On the temperature dependence of organic reactivity , nitrogen oxides , ozone production , and the impact of emission controls in San Joaquin Valley

The San Joaquin Valley (SJV) experiences some of the worst ozone air quality in the US, frequently exceeding the California 8 h standard of 70.4 ppb. To improve our understanding of trends in the number of ozone violations in the SJV, we analyze observed relationships between organic reactivity, nitrogen oxides (NOx), and daily maximum temperature in the southern SJV using measurements made as part of California at the Nexus of Air Quality and Climate Change in 2010 (CalNex-SJV). We find the daytime speciated organic reactivity with respect to OH during CalNex-SJV has a temperature-independent portion with molecules typically associated with motor vehicles being the major component. Published by Copernicus Publications on behalf of the European Geosciences Union. 3374 S. E. Pusede et al.: Temperature-dependent impacts of emission controls At high temperatures, characteristic of days with high ozone, the largest portion of the total organic reactivity increases exponentially with temperature and is dominated by small, oxygenated organics and molecules that are unidentified. We use this simple temperature classification to consider changes in organic emissions over the last and next decade. With the CalNex-SJV observations as constraints, we examine the sensitivity of ozone production ( PO3) to future NOx and organic reactivity controls. We find that PO3 is NOx-limited at all temperatures on weekends and on weekdays when daily maximum temperatures are greater than 29 C. As a consequence, NOx reductions are the most effective control option for reducing the frequency of future ozone violations in the southern SJV.


Introduction
California's San Joaquin Valley suffers extremely poor air quality and has so for decades; violations of health-based ozone standards remain frequent despite statewide ozone control measures (Cox et al., 2009).Understanding trends in frequency of these violations requires knowledge of trends in the chemical drivers of ozone production (P O 3 ) -nitrogen oxides (NO x ≡ NO + NO 2 ), the reactivity of volatile organic compounds with OH (VOCR), and the production rate of HO x (HO x ≡ OH + HO 2 + RO 2 ) radicals -along with trends in the sensitivity of local and regional P O 3 to these parameters.The sequence of ozone forming reactions is generally known; however, determining the response of P O 3 in any location to NO x and organic emission reductions is challenging, as P O 3 is a nonlinear function of each precursor.Meteorological effects impart variability in the number of O 3 violations within a given O 3 season but do not, in the average, impact the interannual trends.
In Pusede and Cohen (2012), we examined relationships between nitrogen oxides, the frequency of O 3 exceedances, and temperature, which we used as a proxy for VOCR, over the period 1996-2010, inferring from the observed nonlinear nitrogen oxide dependence that chemical production rather than meteorology dominated the statistics of violations in the San Joaquin Valley (SJV).We showed that in the northern SJV (Stockton plume) and central SJV (Fresno plume) P O 3 has largely been sensitive to VOCR and that organic emission reductions led to decreases in the frequency of O 3 violations.At high temperatures, P O 3 in these plumes is presently NO x sensitive.In the southern SJV (Bakersfield plume) at moderate temperatures P O 3 has been sensitive to VOCR for most of the record, with organic emissions controls resulting in fewer annual O 3 violations; simultaneous NO x reductions shifted chemistry to peak P O 3 with respect to NO x in 2007-2010.At high temperatures, southern SJV P O 3 has been at peak production for most of the last 15 yr, becoming NO x sensitive in 2007-2010; at these temperatures, VOCR reduc-tions have made little impact.This temperature dependence in the observed effects of VOCR controls thus suggested two distinct categories of organic reactivity in the Bakersfield plume: one source that had decreased over the last decade and was more important at moderate temperatures and a second source, dominating at high temperatures, that had not substantially changed over the same time period.
During the California at the Nexus of Air Quality and Climate Change experiment (CalNex-SJV), 18 May-29 June 2010, we measured nitrogen oxides, OH reactivity, HO x source molecules, and a wide suite of individual volatile organic compounds (VOCs) at the San Joaquin Valley supersite allowing new insight into the production rate of ozone in the region.Here, we use these observations to test the response of P O 3 to variations in temperature, VOCR, and NO x .We begin our analysis by describing the temperature dependence of the daytime organic reactivity and categorizing one component of the VOCR that is independent of temperature and a second that increases exponentially with temperature.This simple temperature-response framework allows us to consider how organic emission controls have impacted the total organic reactivity over the last decade and how they will do so over the next 10 yr.We use the CalNex-SJV observations to constrain a model of instantaneous P O 3 and then to guide our thinking about the frequency of O 3 violations in the complete 2010 O 3 season.We test various emission reduction scenarios and show both the magnitude and sign of the effects of controls on emissions are temperature dependent.
2 Chemical production of O 3 2.1 P O 3 reaction sequence P O 3 is initiated by the oxidation of gas-phase organic molecules.During the daytime, this usually involves reaction with the hydroxyl radical (OH) leading to formation of an alkyl radical (R), which immediately reacts with O 2 to form a peroxy radical (RO 2 ).In the typical chain-propagating branch of this sequence, RO 2 reacts with NO to form an alkoxy radical (RO) and NO 2 .In the presence of O 2 , the RO leads to formation of HO 2 and an aldehyde or ketone.The HO 2 in turn reacts with NO to form a second NO 2 molecule and regenerate OH.Hence Reactions (R1-R5) are catalytic, being HO x neutral while converting NO to NO 2 .
The importance of any individual VOC in the HO x cycle depends on its reaction rate with OH (k OH ).The rate at which a single organic species reacts with OH is defined as VOCR i ≡ k OH+VOC i [VOC i ] (s −1 ); for the entire VOC mixture, we define the total organic reactivity (VOCR).Throughout this paper we use i VOCR i to refer to the reactivity summed from individually measured compounds; we use VOCR both generally and to refer to the VOCR obtained from direct measurements of the total OH reactivity.In the continental boundary layer, representative daytime VOCR values are 7-12 s −1 in suburban Nashville, Tennessee (Kovacs et al., 2003); 10-30 s −1 in the polluted megacity Beijing (Lou et al., 2010); and 1-13 s −1 at a rural forest site in Michigan and covarying with daytime temperatures of 10-27 • C (Di Carlo et al., 2004).
The photolysis of NO 2 gives NO (NO x is conserved on timescales of the interconversion between NO and NO 2 ) and a ground-state oxygen atom, which rapidly combines with O 2 to form O 3 (Reactions R6,R7).
For a typical organic molecule, used as a basis for Reactions (R1-R6), the net reaction yields two O 3 from one OH reaction.Variations in this mechanism include compounds that produce one (e.g., carbon monoxide), three (e.g., acetaldehyde), or more O 3 molecules.While P O 3 radical chemistry is propagated by reactions involving HO x and NO x , it is terminated by reactions between the same species (Reactions R9-R13, R3').Herein arises the nonlinear dependence of P O 3 on NO x and VOCR.
Reaction (R3') is the minor channel of R3 and, depending on the identity of the VOC, its branching, k R3 / (k R3 + k R3 ), is 0-35 %.Ambient air organic mixtures are observed to have average Reaction (R3') branching ratios in the range of 2-8 % (Perring et al., 2013).At low NO x , increasing NO drives Reaction (R3) and Reaction (R5), and therefore the oxidation of VOCs and P O 3 .Chemistry in this regime is NO x -limited and increases in VOCR have little effect on P O 3 .At higher NO x levels, NO 2 begins to outcompete organics for OH (Reaction R13) and the P O 3 rate slows.Chemistry here is VOC-limited (also known as NO x -suppressed or NO x -saturated).At a given VOCR, P O 3 is maximized near the transition point between the NO x -and VOC-limited regimes.At peak P O 3 , the production of alkyl nitrates by Reaction (R3') and peroxy nitrates also reach a peak.The formation of these nitrates is important to the absolute P O 3 but has little effect on P O 3 's functional form (Farmer et al., 2011;Perring et al., 2013).When chemistry is VOC-limited, the impact of increasing the VOCR is twofold: (a) it increases the magnitude of P O 3 and (b) it shifts the position of peak P O 3 to higher NO x .Thus, the NO x concentration at which P O 3 transitions from NO xlimited to VOC-limited is VOCR dependent -a dependence that has important implications in the design of organic and NO x emission control strategies.
At VOC-limited and peak P O 3 , an increase in P HO x enhances P O 3 nearly one to one.NO].Decreasing P HO x also shifts peak P O 3 to lower NO x but to a smaller extent than the equivalent shift from decreasing VOCR.For example, at 10 s −1 VOCR and 1 ppt s −1 P HO x , a 50 % reduction in VOCR shifts peak P O 3 to ∼30 % lower NO x while the same P HO x reduction has half the effect.The photolysis of nitrile chloride (ClNO 2 ) also initiates radical chemistry in the early morning (e.g., Osthoff et al., 2008;Thornton et al., 2010;Mielke et al., 2011).ClNO 2 is not included in our model (described below) as we focus on daytime chemistry from 10:00-14:00 LT (local time) when ClNO 2 is no longer present in relevant amounts.

Analytical model of P O 3
We compute P O 3 with an analytical model similar to that described by Murphy et al. (2007) and Farmer et al. (2011).We summarize our specific formulation here.This calculation assumes HO x is in steady state (Eq.1).
Equations ( 1) and ( 2) are combined to solve for OH and HO 2 with Details are given in Appendix A.
The observational inputs to the calculation are NO and NO 2 , the total VOCR, P HO x , α, and temperature.In our analysis, VOCR, P HO x , and the NO to NO x ratio are fit as exponential functions of temperature and we use these fits to constrain the model; VOCR, P HO x , and α are discussed below.
Total VOCR is equal to the measured OH reactivity minus the OH reactivities of NO, NO 2 , HONO, HNO 3 , ammonia (NH 3 ), and sulfur dioxide (SO 2 ).
P HO x is the summed HO x production rates of O 3 , H 2 CO, HONO, and H 2 O 2 photolysis and O 3 reactions with alkenes.The contributions of O 3 and H 2 CO were 34 and 26 %, respectively.Noontime HONO concentrations during CalNex-SJV were between 30 and 250 ppt (Ren et al., 2011;Van-denBoer et al., 2013) making HONO an important daytime radical source (32 %).H 2 O 2 photolysis (1 %) and O 3 + alkene reactions (7 %) are less important to the daytime total.Photolysis frequencies are computed with the TUV (total ultraviolet) calculator (http://cprm.acd.ucar.edu/Models/TUV/Interactive_TUV) (Madronich, 1987) at 1 h resolution for a clear sky day (7 June 2010) and used to scale the 1 min measured photosynthetically active radiation to capture changes with cloud cover.The O 3 optical depth used is 300 DU, measured by OMI (ozone monitoring instrument) on 7 June 2010 (http://www.temis.nl/protocols/O3total.html)(Veefkind et al., 2006).The measurement altitude of the CalNex-SJV site is 0.14 km a.s.l.The surface albedo, 0.08, was determined using the MODIS 16 day 500 m product in band 3 (459-479 nm) for 18 June 2010.The product was reprojected at a fine resolution (0.001 • , ∼10 m) for the CalNex-SJV site-specific surface reflectivity.The photolysis rates were reduced by 10 % so that calculated P HO x ∼LHO x .We attribute the adjustment in photolysis frequencies to error associated with selection of TUV model input parameters.P HO x is averaged over the 10:00-14:00 LT window.
An effective α value is inferred from observations of O x (O x ≡ O 3 + NO 2 ) and total alkyl nitrates ( ANs) as two divided by the slope of their correlation (Perring et al., 2013).Deriving α this way treats O x and ANs concentrations as proxies for their respective production rates and assumes that, on average, two O 3 are produced per unit VOCR.During CalNex-SJV the daily average α is 2-3 % (2σ ) and we use 2.5 % as a constant in our calculation.The NO x value at peak P O 3 is sensitive to α, where a lower α value gives ∂P O 3 / ∂NO x equal to zero at higher NO x .Variability in α of ± 0.5 % is not large enough to affect our results.

CalNex-SJV measurements
CalNex was a multi-site, multi-platform field intensive that took place in the spring of 2010 across California (Ryerson et al., 2013).The San Joaquin Valley supersite (CalNex-SJV) was located at the University of California Extension -Kern County (35.346 • N, 118.965 • W).Daytime (10:00-14:00 LT) winds to the site were consistently from the northwest and at on average 2-3 m s −1 (∼1σ ).The CalNex-SJV site was 6 km southeast of downtown Bakersfield and typically less than an hour downwind.
Continuous in situ observations of over 100 VOCs, H 2 CO, peroxy acyl nitrate (PAN), peroxypropionyl nitrate (PPN), C 1 -C 3 organic acids, glyoxal, carbon monoxide (CO), methane (CH 4 ), total OH reactivity, NO, NO 2 , HNO 3 , NH 3 , SO 2 , ANs, O 3 , H 2 O 2 , and meteorological parameters (e.g., temperature, relative humidity, and photosynthetically active radiation) were collected from 18 May to 29 June 2010 atop an 18 m walk-up tower.HONO was measured at a lower tower level 15.3 m above ground.Table B1 lists the measurement uncertainty, instrumental technique, and associated reference(s) for each species used in this analysis.All observations are at 1 min time resolution with the following exceptions: VOCs (1 h), NH 3 and SO 2 (1 h), OH reactivity (10 min), and HONO (2 min).Acetaldehyde, the firstgeneration isoprene oxidation product methyl vinyl ketone (MVK), ethane, and benzene were not measured.Their concentrations were calculated via relationships with observed species (see Appendix B for details).
Throughout this analysis we use daily average concentrations (10:00-14:00 LT), as our focus is the day-to-day rather than time-of-day variability.Daily averages of hourly VOC data are 10:00-14:30 LT in order to include five data points.We use the daily maximum 1 h temperature at the CalNex-SJV site (the maximum temperature over the entire day not just the 10:00-14:00 LT window) unless otherwise noted.The daily maximum temperature is well correlated with the daily average temperature: slope = 0.91(± 0.03), y intercept = 4.72(± 0.70) • C, and R 2 = 0.99.Skies were mostly cloudless during CalNex-SJV with the measured daily average photosynthetically active radiation (PAR) independent of temperature at daily maximum temperatures greater than 22 • C.

O 3 season routine monitoring observations
We include monitoring network measurements of the daily 1 h maximum temperature, hourly O  Edison (35.346 • N, 118.852 • W) sites.We discuss the O 3 data as the daily average (10:00-14:00 LT) and in terms of the frequency of exceeding the California 8 h maximum average standard of 70.4 ppb.
Throughout this manuscript we define low, moderate, and high temperature regimes as follows: low -daily maximum temperatures of 17-27 • C (days between 1 May-31 October), moderate -daily maximum temperatures of 28-33 • C, and high -daily maximum temperatures of 34-45 • C. In 2010 there were 38 low temperature days, 54 moderate temperature days, and 80 high temperature days.These temperature cutoffs were selected in order to be consistent with Pusede and Cohen (2012).We also refer to the 2010 average low, moderate, and high temperatures, 24.7 • C, 30.8 • C, and 36.4 • C, which are averages of the 2010 1 h daily maximum temperature record and not of the 1 h daily maximum temperatures observed during the cooler, springtime CalNex-SJV study period.We use the average temperatures to streamline the discussion of our results.All figures in the manuscript are plotted along the x axis of daily maximum temperature spanning 15-40 • C.There were three days in 2010 when the daily maximum temperature exceeded 40 • C.

Organic reactivity and temperature
Organic and NO x emissions arise from multiple sources.Some sources are largely independent of temperature, for example emissions from motor vehicle exhaust and baseload power plants; others are known to increase exponentially with temperature such as biogenic emissions and the evaporation of gasoline.We examine observed daytime (10:00-14:00 LT) relationships between reactivity and temperature of more than 120 individual organic compounds and find that reactivities can be broadly categorized as either temperature independent or temperature dependent.We emphasize the observed temperature dependence is driven primarily by temperature's effect on VOC concentrations (presumably emissions) rather than by acceleration of the OH reaction rates.In Fig. 1 we show the empirically determined fits of all observed organic molecules (10:00-14:00 LT) as a function of the daily maximum temperature.The i VOCR i are grouped i VOCR i (s −1 ) versus daily maximum temperature ( • C).Data points for each curve are shown in Fig. B1.Colors are as follows: C 1 -C 2 aldehydes (red), C 1 -C 3 alcohols (pink), known biogenic VOCs (green), CH 4 (purple), temperature-dependent alkanes (orange), C 1 -C 3 organic acids (cyan), CO (gray), and temperatureindependent AVOCs (blue).For a listing of the mean VOCR i of each molecule see Table B2.
by chemical functionality when possible and we prioritize simplicity (the fewest number of groupings) over groupings by source category 1 .Figure 1 also includes reactivity contributions from four unmeasured species using computed concentrations.Each VOC included in Fig. 1 is listed in Table B2 (see Appendix B for details).The data points summarized in Fig. 1 are shown in Fig. B1.
In Fig. 1 we find that the daily average i VOCR i of certain anthropogenic VOCs (AVOCs) and CO is independent of temperature; the reactivities of C 1 -C 2 aldehydes, C 1 -C 3 alcohols, known biogenic VOCs, CH 4 , a subset of alkanes 1 For example, ethanol and methanol both have hydroxyl functional groups; however, their emission sources are distinct.Ethanol sources include evaporative emissions from decaying organic matter on farms (e.g., Alanis et al., 2010;Howard et al., 2010a, b;Malkina et al., 2011) and vehicle combustion (e.g., de Gouw et al., 2012); methanol sources include emissions from dairy wastes and also from living plants in forests and fields (e.g., Karl et al., 2001;Schade and Goldstein, 2001;Galbally and Kirstine, 2002;Fares et al., 2012).We observe that the alcohols' temperature dependence are also different, the exponential temperature dependence of methanol being more pronounced than that of ethanol (solid and dashed lines, respectively, in Fig. B1b).A second example is in the known biogenic grouping of isoprene, various monoterpenes, oxidation products of these emissions, et alia (Table B2).Isoprene and monoterpenes are emitted from different plant species, are transported to the CalNex-SJV site from different distances, and exhibit unique light, humidity, and (very high) temperature dependence (Guenther et al., 1993).
and C 1 -C 3 organic acids increase exponentially with increasing temperature.We find that at low and moderate temperatures, temperature-independent AVOCs and CO represent the largest portion of i VOCR i .At high temperatures, the small oxygenates, H 2 CO, acetaldehyde, and C 1 -C 3 alcohols, dominate.The known biogenic source is only a minor contributor to VOCR in the southern SJV even at the highest temperatures (10 % of i VOCR i and 5 % of VOCR at 36.4 • C).
In Fig. 2 i VOCR i and VOCR are organized according to temperature dependence: temperature-independent i VOCR i (blue), known temperature-dependent i VOCR i (yellow), and unknown temperature-dependent reactivity (green).The temperature-independent i VOCR i has a mean value of 2.1 s −1 ; it is equal to the sum of temperatureindependent AVOCs, CO, 0.2 s −1 H 2 CO, and 0.15 s −1 temperature-dependent alkane i VOCR i .Light-duty vehicles are known to emit these molecules, the largest portion of this vehicular source (70-90 %) being temperatureindependent tailpipe emissions (Pierson et al., 1999;Rubin et al., 2006;Gentner et al., 2009).An aside: the temperature-dependent alkane i VOCR i category has been further divided into temperature-dependent and temperatureindependent components, henceforth associating the ∼yintercept value of 0.15 s −1 (Fig. 1 orange line) with vehicle emissions (i.e., temperature-independent i VOCR i ) and the remaining reactivity, having a distinctly temperature-driven source, with evaporative emissions from gasoline vehicles and fugitive emissions from petroleum operations (Gentner et al., 2013a).
The temperature-dependent i VOCR i and temperaturedependent VOCR, the latter defined here as the measured OH reactivity minus the temperature-independent i VOCR i and OH reactivities of NO, NO 2 , HONO, HNO 3 , NH 3 , and SO 2 , are also shown.A comparison of these data indicates a portion of the reactivity is unaccounted for by VOC observations.This unknown reactivity is temperature dependent, increasing from ∼zero at low temperatures to ∼5 s −1 at high temperatures.Our estimate for VOCR acetaldehyde is uncertain (Appendix B).We include VOCR acetaldehyde in Figs. 1 and  2, as we are interested in the magnitude of the unknown as opposed to the unmeasured VOCR.Measurement uncertainties in HONO, HNO 3 , NH 3 , and SO 2 cannot fill the observed gap in reactivity at high temperatures.The total mean hightemperature reactivity of these species is 0.15 s −1 ± 21 %, or at most 0.18 s −1 , as > 85 % of this reactivity is from NH 3 (Table B1).
Identifying what molecules/sources are responsible for the unknown organic reactivity is not the aim of this work but we are able to establish the temperature response of the unknown source(s).We note that it has been previously speculated that a missing source of VOCR is important when O 3 is high in the SJV (Steiner et al., 2008;Hu et al., 2012;Pusede and Cohen, 2012).Organic sources with this temperature dependence include molecules with emission rates controlled by their vapor pressures, biogenic VOCs from forests ) daily maximum temperature ( o C) Fig. 2. Daily average (10:00-14:00 LT) organic reactivity (s −1 ) versus the daily maximum temperature ( • C) with fits: temperature-independent i VOCR i (blue), temperature-dependent i VOCR i (yellow), and temperature-dependent VOCR (green).The temperature-dependent VOCR is equal to the measured OH reactivity minus the temperature-independent i VOCR i and minus the OH reactivities of NO 2 , NO, HONO, HNO 3 , NH 3 , and SO 2 .The mean temperature-independent i VOCR i is 2.1 ± 0.3 1 (1σ ) (light blue line).The unknown reactivity is thus temperaturedependent and is the difference between the yellow and green curves.The error bars are the sum of the squared y residuals (divided by the number of points less 2) and are ± 0.3 4 (yellow) and ± 1.5 (green).
In Pusede and Cohen (2012) -using temperature as a proxy for VOCR -we showed that the frequency of 8 h O 3 violations in Bakersfield and the southern SJV region at moderate temperatures fell substantially from 1999-2010 because of VOCR reductions; the same was not true at high temperatures and we concluded that there were two reasons for Fig. 3.The effects of organic emission reductions on the total VOCR versus daily maximum temperature ( • C).We assume here that the temperature-independent i VOCR i of 2.1 s −1 has decreased and will continue to decrease at a rate of 6 % yr −1 and that the temperature-dependent reactivity has been and will continue to be unchanged.The percent difference (higher) in the total VOCR in 2000, (VOCR 2000 -VOCR 2010 ) / VOCR 2000 , is in black.The percent change (lower) in the total VOCR in 2020, (VOCR 2020 -VOCR 2010 ) / VOCR 2010 , is in periwinkle.The turquoise line is the year 2010 at 0 %.this2 .(1) At moderate temperatures O 3 chemistry was VOClimited for most of the record and, as such, the exceedance frequency was sensitive to changes in VOCR.At high temperatures, O 3 chemistry was near peak P O 3 , and recently (2007-2010) NO x -limited on weekends, and thus less sensitive to VOCR decreases.(2) There were two distinct categories of organic reactivity in the southern SJV -one source that had decreased over the last decade and was more important at moderate temperatures and a second source that dominated at high temperatures and had not substantially changed.
In light of the CalNex-SJV i VOCR i and VOCR observations and well-documented efforts to control motor vehicle emissions (e.g., Kirchstetter et al., 1999a, b;Parrish et al., 2002;Harley et al., 2006;Parrish, 2006;Bishop and Stedman, 2008) we speculate that it is the temperatureindependent reactivity that has decreased over the last decade.At the same time, we know of no deliberate attempt to control the molecules dominating the reactivity at high temperatures -small aldehydes, alcohols, and the unknown VOCR.It is therefore likely that the temperatureindependent i VOCR i of 2.1 s −1 is the endpoint of longterm efforts at organic emission reductions.
Routine summertime VOC canister sampling in the southern SJV (Appendix B) suggests a decrease of ∼6 % yr −1 in temperature-independent emissions.In what follows, we use this −6 % yr −1 rate in the temperature-independent i VOCR i and assume no change in temperature-dependent emissions to illustrate the effects of VOC controls from 2000 to 2010 on the total VOCR (Fig. 3).We show that over the last decade the impact of reductions on the total VOCR (black) have been greatest at low temperatures (50-35 %) and smallest at high temperatures (20-15 %).We show the predicted percent change in total VOCR in the next decade, assuming the temperature-independent i VOCR i continues to decrease at the same rate (periwinkle).Again, the impacts are largest at low temperatures (55-25 %) and smallest at high temperatures (15-10 %).An interesting feature in Fig. 3 is that the temperature dependence becomes more prominent in the future as the temperature-independent component declines.

Combined NO x and temperature dependence
NO x decreases by ∼60 % from weekdays (5.3 ppb) to weekends (2.3 ppb) and is observed to be temperature independent (Fig. 4a).In Fig. 4b, we combine relationships between temperature, total VOCR, and NO x (as NO x R ≡ NO x reactivity with OH).On weekends, VOCR/NO x R is more than double that on weekdays.On weekdays and weekends the observed VOCR/NO x R increases across the temperature range of 15-40 • C by a factor of 5 and 10, respectively.
Figure 4b suggests that the combined NO x and VOCR relationships with temperature impart a dependence on temperature of the sensitivity of P O 3 to its precursor emissions.To investigate this, we calculate P O 3 as described in Sect.2.2 using the temperature relationships of VOCR, P HO x , and NO / NO x observed during CalNex-SJV (Figs. 5a-c).P HO x increases with increasing temperature.This temperature dependence is driven by the HO x source molecules O 3 , HONO, and H 2 CO, each increasing with increasing temperature so that the relative contributions are similar across the 15-40 • C range.NO / NO x decreases with increasing temperature, this ratio being set by the O 3 concentration (Fig. 5d).For context, we also show the daily average (10:00-14:00 LT) O 3 during CalNex-SJV (magenta) and over the entire 2010 O 3 season (1 May-31 October) (gray).Fitting the 2010 O 3 (gray) versus daily maximum temperature to a line (not shown) yields a weekday slope that is 28 % steeper on weekdays than on weekends.
Large changes in NO x , like those occurring weekday to weekend, impact OH concentrations; this change in OH has a feedback on VOCR and NO x .Higher OH reduces total VOCR, as oxidized organic molecules are usually less reactive than their parent VOC.OH has the same functional form as P O 3 and, analogously, is least responsive to changes in NO x near peak P O 3 .Higher average weekday O 3 (gray) at high temperatures indicates chemistry is NO x -limited (a   point examined in detail in Sect.4.3).Equal weekday-toweekend percent decreases in OH occur alongside decreases in P O 3 .An equivalent reduction in the OH-reaction removal rate of organic emissions is implied; however, enhanced weekend VOCR at high temperatures is not observed (Fig. 5a).A compensating increase in organic emissions on weekends is unlikely, as high-temperature reactivity is dominated by molecules with emission rates controlled by temperature and not by human activity (Sect.4.1).Temperatureindependent i VOCR i is also invariant with day of week (not shown).We are not able to definitively explain the absence of a day-of-week dependence of VOCR but suggest it is a combined effect of (a) the specific VOCR mixture in the southern SJV, (b) the limited statistics, and (c) the variability with temperature.

Ozone production rates and the impacts of emissions controls
Ozone production rates calculated using Eqs.(1-2) are useful for interpreting the observed frequency of O 3 violations and the sensitivity of P O 3 chemistry to emission changes.In Fig. 6 we map P O 3 (ppb h −1 ) as a function of NO x and temperature, with the latter as a surrogate for VOCR (the fit from Fig. 5a).The average weekday (solid line) and week-  at 4, 8, 12, 16, 20, and 24 ppb h −1 as a function of NO x mixing ratio versus the daily maximum temperature ( • C).Temperature here is a surrogate for VOCR.P O 3 is calculated with an analytical model constrained with the CalNex-SJV observations (see text for details).NO x (ppb) is shown for weekdays (solid brown line) and weekends (dashed brown line).The equivalent NO x R are 1.5 and 0.6 s −1 , respectively.The symbols are NO x mixing ratios at 2010-average low (purple), moderate (blue), and high (red) temperatures on weekdays (closed circles) and weekends (open diamonds) for reference.The arrow along the right-hand side represents the prevailing daytime (10:00-14:00 LT) wind direction with respect to NO x to the CalNex-SJV site, i.e., the wind travels from the higher NO x Bakersfield city center to the lower NO x measurement site.end (dashed line) NO x mixing ratios are shown.In Table 1 we report P O 3 at average low (purple), moderate (blue), and high (red) temperatures.Note the exponential form of the temperature dependence of the transition point between NO x -limited and VOC-limited regimes.At low temperatures (the 2010 average), P O 3 is VOC-limited and increases weekday to weekend.At moderate temperatures, weekday P O 3 is VOC-limited (1 ppb NO x right of peak P O 3 ) but decreases by 17 % on weekends.At high temperatures, P O 3 is at the peak with respect to NO x on weekdays and decreases by 42 % on weekends, when P O 3 is NO x -limited.
We compare the results of our CalNex-SJV parameterized P O 3 calculation to the complete Bakersfield 2010 O 3 record (Table 1).At low temperatures, there were no exceedances of the 8 h O 3 standard on either weekdays or weekends.At moderate temperatures there were 0.22 ± 0.09 exceedances per day on weekdays and 0.25 ± 0.11 exceedances per day on weekends.This is a 14 % increase in the frequency of O 3 exceedances weekday to weekend.At high temperatures there were 0.76 ± 0.07 exceedances per day on weekdays and 0.43 ± 0.10 exceedances per day on weekends, a 43 % reduction in the frequency of violations in response the changes in emissions, particularly the 60 % NO x reduction on weekends.Uncertainties in exceedance frequency are treated as counting errors as 0.5(N) 1/2 /N , with N being the total numbers of days in that day-of-week and temperature bin.We see that at high temperatures, the weekday-to-weekend change in exceedance probability (43 %) is comparable to the 42 % decrease in calculated P O 3 .
To make sense of the comparison at moderate temperatures and between temperature regimes requires consideration of an additional variable -that the O 3 concentration is an integrated function of P O 3 across the span of an urban plume.In Fig. 6, placing NO x on the y axis illustrates this: directly upwind of the CalNex site was the Bakersfield city center, where NO x emissions are higher and, according to Fig. 6, P O 3 is higher as well.The CalNex-SJV site was located at the plume edge (OMI satellite imagery) where gradients are steep and the upwind impact is large.At high temperatures, because weekday chemistry is near peak P O 3 and because peak P O 3 spans the widest range in NO x at the highest VOCR, upwind and local NO x control over production are similar, as such, P O 3 and the exceedance frequency are in close agreement.At moderate temperatures, we compute that a 60 % decrease in NO x is enough to transition the calculation of instantaneous P O 3 to NO x -limited chemistry at the CalNex-SJV site but the exceedance data suggest this is not true upwind or for the integrated P O 3 .Likewise, violations are approximately 40 % more frequent at the equivalent P O 3 at high temperatures than at moderate temperatures.
We use our constrained P O 3 calculation to test the effects of possible controls on emissions (Fig. 7).We consider the following three strategies: −50 % NO x (panels a and d), −50 % temperature-independent i VOCR i (panels b and e), and −50 % NO x and −50 % temperature-independent i VOCR i (panels c and f).If temperature-independent reactivity continues to decrease at a rate of 6 % yr −1 then the 50 % emission decreases are expected by ∼2021.Likewise, NO x decreases in Bakersfield have been observed (from space) at approximately the same rate (Russell et al., 2012).NO x controls on heavy-duty diesel trucks that are in the implementation phase (McDonald et al., 2012) will reduce NO x emissions without concomitant VOC decreases and bring about the −50 % NO x scenario relatively quickly.Figure 7 shows the absolute P O 3 (ppb h −1 ) and percentage changes calculated during CalNex (black) on weekdays (solid line) and weekends (dashed line) for the three scenarios.We list these results at the 2010 average daily maximum temperatures in Table 1 and find our results in each case to be strongly temperature dependent.
On weekdays, a 50 % reduction in NO x increases P O 3 at temperatures lower than 29 • C (P O 3 is VOCR-limited) and decreases P O 3 at temperatures above 29 • C (P O 3 is NO xlimited).On weekends, a 50 % reduction in NO x decreases P O 3 at all temperatures (>16 • C) with the largest impacts at high temperatures.
A 50 % decrease in the temperature-independent i VOCR i has the largest impact at low temperatures (Fig. 7b, e).This is both because VOCR / NO x R is lowest here (Fig. 4b) and because the temperature-independent VOCR dominates at low temperatures (Fig. 3).Only at low temperatures are controls on temperature-independent i VOCR i more effective than equivalent NO x decreases at reducing P O 3 .
In the third scenario of simultaneous 50 % NO x and 50 % temperature-independent i VOCR i reductions (Fig. 7c, f), we find that on weekdays P O 3 increases 0-9 % at 20-27 • C and otherwise decreases at all temperatures.On weekends, P O 3 decreases by at least 28 % at all temperatures.The difference between simultaneous NO x and temperatureindependent i VOCR i decreases and NO x reductions alone are minimal at moderate and high temperatures: < 2 % more effective.
In summary, NO x controls will be immediately effective at decreasing P O 3 under the conditions when violations of the O 3 standard occur -at moderate and high temperatures.Organic emission reductions will be of limited benefit to reducing P O 3 except at low temperatures, when there were zero O 3 violations in 2010; this benefit will be further attenuated alongside co-occurring NO x decreases.A 50 % reduction in the temperature-dependent i VOCR i on top of the 50 % reduction in temperature-independent i VOCR i (not shown) reduces high-temperature (36.4 • C) P O 3 by an additional 6 % on weekdays and 1 % on weekends.A 50 % reduction in the total VOCR, including reactivity from unknown molecules, decreases P O 3 at high temperatures by 25 % and 6 % on weekdays and weekends, respectively.However, VOCR reductions of equal effect to equivalent NO x controls require development of new approaches, as existing controls do not target the molecules dominating the reactivity at high temperatures and existing measurement networks do not observe them.Planning and executing reductions in this VOCR will likely require years of effort -a time period during which NO x controls on diesel trucks will be fully implemented and other strategies for NO x reduction could be developed with knowledge of how to monitor their success.
One consequence of the local photochemistry moving to a NO x -limited regime is that in the future the temperature dependence of P O 3 will diminish; at low enough NO x levels P O 3 will be temperature independent, as temperature-driven increases in VOCR will not increase P O 3 .This is visualized in Fig. 7 in the comparison of the black solid (higher NO x emissions) and brown dashed lines (lower NO x emissions).In the future, we therefore expect less variability in P O 3 and, by extension, less variability in the frequency of O 3 exceedances with temperature.

Summary
Using CalNex-SJV observations of organic molecules, OH reactivity, O 3 , and nitrogen oxides we describe relationships between temperature, i VOCR i , VOCR, NO x , and P O 3 .We find the i VOCR i in the southern San Joaquin Valley (SJV) has a temperature-independent component with a reactivity of 2.1 s −1 and a temperature-dependent component that increases exponentially with temperature.The temperatureindependent i VOCR i is composed of organic molecules associated with motor vehicle emissions.The known molecules contributing to the temperature-dependent VOCR are dominated by small aldehydes and alcohols; however, unidentified molecules are the largest source of reactivity at high temperatures.The OH reactivity of NO x is temperature independent and decreases by ∼60 % from weekdays to weekends.
We compute P O 3 using an analytical model constrained by the CalNex-SJV measurements.In response to the observed NO x decreases (at ∼constant VOCR) occurring on weekends, we calculate that P O 3 increases at low temperatures, decreases by 17 % at moderate temperatures, and decreases by 42 % at high temperatures (Fig. 6).We show consistency between weekday-to-weekend percent changes in the frequency of exceedances of the California 8 h O 3 standard (70.4 ppb) and the constrained model predictions suggesting the results of this short-term field experiment, viewed in the framework of temperature, give insight into the entire O 3 season in the southern SJV.
Using this same P O 3 model we estimate the effects of possible emission control scenarios and show the impacts are variable with temperature in both sign and magnitude.We conclude that NO x reductions will be immediately and incrementally productive at reducing P O 3 on weekends at moderate temperatures and on both weekdays and weekends at high temperatures, which is when exceedances are most frequent.Reductions of the temperature-independent organic reactivity will be most effective at low temperatures; however, there were no violations of the California 8 h O 3 standard in 2010 at these temperatures.The impact of reductions of organic emissions on O 3 violations at moderate and high temperatures will be further diminished with co-occurring NO x decreases.

CalNex-SJV measurements and i VOCR i
Table B1 lists each observation/class of observed compounds, the measurement uncertainty, analytical technique, and associated reference included in our analysis.
In Fig. B1 we show the daily averaged (10:00-14:00 LT) i VOCR i versus daily maximum temperature for C 1 -C 2 small aldehydes (panel a), C 1 -C 2 alcohols (panel b), known biogenic VOCs (panel c), CH 4 (panel d), temperaturedependent alkanes (generally those with <C 10 ) (panel e), C 1 -C 3 organic acids (panel f), the sum of all alkanes, alkenes, aromatics, and carbonyls observed to be independent of temperature (panel g), and CO (panel h).In panels a-f gray lines are the empirical exponential fits of reactivity as a function of the daily maximum temperature.In panels g-h the gray lines are the average reactivity.Linear fits to these data give slopes that are not meaningfully different than zero.The C 1 -C 2 aldehydes are H 2 CO and acetaldehyde (acetaldehyde's concentration was estimated, not measured; see below).The C 1 -C 3 alcohols are methanol, ethanol, and isopropanol.Key molecules in the known biogenic grouping are isoprene, methacrolein, MVK, glyoxal, limonene, and α-and β-pinene.C 1 -C 3 organic acids are formic acid, acetic acid, and propionic acid.All alkanes (panels e and h), alkenes, aromatics, and carbonyls, along with all other species included in Figs. 1 and B1 are listed in Table B2.Table B2 includes each compound's mean reactivity at low, moderate, and high daily maximum temperatures.If a temperature dependent OH rate expression is available we use the time varying temperature over the same time interval, not the daily maximum temperature.If no rate constant is available one is assigned by analogy to a compound with similar molecular structure, indicated with a double asterisk and an explanatory note.In order to sum the reactivities of many species, null daily average concentrations were filled using the temperature-dependent fit or average value as dictated by the observed temperature relationship.Methyl ethyl ketone is included in the temperature-dependent alkane i VOCR i , as it exhibits clear temperature dependence; this was done so as to avoid introducing another temperature-dependent grouping.In a handful of instances, individual daily average VOCR i outliers were removed prior to computing the temperature-dependent fit or mean.
The mixing ratio of acetaldehyde is estimated assuming (a) acyl peroxy nitrate products and aldehyde sources are co-located, i.e., that PPN : propanal is equal to PAN : acetaldehyde, and (b) that acetaldehyde is the only PAN source molecule.Methacrolein, biacetyl, MVK, and methyl glyoxal are also known PAN sources.Biacetyl, MVK, and methyl glyoxal were not measured during CalNex-SJV, so to test this second assumption, we approximated their concentrations via ratios to methacrolein, as observed during the Biosphere Effects on AeRosol and Photochemistry Experiment (BEARPEX) (LaFranchi et al., 2009), which took place 15 June-31 July 2007 at a site located ∼3 h downwind from an isoprene source region.Using the calculated acetaldehyde (Fig. B1a), approximated biacetyl, MVK, and methyl glyoxal, and the measured methacrolein we find acetaldehyde to be ∼80 % of the PAN source (compared to <40 % at 30 • C at the BEARPEX site).The dominance of acetaldehyde to PAN formation is not unexpected in the southern SVJ given the low VOCR isoprene and VOCR methacrolein .Alternatively, acetaldehyde can be calculated assuming steady state relationships with PAN, HO 2 , RO 2 , NO, and NO 2 .We find this approach yields an estimated VOCR acealdehyde that is ∼3 times greater than the one we use in this analysis but with the same temperature dependence as we report.The larger acetaldehyde value would reduce the inferred missing VOCR by ∼20 % but would not affect any other conclusion in this analysis.
MVK is estimated as 3 times the methacrolein concentration.This is the same ratio of MVK to methacrolein observed during BEARPEX-2007(LaFranchi et al., 2009).We do not include an estimate for second-generation isoprene oxidation products.Due to the low levels of isoprene and methacrolein we expect second-generation oxidation product concentrations to also be small.
Ethane and benzene were estimated via observed relationships with propane and toluene, respectively, using the PAMS data record (Appendix C).    (i) 13

Fig. 4 .
Fig. 4. (a) Daily average (10:00-14:00 LT) NO x R (s −1 ) versus the daily maximum temperature ( • C) separated by weekdays (closed circles) and weekends (open diamonds).Weekdays are Tuesdays-Fridays.Weekends are Saturdays, Sundays, and Memorial Day (31 May 2010).Mondays and Saturdays are considered transition days, as they are influenced by carryover from the previous day.We omit Mondays for this reason but retain Saturdays to improve weekend statistics.The solid line is the weekday average value of 1.5 s −1 and the dashed line is the weekend average value of 0.6 s −1 .(b) Total VOCR per unit NO x R for weekdays (closed circles) and weekends (open diamond) versus the daily maximum temperature ( • C).The curves are exponential fits to the weekday (solid gray line) and weekend (dashed gray line) data.

Fig. 5 .
Fig. 5. (a) Total VOCR (s −1 ) (measured OH reactivity minus the OH reactivity contributions of NO 2 , NO, HONO, HNO 3 , NH 3 , and SO 2 ; this is equivalent to the sum of the green and blue data in Fig. 2).(b) P HO x (ppt s −1 ).(c) NO / NO x .(d) O 3 (ppb) during CalNex-SJV (magenta) and in Bakersfield during the entire 2010 O 3 season (gray).All data are daily averages (10:00-14:00 LT) versus daily maximum temperature ( • C), are shown for weekdays (closed circles) and weekends (open diamonds), and include their exponential fits.The 2010 Bakersfield O 3 data are fit separately for weekdays (solid black line) and weekends (solid gray line).

Fig. 6 .
Fig.6.P O 3at 4, 8, 12, 16, 20, and 24 ppb h −1 as a function of NO x mixing ratio versus the daily maximum temperature ( • C).Temperature here is a surrogate for VOCR.P O 3 is calculated with an analytical model constrained with the CalNex-SJV observations (see text for details).NO x (ppb) is shown for weekdays (solid brown line) and weekends (dashed brown line).The equivalent NO x R are 1.5 and 0.6 s −1 , respectively.The symbols are NO x mixing ratios at 2010-average low (purple), moderate (blue), and high (red) temperatures on weekdays (closed circles) and weekends (open diamonds) for reference.The arrow along the right-hand side represents the prevailing daytime (10:00-14:00 LT) wind direction with respect to NO x to the CalNex-SJV site, i.e., the wind travels from the higher NO x Bakersfield city center to the lower NO x measurement site.

Fig. 7 .
Fig. 7. (a-c) P O 3 (ppb h −1 ) on weekdays (solid line) and weekends (dashed line) versus daily maximum temperature ( • C) calculated with CalNex-SJV observations (black) and in (a) with a 50 % NO x reduction (brown), in (b) with a 50 % reduction in the temperature-independent i VOCR i (green), and in (c) with both reductions applied (purple).(d-f)the percent change between P O 3 on weekdays (solid line) and weekends (dashed line) during CalNex-SJV and that computed after emission reduction versus daily maximum temperature ( • C).
Fig. B1.Daily average i VOCR i (s −1 ) versus the daily maximum temperature ( • C) for (a) H 2 CO (closed circles with black outline) and calculated acetaldehyde (open circles), (b) methanol (closed circles with black outline) and ethanol (open circles), c) known biogenic VOCs, (d) CH 4 , (e) temperature-dependent alkanes, (f) formic acid (closed circles with black outline) and acetic acid (open circles), (g) all temperature-independent AVOCs, and (h) CO.Gray solid and dashed lines are empirically derived exponential fits for (a-f).The solid lines in(a, b, f) are fits to the C 1 molecules; the dashed lines are fits to the C 2 molecules.In (g) and (h), gray lines are means.Acetaldehyde, MVK, ethane, and benzene are estimated (see Appendix B for details), otherwise VOCR i are derived from measured concentrations.In (a), the red dashed line shows the fit to the acetaldehyde computed from steady-state relationships, as discussed in Appendix B.

Table 1 .
P O 3 (ppb h −1 ) computed with an analytical model parameterized with CalNex-SJV observations (the organic reactivity is equal to VOCR, not i VOCR i ).Row 1: P O 3 calculated at the 2010 average low (24.7 • C), moderate (30.8 • C), and high (36.4• C) temperatures.Row 2: number of violations of the 8 h California O 3 standard in Bakersfield in 2010 (day −1 ) by temperature and by weekday/weekend.Uncertainties are computed as counting errors as 0.5(N ) 1/2 /N , where N is the total numbers of days in that temperature and day-of-week regime.Rows 3-5: P O 3 (ppb h −1 ) calculated with 50 % lower NO x , with 50 % lower temperature-independent i VOCR i , and with both 50 % lower NO x and 50 % lower temperature-independent i VOCR i .Also shown are the percent differences (%) between the 2010 CalNex constrained P O 3 and that computed with the emissions reduction.P O 3 is rounded to the nearest parts per billion per hour ( ppb h −1 ).The relevant uncertainty is not in P O 3 itself but in the derivative of P O 3 with respect to NO x .

Table B1 .
Species, measurement accuracy, analytical technique, and reference for each CalNex-SJV observation included in our analysis.Many species are measured with higher precision than accuracy.See original references for details.All CalNex data are accessible to the public at http://www.esrl.noaa.gov/csd/groups/csd7/measurements/2010calnex/.

Table B2 .
Mean VOCR i of all species included in i VOCR i ± 1σ , indicating the observed variability, over low (17-27 • C), moderate (28-33 • C), and high (34-45 • C) daily maximum temperatures.Species are arranged by Fig.1grouping and listed greatest to smallest hightemperature per second.The VOC i + OH rate constants or temperature-dependent rate expressions are also listed with associated references.The concentrations of the starred species, acetaldehyde, MVK, ethane, and benzene, were estimated (Appendix B).Double starred k OH values were estimated by analogy to compounds with similar molecular structure and these include explanatory notes.