Projected ozone trends and changes in the ozone-precursor relationship in the South Coast Air Basin in response to varying reductions of precursor emissions

ABSTRACT

This study examined the effects of varying future reductions in emissions of oxides of nitrogen (NO_x) and volatile organic compounds (VOC) on the location and magnitude of peak ozone levels within California’s South Coast Air Basin (SoCAB or Basin). As ozone formation is currently VOC-limited in the Basin, model simulations with 2030 baseline emissions (−61% for NO_x and −32% for VOC from 2008) predict 10–20% higher peak ozone levels (i.e., NO_x disbenefit) in the western and central SoCAB compared with the 2008 base simulation. With additional NO_x reductions of 50% beyond the 2030 baseline emissions (−81% from 2008), the predicted ozone levels are reduced by about 15% in the eastern SoCAB but remain comparable to 2008 levels in the western and central Basin. The Basin maximum ozone site shifts westward to more populated areas of the Basin and will result potentially in greater population-weighted exposure to ozone with even a relatively small shortfall in the required NO_x reductions unless accompanied by additional VOC reductions beyond 2030 baseline levels. Once committed to a NO_x-focused control strategy, NO_x reductions exceeding 90% from 2008 levels will be necessary to attain the ozone National Ambient Air Quality Standards (NAAQS). The findings from this study and other recent work that the current VOC emission estimates are underestimated by about 50% suggest that greater future VOC reductions will be necessary to reach the projected 2030 baseline emissions. Increasing the base year VOC emissions by a factor of 1.5 result in higher 2008 baseline ozone predictions, lower relative response factors, and about 20% lower projected design values. If correct, these findings have important implications for the total and optimum mix of VOC and NO_x emission reductions that will be required to attain the ozone NAAQS in the SoCAB.

Implications: Results of this study indicate that ozone levels in the western and central SoCAB would remain the same or increase with even a relatively small shortfall in the projected NO_x reductions under planned NO_x-focused controls. This possibility, therefore, warrants a rigorous analysis of the costs and effects of varying reductions of VOC and NO_x on the formation and combined health impacts of ozone and secondary particles. Given the nonlinearity of ozone formation, such analyses should include the implications of gradually increasing global background ozone concentrations and the Basin’s topography and meteorology on the practical limits of alternative emission control strategies.

Introduction

California’s South Coast Air Basin (SoCAB or Basin) has historically experienced the most severe ground-level ozone (O₃) pollution in the United States. The Basin encompasses 10,743 square miles and is home to over 17.8 million people (2010 U.S. Census). Prior to the implementation of emission reduction measures in the early 1950s, hourly averaged ozone mixing ratios in the SoCAB approached 700 ppb and Stage III episodes (ozone exceeding 500 ppb) were relatively frequent events in the 1960s. Four decades of progressively more stringent controls of volatile organic compound (VOC) and oxide of nitrogen (NO_x) emissions have reduced the number of exceedances of the 2008 8-hr ozone National Ambient Air Quality Standard (NAAQS) of 75 ppb from 211 days in 1975 to 88 days in 2013 and the Basin design value (DV) from 275 to 107 ppb. Although substantial progress has been made, the ozone NAAQS is still exceeded in the Basin on most summer days and attainment of the standard remains a long-term challenge that has become more difficult with the recent revisions of the ozone NAAQS from 75 to 70 ppb (U.S. Environmental Protection Agency [EPA], 2015).

High concentrations of ozone and fine suspended particles (particulate matter with an aerodynamic diameter <2.5 μm; PM_2.5) in the SoCAB result from the combination of high mountains that constrain dispersal of air pollutants, adverse meteorology that results in low mixing layers that limit atmospheric dispersion, coupled with emissions from the second most populated urban area in the United States. The SoCAB is within the semipermanent high-pressure zone of the eastern Pacific. Frequent and persistent temperature inversions are caused by subsidence of descending air that warms when it is compressed over cool, moist marine air. These inversions often occur during periods of maximum solar radiation, which create daytime mixed layers of less than 1000 m thickness. Summertime flow patterns in the SoCAB are from the west and south during the morning, switching to predominantly westerly winds by the afternoon. On-shore breeze is strong during the day, and winds are calm overnight with a weak land-sea breeze. The land-sea breeze circulation moves air back and forth between the SoCAB and the Pacific Ocean, as well as along the coast to other air basins. Heating of the San Gabriel and San Bernardino mountains during the daytime causes upslope flows that can transport pollutants from the surface into the upper parts of, and sometimes above, the mixed layer. When the slopes cool after sunset, the denser air flows back into the SoCAB with pollutants entrained in it. Although local topography, meteorology, and pollutant emissions are the causative factors of peak ozone levels in the Basin, much of the difficulty in addressing the ozone problem in the SoCAB is due to the complex chemistry of ozone formation, which is a nonlinear function of the amount and relative mix of VOC and NO_x.

The photochemical processes in the atmosphere that lead to the formation of ozone and products of photochemical reactions are well known (Seinfeld, 1986; Finlayson-Pitts and Pitts, 1986; Stockwell et al., 2012; Calvert et al., 2015). Ozone formation is nonlinear because VOC and NO_x compete with one another for hydroxyl (HO) radicals. HO initiates the oxidation of VOC, which produces organic peroxy (RO₂) and hydroperoxy (HO₂) radicals that convert nitric oxide (NO) to nitrogen dioxide (NO₂), which leads to formation of new O₃. The photolysis of NO₂ is the direct source of ozone in the troposphere, and it is recycled during the production of ozone from VOC; thus, NO_x acts as a catalyst in the production of ozone. Too much NO_x reduces the net rate of ozone production because NO rapidly titrates O₃ back to NO₂ and O₂, and NO₂ competes for HO to make nitric acid (HNO₃). HNO₃ is then removed by dry deposition or is converted to ammonium nitrate, which is a major component of PM_2.5 in the SoCAB. At a given level of VOC, there exists a NO_x mixing ratio at which a maximum amount of ozone is produced. This optimum VOC/NO_x ratio, expressed as a molar ratio of VOC in parts per billion carbon (ppbC) to NO_x in ppb, is about 10–12, and corresponds to the ridgeline in an isopleth plot of maximum ozone produced relative to initial concentrations of VOC and NO_x. The HO radical chain length, which is the number of times a newly formed HO radical is regenerated through radical chain propagation before it is removed by forming hydrogen peroxide and organic peroxides, reaches a maximum at this VOC/NO_x ratio.

Reductions of VOC and NO_x have varying effect on the rate and efficiency of ozone formation. In the “NO_x-limited” or “NO_x-sensitive” region below the ridgeline in an ozone isopleth plot (i.e., VOC/NO_x ratios >12), lowering NO_x effectively reduces O₃, whereas reductions in VOC have practically no effect. Ozone is reduced near the ridgeline by simultaneous reductions in VOC and NO_x emissions. In the “VOC-limited” or “VOC-sensitive” region above the ridgeline (i.e., VOC/NO_x ratios <10), lowering VOC effectively reduces O₃. However, NO_x reductions can increase O₃ under VOC-limited conditions by lowering the rate at which OH and NO₂ are removed by the formation of HNO₃. “NO_x disbenefit” refers to this situation. NO_x disbenefit, though counterintuitive, is a well-established aspect of ozone photochemistry and is rooted in the nonlinear ozone chemistry described above. Past studies have also associated the root cause of higher weekend ozone levels (“weekend ozone effect”) to lower weekend NO_x emissions and higher VOC/NO_x ratios, which result in more efficient and rapid O₃ formation on weekends relative to weekdays (Brönnimann et al., 2000; Marr and Harley, 2002a, 2002b; Beirle et al., 2003; Blanchard and Tanenbaum, 2003; Fujita et al., 2003a, 2003b, 2013; Huess et al., 2003; Lawson, 2003; Yarwood et al., 2003; Qin et al., 2004a, 2004b).

The evolution of the ozone-precursor relationship in the SoCAB during the past four decades reflects the corresponding changes in VOC and NO_x emissions. The measured 6–9 a.m. ambient VOC/NO_x ratios in the Basin were near the optimum ratio of 10 in the 1980s (Fujita et al., 1992). Control of either NO_x or VOC alone or simultaneous control of both precursors would have been effective at the time. However, emission control technology development allowed for earlier and more effective control of VOC and CO than for NO_x. The weekday 6–9 a.m. ambient VOC/NO_x ratio decreased steadily with VOC-focused emission controls through the 1990s, reaching a minimum of about 4 in the early 2000s (Fujita et al., 2013). The minima in VOC/NO_x ratios coincided with historically low weekday peak ozone levels at some central Basin locations and maxima in the magnitude of the observed weekend ozone effect (e.g., 25% and 50% higher O₃ levels on Saturdays and Sundays, respectively, at Azusa) (Fujita et al., 2013). The weekday VOC/NO_x ratios have since increased to about 5–6 during the past decade and are projected to increase with greater future emphasis on NO_x controls.

Besides the historic changes due to long-term changes in VOC and NO_x emissions, there are substantial spatial differences in the ozone-precursor relationship within the Basin. Ambient VOC/NO_x ratios initially reflect the mix and type of local emission sources but increase with time during atmospheric transport of emissions because the HO radical reacts more rapidly with NO₂ than VOC. This causes ambient NO_x mixing ratios to decrease during the day relative to VOC, leading to generally higher VOC/NO_x ratios within the Basin. Therefore, the VOC/NO_x ratio at a particular location depends upon time of day, pattern of transport, and the timing and amounts of additional fresh emissions. Consequently, although NO_x reductions can be counterproductive under VOC-limited conditions in and near the source areas (e.g., western and central SoCAB), they can ultimately lead to lower peak ozone levels in downwind receptor locations (e.g., eastern SoCAB). The effects of NO_x reductions in the eastern SoCAB are similar to the eastern United States, where ambient VOC/NO_x ratios are higher due to significant continental background VOC concentrations and abundant sources of biogenic emissions (Aburn et al., 2015). Although ozone levels have continued to decline slowly during the past decade at downwind locations in the eastern Basin, the weekday ozone trends in the central SoCAB have flattened and even increased at some air quality monitoring sites in the central Basin (Fujita et al, 2013). These observations illustrate that the Basin design value (DV) and number of exceedance of the NAAQS can be misleading indicators of Basin-wide ozone trends and mask the recent relative lack of progress in more populated areas of the central Basin compared with the current design value site at Crestline in the San Bernardino Mountains, which currently determines the ozone attainment status for the SoCAB.

The South Coast Air Quality Management District (SCAQMD) and the California Air Resources Board (CARB) are implementing a long-term multipollutant (ozone and PM_2.5) control strategy that is primarily NO_x focused. Preliminary model evaluations by both agencies indicate that reduction of NO_x emissions in excess of 90% from 2008 levels would be required to attain the ozone NAAQS throughout the SoCAB (SCAQMD, 2013, 2015; CARB, 2014). The necessary reductions equate to allowable emissions of NO_x in the Basin (“carrying capacity”) of approximately 85 tons per day (tpd) to meet the 75 ppb 8-hr ozone standard (SCAQMD, 2013) compared with the 2008 emissions of 723 tpd and projected 2030 “baseline” emissions of 284 tpd. The 2030 baseline takes into account the effects of only currently adopted control programs, which include the LEV (Low-Emission Vehicle)-III standards. For perspective, on-road vehicles are projected to account for 95 tpd of NO_x in the 2030 baseline inventory (SCAQMD, 2013). Thus, even eliminating all of the remaining NO_x emissions from on-road vehicles would still leave a residual of 189 tpd, which is more than twice the projected carrying capacity. The projected NO_x emission reductions that would be necessary for attainment of the ozone NAAQS are not feasible with current available technology, and the SCAQMD acknowledges that “it would be the greatest air quality challenge the region has ever faced relative to achieving the additional NO_x emission reductions that would be necessary, and would further necessitate transformational technologies with zero or near-zero combustion emissions” (SCAQMD, 2013). Furthermore, the SCAQMD acknowledges that implementation of a NO_x-focused control strategy will results in an interim period of higher ozone levels in the western Basin unless there are additional VOC reductions beyond the 2030 baseline (SCAQMD, 2015).

The nonlinear nature of ozone photochemistry combined with local topographical and meteorological conditions that are conducive to adverse air quality make attainment of the ozone NAAQS a prolonged intractable problem in the SoCAB. The Coordinating Research Council (CRC) Project A-91 was conducted to forecast the effects of varying reductions of NO_x and VOC emissions on the location and magnitude of peak ozone levels within the SoCAB. The objective of the study was to determine the persistence, magnitude, and spatial extent of the interim period of elevated ozone levels that could result from a NO_x-focused control strategy prior to transition from VOC-sensitive to NO_x-sensitive ozone formation. We also examined the implications of underestimating current inventories of VOC emissions on the relative effectiveness of VOC or NO_x reductions on ambient ozone levels and ozone NAAQS attainment.

Approach and methods

Long-term changes in the spatial and temporal variations of peak ozone levels

As an update of our prior work (Fujita et al., 2003, 2013), we examined the trends in daily maximum 8-hr average ozone mixing ratios from 1980 to 2013 by day of the week for six SCAQMD monitoring stations in the SoCAB. Sampling locations include downtown Los Angeles in the upwind western Basin, three central Basin sites in Azusa, Pomona, and Upland, a downwind eastern Basin site in Riverside-Rubidoux, and a far downwind site at Crestline/Lake Gregory (1300 m elevation) in a forested area in the San Bernardino Mountains. Figure 1 shows the locations of the relevant monitoring sites. The historic trends (1980–2013) in spatial and day-of-week variations of peak ozone levels in the SoCAB were related to corresponding changes in mean daily maximum mixing ratios of NO_x and VOC (or carbon monoxide as surrogate). Future trends in NO_x and VOC mixing ratios were projected based on the expected changes in emissions from 2010 to 2030.

Figure 1. Map of the South Coast Air Basin with approximate locations of relevant air quality monitoring stations.

Projected changes in the magnitude and spatial variations of peak ozone levels in response to varying reductions in precursor emissions

Researchers at the University of Texas El Paso (UTEP) ran the Community Multiscale Air Quality (CMAQ) model (version 4.7) simulations for 2008 and 2030 using the meteorological and emissions inputs and model setup used by the South Coast Air Quality Management District (SCAQMD) to develop their 2012 Air Quality Management Plan (AQMP) (SCAQMD, 2013). Much of the relevant supporting information, including the modeling domain and CMAQ model setup and parameters, are described in Appendix V (Modeling and Attainment Demonstrations) of the 2012 AQMP (SCAQMD, 2013). The simulations were made with the same spatial grid resolution (4 × 4 km) and chemical mechanisms (SAPRC99 photochemical, 5-aerosol module, and saprc99_ae5_aq aqueous chemistry) as those used by the SCAQMD. In this project, the existing script codes from SCAQMD were enhanced to run CMAQ in parallel; 96 processors were used for the new simulations. For the multiday simulations, SCAQMD Profile files were used to initialize the model and subsequently daily CMAQ generated output files were used as input for the initial conditions of the following day.

The simulations made by UTEP were compared with the CMAQ output files provided by the SCAQMD for 91 days in summer 2008 (June 1 to August 30). Following successful reproduction of the SCAQMD 2008 base year simulations, a series of simulations were made by adjusting the 2008 and 2030 emission inputs by the factors shown in Table 1 to forecast the effects of varying incremental NO_x reductions and potential underestimations of reactive organic gas (ROG) emissions. Same number of days, meteorology, and model setup were used in each of the model runs. Sensitivity simulations were made to determine whether model performance would be improved with upward adjustment to the 2008 base ROG emissions by factors of 1.5 and 2.0. The model sensitivity analysis examined the spatial variations of peak ozone levels within the SoCAB resulting from varying incremental NO_x emission reductions beyond the projected 2030 baseline inventory. Effect of adjustments to the base ROG emissions on the relative response factors (RRF), projected design values (DVs), and extent of emission reductions necessary to show attainment of the ozone NAAQS were examined.

Table 1. 2008 and 2030 CMAQ model simulations with baseline total ROG and NO_x emissions (highlighted in gray) and sensitivity cases with varying adjustments to baseline emissions.

Year	Case	ROG				NOx				ROG/NOx
		Total ROG (tpd)	Adj Factor	Δ 2008 Base	Δ 2030 Baseline	Total NOx (tpd)	Adj Factor	Δ 2008 Base	Δ 2030 Baseline
2008	0	1277	2.0	100%	192%	723	1.0	0%	155%	5.9
	1.5R1N	958	1.5	50%	119%					4.4
	1R1N	639	1.0	0%	46%					2.9
2030	1R’2.5N’	437	1.0	−32% from 1R −54% from 1.5R	0%	710	2.5	−2%	150%	2.1
	1R’1.75N’					497	1.75	−31%	75%	2.9
	1R’1N’					284	1.0	−61%	0%	5.1
	1R’.75N’					213	0.75	−70%	−25%	6.8
	1R’.5N’					141	0.5	−81%	−50%	10.4
	1R’.3N’					85	0.3	−88%	−70%	17.1
2030	1.5R’2.5N’	655	1.5	3%	50%	710	2.5	−2%	150%	3.1
	1.5R’1.75N’					497	1.75	−31%	75%	4.4
	1.5R’1N’					284	1.0	−61%	0%	7.7
	1.5R’.75N’					213	0.75	−70%	−25%	10.2
	1.5R’.5N’					141	0.5	−81%	−50%	15.5
	1.5R’.3N’					85	0.3	−88%	−70%	25.7
	.5R’2.5N’	218	0.5	−66%	−50%	710	2.5	−2%	150%	1.0
	.5R’1N’					284	1.0	−61%	0%	2.6

Simulation of the chemistry of the ozone-precursor relationship

Chemical box modeling simulations were performed to examine the retrospective and prospective changes in the maximum ozone mixing ratios and rates of ozone formation. The modeling procedure examines changes in ozone photochemistry independently rather than the combined effects of chemistry and pollutant transport. The response of ozone concentrations to changes in VOC and NO_x were reconciled with past ambient ozone trends and used to project the effects of projected changes in precursor emissions in 2030.

These simulations were performed with the chemical box model (SBOX; Seefeld and Stockwell, 1999). The SBOX model employed the Regional Atmospheric Chemistry Mechanism, version 2 (RACM2 mechanism; Goliff et al., 2013). RACM2 is an EPA recognized mechanism that is an option in the Community Multi-scale Air Quality Model (Sarwar et al., 2013). SAPRC and RACM2 both use a lumped molecular approach to organic chemistry. RACM2 consists of 363 chemical reactions, including 33 photolytic reactions, and uses 120 chemical species to describe atmospheric chemistry. Time-dependent photolysis rate coefficients were calculated from spectrally resolved actinic flux calculated according to Madronich (1987) with absorption cross-sections and quantum yields according to Goliff et al. (2013).

The most recently available VOC data for the SoCAB from photochemical assessment monitoring stations (PAMS) were used to create a representative contemporary mix of VOC. The PAMS organic compound mixing ratios for six stations (at Azusa, Banning-Hathaway, Burbank-West Palm, Los Angeles-West, Pico Rivera, and Upland-San Bernardino) were averaged and assigned to species groupings used in the RACM2. The organic compounds were allocated according to the assignments for alkanes, alkenes, alkynes, and carbonyls in Table S-1, and for aromatic hydrocarbons in Table S-2, and used the percentage composition for the RACM2 species from the VOC profile given in Table 2. Updated ozone isopleth plots were derived from a matrix of 2000 simulations with varying initial NO_x and VOC mixing ratios shown in Table S-3 for the years 1995–2011. The VOC and NO_x mixing ratios for 2008 (or the closest available year) were multiplied by the emission reduction factors given in Table 1 to project the VOC and NO_x concentrations in 2030. The initial and boundary conditions and other run conditions (e.g., mixing height, temperature, time-dependent photolysis rate constants) (Table 3) were the same from run to run with the exception of the initial VOC and NO_x mixing ratios.

Table 2. VOC species profile used for Box Model simulations.

RACM2 Species	Average (ppb C)	RACM2 Species	Average (ppb C)
Alkanes		Alkynes
ETH	15.0 ± 7.0	ACE	3.2 ± 1.2
HC3	36.7 ± 23.6
HC5	34.1 ± 14.9	Aromatics
HC8	7.6 ± 4.3	BEN	2.2 ± 1.0
		TOL	15.7 ± 18.8
Alkenes		XYM	3.4 ± 2.3
ETE	4.6 ± 1.6	XYP	3.5 ± 2.9
OLT	2.4 ± 1.1	XYO	3.7 ± 2.9
OLI	0.7 ± 0.3
ISO	0.9 ± 0.7	Aldehydes
		HCHO	6.7 ± 4.6
		ACD	9.5 ± 9.8

Notes: The average values were used for the simulations.

Table 3. Initial conditions used for the Box Model simulations.

Initial Condition	Value
Start time	6:00
Duration	18 hr
Temperature	298 K
Pressure	1013.25 mbar
Date for photolysis calculation	June 21
Ozone (initial)	30. ppbv
NO/NOx (initial)	0.95
H2	550

Results and discussion

Long-term changes in the spatial and temporal variations of peak ozone levels in the SoCAB

The trends in the 3-yr running seasonal (June 1 to September 30) mean daily maximum hourly ozone concentrations in the SoCAB from 1982 to 2012 were examined by location and day of week. Figure 2 shows the trends in summer mean daily maximum 8-hr ozone by day of week. The trends for weekends are shown with filled circles for Sundays and open circles for Saturdays. Weekday trends are shown as broken lines. The weekday trends show rapid and steady decline in ozone levels from 1982 to 2000, but a relatively flat trend during the last 12 yr. Weekends show a relatively slow decline during the first 12 yr, followed by 6 yr of rapid decrease from mid-1990s to 2000, and back to a slow-declining trend during the last 12 yr. Weekend-weekday differences were minimal in the early 1980s, increased during the late 1980s in the western half of the SoCAB, and gradually spread further east during the 1990s. Although ozone levels decreased slowly during the past decade at Crestline for both Sunday and weekdays, the weekday ozone levels at the central Basin sites show slightly increasing trends in the past decade. The minima in the ozone trends at the central Basin sites coincided with minima in the trend of the ratio of ROG to NO_x emissions shown in Figure 3. The magnitude of the weekend ozone effect (Sunday–Weekday) was at historic highs during this period and has eased in recent years with increases in the ambient VOC/NO_x ratios (Fujita et al., 2013).

Figure 2. Thirty-year trend in 3-yr running seasonal (June 1 to September 30) mean daily ambient maximum 1-hr ozone by day of week for Los Angeles, Pomona, Upland, and Rubidoux.

Figure 3. Historic and projected trends in the ratios of ROG to NO_x emissions. Also shown are ROG/NO_x ratios corresponding to further incremental NO_x reductions from the 2030 baseline NO_x emissions (N’ = 284 tpd, −61% from 2008) of 0.75 × N’ (213 tpd, −70%), 0.5 × N’ (141 tpd, −81%), and 0.3 × N’ (85 tpd, −88%) while holding ROG emissions at 2030 baseline levels (R’ = 437 tpd, −32% from 2008 base ROG emissions).

The ambient ozone concentration trends are alternatively displayed in Figure 4 as cumulative ozone reduction trends (trend max − annual max)/(trend max − trend min), which show the incremental progress toward the total reduction in ozone over the 30-yr trend period. On weekdays, over 90% of the total reduction of ozone during the past 30 yr was achieved by the late 1990s or at the midpoint of the trend period. Very little further progress was made on weekdays during the past 12 yr and progress even reversed at some central Basin locations. The flattening of the ozone trend occurred despite continued steady reductions in VOC emissions and even greater reductions of NO_x emissions during the past decade. Attention to recent trends in the Basin DV alone mask the relative lack of progress in urban areas of the central SoCAB compared with the less populated far-downwind locations, which currently determine the ozone attainment status of the SoCAB.

Figure 4. Cumulative ozone reduction trend in the SoCAB from 1982 to 2012 of (30-yr max − annual max)/(30-yr max − 30-yr min) in percent for Sundays (top) and weekdays (middle). Ratios are based on 3-yr running averages of the annual means of the daily maximum 1-hr ozone mixing ratios. Bottom panel shows trends in total Basin-wide ROG and NO_x emissions in tons per day.

2008 base year simulations

The base year simulations included 91 days in June through August 2008. There were seven well-defined multiday ozone episodes during summer 2008, with 75 total days having daily Basin-wide maximum concentrations of 80 ppb or higher. The SCAQMD assessed the seven episodes for normalized meteorological ozone episode potential using a regression based weighting covering 30 yr of data (1998–2010) (SCAQMD, 2013). Eight days during the 2008 period were ranked above the 95th percentile in the long-term distribution of potentials, and another 19 were ranked between the 90th and 94th percentiles. Figure S-1 shows the time series of measured and CMAQ-predicted (UTEP output) daily maximum 1-hr ozone for all 91 summer days in 2008 (June 1 to August 31). The model results obtained by UTEP were virtually identical to those obtained by the SCAQMD (Figure S-2). Therefore, we reference the performance evaluations reported by the SCAQMD in Appendix V of the 2012 AQMP (SCAQMD, 2013). Model performance was evaluated by SCAQMD using the unpaired peak ratio, paired mean normalized gross error, and paired mean normalized bias. For the model performance evaluation, the Basin was represented by three zones: Zone 3—the San Fernando Valley; Zone 4—the Eastern San Gabriel, Riverside, and San Bernardino valleys; and Zone 5—the Los Angeles and Orange County emission source areas. Of the three areas, Zone 4 has the maximum ozone concentrations and is the primary downwind impact zone. The CMAQ ozone simulations generally met the 1-hr average unpaired peak and normalized error model performance goal in all three zones on most days. Normalized bias tended to be negative, particularly in June. Zone 5, however, showed a tendency for overprediction in all 3 months. Zone 4 displayed the best unpaired peak performance with 54 out of 58 days, meeting the 20% criteria. Unpaired peak performance in Zones 3 and 5 lagged, with only 76 and 79% of the days meeting the criteria.

2008 simulations with upward adjustments to the base ROG emissions

In addition to the base case simulations, two additional sets of simulations were run for the 2008 base year to simulate the effect of increasing the 2008 base ROG emissions by factors of 1.5 and 2.0. The predicted/observed ratios of the daily maximum 8-hr ozone values are given in Figure 5 for the three alternative 2008 base year simulations. The predicted values are the mean 8-hr ozone concentrations for the nine cells (12 × 12 km) surrounding the SoCAB air quality monitoring stations for 14 simulation days (June 15, 2008, to June 21, 2008, and July 2, 2008 to July 8, 2008). The objective of these series of model simulations was to determine whether upward adjustments to the 2008 base year ROG inventory (factors of 1.5 or 2.0) yield better agreement between observed and predicted ozone values.

Figure 5. Ratios (mean ± standard deviations) of predicted/observed daily maximum 8-hr ozone for June 15 to June 20 and July 2 to July 8 with 1.0 (0.79 ± 0.18), 1.5 (1.00 ± 0.20), and 2.0 (1.17 ± 0.22) time 2008 base ROG. Ratios are based upon ozone values are for the nine cells (12 × 12 km) containing the SoCAB air quality monitoring station.

The base ROG inventory with no adjustment yielded predicted ozone values in good agreement with observations near the western edge of the Basin at Los Angeles (average predicted/observed ratios of 0.95 ± 0.21 for 1-hr ozone and 0.97 ± 0.22 for 8-hr ozone), but underpredicts the observed ozone values at all other sites from the central Basin to the far downwind eastern edge of the Basin (average observed/predicted ratios of 0.77 ± 0.19 for 1-hr ozone and 0.79 ± 0.18 for 8-hr ozone). Upward adjustments by factors of 1.5 and 2.0 results in predicted 8-hr ozone values that are 1.19 ± 0.28 and 1.39 ± 0.35 times higher, respectively, than the observed values at Los Angles. A factor of 1.5 adjustment results in good agreement for the other five sites, with average observed/predicted ratios of 0.98 ± 0.2 for 1-hr ozone and 1.00 ± 0.2 for the 8-hr ozone, and a factor of 2.0 results in overprediction, with ratios of 1.18 ± 0.22 and 1.17 ± 0.22 for 1-hr and 8-hr ozone, respectively. These results combined with the findings of the 2010 Van Nuys Tunnel Study (Fujita et. al., 2012) and the most recent top-down emission inventory evaluation (Fujita et al., 2013) support the conclusion that ROG emissions in the 2008 base inventory are underestimated by about a factor of 1.5.

The indications that current estimates of ROG emissions may be underestimated relative to NO_x emissions have important implications for modeled demonstrations that the planned emission reductions will result in ambient concentrations that meet the NAAQS. Attainment is demonstrated using the relative reduction factors (RRFs), which are the ratios of the model’s future to current (baseline) predictions at monitoring locations. Future ozone concentrations (i.e., projected design values) are estimated by multiplying the modeled RRF at a location near the monitoring site by the baseline design value (DV) (eq 1) of the site as follows:

(1)

DV is the 3-yr average of the annual fourth highest monitored daily 8-hr maximum value at each monitoring site. Attainment is deemed to be demonstrated if the projected future year DV is less than the NAAQS. In regions that are VOC-limited with respect to ozone formation, such as the SoCAB, an underestimation of VOC emissions in the base year will result in lower predicted ozone and higher RRFs. Correcting the underestimation will yield a lower RRF and decrease the emission reduction necessary to demonstrate attainment. In other word, underestimation of base year emissions will result in overestimation of the required emission reductions.

Future simulations, relative response factors, and projected design values

Results for the future year simulations are consistent with our earlier prediction of an interim period of flat or increasing ozone trends in the central and western SoCAB with a shift of the peak ozone levels from the eastern portions of the Basin in the San Bernardino Mountains westward toward the more populated area of the Basin (Fujita et al., 2013). With 2030 baseline NO_x emissions (−61% from 2008, 723 to 284 tpd) and 2030 baseline ROG emissions (−32% from 2008 base, 639 to 437 tpd) (left half of Figure 6), the average daily maximum 8-hr ozone concentrations are predicted to increase by 18, 14, 11 and 10% at Los Angeles, Azusa, Pomona, and Upland, respectively, and decrease by 2 and 4% at Rubidoux and Crestline, respectively. Ozone levels are not significantly different with 70% reduction in NO_x emissions (723 to 213 tpd) except for further marginal deceases in the eastern Basin. At 81% reduction in NO_x (723 to 141 tpd), the average 8-hr ozone concentration at Los Angeles remains about 18% higher than 2008 levels and most of the central Basin remains near 2008 levels. However, ozone levels at Rubidoux and Crestline are predicted to be 13 and 17% lower, respectively, relative to 2008 levels, and the location of maximum ozone shifts to more highly populated areas of the central Basin. With 88% reduction in NO_x from 2008, ozone levels at Azusa, Pomona, Upland, Rubidoux, and Crestline decrease by 8, 12, 13, 24, and 27% from 2008, respectively. However, the projected design values exceed the ozone NAAQS at all sites if the DVs are based upon the relative response factors (RRFs) referenced to the 2008 simulations using the base ROG emissions (left half of Figure 7).

Figure 6. Ozone remaining relative to the 2008 base simulation with base ROG emissions (left half) and base ROG × 1.5 (right half) in percent for future year simulations with 2030 baseline ROG and varying incremental reduction in NO_x emissions. R and N denote 2008 base ROG (639 tpd) and NO_x (723 tpd) emissions and R’ and N’ denote 2030 baseline ROG (437 tpd) and NO_x (284 tpd) emissions.

Figure 7. 2008 observed (baseline) design value and 2030 projected future design values (ppb) based on relative response factors reference to 2008 simulation with base ROG emissions (left half) and to the 2008 simulation with base ROG × 1.5 (right half). R and N denote 2008 base ROG and NO_x emissions and R’ and N’ denote 2030 baseline ROG and NO_x emissions.

The results of underestimating ROG emissions in the 2008 base year simulation are apparent by comparing the left and right halves of Figures 6 and 7. The right halves show the changes in ozone levels relative to 2008 and the projected design values if the future year simulations are referenced to the 2008 base year simulation with 1.5 factor increase in ROG emissions (i.e., ROG reduction in 2030 is −54% rather than −32% from 2008). The resulting lower relative response factors yield about 20% lower projected design values, which are at or below the 2008 8-hr ozone standard of 75 ppb at all sites, except at Crestline, with VOC-only controls and NO_x remaining at 2008 levels (case 1 in right half of Figure 7). Attainment at Crestline would require close to 90% reduction in NO_x from 2008 levels (case 2 in Figure 7) and failure to reach 90% NO_x reduction by even a relatively small amount would leave the central Basin in nonattainment of the ozone NAAQS. Moreover, ozone levels would change very little from 2008 in the central Basin and would be higher in the western Basin even if the target NO_x reductions are achieved.

Historic and future evolution of the ozone-precursor relationship

Isopleths from the box model simulations for daily average ozone daily average ozone, VOC/NO_x ratios, nitric acid, and ozone production efficiency as a function of initial concentrations of NO_x and VOC are given in Figure 8 for Upland-San Bernardino. Although the isopleths depend somewhat on the VOC mixture, they should represent reasonably well the atmospheric chemistry across the SoCAB. Additionally, the observed HNO₃ concentrations would not be as high as those presented on the isopleths due to dry deposition and other losses that are not included in the model, so the concentrations on the isopleths represent the total HNO₃ production over 1 day. The average weekday (solid diamonds) and weekend (open circles) concentrations of VOC and NO_x for the years 1995–2011 are shown on the ozone isopleths in comparison with the ridgeline. Projections for 2030 are also plotted as rectangles on the isopleths where the solid line rectangle represents weekdays and the dash line rectangle represents weekends. The solid and dashed lines indicate the trend lines for the weekday (solid) and weekend (dash) VOC and NO_x concentrations. These isopleths should not be confused with emission control isopleths where the upper right hand corner usually represents current emission conditions and the axes represent varying degrees of NO_x and VOC control. Emission control isopleths depend on location and meteorological conditions and vary significantly across the SoCAB (Milford et al., 1989).

Figure 8. Isopleths for daily average ozone, VOC/NO_x ratios, and nitric acid and ozone production efficiency for Upland-San Bernardino. Plotted on the isopleths are average values of VOC and NO_x for the years 1995–2011. The solid diamonds are weekday averages and the open circles are weekends. Projections for 2030 are plotted as rectangles on the isopleths where the solid-line rectangle represents weekdays and the dashed-line rectangle represents weekends. The trend lines for the VOC and NO_x weekday (solid line) and weekend (dashed line) concentrations are plotted.

The ozone isopleth for Upland-San Bernardino shows that there was little difference in ozone levels in 1995 between weekdays and weekends. In 1995, the concentrations of VOC and NO_x were greater on weekdays than on weekends, but their downward trends of VOC and NO_x have crossed, though within the range of the 2030 projected VOC and NO_x concentrations. Differences between weekend and weekday emissions show the effects of NO_x emission reductions on ozone concentrations. An increase in ozone during weekends indicates that an ozone disbenefit is likely to result from modest NO_x reductions. However, sufficiently large NO_x reductions may result in an ozone benefit following the disbenefit.

The rate of ozone production is affected by the concentrations of VOC and NO_x. If the concentration of NO_x is too high or too low relative to VOC, then ozone formation is inhibited because NO_x is both the catalyst for the production of ozone and it is a major source of termination of the peroxy radicals that lead to the production ozone. This relationship between VOC and NO_x makes the VOC/NO_x ratio an important indicator for the potential to form ozone. Initially, emission controls for VOC decreased the weekday VOC/NO_x ratio from 10 to 6 ppbC/ppbN, and during that same time period, the weekend VOC/NO_x ratio varied from 10 to about 7 ppbC/ppbN. New NO_x controls are expected to increase the VOC/NO_x ratios for both weekdays and weekends, and by 2030, they could return to the value of 1995, 10 ppbC/ppbN, and may even reach 14 ppbC/ppbN or more.

The NO_x chain length is a measure of ozone production efficiency. Chain length is the rate of the major propagation reaction divided by the rate of termination reactions (or the rate of the initiation reactions). An approximation of the NO_x chain length is the ozone production divided by the NO_x converted to NO_y, where NO_y is the sum of the HNO₃, PAN, and organic nitrate. As the atmospheric concentration of NO_x decreases, the ozone production efficiency should increase due to lower rates of termination. Figure 8 shows the isopleth for the ozone production efficiency/chain length with the trend for Upland-San Bernardino. The chain length is highly correlated with the VOC/NO_x ratio at all concentration levels. For Upland-San Bernardino in 1995, there was not much difference in the chain length between weekdays and weekends, both were near 4. However the chain length increased to 6 by 2009. The projected emissions lead to a widening of the difference in chain length by 2030. The weekend chain length is estimated to be about 15 ± 2, and on weekdays it increases to 17 ± 3 at Upland-San Bernardino.

Both Azusa and Upland-San Bernardino have followed a relatively low NO_x pathway. Reductions in initial VOC lead to reductions in the daily average ozone concentrations, although the reductions are lower where the VOC/NO_x ratio is high. The response of ozone to VOC is relatively low for initial NO_x concentrations less than 30 ppb. The points for the future year 2030 are at the lowest VOC and NO_x, but very low VOC emissions are required to reduce ozone to acceptable levels for these low levels of NO_x. The response of the daily average ozone to changes in initial NO_x is more complicated. Reductions in initial NO_x may lead to increases in the daily average ozone. As the initial VOC is decreased, the lower the level of NO_x that is required to reach the point where reductions in NO_x lead to lower daily O₃ concentrations. Both Azusa and Upland-San Bernardino have not followed an optimum path to lower daily average O₃ concentrations.

Conclusions

Attainment of the ozone NAAQS in the SoCAB is complicated by the chemistry-driven fact that ozone production becomes more efficient as the initial NO_x concentrations are reduced. The current NO_x-focused control strategy for the SoCAB is based upon an implicit expectation that with sufficient NO_x reductions, ozone formation will transition from VOC to NO_x-limited ozone formation. As ozone formation is currently VOC-limited in the SoCAB, model simulations with the 2030 baseline emissions (−61% for NO_x and −32% for ROG from 2008) predict 10–20% higher peak ozone levels in the western and central Basin compared with 2008 levels and minimal change in the eastern Basin. With additional NO_x emission reductions of 50% beyond the 2030 baseline (−81% from 2008), ozone levels are predicted to be about 15% lower than 2008 levels in the eastern SoCAB but would remain comparable to 2008 levels in the western and central Basin. The Basin maximum ozone site will shift westward to more populated areas of the Basin and will result potentially in greater population-weighted exposure to ozone with even a relatively small shortfall in the required NO_x reductions unless accompanied by additional VOC reductions beyond 2030 baseline levels.

The current study supports the underestimation of ROG emissions reported in recent works by Fujita et al. (Fujita et al., 2012, 2013) of about 50%. Thus, future VOC reductions could be greater than currently estimated (−54% rather than −34% from 2008 levels in 2030) assuming that the unaccounted emissions are eventually controlled by currently adopted regulations. The higher 2008 baseline ozone predictions with 1.5 factor adjustments to the base year ROG emissions result in lower relative response factors and about 20% lower projected design values. If correct, these findings have important implications for the ozone NAAQS attainment demonstration, the total emission reductions required, and possibly the optimal mix of VOC and NO_x reductions.

The results of this study indicate a prolonged period of higher ozone levels in the western and central SoCAB resulting from a NO_x-focused control strategy. This forecast and potential underestimation of current VOC emissions, therefore, warrants a rigorous analysis of the costs and effects of varying reductions of VOC and NO_x, either alone or in combination, on the formation and combined health impacts of ozone and secondary particles (e.g., nitrates and organic aerosols) and the implications of gradually increasing global background ozone concentrations (Cooper et al., 2014) (e.g., 40–44 ppb_v at Lassen Volcanic National Park in northern California). As a final note, given the evidence that motor vehicle ROG emissions were underestimated in past emission inventories (Fujita et al., 2013), steps should be taken to verify and ensure that the projected reductions of on-road vehicle ROG emissions from the LEV-III emission standards, especially for evaporative emissions, are fully achieved and that the disproportionate contributions of the remaining high-emitting vehicles are minimized.

Funding

This project was funded by the Coordinating Research Council (CRC) and the Truck and Engine Manufacturers Association (EMA). In addition, Stockwell, Fitzgerald, and Saunders were partially supported by the National Oceanic and Atmospheric Administration, Educational Partnership Program, U.S. Department of Commerce, under Agreement No. NA11SEC4810003. The contents of this paper are solely the responsibility of the authors and do not necessarily represent the views of the CRC, EMA, and NOAA.

Acknowledgment

The authors gratefully acknowledge the South Coast Air Quality Management District for providing the CMAQ modeling inputs and output files.

Supplemental Material

Supplemental data for this article can be accessed on the publisher’s website.

Additional information

Funding

This project was funded by the Coordinating Research Council (CRC) and the Truck and Engine Manufacturers Association (EMA). In addition, Stockwell, Fitzgerald, and Saunders were partially supported by the National Oceanic and Atmospheric Administration, Educational Partnership Program, U.S. Department of Commerce, under Agreement No. NA11SEC4810003. The contents of this paper are solely the responsibility of the authors and do not necessarily represent the views of the CRC, EMA, and NOAA.

Notes on contributors

Eric M. Fujita

Eric M. Fujita is a research professor emeritus and David E. Campbell is an associate research scientist in the Division of Atmospheric Sciences at the Desert Research Institute (Nevada System of Higher Education).

William R. Stockwell

William R. Stockwell is a professor in the Department of Chemistry and the Atmospheric Science Program of Howard University.

Emily Saunders

Emily Saunders is a doctoral student in chemistry at Howard University.

Rosa Fitzgerald

Rosa Fitzgerald is a professor in the Department of Physics of the University of Texas El Paso (UTEP) and Roberto Perea is a Ph.D. student at UTEP.