modeling system used for regulatory applications in California. Because the EPA guidance for photochemical modeling requires that model results be used in a relative rather than absolute sense,18 the relative response of predictions to emission reductions is considered here. Although PM is a significant concern in California, this study focuses only on gas-phase chemistry, and particle formation is not simulated.

The objectives of this study are as follows: (1) to implement a detailed version of the SAPRC07 chemical mechanism (SAPRC07C) into CMAQ version 4.6 (CMAQv.4.6) and simulate gas-phase chemistry in California during a summer episode, (2) to compare predictions of SAPRC07C with those of SAPRC99 and characterize conditions where predictions differ, and (3) to evaluate differences in the relative response of predictions of SAPRC07C and SAPRC99 to emission reductions and their implications for air quality management. Results of this study provide insight to air quality modelers and managers on potential impacts of upgrading CTMs to the latest SAPRC mechanism.

MECHANISM DESCRIPTION

SAPRC99 has recently been updated to SAPRC07 to reflect an improved understanding of atmospheric chemistry.19,20 In brief, many rate constants and reactions have been modified to match the current state of the science. The aromatic mechanisms have been reformulated and are less parameterized. The ability of the mechanism to be adapted to secondary organic PM modules has been improved by incorporating a more explicit representation of peroxy-peroxy reactions and hydroperoxide formation so that effects of changes in nitrogen oxide (NO_x) concentration on organic product formation can be more accurately represented. The number of volatile organic compound (VOC) species increased in the SAPRC07 update to include more of the compounds available in the emissions inventory.

Rate constants are updated in SAPRC07 for reactions involving nitrogen dioxide (NO₂), hydroxyl radical (HO·_, hydroperoxyl radical (HO₂·_), nitric acid (HNO₃), formal-dehyde (HCHO), peroxyacetyl nitrate (PAN), and other peroxy radicals that may significantly impact simulated O₃ concentrations. Key rate constant changes relative to SAPRC99 are provided in Table 1. See Carter19,20 for other important changes.

Table 1. Key rate constant or photolysis rate changes relative to SAPRC99^a

	Rate Constant or Photolysis Rate^b
Reaction^c	SAPRC07	SAPRC99	Change (%)
NO2 + HV = NO + O3P	7.23 × 10^−1	6.69 × 10^−1	8
O3P + NO = NO2	1.64 × 10^−12	2.45 × 10^−12	−33
O3P + NO2 = NO + O2	1.03 × 10^−11	9.70 × 10^−12	6
O3P + NO2 = NO3	3.24 × 10^−12	1.79 × 10^−12	81
O3 + NO = NO2 + O2	2.02 × 10^−14	1.87 × 10^−14	8
NO2 + NO3 = N2O5	1.24 × 10^−12	1.53 × 10^−12	−19
N2O5 = NO2 + NO3	5.69 × 10^−2	6.74 × 10^−2	−16
N2O5 + H2O + H2O = #2 HNO3 + H2O	1.80 × 10^−39
NO3 + HV = NO + O2	1.91 × 10^0	1.59 × 10^0	20
O1D + H2O = #2 OH	1.99 × 10^−10	2.20 × 10^−10	−10
O1D + M = O3P + M	3.28 × 10^−11	2.87 × 10^−11	14
HONO + HV = OH + NO	1.14 × 10^−1	1.27 × 10^−1	−10
HONO + HV = HO2 + NO2	—	1.60 × 10^−2
OH + HONO = H2O + NO2	5.95 × 10^−12	6.42 × 10^−12	−7
OH + NO2 = HNO3	1.05 × 10^−11	8.81 × 10^−12	19
OH + CO = HO2 + CO2	2.28 × 10^−13	2.08 × 10^−13	10
OH + O3 = HO2 + O2	7.41 × 10^−14	6.78 × 10^−14	9
HO2 + NO = OH + NO2	8.85 × 10^−12	8.36 × 10^−12	6
HO2 + NO2 = HNO4	1.12 × 10^−12	1.36 × 10^−12	−18
HO2 + O3 = OH + #2 O2	1.69 × 10^−15	1.89 × 10^−15	−11
HO2H + OH = HO2 + H2O	1.80 × 10^−12	1.70 × 10^−12	6
MeO2 + HO2 = COOH + O2	4.65 × 10^−12	5.12 × 10^−12	−9
MeO2 + HO2 = HCHO + O2 + H2O	4.50 × 10^−13
MeO2 + MeO2 = MeOH + HCHO + O2	2.16 × 10^−13	2.61 × 10^−13	−17
MeO2 + MeO2 = #2 {HCHO + HO2}	1.31 × 10^−13	1.08 × 10^−13	21
Peroxy + HO2 = products	7.63 × 10^−12	1.45 × 10^−11	−47
MeCO3 + NO2 = PAN	9.37 × 10^−12	1.04 × 10^−11	−10
PAN = MeCO3 + NO2	6.27 × 10^−4	7.04 × 10^−4	−11
PAN + HV = products	6.12 × 10^−5
MeCO3 + NO = MeO2 + CO2 + NO2	1.97 × 10^−11	2.12 × 10^−11	−7
MeCO3 + NO3 = products	2.30 × 10^−12	4.00 × 10^−12	−43
MeCO3 + MeO2 = products	1.06 × 10^−11	9.53 × 10^−12	11
MeCO3 + peroxy = products	1.56 × 10^−11	7.50 × 10^−12	108
PAN2 = RCO3 + NO2	5.48 × 10^−4	5.90 × 10^−4	−7
RCO3 + NO = products	2.08 × 10^−11	2.78 × 10^−11	−25
BZO + HO2 = CRES	7.63 × 10^−12	1.45 × 10^−11	−47
HCHO + HV = #2 HO2 + CO	2.76 × 10^−3	2.32 × 10^−3	19
HCHO + OH = HO2 + CO + H2O	8.47 × 10^−12	9.19 × 10^−12	−8
CCHO + OH = MeCO3 + H2O	1.49 × 10^−11	1.57 × 10^−11	−6
RCHO + NO3 = HNO3 + RCO3	6.74 × 10^−15	3.82 × 10^−15	76
Acetylene + OH = products	7.56 × 10^−13
Acetylene + O3 = products	1.16 × 10^−20
Benzene + OH = products	1.22 × 10^−12
Notes: ^aFrom Carter.18 ^bRate constants are for temperatures of 300 K and are in units of molecule, cm3 sec^−1. ^cNO = nitric oxide, O3P = ground state oxygen atmos, O2 = oxygen, NO3 = nitrate radical, N2O5 = nitrogen pentoxide, O1D = excited oxygen atoms, HONO = nitrous acid, M = air, H2O = water, HNO4 = peroxynitric acid, HO2H = hydrogen peroxide, MeO2 = methyl peroxy radicals, MeOH = methanol, MeCO3 = acetyl peroxy radicals, PAN2 = PPN and other higher alkyl PAN analogues, RCO3 = peroxy propionyl and higher peroxy acyl radicals, BZO = phenoxy radicals, CRES = phenols and cresols, CCHO = acetaldehyde, RCHO = lumped C3+ aldehydes.

The SAPRC07 mechanism has uncondensed (i.e., detailed), condensed, and toxic versions. The standard detailed version is SAPRC07B, and SAPRC07C is one of its expanded versions. SAPRC07C uses the same chemistry and emissions assignments as SAPRC07B, but it also includes reactions for peroxy radical operators. An explicit representation of these processes is aimed at improving simulations of secondary organic aerosol formation. Details on different versions of SAPRC07 are provided in Carter.19–21 SAPRC07C is used in this study because it is the only detailed version of SAPRC07 that is compatible with current CMAQ chemical compilers and explicitly represents peroxy + peroxy reactions and hydroperoxide formation.

The SAPRC07C mechanism used here includes all reactions listed in Tables A-1 and A-2 of Carter.20 The consumption and production of species from the three classes of chemical operators (i.e., xPROD, yPROD, and zRNO₃) are represented by 10 reactions per operator (e.g., see note 32, Table A-1 in Carter20). In total, SAPRC07C treats 118 species and 599 reactions. For comparison, the version of SAPRC99 used here treats 76 species and 214 reactions. The larger number of species and reactions in SAPRC07C resulted in a factor of 2 increase in computational time for the CMAQ simulations with SAPRC07C compared with those with SAPRC99.

MODEL APPLICATION

In this study, CMAQv.4.6 with the SMVGEAR numerical solver was used for the simulations with SAPRC07C and SAPRC99. The CMAQ model domain is illustrated in Figure 1. The domain covers the area from (32.36° north, 125.76° west) to (42.18° north, 113.16° west) with 4- by 4-km horizontal grid cells (269 ξ 269 grid points). In the vertical, the model has 15 layers, which extend from the surface to approximately 100 mb, with the lowest layer nominally 30 m deep. The simulation time period is July 25 to August 3, 2000, with a 2-day spin-up run over July 23 and 24. This episode occurred during the 2000 Central California Ozone Study (http://www.arb.ca.gov/airways/) and has been the focus of previous modeling studies.22,23 Meteorological conditions during the episode are typical of California in summertime, and stagnant conditions conductive to O₃ formation and accumulation occurred from July 29 to August 2. Lehrman et al.24 provide a detailed description of synoptic meteorological conditions during the time period.

Figure 1. (a) Difference in mean daily maximum 8-hr O₃ concentration for predictions with SAPRC99 and SAPRC07C, (b) mean daily 8-hr maximum O₃ concentration for predictions with SAPRC07C. Comparison sites are indicated by markers or numbers. Numbers (1–6) correspond to sites (left–right) in Figures 2 and 3: 1 = FLN, 2 = LIVR, 3 = PLR, 4 = ARV, 5 = RUBI, and 6 = ALPN (see Table 2 for site descriptions).

Figure 2. Hourly concentration predictions for simulations with SAPRC07C compared with those for SAPRC99. Columns (left–right) correspond to numbered sites (1–6) in Figure 1.

Figure 3. Mean diurnal concentration profiles for simulations with SAPRC07C and SAPRC99. Columns (left–right) correspond to numbered sites (1–6) in Figure 1. Hour 0 corresponds to 12:00 A.M. PST.

The meteorological fields were simulated using MM5, the Fifth-Generation Pennsylvania State University/National Center for Atmospheric Research (NCAR) Mesoscale model.25 Analysis and evaluation of MM5 meteorological simulations for the episode were described in a previous study.26 MM5 outputs were processed using the Meteorology–Chemistry Interface Program (MCIP version 3.3) to create CMAQ-ready meteorological inputs.

Emission inputs were created based on the 2000 emission inventories prepared by CARB. Biogenic and on-road mobile source emissions were calculated using day-specific observed air temperature fields generated using an objective analysis. The emission inventories also include day-specific emissions for area and point sources. VOC inventories exclude methane, ethane, and acetone, which have negligible photochemical reactivity in the boundary layer for regional air quality modeling conditions.27 Emissions for the lumped VOC species modeled in SAPRC07C and SAPRC99 were processed separately based on the VOC lumping assignments for the two mechanisms as described at http://www.engr.ucr.edu/~carter/emitdb/. It is worth noting that acetylene and benzene are represented explicitly in SAPRC07C but are included in lumped categories in SAPRC99. The emission inventories used in this study do not contain emissions for the state of Nevada, which is also included in the modeling domain. The lack of emissions for Nevada should have a negligible impact on these air quality simulations for California. Spatial distributions of NO_x and VOC emissions are provided in the supplemental material (published at http://secure.awma.org/onlinelibrary/samples/10.3155/1047-3289.61.5.559_supplmaterial.pdf; Figure S1). Anthropogenic emissions are predominantly located near metropolitan areas (e.g., Los Angeles, San Francisco, etc.) and major highways. VOC emissions tend to be high in the foothills of the Sierra Nevada and coastal mountains because of biogenic sources in those regions.

The chemical boundary conditions were calculated by mapping monthly average outputs from the MOZART-4 global CTM28 onto the CMAQ model domain. Using monthly average concentrations for chemical boundary conditions is suitable for the model intercom-parison purposes of this study. Initial concentrations for the spin-up simulation were based on CMAQ default values and varied spatially.

The analysis of the CMAQ simulations in the following sections focuses on the 25 sites listed in Table 2 and shown on the map in Figure 1. These sites cover the key air basins of California, including the San Joaquin Valley, South Coast, Sacramento Valley, San Francisco Bay Area, South Central Coast, and San Diego air basins. The sites also experience air masses with significantly different chemical characteristics, as evident from the model-predicted VOC/NO_x (ppbC/ppb; NO_x = NO + NO₂) ratios shown in Table 2. The model intercomparison of CMAQ SAPRC07C versus CMAQ-SAPRC99 at these sites provides insight into how the chemical mechanism update affects concentration predictions in different chemical regimes.

Table 2. Site locations and averaged VOC/NO_x (ppbC/ppb) ratios for daytime hours from 10:00 a.m. to 5:00 p.m. (PST)

							VOC/NOx (ppbC/ppb)
Site	Address	Air Basin	Latitude	Longitude	SLE (m)	Category	SAPRC99	SAPRC07C
LANM	1630 North Main Street, Los Angeles	South Coast	34.07	−118.23	87	Urban	3.39	3.46
SUN	Sunol	San Francisco Bay Area	37.59	−121.88	140	Rural	3.53	3.59
AZSA	803 North Loren Avenue, Azusa	South Coast	34.14	−117.92	183	Urban	3.81	3.93
BGS	1128 Golden State Highway, Bakersfield	San Joaquin Valley	35.39	−119.01	151	Urban	4.26	4.34
ULDS	1350 San Bernardino Road, Upland	South Coast	34.10	−117.63	369	Urban	4.31	4.48
MRM	2334 M Street, Merced	San Joaquin Valley	37.31	−120.48	53	Suburban	5.10	5.20
SIM	5400 Cochran Street, Simi Valley	South Central Coast	34.28	−118.68	310	Urban	5.25	5.42
SJO	13030 Murphy Avenue, San Martin	San Francisco Bay Area	37.08	−121.60	87	Urban	5.34	5.50
NAT	3801 Airport Road, Sacramento	Sacramento Valley	38.64	−121.51	5	Urban	5.44	5.48
FSF	3425 North First Street, Fresno	San Joaquin Valley	36.78	−119.77	96	Urban	5.61	5.69
GLA	26 Castilian Drive B-23, Goleta	South Central Coast	34.43	−119.86	30	Urban	6.21	6.25
LIVR	793 Rincon Avenue, Livermore	San Francisco Bay Area	37.68	−121.78	145	Rural	6.67	6.90
RUBI	5888 Mission Boulevard, Rubidoux	South Coast	34.00	−117.42	250	Urban	6.89	7.30
GNBY	Granite Bay	Sacramento Valley	38.74	−121.20	146	Suburban	7.00	7.05
ALPN	2300 Victoria Drive, Alpine	San Diego	32.84	−116.77	603	Urban	8.46	8.89
BANN	200 South Hathaway Street, Banning	South Coast	33.92	−116.86	473	Urban	8.73	9.57
RDS	Johnson Farm, Edison	San Joaquin Valley	35.35	−118.85	128	Urban	9.22	9.39
PGV	7310 Pacific Avenue, Pleasant Grove	Sacramento Valley	38.77	−121.52	10	Rural	10.06	10.54
FLN	50 Natoma Street, Folsom	Sacramento Valley	38.68	−121.16	98	Urban	10.44	10.65
RIVY	695 Young Street, San Jacinto	South Coast	33.80	−116.97	459	Suburban	11.02	12.36
PLR	9240 South Riverbend Avenue, Palier	San Joaquin Valley	36.60	−119.50	78	Rural	11.50	11.90
KCW	Omaha Avenue, Kettleman	San Joaquin Valley	36.10	−119.95	69	Rural	17.75	20.11
ARV	20401 Bear Mountain Boulevard, Arvin	San Joaquin Valley	35.21	−118.78	145	Suburban	22.30	22.71
FEDL	8555 South Valentine, Fresno	San Joaquin Valley	36.61	−119.86	76	Rural	28.11	30.63
ANGI	36078 Fourth Avenue, Corcoran	San Joaquin Valley	35.95	−119.53	60	Rural	31.81	35.90
Notes: SLE = sea level elevation. All of the sites listed are ground sites.

RESULTS AND DISCUSSION

O₃ simulations with SAPRC99 and SAPRC07C were evaluated against available observations from CARB's air quality database at 95 sites. The mean normalized bias (MNB), was computed for daily maximum 8-hr O₃ predictions for days when the observed value exceeded 60 parts per billion (ppb). Over all sites, the MNBs for both simulations are within -7%. In Table S1 of the supplemental materials, averaged daily maximum 8-hr O₃ concentrations from the SAPRC99 and SAPRC07C simulations are compared with observations at all 95 sites. This comparison demonstrates that O₃ predictions from both mechanisms are of similar and good quality. Note that the model evaluation cannot provide reliable conclusions on which mechanism more accurately reflects atmospheric chemistry for several reasons. First, many differences in predictions for the two mechanisms are small and probably within the uncertainty inherent in comparing grid-based model predictions with point measurements.29 Second, observations for many important species (e.g., radicals) are not available, and so an exhaustive model evaluation is not possible. Therefore, the remainder of this paper focuses on gaining insight into the new SAPRC07C mechanism by intercomparing predictions and model response for CMAQ simulations with SAPRC07C and SAPRC99 under a wide range of conditions.

O₃ and Selected Species

The difference in mean daily maximum 8-hr O₃ concentration predicted using SAPRC99 and SAPRC07C is illustrated in Figure 1a for the modeling domain. Color versions of all spatial plots are available in the supplemental material. The 25 comparison sites described in Table 2 are indicated by markers or numbers in Figure 1. The sites indicated with numbers have historically had some of the highest daily maximum 8-hr O₃ concentrations in the respective air basins.

Over some parts of northern and eastern California and the ocean, O₃ concentrations simulated with SAPRC07C are slightly higher than those simulated with SAPRC99, as indicated by the negative values in Figure 1a. However, O₃ predictions for the SAPRC07C simulation are generally lower than those of the SAPRC99 simulation in the key areas of central and southern California with a maximum difference approaching 10 ppb. The mean daily maximum 8-hr O₃ concentration predicted using the SAPRC07C chemical mechanism is shown in Figure 1b. In areas of California where the predicted O₃ concentrations are greater than the federal 8-hr O₃ NAAQS of 75 ppb, SAPRC99 predictions are generally greater than those of SAPRC07C, as indicated by the positive values in Figure 1a. Note that at the sites FSF, PLR, and SIM, the daily maximum 8-hr O₃ concentrations based on SAPRC99 are greater than the 75-ppb O₃ NAAQS, whereas those based on SAPRC07C are less than 75 ppb (Table 3).

Table 3. Summary statistics for O₃ predictions based on SAPRC07C compared with those based on SAPRC99 at 25 sites in California

	Mean Daily Maximum 8-hr O3 (ppb)		Comparison over All Hours
Site ^a	SAPRC99	SAPRC07C	R ^b	NMB ^c	NMD ^d	RMSD ^e
BANN	101.3	94.8	0.99	−6.7	6.8	6.18
RIVY	100.7	94.5	1	−7.6	7.6	5.62
ARV	95.0	85.8	0.99	−6.9	7	6.42
RUBI	83.3	75.6	1	−10.2	10.2	5.51
RDS	82.7	76.3	0.99	−6.9	7	5.01
ALPN	82.0	76.4	0.99	−4.7	4.7	3.79
FLN	79.6	76.2	1	−1.9	2.8	2.4
FSF	78.6	73.6	1	−6.3	6.3	4.01
PLR	78.3	73.7	1	−6	6	3.65
SIM	76.5	71.7	1	−4.3	4.5	3.33
FEDL	74.6	71.4	1	−4.1	4.2	2.68
SJO	71.4	67.8	0.99	−3.2	4.5	3.08
GNBY	71.2	68.4	1	−1.6	2.8	2.11
BGS	69.9	65.1	1	−7.9	7.9	3.53
MRM	69.3	63.8	1	−5.7	6.3	3.71
GLA	68.7	65.7	0.99	−2.6	3.2	2.23
ANGI	67.0	63.8	1	−3.8	3.9	2.41
KCW	66.0	62.9	0.99	−3.3	3.5	2.31
LIVR	65.3	62.1	1	−2.1	4	2.48
ULDS	64.9	58.5	1	−12.1	12.1	4.78
NAT	60.0	58.3	1	−1	3.1	1.47
SUN	59.6	55.6	1	−2.6	4.7	2.65
PGV	55.9	54.8	1	−0.9	2.2	1.22
AZSA	55.0	49.1	0.99	−8.2	8.3	3.94
LANM	40.3	36.5	1	−9.7	9.8	2.55
Notes:
^aSee Table 2 and Figure 1 for site locations; ^bPearson correlation coefficient for predictions with SAPRC07C and SAPRC99; ^c ; ^dNormalized mean difference ^eRoot mean square difference (ppb) .

In the top row of Figure 2, predicted hourly O₃ concentrations based on SAPRC07C and SAPRC99 are compared for the six sites indicated by numbers in Figure 1a. Summary statistics for all 25 sites are given in Table 3. O₃ predictions based on SAPRC07C are generally biased low compared with predictions based on SAPRC99 for the sites considered in this study. For the ARV, RUBI, and ALPN sites, the low bias is more pronounced at relatively high O₃ concentrations (Figure 2). This trend suggests that differences in the chemical mechanisms are enhanced during the peak hours of daytime photochemistry.

Lower concentrations of HO· and HO₂· for simulations with SAPRC07C than SAPRC99 are evident from Figure 2 (bottom two rows). This behavior results from the approximately 19% increase in the rate constant of the reaction of HO· with NO₂ as well as other updates to the reactions involving radicals shown in Table 1. Differences in predicted concentrations between the mechanisms for other species appear to be consistent with the differences in HO· and HO₂· predictions. For instance, HCHO concentrations are generally lower in simulations with SAPRC07C than with SAPRC99. HCHO is produced following the reaction of HO· with hydrocarbons, and so lower HO· concentrations predicted by SAPRC07C can produce lower secondary HCHO concentrations. The lower HCHO concentrations in the SAPRC07C simulation may also result from the 19% increase in the HCHO photolysis rate constant. The decreased rate constant (by 8%) in SAPRC07C for the reaction HCHO + HO· 3 HO₂· + carbon monoxide [CO] + H₂O and the lower HO· concentration would tend to increase the HCHO concentration, but this increase appears to be countered by the lower secondary production rate and the increased photolysis rate. Differences in predictions of NO and NO₂ concentration between mechanisms are consistent with the lower HO₂· concentrations for the SAPRC07C simulation. Because peroxy radicals convert NO to NO₂, the lower HO₂· concentration and changes in rates for reactions involving peroxy radicals might explain the higher NO and lower NO₂ predictions for the SAPRC07C simulation (Figure 2, rows two and three). The generally lower O₃ concentrations for the SAPRC07C simulation are also consistent with lower radical and NO₂ concentration.

Differences in reactive nitrogen (NO_y) partitioning between the mechanisms are evident in Figure 2. For instance, compared with SAPRC99, SAPRC07C appears to preferentially convert NO_x to HNO₃ rather than PAN, which is consistent with the increased rate constant for the HO· + NO₂ reaction. The normalized mean bias (NMB) for comparison of SAPRC07C predictions of HNO₃ with those of SAPRC99 ranges from 7.3 to 23.1% for the six sites in Figure 2, whereas the range is -18.4% to -24.8% for PAN. Also, dinitrogen pentoxide (N₂O₅) concentrations for the SAPRC07C simulation tend to be biased low compared with those for the SAPRC99 simulation, and the correlation coefficient (R) for HNO₃ predictions is relatively low compared with those of other species. These differences can be understood by considering diurnal concentration profiles.

Average diurnal profiles for the simulation period are shown in Figure 3 for the species and sites depicted in Figure 2. The average HNO₃ concentration is generally greater for the simulations with SAPRC07C than SAPRC99 at night (e.g., hour 0), but this difference diminishes during the day as the HO· concentration increases. As the day progresses, the mean HNO₃ concentration for the SAPRC99 simulation becomes similar to that for the SAPRC07C simulation before the trend reverses as the HO· concentration drops at night (Figure 3). The similarity in the HNO₃ concentration for SAPRC99 and SAPRC07C during the day likely results from compensating effects of the greater rate constant for the HO· + NO₂ reaction but lower HO· and NO₂ concentrations in SAPRC07C during the early hours of the day.

Considering that HO· concentrations are negligible at night, the lower nighttime HNO₃ concentration for the SAPRC99 compared with the SAPRC07C simulation is likely because of the differences in N₂O₅ chemistry. A possible explanation for the greater nighttime HNO₃ concentration for SAPRC07C is the third-order, gas-phase N₂O₅ hydrolysis reaction, N₂O₅(g) + H₂O(g) + H₂O(g) → 2HNO₃(g) + H₂O(g), which has been added in SAPRC07C. The high concentration of gas-phase water and production of two HNO₃ molecules per N₂O₅ molecule consumed may partially compensate for the small rate of this reaction (1.8 ξ 10^-39 cm6 mol^-2 sec^-1). Wahner et al.30 indicate that this reaction can be important under conditions of low particle loading, which is true for these simulations (recall that particles are not simulated here). Therefore a lower conversion rate of N₂O₅ to HNO₃ by SAPRC99 compared with SAPRC07C potentially explains the lower HNO₃ and greater N₂O₅ concentrations at night in the SAPRC99 simulations. However, differences in the N₂O₅ conversion rate may be smaller in simulations that consider heterogeneous reactions of N₂O₅ on particles.31 Another reason for lower N₂O₅ predictions for SAPRC07C is the lower O₃ and NO₂ concentrations in that simulation. Finally, the diurnal differences in HNO₃ predictions by SAPRC99 and SAPRC07C help explain the relatively low R values for the HNO₃ predictions shown in Figure 2. The bifurcating pattern in the HNO₃ scatterplots (e.g., see PLR and LIVR, Figure 2) is because of the differences in daytime and nighttime chemistry and leads to relatively low R values for HNO₃ compared with those for other species.

O₃ and OPE as Functions of VOC/NO_x Ratio

The ratio of VOC (ppbC) to NO_x (ppb) is an important measure of the overall nature of the O₃-NO_x-VOC system. The relative behavior of VOCs and NO_x in O₃ formation can be understood in terms of competition for HO·_.32 Previous studies show that the rate of O₃ formation and the amount of O₃ formed per NO_x molecule oxidized depend on the VOC/NO_x ratio and the hydrocarbon compounds that constitute the VOC mixture.33–37 In this section, the impact of the chemical mechanism updates on modeled O₃ concentrations and the amount of O₃ formed per NO_x molecule oxidized is investigated. Such investigation provides a basic understanding of the model's response relative to the O₃-NO_x-VOC precursor relationship in the two mechanisms.

O₃ Versus VOC/NO_x Ratio

Carter20 used the empirical kinetic modeling approach box model38 to simulate a series of single-day scenarios for 39 nonattainment urban sites with initial reactive organic gases (ROG)/NO_x ratios between 0 and 12. The box model results show that the mechanism updates in SAPRC07C generally reduce the predicted maximum O₃ concentrations, with the changes being small at ROG/NO_x ratios close to 11 and increasing as the ROG/NO_x ratio is decreased. The reduction in maximum O₃ is as much as approximately 25% in some scenarios with ROG/NO_x ratios close to 3. The ROG used by Carter20 is equivalent to the VOC used in this study.

The effect of the mechanism update may be different in more realistic three-dimensional (3D) simulations than in the box model simulations.20 Here, differences in O₃ predictions for SAPRC07C and SAPRC99 as functions of the VOC/NO_x ratio for 3D CMAQ model simulations are investigated. The investigation considered 180 comparison sites (including the 25 sites previously discussed) throughout California to cover a nearly full range of VOC/NO_x ratios (∼1 to ∼450) in the state. This allows for characterization of the differences in O₃ predictions for the mechanisms at locations with extremely low and high VOC/NO_x ratios in addition to values typical of urban areas.

The dependence of the modeled concentration of O₃ on the natural logarithm of the VOC/NO_x ratio is shown in Figure 4a for the 180 sites for SAPRC99 and SAPRC07C. The normalized O₃ difference between the mechanisms is shown in Figure 4b as a function of ln(VOC/NO_x) from the SAPRC99 simulation. VOC/NO_x values from SAPRC07C are generally within [H11006]5% of those from SAPRC99. The O₃ and VOC/NO_x values in these plots are averages of all daytime hours from 10:00 a.m. to 5:00 p.m. Pacific Standard Time (PST) for each site. Midday hours are selected because they represent the photochemically active period with the most intense vertical mixing.

Figure 4. (a) Scatterplots of O₃ vs. ln(VOC/NO_x) for 180 sites for SAPRC99 and SAPRC07C. (b) Scatterplots of normalized O₃ difference 100% × (O3,SAPRC07C - O3,SAPRC99)/O3,SAPRC99 vs. ln(VOC/NO_x). (c) Scatterplot of OPE vs. ln(VOC/NO_x). (d) Scatterplot of normalized OPE difference 100% × (OPESAPRC07C - OPESAPRC99)/OPESAPRC99 vs. ln(VOC/NO_x). Note: O₃ and VOC/NO_x are averaged over daytime hours from 10:00 a.m. to 5:00 p.m. for each site. VOC/NO_x values are from the SAPRC99 simulation. OPE is the slope of O₃ vs. NO_z regression for the hours from 10:00 a.m. to 5:00 p.m. at each site. Only the sites with O₃ vs. NO_z R [H11022] 0.7 are considered for panels c and d.

O₃concentrations predicted by SAPRC99 and SAPRC07C are generally low for sites with very low VOC/NO_x ratios (Figure 4a). O₃ concentration increases with the increasing VOC/NO_x ratio and reaches the maximum at a site (BANN; see Table 2 for the site information) with an average daytime VOC/NO_x ratio of 8.7 (ln(VOC/NO_x) = 2.16) for SAPRC99 and SAPRC07C. Although SAPRC99 and SAPRC07C simulate similar VOC/NO_x ratios at the BANN site, the two mechanisms simulate differences in the average daytime O₃ concentrations at this site (98 ppb for SAPRC99 and 92 ppb for SAPRC07C). For low and moderate VOC/NO_x ratios, the O₃ concentration tends to increase with the VOC/NO_x ratio because the efficiency of O₃ production tends to increase when the concentration of peroxy radicals increases relative to the NO_x concentration. The flattening of the O₃ concentrations at high VOC/NO_x ratios suggests that peroxy radical concentrations are already ample for conversion of NO to NO₂ and additional increases in VOC concentrations have a reduced effect on O₃ formation. Results in Figure 4b indicate that SAPRC07C generally predicts less O₃ than SAPRC99 at sites with VOC/NO_x ratios less than 13, with a few exceptions. This behavior is consistent with the findings of Carter.20 For these simulations, the largest O₃ reduction for SAPRC07C compared with SAPRC99 (∼11%) occurs at a site with a VOC/NO_x ratio of 3.8 (ln(VOC/NO_x) = 1.3), whereas the largest O₃ reduction reported by Carter21 (∼20%) in the box model is for an initial VOC/NO_x ratio of 3. Positive O₃ differences (higher O₃ in SAPRC07C), up to approximately 5%, usually occur at sites with a VOC/NO_x ratio greater than 13.

OPE Versus VOC/NO_x Ratio

The relationship between O₃ and NO_z (the sum of all of the oxidation products of NO_x) concentrations has been used to quantify the impact of NO_x emissions on O₃ production. In photochemically aged air, O₃ concentration has been found to correlate well with the NO_z concentration.37,39 The slope of the linear portion of the NO_z versus O₃ correlation is defined as the OPE and is a measure of the number of O₃ molecules produced per NO_x molecule consumed.

NO_z versus O₃ linear regression analysis was performed for all 180 sites for predictions corresponding to daytime hours from 10:00 a.m. to 5:00 p.m. PST. Although the NO_z and O₃ concentrations are significantly correlated at many sites (mostly with relatively aged air masses), there is considerable scatter and even curvature in the NO_z versus O₃ plots for some sites impacted by fresh emissions (data not shown here). Significant curvature in NO_z versus O₃ scatterplots was also reported by Arnold et al.,40 who caution against reliance on the traditional regression relationship between NO_z and O₃. In this study, sites were selected with NO_z versus O₃ R values greater than 0.7 for SAPRC99 and SAPRC07C and the slope was calculated as the OPE. This criterion reduces the number of sites from 180 to 46.

The predicted OPE and normalized OPE differences between the two mechanisms are shown in Figure 4, c and d, as functions of ln(VOC/NO_x) for the 46 sites. These sites cover a wide range of OPE from approximately 2 to 13. Although correlations are low, there is a trend of OPE increase with VOC/NO_x increase. This trend is consistent with previous findings.33–35 Because sites with higher VOC/NO_x ratios are mostly dominated by biogenic emissions with high isoprene concentrations (data not shown here), these results are also consistent with the finding of Jacob et al.34 that higher OPE would be expected in regions with higher isoprene and lower NO_x emissions. The normalized OPE difference between mechanisms, 100% × (OPE _SAPRC07C - OPE _SAPRC99)/OPE _SAPRC99, ranges from -22% to 3%, with a zero or negative difference occurring at only 44 of the 46 sites (Figure 4d). Because the two model simulations use identical emission and deposition schemes, the generally lower OPEs for SAPRC07C are consistent with the lower HO_x· concentration for SAPRC07C shown in Figure 3. Note that previous studies (e.g., Luecken et al.16) have shown that OPE for SAPRC99 is generally greater than that for the carbon bond mechanisms (CB05 and CB-IV).

MODEL RESPONSE TO EMISSIONS CHANGES

In the section on O₃ and selected species, the difference in simulated average daily maximum 8-hr O₃ between the SAPRC99 and SAPRC07C mechanisms (SAPRC07C - SAPRC99) was shown to be as much as -10 ppb. Although the differences in O₃ predicted by the two mechanisms can be large in an absolute sense, the importance of absolute concentrations is reduced in regulatory applications in which air quality models are used in a relative sense. For example, Hogrefe et al.41 showed that the relative response to emissions changes between the CMAQ model and the Comprehensive Air Quality Model with extensions (CAMx), when using the CB-IV chemical mechanism, were within a few ppb of each other although absolute model differences reached 20 ppb.

To quantify model response to emission reductions, simulations utilizing the two mechanisms were made for three cases: a baseline, a 20% NO_x reduction, and a 20% anthropogenic nonmethane VOC (NMVOC) reduction. Ratios of averaged daily maximum 8-hr O₃ from the 20% NO_x or 20% NMVOC reduction simulations to that from the baseline simulation for each mechanism were calculated. The incremental change (IC) in a uniform concentration of 75 ppb as a result of a 20% reduction in NO_x or a 20% reduction in NMVOC emissions for SAPRC99 is shown in Figure 5. IC is calculated as follows:

Figure 5. IC in a uniform O₃ concentration of 75 ppb as a result of (a) 20% reduction in NO_x and (b) 20% reduction in VOC emissions for the SAPRC99 mechanism (see eq 1 for the definition of IC).

(1)

where C _reference is the reference concentration (i.e., 75 ppb), Csensitivity is the averaged daily maximum 8-hr O₃ for the emission-reduction sensitivity run, and Cbaseline is the averaged daily maximum 8-hr O₃ for the baseline run. Negative values (shown outside of the black contour lines; see supplemental material for color plot) in Figure 5 indicate that the emission reduction reduced the averaged daily maximum 8-hr O₃ concentration, whereas positive values (shown inside contour lines) indicate that the emission reduction increased the averaged daily maximum 8-hr O₃ concentration. Generally, NO_x reductions in the SAPRC99 mechanism lead to a decrease in O₃ concentration for all regions except the large urban centers of Los Angeles, San Diego, and the San Francisco Bay Area. The increase in the O₃ concentration in these three regions is likely due to limits on the available radical concentrations, such that a decrease in NO_x results in an increase in radical levels and a corresponding increase in O₃.32 Reductions of NMVOC emissions resulted in reduced or no change in O₃ for all regions.

The average ratios for daily maximum 8-hr O₃ from the 20% NO_x or 20% NMVOC emission reduction simulations are shown in Table 4 for the 25 sites. Ratios less than 1 indicate a decrease in O₃ concentration, those equal to 1 indicate no change, and ratios greater than 1 indicate an increase in O₃ concentration as a result of the emission reductions. Daily maximum 8-hr O₃ ratios for SAPRC99 and SAPRC07C are similar at all sites for the 20% NMVOC emissions reduction scenario, although the SAPRC07C mechanism is slightly more responsive. In contrast, the two mechanisms predict a marked difference in response to 20% NO_x emissions reductions. These differences can be grouped into three cases. In case 1, in which both mechanisms show a decrease or no change (ratio [H11349] 1) in daily maximum 8-hr O₃ because of NO_x emissions reductions, SAPRC99 is more responsive (smaller ratio) than SAPRC07C. Conversely, in case 2, in which NO_x emissions reductions increase the average daily maximum 8-hr O₃ in both mechanisms (ratio [H11022] 1), SAPRC07C is generally more responsive (greater ratio) than SAPRC99 and predicts a larger relative increase in O₃ concentration. In contrast to cases 1 and 2, in which the model response was in the same direction for both mechanisms, for case 3 SAPRC99 predicts a decrease in daily maximum 8-hr O₃, whereas SAPRC07C predicts an increase. A case in which SAPRC07C showed a decrease in daily maximum 8-hr O₃ and SAPRC99 predicted an increase was not observed.

Table 4. Ratio of simulated average daily maximum 8-hr O₃ between the 20% emission reduction and base case simulations for the SAPRC99 and SAPRC07C mechanisms

		O3 Response to 20% NOx Reduction		O3 Response to 20% NMVOC Reduction
Case	Site	SAPRC99	SARPC07C	SAPRC99	SARPC07C
1	ANGI	0.94	0.97	0.97	0.95
	FEDL	0.93	0.97	0.97	0.96
	FLN	0.93	1.00	0.97	0.96
	GNBY	0.95	1.00	0.97	0.96
	KCW	0.94	0.98	0.97	0.96
	NAT	0.95	0.99	0.98	0.97
	PGV	0.93	0.97	0.99	0.97
	PLR	0.94	1.00	0.96	0.93
2	ALPN	1.00	1.05	0.93	0.92
	AZSA	1.29	1.34	0.84	0.82
	BANN	1.06	1.11	0.88	0.87
	LANM	1.36	1.30	0.88	0.87
	LIVR	1.01	1.05	0.96	0.95
	MRM	1.01	1.07	0.94	0.91
	RIVY	1.04	1.07	0.90	0.89
	RUBI	1.18	1.23	0.85	0.83
	SJO	1.02	1.06	0.95	0.94
	SUN	1.05	1.11	0.94	0.94
	ULDS	1.30	1.35	0.82	0.80
3	ARV	0.95	1.04	0.93	0.89
	BGS	0.97	1.02	0.95	0.92
	FSF	0.97	1.02	0.95	0.93
	GLA	0.98	1.03	0.97	0.95
	RDS	0.97	1.03	0.94	0.91
	SIM	0.97	1.06	0.94	0.93

The difference in relative response between SAPRC99 and SAPRC07C for cases 1 and 2 could be due to regional and mechanism constraints on the available radical concentrations. Ambient radical levels are generally lower in SAPRC07C than in SAPRC99 (Figures 2 and 3). Therefore, radical-limited regions (case 2; shown within contour lines in Figure 5a) tend to be “more limited” in SAPRC07C. In the more radical-limited regions simulated with SAPRC07C, the relative increase in O₃ appears to be enhanced compared with SAPRC99. In regions where the available radical concentrations are not the limiting factor (case 1; shown outside of the contour lines in Figure 5a), SAPRC99 generally has a higher OPE than does SAPRC07C (results not shown), such that a reduction in NO_x emissions has a larger relative impact on O₃ levels in SAPRC99. A spatial plot of the three cases is shown in Figure 6. From Figure 6 it is clear that case 3 occurs in the region of transition from radical-limited to radical-sufficient regimes (shown as gray regions in Figure 6). Case 3 arises because of the expansion of the radical-limited regions in SAPRC07C. The expansion of these radical-limited regions is most likely because of the overall lower levels of radicals in SAPRC07C, so that the transition from radical limited to radical sufficient occurs more slowly and over a larger area.

Figure 6. Comparison of the ratios of simulated average daily maximum 8-hr O₃ between the 20% NO_x emission reduction and base case simulation for the SAPRC99 and SAPRC07C mechanisms. Case 1: SAPRC99 ratio [H11349] 1 and SAPRC07C ratio [H11349] 1; case 2: SAPRC99 ratio [H11022] 1 and SAPRC07C ratio [H11022] 1; case 3: SAPRC99 ratio [H11021] 1 and SAPRC07C ratio [H11022] 1.

In Figure 7, the difference in the changes of a uniform O₃ concentration of 75 ppb that results from 20% NO_x (Figure 7a) and 20% VOC (Figure 7b) emissions reduction is shown for the two mechanisms (i.e., ICSAPRC99 - ICSAPRC07C; essentially the difference between Figure 5 and a similar figure for SAPRC07C). Positive values (within solid black contour lines) indicate a greater SAPRC99 ratio whereas negative values (outside of the contour lines) indicate a greater SAPRC07C ratio. For the episode simulated here, SAPRC07C is marginally more responsive to VOC emission reductions than SAPRC99 for most of California with the exception of the region directly downwind (east) of Los Angeles, where SAPRC99 is marginally more responsive than SAPRC07C. In the NO_x emissions reduction case, SAPRC07C ratios are greater than SAPRC99 ratios for most of California, with the exception of the Los Angeles area and small regions near San Francisco and San Diego.

Figure 7. Difference in the IC in O₃ between the SAPRC99 and SAPRC07C mechanisms in response to (a) 20% NO_x or (b) 20% VOC emissions reductions on the basis of a uniform O₃ concentration of 75 ppb (ICSAPRC99 - ICSAPRC07) (see eq 1 for the definition of IC).

The magnitude of the difference in Figure 7 is a measure of the difference in mechanism response to emissions reductions from a uniform O₃ concentration of 75 ppb. For instance, in the NO_x emissions reduction case, a difference of approximately 3 ppb at FEDL (case 1) indicates that SAPRC99 would simulate a reduction in O₃ from 75 to 70 ppb, whereas SAPRC07C would only simulate a reduction to 73 ppb. In contrast, a difference of approximately 3 ppb at BANN (case 2) indicates that although SAPRC99 would simulate an increase in O₃ from 75 to 80 ppb, SAPRC07C would simulate an increase to 83 ppb. Finally, a difference of approximately 4 ppb at BGS (case 3) indicates that although SAPRC99 would simulate a decrease in O₃ from 75 ppb to 73 ppb, SAPRC07C would simulate an increase in O₃ to 77 ppb. Generally, differences are within approximately 5 ppb of each other, with some localized areas experiencing larger differences.

It is important to note that the SAPRC07C mechanism used in this study is one of the several versions available. The final version that will be used for future regulatory applications may be different from this version because of the need to include PM formation and better quantification of the rates for the photolysis of NO₂ and the HO·₊ NO₂ reaction.

CONCLUSIONS

SAPRC07C was implemented into CMAQv.4.6. CMAQ simulations using SAPRC07C and SAPRC99 were performed and compared for a summer episode in California. O₃ predictions of the SAPRC07C simulation are generally lower than those of the SAPRC99 simulation in the key areas of central and southern California. The lower O₃ concentrations predicted by SAPRC07C were especially noticeable in areas where modeled concentrations are greater than the federal 8-hr O₃ standard of 75 ppb. Radical and NO_x concentrations were lower for the SAPRC07C simulations than for SAPRC99 in these areas. This difference is due in part to the increased rate constant for the radical termination reaction HO· + NO₂ 3 HNO₃ in the updated mechanism. The lower predictions of O₃ concentrations for the SAPRC07C simulation are consistent with its lower predictions of HO_x radical and NO₂ concentrations.

Further analysis at 180 sites with a wide range of VOC/NO_x ratios indicates that SAPRC07C generally predicts lower daytime average O₃ than SAPRC99 at sites with VOC/NO_x ratios less than 13. The largest O₃ reduction for SAPRC07C compared with SAPRC99 (∼11%) occurs at a VOC/NO_x ratio of 3.8. Higher O₃ concentrations for the SAPRC07C simulation (up to 5%) generally occur for VOC/NO_x ratios greater than 13. The relative changes of OPE against the VOC/NO_x ratio at 46 sites indicate that the OPE is reduced in SAPRC07C compared with SAPRC99 at most sites and the reduction of OPE is as much as approximately 22%. The findings are generally consistent with the results of the box model simulations by Carter.21 However, because O₃ prediction involves the complex and nonlinear O₃-NO_x-VOC system, the detailed explanation of these changes requires further investigation.

Model response to 20% reductions of VOC emissions is similar for the SAPRC99 and SAPRC07C mechanisms, with SAPRC07C being slightly more responsive. However, the two mechanisms have significantly different responses to 20% NO_x emission reductions. In regions where both mechanisms predict a decrease or no change in daily maximum 8-hr O₃ concentration, simulated O₃ in SAPRC07C is less responsive to NO_x emissions reductions. In contrast, SAPRC07C is generally more responsive to NO_x emissions reductions in regions where both mechanisms show an increase in O₃ with decreasing NO_x emissions. In these regions, O₃ increases due to reduced NO_x emissions tend to be larger in SAPRC07C than in SAPRC99. In some areas, SAPRC07C simulates an increase in O₃ in response to reduced NO_x emissions whereas SAPRC99 simulates a decrease in O₃ for the same region. That is, for SAPRC07C, the areas where NO_x controls would be disbeneficial are spatially expanded compared with that for SAPRC99. The difference in mechanism response in these regions appears to be due to limitations in the available radical concentrations, where SAPRC07C tends to be more radical limited than SAPRC99. The difference in mechanism response to 20% NO_x emissions reductions can deviate by as much as 8 ppb when an initial concentration of 75 ppb is assumed.

The results presented in this work should not be interpreted in a strict regulatory sense because they are based on a limited number of simulation days that were not subject to the rigorous guidelines set forth by EPA's guidance for using numerical models in demonstrating air quality attainment.3 Also, the simulations did not consider aerosol processes. Nevertheless, the results presented here provide insight into differences in concentration predictions and model response for CMAQ simulations with the SAPRC99 and SAPRC07C chemical mechanisms and are a basis for future studies on this topic.

ACKNOWLEDGMENTS

The authors thank Drs. Kemal Gürer and Dazhong Yin (CARB) for the MM5 fields and Dr. Louisa Emmons (NCAR) for the MOZART-4 fields used in this study. The authors also gratefully acknowledge helpful discussions with Drs. Deborah Luecken and William Hutzell (EPA) and Dr. William Carter (University of California–Riverside). This publication is a part of the SAPRC07 peer review coordinated by Dr. Ajith Kaduwela of CARB. Peer reviewers included Drs. Merched Azzi, Richard Derwent, Robert Harley, and William R. Stockwell. This paper has been reviewed by the CARB staff and has been approved for publication. Approval does not signify that the contents necessarily reflect the views and policies of CARB, nor does mention of trade names or commercial products constitute endorsement or recommendation for use.