We conducted a first-order sensitivity analysis to investigate ozone responses to precursor emissions and source contributions (local versus upwind) for California's San Joaquin Valley (SJV) under four distinct meteorology conditions of summer 2000 using a three-dimensional photochemical transport model. Ozone-limiting reagents, nitrogen oxides (NOx), or anthropogenic volatile organic compounds (VOCs) (AVOCs) and their transition regime were determined from ozone sensitivity coefficients and delineated spatially at high-ozone locations in the SJV. In general, AVOC-limited areas were located near urban centers, while NOx-limited areas were located farther downwind. However, the spatial extent of AVOC-limited areas varied with meteorology. Meteorological dependence of predominant ozone-limiting precursors was found to vary significantly among different subregions within the SJV. Specifically, weaker dependences were identified for regions of the southern SJV located farther away from emission sources, where ozone chemistry was mostly limited by NOx for the episodes considered. Stronger dependences were identified for the central and northern SJV, where ozone chemistry can be limited by NOx or AVOC depending on meteorology. Source contributions to ozone sensitivities in the SJV were also investigated. Local sources were important for the eastern side of the central SJV, while upwind sources were also important (from ~40% to more than 50% of the total ozone sensitivities) for the western side of the valley, except for the most stagnant episode. Different contributing source regions were identified for the same VOC-limited areas in the northern SJV, and these depended on the flow characteristics. The predominant ozone-limiting reagent was found to exhibit less dependence on meteorology in the central and southern SJV as the baseline NOx emissions were reduced, ultimately causing ozone formation to be limited by NOx. In contrast, the VOC-limited areas in the northern SJV continued to be influenced by meteorology for two of the episodes.
 Tropospheric ozone is a secondary pollutant formed by photochemical reactions involving nitrogen oxides (NOx) and volatile organic compounds (VOCs). Due to its adverse health effect, ozone is designated as a criteria pollutant and regulated by the U.S. Environmental Protection Agency (EPA). Ambient levels of ozone depend not only on its precursor (NOx and VOCs) emissions but also on meteorological conditions that contribute to air quality through various physical and chemical processes influencing the evolution of emissions and photochemical products [Seaman, 2000]. For example, wind circulation alters pollutant transport and accumulation, boundary layer height limits the degree of pollutant dilution, and solar radiation and temperature change chemical reaction rates and biogenic emissions [Baertsch-Ritter et al., 2004; Dawson et al., 2007; Tao et al., 2007]. Important meteorological processes that lead to various types of spatial and/or temporal distributions of elevated ozone concentrations have been investigated in many regions with air pollution problems [e.g., Sillman and Samson, 1995; Dayan and Levy, 2002; Edwards et al., 2004; Kleeman, 2008; Bloomer et al., 2009; Katragkou et al., 2010; Jin et al., 2011]. Not only are ozone concentrations sensitive to meteorological changes but also ozone responses to emission reductions of its precursor species vary with meteorology (as discussed in Lei et al. ). Such information is key to designing effective ozone control strategies for regions with diverse meteorological conditions but is a topic covered by fewer studies.
 Ozone production regimes are defined by ozone response to changes in emissions of NOx and VOCs, which we refer to as the characteristics of the ozone limitation chemistry. Theoretically, meteorological parameters (mixing height, temperature, humidity, wind speed, etc.) can be perturbed one at a time to simulate their individual effects on the characteristics of ozone limitation chemistry [e.g., Baertsch-Ritter et al., 2004]. Studies like this greatly enriched our understanding of the dependence of ozone-precursor relationships on meteorology. Practically, there are limitations to considering only the individual effects of meteorological parameters. First, individual meteorological variables associated with different types of ozone episodes generally do not vary independently but are related to one another at the regional level to form “meteorological regimes” governed by synoptic and mesoscale phenomena [Beaver and Palazoglu, 2009; Jin et al., 2011]. Moreover, compensating effects may exist on ozone limitation chemistry for different meteorological variables. For example, stagnant conditions, conducive to high ozone levels, are often associated with lower wind speed and higher temperature, while Raertsch-Ritter et al.  found that NOx-limited areas decrease with higher temperature and increase with lower wind speed. Finally, wind directions and flow patterns are important to regional pollutant transport, which affects the source-receptor relationships and ozone chemical regimes along the transport paths [Fast et al., 2002]. However, it is difficult to perturb these variables systematically in air quality models for quantifying their effects on ozone sensitivities.
 Investigating the combined meteorological effects on ozone responses to emission controls in a specified region is important. From a regulatory perspective, the key issue is to understand the extent to which the meteorological conditions, commonly occurring (present or future) in the study region, alter the characteristics of the ozone limitation chemistry. Palacios et al.  showed that ozone responses were influenced by different transport and dispersion patterns established in the Greater Madrid area. Lei et al.  characterized ozone production and responses under three distinctive meteorological categories in Mexico City and found weak dependence of the VOC limitation of ozone production in the urban areas, while Song et al.  found that the range of VOC- or NOx-limited areas depended on meteorology in this area in a more recent campaign year. Liao et al.  simulated ozone sensitivities to emissions over the continental U.S. under meteorological conditions of present and future climates and also found a weak dependence of ozone control options (NOx versus VOC) on meteorological conditions.
 California's San Joaquin Valley (SJV) has suffered from some of the worst air in the country and is designated as an “extreme” nonattainment area for the federal 8 h ozone standard. While ozone levels in much of California have fallen steadily over the years, progress in the SJV has been slower with ambient ozone exceeding the 8 h standard more than 100 days a year [Hall et al., 2008]. EPA [Federal Register, 2008] has since strengthened the standard to 75 ppb from the previous value of 84 ppb for 8 h ozone, which makes it even more challenging to bring SJV into compliance.
 The trough-like topography of the SJV favors pollutant accumulation. Diverse emission sources from both local and occasionally upwind regions, namely, the San Francisco Bay (SFB) area, the Sacramento Valley (SV), and some coastal air basins, further complicate source and receptor relationships and make the SJV ozone control a regional problem [Pun et al., 2000]. A first-order sensitivity analysis conducted previously over a 5 day ozone episode [Jin et al., 2008] delineated the spatial variations in the ozone control options (NOx versus VOC) in the SJV and found that (1) NOx control is, overall, more effective for attaining the 8 h ozone standard in the region and (2) ozone sensitivities, especially in the northern part of the valley, were influenced by emissions from upwind air basins, while local contributions dominated most of the SJV. Modeling studies have since been expanded to a full summer season [Jin et al., 2010, 2011] with additional seasonal perspectives gained by identifying statistically determined meteorological regimes that give rise to different spatial distributions of high ozone levels. Whether or not the ozone control options and interbasin contributions derived previously [i.e., Jin et al., 2008] from one ozone episode can be applied to other meteorological regimes has not been investigated. Furthermore, occurrence frequencies of air quality related meteorological conditions in this region are expected to change in future climate [Zhao et al., 2011]. Consequently, understanding the meteorological dependence of ozone sensitivities to emission changes and the local versus upwind contribution is needed to achieve a more comprehensive and effective regional ozone control strategy designed for the SJV for both present and future climates.
 In this study, employing the Community Multiscale Air Quality (CMAQ) modeling system [Byun and Schere, 2006], we extend the first-order ozone sensitivity analysis to include all four ozone episodes representative of the meteorological conditions found in both observational and modeling studies in the central California region (in Fujita et al.  and Jin et al. , respectively). The objective of this study is to understand the extent to which meteorological regimes alter the characteristics of the ozone limitation chemistry and the importance of local versus upwind emission contributions to different parts of the SJV. For any given meteorological conditions, simulated ozone production regimes are affected by underlying emission inputs that are subject to uncertainties and temporal changes (e.g., through weekly cycle and/or future emission controls). In response to this, first-order ozone sensitivities are also evaluated for alternative baseline emissions to understand how the meteorological dependence of ozone sensitivities interacts with different emission scenarios. Section 2 introduces the study domain and base case simulations conducted in Jin et al.  as well as the model inputs. The selection and characteristics of the four meteorologically representative ozone pollution episodes are described, followed by presentations of sensitivity analysis methods, definitions of ozone control options, and alternative baseline emission cases. Section 3 presents the spatial distributions of first-order ozone sensitivities, ozone control maps, local versus upwind contributions, and changes in the characteristics of ozone-limiting chemistry under alternative emission cases. A summary of conclusions is provided in section 4.
2 Data and Methods
2.1 Study Domain and Data
 CMAQ v4.5, with the same configuration used in Jin et al. [2010, 2011], is applied to simulate hourly ozone mixing ratios over the Central California Ozone Study domain (shown in Figure 1). The domain is gridded into 185 by 185 cells, with a horizontal grid spacing of 4 km. Vertically, the domain is divided into 27 layers from the surface to about 17 km. The Central Valley is surrounded by the Sierra Nevada and coastal mountains ranges. The San Francisco Bay (SFB) area and Sacramento Valley (SV) are the major upwind emission sources affecting air quality in the San Joaquin Valley (SJV). The SJV is further divided into northern (NSJV), central (MSJV), and southern (SSJV) parts as shown on the map. Hourly gridded emission inputs were obtained from the California Air Resources Board (CARB) with day-of-week differences in anthropogenic sources and day-specific biogenic emissions. Hourly meteorological fields were simulated with the PSU/NCAR mesoscale model (MM5). More detailed descriptions of the input data, model configuration, and performance evaluation are reported in Jin et al.  for the summer of year 2000 (June to September).
 To isolate the meteorological effects, emission inputs in this study are modified as were done in Jin et al. : Anthropogenic emissions are not varied day to day and instead use a profile identical to typical Monday emissions, light- and/or temperature-sensitive biogenic emissions (mainly isoprene and terpenes) are allowed to vary with meteorology, and fire emissions are excluded. Summerlong CMAQ simulations with this emission setup have been conducted by Jin et al. , and their model outputs are used as the study base case.
 Summer averaged 8 h ozone and surface wind fields are shown in Figure 1 for the [11 A.M., 6 P.M.] hour interval, when most daily 8 h ozone maxima occur (Figure S1 in the supporting information). Our analysis focuses on the SJV basin, where high ozone levels are concentrated and both CMAQ and MM5 perform better [Jin et al., 2010] than other air basins. The average wind fields (Figure 1) reveal that the prevailing daytime summer flow is driven largely by mesoscale phenomena. Thermally induced pressure gradients between land and sea force the marine air to move through the SFB into the Central Valley, where orographic effects split the flow with the majority directed into the SJV. The down-valley flow exits at the southern end into the Mohave Desert. Within the SJV, thermal contrast between the Sierra Mountains and valley surface creates pressure gradient to drive upslope flow during the day. These flow patterns directly influence the transport of pollutants when they are photochemically active.
2.2 Ozone Episodes
Jin et al.  (hereafter referred to as “the cluster paper”) applied cluster analysis to model simulation data for California's San Joaquin Valley (SJV) for the purpose of identifying meteorologically representative ozone pollution conditions for the summer of year 2000. Summer days were categorized according to their spatial distributions of ozone concentrations in the SJV, which were largely driven by meteorology. Ozone clusters in the SFB and SV were also identified for investigating the interbasin relationships between the SJV and its major upwind air basins (see Figure S2 for daily cluster memberships). Four SJV ozone clusters associated with moderate to high ozone levels were determined: three of them with elevated ozone concentrations relatively to the north, south, and west of the SJV and the fourth one with higher ozone levels throughout the valley. For convenience, these four conditions are referred to as “O3-North,” “O3-South,” “O3-West,” and “O3-All.” Despite the variability of ozone episodes within each cluster, the four episodes to be chosen in this study are intended to provide different “modes” of the joint distribution of individual meteorological variables that are representative of various ozone-forming conditions in the San Joaquin Valley. As shown in Jin et al. , the O3-North, O3-South, and O3-West conditions are statistically distinctive from each other by the flow-induced spatial variations in ozone concentrations, while O3-All has statistically higher ozone associated with higher temperatures throughout the domain. Four ozone episodes representative of these conditions are identified: 23 to 25 June (O3-North), 14 to 17 August (O3-South), 17 to 20 September (O3-West), and 29 July to 2 August (O3-All), according to their cluster memberships, relationships to the SFB, and SJV ozone levels, as well as their occurrences in historical ozone observations, which will be described next.
 Ozone anomalies (i.e., episode average − seasonal average) and prominent circulation patterns are illustrated in Figure 2 for the four episodes. Synoptic conditions in these episodes give rise to the deviations of temperature, pressure, and wind fields from their seasonal averages at the regional scale that result in different spatial patterns of elevated ozone anomaly levels.
 The O3-North episode occurs under a pressure trough associated with relatively ventilated conditions. Higher temperatures are observed in the SV and the northern part of the domain (Figure S4). The sea surface pressure anomalies in this episode (Figure S5) indicate a pressure gradient (from south to north) that weakens northerly winds and strengthens southerly flow in the SV and creates a stronger westerly flow through the SFB. As a result, pollutant concentrations increase in the northern part of the SJV and in the SV basin.
 The O3-South episode occurs under a “western U.S. high” anticyclone system located inland with higher temperatures in the eastern and southern parts of the domain (Figure S4). The sea surface pressure anomalies under this synoptic condition give rise to a strong northern westerly flow (Figure S5), enhancing the down-valley and upslope ventilation flows in the SJV. As a result, pollutant concentrations increase in the southeastern part of the valley and the SV and SFB are relatively clean.
 The O3-West episode is under the influence of an “eastern Pacific high” system which heats up the ocean and coastal areas where higher temperature levels are observed (Figure S4). An enhanced surface pressure gradient from land to sea (Figure S5) leads to very stagnant conditions. With weakened onshore flows (from west to east), pollutant concentrations tend to increase on the western side of the SJV and in the SFB coastal regions.
 The above three episodes with their ozone concentration levels in the three air basins (SJV, SFB, and SV) have temperature and flow conditions that are also found to be representative of those observed frequently in historical years (see Fujita et al. , with a summary in the supporting information).
 The O3-All episode features a persistent western U.S. high system with light winds and much higher temperatures over the land than other episodes. The flow pattern resembles the summer mean because the wind anomalies (episode average − seasonal average) are generally small (Figure S6) with slightly enhanced southward flow in the SV. Elevated ozone levels are observed in much of the SJV, SFB, and SV metropolitan areas.
 Surface temperature, pressure, and wind fields as well as the 500 mbar geopotential heights representing synoptic conditions are averaged for the four episodes and provided in Figures S3–S7.
 Sensitivity simulations were previously conducted for the O3-All episode for a smaller domain [Jin et al., 2008] to investigate ozone formation and transport. The study domain has since been expanded to include more areas upwind of the SJV. In this study, we extend our sensitivity analysis to include all four episodes. The four episodes are expanded to 5 days (Table 1) by including additional days at the beginning and using the first 2 days for model spin-ups. The meteorological parameters (temperature, wind, mixing layer height, humidity, etc.) and chemical simulation results presented in this study are 8 h averages over [11 A.M., 6 P.M.] (referred to as “afternoon” for simplicity) for the last 3 days of each simulated episode.
Table 1. Definition of Modeling Periods
June episode: 23 to 25 June (days 175–177)
21 to 25 June
August episode: 14 to 17 August (days 227–230)
13 to 17 August
September episode: 17 to 20 September (days 261–264)
16 to 20 September
July–August episode: 29 July to 2 August (days 211–215)
29 July to 2 August
2.3 First-Order Ozone Sensitivity Coefficient
 The first-order seminormalized ozone sensitivity (defined below) to emitted NOx or anthropogenic volatile organic compounds (AVOCs) from the whole domain or selected subregions (SFB, SJV, and SV) is approximated by a brute force perturbation of the respective emissions by +10% relative to their nominal values for the four ozone episodes. So here, ∂pi = 10% × Pi and ε = 10% in equation ((1)):
where Pi is the emission parameter (i.e., emissions from the whole domain or selected air basins), whose perturbation pi is considered in a relative sense by defining a scaling variable εi, with its nominal value (unperturbed) being 1; C is the species concentration with the base case emissions; and C+10% is the concentration simulated by perturbing the emissions by +10%. These first-order seminormalized sensitivity coefficients represent ozone concentration changes per unit change in emission parameter if the system is linear. It can be used to calculate pollutant responses to reductions in emissions by up to 25% in general [Cohan et al., 2005] when the linearity holds. For example, (ppb) implies that a ±25% change in the emissions would cause (±0.25α) ppb change in the ozone concentration while all other parameters are held constant.
 Ozone sensitivities to domain-wide anthropogenic NOx emissions and to anthropogenic VOC (AVOC) emissions are used to determine ozone-limiting reagents, while ozone sensitivities to emissions from individual air basins are used to evaluate local versus upwind contributions to the SJV ozone sensitivities.
2.4 Ozone Control Options
 Three ozone control options are defined, based on the relationship between the two sensitivity coefficients and following Jin et al. :
 The “VOC control” option is preferred when reducing AVOC reduces ozone concentration and reducing NOx emissions would increase ozone concentrations (NOx disbenefit). The “NOx control” option is preferred when a percentage reduction in NOx emissions results in larger decreases in ozone concentrations than the same percentage reduction in AVOC emissions. Between these two options, the third option is in a transition regime, where reducing NOx emissions can reduce ozone concentrations but is not as effective as reducing AVOC by the same percentage.
2.5 Alternative Base Case Emissions
 The relationship between meteorological conditions and ozone sensitivities to precursor emissions can interact with the underlying nominal emission levels (base case) that are subject to uncertainties and temporal changes. CMAQ predictions of ambient AVOC and total reactive nitrogen for the whole summer period were found to have domain-wide normalized biases of 23% and 3%, respectively, when compared to observations. This most likely reflects the uncertainties in both modeled results and observations, but it does indicate uncertainties in emission inputs [Jin et al., 2010]. This study employs identical anthropogenic daily emissions (Monday) to isolate meteorological effects. Weekend human activities are different from those on weekdays [e.g., Tonse et al., 2008]. About 16% and 25% lower NOx emissions, respectively, from area and mobile sources are inventoried on weekends than on weekdays for the modeling period [Jin et al., 2010]. Emissions have also experienced longer-term changes since our modeling year of 2000. A recent study has shown a downward trend of NOx emissions in recent years from the SJV mainly due to the economic downturn, with 20% lower NOx emitted in 2010 than in 2000 [McDonald et al., 2012].
 To understand whether the meteorological dependence of ozone control options may hold under different emission scenarios, ozone sensitivities are also calculated after altering the base case emissions by 20% (as defined below) in the four ozone episodes. Ozone control options are evaluated based on the sensitivities calculated under these alternative emission cases.
 The new alternative base cases (nominal values) are defined as follows:
 Case “mAVOC”: AVOC emissions are reduced by 20%.
 Case “mNOx”: NOx emissions are reduced by 20%.
 Case “mBoth”: Both AVOC and NOx emissions are reduced by 20%.
3 Results and Discussions
3.1 First-Order Ozone Sensitivities and Control Options
 To evaluate ozone control options for the SJV, first-order seminormalized ozone sensitivities to the domain-wide emissions of AVOC and NOx are calculated. Ozone sensitivities to AVOC (Figure 3, second column) are positive throughout the domain, which indicates that reducing AVOC leads to lowered ozone levels. Greater ozone sensitivities to AVOC tend to be located in the areas where NOx emissions are also large. Ozone sensitivities to NOx change signs (Figure 3, first column). The negative sensitivities indicate a NOx disbenefit. In general, the urban areas, namely, the Bay area, Sacramento, and urban areas in the SJV, such as Fresno and Bakersfield, all exhibit a NOx disbenefit for the various meteorological conditions associated with the different episodes. The spatial extent of the NOx disbenefit areas changes from episode to episode, especially for the northern part of the SJV.
 Ozone control options maps (Figure 3, third column) are evaluated for the cells where the afternoon 8 h average ozone concentrations are over 75 ppb (referred to as “high ozone” hereafter). This ozone level is the current national ambient air quality standard for ground-level ozone and is referred to as the 2008 standard. The white areas are those where the 8 h average ozone concentrations do not exceed the 2008 standard, and the colored areas indicate the extent of ozone exceedances. All episodes except the O3-North episode exhibit extensive ozone exceedances especially for the O3-All condition. The spatial trend in ozone control options is similar for the episodes: VOC control near urban and emission centers and NOx control in rural areas. However, the extent of the NOx disbenefit areas (i.e., VOC controlled, colored in red) varies across the meteorological conditions associated with the various episodes. Focusing on a single ozone episode (July–August, O3-All), in a previous study, Jin et al.  found that NOx control was, overall, more beneficial for reducing 8 h ozone in the SJV. By considering different meteorological conditions, we can see that NOx control is the dominant option for the O3-South and O3-All episodes, when it accounts for 64% and 63%, respectively, of the high-ozone locations. Increased importance in AVOC control (colored in red) is identified in the O3-North and O3-West conditions, when NOx-controlled high-ozone locations are reduced to 45% and 47%, respectively, in the SJV.
 Meteorological dependence of ozone control via NOx or AVOC limitation varies greatly by location (Figure S8). The dominant control options for the southern SJV show weak dependence on the meteorological conditions considered here, with NOx control benefiting the major portion (58% to 78%) of the area throughout all episodes while improvements from VOC control are limited to the urban centers. The control options in the northern and central SJV depend on meteorology. Ozone chemistry (of locations with high ozone) in the northern SJV is generally VOC limited (54~83% VOC controlled in three episodes), and the situation changes under the O3-All condition (20% areas are VOC controlled) with an increased percentage of NOx-controlled grids as high ozone levels are extended to more rural areas. High ozone levels in the central SJV (MSJV) are generally limited by NOx (56~70% areas are NOx controlled in three episodes) except for the emission centers in the urban areas and along the highways. However, VOC control extends farther into the rural areas in the MSJV under the O3-West conditions, while the percentage of NOx-controlled grids is reduced (45% areas are NOx controlled). In this case, NOx control that works well for other episodes would result in an increase in ozone concentrations in some of the rural locations.
 Since anthropogenic emissions are held constant across the episodes, the aforementioned differences in ozone sensitivities and control options among episodes result from changes in meteorology. Effects of isolated meteorological variables on ozone sensitivities and NOx versus VOC limitation were investigated in past studies with findings summarized in Table 2.
Table 2. Effects of Meteorology-Related Parameters on Ozone Sensitivities to NOx and AVOC Identified in the Past Literature
Increasing Meteorological Related Variables
Effects on Ozone Sensitivities to NOx Versus AVOC
Increases ozone sensitivity to AVOC and decreases sensitivity to NOx [Jin et al., 2010]; increases VOC-limited areas [Baertsch-Ritter et al., 2004].
Increases NOx-limited areas and decreases VOC-limited areas [Baertsch-Ritter et al., 2004].
Temperature and humidity
Increases NOx-limited areas and decreases VOC-limited areas [Baertsch-Ritter et al., 2004].
Temperature and biogenic emissions
Increases NOx limitation (Jin et al.  and Baertsch-Ritter et al.  both found small effects).
Depends on treatment of point sources in the model [Baertsch-Ritter et al., 2004].
Increases ozone sensitivity to NOx and shifts chemistry to NOx limitation [Jin et al., 2010]; increases NOx-limited areas and decreases VOC-limited areas [Baertsch-Ritter et al., 2004].
Ventilation rate (= wind speed × mixing height)
Shifts ozone chemistry to NOx limitation [Sillman, 1995; Biswas and Rao, 2001].
 Higher rates of peroxyacetylnitrate (PAN) decomposition were found to be the most important reason for the enhanced ozone production under higher temperatures [Baertsch-Ritter et al., 2004]. Increased availability of NO2 associated with decreased PAN formation shifts ozone chemistry toward VOC limitation with higher temperatures.
 Quantities such as absolute humidity and biogenic emissions that vary with temperature can also alter the ozone-limiting chemistry. Higher absolute humidity associated with higher temperature enhances the formation rate of OH radicals, which leads to faster removal of NO2 by termination reactions as well as increased formation of peroxy radicals through VOC + OH oxidation. As a result, ozone chemistry is shifted toward NOx limitation under higher absolute humidity, which counteracts the temperature effects. Increased biogenic emissions that occur with higher temperature generally enhance NOx limitation. Biogenic emissions are similar for the O3-North, O3-South, and O3-West episodes and are higher (~25% higher in isoprene and ~8% higher in terpene) in the O3-All episode due to its higher temperature. However, for the O3-All conditions, biogenic emissions were found to exert rather small effects on ozone sensitivities to NOx as they are largely not collocated with anthropogenic sources [Jin et al., 2008]. In addition, the study conducted by Steiner et al.  under the O3-All conditions found that the biogenic VOC emissions (isoprene and terpenes) were only a small fraction of the total VOC reactivities in the SJV region.
 Wind speed, mixing height, and ventilation rates affect the rate at which pollutants are diluted. Under more ventilated conditions, NOx concentrations are reduced and ozone chemistry is shifted toward NOx limitation, while under very stagnant conditions, concentrated NOx titrates ozone over larger areas, leading to an increase in the VOC-limited locations.
 As discussed previously, individual meteorological variables associated with different types of ozone episodes generally do not vary independently but are related to one another at the regional level. The ozone episodes studied here provide different modes of the joint distribution of individual meteorological variables that are representative of the ozone-forming conditions in the SJV. Table 3 summarizes the episodic averaged meteorological anomalies in different parts of the SJV as well as their summer averages including surface temperature and humidity, mixing height, and ventilation rate (mixing height multiplied by surface wind speed).
Table 3. Summary of Afternoon Averaged Meteorological Variables for Different Parts of the SJVa
Surface Temperature (K)
Surface Humidity (g/kg)
Mixing Height (m)
Ventilation Rate (m2/s)
aEpisodic anomalies (episodic average − summer average) for four episodes as well as the summer averages are presented.
 All the episodes are warmer than the seasonal averages. The O3-All episode exhibits the highest temperature and absolute humidity levels, while O3-South exhibits the lowest. The atmospheric condition is the most stagnant in the O3-West episode and the most ventilated in the O3-North episode as indicated by mixing heights and ventilation rates.
 For the central and southern SJV, the differences in ozone control options observed in Figure 3 can largely be explained by the joint effects of these scalar meteorology quantities. For example, the VOC-limited area expands significantly under the most stagnant O3-West condition, while it is more confined to urban centers under the most ventilated O3-North condition. Higher humidity in the O3-All episode produces much higher OH levels than those in the O3-South episode (Figure S9) that shifts the overall ozone chemistry slightly in the central SJV toward NOx limitation (with slightly reduced VOC-controlled areas) than the O3-South condition, despite the O3-All higher temperatures.
 For the northern SJV, meteorological conditions described by the scalar quantities in Table 3 might suggest to an opposite ozone control option from that indicated in Figure 3. VOC limitation dominates the O3-North episode while the scalar quantities presented here, namely, lower temperature, moderate humidity, and more ventilation, all provide favorable conditions for NOx limitation. Besides these scalar quantities, wind directions or flow patterns provide additional constraints on the ozone-limiting chemistry in this region.
3.2 Local Versus Upwind Contributions
 The transport patterns (wind flows) affect the source-receptor relationships and ozone chemical regimes along the transport paths, which differ from episode to episode. The interbasin transport effect is especially important for the SJV ozone response to NOx emissions due to the nonlinearity characteristics of the chemistry (i.e., changes signs). NOx emissions from large upwind sources (e.g., in an urban plume) first titrate ozone, resulting in negative ozone sensitivities to NOx. As NOx emissions are transported farther downwind, they contribute to ozone formation by increasing the supply of odd oxygen (NO2 + O3), resulting in positive ozone sensitivities to NOx [Jin et al., 2008]. To illustrate the local and upwind contributions to the SJV ozone responses to NOx emissions, the total sensitivity is decomposed into contributions from the local and upwind air basins and . The episodic averages are plotted (Figure 4) at grid cells along two transects (illustrated in Figure 3) to illustrate their spatial evolution along the eastern and western sides of the SJV, where high ozone levels occur. Both transects start from the same location at the border between the SFB and SJV in the northern SJV and move downwind toward the southern SJV. The western side of the SJV is largely limited by NOx emissions except the northern part, which is closer to the emissions from the SFB and SV. The eastern transect runs along Interstate 5 and urban centers where NOx disbenefits occur . The sum of sensitivities from the three air basins (SJV, SFB, and SV) generally does not equal the total sensitivity (black dotted line) because emissions from the other air basins also play a role.
 The SFB NOx emissions generally reduce ozone levels through titration in the northern SJV and contribute positively to ozone levels farther downwind, while the SV NOx emissions generally exhibit positive contributions throughout the SJV because the transects start farther away from it. and provide direct means to quantify the importance of the effects of transported pollutant on the SJV ozone sensitivities with their magnitude and spatial extent depending on the flow characteristics of the episodes.
 Under O3-North conditions, enhanced eastward transport leads the SFB emissions to penetrate farther into the northern SJV along the eastern transect. As a result, SFB emissions contribute to extensive NOx disbenefit areas as we have seen in Figure 3, a phenomenon that was not explained by the scalar meteorological parameters considered previously. The weakened southward flow in this episode reduces the impact of SFB emissions farther downwind, and the SV emissions show minimum effects on the SJV ozone. Large gaps along the western transect between the total sensitivity (black dotted line) and the sum of subregional sensitivities considered here indicate important contributions from other coastal emissions in this episode due to the strengthened westerly flows.
 The enhanced southward transport under O3-South conditions causes a wider spread of NOx disbenefit areas of SFB emissions in the northern SJV along the western transect than other episodes and increases the positive contributions of both the SFB and SV farther south which account for more than 50% of the total sensitivities. The NOx disbenefit areas along the eastern transect are also slightly enhanced by the SFB emissions in the northern SJV while positive contributions of both SFB and SV spread farther south.
 Stagnant conditions during the O3-West episode feature the weakest eastward transport as well as slower southward transport within the SJV, which greatly reduces the influence of SFB emissions on the SJV ozone. Local (SJV) emissions dominate ozone sensitivities along the eastern transect with an extensive spread of the NOx disbenefit areas. Increased contributions of the SV emissions are only observed along the western transect but are limited to the northern half of the SJV.
 The regional flow pattern in the O3-All episode resembles the O3-South episode but with a weaker southward flow than the latter in the SJV. As a result, the NOx disbenefit areas caused by SFB emissions are reduced in the northern SJV and the sensitivities are smaller in this episode than in O3-South.
 In general, along the eastern transect, only the central SJV exhibits weak dependence on meteorology, where local contributions always dominate and determine the signs of the total sensitivities, while the northern and southern SJV both depend on the meteorology where interbasin contributions play important roles except for the O3-West condition. Along the western transect, upwind sources exert greater influences on the total sensitivities especially in the northern SJV. Decreases in NOx emissions of both local and upwind sources can benefit the ozone control in the NOx-limited regions on the western side of the valley. Important upwind source regions also change with meteorology. The SFB and SV are largely the major upwind contributors. However, under O3-North conditions, the SV contribution diminishes and the increased SFB contribution is limited to the northern SJV, while coastal contributions increase greatly along the western transect throughout the SJV.
 Control of AVOC emissions is important for the northern SJV as well as for the urban areas in the central and southern SJV. VOCs have a longer lifetime than NOx [Finlayson-Pitts and Pitts, 1997] in the air, so they are more likely to be transported farther downwind and can capture better how wind fields influence the transport of pollutants. As ozone sensitivities to AVOC are always positive (Figure 3), the importance of local emissions versus interbasin transport can be conveniently presented with the metric “contribution of SJV AVOC emissions” defined below:
where SSJV,AVOC is the sensitivity to SJV AVOC emissions and SAVOC is the ozone sensitivity to domain-wide AVOC emissions. A “contribution map” is shown in Figure 5: Contributions greater than 50% indicate a major local influence, while contributions less than 50% indicate a greater contribution from upwind air basins, i.e., the sum of all upwind air basins except the SJV. In this manner, we can develop the notion of a “contribution pattern,” i.e., how local and upwind contributions spatially evolve within an episode, and evaluate the variability of this pattern with meteorology.
 The contribution maps generally present a contrast between west and east with the upwind influences dominating the western side of the SJV while local influences dominate the eastern side where VOC limitation usually occurs. Distinctive contrast is seen for the O3-North and O3-West conditions. In the O3-North episode, the local contributions (%) are the smallest throughout the SJV among all the episodes, and the upwind contributions dominate the entire northern SJV including those VOC-limited locations where local sources exert greater influences in other episodes. Effective ozone control here would require reduction of both local and upwind AVOC emissions. In the O3-West episode, however, the local contributions increase greatly and expand farther west, covering part of the northern SJV and the entire southern SJV. The ozone control in the VOC-limited areas here can be well achieved by reducing local AVOC emissions. Increased importance of VOC controls was previously identified for these two episodes (Figure 3), and the results presented here have highlighted the differences in their targeted source regions governed by the distinctive flow characteristics.
 The flow characteristics that Zhao et al.  simulated under future climate showed overall decreased ventilation rates in the SJV and identified a wind anomaly from the southwest (i.e., southwesterly winds) that enhances the sea breezes as the result of higher temperature rise over the land than ocean. The enhanced southwesterlies can reduce the contributions from the SV to the SJV while increasing the contributions from the SFB and coastal areas to the northern and western parts of the SJV, like the behavior seen under the O3-North flow pattern. The local contributions within the eastern side of the valley would be strengthened with the increasing overall stagnation.
3.3 Alternative Base Case Emissions
 The spatial distribution of both high ozone levels and their sensitivities to emission reductions under a given set of meteorology is also a function of the underlying emissions, which are subject to uncertainties and changes. First-order sensitivities to domain-wide emissions are calculated under alternative baseline emissions, where emitted AVOC, NOx, or both are reduced by 20% (mAVOC, mNOx, and mBoth, respectively). The derived control options at high-ozone locations under these alternative emission cases are compared to the ones under the original base case (hereafter called “OrigBase”) to investigate the influences of baseline emissions on the meteorological dependence of choices of ozone control in the SJV.
 For our studied periods and domain, reducing AVOC emissions (i.e., mAVOC) has a smaller impact on the ozone sensitivities than reducing NOx or reducing both (i.e., mNOx or mBoth), which can be seen in the difference maps shown in Figure S10. In particular, increases greatly as baseline NOx is reduced in the mNOx and mBoth cases. Throughout all the alternative baseline emission cases, the percentage of VOC-controlled areas in the SJV under the O3-North and O3-West conditions is greater than that under the O3-South and O3-All conditions, while the opposite is true for the NOx control options (Figure S11), which is similar to the behavior observed in the OrigBase case. In an absolute sense, a reduction in AVOC emissions exerts small effects on control options as we can see that mAVOC is similar to OrigBase and mNOx is similar to mBoth (Figures S10 and S11). On the other hand, reductions of NOx (in mNOx or mBoth) simultaneously increase NOx-limited areas and decrease VOC-limited areas, which leads to the majority of the high-ozone locations (55%~80%) in the SJV requiring NOx control for all the meteorological conditions (Figure S11).
 Spatially (Figure S12), the resulting changes in control options are limited to regions close to large emission sources (Fresno, Bakersfield, or the northern SJV). Similar distributions of control options are seen between the mAVOC and OrigBase cases. On the other hand, NOx control and transition areas expand as VOC-controlled areas reduce greatly under the mNOx or mBoth cases. This is especially true in the central and southern SJV where VOC limitation is now confined to the urban locations across all episodes, accounting for no more than 10% of the high-ozone locations (Table 4). Consequently, the control options in the central and southern SJV both exhibit weak dependence on meteorology after the baseline NOx emissions are reduced. These modeling results suggest that the chemistry has moved toward NOx limitation and NOx control in the central and southern SJV has become more effective in recent years as the result of reduced NOx emissions from the level of year 2000. Such behavior is also observed in the ambient measurements [Pusede and Cohen, 2012].
Table 4. Percentage of VOC-Controlled High-Ozone Grid Cells in the Three Subregions of SJV Under Different Baseline Emission Cases and Meteorological Conditions
 The control options in the northern SJV, however, still vary greatly with meteorology for all baseline emission cases. With reduced NOx emissions (mNOx or mBoth), VOC control remains important in this region under the O3-North and O3-West conditions (Table 4). The northern SJV is closest to large emission source regions (the SFB and SV) upwind of the SJV. When designing an ozone control strategy for this subregion, it would be important to consider the meteorological influences, especially those associated with flow characteristics, on the limiting precursor emissions and targeted source regions for emission scenarios in both year 2000 and more recent years.
3.4.1 Comparison to Past Studies
 Extensive studies reported in the literature have investigated ozone formation in central California [e.g., Pun et al., 2000; Liang et al., 2006; Steiner et al., 2006; Tonse et al., 2008; Jin et al., 2008; Beaver and Palazoglu, 2009; Zhao et al., 2011; Pusede and Cohen, 2012]. While some of these investigated the role of meteorology [e.g., Beaver and Palazoglu, 2009; Zhao et al., 2011], they focused on the influence of meteorology on ozone concentrations rather than on ozone sensitivities to precursor emissions and source-receptor relationships. The latter is the focus of our study.
 In practice, several representative episodes were selected for developing ozone control strategies for the San Joaquin Valley State Implementation Plan (SIP) (CARB 2007) submitted to the U.S. Environmental Protection Agency by the California Air Resources Board (CARB). The CARB submittal (CARB 2007) relied on three representative episodes: 7–13 July 1999, 29 July to 2 August 2000, and 17–21 September 2000, and modeling was conducted for the first two episodes. Our study focused on summer 2000, and thus, the episode in 1999 was not included. The other two CARB episodes were also included in our work (O3-All and O3-West). Note that ozone concentrations in excess of 75 ppb in these two CARB episodes were more extensive than in the other two episodes that CARB did not include (Figure 3, third column). Hence, the CARB episodes may represent the worst-case scenarios for ozone pollution.
 Selection of the four episodes for this study was based on the ozone spatial patterns in addition to ozone levels modeled for summer 2000 [Jin et al., 2011] and observed historically [Fujita et al., 1999]. As a result, despite their lower ozone levels than the “worst cases,” O3-South and O3-North were also selected due to their distinctive flow patterns. Nevertheless, ozone levels exceeding 75 ppb were observed in a number of places during these two (non-CARB) episodes. The 8 h standard has been reduced from 84 ppb to 75 ppb a year after the submittal of CARB SIP (CARB 2007) and will continue to be reduced further. It will become increasingly important to consider these two (non-CARB) episodes because they also contribute to the nonattainment of the current and future ozone standards.
 For the O3-All episode, our findings are similar to the CARB SIP submittal. NOx is found to be the predominant limiting reagent. However, by including more types of episodes, our work indicates meteorology-induced variabilities in the ozone limitation chemistry as well as in the importance of source contributions. VOC control can be important especially for the northern part of the SJV during two episodes (O3-North and O3-West). The importance of contributing source locations to the SJV ozone is found to be greatly affected by flow patterns. For example, the upwind contribution is enhanced during the more ventilated O3-North episode in the northern SJV, while local source contribution dominates ozone formation during the more stagnant O3-West episode.
 Model-simulated ozone sensitivities to precursor emissions can be influenced by the quality of emission inputs. We have addressed the effects of uncertainties and variations in emissions on the ozone limitation chemistry by computing ozone sensitivities under alternative emission cases and compared them to the original case. Changes in AVOC emissions are found to exert small effects on ozone sensitivities and control options. Changes in NOx emissions have greater effects; ozone chemistry was shifted toward NOx limitation with decreasing NOx. With 20% reduction in NOx, NOx control became, overall, beneficial for the central and southern parts of the SJV throughout all four episodes.
 Analysis in this study for control options has been focused on “high-ozone” locations. As ozone sensitivity to NOx changes sign, it is possible that NOx control may cause other areas to become noncompliant while cleaning up some of the already noncompliant areas and thus affect the efficacy of emission controls. This concern can be addressed by examining changes in the number of grid cells exceeding the ozone standard under alternative emission cases (with reduced precursor emissions) relative to those having the original base case emissions. The number of grid cells exceeding the 75 ppb ozone standard is decreased for all alternative emission cases relative to the original base case (Table S1). The effectiveness of VOC versus NOx reductions varies with episode in terms of reducing the spatial extent of ozone exceedances and is consistent with the control options described in section 3.1. The O3-All and O3-South episodes benefit more from NOx reductions, while the O3-North and O3-West episodes benefit more from AVOC reductions.
 Some uncertainties are not addressed in this work which could also influence the model-simulated ozone sensitivities.
 Simulated ozone sensitivities can be influenced by the choice of chemical mechanism used in the model. A recent study indicated greater ozone sensitivities to NOx simulated by the newly released Statewide Air Pollution Research Center '07 SAPRC07 mechanism [Shearer et al., 2012] than by the Statewide Air Pollution Research Center '99 SAPRC99 mechanism that is used by this study. Cai et al.  also indicated that the NOx disbenefit areas are spatially expanded using the condensed version of SAPRC07. Using observable indicators [Liang et al., 2006] and ozone production efficiencies calculated by ambient concentration measurements may provide another independent measure of ozone sensitivities to precursors.
 Forest fire emissions were found to be important ozone precursors and can interact with urban emissions [Pfister et al., 2008; Bein et al., 2008; Jaffe and Wigder, 2012; Strada et al., 2012; Singh et al., 2012]. Forest fire emissions are not included in this study and need to be investigated in future research.
 The choice of model year can also influence the application of our results to other years. Summer 2000 (June to September) was found to be statistically cooler than its 30 year climatology [Fujita et al., 1999]. Extending a similar modeling study to a different summer will be an additional important step for verifying the effects of temperature on model-simulated ozone sensitivities.
 We have investigated the meteorological dependence of ozone chemical limitation characteristics and local versus upwind contributions in the SJV by applying a three-dimensional photochemical model to simulate four ozone episodes, i.e., O3-North, O3-South, O3-West, and O3-All, that were found to be representative of the regional ozone pollution meteorology. Spatial distributions of ozone control options via NOx or AVOC and the transition regime have been delineated at high-ozone locations (i.e., with afternoon 8 h average ozone greater than 75 ppb) according to their first-order ozone sensitivity coefficients. Despite similar spatial trends with VOC-controlled areas that are found near emission centers and NOx-controlled areas located farther downwind, the spatial extent of the NOx disbenefit areas changes with meteorology, especially for the northern SJV. In contrast to previous findings obtained by studying a single ozone episode (O3-All) which revealed that NOx control was, overall, more beneficial for reducing 8 h ozone in the SJV [Jin et al., 2008], our current study results have suggested an increased importance of VOC control for the O3-North and O3-West conditions relative to the O3-All and O3-South conditions.
 The meteorological dependence of ozone control via NOx or AVOC varies significantly by location. Specifically, weaker dependences have been identified for the southern SJV and stronger dependences for the northern and central SJV. The effects of the scalar meteorological quantities considered here, namely, temperature, humidity, mixing height, and ventilation rate, can explain the ozone limitation chemistry characterized for the four episodes in the central and southern SJV. The northern SJV, being closest to large emission source regions (SFB and SV) upwind of the SJV, is subject to additional constraints by wind direction and flow characteristics that influence the choices of its control options. As has been revealed in the transect analyses, the extensive NOx disbenefit (or VOC control) areas identified under O3-North in the northern SJV are influenced by the transport of SFB pollutants.
 The importance of contributing source regions to the ozone formation in the SJV is also dependent on meteorology and is different from location to location along the flow paths. In general, the local contributions are more important for the eastern side of the central SJV, while upwind sources are also important (from ~40% to more than 50% of ozone sensitivities) for the western side of the valley, except for the most stagnant O3-West episode, when the SJV ozone is mostly sensitive to local sources at high-ozone locations. The VOC control was found to be important for the northern SJV for both the O3-West and O3-North conditions, and the results here have highlighted the differences in their VOC source regions, local SJV and upwind SFB, respectively, which are governed by the flow differences.
 The meteorological dependence of ozone control options identified in this study is also a function of the underlying emissions that influence the atmospheric composition and chemistry. Changes in AVOC emissions have been found to exert much smaller influence than changes in NOx emissions on the determination of ozone control options. With reduced baseline NOx emissions, the spatial extent of NOx disbenefit areas (VOC control) is reduced greatly in the central and southern SJV. This leads to a weaker dependence of the ozone limitation chemistry characteristics on meteorology in these two subregions, and NOx control is always the dominant option. The northern SJV, which is closer to the large upwind source regions (SFB and SV) than the rest of the valley, remains influenced by meteorology with VOC control dominant under the O3-North and O3-West conditions. When designing an ozone control strategy for this subregion, it would be important to consider the meteorological influences, especially those associated with flow characteristics, on the limiting precursor emissions and targeted source regions for various emission scenarios.
 The study summarized in this paper was performed with support from the California Energy Commission's PIER Environmental Program. It does not necessarily represent the views of the Energy Commission, its employees, or the state of California. We thank Marla Mueller in the California Energy Commission for managing the project. We also thank Robert A. Harley for the valuable technical advice and the three anonymous reviewers' help with improving the clarity of the manuscript.