On the observed response of ozone to NO x and VOC reactivity reductions in San Joaquin Valley California 1995 – present

We describe the effects of nitrogen oxide (NO x) and organic reactivity reductions on the frequency of high ozone days in California’s San Joaquin Valley. We use sixteen years of observations of ozone, nitrogen oxides, and temperature at sites upwind, within, and downwind of three cities to assess the probability of exceeding the California 8h average ozone standard of 70.4 ppb at each location. The comprehensive data records in the region and the steep decreases in emissions over the last decade are sufficient to constrain the relative import of NO x and organic reactivity reductions on the frequency of violations. We show that high ozone has a large component that is due to local production, as the probability of exceeding the state standard is lowest for each city at the upwind site, increases in the city center, is highest at downwind locations, and then decreases at the receptor city to the south. We see that reductions in organic reactivity have been very effective in the central and northern regions of the San Joaquin but less so in the southern portion of the Valley. We find evidence for two distinct categories of reactivity sources: one source that has decreased and dominates at moderate temperatures, and a second source that dominates at high temperatures, particularly in the southern San Joaquin, and has not changed over the last twelve years. We show that NOx reductions are already effective or are poised to become so in the southern and central Valley, where violations are most frequent, as conditions in these regions have or are transitioning to NO x-limited chemistry when temperatures are hottest and high ozone most probable.


Introduction
Ozone formation is a nonlinear function of nitrogen oxides (NO x ) and the reactivity of gas phase organic molecules and consequently, reductions in the emissions of these precursors can decrease, increase, or leave unchanged the rate of ozone production.Emissions control policies aimed at improving ozone (O 3 ) air quality therefore require sufficient information on how the chemical system at a given location will respond to reductions in precursor concentrations.Over the last decade there have been dramatic reductions in NO x concentrations across North America and Europe (e.g.Richter et al., 2005;Kim et al., 2006Kim et al., , 2009;;Stavrakou et al., 2008;van der A et al., 2008;Konovalov et al., 2010;Russell et al., 2010Russell et al., , 2012)).At many locations there are reports of decreases in organic emissions (e.g.Environmental Protection Agency, 2003;Parrish, 2006;Bishop and Stedman, 2008;Monks et al., 2009;Wilson et al., 2012) but changes to the total organic reactivity are not well documented.These precursor changes are predicted to have substantially affected the photochemical ozone production rate and thus the probability of exceeding health-based standards.Reports of improved air quality are mixed and there has been little success in attributing quantitative measures of changes in ozone concentrations to the reductions of specific emissions.
A variety of observational and modeling approaches have been used to evaluate ozone's sensitivity to NO x and organic reactivity.These include analyses of ratios of peroxides to nitric acid (e.g.Sillman et al. 1995;Sillman et al., 1997), relationships between measured nitrogen oxides and organic molecules (e.g.Kleinman et al., 2000Kleinman et al., , 2005;;Trainer et al., 2000;Martin et al., 2004;Stephens et al., 2008;Pollack et al., Published by Copernicus Publications on behalf of the European Geosciences Union.2012), rates of ozone production derived from observed reactant concentrations (e.g.Thornton, et al., 2002;Martinez et al., 2003;Ren et al., 2003), and, very recently, the direct measurement of the instantaneous ozone formation rate (Cazorla and Brune, 2010;Cazorla et al., 2012).These methods each work to constrain the chemistry of ozone production at the specific local NO x and organic reactivity.Predictions of the effects of emissions reductions are usually based on models that hindcast a small subset of historical high ozone episodes.These studies typically implement a given percentage reduction in NO x and/or organic emissions and calculate whether O 3 would have indeed been reduced during that episode.However, the short-time and/or limited-spatial scales of these measurement and modeling analyses make it difficult to assess the accuracy of the predictions.For example, we know of no case where a quantitative prediction of the reduction in the number of annual violations of a health-based standard was made in advance of a policy and then explicitly verified with observations after the fact.
Growth in the observational database and the increase in computational power have made it possible to think about ozone statistics over wide regions of space and over long periods of time instead of focusing on individual episodes.For example, Gilliland et al. (2008) examined models and observations before and after the implementation of controls on electric generating utilities in the eastern US and used the ensemble to suggest that air quality models underestimated the benefits of the NO x reductions.In this paper, we describe changes in the frequency of high ozone days and show that the existing routine observations of O 3 , nitrogen oxides, and temperature can provide direct insight into the probabilistic response of ozone to emission reductions.We develop our methodology using the example of California's San Joaquin Valley (SJV), a region competing with the Los Angeles basin for the most frequent high ozone days in the United States (American Lung Association, 2011) and where ambient O 3 concentrations persistently violate health-based air quality standards (Cox et al., 2009) despite sustained scientific attention (Venkatram et al., 1994;Andreani-Aksoyoglu et al., 2001;Marr et al., 2002a;Marr et al., 2002b;Steiner et al., 2006Steiner et al., , 2008;;Lin et al., 2008;Howard et al., 2008Howard et al., , 2010a,b;,b;Hu et al., 2012) and regulatory efforts at both the local (e.g.San Joaquin Valley Air Pollution Control District, 2007) and state level (California Air Resources Board, 2011).We use the results from our statistical approach to make policyrelevant conclusions about how the frequency of high O 3 in the SJV will respond to future NO x and organic reactivity emissions reductions.We note that the data to support this type of analysis are available at many locations in North America and Europe.

Ozone production
Photochemical ozone production results from a pair of catalytic cycles initiated by creation of odd-hydrogen (OH or HO 2 ) or organic peroxy radicals (RO 2 ), collectively referred to as HO x (HO x ≡ OH + HO 2 + RO 2 ) (Fig. 1).Entering the HO x cycle, a generic organic molecule is oxidized by OH, forming RO 2 , then HO 2 , and subsequently regenerating OH (Fig. 1a).This cycle drives the oxidation of NO to NO 2 twice (Fig. 1b).The photolysis of NO 2 is rapid and the product oxygen atom combines with O 2 to yield O 3 .During the daytime, HO x chain lengths are long enough that the ratio of HO 2 to RO 2 is near one.
Figure 2 shows the nonlinear dependence of the instantaneous rate of O 3 production (P O 3 ) on NO x (NO 2 + NO) and the organic reactivity (VOCR).Moving left to right, i.e. from a scenario of remote continental to urban photochemistry, P O 3 grows steeply with increasing NO x abundance, reaches a peak, and then decreases with continued NO x increases.This initial rise results from NO x 's role as modulator of the (HO 2 + RO 2 ) to OH ratio.At low NO x , adding NO enhances OH via reactions between NO and HO 2 or RO 2 and thereby the oxidation rate of organic molecules (NO x -limited chemistry).Because OH is typically 100 times less abundant than HO 2 or RO 2 , this has little effect on the comparatively large HO 2 + RO 2 reservoir.At high NO x , OH reacts with NO 2 to form nitric acid reducing the HO x radical pool (NO x -suppressed chemistry).In the intermediate regime, reactions forming alkyl and peroxy nitrates are important to the absolute rate but do not strongly affect the shape of the curves (Farmer et al., 2011).
Participating organic molecules are commonly referred to as volatile organic compounds (VOCs) distinguishing them from low vapor pressure species that are instead more likely to condense onto aerosol surfaces.The impact of any individual VOC to ozone production depends mainly on its reaction rate with OH (except for a small subset of VOCs that are photolabile); rapidly reacting molecules such as alkenes and aldehydes are disproportionally important compared to less reactive alkanes, acids, and ketones.The rate at which the sum of all VOCs reacts with OH is defined as the VOC reactivity (VOCR).This is a condensed parameter summarizing the integrated effects of the local VOC mixture.In Fig. 2, we show P O 3 calculated with three different VOCRs: a base case, twice the base VOCR, and three times the base VOCR.Note that at the left of Fig. 2 (low NO x ), the VOCR has no effect on the rate of O 3 production, while at the right, P O 3 increases with VOCR almost linearly (VOC-limited chemistry).
Just as decreases in VOCR decrease P O 3 , so will reductions in the rate of HO x production (P HO x ), as a shrinking HO x pool will slow VOC oxidation rates (not shown).P O 3 scales nearly linearly with P HO x , its response smaller at low NO x than high.Net sources of HO x include the photolysis of O 3 , formaldehyde and other aldehydes, nitrous acid, and nitryl chloride, reactions between O 3 and alkenes, and organic radical reactions that amplify rather than merely propagate OH and HO 2 .P HO x and VOCR are linked.For example, formaldehyde is both a primary anthropogenic emission and is an oxidation product of virtually every gas phase organic molecule.Formaldehyde is also reactive with OH and, after oxidation, enters the HO x cycle at HO 2 formation directly.Emissions reductions targeting formaldehyde and/or any of its precursors will have the combined effect of simultaneously reducing P HO x and VOCR.In addition, VOC emission controls that improve O 3 air quality will also decrease P HO x .The photolysis of O 3 is the single largest HO x source in many locations and lower O 3 concentrations impact P HO x in a positive feedback resulting in further decreased ozone production rates.That said, in the SJV the average Valleywide summertime (June-August) 8-h O 3 has varied by less than 16 ppb in the last twelve years (it was 70.2 ppb in 1999 and 66.4 ppb in 2010).In the analysis that follows, we make no attempt to tease apart the effects of P HO x from those of VOCR as data do not exist with which to do this; we acknowledge that our "VOCR" likely includes a component due to changes in HO x sources.
We illustrate the change in ozone production in response to three scenarios of NO x and/or VOCR reductions with dashed lines in Fig. 2: Scenario A decreases NO x at constant VOCR (1 → 2 → 3).NO x reductions initially increase P O 3 at high NO x (1 → 2) followed by a decrease in P O 3 at low NO x (2 → 3).This scenario occurs on weekends in locations where dramatic reductions in diesel truck traffic result in lower NO x emissions alongside small changes in VOCR.
Scenario B decreases VOCR at constant NO x (2 → 4).VOC reductions have the effect of proportionally reducing P O 3 at high NO x and of negligibly changing P O 3 at low NO x .This scenario occurs in regions where NO x emissions are constant and VOC emissions are exponential with temperature.One such example is in forested regions downwind S. E. Pusede and R. C. Cohen: Ozone, NO x , and VOC reactivity in the San Joaquin Valley of cities where VOCR is largely biogenic and higher at hotter temperatures (e.g.LaFranchi et al., 2011).
Scenario C reduces NO x and VOCR simultaneously (2 → 5).This transition is typical of what has occurred over the last decade in cities where vehicular emissions dominate both NO x and VOCR.

Ozone production, O 3 concentration, and the frequency of high O 3 days
The atmospheric O 3 concentration is a function of the timeintegrated effects of P O 3 , chemical and depositional loss, and mixing.All of these terms vary and often co-vary.Over the time interval of our study, we expect no significant changes in the chemical or depositional loss terms or in the frequency of stagnation in the SJV.Trends in the mean, median, and width of the distribution of ozone concentrations -observed to be Gaussian in our dataset -are thus dominated by the statistics of changes in P O 3 .Moreover, O 3 exceedances varying in the nonlinear manner shown in Fig. 2, as we will show they do, bolster the notion that production is the principal term changing over time.To make the association between the O 3 concentration and the frequency of high ozone days, we take advantage of the statistical properties of normal distributions.Specifically, the cumulative probability of the portion of a normal distribution above a particular threshold varies linearly with shifts in the mean (assuming the width is constant) so long as the threshold is within one standard deviation of the mean, or between approximately 15 % and 85 %.On this basis, we hypothesize that the curves representing P O 3 in Fig. 2 also describe the statistics of high ozone days and use this conceptual framework, which in our analysis we support empirically, to interpret observed changes in the probability of high ozone defined as the fraction of days exceeding the 8-h O 3 California Ambient Air Quality Standard (CAAQS) of 70 ppb (>70.4 ppb).

NO x
NO x abundances across California have fallen at near constant rates over the last decade; this is consistent with our understanding of trends in emissions (Cox et al., 2009;Millstein and Harley, 2010;Dallmann and Harley, 2010) and supported by surface measurements (Ban-Weiss et al., 2008;Lafranchi et al., 2011;Parrish et al., 2011) and space-based observations (Kim et al., 2009;Russell et al., 2010Russell et al., , 2012)).These NO x decreases have had led to striking improvements in ozone air quality in the Sacramento Valley (Lafranchi et al., 2011) but less so in the Los Angeles basin, where chemistry remains NO x -suppressed and the dramatic improvements of the 1980's and 1990's have slowed (e.g.Pollack et al., 2012).In the SJV, both satellite NO 2 and the groundbased nitrogen oxide data records indicate steady decreases of approximately 5 % per year Valley wide (Russell et al., 2010(Russell et al., , 2012)).
In addition to long-term reductions, NO x concentrations have a well known day-of-week dependence.In the SJV, NO x is typically 30-50 % lower on weekends than weekdays, a phenomenon largely due to reduced weekend heavyduty diesel truck traffic (e.g.Marr et al., 2002b;Millstein and Harley, 2010).Meteorological and chemical conditions, such as VOCR, are far less day-of-week dependent than are changes in NO x and, as a result, comparison of weekdays to weekends is an effective and widely used tool to study the NO x dependence of O 3 formation (e.g.Murphy et al., 2006Murphy et al., , 2007;;Stephens et al. 2009;LaFranchi et al., 2011;Pollack et al., 2012).
In this work, we consider both annual and day-of-week NO x trends comparing curves describing weekday and weekend O 3 CAAQS exceedance probabilities over the past sixteen years.We note that the NO 2 data presented here are obtained by chemiluminescence coupled with a heated molybdenum catalyst, a technique with a known positive interference from the higher oxides of nitrogen (alkyl and peroxy nitrates and nitric acid).We refer to measured "NO 2 " as NO 2 * hereafter (a more detailed description of all measurements is found in the Appendix).To a reasonable approximation NO 2 is a constant fraction of NO 2 * at a given location at a given time of day (Dunlea et al., 2007).

VOCR and temperature
Tailpipe emissions from vehicles are only weakly temperature dependent, for example due to the increase in fuel consumption for air conditioning on hot days.By contrast, biogenic VOCs from forests (e.g.Guenther et al., 1993;Schade and Goldstein, 2001) and agriculture (e.g.Ormeño et al., 2010) are emitted as an exponential function of temperature until, for certain species, inhibited by extreme heat.Vapor pressures rise exponentially with temperature and so evaporative emissions, such as from fuels and farm residues, are also strongly temperature dependent (Rubin et al., 2006).Temperature also influences the rates of reaction of organic molecules with OH and of radical cycling, but this effect is much smaller than that due to the increase in VOC abundance (Steiner et al., 2006).
There is evidence for decreases in both the concentrations (Harley et al., 2006) and emissions (Cox et al., 2009) of some VOCs in the SJV over the last twenty years.Observations in other locations indicate VOCs and NO x emissions from passenger vehicles have decreased in tandem (e.g.Parrish et al., 2002;Parrish, 2006).However, how or if these reductions have broadly translated to decreases to total reactivity is not known as VOC measurements do not necessarily include VOCR's major components.Observations of VOCR are not generally available because techniques for the direct measurement of OH reactivity have only recently been developed (Kovacs et al., 2001;Sadanaga et al., 2004a;Sinha et al., 2008;Ingham et al., 2009).The use of these techniques is still limited to large-scale field experiments and at most Northern SJV (d bottom).13 sites observations of individual VOCs do not add up to the total VOCR measured (e.g.Kovacs et al., 2003;Di Carlo et al., 2004;Sinha et al., 2008;Ingham et al., 2009;Lou et al., 2010;Sinha et al., 2010).In the SJV, we show temperature is a useful surrogate for VOCR insofar as we recreate distinct curves analogous to Fig. 2 by organizing observations by temperature (details to follow).Meteorological conditions conducive to high ozone, including stagnation events and clear skies, correlate with increasing temperature.We group data into two temperature regimes, high (34-45 • C) and moderate (28-33 • C); we find these ranges are sufficiently distinct to identify differences in production of ozone (see below) while still maintaining sufficient statistics to characterize the ensemble of O 3 observations at each site.We note that in the SJV, boundary layer dynamics are strongly influenced by mountain valley flow and as a result we do not expect meteorological factors (e.g.wind speeds) that are particularly different between high and moderate temperatures.

The San Joaquin Valley
The SJV is characterized by regular airflow from north to south during ozone season (∼May-October) with background O 3 well mixed Valley wide (Zhong et al., 2004).Here we divide the SJV (Fig. 3a) into three distinct urban photochemical plumes each captured by California Air Resources Board (CARB) monitoring stations and refer to these three regions as Southern SJV (Fig. 3b), Central SJV (Fig. 3c), and Northern SJV (Fig. 3d).Within each plume we identify an upwind, city center, and downwind location all along the axis of air movement (nine locations total).We see the lowest exceedance probabilities at upwind sites (Figs.4-10 panels a), increased probabilities across the city center (Figs.4-10 panels b), and the highest probabilities at locations downwind (Figs.4-10 panels c).At the upwind site of the adjacent study regions to the south, the likelihood of a violation is again at a minimum.This is evidence for the production of ozone within each transect (details in Sect.4.4).
The bottom panels in Fig. 3 show NO 2 observations from the Ozone Monitoring Instrument (OMI) averaged for weekdays in June-August in 2007-2010 using the Berkeley High-Resolution (BEHR) product (Russell et al., 2011) 1995-1998, 1999-2002, 2003-2006, and 2007-2010. 7 Error bars are uncertainties in the four-year median exceedance probabilities, are calculated as 8 counting errors, and are typically less than ±0.04 (1σ) for weekdays and ±0.06 (1σ) for 9 weekends.The average number of days per year over the past sixteen years for both temperature 10 regimes is 66 (high) and 72 (moderate).11  1998, 1999-2002, 2003-2006, and 2007-2010.Error bars are uncertainties in the four-year median exceedance probabilities, are calculated as counting errors, and are typically less than ±0.04 (1σ ) for weekdays and ±0.06 (1σ ) for weekends.The average number of days per year over the past sixteen years for both temperature regimes is 66 (high) and 72 (moderate).
images highlight three separate NO 2 plumes in our three study areas and point to the local nature of NO x emissions (and presumably some component of VOCR) in the SJV.In what follows, we discuss each region in turn, starting in the south and moving north.

Southern San Joaquin Valley
In Fig. 4 we show the Southern SJV 8-h O 3 CAAQS exceedance probability vs. NO 2 * and in Fig. 5 we show the trend in this probability vs. year (year increases right to left analogous to NO 2 * concentration).The red symbols are statistics for high temperatures (34-45 • C) and the blue for moderate temperatures (28-33 • C).Solid symbols are weekdays (Tuesday-Friday) and open diamonds are weekends (Saturday-Sunday).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 statistics for weekends.Uncertainties in exceedance probabilities are treated as counting errors and computed as 0.5(N ) 1/2 /N, where N is the total number of days in that bin.Uncertainties are typically less than ±0.09 (1σ ) for weekdays and ±0.12 (1σ ) for weekends.Uncertainties in the four-year median probabilities are less than ±0.04 (1σ ) for weekdays and ±0.06 (1σ ) for weekends.
At high temperatures, the probability of an ozone violation at the upwind site, Shafter, decreased from 80 % on weekdays when NO 2 * was 9.8 ppb in 1996 to 30 % on weekends in 2010 when NO 2 * was 4.6 ppb (Fig. 4a).In Bakersfield, the exceedance probability fell from greater than 90 % on weekdays at 10.7 ppb NO 2 * to 75 % on weekdays at 5.7 ppb NO 2 * and 50 % on weekends at 4.0 ppb NO 2 * in 2010 (Fig. 4b).Downwind in Arvin, the probability held constant and near unity on weekdays despite an NO 2 * decrease from 9.2 to 4.4 ppb over the window of the measurements; in the last two years it fell to 60-70 % on weekends at ∼3.7 ppb NO 2 * (Fig. 4c). 1  A key observation from Figs. 4b and 4c is that the probability of an exceedance on weekends, when NO x is 30-50 % lower within a given year, is essentially identical to the weekday probability years later when the same NO x decrease is achieved.This can only occur if VOCR remained constant over that same interval (Scenario A).From the shape of the curves in Fig. 4a, we infer that P O 3 in Shafter is presently NO x -limited (to the left of peak production).In Bakersfield, the exceedance probability is NO x -limited on weekends and appears to have recently transitioned to NO x -limited chemistry on weekdays at NO 2 * less than ∼9 ppb.In Arvin, while the weekday probability of exceeding the state ozone standard has been at or near unity for the last sixteen years, we do observe a small decrease in the probability of high ozone on weekends at NO 2 * less than ∼4 ppb.We interpret the shape of these curves to indicate that we are at or near the peak of ozone production as a function of NO x in Bakersfield and Arvin.Consequently, reductions in the frequency of ozone exceedances have been slow to accrue despite a more than two-fold decrease in NO 2 *.
We estimate the effects of future NO x reductions from weekend observations (Fig. 5).Regionally, over the past four years, exceedances are less likely on weekends than weekdays at high temperatures (20 % in Shafter, 25 % in Bakersfield, and 20 % in Arvin), indicating that at each point along the Southern SJV transect at these temperatures the frequency of exceedances has indeed crossed the peak in probability and is now in a regime of NO x -limited chemistry on 1 Titration of O 3 by NO can affect the frequency of violations even when the odd oxygen, O x (O x = O 3 + NO 2 ), is constant.We checked our results using O x instead of O 3 and found no significant differences.
weekends.Although NO x decreases substantially larger than those occurring on weekends are required to eliminate violations, those reductions that do occur will be immediately effective on weekdays and even more so on weekends.
At moderate temperatures, although NO 2 * is unchanged, the observed exceedance probabilities are lower than at high temperatures.This is evidence that temperature is a proxy for VOCR (Figs. 4 and 5).A second piece of evidence is that the weekday and weekend curves vs. NO 2 * do not overlap in this temperature regime (Fig. 4).Rather, we see a different functional dependence in the probability of violations by day of week.When weekday NO 2 * matches the weekend value of several years earlier the probability of violations is noticeably lower.This implies that annual NO x reductions are attended by year-to-year changes in VOCR (Scenario C) at moderate temperatures.As shown in Fig. 5, exceedances were much more frequent on weekends than weekdays for 1995-1998 and 1999-2002, placing regional ozone chemistry to the right of peak P O 3 (NO x -suppressed).At all three locations in the last four years, the probability of a high ozone day is almost identical on weekdays and weekends indicating that at moderate temperatures Southern SJV ozone chemistry is near the peak, where the derivative with respect to NO x at the current VOCR is small.
Another perspective on the impact of NO x and VOCR reductions is shown in Fig. 6.Here, four-year median exceedance probabilities are shown as a function of NO 2 * with lines tethering weekday (solid circles) and weekend (open diamonds) conjugates.For each measurement point shown, because day-of-week variability in VOCR and meteorology is small, the weekday-weekend pair describes the NO x dependence along a single P O 3 curve.For visual aid, we have included a set of dashed lines as a qualitative description of the P O 3 curves corresponding to the data, which were created with the same equations (with tuned parameters) used to draw the curves in Fig. 2. If inter-annual decreases in NO x have occurred without simultaneous changes in VOCR, as in Scenario A, consecutive yearly weekday-weekend pairs would trace a single curve.This is what we observe at high temperatures.If VOCR changes occurred in concert with NO x reductions, as in Scenario C, the weekday-weekend pairs will each lie on separate curves.This is what we observe at moderate temperatures.We also see in Fig. 6 that the relationship between high-and moderate-temperature curves is consistent with overall lower VOCR at moderatetemperatures. We observe a shift of peak ozone production to lower rates and that the peak occurs at lower NO x concentrations.
When temperatures are highest, Fig. 6 reinforces the conclusions drawn from Fig. 4 that VOCR in Bakersfield and Arvin has been almost constant over the last twelve years, as subsequent weekday-weekend pairs each trace the same curve.Decreases in the frequency of violations are recent and appear to be solely a result of sustained NO x reductions.In contrast in Shafter VOCR reductions appear to  and Arvin (c).Uncertainties in the probability of violations (by counting statistics) are typically less than ±0.04 (1σ ) for weekdays and ±0.06 (1σ ) for weekends.Curves (dashed grey lines) are included for visual aid and are not meant to be quantitative; the lines were generated with an analytical model where only VOCR was tuned and P O 3 was then scaled to fit.have influenced the trends over time.At high temperatures, throughout the metropolitan region spanned by these three sites, conditions have transitioned to NO x -limited chemistry on weekends as depicted by the steep positive slopes of the most recent conjugates (green).
With the near unity exceedance probabilities observed in Arvin, it is possible that the O 3 concentration did actually decrease but that the normal distribution did not shift sufficiently to move any of the population below the threshold of 70.4 ppb.If this is the case then the VOCR may have also decreased.To check our conclusion in the Southern SJV, we use exceedance thresholds of 80.4 and 90.4 ppb, where the probability of exceeding these higher standards is low enough (with maximum values of 83 % and 63 %, respectively) that we expect a linear response in violations to changes in P O 3 .In Shafter, we find no difference in the slopes and in the NO 2 * and VOCR relationships depicted in Fig. 6 for either the 80.4 or 90.4 ppb standard.In Bakersfield, the shape of the curves for the 90.4 ppb standard is the same as for the 70.4 ppb; however, we find some evidence for VOCR decreases using the 80.4 ppb standard.We attribute this behavior to Bakersfield's transitional location within the plume between upwind Shafter and downwind Arvin.In Arvin (perhaps most importantly) the slopes of the three weekend-weekday conjugates and the chemical conditions they describe are unchanged; using either the 80.4 or 90.4 ppb standard we find no evidence for VOCR reductions.
In contrast to the high temperature observations, at moderate temperatures the O 3 exceedance probability has been largely NO x -suppressed over the past twelve years.In Shafter and Bakersfield, ozone production has remained NO x -suppressed since 1999 (negative slopes) with P O 3 nearing the peak (small slopes) in the last eight years (Fig. 6a  and b).In Arvin, early in the data record the sign of the slope fluctuated at constant NO x ; if the NO x level corresponds to peak ozone production then the slope is more sensitive to changes in VOCR.In 2007-2010, the slope is near zero and ozone chemistry close to peak production.Although it appears from Fig. 4 that at moderate temperatures the percentage of violations has fallen because of decreasing NO 2 *, Fig. 6 shows that VOCR reductions are the primary cause of the smaller observed exceedance probabilities at moderate temperatures.This situation is best described by Scenario C, where VOCR reductions decrease the frequency of violations and also shift peak P O 3 to lower NO x .
Taken together, distinct behavior in the two temperature regimes provides evidence for two classes of VOCR sources in the Southern SJV.One class has decreased over the last twelve years and is a large VOCR source at moderate temperatures.Another class dominates at high temperatures, has not decreased, and at high temperatures far exceeds the moderate temperature source.

Central San Joaquin Valley
The past decade has seen the 8-h O 3 CAAQS exceedance probability in the Central SJV fall by almost 50 % both on weekdays and weekends when temperatures are highest (Fig. 7; note that year increases right to left in analogy to 37  1995-1998, 1999-2002, 2003-2006, and 2007-2010 1999-2002 (black), 2003-2006 (brown), and 2007-2010 (green).Data are separated into high-(top) and moderate-(bottom) temperature regimes for Madera (a), Fresno (b), and Parlier (c).Uncertainties in the probability of violations are computed with counting statistics, are typically less than ±0.04 (1σ ) for weekdays and ±0.06 (1σ ) for weekends.Curves (dashed grey lines) were produced with an analytical model as for Fig. 6.
the NO 2 * concentration).In the last four years, at high temperatures exceedances became slightly less likely on weekends at all locations in the Central SJV suggesting O 3 conditions are transitioning to NO x -limited chemistry.Unlike in the Southern SJV, in the Central SJV there is evidence for a significant role played by VOCR reductions in decreasing the number of violations at high temperatures (Fig. 8).In Fresno, we infer VOCR decreases from 1999-2002 to 2003-2006  At moderate temperatures, the frequency of violations has decreased dramatically.In 2007-2010, the probability was less than 25 % at all three locations, with the largest changes in Parlier, where violations occurred at a frequency of more than 75 % on weekends a decade ago.We show this decrease in the exceedance probability is due to VOCR decreases, as exceedances are more likely on weekends (Fig. 7, bottom panel) and as probabilities consistently exhibit negative dayof-week slopes vs. NO 2 * (Fig. 8, bottom panel).Figure 8 suggests that the magnitude of the decrease in the likelihood of violations from 1999-2002 to 2003-2006 is approximately twice that at high temperatures.This is similar to the results for the Southern SJV (Fig. 6 1995-1998, 1999-2002, 2003-2006, and 2007-2010  at moderate temperatures, the controlled class is a larger fraction.These changes are explained if we assume that at high temperatures VOCR is a mixture of a controlled class and an uncontrolled class with both terms being important.

Northern San Joaquin Valley
From 2007-2010 in Stockton, the upwind location of the Northern SJV region, there is a less than 10 % probability that ozone concentrations will exceed the 8-h CAAQS at high temperatures on either weekdays or weekends (Fig. 9a).Downwind, probabilities are higher.At all three sites, there have been steep weekday decreases in the last sixteen years: in Stockton from 20 % to 5 %, in Turlock from 75 % to 35 %, and in Merced from 95 % to 55 % (Fig. 9b and c).In Stockton and Turlock, more frequent weekend exceedances (Fig. 9a  and b) and negative day-of-week slopes vs. NO 2 * (Fig. 10a  and b) show these locations are in a NO x -suppressed chemical regime.In contrast, in Merced at high temperatures, chemistry became NO x -limited in the last four years.Overall, in the Northern SJV the observed decreases in the frequency of high O 3 apparently are due to VOCR reductions.However, Fig. 10c (top panel) indicates that the frequency of high O 3 in Merced will fall with continued NO x reductions and Fig. 10b (top panel) shows that Turlock is near the threshold where NO x reductions become effective.
At moderate temperatures, exceedances from 2007-2010 were highly unlikely, occurring on fewer than 10 % of days at any of the three locations in the Northern SJV (Figs. 9 and 10).Violations were more frequent earlier in the record (e.g.Fig. 10c, bottom panel) and we infer the observed decreases are due to reductions in VOCR.

Evidence for local ozone production
There are two pieces of evidence that support local ozone production to be a large contributor to the frequency of high ozone days in the SJV.First, the observed exceedance proba-bility is lowest for each of the upwind sites, Shafter (Southern SJV), Madera (Central SJV), and Stockton (Northern SJV), increases along the plume transect (in Bakersfield, Fresno, and Turlock), and is highest at the corresponding downwind locations, Arvin, Parlier, and Merced, respectively.In the Southern SJV in 2007-2010 at high temperatures, we see an increase in the probability of a violation by 45 % on weekdays and by 40 % on weekends between Shafter and Arvin.In the Central SJV, over the same time period and in the same temperature regime, the percentage of violations is shown to increase by 20 % on weekends and 35 % on weekdays from Madera to downwind Parlier.In the Northern SJV in 2007-2010 at high temperatures, the probability increases by 35 % on weekdays and 45 % on weekends between Stockton and downwind Merced.The second piece of evidence is that there is a ∼10 % drop in the exceedance percentage between Parlier (downwind Central) and Shafter (upwind Southern) and a ∼20-35 % decrease between Merced (downwind Northern) and Madera (upwind Central).If local production were not important, we would expect to observe a single Valley-wide ozone plume and therefore to see the exceedance probability to smoothly rise (or fall) the length of the SJV.This is not the case however.Rather, the exceedance probably increases across each sub-region but then decreases again at the next site to the south (at the upwind sites Shafter and Madera).Exceedances are presently unlikely at moderate temperatures in the Central and Northern SJV but a comparison of past fouryear median exceedance probabilities also illustrates this effect.

Discussion
From 1995-2010, reductions in NO x emissions in California have been mostly due to more stringent standards on stationary sources and light-duty vehicles.In contrast, emissions from heavy-duty diesel engines, the largest source of NO x emissions in the SJV, have increased over the past fifteen 40 Uncertainties in the probability of violations are smaller than the observed year-to-year variability at ±0.04 (1σ ) for weekdays and ±0.06 (1σ ) for weekends.Curves (dashed grey lines) were produced as for Fig. 6.
years (Cox et al., 2009;Dallmann and Harley, 2010).Nationally new rules require heavy-duty diesel engines to meet more stringent NO x emissions standards (Environmental Protection Agency, 2000); however, these engines have long service lifetimes and slow fleet turnover rates.In California, in an effort to expedite benefits from new diesel engine regulations, the California Air Resources Board (CARB) is requiring all vehicle owners to retrofit or replace older diesel engines by 2023 and half of the in-use heavy duty-engines in large fleets must meet new NO x standards by 2014 (California Air Resources Board, 2007).Millstein and Harley (2010) show that in Los Angeles, as a result of this accelerated engine retrofit/replacement program, reductions in summertime diesel NO x emissions could be greater than 50 % over the five years from 2010 to 2015, with slower reductions (-20 % in tons day −1 ) predicted in the following ten years from 2015 to 2025.Additionally, the SJV Air Pollution Control District is also partnering with the Environmental Protection Agency (EPA) under the National Clean Diesel Campaign to replace diesel locomotives and diesel engines on agricultural equipment (Environmental Protection Agency, 2012a).NO x emissions reductions can still be expected from cars and light-duty trucks in the next twenty-five years.In 2012 CARB announced the Advanced Clean Cars Program, which aims to further reduce these NO x emissions by 75 % from 2014 levels through new emissions standards (in the 2015 model year) and by requiring one in seven news cars sold in California be zero-emission or plug-in hybrid vehicles by 2025(Environmental Protection Agency, 2012b).
In summary, policymakers at the local (San Joaquin Valley Unified Air Pollution Control District), state (CARB), and federal level (EPA Region 9) have expressed a commitment to reducing NO x emissions in the SJV and so we expect NO x concentrations to continue to decrease Valley wide.
The outlook for VOCR in the SJV is less clear.We show that at moderate temperatures, VOCR throughout the SJV has decreased over the last twelve years and that these decreases have resulted in fewer high O 3 days.This implies that the dominant sources of organic reactivity in this temperature regime are currently being controlled.VOC emissions from mobile sources have been thought to be largest source of O 3 forming organic precursors in the Valley (Hu et al., 2012).Regulatory efforts during our study window have focused on VOC emissions from light-duty vehicles and reduced these emissions through a combination of stricter standards and gasoline reformulation (Kirchstetter et al., 1999;Harley et al., 2006).At high temperatures in the Central and Northern SJV, we also show that reductions in VOCR have significantly decreased the frequency of violations.However, in the Central SJV these decreases in VOCR are smaller than those observed at moderate temperatures.This same temperature dependence is seen to a more dramatic extent in the Southern SJV, where over the last twelve years at high temperatures the VOCR in Bakersfield and Arvin has not changed.In this temperature regime, we therefore infer the existence of a VOCR source that both overwhelms the moderate-temperature source and that has gone unregulated over the last twelve years.
Recent model calculations have indicated non-mobile VOCR sources are important to P O 3 in the SJV, but to our knowledge this manuscript provides the first direct observational evidence.For example, Steiner et al. (2009) computed S. E. Pusede and R. C. Cohen: Ozone, NO x , and VOC reactivity in the San Joaquin Valley the total reactivity in the SJV, finding that the biogenic VOC emissions important in most other locations, such as isoprene and monterpenes (α-pinene), were only a small fraction of the total VOCR in this region.The authors suggested that the regional reactivity was dominated by oxygenates, although they noted that the sources of these species were very poorly quantified.VOC emissions from animal feeds have been proposed to be a large component of SJV VOCR (Alanis et al., 2008;Howard et al. 2010a;Howard et al. 2010b;Malkina et al., 2011).This source is not currently included in official inventories.In a first step toward understanding their impacts, inclusion of animal feed emissions in a regional air quality model (focusing on a single O 3 episode 24 July-2 August 2000) found that they were less important than mobile source VOC emissions to P O 3 , that P O 3 was still under-predicted, and that there is still likely missing VOCR (Hu et al., 2012).Clearly more research is needed to identify this VOCR, but whatever the source, our analysis suggests it has been unchanged over the last decade.
With this background on the expected changes in San Joaquin emissions, we present policy-relevant conclusions for the Southern, Central, and Northern SJV below and address the impacts of additional NO x and VOCR reductions on the frequencies of future CAAQS 8-h O 3 exceedances in the region.

Southern San Joaquin Valley
When temperatures are hottest, ozone production in Bakersfield and Arvin has been at peak for much of the last sixteen years and at constant VOCR.This explains why, despite a decade of NO x emission reductions, violations remain highly probable.At both sites, ozone production has recently transitioned to NO x -limited chemistry and, as a result, continued NO x controls are poised to improve O 3 air quality.Sizable NO x reductions are required before gains are seen in Arvin, as the exceedance probability at this site is still at peak on weekdays and very near unity.Current decreases in the hightemperature exceedance percentage in Arvin from 90 % on weekdays to 70 % on weekends suggest there will be 20 % fewer weekday violations in response to the next 50 % NO x reduction.Fifty percent NO x reductions will reduce the frequency of high ozone on weekdays in Bakersfield to 50 % and in Shafter to 30 %.At all three locations at moderate temperatures, ozone production is still at peak P O 3 or slightly NO x -suppressed (with a small slope) and so NO x reductions in this temperature regime will not immediately improve local O 3 air quality but will also not exacerbate it.
At the highest temperatures, observations suggest VOCR has not appreciably changed in the past decade.New strategies are therefore needed both to identify what organic molecules drive at the hottest temperatures and to reduce these precursor species.That said, because Southern SJV ozone production has transitioned to NO x -limited chemistry at high temperatures, additional VOCR reductions will provide diminished returns.At moderate temperatures, there is still the potential for VOCR reductions to decrease the frequency of violations.

Central San Joaquin Valley
At high temperatures, the exceedance probability has in the last four years transitioned to NO x -limited chemistry.It is difficult to be quantitative, but a comparison of the steepness of the 2007-2010 high-temperature slopes in Fig. 8 and Fig. 6 shows ozone chemistry in this region nearer to peak production than in the Southern SJV.As such, NO x controls will improve O 3 air quality but gains will lag those anticipated in the south.At moderate temperatures, NO x reductions will be slow to decrease the frequency of exceedances because chemistry is still NO x -suppressed.
VOCR reductions have been a powerful force in decreasing the exceedance probability under both high-and moderate-temperature conditions.Continued controls on mobile source emissions will further reduce the frequency of violations in both regimes but the impact of further controls is checked by the onset of NO x -limited ozone chemistry and by the fraction of VOCR that is due to uncontrolled sources.This fraction is important at high temperatures.

Northern San Joaquin Valley
In Stockton, NO 2 * abundances are high, the frequency of violations is NO x -suppressed, and high O 3 days are uncommon.As a result, NO x controls will not improve local O 3 air quality in this location.In Turlock under both highand moderate-temperature conditions, the exceedance probability remains NO x -suppressed.The payoff from continued NO x reductions will be delayed until a transition to NO xlimited chemistry takes place.The difference in the percentage of violations on weekdays and weekends is small and so chemistry is proximate to peak P O 3 .This gives confidence that NO x controls will not degrade Turlock O 3 air quality.In Merced, at high and moderate temperatures, P O 3 is NO x -limited as of 2007-2010.We anticipate continued NO x reductions will decrease the exceedance probability at this location and note that NO x reductions upwind in Stockton and Turlock are important to decreasing NO x abundances in Merced.
At high temperatures, continued reduction of VOC emissions is expected to decrease the frequency of high ozone days in Turlock.We predict that the impact of VOC emission reductions will be smaller than previously seen, as the decrease in O 3 exceedance probability in the last four years was only half that seen earlier in the decade.In Merced, in both temperature regimes, VOCR reductions have made profound improvements to O 3 air quality.At moderate temperatures, exceedances are below 15 %.At the high temperatures, VOCR reductions have resulted in exceedances being 50 % less probable than a decade ago.Now that ozone production is NO x -limited further VOCR reductions will be unable to drive substantial decreases in the number of violations.

Conclusions
We describe ozone's dependence on NO x and organic reactivity (VOCR) in San Joaquin Valley California using sixteen years of routine measurements of O 3 , NO 2 *, and temperature.
We show that local ozone production plays a large role in the frequency of high ozone days, as the exceedance percentage is seen to increase from upwind to downwind within each of our study regions and because the probability of a violation between regions is, in each case, higher at the downwind site to the north than at the receptor city to the south.This underscores the importance of controlling precursor emissions from local sources in the SJV.
We present location-specific policy-relevant conclusions for the Southern SJV, Central SJV, and Northern SJV in Sect.5.1, 5.2, and 5.3, respectively.Broadly speaking, we show that in the Central and Northern SJV, decreases in VOCR have dramatically reduced the frequency of violations.We report a temperature dependence in the effects of VOCR reductions in the Central SJV, finding they are larger at moderate-range temperatures than at high.This is likewise true in the Southern SJV, where reductions in the VOCR have decreased the frequency of exceedances at moderate temperatures but have made no impact when temperatures are hottest.That the VOCR has remained unchanged over the past twelve years at high temperatures in one region but not in the others reveals a need for detailed high-spatial resolution VOC emissions inventories in the SJV and a thorough analysis of the temperature dependence of each source.This evidence for two distinct types of VOCR sources frames an outstanding question for future research.What organic molecules drive the temperature dependence of VOCR both within each region and Valley wide?
We find that NO x reductions are poised to improve ozone air quality where violations are most frequent-the Southern and Central SJV.We see that these regions have or soon will transition to NO x -limited conditions when temperatures are highest and the likelihood of high ozone is greatest.We show that exceedances in the Southern SJV have remained highly probable despite NO x emissions control efforts because the ozone chemistry in Bakersfield and Arvin has been near peak P O 3 and at constant VOCR for more than a decade.
Ozone, NO 2 *, and temperature measurements have been collected across North America and around the world for more than a decade.We expect that the statistical approach described herein should be applicable to other isolated urban plumes.Even if wind directions are not as persistent as in the SJV, we imagine an analysis at the city center alone or one sorted by wind direction in addition to temperature will be interesting.We look forward to such analyses provid-ing broader observational perspective on the effectiveness of NO x and VOCR controls in other locations.

Measurements
CARB maintains an extensive network of ground-based monitors statewide.In this paper we use the 8-h maximum O 3 and hourly NO 2 data from thirteen CARB sites in the San Joaquin Valley Air Basin.Alphabetically, these sites are Arvin, ) (this site closed in November 2010); California Avenue, ); Clovis, Clovis-N Villa Avenue (36.819,Edison,Edison (35.346,Drummond,First Street,Madera,Merced,Parlier,Parlier (36.597,Shafter,Skypark,Stockton,and Turlock,."Bakersfield" is the median of the California Avenue and Edison stations and "Fresno" is the median of the Skypark, First Street, Drummond, and Clovis stations.Data at Madera-Pump Yard are available starting in 1998 and data from Clovis in 2008 were not reported.Data at Merced-S Coffee Avenue are not available in 2000 (NO 2 *) and 2006 (O 3 ).All data and detailed information about the location of each monitor are available for download on the CARB website: http://www.arb.ca.gov/adam/index.html.
We removed any concentration data exactly equal to 0.000 ppm believing this to be a physically unreasonable daytime concentration for either the 8-h maximum O 3 or the hourly NO 2 *.The daytime NO 2 * concentration is the daily mean value between 10 am and 2 pm local time.The average NO 2 * is not very sensitive to a change in this window and our work uses relative rather than absolute NO 2 * concentration.For Fresno and Bakersfield we use medians of the individual sites and in the absence of data at a single site for a given day that day is omitted.Yearly NO 2 * data are averaged for weekdays (Tuesdays-Fridays) and weekends (Saturdays-Sundays).
CARB NO 2 * is measured by chemiluminescence coupled with a heated molybdenum catalyst.NO 2 measurements with this technique are attended by a known positive interference from higher oxides of nitrogen, for example organic nitrates and nitric acid, which also thermally decompose (Williams et al., 1998;Dunlea et al., 2007).Ammonia (NH 3 ) has also been seen to positively interfere (0-10 %) with NO 2 chemiluminescence (Williams et al., 1998;Dunlea et al., 2007).NH 3 concentrations in the SJV are high (Clarisse et al., 2010) S. E. Pusede and R. C. Cohen: Ozone, NO x , and VOC reactivity in the San Joaquin Valley but we take confidence in the usefulness of the CARB NO 2 * data, as the NO 2 * abundances are decreasing across the Valley at rates similar to those observed from space by OMI (Russell et al., 2010).NO 2 * data are reported by CARB to be accurate to at least 15 %.
Temperature data are the 1-h maximum daily temperatures and data are used from three sites, Merced-S Coffee Avenue (37.282,Fresno Air Terminal (36.776,and Bakersfield Airport (35.325,; one site in each of our three study areas.The average maximum temperature is not statistically different from 1995 to 2010.We do not separate NO 2 * by temperature finding no significant temperature dependence in its concentration by day of week between high and moderate conditions.

Fig. 1 .Fig. 2 .
Fig. 1.Schematic of photochemical production of two new O 3 molecules from the oxidation of one generic organic molecule at the overlap of the HO x (a) and NO x (b) catalytic cycles.Only the NO x termination channels are shown.HO x chain terminations are reactions among peroxy radicals and OH.

Figure 3 .
Figure 3. Map of the California San Joaquin Valley (SJV) (a top) and details of each region for 3 this study: Southern SJV (b top), Central SJV (c top), and Northern SJV (d top).CARB 8-hour 4 maximum average O 3 and NO 2 * data are used from thirteen CARB sites: Shafter (upwind), 5 Bakersfield, and Arvin (downwind) (white circles), where Bakersfield is the median of the 6 California Avenue and Edison stations (grey circles); Madera (upwind), Fresno, and Parlier 7 (downwind) (white circles), where Fresno is the median of the Skypark, First Street, 8 Drummond, and Clovis stations (grey circles); Stockton (upwind), Turlock, and Merced 9 (downwind) (white circles).OMI NO 2 columns (molecules cm -2 ) are shown over the same 10 regions.These images are June-August weekday averages from 2007-2010 for the California 11 San Joaquin Valley (a bottom), Southern SJV (b bottom), Central SJV (c bottom), and 12

Fig. 3 .
Fig. 3. Map of the California San Joaquin Valley (SJV) (a top) and details of each region for this study: Southern SJV (b top), Central SJV (c top), and Northern SJV (d top).CARB 8-h maximum average O 3 and NO 2 * data are used from thirteen CARB sites: Shafter (upwind), Bakersfield, and Arvin (downwind) (white circles), where Bakersfield is the median of the California Avenue and Edison stations (grey circles); Madera (upwind), Fresno, and Parlier (downwind) (white circles), where Fresno is the median of the Skypark, First Street, Drummond, and Clovis stations (grey circles); Stockton (upwind), Turlock, and Merced (downwind) (white circles).OMI NO 2 columns (molecules cm −2 ) are shown over the same regions.These images are June-August weekday averages from 2007-2010 for the California San Joaquin Valley (a bottom), Southern SJV (b bottom), Central SJV (c bottom), and Northern SJV (d bottom).

2
Figure 4. Southern SJV, Shafter (a), Bakersfield (b), and Arvin (c), exceedance probabilities vs. 3 NO 2 * at high (34-45 o C) and moderate (28-33 o C) temperatures in red and blue, respectively.4 Data from weekdays (closed circles) and weekends (open diamonds) are shown as separate 5 symbols.NO 2 * are averages (10 am-2 pm local time) of hourly data at each site.Uncertainties 6are typically less than ±0.09 (1σ) for weekdays and less than ±0.12 (1σ) for weekends.NO 2 * 7 data are reported by CARB to be accurate to at least 15%.Black lines connect the median 8 percentage of violations at every 5 th NO 2 * data point.Over the past sixteen years, the average 9 annual number of days per year (rounded up) in the Southern SJV with a maximum temperature 10 in the high-temperature range is 66 and the average number in the moderate-temperature range is 11 72.12

Fig. 5 .
Fig. 5. Four-year median exceedance probabilities of the 8-h O 3 CAAQS vs. year (increasing right to left) in the Southern SJV: Shafter (a), Bakersfield (b), and Arvin (c).Data are shown for two temperature regimes: high (34-45 • C) and moderate (28-33 • C) in red and blue, respectively and divided into weekdays (closed circles) and weekends (open diamonds).The exceedance probabilities are shown for 1995-1998, 1999-2002, 2003-2006, and 2007-2010.Error bars are uncertainties in the four-year median exceedance probabilities, are calculated as counting errors, and are typically less than ±0.04 (1σ ) for weekdays and ±0.06 (1σ ) for weekends.The average number of days per year over the past sixteen years for both temperature regimes is 66 (high) and 72 (moderate).

1 2Figure 7 .Fig. 7 .
Figure 7. Central SJV four-year median exceedance probabilities of the 8-hour O 3 CAAQS vs. 3 year.Colors and symbols as in Figure 5.Over the past sixteen years there were on average 76 4 high-temperature days and 68 moderate-temperature days, 5
have amounted to 20 % fewer O 3 violations at the same NO 2 * (Fig. 8b, top panel).From 2003-2006 to 2007-2010, VOCR changes again contributed a 20 % decrease in O 3 exceedances.Similar trends are seen upwind in Madera and downwind in Parlier.

Figure 9 .Fig. 9 .
Figure 9. Northern SJV four-year median 8-hour O 3 CAAQS exceedance probabilities vs. year 3 (increasing right to left) are plotted separated into high-(34-45 o C) and moderate-(28-33 o C) 4 temperature regimes in red and blue, respectively and into weekdays (circles) and weekends 5 (diamonds) for Stockton (a), Turlock (b), and Merced (c).The average number of days with a 6 maximum temperature in each temperature range over the past sixteen years is 48 (high) and 79 7 (moderate).Error bars are the uncertainties calculated as counting errors and are typically ±0.04 8(1σ) for weekdays and ±0.06 (1σ) for weekends for four-year averages.9 Fig.9.Northern SJV four-year median 8-h O 3 CAAQS exceedance probabilities vs. year (increasing right to left) are plotted for1995-1998,  1999-2002, 2003-2006, and 2007-2010  separated into high-(34-45 • C) (red) and moderate-(28-33 • C) (blue) temperature regimes and into weekdays (circles) and weekends (diamonds) for Stockton (a), Turlock (b), and Merced (c).The average number of days with a maximum temperature in each temperature range over the past sixteen years is 48 (high) and 79 (moderate).Error bars are the uncertainties calculated as counting errors and are typically ±0.04 (1σ ) for weekdays and ±0.06 (1σ ) for weekends for four-year averages.