The influence of stratosphere-to-troposphere transport (STT) on surface ozone (O3) concentrations in the greater Los Angeles area during the CalNex and IONS-2010 measurement campaigns has been investigated. Principal component analysis (PCA) of surface O3measurements from 41 sampling stations indicates that ∼13% of the variance in the maximum daily 8-h average (MDA8) O3between May 10 and June 19, 2010 was associated with changes of 2–3 day duration linked to the passage of upper-level troughs. Ozonesondes launched from Joshua Tree National Park and airborne lidar measurements show that these changes coincided with the appearance of stratospheric intrusions in the lower troposphere above southern California. The Lagrangian particle dispersion model FLEXPART reproduces most of these intrusions, and supports the conclusion from the PCA that significant transport of stratospheric air to the surface occurred on May 28–30. This intrusion led to a peak 1-h O3 concentration of 88 ppbv at Joshua Tree National Monument near the ozonesonde launch site on May 28, and widespread entrainment of stratospheric air into the boundary layer increased the local background O3 over the entire area to ∼55 ppbv on May 29–30. This background was 10–15 ppbv higher than the baseline O3 in air transported ashore from the Pacific Ocean, and when combined with locally produced O3 led to several exceedances of the current National Ambient Air Quality Standard (NAAQS) on the following day.
 Ozone (O3) formed through photochemical reactions of nitrogen oxides (NOx) and volatile organic compounds (VOCs) from anthropogenic sources is a serious secondary pollutant in many areas, including the Los Angeles Basin [National Research Council (NRC), 1991]. The adverse effects of high O3concentrations on human health and vegetation have long been recognized, and the U.S. Environmental Protection Agency (EPA) first established a 1-h National Ambient Air Quality Standard (NAAQS) for O3in 1971. The current 8-h standard of 75 parts-per-billion by volume (ppbv) was implemented in 2008 [U.S. Environmental Protection Agency, 2006], and the EPA is considering a further reduction of the NAAQS to a value between 60 and 70 ppbv (U.S. EPA, National Ambient Air Quality Standards for Ozone, 2010, Federal Register FRL–9102–1, pp. 2938–3052). This lower standard would be comparable to the 8-h standard of 66 ppbv adopted by the European Commission in 2010, but would still be higher than the exposure limit of 51 ppbv recommended by the World Health Organization and adopted in the UK.
 Tropospheric O3 has both natural and anthropogenic sources, and the adoption of more stringent standards will increase the need to better understand the contribution of “background” or “baseline” O3to ambient concentrations in non-attainment areas [NRC, 2009]. Baseline O3, as defined by the Task Force on Hemispheric Transport of Air Pollution [Dentener et al., 2010], is the O3 observed at a site when it is not influenced by recent local anthropogenic emissions and production. For the western United States, this includes O3 originating from anthropogenic sources across the Pacific in addition to that produced by photochemical reactions of NOx, VOCs, and carbon monoxide (CO) emitted from biogenic sources and wildfires, formed by lightning, or introduced into the troposphere from the stratosphere.
 Recent studies have shown that baseline O3 along the coast of the western United States is relatively high compared to the 75 ppbv standard, and is a significant portion of ambient O3 at inland locations. For example, sea level baseline O3 averages 42 ppbv in spring (March, April, May) and 27 ppbv in summer (June, July, August), which is 56% and 36% of the current NAAQS, respectively [Parrish et al., 2009]. Quantifying the amount of O3 throughout the depth of the daytime boundary layer during the late spring of 2010, Cooper et al.  found that baseline O3 equates to more than 80% of average O3 within California's Central Valley and up to 76% of the average O3 above the Mojave Desert and the Los Angeles Basin. Observations also show that baseline O3 at the surface and in the free troposphere above the western United States has increased significantly over the last two decades [Parrish et al., 2009; Cooper et al., 2010]. Most of this increase is likely due to growing anthropogenic emissions of O3 precursors in Asia [Jacob et al., 1999], but stratosphere-to-troposphere transport (STT) also contributes to baseline O3, and this input may increase with the recovery of O3 in the lower stratosphere and changes in circulation patterns associated with climate change [Shindell et al., 2006; Zeng et al., 2010].
 Model-based global climatologies [James et al., 2003; Sprenger and Wernli, 2003; Stohl et al., 2003a, 2003b; Wernli and Bourqui, 2002] have shown that there are preferred regions for episodic deep STT, including Baja California and the west coast of the United States south of the North Pacific storm track exit region during winter and spring. Given the frequency with which deep stratospheric intrusions are predicted to occur in the vicinity of southern California, it seems likely that O3-rich air from some of these intrusions could be entrained into the boundary layer over the Los Angeles Basin and influence surface O3. However, to the best of our knowledge, no previous measurement campaigns have examined STT over this region and attempted to quantify its effect on surface O3.
 In this paper, we address this question using measurements from surface stations, ozonesondes, and airborne lidar that deliberately targeted forecast intrusions, together with calculations from the FLEXPART Lagrangian particle dispersion model, during the CalNex (Research at the Nexus of Air Quality and Climate Change) and IONS-2010 (Intercontinental Chemical Transport Experiment Ozonesonde Network Study) measurement campaigns in May–June 2010. We use the ozonesonde and lidar profiles to characterize the distribution of O3 in the lower troposphere and boundary layer during the passage of several upper level low pressure systems, and compare the results to the FLEXPART calculations, which estimate the amount of O3transported to the Los Angeles area from the lower stratosphere and free troposphere, including long-range transport from Asia and biomass burning. We then examine the surface O3 concentrations, and use principal component analysis (PCA) to resolve the influence of synoptic scale transport, including STT, from that of local and regional photochemical production and transport. Finally, we estimate the influx of stratospheric O3 to the surface in the Los Angeles area during a particularly deep STT event (May 28–30), and discuss the implications for surface air quality.
 The potential impact of STT on surface O3 air quality in urban areas is even more uncertain. Langford et al. [2009a] recently described a case where STT directly contributed to an exceedance of the O3 NAAQS in the Denver, Colorado metropolitan area, and Akritidis et al.  used FLEXPART to show that STT could explain an unusual nighttime enhancement in surface O3 within the near sea level megacity of Athens, Greece. Although it has not been previously reported, STT can also potentially decrease surface O3 in heavily polluted areas such as the Los Angeles Basin where surface concentrations can be higher than those measured in some deep tropopause folds. Surface O3 can also be depressed by cold fronts that flush the boundary layer with clean free tropospheric air [Kunz and Speth, 1997; NRC, 1991] ahead of intrusions, or reduce the rate of local photochemical production through increased cloudiness and cooler temperatures.
3. Upper Air Measurements
 The spring of 2010 was unusually cool in southern California with frequent cold fronts and upper air disturbances. Figure 1 plots 700 hPa (∼3100 m ASL) geopotential surfaces from the NCEP/CDAS (National Centers for Environmental Prediction/Climate Data Assimilation System) Reanalysis [Kalnay et al., 1996] showing two of the six deep upper level troughs that passed over the western United States during the six-week long IONS-2010 campaign. The dotted black lines mark the locations of jet streaks (>50 m s−1) in the 300 hPa circulation, and the heavy solid lines show surface cold fronts associated with the low pressure systems (http://weather.unisys.com/archive/sfc_map/). Figure 1ashows the first of these low-pressure systems, which passed through the region on May 11. The plot shows the center of the low to be closed off from the main circulation at 700 hPa. In addition to tropopause folding [Langford et al., 1996b], closed lows can facilitate STT through convective and turbulent erosion of the stratospheric air trapped within the vortex [Lamarque and Hess, 1994; Price and Vaughan, 1993]. The low-pressure system on May 28 inFigure 1b was shallower, but had a very large jet streak on the eastern flank.
 The O3, temperature, and frost point distributions in the free troposphere and boundary layer were measured by near daily ozonesondes equipped with electrochemical cell (ECC) sensors launched from six locations in California as part of IONS-2010 [Cooper et al., 2011]. Two of these launch sites were located in Southern California: San Nicolas Island (SN), ∼140 km southwest of Los Angeles, and Joshua Tree (JT), located in the Mojave Desert on the northwest edge of Joshua Tree National Park. The next closest launch site was Point Sur (PS), ∼400 km northwest of Los Angeles on the coast. Approximately 36 ozonesondes were released from each site between May 10 and June 19. Sondes were launched six afternoons per week, with no soundings on Sundays (May 16, May 23, May 30, June 6, and June 13). The sondes were typically released at ∼2200 UT (1400 Pacific Standard Time, PST). The IMET relative humidity measurements by the pre-2010 sensors used in most of the soundings appear to be inaccurate below ∼5% RH [Hurst et al., 2011]. Other instrumental problems limited the number of useful profiles from San Nicholas Island. The latter will not be discussed here.
Figure 2 plots the O3, potential temperature, and frost point profiles measured by the Point Sur and Joshua Tree ozonesondes ∼10 h after the synoptic conditions shown in Figure 1. The May 11 soundings from both sites (Figures 2a and 2b) show extremely dry layers with high O3(>150 ppbv) and high static stability 4.5 to 6.5 km above mean sea level (ASL) consistent with a large stratospheric intrusion. The NOAA WP-3D research aircraft also intercepted this intrusion over the Central Valley, measuring more than 180 ppbv of O3 and less than 120 ppbv of CO between 4.5 and 5.5 km ASL [Neuman et al., 2012]. The Point Sur sounding from May 28 (Figure 2c) shows a similar layer with more than 180 ppbv of O3 near 5 km ASL. However, in contrast to the May 11 sounding, the Joshua Tree profile (Figure 2d) shows that at least part of this intrusion approached to within 2 km of the ground.
3.2. Airborne Ozone Lidar
 Stratospheric intrusions associated with upper level disturbances were also detected by the Joshua Tree ozonesondes on May 19, May 23, May 29, June 11, and June 16. There was no sounding during the passage of a second deep closed low on May 23, but an intrusion was detected by the Tunable Optical Profiler for Aerosols and oZone (TOPAZ) differential absorption lidar (DIAL) [Alvarez et al., 2011; Langford et al., 2011] flown aboard the NOAA Twin Otter aircraft during CalNex. This downward-looking instrument was operated on 46 flights in Southern and Northern California between May 23 and July 18, 2010, flying from the Ontario Airport in San Bernardino County from May 23 to June 7, and again from June 30 to July 18.Figure 3a shows a curtain plot of the O3vertical distribution measured by TOPAZ with a 90-m vertical range resolution (smoothed with a 450-m running filter) and 600-m horizontal resolution (10-s integration) along an east-to-west flight leg ∼10 km south of Joshua Tree on the afternoon of May 23. The lidar measured O3 concentrations of up to ∼250 ppbv at 4.6 km ASL, while the in situ monitor aboard the Twin Otter measured up to ∼300 ppbv at 5.3 km ASL. These high concentrations suggest an unusually large tropopause fold, consistent with the deep closed low. The high O3 layer remained above 4 km ASL throughout the flight, however.
 TOPAZ also mapped the intrusion associated with the upper level trough shown in Figure 1b between 0930 and 1300 PST on May 29. Figure 3bshows the measurements made along a north-to-south flight leg ∼10 km west of Joshua Tree. The measurements show the intrusion to be split north of ∼34 °N, with concentrations exceeding 100 ppbv in a lower tongue less than 1.5 km above the surface in the Mojave Desert, well below the tops of the highest peaks (∼3.5 km ASL) in the San Bernardino Mountains. This structure is similar to that seen in the ozonesonde profile launched ∼2 h later (Figure 3b, black trace). Measurements made earlier in the same flight show the high O3 layer extended westward from the desert across the Los Angeles Basin to the Pacific Ocean. A second flight (∼1700–2000 PST) over the Basin later the same afternoon found in situ O3 concentrations as high as 110 ppbv near 4 km ASL, with concentrations between 50 and 60 ppbv throughout most of the boundary layer. These measurements (not shown) also found a narrow layer with ∼80 ppbv that appeared to be lifted from the surface to ∼1.5 km ASL over the Basin by the sea breeze. The TOPAZ measurements and the morphology of the May 23 and May 29 intrusions will be described in more detail elsewhere.
4. FLEXPART Model
 The FLEXPART Lagrangian particle dispersion model (version 8.1) [Stohl et al., 2005] was used to estimate how much of the O3in and above the Los Angeles Basin and surrounding areas during IONS-2010 was transported from the free troposphere, the lower stratosphere, across the Pacific from Asia [Cooper et al., 2010] or from biomass burning sources [Jaffe, 2011]. The model calculates the forward trajectories of a multitude of particles carried by the global NCEP GFS winds with a temporal resolution of 3 h (analyses at 0000, 0600, 1200, and 1800 UTC; 3-h forecasts at 0300, 0900, 1500, and 2100 UTC), a horizontal resolution of 0.5° × 0.5°, and 26 vertical levels. Particles are transported both by the resolved winds and by parameterized sub-grid motions. The paths of the individual particles are not retained. To account for convection, FLEXPART uses the parameterization scheme ofEmanuel and Zivkovic-Rothman , which is implemented at each 15-min model time step. This scheme includes entrainment and mixing, cloud microphysical processes, and large-scale control of ensemble convective activity, including the interaction between convective downdrafts and surface fluxes. Mean tracer concentrations are calculated for three 1° × 2° receptor domains centered over Santa Barbara County, the Los Angeles Basin, and the Mojave/Colorado Deserts, using 0 to 3 and 3 to 6 km ASL altitude bins. The mean concentrations are smoothed with an 8-h running average and the concentrations at 2200 UT (1400 Pacific Standard Time, PST) compared to the measured surface and ozonesonde concentrations. The latter were launched at about this time to coincide with the peak in surface O3, and autocorrelations confirm that the best agreement between the model results and measurements is obtained for this time of day.
 The stratospheric tracer masses were carried by particles released in the stratosphere (>2 potential vorticity units, PVU) over the North Pacific Ocean between 140 and 90°W, and between 25 and 55°N. The mass of O3 carried by the particles is calculated using a linear relationship between O3 and potential vorticity (60 ppbv/PVU for PV higher than 2 PVU) at the trajectory origin in the stratosphere. The mass of O3 is conserved and carried along the trajectory for up to 10 days; similar results are obtained if the trajectories are shortened to 5 days. The Asian tracer is based on the amount of CO released into the boundary layer from anthropogenic sources using the EDGAR 3.2 fast track inventory [Olivier et al., 2005], and is carried over 20 days. The biomass burning CO emissions were calculated using the algorithm of Stohl et al. , which uses MODIS fire detection data, information on land use, and published emission factors. Instead of prescribing an injection height, we used a probability density function relative to the local PBL height as described by Brioude et al. . The biomass burning CO tracer includes both North American and Asian sources.
 The amount of O3 produced in the anthropogenic emission and biomass burning plumes is quite variable, but typically ranges from ∼0.2 to 0.5 mol mol−1 of CO [Bertschi et al., 2004; Heald et al., 2003; McKeen et al., 2002; Price et al., 2004; Zhang et al., 2008]. We assume average values of 0.33 and 0.2 mol mol−1 for the Asian transport and biomass burning plumes, respectively, in our calculations. The former leads to mean contributions of ∼8 and 6 ppbv O3 to the lower free troposphere and boundary layer from anthropogenic sources in Asia, in good agreement with other estimates [Zhang et al., 2008]. Finally, the free tropospheric tracer is carried by particles released above the boundary layer, but below the tropopause (defined as the 2 PVU isentropic surface) over the domain from 140 to 90°W, and 10 to 50°N. This tracer is also followed for 10 days. A mean free tropospheric O3 concentration of 60 ppbv (including the contribution from Asia) [Cooper et al., 2011] is assumed for the entire domain.
4.1. FLEXPART Results
Figure 4a plots time series of the mean lower free troposphere (3 to 6 km ASL) O3concentrations calculated by FLEXPART for the Mojave/Colorado Desert target domain, which includes the Joshua Tree ozonesonde launch site, during IONS-2010. The contributions from the free tropospheric, stratospheric, Asian transport, and biomass burning tracers are plotted along with their sum. The dotted blue lines mark the presence of cold fronts, and the plot shows that the stratospheric tracer increases dramatically following several of the frontal passages. Peak concentrations between 20 and 105 ppbv occurred on May 11, 23, and 29, and again from June 12 to 14. The largest peaks are associated with the deep closed lows on May 11 (64 ppbv) and 23 (105 ppbv).
 FLEXPART also shows that there was significant transport of O3 from Asia into the free troposphere above the Mojave Desert. These pollution plumes often become commingled with descending stratospheric air in cyclonic systems over the Pacific [Cooper et al., 2004b; Lin et al., 2012]. Although the FLEXPART Asian tracer concentrations in the lower free troposphere were much smaller than the stratospheric peaks, some influx occurred nearly every day and the mean contribution during IONS-2010 (∼8.3 ppbv O3) was similar to that of the stratospheric tracer (∼8.6 ppbv O3). The large Asian transport events on May 16–20 and June 17–19 in the intermountain west investigated by Lin et al. using a new high-resolution (∼50 × 50 km2) global chemistry-climate model (GFDL AM3) and satellite measurements of CO and total O3 columns, are not prominent in the southern California FLEXPART time series. The mean biomass burning contribution to free tropospheric O3during IONS-2010 was <2 ppbv. Similar results were found for the Santa Barbara and Los Angeles domains.
Figure 4b is similar to Figure 4a, but shows the FLEXPART results for the surface up to 3 km ASL (∼2 km AGL), a typical boundary layer depth over Joshua Tree during late spring. The stratospheric contributions to boundary layer O3 are generally much smaller than the 3 to 6 km contributions seen in Figure 4a, with maximum concentrations of up to 27 ppbv during IONS-2010. The difference is particularly striking on May 23, when the boundary layer contribution was only ∼7 ppbv compared to 105 ppbv in the free troposphere. This is consistent with the TOPAZ observations, which show the intrusion remaining above 4 km ASL throughout the flight (seeFigure 3a), and with the unseasonably low surface temperatures (T < 14°C) that imply a relatively shallow boundary layer over Joshua Tree. The stratospheric contribution to boundary layer O3 in the Mojave Desert on May 11 (∼26 ppbv) was also much smaller than the corresponding free tropospheric concentration (∼64 ppbv), consistent with the ozonesonde profile in Figure 2b, which shows the bulk of the intrusion to be above 5 km ASL. The contribution of the stratospheric tracer to boundary layer O3on May 11 was much smaller in the Los Angeles (∼8 ppbv) and Santa Barbara (∼4 ppbv) domains (not shown). FLEXPART shows an average stratospheric contribution to the boundary layer over the Mojave Desert of 5.1 ppbv during IONS-2010. The mean Asian transport tracer is similar (6.0 ppbv), but with peak values much smaller than the stratospheric tracer and no clear peaks associated with the transport events of May 16–20 and June 17–19 mentioned above. The mean value of the biomass burning tracer is <3 ppbv.
 The FLEXPART results in Figure 4b suggest that the most significant influx of stratospheric air into the boundary layer over the Mojave Desert occurred on May 29–30 and June 14–15, with peak O3 concentrations of ∼27 and ∼16 ppbv, respectively. In both cases, the model shows increased O3 persisting over several days with similar contributions in the Mojave Desert, Los Angeles, and Santa Barbara target domains. The large boundary layer contribution on May 29–30, in particular, agrees well with the ozonesonde and TOPAZ measurements in Figures 2 and 3.
4.2. Comparison to Upper Air Measurements
Figure 5 compares the FLEXPART tracers with the ozonesonde and lidar measurements. Figure 5a plots the average O3 between 3 and 6 km ASL from the ozonesondes against the sum of the Asian, biomass burning, stratospheric, and free tropospheric tracers in Figure 4a. The filled circles represent the measurements from the 14 days when FLEXPART showed a stratospheric contribution of at least 0.5 ppbv; the two circled points correspond to the measurements from May 11 and 29. Linear regression of the ozonesonde data gives a slope of 1.07 ± 0.03 (R = 0.84) when the fit is constrained through zero, and a slope of 0.69 ± 0.13 with an intercept of 25 ± 9 ppbv when it is not. The crosses represent the 22 days when FLEXPART found no stratospheric contribution. Many of these days are associated with southwesterly flow, and the low bias arises because the constant 60 ppbv free tropospheric background concentration assumed in the FLEXPART calculations overestimates the O3 concentrations in the subtropical air masses transported to the region on many of these days.
 The red triangles in Figure 5a show TOPAZ measurements made within 200 km of the Joshua Tree launch site and within 4 h of the ozonesonde launches on May 23 and 29. These points represent averages from 3.0 km ASL to the upper altitude limit of 4.8 km ASL. Only data from these two days are shown since the Twin Otter did not fly above 4.5 km ASL during any of the other flights. These results are only in fair agreement with the FLEXPART calculations, since the TOPAZ measurements underestimate the 3 to 6 km concentration on May 23 when the aircraft flew below the O3 maximum (see Figure 3a), and overestimate the 3 to 6 km mean concentration on May 29 since the average does not include lower O3 concentrations of 40–60 ppbv measured by the ozonesonde between 5 and 6 km ASL (see Figure 3b).
Figure 5a shows that FLEXPART reproduced the measured free tropospheric O3concentrations quite well when the large-scale flow was northwesterly. A notable exception occurred on May 19 (large square inFigure 5a) when the Joshua Tree ozonesonde detected an intrusion with up to 129 ppbv of O3at ∼4.5 km ASL following a frontal passage associated with a shallow upper level disturbance. This intrusion was also sampled by the NOAA WP-3D near Joshua Tree, but was missed by FLEXPART. This may have resulted from the O3 filament being finer than the 0.5° GFS model resolution; this explanation is supported by the San Nicholas and Point Sur soundings, which showed no evidence of an intrusion on that day. This intrusion was evident in the GFDL AM3 results [Lin et al., 2012], however.
 A comparison between the FLEXPART results and the boundary layer measurements is shown in Figure 5b. TOPAZ measurements from 8 flights are shown, along with the ozonesonde measurements. The ozonesonde averages from 1.2 to 3.0 km ASL (0 to 1.8 km AGL) and the TOPAZ averages from 1.7 to 3.0 km ASL (0.5 to 1.8 km AGL) for the 7 common measurement days agree quite well, with O3[TOPAZ] = 1.09 ± 0.26(O3[sonde]) − 4.0 ± 17.5 ppbv and R = 0.88. Most of the measurements lie from 0 to 40 ppbv above the 1:1 line in Figure 5b since FLEXPART does not account for O3 transported into the desert from the Los Angeles Basin. For this reason, the accuracy of the FLEXPART stratospheric contribution to boundary layer O3 cannot be directly evaluated from the Joshua Tree ozonesonde or lidar measurements.
5. Surface Ozone Measurements
 The FLEXPART boundary layer predictions can also be compared to measurements from the extensive network of surface O3 monitoring sites maintained by the California Air Resources Board (CARB) in cooperation with other public and private agencies. Figure 6shows the locations of the 57 sampling stations (color-coded by county) operational during the spring of 2010 in the five counties composing the Greater Los Angeles Area (Ventura, Los Angeles, Orange, San Bernardino, and Riverside counties), along with Santa Barbara County to the northwest. The EPA classifies the South Coast Air Basin (SoCAB), which includes Orange, and portions of Los Angeles, Riverside and San Bernardino counties, as an extreme nonattainment area for O3.
 Ozone is measured continuously at the CARB sites by commercial ultraviolet absorption analyzers and the hourly, daily, and daily maximum 8-h average (MDA8) values are reported on the ARB website (http://www.arb.ca.gov/aqmis2/aqdselect.php). We base our analysis mainly on the MDA8 values since this is the primary metric used to determine O3 air quality. Although significant changes are not expected, the posted data are considered preliminary and subject to additional quality assurance testing for up to two years after their acquisition. Measurements from the eight sites enclosed by large diamonds will be discussed in detail.
5.1. Background Ozone
 The prevailing northwesterly flow during spring and early summer ensures that sites on the west coast of Santa Barbara County often sample baseline marine boundary layer air transported ashore from the North Pacific Ocean. This background is represented here by measurements from an open headland less than 1 km from the Pacific Ocean at Vandenberg Air Force Base (VBG); these are plotted as dotted black lines in Figure 7. Since these measurements have not been filtered to remove the influences of coastal ship traffic or local recirculation [Parrish et al., 2009], they represent an upper limit for the marine boundary layer baseline. The VBG measurements exhibit a pronounced spring (mid-April) maximum consistent with baseline O3 measurements from other low elevation locations in the northern hemisphere [Monks, 2000; Oltmans et al., 2008; Parrish et al., 2009]. The MDA8 O3 at VBG slowly declined following the spring maximum in April, and the time series in Figure 7is punctuated by a one-week period with very low O3 (MDA8 ≈ 20 to 25 ppbv) at the beginning of June. Trajectory calculations show that these unusually low O3 concentrations were associated with air transported from the subtropical North Pacific [Oltmans et al., 2008; Parrish et al., 2009]. Low values were also observed at sites up to 50 km inland along the west coast, but not along the south coast at Goleta (GOL, Figure 7a), or Thousand Oaks (THO, Figure 7b), which are often influenced by polluted air recirculated inland by the land-sea breeze.
5.2. Inland Measurements
 The interaction of the large-scale flow with the land-sea breeze and local circulations created by orographic heating creates complex transport patterns within the Los Angeles Basin [Lu and Turco, 1994]. The sea breeze transports both primary pollutants and their photochemical products north and east across the Basin where they accumulate near the San Gabriel and San Bernardino Mountains. Some of these pollutants are transported northward from Los Angeles up the San Fernando Valley to Santa Clarita (SCL, Figure 7c). Although the mountains trap much of the pollution within the Basin, some is transported into the Colorado and Mojave deserts or lifted up the mountain slopes by the orographic flows to be advected back over the Basin or vented into the free troposphere [Langford et al., 2010]. Thus, Crestline (CRL, Figure 7d), which lies ∼1400 m above the Basin in the San Bernardino Mountains often records the highest measured O3concentrations in the Los Angeles area. Indeed, Crestline led the South Coast Air Basin (and the state of California) with 74 exceedance days in 2010, 14 of which occurred during IONS-2010. Moreover, the peak MDA8 O3 measured at Crestline on June 5 (123 ppbv) was the highest recorded in all of California during 2010. This high O3 episode, which persisted from June 3 to 6, would have been even more severe had it not coincided with the unusually low background O3 concentrations (∼23 ppbv) in the air transported ashore from the Pacific.
Figure 7 shows that MDA8 O3 concentrations comparable to the current 75 ppbv NAAQS also occurred at Santa Clarita and Crestline on May 16 and June 17 when the GFDL AM3 model found large Asian transport events [Lin et al., 2012], and on May 30–31 and June 13–14 when FLEXPART showed significant transport of stratospheric O3 into the boundary layer. The lowest concentrations at both sites were measured during the passage of cold fronts (light blue vertical lines), including May 23 when the large stratospheric intrusion shown in Figure 3a passed overhead. The descending free tropospheric air and cool temperatures associated with these fronts suppressed vertical mixing across the Basin, and often decreased the surface O3 concentrations to values near the marine boundary layer concentrations measured at Vandenberg.
6. Principal Component Analysis
 Day-to-day variations in surface O3are created both by local and regional scale processes (i.e., chemistry, transport, and deposition) and by transport associated with large-scale circulations and weather systems. These processes create different spatial and temporal patterns in the surface O3 distributions. We use principal component analysis (PCA) to resolve the contributions of different processes to the overall variance in the surface O3 concentrations. We then seek to identify the responsible processes by comparing the spatial and temporal variability of the principal components to meteorological data from the NCEP/CDAS (National Centers for Environmental Prediction/Climate Data Assimilation System) Reanalysis [Kalnay et al., 1996]. Since we do not combine the O3 measurements and meteorological parameters in the PCA itself, we can interpret the PCA results more quantitatively [Langford et al., 2009b].
 PCA is a non-parametric technique used to identify patterns of common variance in large data sets. Unlike factor analysis, PCA requires no assumptions about the data distribution. The principal components are simply linear combinations of the original variables weighted by appropriate coefficients (or loadings) that range from −1 to +1. Thus the number of components required to account for all of the variance in the data set is equal to the original number of variables. However, if many of the variables are correlated, as would be the case if common processes influence their behavior, most of the variance can be explained by a much smaller number of mutually uncorrelated or orthogonal principal components. Although the PCA tells us nothing about the physical processes responsible for the variance, the cause can often be inferred for the most significant components using other information.Langford et al. [2009b] used PCA to isolate locally produced O3 from the regional background in the Greater Houston area, and we use a similar approach in this study.
6.1. PCA Results
 PCA was applied to the 43-day (May 9 to June 20, 2010) time series of the MDA8 O3 values from 41 of the 57 surface monitoring stations shown in Figure 6 and located in eastern Santa Barbara County, the Los Angeles Basin, and the nearby Mojave and Colorado Deserts. Only those stations with complete data records are used in the analysis, which was performed on the correlation matrix with no further rotation of the principal components. Urban sites that appear to be heavily influenced by local sources are excluded from the analysis, as are coastal sites that often lie within the marine boundary layer. The PCA shows that more than 78% of the variance in the data set is explained by only two of the 41 principal components. The first component, PC1, accounts for 66% of the total variance, while PC2 accounts for another 13%. The next largest component explains less than 5% of the total variance and is not considered further.
 The results for the first two principal components are summarized in Figure 8: the filled circles in Figures 8a and 8b also show the 41 sampling stations used in the PCA (crosses represent the remaining stations from Figure 6). The colors of the circles show the coefficients (loadings) for PC1 (Figure 8a) and PC2 (Figure 8b), with positive coefficients shown in red and negative coefficients shown in blue. The relative sizes reflect the fraction of the total variance (i.e., square of the coefficient) at each station explained by that component. For any given day, the contribution of each principal component to the overall variance may be larger or smaller than the mean represented by the coefficient; this time dependence is given by amplitudes (or scores) obtained by projecting the measurements onto new axes defined by the principal components. These are plotted in Figures 8c and 8d. Like the coefficients, the amplitudes can be either positive or negative, but they must sum to zero over the sample time series. The product of the coefficient and amplitude will be positive if the MDA8 on a given day is larger than the time series mean and vice versa.
Figure 8a shows that PC1 has large positive coefficients in Los Angeles, San Bernardino, and Riverside counties, with smaller coefficients near the coast. Although PC1 accounts for only ∼66% of the overall variance in the data set, it explains more than 90% of the variance at many of the stations in the eastern Basin, and the amplitude time series plotted in Figure 8c closely resembles the MDA8 time series from Crestline plotted in Figure 7c, with peaks during all of the high O3 episodes. Since all of the loadings are positive, PC1 reflects processes that simultaneously increase or decrease O3 at all 41 sites.
Figure 8b shows that PC2 has large positive coefficients in Santa Barbara County, with smaller positive coefficients along the coast in Ventura, Los Angeles, and Orange counties. The amplitude time series plotted in Figure 8d resembles the Goleta MDA8 O3 time series in Figure 7a, with a large peak on May 30 and a smaller peak on June 12. The PC2 coefficients approach zero further inland, and become negative at stations near the mountains and in the desert, including both Santa Clarita and Crestline. This transition implies that PC2 is associated with processes that increase O3 (relative to the time series mean) at sites near the coast, while decreasing it (relative to the time series mean) at sites further inland.
6.2. Interpretation of the PCA Results
 Since photochemistry dominates the O3 variance over most of the Los Angeles Basin, it is no surprise that the spatial and temporal distribution of PC1 is consistent with local photochemistry and transport. All of the high O3 episodes in Figure 7 have corresponding positive peaks in the PC1 amplitude time series, with the largest amplitude coinciding with the highest O3 on June 5. The negative excursions in PC1 coincide with the passage of cold fronts. The amplitudes generally increase from May to June, and Figure 8cshows that the amplitudes are positively correlated (R = 0.73, P < 0.001) with the mean surface temperature (dotted black line) from the NCEP/CDAS Reanalysis. The PC1 amplitudes are also negatively correlated (R = −0.63, P < 0.001) with the mean wind speed at 850 hPa, and with the fractional low and mid-level cloud cover (R = −0.49, P < 0.001) over the same area (not shown). These correlations show that PC1 is associated with warm, sunny, and stagnant conditions, consistent with photochemistry.
 The NCEP/CDAS Reanalysis data also support the conclusion that PC2 is associated with synoptic and large-scale transport, including STT. The amplitudes inFigure 8d generally decrease from May to June, with positive excursions from this trend on May 30 and June 12. The amplitudes have a weak, but significant, negative correlation with temperature (R = −0.36, P < 0.02) due to the cold fronts, and are negatively correlated (R = −0.42, P < 0.003) with the zonal winds at 700 hPa (not shown), which shift to the east as the cyclonic circulations associated with the upper level lows pass through. The association with cyclonic systems also leads to a positive correlation (R = 0.52, P < 0.001) with the Reanalysis absolute vorticity at 700 hPa, which is plotted on the right side of Figure 8d.
 The positive correlation of the PC2 amplitudes with upper level cyclonic activity is consistent with downward transport of O3 to the surface by these systems. There are no peaks in PC2 corresponding to the Asian transport events of May 16–20 and June 15–19. However, a direct link between PC2 and STT is supported by the correspondence between the dominant peaks in the PC2 amplitudes in Figure 8d and the FLEXPART stratospheric tracer time series in Figure 4b. The correlation between the two time series is strongest (R = 0.62, P < 0.001, 24 h lag) for the Santa Barbara (0–3 km ASL) tracer, consistent with the large positive loading of PC2 in Santa Barbara County (Figure 8b).
6.3. Deconvolution of the MDA8 Time Series
 Since we include only the MDA8 O3 measurements in our principal component analysis, the results can be used to quantitatively resolve each MDA8 time series into the contributions related to local photochemistry (PC1), and to STT (PC2). Deconvolution of these component time series might not be possible if other types of data (e.g., meteorological parameters) nonlinearly related to the O3 concentrations are incorporated in the analysis. In the present analysis, the product of the time series standard deviation, component loading, and component amplitudes at each site approximately gives the individual component contributions as deviations from the time series mean [Langford et al., 2009b]. Figures 9a–9d plot these deviations for the Goleta, Thousand Oaks, Santa Clarita, and Crestline time series from Figure 7. The contribution of each component to the overall variance is indicated in each panel. Note that the PC1 time series appear very similar since each is just the PC1 amplitude time series from Figure 8c scaled by the standard deviation of the mean MDA8 for each site and the site specific component loading. The same is true for PC2, except that in the eastern Basin the loading becomes negative (see Figure 8b) and the product time series appear inverted.
Figure 9a shows that the high (relative to the time series mean) MDA8 values recorded at Goleta on May 30 (∼11 ppbv), and to a lesser extent June 13 (∼5 ppbv), correspond to the peaks in PC2. Slightly larger, ∼16 and 6 ppbv, and smaller peaks ∼8 and 3 ppbv, respectively, are found for nearby Carpenteria and Thousand Oaks (cf. Figure 9b). The net contribution of PC2 to the MDA8 values diminished further inland, becoming negative in the eastern Basin and decreasing the MDA8 (again relative to the time series mean) by ∼10 ppbv at Santa Clarita and by ∼20 ppbv at Crestline on May 30. The behavior on June 13 is similar, but smaller in amplitude. The maximum contribution of PC2 to surface O3 on May 11 was less than 4 ppbv at all sites.
Figures 9c and 9d show that the decreases in MDA8 at Santa Clarita and Crestline on May 30 are offset by increases in PC1. In contrast, only PC1 appears to be important during the high O3 episode of June 3–6. The time series also show that negative PC1 amplitudes are associated with the low MDA8 values measured at inland sites during the passage of cold fronts (e.g., May 18 and 27). This is consistent with decreased photochemical production associated with the cooler temperatures and increased cloudiness on these days.
 The adequacy of the first two principal components in explaining the variance in the MDA8 time series is seen in Figures 9e–9h, which plot the sum of the PC1 and PC2 contributions from Figures 9a–9d and the time series means. These synthesized time series closely correspond to the originals, particularly at the inland sites such as Santa Clarita and Crestline where the first two components explain 87% of the overall variance. The adequacy of the first two components decreases toward the coast, where changes in the marine boundary layer background contribute to the variance.
7. Case Study: May 27–30, 2010
 The upper air measurements, FLEXPART trajectories, and principal component analysis all indicate that the upper level trough and cold front that passed through Southern California between May 27 and 30 (see Figure 1b) had the greatest influence on surface O3. Although this trough was less deep or extensive than the closed lows that passed through the area on May 11 and May 23, the associated intrusion (or intrusions) penetrated deeper into the troposphere, and remained over the region for several days.
7.1. Ozonesonde Profiles
Figure 10 shows the lowest 2.5 km of the Joshua Tree ozonesonde profiles from May 27, 28, and 29. The boundary layer (dotted line) behind the cold front on May 27 (Figure 10a) is quite shallow (∼0.8 km), cool, and moist (Tf ∼ 7°C), with a mean O3 concentration of only 40 ppbv and an entrainment zone (determined from the potential temperature gradient) ∼0.1 km deep (dashed line). Descending upper tropospheric/lower stratospheric air behind the surface front is seen in the O3 and Tf gradients. The O3peaks at ∼90 ppbv near 3.5 km ASL (∼2.3 km AGL). The southwesterly winds show that the low-pressure center was still to the west of Joshua Tree.
 The stratospheric intrusion is seen in the May 28 sounding (Figure 10b) with an O3 peak of 120 ppbv only 1.5 km AGL. The mean boundary layer O3 concentration increased to ∼80 ppbv while the frost point decreased to −9°C, indicating that air from the upper troposphere and/or lower stratosphere air had been mixed into the boundary layer. The “noisy” appearance of the boundary layer profiles suggests significant horizontal variability. The winds were almost westerly above the top of the boundary layer (Figure 10e) as the trough passed overhead, and since the boundary layer was still cool and shallow, the entrainment was most likely driven by turbulent mixing associated with the strong vertical wind shear (>10 m s−1 km−1) seen in Figure 10e.
 The winds had shifted to the north by May 29 (Figure 10f) as the low-pressure center moved to the east. The boundary layer was much warmer, growing to ∼1.2 km (Figure 10c) and the entrainment zone nearly 0.3 km deep, extending into the bottom of the high O3 tongue. The smooth profiles show the boundary layer to be much better mixed than on the previous day. Although the O3 tongue is clearly being entrained and mixed into the boundary layer and the frost point remained near −9°C, the mean O3 mixing ratios were lower than on the 28th (∼55 ppbv). There was no sounding May 30.
7.2. Surface Measurements
 The evolution of boundary layer O3 seen in the ozonesonde profiles is also reflected in surface measurements made over a wide area. Figure 11a plots time series of the hourly O3 measurements from Thousand Oaks, Santa Clarita, and Crestline, which lie around the periphery of the Basin, as well as Riverside in the eastern Basin and Joshua Tree National Monument in the Mojave Desert. The vertical dashed lines mark the launch times of the ozonesondes from Figure 10, and the gray boxes bracket the period from 1000 to 1700 PST when the boundary layer is most likely to be well-mixed.Figures 11b and 11c are similar, but show the corresponding PM2.5 and dew point measurements (PM2.5 was not measured at Joshua Tree). Since some stations reported relative humidity instead of dew point, these measurements were converted to dew point (http://www.srh.noaa.gov/images/epz/wxcalc/rhTdFromWetBulb.pdf) for consistency.
Figure 11a shows that low daytime O3 concentrations (∼40 ppbv) similar to those measured by the ozonesonde (see Figure 10a) were recorded at most of the sampling sites during the frontal passage on May 27. The dew points and PM2.5 concentrations were high as the cold front brought scattered clouds and light precipitation to Santa Barbara County and the northwestern Basin.
 The skies cleared on May 28 and the daytime O3 concentrations at most of the sites increased to ∼50 ppbv with slightly lower dew points. High PM2.5was measured at the higher elevation sites. The hourly measurements from Joshua Tree, however, show a short-lived peak of much higher O3 (88 ppbv) and very low dew points (−10°C) in good agreement with the ozonesonde (Figure 10b). Despite the southwesterly winds, this dry, O3-rich air could not have been transported from the western and central Basin since the O3concentrations there were much lower, and must have been entrained into the boundary layer from the low-lying stratospheric intrusion shown inFigures 2d and 10c. The localized nature of this transport is reflected in the much smaller peak (∼68 ppbv) only 30 km away at Palm Springs (171 m ASL), and the much lower concentrations (<55 ppbv) at sites further west. Thus it appears that transport of stratospheric air into the boundary layer was confined primarily to the high Mojave Desert on May 28, and did not impact the Los Angeles Basin or coastal areas. This is consistent with the shallow boundary layers and weak mixing over the Basin immediately following the cold front. Since the peak in surface O3 on May 28 affected only a few of the stations used in the PCA, the PC2 amplitude in Figure 8d is small.
Figure 11 shows that stratospheric air affected a much larger area on May 29. This is reflected in the near uniform daytime O3 mixing ratios of 50 to 60 ppbv (Figure 11a) and low dew points (<−5°C) at all of the surface stations. Figure 11b shows that these transitions were also accompanied by a decrease in PM2.5 concentrations to <3 μg m−3 and in many cases zero. This indicates that the descending air was unlikely to have been influenced by nearby urban areas or biomass burning. The low O3, Td, and PM2.5 values characteristic of the descending air are also evident in the late morning and early afternoon at Riverside (250 m ASL). Local influences are seen in the measurements from most of the sites by the afternoon of May 30.
7.3. O3-H2O Mixing Plots
 Many studies e.g. [Hintsa et al., 1998; Pan et al., 2004] have shown that mixing between the lowermost stratosphere and uppermost troposphere leads to well-defined curves in scatterplots of O3 and tropospheric tracers such as H2O or CO. For the cool and clear conditions that followed the passage of the cold front on May 27, it seems possible that the mixing between the dry, O3-rich stratospheric intrusion with the moist, O3-poor marine boundary layer advected inland from the coast might be similar. CO would provide a more conservative tracer for tropospheric air, but precise CO measurements were not made by the CARB stations or by the ozonesondes. Water vapor should also be reasonably well conserved since there was no significant cloud formation from May 28 to 30. Since accurate pressure measurements are not available to convert the dew points measured (or calculated from measured relative humidities) at the surface sites to H2O mixing ratios, we examine the correlation between O3 and dew point (Td), which varies almost exponentially with the H2O mixing ratio for dew points above about −40°C. The frost points (Tf) measured by the ozonesondes are equivalent to Td for temperatures above 0°C, or altitudes below ∼4 km ASL for the late May Joshua Tree soundings.
Figure 12 compares the O3 and Td measurements from the boundary layer and entrainment zone for May 28 to 30. Gray crosses represent the ozonesonde measurements, while the filled colored symbols represent the hourly surface measurements between 1000 and 1700 PST (1800–0100 UT) (see Figure 11). Figure 12a plots the surface measurements from May 28 along with mixing curves derived from both the May 28 (upper curve) and May 29 (lower curve) soundings. Both mixing curves are well represented by straight lines. Although the frost point measurements are only good down to about −25°C, the upper curve extrapolates to ∼115 ppbv O3 at a frost point of −43°C, the value measured at 3.0 km ASL two hours later by the National Weather Service 0000 UT May 29 San Diego sounding. This extrapolation agrees well with the measured O3 in Figure 10b. The lower curve extrapolates to ∼95 ppbv, in reasonable agreement with the May 29 sounding (Figure 10c). The difference between the two soundings suggests that the May 28 sounding intercepted a filament of air from higher in the stratosphere than the May 29 sounding. Most of the surface measurements fall on or near the lower mixing curve, between the ozonesonde measurements and the Vandenberg measurements, which represent the marine boundary layer. The measurements from Joshua Tree range between the two mixing curves, consistent with the localized short-lived O3 peak at Joshua Tree in Figure 11a, and the apparent small-scale of the higher O3 filament sampled by the ozonesonde. Although the Riverside measurements also are 5 to 10 ppbv above the May 29 mixing line, the lack of a correlation with dew point suggests that local photochemistry is responsible.
Figure 12b plots the surface measurements from May 29. All of the measurements from sites outside the Basin now lie on the mixing line, consistent with entrainment of stratospheric air into the boundary layer over a large area from Vandenberg to Joshua Tree. Only the measurements from Glendora, and the late afternoon measurements from Thousand Oaks and Riverside deviate significantly from the mixing line. These measurements lie above the line and show little dependence on dew point. The vertical span of the Glendora measurements suggests that ∼35 ppbv of locally produced O3 was added to the stratospheric background of ∼50 ppbv. The 1000–1100 PST measurements at Goleta lie below the line suggesting that the boundary layer near the coast was not yet well mixed.
Figure 12c plots the surface O3 and dew point measurements for May 30. Since there was no Joshua Tree sounding, the measurements from the May 29 and May 31 soundings are shown. The surface measurements from May 30 initially lie near the mixing line defined by the May 29 sounding, but shift up and to the right in the early afternoon as photochemical activity increases and locally produced O3 is added to the stratospheric background. This transition is also seen in the time series measurements from Figure 11, with delayed O3 peaks at Crestline and Joshua Tree consistent with transport of locally produced O3 from the Basin. The end of the episode is also evident in the curve defined by the May 31 sounding, which is reversed in slope compared to the May 28 and 29 soundings and the highest O3 in the boundary layer.
 The O3-Td mixing plots can also be used to quantify the net contribution of the stratospheric intrusion on surface O3. Figure 13 shows O3-Td scatterplots for the four sites from Figure 7. The open circles in each plot show all of the daytime (0800–1800 PST) measurements during IONS-2010. The filled circles highlight the measurements over the indicated 8-h daytime window on May 29, and the triangles the same 8-h window on May 30. In each case, the filled circles lie on or near the mixing line from the May 29 Joshua Tree sounding, suggesting that these measurements represent air mixed down to the surface from the stratospheric intrusion, with little modification by local influences. The measurements from May 30 also lie on the mixing line in the morning, but deviate upwards as O3 produced in the Basin is transported to the site. The difference between these measurements and those on the mixing line corresponds to this local contribution, which ranges from ∼20 ppbv at Goleta to ∼30 ppbv at Thousand Oaks, Santa Clarita, and Crestline.
 The dotted horizontal line in each panel of Figure 13shows the 8-h average O3for the May 29 measurements, and the dashed line shows the mean MDA8 for the entire IONS-2010 measurement period. The mean MDA8 does not necessarily correspond exactly to the average of the plotted open circles since the latter include measurements from a 12-h window. For Goleta, the MDA8 on May 29 is nearly 7 ppbv greater than the mean, showing that the stratospheric intrusion increased surface O3compared to the IONS-2010 average. At Crestline, on the other hand, the MDA8 on May 29 was ∼11 ppbv less than the IONS-2010 mean value, which on most days is heavily influenced by O3 of local origin. These differences are consistent with the positive and negative PC2 loadings in Figure 8b. The measurements from Thousand Oaks and Santa Clarita lie much closer to the mean, consistent with the smaller PC2 amplitudes at these sites.
 The springtime air quality in Los Angeles County has improved considerably in recent years (http://www.arb.ca.gov/adam/trends) with the number of days in May and June having MDA8 O3 concentrations exceeding 75 ppbv decreasing from 44 (72%) in 1992 to 26 (43%) in 2009. Within the entire South Coast Air Basin, the corresponding number of exceedance days has decreased from 54 to 33. The number of exceedance days in April decreased from 24 (80%) to 4 (13%) in Los Angeles County over the same period, and from 25 to 5 in the South Coast Air Basin. The peak MDA8 O3 reported in Los Angeles County during May and June has decreased from an average of 94 ppbv in 1992 to 69 ppbv in 2009. The latter value lies below the current state and national standards of 70 and 75 ppbv, respectively, so that only relatively small additions of O3 from the stratosphere or transported from Asia can push the MDA8 values over the NAAQS threshold and create an exceedance. This will be even more likely if the NAAQS is lowered to 65 ppbv, which is less than the mean (May–June) MDA8 of 69 ppbv in 2009.
 The cold fronts and deep stratospheric intrusions that passed through southern California at the end of May 2010 affected surface O3 concentrations across a wide area. The stratospheric intrusion directly contributed to one exceedance of the current state standard of 70 ppbv on May 28, with an MDA8 of 72 ppbv recorded at Joshua Tree National Monument. Most stations, however, measured lower than normal O3 concentrations due to the decrease in photochemical production following the cold front. Photochemical production was still suppressed on May 29, and only two stations exceeded the state standard with 71 and 73 ppbv, respectively at Glendora and Upland. The MDA8 at five stations was greater than 65 ppbv, however, which might have exceeded a revised national standard.
 These two exceedances on May 29 required a local O3 contribution of less than 20 ppbv since entrainment of the stratospheric intrusion into the boundary layer had increased the local background O3 to ∼55 ppbv (see Figure 13), or 10–15 ppbv higher than the baseline concentrations in the marine boundary layer air transported ashore from the Pacific Ocean (Figure 7). These elevated background levels were still present the following day (May 30) when ambient temperatures returned to normal and local photochemical production of O3 increased significantly (see Figure 13), leading to MDA8 values in excess of 75 ppbv at eight stations, 70 ppbv at eleven more, and 65 ppbv at eleven more. None of the exceedances would have occurred if the background had been only 8 ppbv lower, demonstrating that even modest episodic contributions from the stratosphere or Asia can contribute to exceedances of the 8-h NAAQS in the Los Angeles area during late spring. The stratospheric influx will have less impact on air quality standards during winter when photochemical production is slow, even though stratospheric intrusions are more frequent and often larger [James et al., 2003; Sprenger and Wernli, 2003; Stohl et al., 2003a, 2003b]. The stratospheric contribution will also be less significant in summer, when local photochemical production is sufficient to cause exceedances on a near daily basis.
 The authors would like to thank David Parrish for his helpful comments on the manuscript. We also thank Ann Weickmann, Scott Sandberg, Rich Marchbanks, Bob Banta, and Alan Brewer for their help with the airborne O3lidar instrument. We would like to acknowledge support from the NOAA Health of the Atmosphere and Climate Programs and the flight crew of the NOAA Twin Otter research aircraft. We also thank Emrys Hall and Allen Jordan at NOAA ESRL GMD for the use of the Skysonde software and for their assistance in processing the Joshua Tree ozonesondes, and Luke Sabala and Victoria Chang of the Joshua Tree National Park. The NOAA ESRL Health of the Atmosphere Program, NASA Tropospheric Chemistry Program, U. S. Navy, Environment Canada, and NOAA's National Air Quality Forecast Capability provided funding for the IONS-2010 field campaign. The European Commission, Joint Research Centre (JRC)/Netherlands Environmental Assessment Agency (PBL) provided the EDGARv4.1 global NOx emissions inventory. The global land cover data set, as well as the MODIS fire detection data, was provided by the University of Maryland from their ftp server.