Responses in Ozone and Its Production Efficiency Attributable to

Subscriber access provided by READING UNIV

Article

Responses in Ozone and its Production Efficiency Attributable to Recent and Future Emissions Changes in the Eastern United States Lucas Henneman, Cong Liu, Huizhong Shen, Yongtao Hu, James A Mulholland, and Armistead G Russell Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.7b04109 • Publication Date (Web): 07 Nov 2017 Downloaded from http://pubs.acs.org on November 7, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 22

Environmental Science & Technology

1

RESPONSES IN OZONE AND ITS PRODUCTION EFFICIENCY

2

ATTRIBUTABLE TO RECENT AND FUTURE EMISSIONS CHANGES IN

3

THE EASTERN UNITED STATES

4 5 6

Lucas RF Henneman* Harvard University T.H. Chan School of Public Health Boston, MA, 02445, USA

7 8 9

Huizhong Shen School of Civil and Environmental Engineering, Georgia Institute of Technology Atlanta, GA, 30332, USA

10 11 12

Cong Liu School of Energy and Environment, Southeast University, Nanjing, China

13 14 15 16 17 18

Yongtao Hu, James A Mulholland, and Armistead G Russell School of Civil and Environmental Engineering, Georgia Institute of Technology Atlanta, GA, 30332, USA

19 20

ACS Paragon Plus Environment


21 22

Abstract Ozone production efficiency (OPE), a measure of the number of ozone (O3) molecules

23

produced per emitted NOX (NO + NO2) molecule, helps establish the relationship between NOX

24

emissions and O3 formation. We estimate long-term OPE variability across the eastern United

25

States using two novel approaches: an observation-based empirical method and a chemical

26

transport model (CTM) method. The CTM approach explicitly controls for differing O3 and

27

NOX reaction products (NOZ) deposition rates, and separately estimates OPEs from on-road

28

mobile and electricity generating unit sources across a broad spatial scale. We find lower OPEs

29

in urban areas and that average July OPE increased over the eastern United States domain

30

between 2001 and 2011 from 11 to 14. CTM and empirical approaches agree at low

31

NOZ concentrations, but CTM OPEs are greater than empirical OPEs at high NOZ. Our results

32

support that NOX emissions reductions become more effective at reducing O3 at lower

33

NOZ concentrations. Electricity generating unit OPEs are higher than mobile OPEs except near

34

emissions locations, meaning further utility NOX emissions reductions will have greater per-unit

35

impacts on O3 regionally.

36

Introduction

37

Atmospheric scientists have understood the underlying mechanisms of ozone (O3)

38

production for decades; however, air quality managers continue to have trouble meeting ozone

39

standards, particularly as they have been tightened in response to health studies identifying

40

impacts at lower concentrations1,2. The availability of detailed ambient monitoring data and

41

much expanded modeling capabilities presents an opportunity for a detailed assessment of the

42

effectiveness of past and future air quality improvement efforts. An assessment of this type—

43

investigating changing factors attributable to air quality controls—serves to link accountability

44

assessments (i.e., efforts to quantify the impacts of existing regulatory programs) to proposed air

45

quality programs.

46

Largely as a result of air quality management policies, anthropogenic emissions of two

47

contributors to ozone concentrations—nitrogen oxides (NOX = NO + NO2) and volatile organic

48

compounds (VOCs)—have decreased throughout the eastern United States over the past two

49

decades, particularly in urban areas3,4. Emissions reductions, particularly in electrical generating

50

unit (EGU) and mobile source sectors, have led to reduced levels of ambient NOX, NOX reaction

51

products (NOZ = total atmospheric oxidized reactive nitrogen (NOY) minus NOX), and VOCs5,6.


Page 2 of 22

Page 3 of 22


52

These reductions have been linked to changing O3 concentrations—both reductions (primarily in

53

the summer) and increases (primarily in the winter and in VOC-limited urban areas)1. Multiple

54

studies have found that meteorology has a major impact on daily O3 concentrations5,7,8, and

55

climate variability and synoptic patterns impact longer-term trends9,10.

56

Ozone production efficiency (OPE) has historically been a popular method for reporting

57

the relationship between O3 and NOX. OPE is defined as “the number of molecules of O3 formed

58

per each NOX molecule removed from the system”11. NOX generates O3 as it cycles between NO

59

and NO2, and since each NOX molecule that enters the system must leave, the emissions rate

60

approximates the loss rate across a large scale. The reaction of NO2 with OH—which forms

61

nitric acid (HNO3)—is typically the dominant pathway for NOX loss, though it can combine with

62

organics or form HNO3 via N2O5 hydrolysis (which occurs primarily at night)12. OPE has

63

important policy implications, since it provides an estimate for how much less ozone would be

64

formed for a specified reduction in NOX emissions13,14.

65

Previous studies using both observation-based and chemical transport model (CTM)-

66

based approaches have found increasing OPEs as NOX or NOZ decreases13,15. This increase

67

carries the implication that NOX emissions reductions become increasingly effective. There has,

68

however, been little investigation of OPE at very low NOZ concentrations. We set out to fulfil

69

two objectives: 1) assess multiple observation and CTM-based approaches for estimating total

70

and source-specific OPEs in the eastern United States, and 2) project impacts of continuing NOX

71

emissions reductions on future OPEs and O3 concentrations.

72

Methods

73

Observation-based OPE

74

We use observations from the SouthEastern Aerosol Characterization (SEARCH)

75

network sampling sites16,17, which provide meteorological data, O3, NOX, and NOY species

76

observations at eight locations from 1996-2015 (not all sites were online for the entire period).

77

The eight locations comprise four pairs: three urban-rural and one urban-suburban. Throughout

78

this manuscript, we refer to the sites by their three-letter acronym and a superscript that denotes

79

if the site is in an urban (U), rural (R) or suburban (S) location. The southeastern United States is

80

characterized by hot, humid summers with periods of stagnation interrupted by large

81

thunderstorms. Winters are typically mild and drier, though extended periods of wet weather



82

accompany frontal activity. Isolated city centers and power plants, situated throughout mixed

83

coniferous and deciduous forests and agricultural land, serve as major air pollution sources16.

84

Previous studies have used the relationship between O3 and NOZ as an empirical estimate

85

of OPE; in a single plume, the loss of NOX equals the production of NOZ when the deposition

86

rate of NOX is negligible18. Kleinman et al. (2002) noted that a major limitation to using a fixed-

87

site monitor to estimate OPE is that air masses crossing a monitor have potentially very different

88

histories15. Take, for instance, the case of a monitor near a city; at times, it would be downwind

89

of the city; O3 (and NOZ) levels would be high as the city contributes fresh emissions that lead to

90

high ozone levels, but the NOZ would have had little time to deposit. At other times, the monitor

91

would be upwind, the air masses impacting the monitor would have lower ozone, and much of

92

the NOZ will have been lost to deposition. Relating measured O3 to NOZ in this case is not

93

indicative of the OPE.

94

We controlled for air parcel history by stratifying days based on their photochemical

95

activity using an ensemble of observations across years. Days with high photochemical activity

96

have similar meteorological characteristics, meaning changes in O3 levels should relate to

97

changes in NOZ, though other emissions changes can also play a role. We used long-term

98

meteorological detrending of ozone observations at SEARCH sites to identify the 20% of days

99

with all the relevant data available at each site that were most photochemically active.

100

Photochemical state (PS*, described in detail in Henneman et al. (2017)19) was estimated using

101

the meteorological detrending method presented in Henneman et al. (2015)7. In brief, observed

102

maximum daily average 8-hr (MDA8h) O3 was split into five trends with different periods: long-

103

term (period > 365 days), seasonal (365 days), week-holiday (7-365 days), short-term

104

meteorological (< 365 days), and white-noise (1 day). The long-term and seasonal components

105

were separated using the low-pass moving average KZ filter20. We regressed the remaining

106

fluctuations against similarly filtered daily fluctuations in various metrics: weekday and holiday

107

factors along with 0, 1, and 2-day lags each of solar radiation (total and daily max), temperature

108

(mean and daily maximum), wind speed (morning and daily means), relative humidity, rainfall.

109

We calculate the short-term meteorological trend as the fit of meteorological covariates in this

110

regression. PS* is the sum of the seasonal and short-term meteorological trends (Figure S1), and

111

represents an emissions-independent estimate of daily atmospheric photochemical activity. This


Page 4 of 22

Page 5 of 22


112

metric is positively correlated with temperature and wind speed and negatively correlated with

113

relative humidity.

114

For comparison, we calculated an alternative to the PS* metric (HNO3/NOY), which was

115

found to be likely impacted by long-term emissions changes (SI text and Figures S2 and S3);

116

therefore, PS* was used as the final metric for photochemical plume age. Since PS* was

117

calculated using MDA8h O3, it is indicative of 8-hr averaged photochemical states. We restrict

118

the observation-based OPE analysis to days in the upper 20th percentile for PS*, and use 2-3 pm

119

averages (local time) for daily metrics of all observed species. We investigated the distribution of

120

PS* days and found little evidence that decreasing emissions across the study period impacted

121

PS* levels or their likelihood of meeting the top 20% criteria (Figures S4 and S5). We used only

122

data from the middle afternoon in the primary analysis to approximate the time of maximum

123

photochemical activity21,22. NOZ was calculated as the difference between NOY and NOX.

124

We tested multiple models for the O3-NOZ relationship, including linear and log models.

125

Models of these types, however, are limited in their application to very low NOZ values, where

126

the slope ( ⁄ —estimated with a spline-based regression model) increases and data is

127

sparse. Spline models separate the domain of the independent variable into sections, and fit

128

curves to each section23. The sections are divided at ‘knots’, and fitted curves in each of the

129

sections are required to be continuous at the knots. We applied cubic splines, which extend the

130

constraints to include equal first and second derivatives at each knot (Figure 1). We take the

131

marginal OPE—i.e., the additional amount of ozone formed due to an additional amount of

132

NOZ—as the derivative of fitted O3 vs. NOZ models at each site. The spline model captures slope

133

changes at multiple points and does not result in nonphysically high slopes at low NOZ (which

134

were found using a log-based regression model). Slopes (i.e., OPEs) are constant outside the

135

outermost knots. We employed the the rcs and ols commands from the rms package in R version

136

3.3.3 for the spline modeling24.



137 138 139 140 141 142

Figure 1. Empirical OPE fits using natural spline with 3 knots. Urban sites are on the left, and rural/suburban sites are on the right; each urban/rural (or suburban) pair is matched right to left. Points are colored by the slope of the calculated curve at that NOZ level (i.e., their OPE). Models are fit for days with highest 20% PS*. Chemical Transport Model-based OPE

143

We used the Community Multiscale Air Quality (CMAQ) model with the direct

144

decoupled method (DDM)25 to simulate air quality and first-order sensitivities to emissions over

145

a 12 km domain covering the eastern United States (Figure 2). The model setup and application

146

was first described in Henneman et al. (2017) and evaluated in detail in Henneman et al.

147

(2017b)19,26. We used results from the 2001 and 2011 simulations described in our previous work

148

along with two additional month-long cases with July 2011 meteorology and domain-wide NOX

149

emissions reduced by 50% and 90%. The latter simulations serve to control for meteorology and

150

all other emissions sources completely while reducing NOX emissions. We preserve consistency

151

with observations by using 2-3 p.m. averages, and focus on July because it is a summer month

152

with typical conditions suitable for O3 formation.


Page 6 of 22

Page 7 of 22


153 154 155 156 157 158 159

Figure 2. CMAQ modeling domain and SEARCH sites colored by spline fit intercept (i.e., regional background O3). Acronyms: Jefferson Street in Atlanta, Georgia (JST); Yorkville, Georgia (YRK); Birmingham, Alabama (BHM); Centerville, Alabama (CTR), Gulfport, Mississippi (GFP); Oak Grove, Mississippi (OAK); Pensacola, Florida (PNS); and Pensacola outlying aircraft landing field, Florida (OLF). Superscripts denote urban (U), rural (R), and suburban (S) sites.

160

For a first estimate of CTM-simulated OPE, we used a brute force (BF) method that

161 162

assumes linear OPEs across changing NOZ concentrations similar to previous approaches21,27: ∆

∆


Equation 1


Page 8 of 22

163

Where NOZ is the sum of NOX reaction products—peroxyacetyl nitrate (denoted PAN and

164

PANX in the default CMAQ CB05 chemical mechanism setup28), peroxynitric acid (PNA),

165

organic nitrate (NTR), nitric acid (HNO3), nitrous acid (HONO), nitrate radical (NO3), aerosol

166

nitrate in three modes (ANO3I, ANO3J, ANO3K), and dinitrogen pentoxide (N2O5)—as N (e.g.,

167

N2O5 contains two N molecules). The changes in O3 and NOZ (∆) are matched by site and date

168

across CMAQ simulations. We produced two OPEBF estimates per day at each grid cell in the

169

domain: one each for the differences (2011 − 2011% ) and (2011% −

170

2011% ).

171 172

We compare with a second estimate using the ratio of O3 and NOZ sensitivities to on-road mobile (MOB) and electricity generating unit (EGU) source emissions:

173

!"#$

% ,'%( )% ,*+,

Equation 2

-% ,'%( )-% ,*+,

174

Where each sensitivity term (S) is in ppm. This equation estimates the ratio of the number of O3

175

molecules produced per NOZ molecule attributable to emissions from two sources. We focus on

176

EGU and mobile sources because they make up the majority of anthropogenic NOX emissions.

177

Equation 2, however, has one important omission; it does not account for O3 and NOZ produced

178

that deposits out (indeed, none of the empirical or chemical transport model methods described

179

to this point explicitly account for deposition). We rectify this by adding source-specific

180

deposition terms:

181

/*0 /*0 .% ,'%( )% 1).% ,*+, )% 1 ,'%( ,*+,

Equation 3

/*0 /*0 .-% ,'%( )-% ,'%( 1).-% ,*+, )-% ,*+, 1

"#$

182

Where 23,4 is the sensitivity of species i deposition to emissions from source j (again, all

183

sensitivity terms are in ppm). CMAQ-DDM calculates the total (wet and dry) O3 deposition (e.g.,

184

5 ∗ for O3) and the sensitivities to total deposition from mobile (289 ) and EGU (289 ) ,7 ,:;

185

sources. Deposition sensitivities, however, are in units of mass deposited per area; therefore, they

186

require a conversion to concentration units. We achieve this for each source and species by

187

) by the total deposition (e.g., dividing the source-specific deposition sensitivity (e.g., 289 ,7

188

5 ∗ ) and multiplying by the concentration. For example:

189

∗

∗

∗

289 ,7

∗

/*0 % ,'%(

89%∗

∗


Equation 4

Page 9 of 22

190


The CMAQ deposition-corrected sensitivities method presents a further opportunity—

191

estimating source-specific OPEs—achieved by calculating ratios of mobile and EGU

192

sensitivities, respectively: 7

193

:;

194

/*0 .% ,'%( )% 1 ,'%(

/*0 .-% ,'%( )-% ,'%( 1

Equation 5

/*0 .% ,*+, )% 1 ,*+,

/*0 .-% ,*+, )-% ,*+, 1

Equation 6

195

Where Equations 2-3 and 5-6 produce daily OPE estimates at each grid cell for each CMAQ

196

simulation.

197

Approach limitations

198

Various assumptions are inherent in the methods discussed in this study. The empirical

199

method effectively uses an Eulerian framework, i.e., stationary measurements of plumes with

200

different histories and degrees of photochemical activity. The approach is subject to varying

201

background pollutant levels, nighttime nitrate chemistry, air parcel transport, varying NOZ and

202

O3 deposition rates, and clouds14. A Lagrangian approach, i.e., when observations are made

203

while traveling with a specific air mass, does not suffer from effects from air parcels with

204

varying histories, and has been undertaken using aircraft14,15,29–31. The Lagrangian approach

205

typically has led to lower OPEs than those found using Eulerian approaches, although this may

206

be due to higher NOZ concentrations in plumes, a more limited time to react, or because the

207

typical approach is to correlate O3 with NOZ instead of taking the nonlinearity of the system into

208

account18,30,32,33.

209

The empirical approach is further limited because it does not control for changing

210

emissions of other species involved in ozone formation (e.g., VOCs and CO) or long term

211

changes in the meteorology leading to high ozone days (e.g., due to climate change). Sensitivity

212

studies find that O3 is primarily sensitive to NOX in the Southeast on high ozone days34,35, though

213

results in downtown areas suggest negative sensitivity to local NOX emissions and VOC

214

sensitivity even on high ozone days1,35,36. Urban NO emissions inhibit O3 production, albeit

215

primarily on days with low photochemical activity11.

216 217

Recent work has shown that a portion of NOX in the Southeast may react with organics and hydrolyze to form organic nitrogen aerosol. Fisher et al. (2016) found that 21% of NOX was



218

lost to RONO2, and that this fraction increases with decreasing NOX emissions37. 59% of RONO2

219

was hydrolyzed to aerosol in the southeastern US in August-September 2013. This portion of

220

NOZ is not present in NOY measurements, and would lead to increasing bias in the empirical

221

OPE model at low NOZ concentrations.

222

A limitation of the CMAQ approach is the use of concentrations, deposition, and

223

sensitivities to anthropogenic emissions that are found at the same location/cell (an Eulerian

224

frame), whereas each of these would be impacted by transport and deposition between the

225

NOX source location and the location of interest. Total O3 and NOZ deposition fields are spatially

226

similar to, but smoother than, their respective concentration fields—with the exception of

227

localized NOZ deposition hot-spots associated with wet deposition of aerosol NO3—suggesting

228

this issue is of limited importance when using sensitivity-based approaches, particularly if

229

deposition is accounted for in the OPE calculations (Figures S6 and S7 show modeled O3

230

concentrations and deposition).

231

Results and Discussion

232

Observation-based OPE

233

We compared mean 2-3 p.m. O3 against mean 2-3 p.m. NOZ for days with PS* greater

234

than the 80th percentile. The resulting relationships have positive slopes and negative second

235

derivatives, corroborating results from previous studies38,39 (Figure 1, Figure S8 for results

236

obtained selecting a wider PS* range). NOX and NOZ levels decreased across the study period,

237

driven by reductions in NOX emissions5,19 (Figure S9). OPE results from linear and log models

238

are presented in the supporting information (Figure S10; further, Blanchard et al. (2017)

239

investigate OPE at SEARCH sites using linear models and other methods22); here we focus on

240

the cubic spline model results.

241

Empirical model intercepts at each of the sites represent an estimate of local background

242

concentrations, i.e., expected concentrations when NOZ levels are quite small due to low local

243

NOX emissions and/or nearly complete NOZ deposition. Local background ozone results from

244

global and regional emission sources—not local sources—although it can be generated nearby.

245

For example, urban sources may affect surrounding background ozone, resulting in elevated rural

246

background ozone nearby and increased titration in cities further downwind. These intercepts

247

likely represent an upper bound of the background for two reasons. First, the spline model’s

248

restriction to linearity outside the outermost knots is rather stringent, and may miss small slope


Page 10 of 22

Page 11 of 22


249

increases at the lowest NOZ concentrations. Second, differences between O3 and NOZ deposition

250

are magnified at low concentrations; accounting for this would shift the points plotted in Figure

251

S1 rightward, thus creating a lower intercept.

252

The estimated level that O3 levels would reach in the absence of local NOX emissions

253

varies for each site with meteorological conditions and long-term climate and emissions changes.

254

Using a Monte Carlo sampling technique, we fit multiple spline models at each site using

255

intercepts selected from the normal distribution around the mean estimated in the original model

256

fit, and calculated an uncertainty distribution of the slopes at each NOZ concentration (Figure 3

257

and S11). This method was chosen because of the high uncertainty in background O3 and OPE at

258

low NOZ concentrations and the difficulty in assessing individual model parameter error in spline

259

models. Uncertainty in OPE is larger at sites closer to the Gulf of Mexico.

260 261 262 263 264 265

Figure 3. OPE at each SEARCH site estimated by the empirical spline and CMAQ depositioncorrected sensitivities approaches. Empirical values represent all days with PS* > 80th percentile, and CMAQ values are days in July. Grey shading is the 95% confidence interval around the spline model. All O3-NOZ spline model intercepts fall between 30.2 (GFPU) and 47.2 ppb (OAKR)

266

(Figure 2). Urban sites tend to have lower intercepts than rural sites, and the two sites closest to

267

the Gulf of Mexico (GFPU & PNSU) have the lowest intercepts (i.e., local background O3) with



268

the greatest uncertainty. CTRR, in Alabama, stands out as the rural site with the lowest intercept.

269

Background estimates align with the upper end of previous estimates of the levels that ozone

270

would reach in the absence of NOx emissions in the Southeast. Emery et al. (2012) found very

271

few days on which policy-relevant background—i.e., absent North American emissions—

272

MDA8h O3 exceeded 40 ppb in 200640, and Lefohn et al. (2014) found estimates of background

273

levels to be less than about 30 ppb in Atlanta41. These previous estimates used air quality models

274

to estimate background O3; the present approach uses observed data to estimate regional

275

background levels, which are expected to be somewhat more variable31. Elevated intercepts at

276

rural sites (e.g., YRKR and OAKR) suggest impacts from nearby urban centers, which

277

corroborates previous findings by Blanchard et al. (2013, 2014)34,42, though the lack of urban-

278

rural difference between BHMU and CTRR again suggests some level of uncertainty.

279

Empirical OPEs at JSTU and YRKR, two sites in Georgia, have similar shapes (Figure 3),

280

and OPEs at the rural site are slightly higher than OPEs at the urban site. This trend repeats at

281

BHMU and CTRR, with the rural site (CTRR) exhibiting greater OPEs across all NOZ levels. The

282

remaining sites (GFPU, OAKR, PNSU, OLFS) have OPEs that differ from the urban-rural

283

relationships at the four inland sites. Limited data availability for these sites contributes to larger

284

uncertainty bounds. GFPU, OLFS and PNSU are near the Gulf of Mexico; the air from the Gulf

285

causes OPEs at these sites to increase dramatically at low NOZ because ozone is efficiently

286

transported over the water with relatively little deposition, while nitric acid is very soluble and

287

reacts with sea salt, leading to larger particles that readily deposit. Neuman et al. (2009) suggest

288

high variability in regional background depending on wind direction in Houston; these results

289

reflect large uncertainties in coastal sites compared to inland found in the present study31. OLFS,

290

a suburban site somewhat inland from PNSU, has a higher OPE at elevated NOZ levels and lower

291

OPE at low NOZ levels than PNSU. The two sites nearest the Gulf of Mexico (GFPU & PNSU)

292

have the lowest intercepts and the highest OPEs at low NOZ.

293

Chemical Transport Model-based OPE

294

For two July 2011 comparisons—i.e., between the change from base emissions to 50%

295

NOX emissions and from 50% NOX emissions to 10% NOX emissions—correlation analysis of

296

OPEnodep versus OPEBF finds a slope near one and R2 of 0.39 and 0.27, respectively (Figure S12).

297

The comparison controls for meteorology, but neither controls for deposition. A potential source

298

of bias in OPEnodep is the exclusion of higher order sensitivities; Cohan et al. (2005) showed that


Page 12 of 22

Page 13 of 22


299

second order effects could contribute ~30% bias. Still, the slopes near one support the

300

sensitivity-based approach.

301

Without correcting for deposition, over-land average July 2001 and 2011 OPEs were 10.9

302

and 13.8, similar to the with-deposition values (Table 1). Remaining discussions focus on

303

deposition-corrected OPEs (Equations 3, 5, and 6); all summary statistics are reported for the

304

portion of the domain encompassed by the superimposed polygon in Figure 4 to exclude

305

boundary effects and impacts of high OPEs over water.

306 307 308

Table 1. Information on the modeling runs. Emissions are summed over the entire domain, while concentrations and OPEs are averaged over the polygon superimposed on the inland region of each of the plots in Figure 4 to exclude boundary effects and high OPEs over water. July NOX Emissions, 103 tons ALL

MOB

EGU

Mean July Concentrations, ppb (std. dev.) O3

NOZ

Mean July OPE (std. dev.)

NOX

OPE

OPEnodep

OPEMOB

OPEEGU

2001

1,364

773

432

54.7 (4.7)

2.7 (0.9)

1.6 (2.7)

11.2 (2.9)

10.9 (3.1)

10.7 (3.0)

12.7 (3.1)

2011

679

364

169

48.1 (5.5)

1.9 (0.6)

1.1 (1.3)

14.0 (3.3)

13.8 (3.3)

13.8 (3.5)

15.3 (3.6)

2011 50% NOx

339

182

85

39.5 (4.7)

1.1 (0.4)

0.5 (0.6)

15.7 (2.3)

15.7 (2.4)

2011 10% NOx

68

36

17

28.6 (3.6)

0.3 (0.1)

0.1 (0.1)

16.4 (1.7)

16.4 (1.7)

309

310 311 312

A) 2001 OPE

B) 2001 OPEMOB

C) 2001 OPEEGU

D) 2011 OPE

E) 2011 OPEMOB

F) 2011 OPEEGU

Figure 4. CMAQ-modeled mean daily July deposition-corrected OPE (2-3 p.m.) values for 2001 (A-C) and 2011 (D-F), including total OPE (A&D), mobile OPE (B&E), and EGU OPE (C&F).



313

CMAQ simulations yielded consistently higher OPEs in rural areas than urban areas, and

314

generally decreasing OPE with increasing latitude (Figure 4). Between 2001 and 2011, average

315

over-land OPEs increased from 11.2 to 14.0 (Table 1), while OPEs at SEARCH locations

316

increased from 9.6 to 12.7 (Figure S13, Table S1). Between 2001 and 2011, estimated domain-

317

wide NOX emissions decreased 50% and mean overland OPE increased 25%. A hypothetical

318

50% decrease in NOX emissions would lead to only an additional 15% increase in OPE, and a

319

further 80% emissions cut (90% total reduction from base 2011) would increase OPE just 4%

320

more (Figure S14). Average July overland OPEs do not exceed 25, even at the most extreme

321

(90%) NOX emission reduction (Figure S15). Even on July days with the 2nd-highest simulated

322

O3 concentrations in each location, OPE does not exceed 27. These results show that OPE is

323

expected to increase at a decreasing rate with further emission reductions, and corroborate the

324

empirical results that show that OPE increases to location-specific upper limits at lower NOX.

325

Mobile and EGU OPEs differ spatially. Mobile OPEs are slightly lower than the total

326

OPEs in 2001 and 2011 (Table 1). Their spatial pattern generally mirrors that of total OPE, with

327

lower values near concentrated mobile emissions in urban centers and along interstates (Figure

328

4). EGU OPEs are generally greater than the total OPEs, with low spots centered near individual

329

plants.

330

CMAQ OPE results agree with empirical OPEs at low NOZ concentrations for most sites,

331

although CMAQ OPE estimates at rural sites are slightly higher (Figure 3). Note that in this

332

comparison, empirical results use days across the entire period with PS* > 80th percentile, and

333

CMAQ values use only July days in 2001 and 2011. Further, the three coastal sites (GFPU, OLFS

334

and PNSU) are not located within the overland domain used for analysis (Table 1), and direct

335

comparisons between the methods are inherently problematic because both are models with

336

uncertainties and bias.

337

CMAQ-modeled domain-wide average O3 concentration in the 10% NOX scenario, at

338

28.6 ppb, is lower than the background O3 concentrations estimated by the empirical approach.

339

While the comparison here is not direct, it does again suggest that the empirical methods

340

estimate a regional (instead of a global) background that is still influenced by non-local

341

emissions.


Page 14 of 22

Page 15 of 22

342 343


Implications The empirical and CMAQ-based methods applied here were designed to reduce biases

344

associated with the limitations discussed above. In the empirical method, the use of

345

meteorologically similar days with high photochemical activity (PS*) across multi-year datasets

346

limits (but likely does not remove) effects of plumes with markedly different origins. We

347

assessed OPE uncertainty based on a distribution of statistically modeled background O3 levels,

348

and found the largest impacts at coastal sites. Using CMAQ-DDM, we controlled for air parcel

349

transport and differing O3 and NOZ deposition rates by applying source-specific concentration

350

and deposition sensitivity ratios instead of just concentrations.

351

Rising OPEs with decreasing NOZ concentrations suggests a NOX-limited regime, with

352

increasing effectiveness for reducing O3 concentrations for each subsequent avoided NOX

353

molecule emitted. We show this for the empirical models by integrating under the OPE curves in

354

Figure 3. For a 0.5 ppb NOZ concentration decrease, the expected change in O3 concentrations

355

(∆O3) increases with decreasing starting NOZ concentrations for all sites (Figure 5). That ∆O3

356

reaches a maximum at the lowest NOZ concentrations is somewhat a function of the choice of

357

spline model; however, CMAQ results provide evidence for a maximum OPE (and subsequently

358

a maximum ∆O3 in response to emissions decreases). 2011 OPEs across the domain and at all

359

SEARCH sites increased at a decreasing rate with NOX emissions reductions (Table S1, Figure

360

S13); at CTRR, GFPU, and OAKR, the reduction from 50% to 10% NOX emissions led to slight

361

decreases in OPE, suggesting a plateauing OPE.



362 363 364

Figure 5. Change in O3 expected from a 0.5 ppb reduction in NOZ from observed NOZ at each site for days in July.

365

We tested the agreement between the empirical spline model and CMAQ-modeled

366

changes in O3 for a 50% NOX emissions reduction at the SEARCH sites (Figure S16). For

367

relative reductions (as opposed to absolute reductions addressed above), both models find

368

decreasing effectiveness at lower starting NOZ concentrations (since relative reductions at lower

369

total emissions refer to increasingly small mass-based emissions). The models generally agree

370

for all sites. The comparison assumes that NOX emissions reductions directly correlate with NOZ

371

concentration reductions, which is a close (though not perfect) approximation; NOZ

372

concentrations are linear with reductions using consistent meteorology, but are nonlinear

373

between 2001 and 2011 due to changing meteorological conditions (Table 1, Figure S14).

374

CMAQ results yield strong OPE differences between rural and urban areas, and a

375

lesser—though still significant—trend with decreasing latitude (Figures S17 and S18). We found

376

negative relationships between OPE and both NOZ and formaldehyde (HCHO—a reaction

377

product of volatile organics). We found a positive relationship between OPE and the organic

378

fraction of NOZ (fN,org), which explains variability consistent with change across latitude (see

379

discussion in the supporting information).

380

Empirical and CMAQ results provide evidence that maximum OPEs are site-specific,

381

reflecting site-specific meteorologies and emissions of species other than NOX. At the lowest


Page 16 of 22

Page 17 of 22


382

NOZ concentrations, the potential for reducing O3 concentrations through NOX emissions

383

reductions levels off; however, O3 concentrations have a positive relationship with NOZ at all

384

NOZ concentrations. This suggests that continuing emissions controls will reduce O3

385

concentrations at increasing rates, but the effect will level off at location-specific NOZ

386

concentrations.

387

EGU OPEs are on average greater than mobile OPEs across the domain. This suggests

388

greater expected O3 reductions per absolute NOX emissions reduction from EGUs than mobile.

389

However, this comparison does not consider the amount of emissions left to reduce—mobile

390

sources emitted more than 2X NOX than EGU sources in 2011. Since modelled 2011 EGU OPE

391

is 10% larger than mobile OPE (15.3 vs. 13.8), more O3 reduction is available through

392

proportional mobile emission reductions than EGU.

393

This study shows that future NOX emissions reductions remain a viable option for

394

reducing the highest O3 concentrations. OPE increases with decreasing NOZ concentrations, but

395

appears to level out or reach a maximum, which implies that incremental NOX emissions

396

reductions will become more effective at reducing O3 up to a point, after which the incremental

397

benefit will level off. Empirical results based on ambient observations suggest that maximum

398

OPE has already been attained in rural areas, and results obtained using the chemical transport

399

modeling approach corroborate this finding.

400

Acknowledgements

401

This material is based upon work supported by Health Effects Institute and the National

402

Science Foundation Graduate Research Fellowship under Grant No. DGE-1148903. This

403

publication was developed under Assistance Agreement No. 83588001 awarded by the U.S.

404

Environmental Protection Agency to Georgia Institute of Technology. It has not been formally

405

reviewed by EPA. The views expressed in this document are solely those of the authors and do

406

not necessarily reflect those of the Agency. EPA does not endorse any products or commercial

407

services mentioned in this publication. We are grateful for discussions with Charlie Blanchard

408

regarding measurements made at SEARCH network sites.

409

Supporting Information Available

410

Supplemental analyses, 18 figures and two tables.



411

References

412

(1)

Simon, H.; Reff, A.; Wells, B.; Xing, J.; Frank, N. Ozone trends across the United States

413

over a period of decreasing NOx and VOC emissions. Environ. Sci. Technol. 2014, 49 (1),

414

186–195.

415

(2)

416 417

U.S. EPA. National Ambient Air Quality Standards for Ozone Final Rule; 2015; Vol. 80, pp 1–7.

(3)

Vijayaraghavan, K.; DenBleyker, A.; Ma, L.; Lindhjem, C.; Yarwood, G. Trends in on-

418

road vehicle emissions and ambient air quality in Atlanta, Georgia, USA, from the late

419

1990s through 2009. J. Air Waste Manage. Assoc. 2014, 64 (7), 808–816.

420

(4)

Xing, J.; Pleim, J.; Mathur, R.; Pouliot, G.; Hogrefe, C.; Gan, C. M.; Wei, C. Historical

421

gaseous and primary aerosol emissions in the United States from 1990 to 2010. Atmos.

422

Chem. Phys. 2013, 13 (15), 7531–7549.

423

(5)

424 425

Hidy, G. M.; Blanchard, C. L. Precursor reductions and ground-level ozone in the Continental United States. J. Air Waste Manag. Assoc. 2015, 65 (10), 1261–1282.

(6)

He, H.; Stehr, J. W.; Hains, J. C.; Krask, D. J.; Doddridge, B. G.; Vinnikov, K. Y.; Canty,

426

T. P.; Hosley, K. M.; Salawitch, R. J.; Worden, H. M.; et al. Trends in emissions and

427

concentrations of air pollutants in the lower troposphere in the Baltimore/Washington

428

airshed from 1997 to 2011. Atmos. Chem. Phys. 2013, 13 (15), 7859–7874.

429

(7)

Henneman, L. R.; Holmes, H. A.; Mulholland, J. A.; Russell, A. G. Meteorological

430

Detrending of Primary and Secondary Pollutant Concentrations: Method Application and

431

Evaluation Using Long-Term (2000-2012) Data in Atlanta. Atmos. Environ. 2015, 119,

432

201–210.

433

(8)

434 435

Camalier, L.; Cox, W.; Dolwick, P. The effects of meteorology on ozone in urban areas and their use in assessing ozone trends. Atmos. Environ. 2007, 41 (33), 7127–7137.

(9)

Shen, L.; Mickley, L. J.; Tai, A. P. K. Influence of synoptic patterns on surface ozone

436

variability over the eastern United States from 1980 to 2012. Atmos. Chem. Phys. 2015, 15

437

(19), 10925–10938.

438

(10)

439 440 441

Shen, L.; Mickley, L. J. Seasonal prediction of US summertime ozone using statistical analysis of large scale climate patterns. PNAS 2017, 114 (10), 2491–2496.

(11)

Seinfeld, J. H.; Pandis, S. N. Atmospheric Chemistry and Physics: From Air Pollution to Climate Change; 2006; Vol. 2nd.


Page 18 of 22

Page 19 of 22

442


(12)

443 444

Russell, A. G.; Cass, G. R.; Selnfeld, J. H. On Some Aspects of Nighttime Atmospheric Chemistry. Environ. Sci. Technol. 1986, 20 (11), 1167–1172.

(13)

Liu, S. C.; Trainer, M.; Fehsenfeld, F. C.; Parrish, D. D.; Williams, E. J.; Fahey, D. W.;

445

Hobler, G.; Murphy, P. C. Ozone Production in the Rural Troposphere and the

446

Implications for Regional and Global Ozone Distributions. J. Geophys. Res. 1987, 92

447

(D4), 4191–4207.

448

(14)

Ryerson, T. B.; Buhr, M. P.; Frost, G. J.; Goldan, P. D.; Holloway, J. S.; Hübler, G.;

449

Jobson, B. T.; Kuster, W. C.; McKeen, S. a.; Parrish, D. D.; et al. Emissions lifetimes and

450

ozone formation in power plant plumes. J. Geophys. Res. 1998, 103 (D17), 22569.

451

(15)

Kleinman, L. I.; Daum, P. H.; Lee, Y. N.; Nunnermacker, L. J.; Springston, S. R.;

452

Weinstein-Lloyd, J.; Rudolph, J. Ozone production efficiency in an urban area. J.

453

Geophys. Res. 2002, 107 (D23), 4733.

454

(16)

Hansen, D. A.; Edgerton, E. S.; Hartsell, B. E.; Jansen, J. J.; Kandasamy, N.; Hidy, G. M.;

455

Blanchard, C. L. The Southeastern Aerosol Research and Characterization Study: Part 1—

456

Overview. J. Air Waste Manage. Assoc. 2003, 53 (12), 1460–1471.

457

(17)

Hansen, D. A.; Edgerton, E.; Hartsell, B.; Jansen, J.; Burge, H.; Koutrakis, P.; Rogers, C.;

458

Suh, H.; Chow, J.; Zielinska, B.; et al. Air quality measurements for the aerosol research

459

and inhalation epidemiology study. J. Air Waste Manag. Assoc. 2006, 56 (10), 1445–

460

1458.

461

(18)

Thornton, J. A.; Wooldridge, P. J.; Cohen, R. C.; Martinez, M.; Harder, H.; Brune, W. H.;

462

Williams, E. J.; Roberts, J. M.; Fehsenfeld, F. C.; Hall, S. R.; et al. Ozone production rates

463

as a function of NOx abundances and HOx production rates in the Nashville urban plume.

464

J. Geophys. Res. 2002, 107 (D12), 1–17.

465

(19)

Henneman, L. R.; Chang, H. H.; Lavoue, D.; Mulholland, J. A.; Russell, A. G.

466

Accountability assessment of regulatory impacts on ozone and PM2.5 concentrations

467

using statistical and deterministic pollutant sensitivities. Air Qual. Atmos. Heal. 2017, 10

468

(6), 695–711.

469

(20)

470 471 472

Zurbenko, I. G. Spectral analysis of nonstationary time series. Int. Stat. Rev. 1991, 39, 163–173.

(21)

Trainer, M.; Parrish, D. D.; Buhr, M. P.; Norton, R. B.; Fehsenfeld, F. C.; Anlauf, K. G.; Bottenheim, J. W.; Tang, Y. Z.; Wiebe, H. A.; Roberts, J. M.; et al. Correlation of ozone



473

with NOy in photochemically aged air. J. Geophys. Res. Atmos. 1993, 98 (D2), 2917–

474

2925.

475

(22)

476 477

Southeastern United States. Atmos. Chem. Phys. Discuss. 2017, in review, 1–33. (23)

478 479

Blanchard, C. L.; Hidy, G. M. Ozone Response to Emission Reductions in the

Hastie, T.; Tibshirani, R.; Friedman, J.; Franklin, J. The elements of statistical learning: data mining, inference and prediction, 2nd ed.; Springer: Stanford, CA, 2009.

(24)

R Development Core Team. R: A Language and Environment for Statistical Computing. R

480

Foundation for Statistical Computing Vienna Austria. Vienna, Austria 2012, p {ISBN} 3-

481

900051-07-0.

482

(25)

Byun, D.; Schere, K. L. Review of the Governing Equations, Computational Algorithms,

483

and Other Components of the Models-3 Community Multiscale Air Quality (CMAQ)

484

Modeling System. Appl. Mech. Rev. 2006, 59 (2), 51.

485

(26)

Henneman, L. R.; Liu, C.; Hu, Y.; Mulholland, J. A.; Russell, A. G. Air quality modeling

486

for accountability research: Operational, dynamic, and diagnostic evaluation. Atmos.

487

Environ. 2017, 166 (2017), 551–565.

488

(27)

Godowitch, J. M.; Gilliam, R. C.; Rao, S. T. Diagnostic evaluation of ozone production

489

and horizontal transport in a regional photochemical air quality modeling system. Atmos.

490

Environ. 2011, 45 (24), 3977–3987.

491

(28)

Whitten, G. Z.; Heo, G.; Kimura, Y.; McDonald-Buller, E.; Allen, D. T.; Carter, W. P. L.;

492

Yarwood, G. A new condensed toluene mechanism for Carbon Bond: CB05-TU. Atmos.

493

Environ. 2010, 44 (40), 5346–5355.

494

(29)

Sillman, S. Ozone production efficiency and loss of NOx in power plant plumes:

495

Photochemical model and interpretations of measurements in Tennessee. Jounal of

496

Geophysical Research. 2000, pp 9189–9202.

497

(30)

Imhoff, R. E.; Valente, R.; Meagher, J. F.; Luria, M. The production of O-3 in an urban

498

plume: Airborne sampling of the Atlanta urban plume. Atmos. Environ. 1995, 29 (17),

499

2349–2358.

500

(31)

Neuman, J. a.; Nowak, J. B.; Zheng, W.; Flocke, F.; Ryerson, T. B.; Trainer, M.;

501

Holloway, J. S.; Parrish, D. D.; Frost, G. J.; Peischl, J.; et al. Relationship between

502

photochemical ozone production and NOx oxidation in Houston, Texas. J. Geophys. Res.

503

Atmos. 2009, 114 (10), 1–12.


Page 20 of 22

Page 21 of 22

504


(32)

Trainer, M.; Ridley, B. a; Buhr, M. P.; Kok, G.; Walega, J.; Hubler, G.; Parrish, D. D.;

505

Fehsenfeld, F. C. Regional Ozone and Urban Plumes in the Southeastern United States -

506

Birmingham, a Case Study. J. Geophys. Res. 1995, 100 (D9), 18823–18834.

507

(33)

Ridley, B. A.; Walega, J. G.; Lamarque, J.-F.; Grahek, F. E.; Trainer, M.; Hubler, G.; Lin,

508

X.; Fehsenfeld, F. C. Measurements of reactive nitrogen and ozone to 5-km altitude in

509

June 1990 over the southeastern United States. J. Geophys. Res. 1998, 103 (D7), 8369–

510

8388.

511

(34)

Blanchard, C. L.; Hidy, G. M.; Tanenbaum, S. Ozone in the southeastern United States:

512

An observation-based model using measurements from the SEARCH network. Atmos.

513

Environ. 2014, 88, 192–200.

514

(35)

Mazzuca, G. M.; Ren, X.; Loughner, C. P.; Estes, M.; Crawford, J. H.; Pickering, K. E.;

515

Weinheimer, A. J.; Dickerson, R. R. Ozone production and its sensitivity to NOx and

516

VOCs: Results from the DISCOVER-AQ field experiment, Houston 2013. Atmos. Chem.

517

Phys. 2016, 16 (22), 14463–14474.

518

(36)

Blanchard, C. L.; Hidy, G. M.; Tanenbaum, S. NMOC, ozone, and organic aerosol in the

519

southeastern United States, 1999–2007: 2. Ozone trends and sensitivity to NMOC

520

emissions in Atlanta, Georgia. Atmos. Environ. 2010, 44 (38), 4840–4849.

521

(37)

Fisher, J. A.; Jacob, D. J.; Travis, K. R.; Kim, P. S.; Marais, E. A.; Miller, C. C.; Yu, K.;

522

Zhu, L.; Yantosca, R. M.; Sulprizio, M. P.; et al. Organic nitrate chemistry and its

523

implications for nitrogen budgets in an isoprene- and monoterpene-rich atmosphere:

524

Constraints from aircraft (SEAC4RS) and ground-based (SOAS) observations in the

525

Southeast US. Atmos. Chem. Phys. 2016, 16 (9), 5969–5991.

526

(38)

Daum, P. H.; Kleinman, L. I.; Newman, L.; Luke, W. T.; Weinstein-Lloyd, J.; Berkowitz,

527

C. M.; Busness, K. M. Chemical and physical properties of plumes of anthropogenic

528

pollutants transported over the North Atlantic during the North Atlantic Regional

529

Experiment. J. Geophys. Res. 1996, 101 (D22), 29029.

530

(39)

531 532

Lin, X.; Trainer, M.; Liu, S. C. On the nonlinearity of the tropospheric ozone production. J. Geophys. Res. 1988, 93 (D12), 15879.

(40)

Emery, C.; Jung, J.; Downey, N.; Johnson, J.; Jimenez, M.; Yarwood, G.; Morris, R.

533

Regional and global modeling estimates of policy relevant background ozone over the

534

United States. Atmos. Environ. 2012, 47 (2012), 206–217.



535

(41)

Lefohn, A. S.; Emery, C.; Shadwick, D.; Wernli, H.; Jung, J.; Oltmans, S. J. Estimates of

536

background surface ozone concentrations in the United States based on model-derived

537

source apportionment. Atmos. Environ. 2014, 84, 275–288.

538

(42)

Blanchard, C. L.; Hidy, G. M.; Tanenbaum, S.; Edgerton, E. S.; Hartsell, B. E. The

539

Southeastern Aerosol Research and Characterization (SEARCH) study: Spatial variations

540

and chemical climatology, 1999–2010. J. Air Waste Manage. Assoc. 2013, 63 (3), 260–

541

275.

542


Page 22 of 22

Responses in Ozone and Its Production Efficiency Attributable to

Recommend Documents