ozone production rates as a function of no abundances and ... › ~thornton › publications_files...

Ozone production rates as a function of NOx abundances and HOxproduction rates in the Nashville urban plume

J. A. Thornton,1 P. J. Wooldridge,1 R. C. Cohen,1,2,3 M. Martinez,4 H. Harder,4

W. H. Brune,4 E. J. Williams,5,6 J. M. Roberts,5,6 F. C. Fehsenfeld,5,6 S. R. Hall,7

R. E. Shetter,7 B. P. Wert,7 and A. Fried7

Received 6 June 2001; revised 14 September 2001; accepted 14 October 2001; published 22 June 2002.

[1] Tropospheric O3 concentrations are functions of the chain lengths of NOx (NOx � NO+ NO2) and HOx (HOx � OH + HO2 + RO2) radical catalytic cycles. For a fixed HOxsource at low NOx concentrations, kinetic models indicate the rate of O3 productionincreases linearly with increases in NOx concentrations (NOx limited). At higher NOxconcentrations, kinetic models predict ozone production rates decrease with increasingNOx (NOx saturated). We present observations of NO, NO2, O3, OH, HO2, H2CO, actinicflux, and temperature obtained during the 1999 Southern Oxidant Study from June 15 toJuly 15, 1999, at Cornelia Fort Airpark, Nashville, Tennessee. The observations are usedto evaluate the instantaneous ozone production rate (PO3) as a function of NO abundancesand the primary HOx production rate (PHOx). These observations provide quantitativeevidence for the response of PO3 to NOx. For high PHOx (0.5 < PHOx < 0.7 ppt/s), O3production at this site increases linearly with NO to �500 ppt. PO3 levels out in the range500–1000 ppt NO and decreases for NO above 1000 ppt. An analysis along chemicalcoordinates indicates that models of chemistry controlling peroxy radical abundances, andconsequently PO3, have a large error in the rate or product yield of the RO2 + HO2 reactionfor the classes of RO2 that predominate in Nashville. Photochemical models and ourmeasurements can be forced into agreement if the product of the branching ratio and rateconstant for organic peroxide formation, via RO2 + HO2 ! ROOH + O2, is reduced by afactor of 3–12. Alternatively, these peroxides could be rapidly photolyzed underatmospheric conditions making them at best a temporary HOx reservoir. This resultimplies that O3 production in or near urban areas with similar hydrocarbon reactivity andHOx production rates may be NOx saturated more often than current modelssuggest. INDEX TERMS: 0317 Atmospheric Composition and Structure: Chemical kinetic andphotochemical properties; 0345 Atmospheric Composition and Structure: Pollution–urban and regional

(0305); 0365 Atmospheric Composition and Structure: Troposphere–composition and chemistry; 0368

Atmospheric Composition and Structure: Troposphere–constituent transport and chemistry; KEYWORDS:

Ozone, NO2, peroxy radicals, Nashville, OH, HO2

1. Introduction

[2] Photochemical O3 production is a complex functionof nitrogen oxide (NOx � NO + NO2) and hydrogen radical

(HOx � OH + HO2 + RO2) abundances, and of thepartitioning of NOx between NO and NO2 and HOxbetween OH, HO2, and RO2 [National Research Council,1991]. Ozone production is initiated by reactions thatproduce HOx. The production rate of new hydrogen radi-cals, PHOx, occurs primarily via O3 and H2CO photolysis:

ðR1Þ O3 þ hn ! O 1D� �

þ O2;

ðR2Þ O 1D� �

þ H2O ! 2OH;

ðR3Þ O 1D� �

þ N2;O2 ! O 3P� �

þ N2;O2;

ðR4aÞ H2CO þ hn þ 2O2ð Þ ! 2HO2 þ CO;

ðR4bÞ H2CO þ hn ! H2 þ CO:

Other sources of hydrogen radicals include photolysis ofH2O2 and HONO. The OH produced in (R2) usually leads

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 107, NO. D12, 10.1029/2001JD000932, 2002

1Department of Chemistry; University of California, Berkeley,California, USA.

2Department of Earth and Planetary Science, University of California,Berkeley, California, USA.

3Energy and Environment Technologies Division, Lawrence BerkeleyNational Laboratory, Berkeley, California, USA.

4Department of Meteorology, Pennsylvania State University, UniversityPark, Pennyslvania, USA.

5Aeronomy Laboratory, National Oceanic and Atmospheric Adminis-tration, Boulder, Colorado, USA.

6CIRES/University of Colorado, Boulder, Colorado, USA.7Atmospheric Chemistry Division, National Center for Atmospheric

Research, Boulder, Colorado, USA.

Copyright 2002 by the American Geophysical Union.0148-0227/02/2001JD000932$09.00

ACH 7 - 1

to the production of peroxy radicals (RO2) via the oxidationof VOC by OH in reactions (R5) and (R6). RO2 thenoxidize NO to NO2 in reaction (R7a) leading to theproduction of O3 in reactions (R8) and (R9):

ðR5Þ RH þ OH ! R� þ H2O;

ðR6Þ R� þ O2 þMð Þ ! RO2;

ðR7aÞ RO2 þ NO ! RO þ NO2;

ðR8Þ NO2 þ hn ! NO þ O 3P� �

;

ðR9Þ O 3P� �

þ O2 þMð Þ ! O3:

The alkoxy radical product (RO) of the RO2 + NO reactionusually reacts rapidly to produce HO2 which then oxidizesNO, regenerating OH and producing a second NO2molecule:

ðR10Þ RO þ O2 ! R0O þ HO2

ðR11Þ HO2 þ NO ! NO2 þ OH:Net R5ð Þ R11ð Þ : RH þ 4O2 ! R0O þ 2O3 þ H2O:

In most cases, the aldehyde or ketone product (R0O) ofthe net reaction continues to be oxidized in the samemanner yielding more than two O3 molecules per VOCemitted.[3] The rate-determining step in the chain propagation is

usually oxidation of NO by HO2 and RO2 in reactions(R7a) and (R11). Under conditions of high peroxy radical(HO2 + RO2) abundances and low NOx the steady stateconcentration of peroxy radicals is insensitive to variationin NOx because the primary chain terminating reactions ofthe HOx catalytic cycle are the HOx + HOx reactions:

ðR12Þ OH þ HO2 ! H2O þ O2;

ðR13Þ HO2 þ HO2 þ Mð Þ ! H2O2 þ O2;

ðR14Þ HO2 þ RO2 ! ROOH þ O2;

O3 production in this chemical regime is said to be NOx-limited and the O3 production rate increases approximatelylinearly with NOx. Reactions of HOx with NOx lead to chaintermination of both the HOx and NOx catalytic cyclesthrough nitric acid and alkyl nitrate formation:

ðR15Þ OH þ NO2 þ Mð Þ ! HNO3;

ðR7bÞ RO2 þ NO þ Mð Þ ! RONO2:

At some point, as NOx increases, these termination stepsbecome faster than the HOx–HOxreactions (R12)–(R14). Inthis regime, O3 production slows, decreasing with increas-ing NOx. O3 production in this high NOx regime is said tobe NOx-saturated (or VOC-limited).[4] In addition to the role of NOx as a control over the

abundance of peroxy radicals, PO3 is affected by the rate of

HO2 and RO2 production. The role of PHOx in the balancebetween NOx-limited and NOx-saturated O3 production isillustrated by considering a constant NOx level. As PHOxincreases, the rate of OH + RH increases. It follows thatorganic peroxy and hydroperoxy radical concentrationsincrease with resulting increases in O3 production rates.As PHOx increases, the corresponding increase in peroxyradicals enhances the relative importance of HOx–HOxcatalytic chain termination steps in reactions (R12)–(R14)over the competing HOx–NOx reactions because the ratesof HOx–HOx reactions increase as the square of HOxconcentrations. This in turn shifts the peak O3 productionrate, which by definition occurs at the crossover betweenNOx-limited and NOx-saturated behavior, to higher levelsof NOx. Of course, in the polluted boundary layer the issueis still more complicated because the assumption of constantNOx does not hold. Increases in HOx abundance canaccelerate the removal of NOx by conversion to the long-lived reservoirs HNO3 and organic nitrates. In the uppertroposphere, Jaegle et al. [1998, 2001] show that PHOx andNOx are strongly correlated, making it challenging to obtainobservations over a range of NOx at a fixed PHOx or viceversa.[5] The crossover point between NOx-limited and NOx-

saturated regimes remains important from both a regula-tory standpoint and as a test of our understanding of thechemistry controlling O3 production. Regulatory concernshave focused on whether NOx-saturated inner cities willhave higher ozone concentrations if modest NOx controlsare enacted [e.g., Cardelino and Chameides, 1990].Analyses of atmospheric models have focused on identi-fying observable quantities that can be used to indicatewhether ozone concentrations at a particular locationresult from NOx-limited or NOx-saturated photochemistry[Sillman, 1995; Kleinman et al., 1997; Sillman et al.,1997; Tonnesen and Dennis, 2000a, 2000b; Trainer et al.,2000].[6] In this article, we use in situ measurements of NO2,

NO, OH, HO2, O3, H2CO, actinic flux, relative humidity,and temperature made at Cornelia Fort Airpark (CFA), 8 kmNE of downtown Nashville, Tennessee, during the 1999Southern Oxidants Study (SOS 99) to quantify instantane-ous O3 production rates and to empirically evaluate theinfluence of NOx and PHOx on O3 production rates. Weexamine the crossover point between NOx-limited and NOx-saturated regimes, and we use chemical coordinates [Cohenet al., 2000; Lanzendorf et al., 2001] to evaluate the relativeaccuracy of the catalytic chain termination steps in currentphotochemical models.

2. Measurements

[7] Continuous in situ measurements of NO2, NO, O3,OH, HO2, H2CO, actinic flux, relative humidity, and temper-ature were made during SOS 99 at CFA. These data and anextensive suite of observations of other chemical and phys-ical parameters may be obtained by contacting the individualinvestigators. A contact list is available at http://www.al.noaa.gov/WWWHD/pubdocs/SOS/SOS99.html. Observa-tions were made 10–12 m above the ground from a 12-mwalkup tower. These measurements were collected at ratesbetween 0.03 and 1 Hz. The analysis presented here uses

ACH 7 - 2 THORNTON ET AL.: PO3 DEPENDENCE ON [NOX] AND PHOx

1-min averages of the data with the exception of OHconcentrations. A five-point running average of the 1-minOH data is used here. The measurements were synchronizedin time using temporal structures observed in the measuredNO2/NO ratio and JNO2.[8] NO2 measurements were made using laser-induced

fluorescence (LIF) [Thornton et al., 2000], NO was meas-ured by an NO-O3-chemiluminescence instrument [Williamset al., 1998], and O3 was measured by UV absorbance[Ridley et al., 1992a]. OH was measured directly by LIFand HO2 was converted to OH by NO followed by LIFdetection of the resulting OH [Mather et al., 1997]. H2CO

measurements were made using tunable diode laser absorp-tion spectroscopy [Fried et al., 1998]. Spectrally resolvedmeasurements of the downwelling solar actinic flux span-ning a wavelength region from 280 to 420 nm were madeusing a 2p dome coupled by optical fibers to a monochro-mator. The downwelling actinic flux measurements togetherwith an in situ measurement of the upwelling radiation wereused to calculate the photolytic rate constants used here[Shetter and Muller, 1999]. The upwelling radiation wasconsistent with an albedo of 4%. Relative humidity andtemperature were measured using a commercial probe(Campbell Scientific).[9] Figure 1 shows measured and derived species versus

fractional day of the year for a typical day (July 9, 1999) atCornelia Fort Airpark. O3 peaks at 60 ppb on this day. Themaximum and minimum peak O3 concentrations observedduring the study were �110 and �30 ppb, respectively. Thesum total peroxy radical concentrations ([HO2 + RO2]PSS,triangles) were derived from the NOx-steady state equationsas described below. Water vapor mixing ratios, plotted inparts per thousand (ppth) in Figure 1, were calculated frommeasured temperature and relative humidity and rangedfrom 10 to 30 ppth during the course of the study.[10] Table 1 shows the measurement method and accu-

racy (1s) for each species used in this analysis. Throughoutour analysis we average over enough observations thatprecision errors from the instrumentation can be neglected.We focus our analysis on daytime data by excluding datawhere JNO2 < 5 105 s1. Reaction rate constants arecalculated point by point as functions of temperature, totalnumber density, and where necessary, water vapor numberdensity.

3. Instantaneous O3 Production Rate

[11] NO and NO2 rapidly interconvert in reactions thatform a null catalytic cycle and in reactions that catalyti-cally produce O3. In the null cycle, O3 oxidizes NO toNO2, and photolysis of NO2 in reaction (R8) is followedby reaction (R9) to regenerate an O3 molecule with a yieldnear unity:

ðR16Þ NO þ O3 ! NO2 þ O2;

ðR8Þ NO2 þ hn ! NO þ O 3P� �

;

ðR9Þ O 3PÞ þ O2 þMð Þ ! O3:�

The rate constants for reactions of HO2 and RO2 with NO toproduce NO2 are nearly 1000 times faster than the rateconstant of NO with O3. Although HO2 and RO2concentrations are �1000 times lower than O3 concentra-tions, these reactions provide a mechanism for theproduction of O3:

ðR11Þ NOþ HO2 ! NO2 þ OH;

ðR7aÞ NOþ RO2 ! NO2 þ RO;

ðR8Þ NO2 þ hv ! NOþ Oð3PÞ;

ðR9Þ Oð3PÞ þ O2ðþMÞ þ O3;

Figure 1. Measured and derived species are plottedversus time for 9 July 1999, a typical day at CorneliaFort Airpark. NO2 (squares), NO (circles), and O3(triangles) measurements are plotted in ppb. JNO2 (thickblack line, left axis, s1) and JO3 ! O1D (gray line, rightaxis, s1) are derived from solar actinic flux measurements.[HO2 + RO2]PSS (triangles, ppt) are derived from the NOx-steady state assumption, and OH (small squares, 5-minrunning average, ppt) and HO2 were measured by LIF.H2CO (squares) and PAN (circles) measurements areplotted in ppb. Temperature (squares) is in degrees celsiusand the calculated water mixing ratio (circles) is in partsper thousand (ppth).

THORNTON ET AL.: PO3 DEPENDENCE ON [NOX] AND PHOx ACH 7 - 3

The gross rate of new O3 produced by this mechanism,PO3, is

PO3 ¼ kHO2þNO HO2½ þXi

kRiO2þNO RiO2½ !

NO½ ð1Þ

where kHO2+NO is the rate constant for reaction of HO2 withNO in reaction (R11), and the sum is over the suite oforganic peroxy radicals with R group Ri and the associatedrate constants kRiO2+NO for reaction (R7a). In the absence offresh NOx emissions or rapid NOx removal a photochemi-cal steady state between NO and NO2 is rapidly established(t � 100 s). The sum total peroxy radical concentration,[HO2 + RO2]PSS, can be inferred using the steady stateequation:

keff ½HO2 þ RO2PSS ½NO þ kNOþO3½NO½O3 ¼ JNO2 ½NO2 ð2Þ

where keff is an effective rate constant for the reactions ofHO2 with NO and of RO2 with NO, kNO+O3 is the rateconstant for reaction of NO with O3 in reaction (R16), andJNO2 is the rate constant for NO2 photolysis in reaction(R8). This inference assumes the chemistry describedabove is complete. The subscript PSS (photostationarystate) is used here to distinguish inferred quantities fromdirect in situ observations. Reactions of NO with mostorganic peroxy radicals to yield NO2 have been shown tooccur at similar rates to those with HO2 [Atkinson, 1994;DeMore et al., 1997]. In this analysis we assumekeff = kHO2+NO. Note that the derived product keff [HO2 +RO2]PSS is expected to be more accurate than the absolutesum peroxy radical concentrations. At CFA, there weremea-surements of HO2 but not of RO2. We use thedifference between the measured HO2 and the calculatedsum total peroxy radical concentrations to estimate theamount of RO2:

RO2½ PSS¼ HO2 þ RO2½ PSS HO2½ OBS: ð3Þ

Typical [RO2]PSS/[HO2]OBS at CFA ranged between 0 and2 with extremes of approximately 1 and �4.[12] Although some of our conclusions depend on the

estimates of RO2 from equations (2) and (3), PO3 can bederived from the observations without any assumptionsabout reactions of HO2 or RO2 so long as the only other

processes involved are NO2 photolysis and the reaction ofNO with O3:

PO3 ¼ JNO2 NO2½ kNOþO3 NO½ O3½ ð4Þ

[13] Observations and equations (1)–(4) have been used toinfer RO2 radical abundances and to estimate instantaneousozone production rates for a range of urban, rural, and remoteenvirons [Kelly et al., 1980; Parrish et al., 1986; Ridley et al.,1992b; Cantrell et al., 1993a; Kleinman et al., 1995; Car-penter et al., 1998;Frost et al., 1998;Patz et al., 2000]. Figure2a shows the calculated PO3 plotted versus time of day, andFigure 2b shows the calculated PO3 at CFA for three consec-utive days, June 25–27, 1999. PO3 inferred from equation (4)peaks in the afternoon and reaches zero in the early morningand late evening as the two terms become nearly equivalent.Negative PO3 is often observed during the period of 0600–0900 local standard time (LST). This could indicate that theair sampled during this time of the day was frequently not insteady state or that other processes not included in equation(4) and involving NOx or O3 take place at a rate sufficient tochange the photostationary state. There is also day-to-day andminute-to-minute variation in the magnitude of PO3 becauseof cloud cover and changes in VOC, NOx, and O3 abundan-ces. Some of the variation is due to noise in themeasurements.Averaging over the large data set eliminates the importance ofinstrument noise in our analysis.[14] The steady state assumption is not valid in the

following situations: (1) when measurements are made closeto NOx sources such that a steady state is not achieved priorto sampling the air mass because of insufficient reactiontime, (2) when there are fast changes in actinic flux, and (3)when the air sampled is sufficiently inhomogeneous so thatmixing drives NO and NO2 out of steady state in an air mass.Cornelia Fort Airpark is nearly 2 km from a major freewayand 8 km from downtown Nashville. These are the closestlarge NOx sources. The time for a well-mixed air mass toreach CFA from these sources is at least 10 min for anaverage wind speed in the range of 1–4 m/s. This is severalNOx lifetimes and consequently, with respect to time forchemistry to occur downwind of sources, the NOx partition-ing in air masses observed at Cornelia Fort should usually bein a steady state. Small NOx sources such as emissions fromthe surface could perturb the NOx partitioning from steadystate, leading to PO3 and [HO2 + RO2]PSS that are biasedhigh compared to the true values. To investigate the magni-tude of this potential bias at CFA, we assume that the

Table 1. Continuous, in Situ, Ground-Based Measurements of NO2, NO, O3, OH, HO2, H2CO, and Solar Actinic Flux Made From June

15 to July 15, 1999, at Cornelia Fort Airpark, Nashville, Tennessee as Part of the SOS99 Campaigna

Species Method Accuracy, % , 1s References

NO2 laser-induced fluorescence (LIF) 5 Thornton et al. [2000]NO O3-chemiluminescence 4 Williams et al. [1998]O3 UV-absorbance 5 Ridley et al. [1992a]OH LIF 20 Mather et al. [1997]HO2 NO titration/OH LIF 20 Mather et al. [997]H2CO tunable diode laser absorption spectroscopy 10 Fried et al. [1998]Solar actinic flux spectral radiometer 5 Shetter and Muller [1999]

aObservations were made at �11 m above ground on a 12-m walkup tower. The measurement method and total uncertainty (1s) for each species arepresented.


observed NO concentrations are a combination of NO that isin steady state and NO that is freshly emitted from thesurface, i.e., [NO]OBS = [NO]PSS + [NO]Surf. A typicalnighttime value of [NO]OBS at CFA is �20 ppt. If this isentirely due to NO emitted from the surface (i.e., [NO]Surf= 20 ppt), and if the source strength is constant over a24-hour period, we estimate that our derived PO3 and[HO2 + RO2]PSS are biased high by �5%. The size of thisbias scales nearly linearly with increasing estimates of NOconcentrations that are not in steady state. Propagating a 5%change in PO3 or [HO2 + RO2]PSS through our analysis has anegligible effect on the conclusions of this paper.[15] During the daytime, NOx at Cornelia Fort was

typically 1–5 ppb. On occasions, NOx rose above 15 ppb,

and these high NOx points are correlated with negative PO3.The arrival of these plumes is also strongly correlated withtimes when gradients in NOx and O3 were larger than 10%/min. NO2 production in these plumes through reaction ofHO2 with NO was positive. At NOx greater than 25 ppb,where the steady state calculation yields [HO2 + RO2]PSSless than 3 ppt, the [HO2]OBS is often significantly largerthan [HO2 + RO2]PSS. The high NOx points were primarilyobserved in the morning, possibly before solar heatingcould begin to effect vigorous turbulence in the boundarylayer. Therefore high NOx seems to be a reasonable indi-cator for mixing of heterogeneous air masses causing NOxto be out of photostationary state. As we are relying on theaccuracy of the photostationary state equation for our

Figure 2. (a) One-minuted averaged PO3 calculated using equation (4) plotted versus hour of day(local standard time). All of the data obtained at CFA is shown. (b) Three consecutive days, 25–27June 1999, of PO3 versus time.


conclusions, we exclude data where NOx > 15 ppb fromfurther analysis. Choosing a cutoff of NOx less than 15 ppbis somewhat arbitrary, excluding data above either 10 or 20ppb does not significantly alter our conclusions. All dataused in the analyses presented from here on have beensubjected to the two selection criteria: JNO2 > 5 105 s1and NOx < 15 ppb. Other selection criteria, when noted, arealways used in addition to these two. For the available PO3data calculated from equation (4), the JNO2 selection criteriaremoves 2175 points (�15) and the 15 ppb criteria elimi-nates another 2317 points (20% of the data that remainsafter the JNO2 filter). We also impose an additional filter tothe PO3 data to remove extreme outliers by requiring eachpoint be within 3 standard deviations from the mean at thattime of day. This filter excludes an additional 102 points(1%) from the PO3 data.[16] The accuracy of PO3 calculated using equation (4)

depends on the accuracy of the products: JNO2[NO2]and kNO+O3[NO][O3]. The accuracy of the productkNO+O3 [NO][O3] is estimated to be ±12%, assuming system-atic errors in themeasurements (see Table 1) and rate constant[DeMore et al., 1997] add in quadrature. Prior studies haveimplicated the accuracy of NO2 measurements as the largestsource of error in PO3 calculated from equation (4) [e.g.,Frost et al., 1998]. The SOS 99 campaign included threedifferent techniques for measuring NO2: laser-induced fluo-rescence [Thornton et al., 2000], photolysis-chemilumines-cence [Williams et al., 1998], and differential opticalabsorption spectroscopy [Alicke et al., 2000]. All threemethods agreed to within ±5% (1s) on average over theentire campaign [Wooldridge et al., 2000]. The LIF NO2measurements used here are therefore thought to be accurateto 5%. The uncertainty in the J value is dominated byuncertainty in the cross-section and quantum yield (±20%[DeMore et al., 1997]). The accuracy of the productJNO2[NO2] is estimated to be ±21% (1s). The maximumsystematic uncertainty in the derived PO3 is therefore ±32%.[17] The systematic error in the difference JNO2[NO2]

kNO+O3[NO][O3] can be reduced by forcing reasonablebehavior at sunset as described below. In the evening athigh solar zenith angles, the rate of O3 production at CFAdrops to near zero as JNO2[NO2] � kNO+O3[NO][O3] and theproduct kNO+O3[NO][O3] is typically 5–10 times higherthan kHO2+NO[NO][HO2]. For this high solar zenith angledata we require that the fractional conversion of NO to NO2by organic peroxy radicals be on average greater than orequal to zero:

kHO2þRO2 RO2½ NO½ kHO2þRO2 RO2½ NO½ þ kHO2þNO HO2½ NO½ þ kNOþO3 O3½ NO½

� 0:

ð5Þ

This constraint is equivalent to requiring [HO2 + RO2]PSS tobe at least equal to [HO2]OBS (see equation (3)) on averageduring the morning and evening, and it couples all of thesystematic errors together. We find that a negative bias in theinferred peroxy radical concentration is present at high solarzenith angles unless we increase the product JNO2[NO2] by11%, decrease the product kNO+O3[NO][O3] by 11% or makean equivalent change in a combination of the two terms. Thisadjustment is well within the uncertainty in the rate constantsand measurements. We extend this adjustment to the entire

data set by increasing JNO2 by 11%. The PO3 data plotted inFigure 2 include this adjustment to JNO2.[18] PO3 is usually the difference between two large

numbers with consequent high uncertainty in any individualvalue. While the random uncertainty in individual PO3values can be averaged away, the average value of PO3carries the effects of any systematic errors in the termsJNO2[NO2] or kNO+O3[NO][O3]. To reduce the influence ofthese systematic errors we focus on the derivatives of PO3with respect to PHOx and NO. These derivatives do notcompletely eliminate the effects of systematic errors. Forexample, the largest term in PHOx is the measured O3concentration and PO3 depends directly on NO. Never-theless, the effect of measurement error is damped byfocusing on the derivatives because systematic errors inthe measurements are additive or multiplicative constantswhich at most shift PO3 with respect to NO or PHOx but donot change the functional form of the derivative @PO3/@NOor @PO3/@PHOx. We also observe a systematic error due to abreakdown the photostationary state assumption that iscorrelated with NO. If our cutoff of 15 ppb NOx is notlow enough, our conclusions could be biased at high NOx.

4. PO3 Dependence on PHOx and NO

[19] We take the total primary HOx radical productionrate to be the sum of the O3 and formaldehyde photolysischannels leading to OH or HO2 via reactions (R1)–(R3) and(R4a). The total production rate for primary HOx radicals,PHOx, at CFA is

PHOx ¼ 2JO3!O1D O3½ k2 H2O½

k2 H2O½ þ k3 N2 þ O2½ þ 2JH2CO H2CO½

ð6Þ

where JO3!O1D and JH2CO are the photolytic rates constantsfor reactions (R1) and (R4a), respectively, and k2 and k3 arethe rate constants for reactions (R2) and (R3), respectively.JO3 and JH2CO derived from solar actinic flux measurements,and measurements of O3, H2CO, relative humidity, andtemperature are used to calculate PHOx explicitly. Values forPHOx ranged from 0 to greater than 1 ppt/s.[20] During the peak of solar radiation at CFA the

photolysis of O3 in reactions (R1)–(R3) comprised �70%on average of PHOx. We assume peroxyacyl nitrate (PAN)was not a net source of peroxy radicals at CFA. Its time rateof change was small compared to its lifetime with respect tothermal decomposition (25 min at 300 K, 760 Torr), andthus PAN was most likely in thermal equilibrium with NO2and the peroxyacyl radical. Furthermore, the reaction of theperoxyacyl radical (PA) with NO2 to reform PAN is onaverage 2 times faster than the reaction PA + NO ! NO2 +products under typical midday conditions at CFA. The rateof PA + NO to yield NO2 corresponds to �5–10% of thetotal [HO2 + RO2]PSS + NO ! NO2 inferred from equation(2). Measurements of HONO [Alicke et al., 2000] at CFAshow that over most of the day this term is small. The largeconcentrations of HONO coincide with the early morningrush hour traffic during the period where we most oftenobserve NOx to be greater than 15 ppb. H2O2 and largeraldehyde (C2–C9 straight-chain aldehydes) photolysistogether are typically �15% of the major HOx sourcesincluded in our analysis. These processes become a larger


fraction in the late afternoon when PHOx is small. However,there are not enough measurements of HONO, H2O2 orlarger aldehydes during the day to include them in thisanalysis. Because we focus on high PHOx, omitting theseprocesses does not affect our conclusions.[21] The PO3 data shown in Figure 3 are the data from

Figure 2a separated into a high (0.5 < PHOx < 0.7 ppt/s),

moderate (0.2 < PHOx < 0.3 ppt/s), and low (0.03 < PHOx< 0.07 ppt/s) PHOx regime. These bins are representative ofthe range of the observations. Observations at other valuesof PHOx are consistent with the data shown here. In themoderate and low PHOx bins we also restrict our analysisto times later than 0900 LST so that we have an accurateestimate of PO3 and PHOx. Consequently, we remove 93

Figure 3. (a) The averaged (solid circles) and the raw (open squares) PO3 from the high PHOx regimeplotted versus [NO]. (b) Averaged PO3 plotted versus [NO] at CFA. PO3 data were placed into threePHOx bins: high (0.5 < PHOx < 0.7 ppt/s, circles), moderate (0.2 < PHOx < 0.3 ppt/s, squares), and low(0.03 < PHOx < 0.07 ppt/s, triangles), and then averaged as a function of NO. All three PHOx regimesdemonstrate the expected generic dependence on NO, PO3 increases linearly with NO for low NO(600 ppt NO). The crossoverpoint between NOx-limited and NOx-saturated O3 production occurs at different levels of NO in thethree PHOx regimes.


points (�8%) from the moderate PHOx bin and 350 points(�37%) from the low PHOx bin. This time of day selectiondoes not change the number of points in the high PHOx binas all PHOx data greater than 0.5 ppt/s occurred after 0900LST. The potential for the exclusion of data to bias ourconclusions is a concern. In this respect, the high PHOxdata are the most reliable, and both the raw and averagedPO3 data in this regime are shown in Figure 3a. Of the1018 observations where PHOx ranges between 0.5 and 0.7ppt/s (subject to the requirement JNO2 > 5 105 s1) theadditional selection criteria (NOx < 15 ppb and PO3 within3-sigma of the mean) remove another 15 points from theanalysis. Also, the lack of specific information about H2O2or HONO contributions to PHOx, and the influence of PANare all minimized for this data because of the largecontribution of O(1D) + H2O to PHOx.[22] The data in each of the PHOx regimes were aver-

aged into bins of NO concentrations ranging in size from10 ppt at the lowest NO concentrations to 250 ppt at thehighest NO levels. Points are shown where five or moreobservations were available prior to averaging. In the highPHOx regime (circles), PO3 increases approximately line-arly with NO at an average slope of �0.008 molecule O3(molecule NO)1 s1 between 100 and 500 ppt, becomingindependent of NO over the range of 750–1100 ppt. Atstill higher NO, PO3 begins to decrease. In the moderate(squares) and low (triangles) PHOx regimes, PO3 increaseswith NO less steeply at slopes of �0.005 molecule O3(molecule NO)1 s1 and � 0.004 molecule O3 (moleculeNO)1 s1, and becomes independent of NO at �400 and�200 ppt, respectively.[23] Qualitatively, the expected dependence of PO3 on

PHOx and NO is demonstrated in Figure 3:1. For fixed PHOx, PO3 increases, levels, and then (most

clearly at high PHOx) decreases as NOx increases.2. For fixed NOx, ozone production rates increase with

PHOx.3. The NO concentration where PO3 stops increasing

linearly with NO shifts to higher NO as PHOx increases.These features of the role of NOx and PHOx on O3production have been demonstrated directly from measure-ments made in the upper troposphere [Jaegle et al., 1998;Wennberg et al., 1998; Jaegle et al., 1999]. The observa-tions we show in Figure 3 are the first to separate NOx andHOx influences under the high NOx and VOC regime of anurban setting and the first to illustrate a complete PO3 versusNOx curve at a single value of PHOx.

5. NOx Limited Versus NOx Saturated PO3

[24] For a single value of PHOx the crossover pointbetween NOx-limited and NOx-saturated behavior corre-sponds to a unique NO concentration where @PO3/@NO = 0The position of the crossover point between NOx-limitedand NOx-saturated photochemistry is an extremely sensitivetest of our understanding of the response of O3 photo-chemistry to NOx and HOx. Models that do not accuratelypredict its location with respect to NOx will likely predictthe incorrect sign of the change in PO3 due to hypotheticalchanges in future NOx and VOC emissions. Under low NOxconditions, HOx self-reactions (R12)–(R14) are the domi-nant chain termination. At CFA, peroxide formation in

reactions (R13) and (R14) represents the most importantself-loss pathway with OH + HO2 typically contributingless than 10% of the HOx self-loss during the daytime.Chain terminating RO2 cross reactions are too slow to beimportant at CFA. Organic hydroperoxide formation, RO2 +HO2 ! ROOH, often dominates the HOx loss becausekHO2+RO2 is approximately twice kHO2 + HO2 and RO2/HO2is often greater than 2. We estimate that the fraction of HOxloss due to surface deposition at CFA is less than 2% of thetotal loss assuming deposition velocities for OH, HO2, andRO2 to be 5 cm/s and a 1-km boundary layer height. Weomit surface deposition of HOx from further analysis.Under high NOx conditions the reactions removing NOx,specifically nitric acid and alkyl nitrate formation in reac-tions (R15) and (R7b), become the dominant sinks for freeradicals. Assuming a 3% yield for (R7b) relative to (R7a) +(R7b), and using the rate constant kHO2+NO, the rate of alkylnitrate formation at CFA under high NOx conditions is�15% of the rate of OH + NO2 in reaction (R15). Low NOxconcentrations (

[HO2 + RO2]PSS/[OH] ratio observed at CFA. Rateconstants used in the model are from Sander et al. 2000,and references therein) (see Table 2). Figure 4 shows themodel results from a calculation where PHOx = 0.6 ppt/s.PO3 (circles, ppt/s, divided by a scale factor of 5) and [HO2](down triangles, ppt, divided by a scale factor of 100) areshown as symbols. PO3 in the model increases linearly atlow NO, begins to slow at �700 ppt, peaks at 1125 ppt, andbeyond 1250 ppt PO3 decreases with further increases inNO. Solid lines show the initial slope of PO3 and thefractional loss of HOx due to HHLoss through H2O2formation (black curve) and the fractional loss due toNHLoss through HNO3 formation (gray curve). Themodeled HHLoss and NHLoss curves are equal at �900ppt NO, which is 25% below the 1125 ppt NO value where@PO3/@NO = 0. We examined a wide range of COconcentrations, varied kHO2+HO2 by a factor of 10 in themodel, and increased PHOx to 1.2 ppt/s and found that thedifference between the point where HHLoss = NHLoss andthe peak PO3 in the model ranged from 17 to 27%. Allowingthe CO concentration to increase proportionally with an

increase in NO shifted the NO concentration whereHHLoss = NHLoss to �30% lower than the position ofthe peak PO3. However, although there is some expectationthat hydrocarbon and NOx sources are proportional to oneanother, there is no evidence that hydrocarbons and NOconcentrations increased proportionally at CFA, and wetherefore use a difference of 25% as a guide. Models withlower HO2 or kHO2+HO2 exhibit PO3 curves that peak at lowerNO. The model does not explicitly treat RO2, consequently,its predictions should only be considered illustrative andthey should not be expected to accurately reproduce both theamplitude and position of the peak PO3 in the observations.[26] Figure 3a shows the PO3 data (open squares) for the

high PHOx regime at CFA plotted versus NO concentrationsalong with the averaged data (solid circles). The data arenoisy, making it difficult to exactly specify the NO con-centration where @PO3/@NO = 0. However, the basic shapeis similar to that of the simple model. PO3 increases at lowNO up to 600 ppt. Beyond 1100 ppt, a decrease clearlyemerges. This decrease is due to a decrease in inferred RO2.The contribution of the HO2 + NO reaction to PO3

Table 2. Rate Constants Used for the Reactions Involved in the Two Models of LHOx Presented in Figures 6 and 7

Chain Terminating Reaction Rate Constant For Figure 6a Rate Constant For Figure 7a

OH + HO2 ! H2O + O2 1.1 1010 (a) 1.1 1010HO2 + HO2 ! H2O2 + O2 6.1 1012 (a) 6.1 1012HO2 + RO2 ! ROOH + O2 1.0 1011 (b) (.08)*1.0 1011RO2 + NO ! RONO2 0 (0.03)*8.1 1012 (a)OH + NO2 ! HNO3 9.0 1012 (c) 9.0 1012

aThe rate constant for RO2 + NO is that recommend for HO2 + NO(a). All rate constants in the table are estimated for 298 K, 760 Torr, and for the case of

HO2 + HO2 for 2% [H2O]. Units are cm3 s1, estimated for 298 K, 760 Torr, and for the case of HO2 + HO2, for 2% [H2O]. Throughout the analysis we

use measured temperature and relative humidity to calculate each rate constant explicitly as functions of temperature, pressure, and water vapor. Seereferences for temperature and pressure dependences used in the analysis: a, DeMore et al. [1997]; b, Atkinson [1994]; and c, Sander et al. [2000].

Figure 4. Modeled PO3 (circles, ppt/s), [HO2] (down triangles, ppt), HHLoss/LHOx (thick black curve),and NHLoss/LHOx (thick gray curve) plotted versus [NO] for a calculation where PHOx = 0.6 ppt/s. Seetext for description of model. O3 was fixed at 75 ppb and CO at 60 * O3. The NO2/NO ratio was heldconstant at 7. PO3 is plotted here reduced by a factor of 5, and [HO2] is plotted reduced by a factor of 100.


increases approximately linearly from 1 ppt/s at 150 pptNO to 3.5 ppt/s at 500 ppt NO. Between 500 and 1000 pptNO, PO3 from the reaction of HO2 + NO increases moreslowly to �4 ppt/s. At higher NO the rate of HO2 + NO isessentially constant while total PO3 is decreasing. Withinthe noise of the measurements the inferred PO3 we presentin Figure 3 is always higher than that calculated directlyfrom HO2 and NO measurements. The maximum in PO3 asshown in Figure 3 occurs approximately at 800–1000 pptNO. Therefore, using Figure 4 as a guide, we expectthe point where HHLoss = NHLoss to be in the range of600–750 ppt NO.[27] A complete evaluation of the HHLoss and NHLoss

terms requires an explicit treatment of the RO2 chemistrythat we omitted from the simple model. We evaluate thebalance between HHLoss and NHLoss using observationsof NO2 and NO, of OH and HO2, and the inferred RO2concentrations in equations (2) and (3). There are fewmeasurements of rate constants for reactions between HO2and RO2 (kRO2+HO2) for R larger than 3 carbons [e.g.,Villenave and Lesclaux, 2001] and even fewer quantitativestudies of product yields. We use the rate constantrecommended by Atkinson [1994] for generic RO2(kRO2+HO2 � 1 1011 cm3 s1 at 298 K) and anROOH yield of unity. The rate constants for the HO2 +HO2 in reaction (R13) and OH + NO2 in reaction (R15)reactions were taken from DeMore et al. [1997] andSander et al. [2000], respectively, and we include theH2O dependence of the HO2 + HO2 reaction (kHO2+HO2 �6.1 1012 cm3 s1 and kOH+NO2 � 9 1012 cm3s1 at 298 K, 760 Torr, 2% water vapor). The yield ofalkyl nitrate formation via reaction (R7b), a, is defined ask7b/(k7a + k7b). To demonstrate the ability of our approach

to discern small errors in the photochemical model, weinitially set a to zero. Figure 5 shows the two fractionsHHLoss/LHOx and NHLoss/LHOx plotted versus NO forthe high PHOx regime (0.5 < PHOx < 0.7 ppt/s) as well asPO3 from the same PHOx regime now scaled by a factorof 7. The data obtained at CFA included air whereHHLoss (open squares) is the primary HOx sink and airwhere NHLoss (gray triangles) is the primary sink.Consequently, the data provides separate and distinctinformation on the accuracy of models in both regimes.If the model represented by equations (8) and (9), if ourderived PO3 are accurate, and if the calculations shown inFigure 4 are a reasonable guide, the range of NOconcentrations at which the two loss rates are equalshould correspond to 25% lower NO than where the peakin PO3 is observed. This crossover should occur at�600–750 ppt. For the high PHOx range shown in Figure5 the HHLoss is equal to the NHLoss at �1000 ppt. Thisis not unambiguously inconsistent with the data, but it isat the high end of the �600–750 ppt range for thecrossover derived from Figure 3 and 4. To investigateand quantify the similarities and differences in the modeland measurement descriptions of the functional depend-ence of PO3 on NOx and PHOx, we examine the HOxbalance along chemical coordinates.

6. Chemical Coordinates

[28] In recent papers, Cohen et al. [2000] and Lanzendorfet al. [2001] have proposed that chemical coordinatesprovide a means to systematically organize a large ensembleof measurements. These papers discuss how chemicalcoordinates can be used to identify the information content

Figure 5. HOx loss rates divided into two categories, HOx-HOx (HHLoss) and HOx-NOx (NHLoss)reactions. The two fractions, HHLoss/LHOx (open squares) and NHLoss/LHOx (gray triangles) are plottedversus [NO] for the high (0.5 < PHOx < 0.7 ppt/s) PHOx observations. The two fractions represent thecontributions of peroxide formation and nitric acid formation, respectively, to the total free radical lossrate, LHOx. PO3 (solid circles) for the same PHOx regime from Figure 3a are shown here divided by a scalefactor of 7.


of a suite of measurements and suggest they are especiallywell suited to determining the rates of fast photochemicalreactions by comparison of a large ensemble of atmosphericobservations with highly constrained models. Chemicalcoordinates also serve the purpose of turning attention awayfrom geophysical variables and long-lived chemical speciesthat play no direct role in the chemistry on the timescale ofinterest and aid in comparing models across gradients ofspecific components of a reaction set. In recent examples weshowed that an ensemble of stratospheric measurements arecapable of (1) reducing the estimated uncertainty in the ratesof reactions controlling partitioning of NOx and NOy, and ofNO and NO2 [Cohen et al., 2000], (2) indicating that thecurrent recommendations for some of the key reactionscontrolling the portioning of OH and HO2 are likely in errorat low temperatures [Lanzendorf et al., 2001], and (3) con-firming the mechanism for NOx control over the partitioningof chlorine between ClO and ClONO2 [Stimpfle et al., 1999].[29] Here, we seek to determine whether the discrepancy

between the modeled and observed crossover between NOx-limited and NOx-saturated ozone production is a significanterror and if so to determine if it arises from modeledperoxide formation rates that are too fast, nitric acid andalkyl nitrate formation rates that are too slow, or somecombination of the two. We compare PHOx and LHOx usingthe following set of chemical coordinates: (1) the primaryHOx production rate (PHOx), (2) HHLoss, (3) the contribu-tion of organic peroxide formation in reaction (R14) to thetotal loss rate, FHO2+RO2, and (4) the contribution of nitricacid formation in reaction (R15) to the total loss rate,FOH+NO2, where

FHO2þRO2 ¼2kHO2þRO2 HO2½ RO2½

LHOx; ð12Þ

FOHþNO2 ¼kOHþRO2 OH½ NO2½

LHOx: ð13Þ

The values of FHO2+RO2 and FOH+NO2 range from 0 to 1.If, for example, data are available where either coordinate,Fx = 1, then those measurements were made underconditions where the process x is the sole reactioncontrolling HOx loss in the model. Comparison of modelsand measurements at that point in the chemical coordinatesis a direct measure of an error in the rate of that process oran indication that a process not represented by the model isimportant. Conversely, data obtained where Fx = 0 containno information about the accuracy of process x in the model.The primary HOx production rate, PHOx, does not isolate asingle variable in the same sense; however, this coordinatehelps to indicate missing chemistry not included in themodel and to provide a general picture of the accuracy ofthe model over a range of the production and loss processes.[30] Under typical daytime conditions the HOx lifetime is

�10. Thus HOx is always expected to be in steady state. Ifthe chemistry included in our description of PHOx and LHOxis correct, and if we are only limited by random noise in themeasurements, we expect a plot of PHOx/LHOx versus anychemical coordinates to be scattered about the number 1with a slope equal to zero. Systematic measurement errorsare expected to lead to systematic shifts in the agreement ofPHOx and LHOx but not, to first order, in the slope of

PHOx/LHOx versus chemical coordinates. Furthermore,focusing on FHO2+RO2, FOH+NO2, and HHLoss/LHOx reducesthe importance of systematic errors because these quantitiesrepresent the importance of individual HOx loss processesrelative to a total loss rate that includes the rate of eachindividual process. Figure 6 shows four panels containingplots of PHOx/LHOx versus the chemical coordinates des-cribed above. In all panels we required [RO2]PSS to begreater than 10% of [HO2]OBS and [HO2]OBS to be greaterthan 0 ppt to minimize the influence of noise from our PSScalculations. These two selections exclude 53 points fromdata where PHOx is greater than 0.5 ppt/s (�9% of the data,Figure 6b–6d).[31] Figure 6a shows PHOx/LHOx versus PHOx including

data with PHOx > 0.1 ppt/s and an additional filter to excludepoints prior to 0900 LST. This filter acts to minimize therole of other potentially large morning HOx sources such asHONO and excludes 233 points (10% of the data for PHOx >0.1 ppt/s) from the data where PHOx is between 0.1 and0.5 ppt/s. The data in Figure 6a are shown as differentsymbols above and below 0.5 ppt/s. The large squares areaverages over data in bins with widths of 0.1 ppt/s. The barsrepresent the standard deviation of the data used in theaverage. The data show no evidence of a trend in the qualityof agreement versus PHOx. Over a decade in the rate of PHOxthe quantity PHOx/LHOx is nearly constant at an averagevalue of �0.6. At still lower PHOx than shown in Figure 6a,PHOx/LHOx decreases, indicating that the model is missingan important source of HOx at higher solar zenith angles.This discrepancy is due in part to omitting the photolysisof H2O2 and aldehydes (C2 and larger) as well as PANdecomposition in our PHOx calculation. However, theremay also be other missing sources. To avoid confusing thedifferent errors, we focus our analysis on the high PHOxregime where O3 and formaldehyde photolysis terms arelarge and where PHOx/LHOx is largely independent ofPHOx.[32] In Figure 6b, PHOx/LHOx is plotted versus the

fraction HHLoss/LHOx. PHOx/LHOx is anticorrelated withthe importance of HOx-HOx reactions to the total HOx loss.When the HOx-HOx reactions dominate the total HOx loss(HHLoss/LHOx � 0.95), PHOx/LHOx approaches 0.25. Alinear fit to the averaged data gives the line, PHOx/LHOx= 1.24 0.91*HHLoss/LHOx, R2 = 0.96. We haveexcluded the point at HHLoss/LHOx = 0.1 in this fit.Including this point degrades the quality of the fit yieldinga slope of 0.53 and an R2 of 0.60. The steep slope shownin Figure 6 implies errors in the HOx budget are stronglycorrelated with the relative importance of the HOx-HOxreactions.[33] In Figure 6c, PHOx/LHOx is plotted versus the frac-

tional contribution of organic peroxide formation to thetotal HOx loss rate, FHO2+RO2. Figure 6c shows that most ofthe model-measurement discrepancy correlated with HOxself-reactions indicated by Figure 6b is specifically asso-ciated with the RO2 + HO2 reaction. PHOx/LHOx is scattered�1 when the contribution of organic peroxide formation toLHOx is negligible (i.e., FHO2+RO2 � 0). It is significantlyless than 1 when LHOx is dominated by organic peroxideformation ranging from 0.25 to 0.33, where FHO2+RO2 =0.80. If we assume a linear model, we find PHOx/LHOx =0.940.83*FHO2+RO2, R2 = 0.97. Extrapolating to


FHO2+RO2 = 1 gives PHOx/LHOx = 0.11. This extrapolationsuggests that the sink of HOx to form organic peroxides is�10 times too fast in the model. As a range of estimates forthe model-measurement differences, we show two dashedlines in Figure 6c that are the maximum and minimumslopes that bound one standard deviation of the data. Theselines indicate modeling the high PHOx data set requires therate of HOx chain termination by HO2 + RO2 to be reducedby a factor between �3 and 12.[34] Figure 6d shows the results of a similar analysis

along the FOH+NO2 coordinate. In this model, because we seta, the yield of alkyl nitrate formation, to zero, FOH+NO2 isthe exact complement of HHLoss/LHOx shown in Figure 6b.When FOH+NO2 approaches 1, the linear regression impliesthat LHOx may be too slow by 25% indicating either thatmodeled HNO3 formation is too slow or that other lossprocesses involving HOx and NOx, such as alkyl nitrateformation, may be important.[35] There are number of possible errors that could cause

the model-measurement discrepancies identified above.

Errors in the measured [HO2], the [RO2]PSS, the rateconstants, or some combination of these three, could beresponsible. However, the uncertainty in the measured[HO2] (±20%) and the inferred [RO2] (±56%) are too smallto have much effect, and uncertainties in [HO2] and[RO2]PSS are anticorrelated. A systematic reduction in[HO2] leads to a systematic increase in the inferred[RO2]PSS and vice versa. Another possibility is that theincrease in JNO2 that we use to reduce the systematic biasesin our analysis is too high. If instead we do not make an11% adjustment to compensate for net negative [RO2]PSS inthe early morning and late evening, the chemical coordinateplots look nearly identical with a mean error of about afactor of 7 instead of a factor of 9 upon extrapolation toFHO2+RO2 = 1. Furthermore, if we neglect the 11% adjust-ment in JNO2[NO2] and systematically increase the productJNO2[NO2] by 21% (the maximum uncertainty), Figure 6clooks nearly identical with FHO2+RO2 extending to �0.9. Asystematic decrease of JNO2[NO2] by 21% reduces the rangeof FHO2+RO2 by nearly a factor 2, and PHOx/LHOx decreases

Figure 6. (a–d) PHOx/LHOx calculated using 1-min averaged observations plotted versus (a) theprimary HOx production rate, PHOx, (b) the fraction of HOx-HOx reactions, HHLoss/LHOx, (c) thefractional contribution of organic peroxide formation to LHOx, FHO2+RO2, and versus (d) the fractionalcontribution of nitric acid formation to LHOx, FOH+NO2. Figure 6a shows PHOx/LHOx over both amoderate (0.1 < PHOx < 0.5 ppt/s, small open squares) and a high PHOx (>0.5 ppts, triangles) regime.In Figures 6b–6d, only data from the high PHOx (PHOx > 0.5 ppt/s) regime is shown. The large squares areunweighted averages of PHOx/LHOx over the moderate (open, Figure 6a only) and high (solid) PHOxregimes. The error bars represent the 1s deviation of the data used in the average. A linear regression (solidline) of the averaged data is shown for each panel. In Figures 6c and 6d, two additional lines (dashed) areshown representing the minimum and maximum slopes allowed by the 1s variation of the data.


with FHO2+RO2 nearly twice as fast. However, we note thatthis systematic reduction of JNO2[NO2] also producesunphysical [HO2 + RO2]PSS where this quantity is smallerthan the measured [HO2]OBS 84% of the time.[36] The possible errors in reactant concentrations noted

above are either unrealistic or do not approach the factor of3–12 required to bring the model and measurements intoagreement where HO2 + RO2 is the dominant HOx sink. Inwhat follows we focus on errors in the model representationof HOx loss processes. We note here that an alternatesolution would be to infer an error in the HOx productionrate that is correlated with the abundance of RO2 or the rateof the RO2 + HO2 reaction. However, such an error seemsless likely, and we have been unable to identify a candidatesource molecule that has a strong correlation with theobserved RO2.[37] The chemical coordinates indicate a large error in the

loss reactions at low NOx and indicate the possibility of anerror at high NOx. Because the HNO3 formation reactionhas been extensively revisited in recent laboratory studiesand because we deliberately set a = 0 when alkyl nitrateyields are known to be significant, we think it more likelythat the source of the potential error at high NOx is due toomitting alkyl nitrate formation. It is possible that a fractionof this error may be due to net peroxynitrate formation, viaRO2 + NO2; however, we assume this term is small andomit it from the analysis. At low NOx, there is an apparenterror associated with the HOx sink through the RO2 + HO2reaction. The rate of this reaction is probably accurate towithin a factor of 2 or 3, although there are few directmeasurements for complex organic peroxy radicals. Theproducts for the reaction of RO2 + HO2 are even lesscharacterized than the rate coefficient. We propose that afraction of the reaction products are free radicals, via RO2 +HO2 ! RO + OH + O2 for example, as opposed to a freeradical chain terminator such as ROOH. We note that afunctional equivalent to this hypothesis is that if the organichydroperoxide is formed as the product, it is rapidlyphotylzed to produce free radicals on a timescale of 1 hour.Although this hypothesis is not supported by some smogchamber data [Miyoshi et al., 1994], the error we observe atlow NOx is most likely a combination of errors in the rateconstant for RO2 + HO2 and the branching ratio for ROOHproduction via this reaction for the types of RO2 at CFA. Wedefine the parameter g that represents the change in overallrate of RO2 + HO2 ! ROOH (where ROOH is presumed tobe long-lived).[38] We optimized (brought as close to 0 as possible) the

slope of PHOx/LHOx versus the two coordinates FOH+NO2and FRO2+HO2 by adjusting the parameters a and g. We findthat the optimum solution has g of between 0 and 0.15 anda between 0.01 and 0.15. Because the role of alkyl nitrateformation is unimportant at low NOx where RO2 + HO2 isthe dominant sink of HOx, the optimum choice of g isindependent of the optimum rate of alkyl nitrate formation.In the rest of this analysis we set g to 0.08, which is thesmallest value allowed by the range of data in the linearregression of PHOx/LHOx versus FHO2+RO2 for high PHOx.We emphasize that g is an empirical parameter that repre-sents the change in the product of the effective rate constantkHO2+RO2 and the branching ratio leading to ROOH for-mation. For example, a value of 0.08 for g could be

achieved by a factor of 4 reduction in the effective rateconstant together with a factor of 3 reduction in the ROOHformation branching ratio. An alkyl nitrate yield of �3%,minimized the slope of PHOx/LHOx versus the chemicalcoordinates and brings the ratio PHOx/LHOx closer to 1 athigh NOx. However, this parameter is not precisely con-strained by the observations and our assumed linear model,and a wide range in the alkyl nitrate yield (1–15%) isconsistent with the measurements.[39] In Figure 7 we show PHOx/LHOx plotted versus the

chemical coordinates PHOx HHLoss/LHOx, FRO2+HO2, andFOH+NO2, using the optimized model developed above.Table 2 summarizes the reactions and corresponding rateconstants and branching ratios used to generate bothFigure 5 and Figure 6. Figure 7a shows that the adjustedmodel does significantly improve PHOx/LHOx in the highPHOx data and clarifies the presence of a growing bias atthe lowest PHOx, which, as we indicated, is possibly due toH2O2 or HONO photolysis and/or PAN decomposition. Asin Figure 6, we continue to focus on data where PHOx isgreater than 0.5 ppt/s in Figure 7b–7d. In Figure 7b weshow the fraction of the total HOx loss that is due to theHOx-HOx reactions, HHLoss/LHOx, in Figure 7c, we showthe data versus the fraction of HO2 + RO2 that leads toROOH production, and in Figure 7d, we show the dataversus the fraction of loss via the reaction of OH withNO2. Note that the range in Figure 7c, 0–0.2, is greatlyreduced compared to that in Figure 6 of 0–1. In all fourcoordinates, PHOx/LHOx is on average closer to 1 with a slopecloser to zero than shown in Figure 6. Linear regressions ofPHOx/LHOx versus HHLoss/LHOx and versus FOH + NO2 yieldslopes of 0.3 (R2 =0.86) and 0.24 (R2 = 0.76), respectively,each nearly a factor of 3 smaller than those shown in Figure 6.We have excluded the same points from these fits as inFigure 6 (at HHLoss/LHOx � 0.1 and at FOH+NO2 � 0.8) asthey continue to degrade the fit. Including these pointsyields slopes of 0.08 (R2 = 0.058) and 0.106 (R2 = 0.179)versus HHLoss/LHOx and FOH+NO2, respectively.[40] We show the fractions HHLoss/LHOx and NHLoss/

LHOx versus NO for the revised model in Figure 8. Again,we include PO3 (solid circles) divided by a scale factor of7. The data are restricted to the narrow high PHOx range(0.5 < PHOx < 0.7 ppt/s). The two fractions cross 0.5 asearly as 400 ppt NO and stay even for NO as high as 700ppt. This result is more consistent with the start of thecrossover region beginning at �600 ppt observed in thePO3 data. The correspondence of the NOx-limited/NOx-saturated crossover point arrived at by two largely inde-pendent analyses lends additional support for our sugges-tion that the rate of HO2 + RO2 ! ROOH + O2 is muchsmaller than current recommendations.

7. Discussion

[41] The photostationary state (PSS) assumption that weused to infer PO3 is frequently used to examine instanta-neous O3 production rates in equation (4) and to inferperoxy radical levels in equations (2) and (3). Surprisingly,to our knowledge, there is no other data set that has shownthe rate of instantaneous ozone production over the fullrange from NOx limited to NOx saturated behavior. Perhapsthe lack of attention to PO3 derived from the PSS analysis


arises from discomfort with using this approach in urbanand continental settings where discrepancies betweenmeasured [HO2 + RO2] and PSS-derived [HO2 +RO2]PSS have been observed. These discrepancies likelyhave a chemical explanation in either the RO2 measure-ments or in the analyses of RO2 + HO2 at low NOx(where RO2 + HO2 ! ROOH + O2) is important. Forexample, where O3, HO2, CH3O2, ClO, and BrO are thedominant oxidants of NO, such as in the remote tropo-sphere or the stratosphere, the photostationary statekinetics have been shown to be accurate to better than20% [e.g., Ridley et al., 1992b; Volz-Thomas et al., 1997;Cohen et al., 2000].[42] In regions with high hydrocarbon abundances, sig-

nificant discrepancies in the PSS method have beenobserved. Peroxy radicals calculated from equation (2) arehigher than measured [RO2 + HO2] by factors of 2–3[Cantrell et al., 1993a; Volz-Thomas et al., 1997; Carpenteret al., 1998]. Explanations for this discrepancy include thepropagation of experimental uncertainty [Cantrell et al.,1993a; Baumann et al., 2000], inaccurate measurements ofthe species used in equations (2) and (3) [Frost et al., 1998;Baumann et al., 2000], or the deviation from a steady state in

the atmosphere [e.g., Calvert and Stockwell, 1983]. What-ever the reason for these discrepancies, we believe the shapeof the curves derived in this study and shown in Figure 3: thepresence of an initial slope that is linear in NOx, a regionwhere the slope decreases to near zero, and a turnover at highNOx cannot be caused by experimental error.[43] The results we present here add to a growing list of

discrepancies observed in the HOx budget under low NOxconditions [e.g., McKeen et al., 1997; Stevens, 1997;Faloona et al., 2000]. The observed discrepancies in thevarious studies are not linked by an obvious error, and whilethey may not be directly comparable, they point to a criticallack of understanding of the HOx budget at low NOx.Stevens [1997] and McKeen et al. [1997] report significantdifferences between observations and model calculatedmagnitudes of OH, HO2, and RO2 concentrations and alsoin the partitioning between these HOx species during theTropospheric OH Photochemistry Experiment. At low NOxthe observationally constrained model employed by Stevens[1997] calculates HO2/OH and RO2/HO2 that are higher byfactors of 3–4 and 4–15, respectively, than those observed.McKeen et al. [1997] report that a similar model over-estimates OH concentrations by a factor of 6–8, while

Figure 7. (a–d) PHOx/LHOx plotted versus the same chemical coordinates as in Figure 6 where LHOxhas been modified to include alkyl nitrate formation (3%; yield) and an HO2 + RO2 ! ROOH ratethat is a factor of 12 slower. (a) PHOx/LHOx over both a moderate (0.1 < PHOx < 0.5 ppt/s, smallopen squares) and a high PHOx (>0.5 ppts, triangles) regime. (b–d) Only data from the high PHOx(PHOx > 0.5 ppt/s) regime are shown. The large squares are averages of PHOx/LHOx for the moderate(open squares) and high (solid squares) PHOx regimes. Linear regressions for the high PHOx averageddata are shown in all panels.


underestimating HO2 concentrations by more than a factorof 3–4. Similarly, Faloona et al. [2000] present results fromthe upper troposphere and lowermost stratosphere that showmodel calculations of HO2 concentrations were often lowerthan those observed. They use a 0-D stationary state boxmodel and a diel steady state model both constrained withobservations and show that the discrepancies between themeasured and modeled HO2 concentrations are stronglycorrelated with measured NOx concentrations [Faloona etal., 2000]. The model errors that our analysis points to are inthe chain termination steps of the HOx-NOx catalytic cycleand substantially modify HOx partitioning at low NOx.Analysis of HOx partitioning over a range of chemicalconditions by using combined data sets should providefurther insight. For example, the HO2 measured in Nashvilleand used for this study decreases more slowly with NO thandoes the RO2 inferred from equations (2) and (3). Con-sequently, PO3 calculated from equation (4) increases withNO up to �800 ppt NO, decreasing for NO > 1100 ppt,while PO3 from the reaction of HO2 + NO increases for NOup to 1000 ppt and is nearly constant at higher NO.[44] Comparison of our results to those from the ROSE

study in rural Alabama is particularly interesting because ofsimilarities in VOC abundance and PHOx. Frost et al.[1998], and Trainer et al. [2000] compare O3 productionrates at Kinterbish, Alabama, calculated by four differentapproaches: (1) the photostationary state assumption, seeequation (3) used here, (2) directly from PO3 = keff [HO2 +RO2]MEAS [NO] using measurements of sum total peroxyradicals made with the chemical amplifier technique [Can-trell et al., 1993b], (3) from a radical budget analysis inwhich peroxy radicals are calculated such that PHOx = LHOx,and (4) from a photochemical box model using inputs ofmeasured O3, NO, NO2, HNO3, CO, VOCs, temperature,relative humidity, and JNO2. Trainer et al. examine theresults as a function of NOx and report PO3 data averagedover the entire Kinterbish campaign. In their analysis,Trainer et al. select data where JNO2 > 0.005 s

1. Using

the PSS method, they calculate O3 production rates as highas 4 (±1, 1s) ppt/s at 1.5 ppb of NOx. There is noobservation of a crossover point, i.e., a deviation of PO3from a linear increase with NOx, in this data. The 1sstandard deviation of the data arises primarily from atmos-pheric variation and not from noise in the measurements.Applying the same J value selection criteria to the CFA dataselects a range of PHOx from 0.05 to 1 ppt/s. We calculatepeak O3 production rates of 7 (±3, 1s) ppt/s at 4–5 ppbNOx. At a NOx concentration of 1.5 ppb, PO3 at CFA is 4(±1.5, 1s) ppt/s, similar to that of Kinterbish. Both data setsexhibit the same slope of PO3 versus NOx at low NOx (whenfiltered identically).[45] The photochemical box model and radical budget

method employed by Frost et al., and Trainer et al., yield O3production rates that are more than a factor of 2 slower thanthose they calculate using the photostationary state method.These two models predict a crossover point between NOx-limited and NOx-saturated regimes near 7 ppb NOx. Frostet al. [1998] note that their model and radical budget resultsdepend strongly on the choice of a peroxide formation rateconstant and that this is a significant uncertainty. If themodel is revised as we suggest, the predicted crossoverpoint for Kinterbish would move to lower NOx but not solow as to conflict with the absence of a crossover in theKinterbish data set. Since both PHOx and PO3 versus NOx atmoderate to high PHOx were similar at CFA and Kinterbish[Frost et al., 1998], it is likely that the crossover point inKinterbish would also be similar to that observed at CFA,4–5 ppb NOx (600–800 ppt NO). This is above the highestNOx observed at Kinterbish.[46] An alternative approach to evaluating PO3 has been

developed by Kleinman [2000]. They have examined NOx-limited and NOx-saturated O3 production chemistry byevaluating estimates of the rates of the competing chaintermination steps (HHLoss and NHLoss). Their methoduses an effective rate constant for total peroxide formationand does not separate the contributions from the reactionsHO2 + HO2 and RO2 + HO2 [Kleinman et al., 1997]. Forexample, Daum et al. [2000], apply this method to calculatePO3 using a box model constrained with data observed ontwo separate days in the Nashville urban plume during the1995 Southern Oxidants Study. At low NOx their modelcalculated that PO3 increases linearly with NOx and at �4ppb of NOx, the calculated PO3 becomes independent ofNOx. For a typical NO2 to NO ratio of 7 this crossover pointis relatively close to the one we derive directly fromobservations at CFA. This agreement is somewhat surpris-ing given the large error we identify in current estimates oforganic peroxide formation.[47] The factor of 10 reduction rate of HO2 + RO2 !

ROOH + O2 that we propose is a large change in models.However, the rate constant for this class of reactions is notwell known for species with complicated organic moietiessuch as the isoprene derivatives [Lee et al., 2000], and wecalculate isoprene-RO2 radicals are more than 40% of thetotal at CFA. Current measurements and structure reactivityestimates are described by Lesclaux [1997] and Wallingtonet al. [1997] and measurements for overall rate constantsfor secondary and tertiary RO2 range from 2 1012 to2 1011 at 298 K, 760 Torr [Lesclaux, 1997]. If the rateconstant for isoprene-RO2 is at the lower end of this range,

Figure 8. PO3, HHLoss and NHLoss, as for Figure 5 withan alkyl nitrate yield of 3% and a rate/yield of RO2 + HO2! ROOH + O2 reduced by a factor of 12.


then our suggestion for reducing g can be interpreted asprimarily a reduction in the rate constant used in models.If, on the other hand, the rate constant is at the high end ofthe range then g must be interpreted as a branching ratiobetween ROOH and radical products. There are fewproduct studies for this class of reaction. The yields oforganic peroxides in these studies are usually observed tobe near unity [e.g., Rowley et al., 1992a, 1992b; Tuazon etal., 1998], a result that clearly conflicts with our interpre-tation of the CFA data. These laboratory results could bereconciled with our interpretation if the peroxides observedare not a primary product of the initial reaction to form anisoprene-ROOH but are produced by subsequent cycles ofRO2 + HO2 in the equilibrium reaction mixture to yieldROOH from smaller, less complex RO2. Alternatively,these peroxides could be rapidly photolyzed under atmos-pheric conditions making them at best a temporary HOxreservoir.

8. Conclusions

[48] In situ observations ofNO,NO2,O3,OH,HO2,H2CO,actinic flux, relative humidity, and temperature are used toseparate theNOx andHOx influences on urbanO3 production.Instantaneous ozone production rates calculated using thephotostationary state assumption exhibit NOx-limited andNOx-saturated behavior over three different ranges of primaryHOx radical production rates. The fastest HOx productionrates yield the fastest O3 production rates. The peak in O3production and the crossover between NOx-limited and NOx-saturated behavior shifts to higher NOx abundances withincreases in PHOx. HOx partitioning between RO2 and HO2is an important and poorly understood factor controlling theNOx dependence of PO3. Further studies of the NOx depend-ence of this partitioning are necessary.[49] We compare the observed crossover point to a model

of the chain termination processes using current recommen-dations for the yield of peroxides from the reaction RO2 +HO2 ! ROOH + O2. This model predicts a crossover pointat NO mixing ratio higher than we observe. A chemicalcoordinate approach to evaluating the accuracy of themodeled fast photochemistry provides a strong indicationthat the model and measurements are inconsistent becauseof some error in the representation of the processes asso-ciated with the RO2 + HO2 reaction. Our analysis suggeststhat any combination of slowing the rate of RO2 + HO2 orreducing the yield of ROOH from this reaction that reducesthe effective rate of chain termination by about a factor of10 will improve model representations of HOx. Even if thisreaction is correctly represented in current models, using theadjusted RO2 + HO2 rate we recommend is likely to accountfor some alternate process not captured in current models(such as rapid photolysis of ROOH) and to be a betterrepresentation of atmospheric observations.[50] These results have significant implications for strat-

egies to reduce urban and regional ozone concentrations.Models that incorporate organic peroxide formation ratesthat are faster than those that occur in the atmosphere willpredict crossover points between NOx-limited and NOx-saturated O3 production that are too high in NOx. Thesemodels will underestimate the NOx reduction required toreduce O3 concentrations in regions that are presently NOx

saturated. However, simply extending our results to predic-tions of the response of the atmosphere to changes inemissions should be made with caution because of thedifference between net O3 production rates or O3 concen-trations and the instantaneous gross PO3 we present in thispaper. Many modeling studies have invoked observation-based indicators to examine instantaneous O3 production, todiscern NOx-limited and NOx-saturated conditions in netozone formation, and to study the sensitivity of O3 concen-trations to changes in NOx emissions [Sillman, 1995; Klein-man et al., 1997; Sillman et al., 1997; Tonnesen and Dennis,2000aTonnesen and Dennis, 2000b]. Understanding the lowNOx implications of our analysis of instantaneous PO3 forthe integrated behavior of O3 should be a focus of futurechemical studies as well as for regulatory studies aimed atguiding reductions in NOx emissions.

[51] Acknowledgments. Observations in Nashville, Tennessee, weresupported by NOAA Office of Global Programs. Joel Thornton gratefullyacknowledges support from a NASA Earth Systems Science FellowshipNASA grant NGT5-30219.

ReferencesAlicke, B., G. Hoennenger, U. Platt, and J. Stutz, Measurements of O3,HCHO, NO2, HONO, and NO3 by differential optical absorption spectro-scopy during the 1999 SOS Field Study in Nashville, Eos Trans. AGU,81(48), F124, Fall Meet. Suppl., 2000.

Atkinson, R., Gas-phase tropospheric chemistry of organic compounds,J. Phys. Chem. Ref. Data, M2, R1, 1994.

Baumann, K., E. J. Williams, W. M. Angevine, J. M. Roberts, R. B. Norton,G. J. Frost, F. C. Fehsenfeld, S. R. Springston, S. B. Bertman, andB. Hartsell, Ozone production and transport near Nashville, Tennessee:Results from the 1994 study at New Hendersonville, J. Geophys. Res.,105(D7), 9137–9153, 2000.

Calvert, J. G., and W. R. Stockwell, Deviations from the O3-NO-NO2photostationary state in tropospheric chemistry, Can. J. Chem., 61,983–992, 1983.

Cantrell, C. A., R. E. Shetter, J. G. Calvert, D. D. Parrish, F. C. Fehsenfeld,P. D. Goldan, W. Kuster, E. J. Williams, H. H. Westberg, G. Allwine, andR. Martin, Peroxy radicals as measured in rose and estimated fromphotostationary state deviations, J. Geophys. Res., 98(D10), 18,355–18,366, 1993.

Cantrell, C. A., R. E. Shetter, J. A. Lind, A. H. McDaniel, J. G. Calvert,D. D. Parrish, F. C. Fehsenfeld, M. P. Buhr, and M. Trainer, An im-proved chemical amplifier technique for peroxy radical measurements,J. Geophys. Res., 98(D2), 2897–2909, 1993.

Cardelino, C. A., and W. L. Chameides, Natural hydrocarbons, urbaniza-tion, and urban ozone, J. Geophys. Res., 95, 13,971–13,979, 1990.

Carpenter, L. J., K. C. Clemitshaw, R. A. Burgess, S. A. Penkett, J. N.Cape, and G. C. McFadyen, Investigation and evaluation of the NOx/O3photochemical steady state, Atmos. Environ., 32(19), 3353–3365, 1998.

Cohen, R. C., Quantitative constraints on the atmospheric chemistry ofnitrogen oxides: An analysis along chemical coordinates, J. Geophys.Res., 105(D19), 24,283–24,304, 2000.

Daum, P. H., L. Kleinman, D. G. Imre, L. J. Nunnermacker, Y. N. Lee,S. R. Springston, L. Newman, and J. Weinstein-Lloyd, Analysis of theprocessing of Nashville urban emissions on July 3 and July 18, 1995,J. Geophys. Res., 105(D7), 9155–9164, 2000.

DeMore, W. B., S. P. Sander, C. J. Howard, A. R. Ravishankara, D. M.Golden, C. E. Kolb, R. F. Hampson, M. J. Kurylo, and M. J. Molina,Chemical kinetics and photochemical data for use in stratosphericmodeling, Eval. 12, NASA Jet Propul. Lab., Calif. Inst. of Technol.,Pasadena, 1997.

Faloona, I., Observations of HOx and its relationship with NOx in the uppertroposphere during SONEX, J. Geophys. Res., 105(D3), 3771–3783,2000.

Fried, A., B. Henry, B. Wert, S. Sewell, and J. R. Drummond, Laboratory,ground-based, and airborne tunable diode laser systems: Performancecharacteristics and applications in atmospheric studies, Appl. Phys. BLasers Opt., 67(3), 317–330, 1998.

Frost, G. J., Photochemical ozone production in the rural southeasternUnited States during the 1990 Rural Oxidants in the Southern Environ-ment (ROSE) program, J. Geophys. Res., 103(D17), 22,491–22,508,1998.


Jaegle, L., D. J. Jacob, W. H. Brune, D. Tan, I. C. Faloona, A. J.Weinheimer, B. A. Ridley, T. L. Campos, and G. W. Sachse, Sourcesof HOx and production of ozone in the upper troposphere over theUnited States, Geophys. Res. Lett., 25(10), 1709–1712, 1998.

Jaegle, L., Ozone production in the upper troposphere and the influence ofaircraft during SONEX: Approach of NOx-saturated conditions, Geophys.Res. Lett., 26(20), 3081–3084, 1999.

Jaegle, L., D. J. Jacob, W. H. Brune, and P. O. Wennberg, Chemistry ofHOx radicals in the upper troposphere, Atmos. Environ., 35, 469–489,2001.

Kelly, T. J., D. H. Stedman, J. A. Ritter, and R. B. Harvey, Measurements ofoxides of nitrogen and nitric acid in clean air, J. Geophys. Res., 85(C12),7417–7425, 1980.

Kleinman, L., Y. N. Lee, S. R. Springston, J. H. Lee, L. Nunnermacker,J. Weinsteinlloyd, X. L. Zhou, and L. Newman, Peroxy radical con-centration and ozone formation rate at a rural site in the southeasternUnited States, J. Geophys. Res., 100(D4), 7263–7273, 1995.

Kleinman, L. I., Ozone process insights from field experiments, part II,Observation-based analysis for ozone production, Atmos. Environ.,34(12-14), 2023–2033, 2000.

Kleinman, L. I., P. H. Daum, J. H. Lee, Y. N. Lee, L. J. Nunnermacker,S. R. Springston, L. Newman, J. WeinsteinLloyd, and S. Sillman,Dependence of ozone production on NO and hydrocarbons in thetroposphere, Geophys. Res. Lett., 24(18), 2299–2302, 1997.

Lanzendorf, E. J., T. F. Hanisco, R. M. Stimpfle, J. G. Anderson, P. O.Wennberg, and R. C. Cohen, Comparing atmospheric [HO2]/[OH] tomodeled [HO2]/[OH]: Identifying discrepancies in reaction rate constants,Geophys. Res. Lett., 28(6), 967–970, 2001.

Lee, M. H., B. G. Heikes, and D. W. O’Sullivan, Hydrogen peroxide andorganic hydroperoxide in the troposphere: A review, Atmos. Environ.,34(21), 3475–3494, 2000.

Lesclaux, R., Combination of peroxyl radicals in the gas phase, in PeroxyRadicals, edited by Z. B. Alfassi, pp. 81–113, John Wiley, New York,1997.

Mather, J. H., P. S. Stevens, and W. H. Brune, OH and HO2 measurementsusing laser-induced fluorescence, J. Geophys. Res., 102(D5), 6427–6436, 1997.

McKeen, S. A., Photochemical modeling of hydroxyl and its relationship toother species during the Tropospheric OH Photochemistry Experiment,J. Geophys. Res., 102(D5), 6467–6493, 1997.

Miyoshi, A., S. Hatakeyama, and N. Washida, OH radical initiated photo-oxidation of isoprene: An estimate of global CO, J. Geophys. Res.,99(D9), 18,779–18,787, 1994.

National Research Council, Rethinking the Ozone Problem in Urban andRegional Air Pollution, Natl. Acad. Press, Washington, D. C., 1991.

Parrish, D. D., M. Trainer, E. J. Williams, D. W. Fahey, G. Hubler, C. S.Eubank, S. C. Liu, P. C. Murphy, D. L. Albritton, and F. C. Fehsenfeld,Measurements of the NOx-O3 photostationary state at Niwot Ridge,Colorado, J. Geophys. Res., 91(D5), 5361–5370, 1986.

Patz, H. W., Measurements of trace gases and photolysis frequencies duringSLOPE96 and a coarse estimate of the local OH concentration fromHNO3 formation, J. Geophys. Res., 105(D1), 1563–1583, 2000.

Ridley, B. A., F. E. Grahek, and J. G. Walega, A small, high-sensitivity,medium-response ozone detector suitable for measurements from lightaircraft, J. Atmos. Oceanic Technol., 9(2), 142–148, 1992.

Ridley, B. A., S. Madronich, R. B. Chatfield, J. G. Walega, R. E. Shetter,M. A. Carroll, and D. D. Montzka, Measurements and model simulationsof the photostationary state during the Mauna-Loa-Observatory Photo-chemistry Experiment—Implications for radical concentrations andozone production and loss rates, J. Geophys. Res., 97(D10), 10,375–10,388, 1992.

Rowley, D. M., R. Lesclaux, P. D. Lightfoot, K. Hughes, M. D. Hurley,S. Rudy, and T. J. Wallington, Kinetic and mechanistic study of thereaction of neopentylperoxy radicals with HO2, J. Phys. Chem., 96(17),7043–7048, 1992.

Rowley, D. M., R. Lesclaux, P. D. Lightfoot, B. Noziere, T. J. Wallington,and M. D. Hurley, Kinetic and mechanistic studies of the reactions ofcyclopentylperoxy and cyclohexylperoxy radicals with HO2, J. Phys.Chem., 96(12), 4889–4894, 1992.

Sander, S.P., et al., Chemical kinetics and photochemical data for use instratospheric modeling, Eval. 13, NASA Jet Propul. Lab., Calif. Inst. ofTechnol., Pasadena, 2000.

Shetter, R. E., and M. Muller, Photolysis frequency measurements usingactinic flux spectroradiometry during the PEM-Tropics mission: Instru-mentation description and some results, J. Geophys. Res., 104(D5),5647–5661, 1999.

Sillman, S., The use of NOy, H2O2, and HNO3 as indicators for ozone-NOx-hydrocarbon sensitivity in urban locations, J. Geophys. Res., 100(D7),14,175–14,188, 1995.

Sillman, S., D. Y. He, C. Cardelino, and R. E. Imhoff, The use of photo-chemical indicators to evaluate ozone-NOx-hydrocarbon sensitivity: Casestudies from Atlanta, New York, and Los Angeles, J. Air Waste Manag.Assoc., 47(10), 1030–1040, 1997.

Stevens, P. S., HO2/OH and RO2/HO2 rates during the Tropospheric OHPhotochemistry Experiment: Measurement and theory, J. Geophys. Res.,102(D5), 6379–6391, 1997.

Stimpfle, R. M., The coupling of ClONO2, ClO, and NO2 in the lowerstratosphere from in situ observations using the NASA ER-2 aircraft,J. Geophys. Res., 104(D21), 26,705–26,714, 1999.

Thornton, J. A., P. J. Wooldridge, and R. C. Cohen, Atmospheric NO2: Insitu laser-induced fluorescence detection at parts per trillion mixingratios, Anal. Chem., 72(3), 528–539, 2000.

Tonnesen, G. S., and R. L. Dennis, Analysis of radical propagationefficiency to assess ozone sensitivity to hydrocarbons and NOx, 1,Local indicators of instantaneous odd oxygen production sensitivity,J. Geophys. Res., 105(D7), 9213–9225, 2000.

Tonnesen, G. S., and R. L. Dennis, Analysis of radical propagationefficiency to assess ozone sensitivity to hydrocarbons and NOx, 2,Long-lived species as indicators of ozone concentration sensitivity,J. Geophys. Res., 105(D7), 9227–9241, 2000.

Trainer, M., D. D. Parrish, P. D. Goldan, J. Roberts, and F. C. Fehsenfeld,Review of observation-based analysis of the regional factors influencingozone concentrations, Atmos. Environ., 34(12-14), 2045–2061, 2000.

Tuazon, E. C., S. M. Aschmann, J. Arey, and R. Atkinson, Products of thegas-phase reactions of a series of methyl-substituted ethenes with the OHradical, Environ. Sci. Technol., 32, 2106–2612, 1998.

Villenave, E., and R. Lesclaux, Kinetic studies of the self-reaction andreaction with HO2 of a tertiary beta-brominated peroxy radical between298 and 393 K, Int. J. Chem. Kinet., 33, 41–48, 2001.

Volz-Thomas, A., et al., Photochemical ozone production rates at differentTOR sites, in Tropospheric Ozone Research, edited by O. Hov, pp. 95–110, Springer-Verlag, New York, 1997.

Wallington, T. J., O. J. Nielsen, and J. Sehested, Reactions of organicperoxy radicals in the gas phase, in Peroxy Radicals, edited by Z. B.Alfassi, pp. 113–172, John Wiley, New York, 1997.

Wennberg, P. O., Hydrogen radicals, nitrogen radicals, and the productionof O3 in the upper troposphere, Science, 279(5347), 49–53, 1998.

Williams, E. J., Intercomparison of ground-based NOy measurementtechniques, J. Geophys. Res., 103(D17), 22,261–22,280, 1998.

Wooldridge, P. J., Informal comparison of in-situ and long path measure-ments of NO2 during the 1999 Southern Oxidants Study, Eos Trans.AGU, 81(48), F125, Fall Meet. Suppl., 2000.

W. H. Brune, H. Harder, and M. Martinez, Department of Meteorology,

Pennsylvania State University; University Park, PA 16802, USA.R. C. Cohen, J. A. Thornton, and P. J. Wooldridge, Department of

Chemistry, University of California, Berkeley, CA 94705, USA.(cohen@cchem. berkeley.edu)F. C. Fehsenfeld, J. M. Roberts, and E. J. Williams, Aeronomy

Laboratory, National Oceanic and Atmospheric Administration, Boulder,CO 80305, USA.A. Fried, S. R. Hall, R. E. Shetter, and B. P. Wert, Atmospheric

Chemistry Division, National Center for Atmospheric Research, Boulder,CO 80305, USA.


ozone production rates as a function of no abundances and ... › ~thornton › publications_files...

Documents