Trends in ozone, its precursors, and related secondary oxidation products in Los Angeles, California: A synthesis of measurements from 1960 to 2010

Authors

  • Ilana B. Pollack,

    Corresponding author
    1. Cooperative Institute for Research in Environmental Sciences, University of Colorado Boulder, Boulder, Colorado, USA
    2. Chemical Sciences Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA
    • Corresponding author: I. B. Pollack, Chemical Sciences Division, Earth System Research Laboratory, National Oceanic and Atmospheric Administration, 325 Broadway, MS R/CSD7, Boulder, CO 80305, USA. (ilana.pollack@noaa.gov)

    Search for more papers by this author
  • Thomas B. Ryerson,

    1. Chemical Sciences Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA
    Search for more papers by this author
  • Michael Trainer,

    1. Chemical Sciences Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA
    Search for more papers by this author
  • J. A. Neuman,

    1. Cooperative Institute for Research in Environmental Sciences, University of Colorado Boulder, Boulder, Colorado, USA
    2. Chemical Sciences Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA
    Search for more papers by this author
  • James M. Roberts,

    1. Chemical Sciences Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA
    Search for more papers by this author
  • David D. Parrish

    1. Chemical Sciences Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA
    Search for more papers by this author

Abstract

[1] Decreases in ozone (O3) observed in California's South Coast Air Basin (SoCAB) over the past five decades have resulted from decreases in local emissions of its precursors, nitrogen oxides (NOx = NO + NO2) and volatile organic compounds (VOCs). Ozone precursors have been characterized in the SoCAB with measurements dating back to 1960. Here we compile an extensive historical data set using measurements in the SoCAB between 1960 and 2010. Faster rates of decrease have occurred in abundances of VOCs (−7.3 ± 0.7% yr−1) than in NOx (−2.6 ± 0.3% yr−1), which have resulted in a decrease in VOC/NOx ratio (−4.8 ± 0.9% yr−1) over time. Trends in the NOx oxidation products peroxyacetyl nitrate (PAN) and nitric acid (HNO3), measured in the SoCAB since 1973, show changes in ozone production chemistry resulting from changes in precursor emissions. Decreases in abundances of PAN (−9.3 ± 1.1% yr−1) and HNO3 (−3.0 ± 0.8% yr−1) reflect trends in VOC and NOx precursors. Enhancement ratios of O3 to (PAN + HNO3) show no detectable trend in ozone production efficiency, while a positive trend in the oxidized fraction of total reactive nitrogen (+2.2 ± 0.5% yr−1) suggests that atmospheric oxidation rates of NOx have increased over time as a result of the emissions changes. Changes in NOx oxidation pathways have increasingly favored production of HNO3, a radical termination product associated with quenching the ozone formation cycle.

1 Introduction

[2] Ozone concentrations have decreased significantly since the 1960s in the California South Coast Air Basin (SoCAB), a region encompassing the Los Angeles urban area. Maximum 8 h average ozone mixing ratios measured basin wide and at individual surface monitoring network stations in the SoCAB decreased by about a factor of 3 (Figure 1; http://www.arb.ca.gov/adam/index.html) between 1973 and 2010. The observed decrease in ozone is attributed to decreases in local emissions of volatile organic compounds (VOCs) [Warneke et al., 2012] and nitrogen oxides (NOx = NO + NO2) [McDonald et al., 2012], the precursors to ozone formation. Reactions of hydrocarbons and carbon monoxide (CO) with hydroxyl radicals (OH), as shown in () and (), initiate ozone formation chemistry [Finlayson-Pitts and Pitts, 2000; Jacob, 1999] through formation of peroxy radicals. Oxidation of NO by HO2 or RO2 via ()–() then generates nitrogen dioxide (NO2) and produces ozone following photolysis () and reaction with molecular oxygen ().

display math(R1)
display math(R2)
display math(R3)
display math(R4)
display math(R5)
display math(R6)
display math(R7)
Figure 1.

Maximum 8 h average ozone mixing ratios from selected Air Quality Monitoring District (AQMD) monitoring network stations in the SoCAB. Each data point represents the maximum 8 h average for a 1 year period. LLS fit of the maximum 8 h average for all sites in the SoCAB (solid gray line) indicates a decrease in ozone by a factor of 2.9 between 1973 and 2010, corresponding to a rate of decrease of 2.8 ± 0.2% yr−1.

[3] Ozone precursors have been extensively studied over the years in the SoCAB. Some of the earliest measurements of NOx, CO, and speciated VOCs in downtown Los Angeles date back to 1960 [Neligan, 1962], after emissions from automobiles and local industry were identified as major contributors to photochemical smog [Haagen-Smit, 1952; Haagen-Smit and Fox, 1954]. These measurements initiated the development of an extensive surface monitoring network in the state of California and motivated future intensive field measurement campaigns. Since 1960, several short-term ground-based field studies have been conducted at selected locations within the SoCAB. Basin-wide measurements from instrumented research aircraft began in the 1970s [Husar et al., 1977], and near-tailpipe measurements from mobile roadside monitors began in the early 1990s [Beaton et al., 1995; Bishop and Stedman, 2008; Gertler et al., 1999; Lawson et al., 1990]. Long-term trends in ozone and emissions of its precursors in the SoCAB have been extensively studied using the data collected in these experiments [Ban-Weiss et al., 2008; Bishop and Stedman, 2008; Dallmann and Harley, 2010; Fortin et al., 2005; Fujita et al., 2003; Fujita et al., 2013; Grosjean, 2003; Harley et al., 2005; McDonald et al., 2012; Parrish et al., 2002; Parrish et al., 2011; Warneke et al., 2012].

[4] Other secondary pollutants such as nitric acid (HNO3), alkyl nitrates (RONO2), peroxides (H2O2 and ROOH), and peroxyacetyl nitrate (PAN; CH3C(O)O2NO2), formed in reactions accompanying those that produce ozone, have also been measured in the SoCAB. Formation of HOx-[BOND]NOx oxidation products, such as nitrates via () and () and peroxides via () and (), effectively removes these radicals from ozone-producing reaction cycles. In contrast, formation of VOC-[BOND]NOx oxidation products, such as PAN via (), produces temporary reservoir species for HOx and NOx due to relatively short thermal decomposition lifetimes for the peroxyacyl nitrates under surface conditions characteristic of summertime in Los Angeles [Roberts et al., 2007; Roberts et al., 1995; Stephens, 1969; Grosjean et al., 2001; Grosjean, 2003; Tuazon et al., 1991]. These reactions propagate the radical chain and lead to continued ozone production.

display math(R8)
display math(R9)
display math(R10)
display math(R11)
display math(R12)

[5] The interdependence of these odd-hydrogen sinks with the chain reactions that produce ozone, reactions ()–(), makes the oxidation products of reactions ()–() a useful tool for attributing a cause and effect relationship between decreasing ozone and decreasing precursor abundances [Kelly, 1992; Roberts et al., 2007; Roberts et al., 1995; Sillman et al., 1990; Sillman, 1991; Trainer et al., 1993]. Even though secondary oxidation products have been measured and studied in SoCAB field experiments since the 1960s and 1970s, relatively few peer-reviewed publications analyze these data over multiple years. These few include reviews of hydrogen peroxide (H2O2) measurements [Lee et al., 2000; Sakugawa et al., 1990], with several measurements in the California SoCAB between 1970 and 1988, and reports of ambient PAN measurements at various sites in the SoCAB between 1960 and 1997 by Grosjean [2003] and between 1975 and 1983 by Temple and Taylor [1983]. Historical analyses of secondary oxidation products other than ozone in the SoCAB have been largely neglected, despite their key role as indicators for understanding the response of ozone production rates and yields following changes in precursor emissions.

[6] In this work, we confirm and extend reported long-term trends in abundances and emission ratios of ozone precursors over the five decades from 1960 to 2010 in the SoCAB using data from surface monitoring network stations, mobile roadside monitors, ground-based field campaigns, and instrumented research aircraft. Abundances and emission ratios of ozone precursors determined from these measurements are also compared to those derived from emission inventories. We further report long-term trends in secondary oxidation product concentrations and extend the measurements to 2010. Measured abundances and enhancement ratios for secondary oxidation products are compared with values predicted by a chemical box model [Fujita et al., 2013]. These long-term trends are described as exponential decreases [e.g., Parrish et al., 2002; Warneke et al., 2012] and quantified by calculating average rates of change per year. We use this approach to define trends in measured precursors and secondary pollutants. Correlations of ozone and related oxidation products with precursor abundances and VOC/NOx ratio are identified using these historical data. Decadal changes in ozone production efficiency and rate of photochemical processing are interpreted from long-term trends in ozone and NOx oxidation products.

2 Methods

2.1 Data Sets

[7] Data for this analysis are compiled from measurements performed in the SoCAB between 1960 and 2010. Measurements are acquired from (1) the South Coast Air Quality Monitoring District (SCAQMD) and Photochemical Assessment Monitoring Stations (PAMS) surface network, (2) mobile roadside monitors, (3) ground-based field studies, and (4) chemically instrumented research aircraft. Figure 2 illustrates the locations of the network monitoring sites and ground-based field study locations as well as example flight tracks from the airborne experiments used in this analysis.

Figure 2.

Measurements for this analysis are compiled from surface network sites (white circles), roadside monitors (cyan squares), ground-based field sites (yellow triangles), and airborne studies (sample flight tracks in pink). Airborne measurements collected in the mixed boundary layer (<1 km) over the SoCAB (dashed box) are interpreted as basin-wide averages.

2.1.1 Surface Network Stations

[8] Trace gas data since the 1970s were acquired from the web-based Air Quality Data Statistics database maintained by the California Air Resources Board (CARB) (http://www.arb.ca.gov/adam/index.html) for select monitoring stations in the SoCAB network. These data include annual 8 h maximum ozone concentrations; hourly measurements of NO, NO2, and CO; and 3 h canister samples of volatile organic compounds (VOCs). We focus on data collected from the Azusa and Upland sites due to their long-term data coverage and proximity to ground-based field study locations; data coverage and measurement techniques at these stations are summarized in Table 1. We use the available NO2 data from these ground sites despite a known sensitivity of the NO2 measurements to organic nitrates and HNO3, which depends upon inlet configuration and thermal operation range of a molybdenum converter [Fehsenfeld et al., 1990; Fitz, 2002; Murphy et al., 2007; Winer et al., 1974]. NOx mixing ratios are calculated as the sum of the reported NO and NO2 measurements. Since precursors in the SoCAB primarily arise from motor vehicle emissions, we focus our analysis on VOC species characteristic of automobile exhaust, including benzene, toluene, ethylbenzene, and o-xylene. Measurements of isoprene are used to assess potential changes in biogenic emissions over time. Although network measurements of VOCs (also called non-methane organic compounds (NMOCs) in the data sets) date as far back as 1990 [Warneke et al., 2012], we only utilize quality-reviewed data since 1994 for this analysis. Network measurements of VOCs below the instrumental limit of detection are reported by CARB as one-half the limit of detection. Measurements of VOCs are converted from units of parts per billion of carbon (ppbC) to parts per billion by volume (ppbv) of compound prior to further analysis.

Table 1. Summary of Trace Gas Data From the Azusa and Upland Network Stations in the SoCAB
MeasurementSiteData CoverageTechniquea
  1. a

    Additional information about measurement techniques and instrumentation are available on the CARB website (http://www.arb.ca.gov/airwebmanual/index.php).

  2. b

    VOC measurements (non-methane organic carbon (NMOC) in the original data files) represent 3 h samples collected on a varying schedule during a 3 month period from July through September. Years represent data coverage for benzene, toluene, ethylbenzene, and o-xylene; measurements of isoprene are only available at Azusa and Upland between 1994 and 2008.

  3. c

    NO2 measurements since 1980 at Azusa and 1975 at Upland include a known sensitivity to organic nitrates and HNO3, which depends upon inlet configuration and thermal operation range of the molybdenum converter [Fehsenfeld et al., 1990; Fitz, 2002; Murphy et al., 2007; Winer et al., 1974]; NOx is calculated as the sum of the NO and NO2 measurements.

VOCbAzusa1974–1975, 1995–2010Canister collection followed by gas chromatography
Upland1975, 1994–2008Canister collection followed by gas chromatography
O3Azusa1978–2010Ultraviolet absorption
Upland1975–2010Ultraviolet absorption
COAzusa1975–2010Nondispersive infrared absorption
Upland1975–2010Nondispersive infrared absorption
NOAzusa1975–1979Colorimetry
Azusa1980–2010Ozone-induced chemiluminescence
Upland1975–2010Ozone- induced chemiluminescence
NO2cAzusa1975–1979Colorimetry using Lyshkow-modified Saltzman reagent
Azusa1980–2010Heated molybdenum converter followed by chemiluminescence
Upland1975–2010Heated molybdenum converter followed by chemiluminescence

2.1.2 Roadside Monitors

[9] Ozone precursor emissions from light-duty and heavy-duty vehicles measured by mobile roadside monitors in California are available from peer-reviewed publications [Bishop and Stedman, 2008; Bishop et al., 2010; Bishop et al., 2012; Stedman et al., 2009] and research reports accessible from the Fuel Efficiency Automobile Test (FEAT) website (http://www.feat.biochem.du.edu). We use measurements of NO, CO, and total unspeciated hydrocarbons (total HC) including methane from light-duty vehicles acquired at the Los Angeles, Van Nuys, and Riverside monitoring sites since 1991 [Bishop and Stedman, 2008; Bishop et al., 2010]; measurement periods and locations are summarized in Table 2. For this analysis, we assume that NOx is primarily emitted as NO [Soltic and Weilenmann, 2003] and that there is negligible chemical processing between emission and detection given the sampling configuration of the instrument [Burgard et al., 2006]; roadside measurements of NO are thus referred to as NOx in the following analysis. Molar emission ratios of CO, NOx, and (total HC) to CO2 are derived from the reported fuel-specific emissions of grams of pollutant per kilogram of fuel [Bishop and Stedman, 2008; Burgard et al., 2006; Singer et al., 1998].

Table 2. Summary of Roadside Measurements in the SoCAB
YearDatesLocationReference
199119 May to 27 JunLos AngelesBeaton et al. [1995]
199915 Oct to 19 OctLos AngelesBishop and Stedman [2008]
28 Jun to 7 JulRiverside 
200030 May to 6 JunRiversideBishop and Stedman [2008]
200115 Oct to 19 OctLos AngelesBishop and Stedman [2008]
6 Jun to 13 JunRiverside 
200327 Oct to 31 OctLos AngelesBishop and Stedman [2008]
200517 Oct to 21 OctLos AngelesBishop and Stedman [2008]
200817 Mar to 21 MarLos AngelesBishop et al. [2010]; Stedman et al. [2009]
201012 Aug to 16 AugVan NuysBishop et al. [2012]

2.1.3 Ground-Based and Airborne Field Studies

[10] The SoCAB has been the focus of many ground-based and airborne air quality field campaigns. Compound-specific VOCs, NOx, and CO and secondary oxidation products such as O3, PAN, and HNO3 have been measured during these studies. The species measured, measurement techniques, data coverage, sampling locations, and references for field studies performed between 1960 and 2010 are summarized in Table 3.

Table 3. Summary of Measurements From Airborne and Ground-Based Field Studies in the SoCAB
YearDatesLocationMeasurementTechniqueReferences
  1. a

    Nitrogen Species Methods Comparison Study (NSMCS) 1985.

  2. b

    Carbonaceous Species Methods Comparison Study (CSMCS) 1986.

  3. c

    Southern California Air Quality Study (SCAQS) 1987.

  4. d

    Los Angeles Atmospheric Free Radical Study (LAAFRS) 1993.

  5. e

    Southern California Ozone Study (SCOS) 1997.

  6. f

    Intercontinental Transport and Chemical Transformation (ITCT) 2002.

  7. g

    CARB phase of Arctic Research of the Composition of the Troposphere from Aircraft and Satellites (ARCTAS-CARB) 2008.

  8. h

    California Research at the Nexus of Air Quality and Climate Change (CalNex) 2010.

  9. i

    Data reported in the cited literature.

  10. j

    Average values reported in the literature reference.

  11. k

    Data requested from CARB.

  12. l

    Data acquired from NOAA website (http://www.esrl.noaa.gov/csd/field.html).

  13. m

    Data acquired from NASA website (http://www-air.larc.nasa.gov/missions.htm).

196018 Aug to 18 NovLos AngelesNOx, COiNearby monitoring stationNeligan [1962]
Benzene, tolueneiGas chromatography 
PANjPolyethylene bag collection followed by long-path Fourier transform infrared spectroscopyRenzetti and Bryan [1961]
1962N/ARiversidePANjElectron capture-gas chromatographyDarley et al. [1963]
1963–196520 Apr to 6 NovLos AngelesBenzene, toluenejGas chromatographyLeonard et al. [1976]
19661 Sep to 30 NovPasadena/NOx, COjNearby monitoring stationLonneman et al. [1968]
Los AngelesToluene, ethylbenzene, o-xylenejTedlar bag collection followed by gas chromatography 
196719 Sep to 17 NovLos Angeles/AzusaToluene, ethylbenzene, o-xylenejTedlar bag collection followed by gas chromatographyAltshuller et al. [1971]
19685 Sep to 13 NovLos AngelesNOxjChromium trioxide paper and Griess-Saltzman reagentKopczynski et al. [1972]; Lonneman et al. [1976]
Toluene, ethylbenzene, o-xylene, COjTedlar bag collection followed by gas chromatography 
PANjGas chromatography-electron capture detectionLonneman et al. [1976]
19707, 10 AugRiversideH2O2jTeflon bag collection followed by colorimetry with Ti (IV)Bufalini et al. [1972]
197123 Aug to 14 OctLos AngelesBenzene, toluenejGas chromatographyLeonard et al. [1976]
19732 Jul to 30 AugLos AngelesBenzene, toluenejGas chromatographyLeonard et al. [1976]
197324 Jul to 26 JulPasadenaPANiLong-path Fourier transform infrared spectroscopyHanst et al. [1975]
197324 Aug to 28 SepWest CovinaO3, NO, NOxiChemiluminescenceSpicer [1977a]; Spicer [1977b]
PANiGas chromatography-electron capture detection 
HNO3iAcid-detecting mast colorimetry 
197719 Jul to 29 JulClaremont/RiversideH2O2jScrubbing coil impinger followed by luminol-based chemiluminescenceKok et al. [1978]
19789 Oct to 13 OctClaremontNO, NO2iChemiluminescenceTuazon et al. [1981]
O3, PAN, HNO3iLong-path Fourier transform infrared spectroscopy 
19799 Apr to 21 AprLos AngelesBenzene, toluene, ethylbenzene, o-xylenejCanister/Gas Chromatography-flame ionization detectorSingh et al. [1981]
PANjGas chromatography-electron capture detection 
198026 Jun to 27 JunLos AngelesCO, NO, NO2, NOx, PAN, HNO3, O3iLong-path Fourier transform infrared spectroscopyHanst et al. [1982]
198231 AugPasadena/RiversidePANjGas chromatography-electron capture detectionRussell et al. [1988a]
1985a11 Sep to 18 SepClaremontO3UV absorption (from nearby monitoring station) 
NO, NO2iChemiluminescenceGrosjean [1988]
PANiGas chromatography-electron capture detection 
HNO3iFilter pack followed by ion chromatography;Winer et al. [1986]
Fourier transform infrared spectroscopy
1986b12 Aug to 21 AugGlendoraHNO3jFourier transform infrared spectroscopy;Anlauf et al. [1991]
Tunable diode laser absorption spectroscopy; 
Filter pack followed by ion chromatography 
H2O2jDual coil impinger followed by dual enzyme fluorescence;Kok et al. [1990]
Impinger with diffusion scrubber followed by dual enzyme fluorescence;Tanner and Shen [1990]
Cryogenic trap followed by dual enzyme fluorescence;Sakugawa and Kaplan [1990]
Tunable diode laser absorption spectroscopyMackay et al. [1990]
1987c19 June to 3 SepClaremontO3, NOx, NOykChemiluminescenceFujita et al. [1992]; Lawson [1990]; Williams and Grosjean [1990]
PANkGas chromatography-electron capture detectionWilliams and Grosjean [1990]
HNO3kTunable diode laser absorption spectroscopy 
CO, benzene, toluene, o-xylene, isoprenekCanister/Gas Chromatography-flame ionization detector 
H2O2jTunable diode laser absorption spectroscopyMackay et al. [1988]
19903 Aug to 5 SepTanbark FlatPANjGas chromatography-electron capture detectionGrosjean et al. [1993]
19915 Aug to 12 SepTanbark FlatPANjGas chromatography-electron capture detectionGrosjean et al. [1993]
1993d1 Sep to 26 SepClaremontO3kUV absorptionMackay [1994]
NO2, NOx, NOykChemiluminescence 
CO, benzenekCanister and DNPH cartridge followed by gas chromatography-flame ionization detector 
PANkGas chromatography/chemiluminescence 
HNO3, H2O2kTunable diode laser absorption spectroscopy 
1997e16 Jun to 15 OctAzusaO3kUV absorptionBlumenthal [1999]; Croes and Fujita [2003]; Fitz [1999]; Fujita et al. [1999]; Grosjean [2003]; Grosjean et al. [2001]
NOykChemiluminescence 
CO, CO2, benzene, toluene, ethylbenzene, o-xylenekCanister and DNPH cartridge followed by gas chromatography-flame ionization detection 
PANkGas chromatography-electron capture detection 
HNO3kTunable diode laser absorption spectroscopy 
1997e16 Jun to 15 OctPiper Aztec aircraftO3kUV absorptionCroes and Fujita [2003]; Fitz [1999]; Fujita et al. [1999]; Grosjean [2003]; Grosjean et al. [2001]
NOykChemiluminescence 
CO, CO2, benzene, toluene, o-xylenekCanister and DNPH cartridge followed by gas chromatography-flame ionization detection 
2002f13 MayNOAA P-3 aircraftO3, NOx, NOylChemiluminescenceParrish et al. [2004]
COlVacuum UV resonance fluorescence 
CO2lNondispersive infrared absorption 
PAN, HNO3lChemical ionization mass spectrometry 
benzene, toluene, o-xylenelWhole air sampler followed by gas chromatography 
2008g13 MayNASA DC-8 aircraftO3, NOx, NOymChemiluminescenceJacob et al. [2010]
COmVacuum UV resonance fluorescence 
CO2mNondispersive infrared absorption 
PAN, HNO3mChemical ionization mass spectrometry 
benzene, toluenemProton transfer reaction mass spectrometry 
2010h1 May to 30 JunNOAA P-3 aircraftO3, NOx, NOylChemiluminescenceRyerson et al. [2013]
COlVacuum UV resonance fluorescence 
CO2lWavelength-scanned cavity ring-down spectroscopy 
PAN, HNO3lChemical ionization mass spectrometry 
benzene, toluene, ethylbenzene, o-xylenelWhole air sampler followed by gas chromatography 
2010h15 May to 15 JunPasadenaO3lUV absorptionRyerson et al. [2013]
NOx, NOylChemiluminescence 
COlVacuum UV resonance fluorescence 
CO2lNondispersive infrared absorption 
PAN, HNO3lChemical ionization mass spectrometry 
benzene, toluene, ethylbenzene, o-xylenelGas chromatography-mass spectrometry 

[11] Our historical analysis of precursor emissions begins with ground-based measurements of VOCs and NOx in downtown Los Angeles in 1960 by Neligan [1962]. This study was followed with additional measurements by the California Air Pollution Control District in 1963 [Leonard et al., 1976], Lonneman et al. [1968] in 1966, Altshuller et al. [1971] in 1967, and Kopczynski et al. [1972] in 1968. Data from these studies were acquired from tables and text within the cited literature.

[12] More extensive ground-based field studies supported by CARB and SCAQMD began in 1987 with the Southern California Air Quality Study (SCAQS) and continued with the Los Angeles Atmospheric Free Radical Study (LAAFRS) in 1993 and the Southern California Ozone Study (SCOS) in 1997. SCOS 1997 incorporated measurements from several instrumented aircraft, although we use measurements only from the Piper Aztec aircraft due to its primary sampling objective to characterize ozone and precursors in the SoCAB [Croes and Fujita, 2003]. Archived data from field projects supported by CARB were acquired directly from CARB. Peer-reviewed publications outlining objectives, measured species, measurement techniques, sampling details, and results are available for most field studies; research reports are available for the remainder on the CARB website (http://www.arb.ca.gov/research/research.htm).

[13] Airborne measurements are also compiled from several field campaigns performed in the SoCAB in the past decade. A chemically instrumented NOAA P-3 aircraft sampled the daytime mixed layer throughout the Los Angeles Basin on 13 May 2002 during the Intercontinental Transport and Chemical Transformation (ITCT) study [Parrish et al., 2004], and the NASA DC-8 aircraft sampled the same general area on 18 June 2008 during the California phase of the Arctic Research of the Composition of the Troposphere from Aircraft and Satellites study (ARCTAS-CARB) [Jacob et al., 2010]. Most recently, the California Research at the Nexus of Air Quality and Climate Change (CalNex) field project in May, June, and July 2010 reported chemical measurements in the Los Angeles mixed layer from the NOAA P-3 aircraft and from a ground site in Pasadena, California [Ryerson et al., 2013]. Archived data for the field studies conducted by NOAA and NASA are publicly available on each institution's website (http://www.esrl.noaa.gov/csd/field.html; http://www-air.larc.nasa.gov/missions.htm).

[14] Airborne and ground-based measurements of secondary oxidation products, such as PAN and HNO3, are available from SCAQS 1987, SCOS 1997, ITCT 2002, ARCTAS-CARB 2008, and CalNex 2010. Additional measurements are reported in the literature for ground-based field studies near Riverside in 1962 [Darley et al., 1963] and 1982 [Russell et al., 1988a]; Los Angeles in 1968 [Kopczynski et al., 1972]; Pasadena in 1973 [Hanst et al., 1982]; West Covina in 1973 by Spicer [1977b]; Claremont in 1978 by Tuazon et al. [1981] and 1985 by Grosjean [1986, 1988]; Glendora in 1986 by Anlauf et al. [1991]; and Tanbark Flats in 1990 and 1991 by Grosjean et al. [1993].

[15] Ground-based field measurements of hydrogen peroxide (H2O2) are available from the Carbonaceous Species Methods Comparison Study (CSMCS) at Glendora in 1986 [Kok et al., 1990; Mackay et al., 1990; Sakugawa and Kaplan, 1990; Tanner and Shen, 1990], SCAQS 1987 [Mackay, 1988], and LAAFRS 1993 [Mackay, 1994]; airborne measurements of H2O2 are available from ARCTAS-CARB 2008 [Jacob et al., 2010]. Hourly measurements of H2O2 were obtained upon request from CARB for LAAFRS, and 15 s H2O2 data were acquired from the NASA website for ARCTAS-CARB. All other data points represent reported averages or maxima from the literature. Historical measurements of H2O2 are summarized in Table 3. Earlier reports of H2O2 dating back to 1970 at Riverside [Bufalini et al., 1972] and 1977 at Claremont [Kok et al., 1978] were later invalidated owing to an interference of the H2O2 collecting solution with ozone and possible aqueous reactions of H2O2 with other gases [Kok et al., 1988; Lee et al., 2000]. Eliminating the data points from 1970 and 1977 leaves too few measurements to identify a significant trend; thus, further interpretation of H2O2 data in the SoCAB is not attempted here.

[16] A historical account of alkyl nitrates is also not reported here owing to a lack of field measurements. For analyses requiring total reactive nitrogen (NOy) in section 4.3, we assume alkyl nitrates are a relatively small contribution to the sum of reactive nitrogen species in the Los Angeles basin. We support this assumption by comparing daytime measurements of gas-phase NOy with the calculated sum of NOy, where ΣNOy = NOx + PAN + HNO3 [Parrish et al., 1993], from four field studies between 2002 and 2010. Quantitative agreement within 1σ uncertainties, which typically range from ±10% to ±20% of the mixing ratio, between measured NOy and ΣNOy suggests little contribution on average from other species to NOy during the most recent decade of our long-term analysis. An earlier account of alkyl nitrate contributions in the SoCAB reported by Grosjean [1983] showed mixing ratios of methyl nitrate up to ~5 ppbv during photochemical smog episodes in Claremont in 1980; however, methyl nitrate contributed <5% to ΣNOy since NOx, PAN, and HNO3 were also significantly enhanced during this time period.

[17] Accuracy of the utilized data is considered in this analysis. For data sets found in the literature, we assume that measurements were subjected to quality analysis and peer review prior to publication and that any concerns with data quality, measurement technique, instrumentation, or sampling were reported with the published data or in subsequent publications. For field study data provided by CARB, we assume that final reported data have been subjected to quality analysis with questionable data points eliminated or flagged appropriately. Where available, information regarding measurement uncertainties, precision, and possible interferences by other species is taken from data file headers and supporting documents.

2.2 Data Analysis

[18] This analysis is focused on measurements during summertime, when ozone production is at a maximum, and weekdays, when precursor emissions are most prominent. Surface network measurements are limited to data collected between 1 May and 30 September; field study data are averaged over the collection periods summarized in Table 3, which take place between May and November and range from a single day to several months. With the exception of the aircraft campaigns where flight times are typically during midday, measurements of VOCs, CO, and NOx are further limited to samples collected on weekday mornings between 0500 and 0900 PDT when precursor emissions are at a maximum and prior to development of a well-mixed planetary boundary layer and extensive photochemical processing. Conversely, measurements of oxidation products, including O3, PAN, and HNO3, are taken from weekday afternoons between 1200 and 1800 PDT when the planetary boundary layer is well developed and accumulated concentrations of secondary pollutants are at a maximum.

[19] To simplify the analysis, we group the available ground-based measurements from various locations throughout the SoCAB into two general geographic areas. The first includes sampling sites on the western side of the SoCAB located near major centers of emissions, such as Los Angeles, Pasadena, Azusa, and Glendora, and the second includes locations generally downwind, such as Upland, Claremont, and Riverside, on the eastern side of the basin. We distinguish data points acquired from the two defined areas for comparison. Airborne measurements from the mixed boundary layer over the SoCAB between 0.2 and 1 km above ground level, 33.6° < latitude < 34.3°, and −118.5° < longitude < −116.8° [Pollack et al., 2012], are retained for this analysis as a representative of basin-wide averages.

[20] Previously reported values for average atmospheric abundances and emissions ratios of ozone precursors and secondary pollutants are taken from the literature where available. Otherwise, we calculate average abundances and emissions ratios from the data sets described above. Average abundances are calculated as the arithmetic mean of all retained measurements for a given substance during the specified collection periods. The corresponding uncertainties represent a confidence limit calculated by dividing the 1σ standard deviation of the mean by the square root of the number of days of observations. Emissions ratios of directly emitted species, and enhancement ratios involving secondary oxidation products, are determined by linear least squares (LLS) [Press et al., 1988] orthogonal distance regression (ODR) [Boggs et al., 1987] to the relevant observations; the slope of the fitted line is interpreted as the emissions or enhancement ratio [Pollack et al., 2012]. Where information is available, the LLS ODR fits are weighted by the imprecision of the measurements, and a total uncertainty for each ratio is calculated by quadrature additions of the uncertainty in the fitted slope and the uncertainties of the respective measurements [Taylor, 1997]. Error bars associated with these ratios reflect measurement uncertainty as well as day-to-day variability. For data sets where instrument accuracy and precision information are not reported, uncertainty in the emissions ratio is reported as the 1σ standard deviation of the slope.

[21] LLS regression analysis of the logarithmic-transformed data is used to quantify the exponential change describing the trends in abundances and emissions ratios of ozone precursors (Table 4) as well as abundances and enhancement ratios of secondary oxidation products (Table 6). Extracting information about the trends from a functional fit to multiyear data minimizes the influence of interannual variability from various measurement sites, platforms, meteorological conditions, and possible pollution episodes. Here long-term changes are quantified and reported in two ways: (1) a constant rate of change, R, reported in units of percent change per year, and (2) a factor of change, f, spanning the range of years of available measurements. We assume an exponential function ((1)) for changes in abundances and emissions ratios over time.

display math(1)
Table 4. Rate of Change (Δ) and Corresponding 1σ Standard Deviation in Units of Percent Change Per Year for Abundances and Emissions Ratios of Ozone and Its Precursorsa
MeasurementField ObservationsbRoadside MonitorsAQMD Network—AzusaAQMD Network—UplandAll Observationsc
Δ (% yr−1)r2NΔ (% yr−1)r2NΔ (% yr−1)r2NΔ (% yr−1)r2NΔ (% yr−1)r2N
  1. a

    A negative rate of change signifies a decreasing trend over time; N represents the number of data points, corresponding to years of data, used for linear regression.

  2. b

    Overall trend since 1960 for ground-based observations of precursor abundances, overall trend since 1960 for airborne and ground-based observations of precursor emissions ratios.

  3. c

    Overall trend since 1960 representing combined data from the available field, roadside, and network observations.

  4. d

    Overall trend since 1973 determined from maximum 8 h average ozone for all sites in the SoCAB (Figure 1).

  5. e

    Overall trend since 1991 representing a combined LLS fit of the available field and roadside observations.

  6. f

    NOy used for emissions ratios from airborne and ground-based field studies where available.

O3 (8 h max)      −3.4 ± 0.20.8833−3.0 ± 0.10.9238−2.8 ± 0.1d0.9138
NOx−5.2 ± 1.00.828   −2.0 ± 0.30.6821−1.8 ± 0.30.5136−2.6 ± 0.30.5665
CO−7.9 ± 0.80.957   −4.8 ± 0.40.8921−5.2 ± 0.40.8726−5.7 ± 0.30.8654
Benzene−8.6 ± 0.80.949   −7.4 ± 0.50.9417−7.6 ± 0.80.8815−8.0 ± 0.30.9441
Toluene−8.0 ± 0.80.9111   −6.6 ± 0.40.9517−5.7 ± 0.60.8616−6.7 ± 0.30.9244
Ethylbenzene−8.6 ± 0.90.956   −5.8 ± 0.60.8517−4.2 ± 1.00.5516−6.9 ± 0.40.8739
o-Xylene−8.7 ± 1.00.947   −8.5 ± 0.60.9317−6.7 ± 0.70.8716−7.5 ± 0.30.9340
Isoprene      0.7 ± 2.10.01140.3 ± 0.90.0115   
CO/CO2−11.6 ± 1.70.926−9.9 ± 1.00.9210      −11.5 ± 1.4e0.8216
NOx/CO2f−5.5 ± 2.60.526−7.0 ± 1.00.869      −6.6 ± 1.7e0.5215
(Total HC)/CO2   −11.6 ± 1.00.8310         
Benzene/CO2−7.7 ± 1.30.886            
NOx/COf5.2 ± 1.40.59125.9 ± 1.30.7793.6 ± 0.70.68175.5 ± 0.40.89264.9 ± 0.40.7167
(Total HC)/CO   −1.8 ± 2.30.0710         
Benzene/CO−1.1 ± 0.60.3010   0.7 ± 0. 60.0817−1.6 ± 0.40.708−1.2 ± 0.40.2035
(Total HC)/NOx   −2.0 ± 2.70.089         
Toluene/NOx−4.4 ± 0.70.869   −3.8 ± 0.70.6817−5.1 ± 0.90.6916−4.6 ± 0.40.7842
Ethylbenzene/NOx−4.5 ± 0.40.975   −3.0 ± 0.90.4417−5.9 ± 2.00.4015−4.2 ± 0.60.6037
o-Xylene/NOx−5.6 ± 0.70.927   −5.5 ± 0.70.8017−5.4 ± 1.80.4015−5.7 ± 0.50.7739

[22] Rates and factors of change for a given species, y, are then derived from the slope, a, derived from a LLS fit to the logarithmic-transformed data points, as described by equations ((2))–((5)):

display math(2)
display math(3)
display math(4)
display math(5)

[23] The R percent change per year (Δt = 1 year) is determined from ((4)), and the f factor of change over the full time period of the historical measurements (e.g., Δt = 50 years for measurements between 1960 and 2010) is determined from ((5)). The LLS fits are not weighted by the uncertainties of the individual data points due to differences in calculating the reported uncertainties for abundances and emissions ratios described above. Uncertainties associated with the rate and factor of change reflect propagation of the 1σ standard deviation of the fitted slope.

[24] This analysis method is also used to determine trends in emissions and emission ratios of precursors in bottom-up emission inventories. Annual average emissions predicted by the inventories are typically reported by mass in units of short tons d−1, except for CO2 from the CARB greenhouse gas inventory which is reported in metric tons d−1. Molar emission ratios are calculated from annual average mass emissions by converting to metric tons then dividing by molecular weight (i.e., 0.028 kg/mol for CO, 0.046 kg/mol for NO2, and 0.044 kg/mol for CO2). Emissions and emission ratios involving unspeciated reactive organic gases (ROG) are not converted to molar values but are instead reported on an arbitrary scale. In sections 3.1 and 3.2, we compare trends in ambient measurements using airborne and ground-based field studies and surface network observations to inventory emission trends estimated by summing all source contributions in the yearly CARB emission inventories for the SoCAB [California Air Resources Board (CARB), 2009, accessed January 2013]. In section 3.2, we specifically compare trends in near-tailpipe measurements of light-duty gasoline-fueled vehicles (LDV) from roadside monitors to the EMFAC2011 on-road mobile emission inventory for this subset of vehicles (Table 5) [CARB, 2011, accessed January 2013].

Table 5. Rate of Change (Δ) and Corresponding 1σ Standard Deviation in Units of Percent Change per Year for Abundances and Emission Ratios of Precursors From the EMFAC2011 Emission Inventory Between 1990 and 2010 [CARB, 2011, Accessed January 2013] and the CARB Emission Inventory for the SoCAB Between 1975 and 2010 [CARB, 2009, Accessed January 2013]a
MeasurementEMFAC2011-LDVEMFAC2011-SGCARB InventoryCARB Inventory
(Light-Duty Gasoline Vehicles)(Light-Duty Gasoline + Heavy-Duty Diesel)(On-Road Sources)(All Sources)
Δ (% yr−1)r2NΔ (% yr−1)r2NΔ (% yr−1)r2NΔ (% yr−1)r2N
  1. a

    A negative rate of change signifies a decreasing trend over time; N represents the number of data points used for linear regression.

  2. b

    NOx, ROG, and CO from all emissions sources in the SoCAB [CARB, 2008a, accessed January 2013].

  3. c

    Gross CO2 emissions from the statewide greenhouse gas inventory [CARB, 2008a, accessed January 2013].

NOx−6.2 ± 0.30.995−4.2 ± 0.70.915−2.2 ± 0.5b0.858−2.1 ± 0.4b0.858
ROG−8.4 ± 0.21.005−8.2 ± 0.21.005−6.2 ± 0.4b0.968−4.4 ± 0.4b0.968
CO−8.5 ± 0.21.005−8.4 ± 0.11.005−5.8 ± 0.5b0.968−4.9 ± 0.4b0.968
CO21.5 ± 0.10.9851.6 ± 0.20.9650.4 ± 0.1c0.74200.9 ± 0.1c0.7420
CO/CO2−9.9 ± 0.31.005−9.8 ± 0.31.005      
NOx/CO2−7.6 ± 0.30.995−5.6 ± 0.60.965      
ROG/CO2−9.6 ± 0.31.005−9.6 ± 0.31.005      
NOx/CO2.5 ± 0.30.9654.6 ± 0.90.9053.7 ± 0.20.9982.9 ± 0.20.998
ROG/CO0.1 ± 0.10.2050.2 ± 0.10.435−0.5 ± 0.10.6280.5 ± 0.10.748
ROG/NOx−2.4 ± 0.40.925−4.2 ± 0.90.875−4.1 ± 0.20.998−2.3 ± 0.20.978

[25] In section 4, trends in secondary oxidation products determined from ambient measurements are compared with simulated mixing ratios [Fujita et al., 2013] predicted using a chemical box model.

3 Trends in Ozone and Its Precursors

3.1 Atmospheric Abundances

[26] Figure 1 illustrates the significant decrease in ozone concentrations observed in the SoCAB since the 1970s. The data points represent maximum 8 h average ozone mixing ratios measured basin wide and at selected individual surface monitoring network stations in the SoCAB for a 1 year period. Basin-wide measurements represent the maximum value reported from any of the individual network sites in the SoCAB in a given year, while selected individual sites are depicted to demonstrate the progression of ozone from coastal to inland sites across the basin. Here we derive the factor and rate of change in ozone from the slope of the LLS regression of the natural log of the basin-wide maximum 8 h ozone values in Figure 1. The data show that ozone maximum concentrations in the SoCAB have decreased at a rate of 2.8 ± 0.2% yr−1, equivalent to a decrease of a factor of 2.9 over nearly four decades. Although the maximum 8 h average ozone from the individual sites is often less than the basin-wide average, LLS fits of the long-term trends from the individual sites show similar rates of decrease in ozone (average of 3.3 ± 0.2% yr−1 for the five sites shown in Figure 1). We find that the basin-wide data are well fit by a constant exponential decrease of 2.8% yr−1 with no indication of a change in that rate between 1973 and 2010, in contrast to recent descriptions of ozone trends [e.g., Fujita et al., 2013; Warneke et al., 2012] that have invoked a slower rate of decrease in ozone in Los Angeles after 1999. Inclusion of a constant term in the fitting expression to represent a nonnegligible ozone background does not significantly improve the residuals to the data points in Figure 1. However, the constant rate of change must slow when future ozone concentrations approach baseline ozone levels transported into the SoCAB.

[27] The observed trend in ozone is positively correlated with trends in its precursors. Time series plots of the annually averaged abundances of NOx and CO (Figure 3) and select VOCs (Figure 4) using data from the literature, ground-based field campaigns, and the Azusa and Upland SCAQMD surface network stations demonstrate large decreases in precursor emissions over the past five decades [McDonald et al., 2012; Warneke et al., 2012]. Considering all data, abundances of NOx have decreased at an average rate of 2.6 ± 0.3% yr−1, resulting in a factor of 3.7 decrease over the 50 years between 1960 and 2010. There are site-specific differences in the observed rates, but we take the combined observations from all locations as the most robust indication of the rate of NOx decrease in the SoCAB. The combined rate of decrease reported here of 2.6 ± 0.3% yr−1 is in agreement with NOx emissions trends in the CARB emissions inventory [CARB, 2009, accessed January 2013], which predict an annual average decrease in NOx of 2.1 ± 0.4% yr−1 for all sources between 1975 and 2010 (Table 5).

Figure 3.

Average abundances of NOx (in units of ppbv) and CO (in units of ppmv) measured at the Azusa (solid black circles) and Upland (open circles, offset by 0.5 years on x axis) surface network sites and ground-based field campaigns near LA/Pasadena (solid blue triangles) and Claremont (open blue triangles) since 1960. Error bars represent confidence limits based on the number of days of observations per year. LLS fits (red lines) represent long-term trends in NOx and CO abundances.

Figure 4.

Average abundances of select NMOCs (in units of ppbv) measured at the Azusa (solid black circles) and Upland (open circles, offset by 0.5 years on x axis) surface network sites and ground-based field campaigns near LA/Pasadena (solid blue triangles) and Claremont (open blue triangles) since 1960. Error bars represent confidence limits based on the number of days of observations per year. LLS fits (red lines) represent long-term trends in abundances of speciated NMOCs. NMOCs related to vehicle exhaust have decreased at an average rate of 7.3% yr−1. No significant change is observed for isoprene, an indicator of biogenic emissions.

[28] Since 1960, CO abundances have decreased at a rate of 5.7 ± 0.3% yr−1, roughly a factor of 2 more rapidly than NOx, corresponding to a factor of 19 decrease in CO between 1960 and 2010 [e.g., Warneke et al., 2012]. The overall rate of change from CO measurements in the SoCAB is consistent with a nationwide decrease in urban CO abundances of 5.2 ± 0.8% yr−1 observed between 1989 and 1999 by Parrish et al. [2002]. Observed CO trends in the SoCAB are consistent with those predicted in the CARB emissions inventory (e.g., annual average decrease in CO of 4.9 ± 0.4% yr−1 from all sources and 5.8 ± 0.5% yr−1 from on-road mobile sources) [CARB, 2009, accessed January 2013].

[29] For anthropogenic VOCs, an average rate of decrease of 7.3 ± 0.7% yr−1, corresponding to a decrease in average abundances by a factor of 44 over the 50 years, is determined using data from the Azusa and Upland SCAQMD stations and the ground-based field measurements since 1960 (Figure 4). This value, derived from two representative sites, is in quantitative agreement with a basin-wide average rate of decrease of 7.5% yr−1 [Warneke et al., 2012]. The annual average decrease in ROG of 6.2 ± 0.4% yr−1 between 1975 and 2010 solely from SoCAB on-road mobile sources in the CARB inventory (Table 5) is in better agreement with the ambient observations than is the annual average decrease of 4.4 ± 0.4% yr−1 determined from all sources in the CARB inventory.

[30] In contrast to anthropogenic VOCs, temporal trends in isoprene (Figure 4) show no statistically significant change since 1987 in the abundance of this biogenic VOC in the SoCAB. Additionally, no significant differences in average abundances of isoprene were observed from the arithmetic mean of the data from summertime weekday mornings (0500 and 0900 PDT) and weekday afternoons (1200 and 1800 PDT), when emissions of biogenic VOCs are most prominent [Guenther et al., 1993].

3.2 Emissions Ratios

[31] Although many factors have contributed to changes in ambient concentrations of NOx, CO, and VOCs over the years, decreasing abundances in the SoCAB are predominantly attributed to decreasing emissions from motor vehicles due to increasingly strict emissions standards in California [Ban-Weiss et al., 2008; Bishop and Stedman, 2008; Fujita et al., 2013; Harley et al., 2005; Lawson et al., 1990; Lawson, 2003; McDonald et al., 2012; Warneke et al., 2012]. Large decreases in motor vehicle emissions have occurred despite a factor of 2.4 increase in population (http://quickfacts.census.gov/qfd/states/06000.html) and a factor of 3 increase in fuel sales in the state of California since the 1960s [Warneke et al., 2012]. In this section, we investigate the relative changes in precursor abundances and demonstrate their connection to changes in motor vehicle emissions by examining changes in observed emissions ratios.

3.2.1 Emissions Ratios to CO2

[32] We start by comparing emissions ratios to CO2 derived from near-tailpipe roadside measurements to airborne and ground-based measurements from field studies. Figure 5 illustrates changes over time in molar emission ratios of CO/CO2, NOx/CO2, and (total HC)/CO2 determined from the roadside observations since 1991 and from ambient observations from field studies since 1997 when ozone precursors and CO2 were concurrently measured. As in previous studies [Murphy et al., 2007; Parrish et al., 2002; Pollack et al., 2012], we use field observations of gas-phase NOy when available as a more conserved tracer than NOx for determining emissions ratios to long-lived atmospheric species such as CO and CO2. Benzene is selected for comparison of ambient measurements of speciated VOC/CO2 to roadside measurements of (total HC)/CO2 due to its relatively slow reactivity with OH, thereby providing information about emissions independent of atmospheric processing or removal.

Figure 5.

Molar emissions ratios of CO/CO2, NOx/CO2, and HC/CO2 (left axes) determined from roadside measurements (green crosses) and emissions ratios of CO/CO2, NOy/CO2, and benzene/CO2 (right axes) from airborne (red squares) and ground-based field measurements (blue triangles). LLS fits are indicated for roadside (dashed black line) and field (solid black line) measurements.

[33] The differences between roadside and ambient measurements reflect contributions from sources other than motor vehicles to total emissions of ozone precursors in the SoCAB. In Figure 5, ambient molar emission ratios of CO/CO2 and NOy/CO2 are smaller than the corresponding emission ratios determined from near-tailpipe roadside measurements by an average percent difference of 37 ± 14% and 36 ± 9%, respectively. This difference indicates smaller relative contributions of on-road motor vehicles to total CO2 emissions than to total CO and NOx. According to 2008 California emission inventories [CARB, 2008a, accessed January 2013, 2008c, accessed January 2013], on-road motor vehicles are responsible for 53% of total CO and 49% of total NOx in the SoCAB but only 35% of the total statewide CO2 emissions. Assuming that the contribution from on-road motor vehicles to total CO2 emissions in the SoCAB is the same as it is statewide (35%), we expect a 34% and 29% difference between roadside and ambient CO/CO2 and NOx/CO2 measurements for the SoCAB. These differences are in close accord with the observations, which show a difference of 37 ± 14% for CO/CO2 and 36 ± 9% for NOx/CO2, suggesting minimal variability between different measurement platforms and demonstrating the relative accuracy of the CARB emission inventories.

[34] Roadside measurements of CO/CO2 and NOx/CO2 between 1991 and 2010 and ambient measurements of CO/CO2 and NOy/CO2 between 1997 and 2010 decreased at similar rates, within the 1σ uncertainties of each determination (Figure 5 and Table 4). Larger differences are observed between trends in roadside measurements of (total HC)/CO2 (−11.6 ± 1.0% yr−1) and ambient measurements of benzene/CO2 (−7.7 ± 1.3% yr−1). Due to its toxicity, benzene has been selectively and progressively removed from fuels over the years, accounting for its disproportional change relative to CO. Regulatory efforts have resulted in a nationwide reduction of benzene relative to acetylene from vehicle exhaust by 40% between 1994 and 2002, corresponding to a reduction in benzene emissions by roughly 56% [Fortin et al., 2005]. Fuel composition changes focused on minimizing benzene emissions between 1994 and 2002 contributed to the factor of 65 decrease in benzene abundances in the SoCAB observed between 1960 and 2010 (Figure 4). This specific focus on benzene approximately accounts for the greater decrease in ambient benzene abundances relative to the factor of 32 decrease in toluene abundances over the same 50 year period. Since benzene has decreased faster than most VOCs, we expect the rate of decrease in emissions ratios of benzene/CO2 to represent an upper limit for decreases in ambient VOC/CO2 ratios over time; however, (total HC) has decreased at a faster rate than benzene with respect to CO2. Discrepancies between the trends may arise from the lack of ambient benzene/CO2 data prior to implementation of regulatory efforts in 1994. Excluding the 1991 roadside data point from the LLS fit gives a rate of change for (total HC)/CO2 of −8.9 ± 2.3% yr−1, which is in better agreement with the decrease in ambient measurements of benzene/CO2 (−7.7 ± 1.3% yr−1). We also compare trends in molar emission ratios to CO2 determined from roadside measurements, which targeted gasoline-fueled passenger vehicles, since 1991 to those predicted by the EMFAC2011-LDV emission inventory during summer between 1990 and 2010 (Table 5) [CARB, 2011, accessed January 2013]. Trends in roadside measurements (Table 4) agree reasonably with those from EMFAC2011-LDV (Table 5), demonstrating consistency between measurements and inventory emissions for light-duty vehicles. A slower decrease in NOx/CO2 ratio determined using EMFAC2011-SG (encompassing all on-road sources) compared to EMFAC2011-LDV shows that diesel-fueled vehicles are a significant contributor to on-road NOx emissions [Ban-Weiss et al., 2008; McDonald et al., 2012]. From the difference in long-term trends from the EMFAC2011 inventory, we estimate that the presence of diesel-fueled vehicles has slowed the decrease of NOx emissions from on-road mobile sources by 2.1 ± 0.7% yr−1 consistent with the ambient measurements that sampled all sources in the SoCAB.

3.2.2 Emissions Ratios to CO

[35] Next we look at trends in emissions ratios of select VOCs and NOx relative to CO (Figure 6). A LLS fit of ambient enhancement ratios shows a statistically significant trend of −1.2 ± 0.4% yr−1 since 1960 in benzene/CO but no significant trend (−1.8 ± 2.3% yr−1) in (total HC)/CO ratio in roadside measurements since 1991 (Table 4). The differences are attributed to the temporal extent of the data record (e.g., ambient measurements since 1960 and roadside data only since 1991) as well as the faster decrease in emissions of benzene resulting from regulatory efforts initiated in the mid-1990s. The EMFAC2011-LDV emission inventory suggests no significant trend (+0.1 ± 0.1% yr−1) for ROG/CO between 1990 and 2010, consistent with the lack of trend in the roadside measurements.

Figure 6.

Ambient emissions ratios of VOC/CO and NOx/CO (in units of ppbv/ppbv) determined from airborne measurements (red squares), ground-based field measurements near Azusa, Pasadena, and LA (solid blue triangles) and Claremont (open blue triangles), and surface network measurements for Azusa (solid black circles) and Upland (open black circles). Emission ratios of (total HC)/CO and NOx/CO (green crosses) from roadside measurements and ROG/CO and NOx/CO from the CARB emission inventory for all sources (purple symbols) and on-road sources (orange symbols) are also shown. Emission ratios of NOx/CO from the various measurements are presented on a common scale. Ambient measurements of VOC/CO correspond with the right axis scale, roadside measurements of (total HC)/CO correspond with the left axis scale, and ROG/CO from the inventory is presented on an arbitrary scale. Solid lines represent the LLS fit to the combined ambient measurements; dashed lines represent LLS fits to the inventory data.

[36] Similarly, the long-term rate of change in ambient observations of benzene/CO can be compared with ROG/CO ratio determined for all sources in the SoCAB from the CARB emission inventory between 1975 and 2010 (Figure 6). Inventory-based ROG/CO ratios are presented on an arbitrary scale in Figure 6 and result in a small but significant rate of increase of 0.5 ± 0.1% yr−1. In contrast, ROG/CO ratio determined for on-road sources results in a small but significant rate of decrease (−0.5 ± 0.1% yr−1). The long-term trends determined from the CARB inventory are inconsistent with the larger decrease in benzene/CO (−1.2 ± 0.4% yr−1) from the ambient measurements, although the latter reflects the additional decrease in benzene due to regulatory efforts. In a similar analysis utilizing another long-lived and well-conserved tracer, Warneke et al. [2012] found little change in the ratio of acetylene/CO (0.3 ± 0.5% yr−1) since 1960 (C. Warneke, personal communication, 2013).

[37] Since regulatory efforts have led to additional reductions in benzene relative to other VOCs and CO from light-duty vehicles, we report trends in toluene, ethylbenzene, and o-xylene to further characterize changes in gasoline-fueled motor vehicle emissions over time. Although these VOCs are shorter-lived, we use data sampled during weekday mornings between 0500 and 0900 PDT to characterize VOC/CO ratios when emissions are at a maximum [Parrish et al., 2002] and prior to extensive photochemical processing, as described in section 2.2. Ambient measurements show a slight increase in VOC/CO ratio over time (e.g., +1.3 ± 0.4% yr−1 for toluene/CO, +0.8 ± 0.5% yr−1 for ethylbenzene/CO, and +0.5 ± 0.5% yr−1 for o-xylene/CO). Assuming these compounds have a common source and should show similar changes, an average of +0.3 ± 1.0% yr−1 between 1960 and 2010 suggests there is no detectable trend in the average gasoline-fueled motor vehicle VOC/CO emissions ratio. This is in general agreement with a trend of −1.7 ± 1.0% yr−1 determined from the rates of change in abundances of VOCs (−7.3 ± 0.7% yr−1) and CO (−5.7 ± 0.3% yr−1) using the combined observations (Table 4). Warneke et al. [2012] found the ratios to CO relatively constant for all VOCs in the LA basin since 1960. The long-term rate of increase of 0.5 ± 0.1% yr−1 for ROG/CO in the SoCAB determined from the CARB emission inventory between 1975 and 2010 is in agreement with the combined observations of VOC/CO (+0.3 ± 1.0% yr−1) from the ambient measurements since 1960.

[38] Emissions ratios of NOy/CO from roadside and ambient measurements are also shown in Figure 6. The observed increase in NOx/CO emissions ratio follows from the faster decrease in CO emissions (average of 5.7% yr−1) compared to the decrease in NOx (average of 2.6% yr−1) even though the atmospheric abundances of both have decreased significantly since 1960. LLS fit of the combined ambient and roadside measurements shows a positive trend in the NOx/CO emissions ratio of 4.9 ± 0.4% yr−1 between 1960 and 2010 (Table 4). This value is larger than the trend of 3.3 ± 0.5% yr−1 based on the relative rates of decrease in NOx and CO abundances. The rate of increase in NOx/CO from the ambient field measurements (5.2 ± 1.4% yr−1) is in agreement with that determined from the roadside measurements but larger than the long-term rate of change in NOx/CO ratio predicted between 1975 and 2010 for the SoCAB by the CARB emission inventory (3.7 ± 0.2% yr−1 for on-road sources and 2.9 ± 0.2% yr−1 for all sources) (Table 5).

[39] Absolute emission ratios of NOx/CO predicted by the CARB emission inventory, which are included in Figure 6, are larger than that determined from the combined observations. Overestimates of NOx/CO ratio from the inventory decrease from a factor of 4.3 for all sources and 2.8 for on-road sources in 1990 to a factor of 1.9 and 1.6, respectively, in 2010. These differences must reflect significant error(s) in the inventory. In contrast, top-down estimates of emissions ratios of NOy/CO for weekdays in 2002 and 2010 in the LA basin using an inverse model [Brioude et al., 2012] show better agreement with absolute emission ratios determined from the field measurements (e.g., NOy/CO ratios simulated for 2002 and 2010 are underestimated by an average factor of 1.5). The long-term rate of change in measured weekday ambient NOx/CO is also in agreement with that determined from the EMFAC2011-SG inventory (4.6 ± 0.9% yr−1), which combines emissions from gasoline-fueled and diesel-fueled vehicles [Pollack et al., 2012]. However, there is less agreement between the long-term increase in roadside measurements of NOx/CO from light-duty vehicles (5.9 ± 1.3% yr−1) and the rate of increase determined from the EMFAC2011-LDV inventory (2.5 ± 0.3% yr−1).

[40] Here we have utilized long-lived, well-conserved species to make direct correlations between abundances and emissions ratios measured in the SoCAB. Consistency between trends derived from near-tailpipe roadside and from ambient measurements confirms on-road motor vehicles as the predominant source of CO, VOC, and NOx emissions in the SoCAB. Agreement between trends in measured emissions ratios and atmospheric abundances demonstrates that changes in ambient concentrations in the SoCAB continue to reflect changes in emission rates of precursor species. This analysis has shown that all of the long-term trends in emissions, except for NOx, are underestimated by the CARB inventory for all sources in the SoCAB. In contrast, all of the long-term rates of change predicted by EMFAC2011-LDV, with the exception of NOx/CO ratio, agree reasonably with the long-term trends determined from the roadside measurements of light-duty vehicles emissions. Differences between long-term trends from the measurements and emission inventories suggest that statistically significant errors still exist in the inventories.

3.2.3 VOC/NOx Ratio

[41] Larger reductions in emissions of VOCs relative to NOx over time have resulted in decreasing VOC/NOx emission ratios in the SoCAB over the past five decades. Data from the monitoring network and from intensive field measurements (Figure 7) show the trends in toluene, ethylbenzene, and o-xylene emissions relative to NOx. These three VOCs are selected as compounds characteristic of vehicle exhaust with OH reaction rate coefficients similar to that for NO2 [Sander et al., 2006], such that VOC/NOx emissions ratios should be approximately preserved during processing by OH. Additionally, as noted above, we minimize chemical processing of the emissions prior to analysis by selectively using data from weekday mornings. A trend of −4.8 ± 0.9% yr−1 for these VOC/NOx ratios is derived from network and field measurements, consistent with the value of −4.5 ± 0.8% yr−1 calculated from observed trends in VOC and NOx abundances. This trend is significantly larger than −2.3 ± 0.2% yr−1 in the unspeciated ROG/NOx emissions ratio predicted by the CARB inventory for all sources in the SoCAB between 1975 and 2010. The discrepancy between observed and inventory values is consistent with Fujita et al. [2013], who reported that ambient measured NMOC/NOx ratios from SCAQS 1987, SCOS 1997, and field measurements in 2009 are about a factor of 2 higher than the emission inventory ratios. Although a trend of −2.4 ± 0.4% yr−1 in ROG/NOx ratio is predicted by the EMFAC2011-LDV emission inventory [CARB, 2011, accessed January 2013] between 1990 and 2010, roadside data since 1991 show no significant trend (2.0 ± 2.7% yr−1) in (total HC)/NOx emissions ratio (Table 4). Trends in ambient VOC/NOx ratios (average of −4.8 ± 0.9% yr−1) are in agreement with changes of −4.2 ± 0.9% yr−1 in ROG/NOx ratio determined from the EMFAC2011-SG emission inventory, which reflects combined emissions from light-duty gasoline-fueled and heavy-duty diesel-fueled vehicles.

Figure 7.

Ambient VOC/NOx ratio (units of ppbv/ppbv, right axes scales) for airborne (red squares), ground-based field measurements near Azusa, Pasadena, and LA (solid blue triangles) and Claremont (open blue triangles), and surface network measurements for Azusa (solid black circles) and Upland (open black circles). Roadside measurements of (total HC)/NOx (green crosses on left axis); emission ratios of ROG/NOx from the CARB emission inventory for all sources (purple symbols) and on-road sources (orange symbols) are presented on an arbitrary scale. Solid lines represent the LLS fits to the combined ambient measurements; dashed lines represent LLS fits to the inventory data.

4 Causes of Decreasing Ozone in the SoCAB

[42] Ozone formation in the SoCAB strongly depends on abundances of anthropogenic VOCs and NOx and on the VOC/NOx ratio [Fujita et al., 2003; Fujita et al., 2013; Lawson, 2003; Pollack et al., 2012]. Figure 8 shows positive correlations between decreasing maximum 8 h ozone concentrations and decreasing VOC/NOx ratio since 1975 from the Azusa and Upland surface network data sets. However, the causal relationships between decreasing ozone and decreasing precursor emissions, and their ratios, remain unclear. In the following sections, we use historical measurements of secondary oxidation products to better identify the major causes of decreasing ozone in the SoCAB. We analyze ambient measurements of secondary pollutants from summertime weekday afternoons between 1200 and 1800 PDT when photochemical processing is well advanced and the planetary boundary layer is well developed. We assume negligible influence from any changes in meteorological conditions that might affect average atmospheric residence times in the SoCAB, dilution, pollutant carryover, transport, and deposition over those 50 years. As discussed in sections 4.2 and 4.3, this approach is supported by airborne measurements from CalNex 2010, ARCTAS 2008, and ITCT 2002, which show no statistically significant difference in NOy/CO and NOy/CO2 emissions ratios throughout the sampling duration of each flight (typically 1000 to 1800 PDT) and between the eastern and western geographic portions of the SoCAB over the past decade.

Figure 8.

Maximum 8 h average ozone (in units of ppbv) versus VOC/NOx ratios (in units of ppbv/ppbv) from the Azusa (solid symbols) and Upland (open symbols) network monitoring sites. Decreasing ozone is correlated with decreasing VOC/NOx ratio from 1994 (red) to 2010 (purple); data from 1975 (black) are also shown where available. Solid black lines represent LLS fits of the data points; also reported are the R2 values determined from each fit.

4.1 Long-Term Trends in NOx Oxidation Products

[43] Atmospheric PAN is a result of HOx radical chemistry and is positively correlated with ozone production [Pollack et al., 2012; Roberts et al., 2007; Roberts et al., 1995; Sillman, 1995]. Thermal decomposition of PAN regenerates RO2 and NO2 and promotes continued ozone production. In contrast, formation of HNO3 removes OH radicals from the chain reactions that produce ozone, and thus atmospheric HNO3 is a result of chemical reactions that terminate the ozone formation cycle. The abundance of PAN relative to HNO3 provides an indication of the balance between propagation and termination of the catalytic ozone formation cycle.

[44] The temporal trends of the average abundances of PAN and HNO3 since 1973 from the available airborne and ground-based measurements are illustrated in Figure 9. PAN maxima prior to 1973 reported in the literature are also shown in the figure, although they are not included in the determination of the long-term trend. Although concentrations of secondary pollutants may vary significantly with location in the SoCAB, the data in Figure 9 show no systematic difference in abundances of PAN measured at these selected locations in the SoCAB. Measured abundances of PAN exhibit an average trend of −9.3 ± 1.1% yr−1 since 1973 (Table 6), corresponding to a factor of 34 decrease between 1973 and 2010. PAN was first identified as a particularly important eye irritant in Los Angeles smog [Leighton, 1961]. PAN abundances have decreased much more rapidly than ozone (e.g., factor of 34 reduction in PAN versus factor of 2.9 reduction in O3 over the same 37 year period), demonstrating much greater progress in reducing some compounds of concern for air quality compared to O3.

Figure 9.

Changes in average abundances of PAN and HNO3 and enhancement ratios of PAN/HNO3 since 1973 using data from airborne (red squares) and ground-based field studies near Los Angeles (solid blue triangles) and Claremont (open blue triangles). Observed PAN maxima between 1960 and 1970 are included in the upper plot but excluded from the LLS fit (solid black lines); however, the extrapolated LLS fit to 1960 (dashed black lines) is in agreement with those observations. Values predicted by box model simulations [Fujita et al., 2013] (green symbols and dashed lines) are shown in comparison to the measurements.

Table 6. Rate of Change (Δ) and Corresponding 1σ Standard Deviation in Units of Percent Change Per Year for Abundances and Enhancement Ratios of Secondary Oxidation Products Determined From the Airborne and Ground-Based Field Measurements Since 1973 and Values Predicted by Fujita et al. [2013] for 1985, 1995, and 2005 Using a Chemical Box Modela
MeasurementField ObservationsModel Predicted
Δ (% yr−1)r2NΔ (% yr−1)r2N
  1. a

    A negative rate of change signifies a decreasing trend over time; N represents the number of data points, corresponding to years of data, used for linear regression.

  2. b

    LLS fit of values predicted for O3/(NOy-[BOND]NOx) [Fujita et al., 2013].

PAN−9.3 ± 1.10.8215−13.5 ± 1.10.993
HNO3−3.0 ± 0.80.4716−5.3 ± 1.00.963
PAN/HNO3−7.3 ± 1.60.7411−8.6 ± 0.11.003
O3/(PAN + HNO3)0.8 ± 1.40.0410−1.1 ± 0.6b0.753
Ox/(PAN + HNO3)−1.4 ± 1.30.149   
NOx/NOy−0.9 ± 0.90.7710   
(PAN + HNO3)/NOy2.2 ± 0.50.6810−3.0 ± 2.21.003
PAN/NOy−2.0 ± 1.40.2010−10.6 ± 0.51.003
HNO3/NOy3.3 ± 1.00.5610−2.3 ± 0.40.973

[45] Previous studies [Kelly, 1992; Russell et al., 1988b; Spicer, 1983] have shown direct correlation between reductions in PAN, which is produced from VOC oxidation in the presence of NOx, and reductions in both VOC and NOx precursors. The observed trend in PAN since 1973 of −9.3 ± 1.1% yr−1 is quantitatively similar to the rate of change in the product of the trends in its precursors (−9.7 ± 0.7% yr−1), calculated from the sum of the slopes of the long-term trends of VOC and NOx since 1960 (Table 4) and using equation (4). The difference may be attributed to our selection of VOCs for the combined trend reported in Table 4, which does not include acetaldehyde, the direct precursor to PAN formation. Analysis of the trend in acetaldehyde results in a decrease of 7.3 ± 1.6% yr−1 in the SoCAB, although this is based solely on measurements after 1980. A similar decrease is observed for formaldehyde, which serves as an indicator for formation of secondary organic products. Analysis of formaldehyde maxima [Grosjean, 1982] dating back to 1960 shows an average trend of −7.6 ± 1.0% yr−1, consistent with that calculated by averaging trends in directly emitted benzene, toluene, ethylbenzene, and o-xylene (−7.3 ± 0.7% yr−1) (Table 4). The observed PAN trend since 1973 (−9.3 ± 1.1% yr−1) is in agreement with the product of the decrease in abundances of formaldehyde and NOx (10.0 ± 1.0% yr−1).

[46] Average abundances of HNO3 have changed at a rate of −3.0 ± 0.8% yr−1 (Table 6), corresponding to a factor of 3.1 decrease between 1973 and 2010. Reductions in HNO3 concentrations have been found to be directly proportional to decreases in NOx [Kelly, 1992; Russell et al., 1988b; Spicer, 1983]. The trend in HNO3 agrees within statistical uncertainty with the average rate of decrease determined for NOx (−2.6 ± 0.3% yr−1). As found for PAN, average HNO3 abundances in Figure 9 show no dependence on the locations selected to represent the SoCAB in this analysis. The faster rate of decrease in PAN relative to HNO3 since 1973 results in a trend in enhancement ratios of PAN/HNO3 of −7.3 ± 1.6% yr−1. Owing to the positive correlations of PAN with ozone production and HNO3 with radical termination, long-term decreases in the relative concentrations of PAN and HNO3 indicate that decreasing ozone concentrations in the SoCAB are a direct result of decreasing photochemical production of ozone due to decreasing abundances of VOC and NOx precursors.

[47] Also shown in Figure 9 are abundances of PAN and HNO3 simulated for 1985, 1995, and 2005 in the SoCAB using a chemical box model [Fujita et al., 2013]. While simulated abundances of PAN are consistent with the measured values, simulated abundances of HNO3 are overestimated by roughly a factor of 3. The model also predicts significantly larger rates of decrease in PAN and HNO3 than the apparent trends in the ambient measurements (Table 6).

4.2 Change in Ozone Production Efficiency

[48] In this section, we investigate the contribution of changes in ozone production efficiency (OPE) to the observed decreases in ozone abundances over time. We take the slope of a LLS fit of O3 versus the sum of the major NOx oxidation products (PAN + HNO3) to provide a measure of ozone produced per NOx oxidized [Trainer et al., 1993]. Figure 10 shows a scatter plot of O3 versus (PAN + HNO3) using the available field measurements in the SoCAB since 1973. LLS fits result in large variability in the slopes determined from each data set. Given the variability in fitted slopes between data sets, this analysis finds no significant trend in derived OPE between 1973 and 2010. LLS slopes from scatter plots of odd oxygen (Ox = O3 + NO2) versus (PAN + HNO3) are also shown in Figure 11. Ox has been interpreted in previous studies as a more conserved measure of net oxidant production [Johnson, 1984; Murphy et al., 2007; Pollack et al., 2012; Tonse et al., 2008] and is used here to isolate the relative contributions from titration and net photochemical production to observed ozone levels. No detectable trend in O3/(PAN + HNO3) and Ox/(PAN + HNO3) suggests that over the 37 years spanned by the data, OPE has remained constant within our ability to quantify from the available data. Model-predicted ratios for O3/(NOy-[BOND]NOx) reported by Fujita et al. [2013], also shown in Figure 11, agree well with the measured ratios and further indicate no detectable change in OPE over time.

Figure 10.

Correlation plots of O3 versus (PAN + HNO3) using airborne and ground-based measurements from field studies between 1973 (red) and 2010 (purple), with the right plot showing an expanded view of the region inside the dashed boxed from the left plot. Ground-based measurements include data reported by Spicer [1977b], Tuazon et al. [1981], Hanst et al. [1982], and Grosjean [1986] and from the SCAQS (1987), LAAFRS (1993), SCOS (1997), and CalNex (2010) field campaigns; airborne measurements include data from ITCT (2002), ARCTAS-CARB (2008), and CalNex (2010). Solid lines represent LLS ODR of each data set corresponding by color; a time series of the slopes is presented in Figure 11.

Figure 11.

Changes in O3/(PAN + HNO3) and Ox/(PAN + HNO3) since 1973 using data from airborne (red squares) and ground-based (blue triangles) field studies near Los Angeles (solid blue triangles) and Claremont (open blue triangles). Values predicted by box model simulations [Fujita et al., 2013] (green symbols and dashed lines) are shown in comparison to the measurements. Assuming no loss of NOy species, LLS fits show no significant change since 1973 (solid black lines).

[49] The above calculations assume no differential loss of NOy species. However, OPE interpreted from these slopes will be overestimated if a significant fraction of any gas-phase NOy species, e.g., HNO3, is removed from the atmosphere prior to measurement [Ryerson et al., 1998; Trainer et al., 1993]. Although minimal loss of NOy species in the SoCAB is justified by the minimal change observed in ratios of NOy/CO and NOy/CO2 over the spatial extent of the basin and the temporal extent of sampling using the airborne data sets, we also consider the influence of loss of HNO3 on the derived OPE. Since HNO3 is highly soluble, it is subject to removal by wet and dry deposition, and it can also form ammonium nitrate aerosol. We have recalculated OPE assuming 50% loss of HNO3 prior to measurement, which results in larger positive biases in OPE for years between 1990 and 2010 where HNO3 is a larger fraction of NOy. Even under this extreme example, HNO3 loss does not significantly affect the long-term LLS fit. With or without HNO3 loss, we detect no significant change in OPE between 1973 and 2010 based on analysis of the ambient data.

[50] The lack of a discernible trend in derived OPE since 1973 (Figure 11 and Table 6) contradicts an expected decrease due to the decreasing VOC/NOx ratio. Studies of weekday-weekend differences in ozone in the SoCAB have shown decreasing OPE with decreasing VOC/NOx ratio on weekdays versus weekends [Fujita et al., 2013; Pollack et al., 2012]. However, the observed changes in VOC/NOx ratio in these studies are primarily driven by weekend reductions in NOx with little weekday-weekend difference in VOCs. Past studies [Lin et al., 1988; Liu et al., 1987; Roberts et al., 2007; Roberts et al., 1995; Ryerson et al., 1998; Ryerson et al., 2001; Trainer et al., 1993] have shown that the amount of ozone produced per NOx oxidized depends nonlinearly on NOx and VOC concentrations, such that OPE at a fixed VOC/NOx ratio increases with decreasing abundances of NOx. Here we suggest that the net effect of decreasing NOx abundances and simultaneously decreasing VOC/NOx ratio over time in the SoCAB have effectively canceled, resulting in minimal change in OPE over the past 50 years.

4.3 Changes in Rate of Photochemical Processing

[51] Assuming atmospheric mixing and residence times in the SoCAB have not changed in the last 50 years, changes in the fraction of oxidized NOx relative to NOy indicate differences in the rate of photochemical processing of NOx over time [Pollack et al., 2012]. Figure 12 presents the data since 1973 in two coupled ways: the measured NOx to NOy ratio in the top panel illustrates changes in the average fraction of unreacted NOx, while measured (PAN + HNO3) to NOy ratio in the bottom panel illustrates changes in the average fraction of NOx oxidation products to NOy. As discussed in section 2.1.3, measured gas-phase NOy data are used where available; the sum of individually measured gas-phase NOy constituent species (∑NOy; typically NOx + PAN + HNO3) is used otherwise. Alkyl nitrates were not measured in most of the field studies used in this analysis and thus are not incorporated into ∑NOy but are assumed to be a relatively small contribution to total NOy in the Los Angeles basin. LLS fit to these data suggests a statistically insignificant trend of −0.9 ± 0.9% yr−1 in the fraction of unreacted NOx in the SoCAB between 1973 and 2010 (Table 6). On the other hand, we observe a significant increase in the fraction of oxidized NOx of 2.2 ± 0.5% yr−1. This result is robust to recalculation of the (PAN + HNO3)/NOy trend assuming 50% loss of HNO3 prior to gas-phase measurement and suggests that this trend is insensitive to the fraction of HNO3 in NOy. In contrast, ratios of (PAN + HNO3)/NOy derived from box model simulations in Fujita et al. [2013] show a decreasing trend in the fraction of oxidized NOx (Figure 12 and Table 6) and are inconsistent with the increasing trend derived from the ambient data.

Figure 12.

Changes in the fraction of unreacted NOx and NOx oxidation products (PAN + HNO3) relative to total reactive nitrogen (NOy) using data from airborne (red squares) and ground-based (blue triangles) field studies near Los Angeles (solid blue triangles) and Claremont (open blue triangles). Values predicted by box model simulations [Fujita et al., 2013] (green symbols and dashed lines) are shown in comparison to the measurements. Assuming no loss of NOy species, LLS fits (solid black lines) of the long-term trends show an average decrease in unreacted NOx (−0.9% yr−1) and an average increase in NOx oxidation products (2.2% yr−1) since 1973.

[52] The ambient data suggest that atmospheric photochemical processing of NOx in the SoCAB has increased over the 37 years spanned by the available measurements. The increase in the fraction of oxidized NOx since 1973 indicates more rapid processing of NOx as a result of the large changes in amount and proportion of emissions in the Los Angeles basin. Along with faster photochemical processing, emissions changes have resulted in increasingly favored production of HNO3 over time, demonstrated by a derived positive trend of 3.3 ± 1.0% yr−1 in HNO3/NOy and a trend of −2.0 ± 1.4% yr−1 in PAN/NOy (Figure 13 and Table 6).

Figure 13.

Changes in the fraction of PAN and HNO3 relative to total reactive nitrogen (NOy) using data from airborne (red squares) and ground-based (blue triangles) field studies near Los Angeles (solid blue triangles) and Claremont (open blue triangles). Values predicted by box model simulations [Fujita et al., 2013] (green symbols and dashed lines) are shown in comparison to the measurements. Assuming no loss of NOy species, LLS fits (solid black lines) of the long-term trends show an average decrease in PAN (−2.0% yr−1) and an average increase in HNO3 (3.3% yr−1) since 1973.

[53] Differences between measured and simulated ratios [Fujita et al., 2013] of HNO3/NOy and PAN/NOy are also illustrated in Figure 13. The differences in PAN/NOy ratio likely reflect the model's overestimate of HNO3.

5 Summary

[54] We quantify long-term trends in abundances and emissions ratios of precursors and secondary oxidation products in the SoCAB using data from a surface monitoring network, roadside monitors, ground-based field sites, and research aircraft. Consistency in average abundances and emission ratios determined from the various platforms over the past 50 years permits quantification of long-term trends in emissions in the South Coast Air Basin. Agreement between near-tailpipe measurements by roadside monitors, basin-wide measurements from aircraft, and local assessments from ground-based measurements and surface network sites confirms motor vehicles as the predominant source of precursor emissions in the SoCAB. Differences between long-term trends determined from the ambient measurements and emission inventories suggest that significant errors still exist in the inventories. A decrease in VOC/NOx ratio is observed since the 1960s and is a direct result of the faster decrease in VOC emissions compared to NOx. Decreasing maximum 8 h ozone concentration in the SoCAB is positively correlated with decreasing ozone precursor abundances and VOC/NOx ratio over the past 50 years. The well-established connection between O3 and PAN production allows use of the trends in NOx oxidation products as a tool for relating long-term changes in ozone production to changes in precursor emissions. The observed decreases in NOx oxidation products, PAN and HNO3, over time are consistent with decreasing VOC and NOx abundances. No change in enhancement ratios of O3 and (O3 + NO2) to (PAN + HNO3) over time suggests no detectable trend in ozone production efficiency since 1973, although there is wide variability in the results from different data sets. The trend in NOx oxidation products demonstrates an increase in the fraction of oxidized NOx since 1973, suggesting that atmospheric oxidation rates of NOx have increased over time as a result of the emissions changes in the SoCAB. Trends in the fractions of PAN and HNO3 relative to NOy show that HNO3 production has been increasingly favored over time.

[55] This analysis demonstrates that significant reductions in secondary pollutants have been the direct result of reduced precursor emissions, with the largest changes related to increased emission standards and improved technology in motor vehicles. Through the decades, VOCs were the initial primary target for emissions reductions in the SoCAB, followed more recently by increasing efforts to control NOx. As shown in our assessment and previous analyses, these control measures have been effective in reducing ambient ozone levels over the past 50 years. Stricter emissions standards for diesel-fueled vehicles are designed to further reduce NOx emissions [CARB, 2008b; McDonald et al., 2012]. Overall, the observations from this analysis suggest that reductions in ozone and other secondary pollutants in the SoCAB over the past 50 years are a direct result of decreasing abundances of VOCs and NOx and VOC/NOx ratio and that ozone continues to be responsive to local emissions control strategies in the Los Angeles basin.

Acknowledgments

[56] The authors thank L. Dolislager, B. Weller, and D. Oda from the California Air Resources Board for providing historical data from the SCAQMD and PAMS surface monitoring networks and various field studies in the SoCAB. We also thank G. Bishop and D. Stedman for help with conversion and interpretation of the roadside measurements.

Ancillary