3.1. Trinidad Head Baseline Ozone Climatology and Comparison to Inland Sites
 To place the IONS-2010 ozonesonde measurements into context we first describe the typical seasonal ozone distribution across western North America, with emphasis on the baseline ozone coming ashore at Trinidad Head, California. Most of the analysis will focus on ozone between 0.0–6.0 km a.s.l. because air entering western North America above 6 km has relatively little impact on the surface of western North America, as will be discussed later. Additional focus will be placed on the 0.0–3.0 km a.s.l. layer which in most instances includes the daytime mixed layer.
 Since 2004 five sites in western North America have provided consistent weekly ozonesonde profiles, with additional spring and summertime ozonesondes launched during the various IONS campaigns. Three of the sites (Figure 1b) are located in southwestern Canada (KE, Stonyplain (SP), Bratt's Lake (BL)), while the two U.S. sites are Boulder (BO) and TH. SP is located in a quasi-rural area 30 km west of Edmonton, Alberta. BL is in a rural area of southern Saskatchewan, just south of Regina. Of the five western North America ozonesonde sites, Boulder, Colorado is exposed to the most pollution as it is situated in the urbanized region of the Northern Colorado Front Range, which exceeds the United States NAAQS for ozone (information available at http://www.epa.gov/region8/air/denverozone.html). TH, located on the coast of northern California, is the only operational ozonesonde site that is ideally situated to measure the baseline ozone flowing into western North America. In this portion of the analysis we assume that TH is representative of baseline ozone mixing ratios, and thus we have not filtered the data to remove measurements associated with recent transport from North American emission sources. As will be shown in Section 3.3, this assumption is valid.
 While we compare TH to the inland sites to gain some understanding of ozone's spatial variation across the continent, TH is only partially representative of the wide range of baseline air masses that influence the inland sites. Therefore comparison of the mass of ozone above TH to an inland site does not necessarily indicate the additional quantity of ozone produced from the time a baseline air mass entered North America until it reached the inland site. Caution is also required when comparing boundary layer ozone at TH to some inland sites. Unlike the IONS-2010 launches, which occurred between 14:00 and 17:00 local time, ozonesondes from TH and BO are generally launched around midday, while SP sondes are launched around 4:30 local time, KE sondes are launched around 15:30 local time, and BL sondes are launched around 10:00 local time. Pre-dawn launches (i.e., SP) have a stronger, low altitude influence from ozone surface deposition due to stable conditions. In contrast, afternoon launches occur during the peak time of day for photochemical ozone production, and they encounter a deeper boundary layer that has greater potential to entrain free tropospheric ozone.
 Figure 1a shows monthly median ozone profiles above TH, with surface ozone peaking in March–April and mid-tropospheric ozone exhibiting a broad maximum from April through August. Focusing first on the May–June period, median ozone above TH increases from 27 ppbv at the surface to 43 ppbv at 1 km, 54 ppbv at 3 km and 64 ppbv at 6 km (Figure 1c). The increase of ozone with altitude is similar at the other North American sites, all of which show substantial enhancements in comparison to tropical Hilo, Hawaii (HI), which is shown to indicate the lower quantity of ozone in the tropics compared to midlatitudes. The most notable difference among the five western North America sites is the ozone enhancement in the lower troposphere above BO. With a surface elevation of 1.6 km a.s.l., BO's surface boundary layer is elevated compared to the other sites, and deep vertical mixing in late spring and summer can push the top of the daytime boundary layer as high as 5 km a.s.l. This analysis does not explore the origin of the lower tropospheric ozone enhancement above BO, but it is likely due to a combination of photochemical ozone production in the boundary layer (from either anthropogenic or biomass burning sources [Jaffe, 2011]) and downward mixing of ozone from aloft.
 Plotting the ozone profiles according to altitude above ground level (Figure 1d) shows that ozone in the 0–2 km layer above TH is similar to ozone above SP (except for deposition in the pre-dawn surface layer) and BL, while KE and BO show surface ozone enhancements relative to TH of 16 and 26 ppbv, respectively. We consulted retroplumes calculated for these sites from a previous study [Cooper et al., 2010] to see if transport from the stratosphere could explain the ozone enhancements at KE and BO compared to TH. During springtime 5%, 6% and 8% of the mass of surface air above TH, KE and BO, respectively, originates in the stratosphere over the previous 15 days. The additional stratospheric influence at KE and BO can only explain a few ppbv of the excess ozone; therefore most of the ozone increase is due to either boundary layer photochemistry or downward transport from the free troposphere.
 Due to the variation in terrain and vertical transport in the lower troposphere across western North America it is difficult to directly compare ozone mixing ratios at particular altitudes among the five monitoring sites. Instead we compare the total mass of ozone between sites for a particular layer of the atmosphere. We begin by identifying the pressure surfaces that on average correspond to a layer of interest, for instance, the 0–6 km layer a.s.l. generally corresponds to the 1025–475 hPa layer. The quantity of ozone in the 1025–475 hPa layer for each ozonesonde profile is summed in Dobson Units (DU), where one DU equals 2.69 × 1016 ozone molecules per square centimeter. To account for elevation differences between sites and variation in surface pressure, each ozone sum is normalized by dividing by the thickness (hPa) of the actual depth of the atmosphere over which ozone measurements were made between 1025 and 475 hPa. Finally the quantity of ozone is reported in units of DU 100hPa−1 which allows the average quantity of ozone in a given layer to be compared between sites. Typical ozone values in the lower troposphere are 2–7 DU 100hPa−1, whereas the full tropospheric ozone column at midlatitudes in springtime falls within the range of 20–60 DU [Newchurch et al., 2003].
 Table 2 shows that the quantity of ozone above TH in the 0.0–6.0 km a.s.l. layer is similar to KE and SP, while BL has 6% more ozone and BO has 20% more ozone. Between 3.0 and 6.0 km a.s.l. there is no statistically significant difference in ozone between TH and the inland sites, as determined by analysis of variance tests. In the 0.0–3.0 km a.s.l. layer BO has 41% more ozone than TH while the remaining inland sites are not significantly different from TH.
Table 2. Quantity of Ozone (DU 100hPa−1) Above Each Long-Term Monitoring Site for Various Layers of the Atmosphere, May–June, 2004–2009a
|Site||Sample Size||5th Percentile||33rd Percentile||50th Percentile||67th Percentile||95th Percentile||Percent Difference From TH||ANOVA p-Value|
|0.0–6.0 km Above Sea Level (1025–475 hPa)|
|3.0–6.0 km Above Sea Level (700–475 hPa)|
|0.0–3.0 km Above Sea Level (1025–700 hPa)|
 So far we have only examined the May–June period which coincides with the IONS-2010 time frame. Figure 2 shows median ozone profiles above the ozonesonde sites for all four seasons. In most cases ozone differences between TH and the inland sites are not large and here we only report differences that are statistically significant. In the 3–6 km range SP differs from TH, with ozone consistently lower by −4%, −5%, −14% and −11% in winter, spring, summer and autumn, respectively. KE is only significantly different (−7%) from TH in summer. Regarding the 0–3 km layer, in autumn and winter all four inland sites have less ozone than TH, ranging from −5 to −17%. During spring ozone differences are +15% at Boulder and −6% at SP, and during summer ozone differences are +57% at Boulder and −11% at SP.
Figure 2. Same as Figure 1 but for (a) Dec-Jan-Feb, (b) Mar-Apr-May, (c) Jun-Jul-Aug, and (d) Sep-Oct-Nov, showing Trinidad head (orange), Boulder (blue), Kelowna (green), Stonyplain (light blue), Bratt's Lake (red) and Hilo (black).
Download figure to PowerPoint
3.2. Influence of Weather Patterns on Ozone During IONS-2010
 Baseline ozone during 2010 was affected by the weather patterns in the eastern North Pacific Ocean during May and June. NCEP Reanalysis [Kalnay et al., 1996] of the 1980–2010 average 500 hPa geopotential height shows that the May–June period is typically distinguished by a weak upper-level ridge over the eastern Pacific Ocean, a trough over California and a ridge over the central USA. This upper-level pattern is accompanied by a strong surface anticyclone over the midlatitude eastern North Pacific Ocean which produces northwesterly surface winds along the U.S. west coast. This same general pattern occurred in May–June 2010 (Figure 3a), except that the ridge-trough pattern over the eastern North Pacific Ocean and western USA was amplified, especially in May when several very deep upper level troughs traversed California. The strengthened ridge over the eastern North Pacific Ocean produced a stronger surface anticyclone, with a tighter pressure gradient between the center of the anticyclone and the California coast, resulting in stronger surface northwesterly winds. The stronger anticyclonic conditions may also have led to increased subsidence during the study period. These upper-level and surface patterns are just the average conditions, with day-to-day weather continually changing as upper-level troughs and their associated midlatitude cyclones impacted the western USA every several days.
Figure 3. (a) May–June 2010 average 500 hPa geopotential height and mean sea level pressure (hPa). (b) Ozone distributions above Trinidad Head for May–June 2004–2009 (black) and May–June 2010 (white) showing from left to right: 5th, 33rd, 50th, 67th and 95th ozone percentiles.
Download figure to PowerPoint
 The enhanced upper-level activity in May–June 2010 affected temperature and precipitation across California. According to the NOAA National Climatic Data Center (see http://www.ncdc.noaa.gov/sotc/) May 2010 was one of the coldest on record for California ranking 13th out of the 116 year (1895–2010) data record, with temperatures below normal in southern California and much below normal in the Central Valley and Northern California. Precipitation during May was above normal in Northern California and below normal in Southern California. In contrast June saw near normal temperatures along the California coast, above normal temperatures across the rest of the state and the driest June on record in southeastern California.
 Further effects of the enhanced upper-level activity in May–June 2010 were the strong ozone enhancements above 7 km, caused by a lower than average tropopause and strong stratospheric ozone intrusions. Using an ozone threshold value of 100 ppbv as an indicator of the tropopause, Figure 3b shows that the tropopause was more than 1 km lower in 2010 than average. Accordingly, the mass of ozone in the 7–10 km range was 39% above average. In contrast there were no statistically significant ozone differences between 2010 and the climatology in the 0–3 km and the 3–6 km layers. While weather conditions in May–June 2010 were more active than average and had a strong impact on baseline ozone in the upper troposphere, they had a relatively small impact on the overall quantity of baseline ozone coming ashore below 6 km.
3.3. Baseline Ozone During IONS-2010
 Figure 4 shows the locations of the seven IONS-2010 ozonesonde sites. TH, RY, PS and SN (transect 1 in Figure 4b) are designed to provide baseline ozone measurements along the California coast. The other three sites, KE, SH and JT are inland and provide indications of the additional amount of ozone produced in western North America. For average conditions, upper tropospheric ozone at northern midlatitudes has a latitudinal gradient, such that concentration decreases from north to south [Logan, 1999]. The IONS-2010 data show that the gradient was very strong during the study period, decreasing from ∼100 ppbv above KE to ∼50 ppbv above SN at 9 km a.s.l. The gradient is due to a stronger influence from a lower tropopause and more frequent stratospheric intrusions at higher latitudes, as well as a longer ozone lifetime at higher latitudes, while low latitudes see greater influence from low-ozone tropical air masses. The low average ozone values in the upper troposphere above SN may be biased: more of the ozonesonde failures occurred toward the end of the experiment when air masses had a high latitude origin. This would induce a bias toward low-ozone tropical air masses. Below 4 km there is little difference among the median ozone profiles with only JT showing strong ozone enhancements above baseline. Among the four coastal sites, there is a latitudinal gradient of ozone below 1 km, with PS and SN having 13% and 26% more ozone than TH, respectively (a statistically significant difference based on the total mass of ozone between 1025 and 900 hPa); RY has more ozone than TH by an insignificant 5%.
Figure 4. (a) Median ozone profiles above the IONS-2010 ozonesonde sites using all available profiles. Line colors correspond to the site label colors in Figure 4b. (b) Locations of the seven IONS-2010 ozonesonde sites. Also shown are the NOAA P3 sampling locations (blue dots) of the measurements used in the Central Valley and LA Basin ozone composite profiles. Gray transects indicate locations of the three ozone vertical cross sections described in the text: 1) coastal, 2) inland, and 3) southern California.
Download figure to PowerPoint
 We next analyze the four coastal sites to determine if they are truly representative of baseline ozone. To qualify as a baseline site the ozone measurements must be unaffected by local emissions. In addition, because we are interested in baseline ozone for all of California and the western USA in general, we also want to know if the sites are affected by pollution plumes originating in the western USA and southwestern Canada.
 Regarding the impact of local emissions, the ozone sites were not equipped with instrumentation that could detect fresh emissions, such as NOx or CO. While these sites are located within a few hundred meters of the shore in regions with low population density they are not completely isolated from local emissions and some influence from local sources is possible. For example TH is located between the shore and the western edge of a small town, RY is between the shore and the main tourist road through Point Reyes National Seashore, PS is between the shore and California Highway 1, and SN is located upwind of a small U.S. Navy airport on remote San Nicolas Island.
 Parrish et al.  and Goldstein et al.  conducted detailed analyses of surface baseline ozone at Trinidad Head Observatory (THO) using measurements of air pollutants from spring 2002. THO is situated at 107 m a.s.l. on top of a rocky promontory extending 500 m into the North Pacific Ocean and located 700 m southwest of the TH ozonesonde launch site. The presence of a diurnal cycle of ozone at THO with a morning minimum and afternoon maximum indicates an influence from North America. Filtering the data for continental influence using trace gas measurements and wind speed and direction shows that the main continental impact occurs in the early morning with a minimum impact during late afternoon when the sea breeze dominates coastal transport. During mid-afternoon, the approximate time of the IONS-2010 ozonesonde launches, Parrish et al.  found no difference in ozone when selecting just those measurements with strong northwesterly winds, which represent baseline air masses. These results indicate that any influence from the low levels of local anthropogenic emissions at the four coastal sites should be minimized at the mid-afternoon launch times of the ozonesondes.
 In the absence of in situ trace gas and particulate matter measurements for identifying locally or regionally emitted pollutants, we use FLEXPART with GFS winds to simulate the transport of an anthropogenic North American NOx tracer. The median value of this passive NOx tracer in the lowest km of the atmosphere above the four coastal sites is 280 pptv. Any ozone measurement with a tracer value above this threshold was flagged as having a strong influence from western North America emissions and removed from the analysis. Figure 5 compares the median ozone profiles above the four coastal sites when all available data are used, to the median profiles that have North American pollution events removed. This filtering removed roughly half of the data below 3 km, but resulted in no statistically significant change in the mass of ozone (±8%) in the 0.0–3.0 km (1025 – 700 hPa) range, as determined by analysis of variance tests. We conclude that regional western North America pollution plumes above the California coast are primarily found below 3 km but have no measureable impact on average ozone at these sites.
Figure 5. (a) Median ozone profiles above the four coastal sites using all available IONS-2010 data; sample sizes shown at right. (b) Same as Figure 5a, but measurements with more than 280 pptv North American anthropogenic NOx tracer have been removed. Numbers in white indicate the percent change in the mass of ozone in the 0–3 km range when air masses with strong North American influence are removed. None of these changes are statistically significant.
Download figure to PowerPoint
 We also considered the influence of biomass burning (BB) plumes on the four coastal sites. Using the Global Fire Emissions Database, version 3 (GFED3) [van der Werf et al., 2010] we compiled the 2000–2009 monthly fire emissions for the western USA (California, Oregon, Washington, Nevada and western Idaho). GFED3 shows a strong seasonal cycle with very low emissions during November–May, and an active summertime fire season that peaks in August. Fire NOx emissions in August range from 3% to 35% of anthropogenic NOx emissions (EDGAR 4.1) across the region, with an average value of 14%. Emissions in May and June are much lower averaging just 0.2% and 1.6%, respectively, of anthropogenic NOx emissions. Fire emissions for May and June 2010 have not yet been compiled by GFED3, but the NOAA National Climatic Data Center (see http://www.ncdc.noaa.gov/sotc/fire/2010/5) reports that the total area burned in the United States in May and June 2010 was 2% below average with no major fires reported during this period within our western U.S. study region. Burn area statistics of large California fires (>300 acres or 1.2 km2) compiled by the California Department of Forestry and Fire Protection (available at http://www.fire.ca.gov/fire_protection/fire_protection_fire_info_redbooks.php) show that the total area burned during May 10-June 19, 2010 (all fires were grassland and all occurred south of San Francisco) was 33 km2, about average for this time of year, with the 2005–2009 burn areas ranging from 16 to 350 km2. In good agreement with the average May–June GFED3 emissions, the mean FLEXPART BB NOx tracer for the IONS-2010 period in the 0–1 km range above the four coastal sites was 1% of the FLEXPART North American anthropogenic NOx tracer. In the 0–3 km range the mean BB NOx tracer was 3% of the North American anthropogenic NOx tracer. Removing coastal ozone measurements associated with relatively high levels of the BB NOx tracer produced no statistically significant changes in ozone in the 0–3 km range. We conclude that during May–June 2010, biomass burning made a minor contribution to the total North America NOx emissions from all combustion sources, and that any plumes that passed over the coastal sites made no measurable impact on the average ozone measured by the ozonesondes.
 Another potential impact on baseline ozone in the lowest few hundred meters of the atmosphere is surface deposition as the air masses come ashore and enter the turbulent daytime continental boundary layer. For example, Oltmans et al.  have shown that during springtime, afternoon surface ozone measured at THO is slightly lower when the air has a North American origin than a marine origin. To address this issue we compare TH ozonesonde measurements below 200 m to concurrent surface ozone measurements at THO. Whereas THO is on a promontory and widely exposed to marine air, the TH ozonesonde launch site is on the edge of a cove and therefore baseline air masses impacting this site have a higher probability of coming into contact with land as they enter the continental boundary layer. Ozone at the TH ozonesonde launch site during May–June 2002–2009 averaged 29.6 ppbv, 8% less than the average value of 32.0 ppbv at THO. An analysis of variance test indicates that the difference between the two data sets is not statistically significant at the 95% confidence limit. However, in May–June 2010, the ozonesonde average of 31.7 ppbv was 9% less than the average value of 34.8 ppbv at THO, a statistically significant difference. These differences are not due to ozonesonde measurement bias as the instruments have been thoroughly compared to the type of UV ozone analyzer used at THO [Johnson et al., 2002]. Instead, this analysis suggests that even when ozonesonde sites are situated on the coast and exposed to onshore flow, influence from the turbulent continental boundary layer (surface deposition and/or local NOx emissions) may decrease baseline ozone in the lowest few hundred meters of the atmosphere.
 The baseline sites of TH, RY, PS and SN are all situated on the coast, with very low local emissions and mid-afternoon exposure to marine air masses. The removal of measurements with a greater likelihood of influence from North American pollution made no statistically significant difference in the quantity of ozone measured below 3 km. Therefore for the remainder of this analysis we treat the median ozone profiles above these sites as representative of baseline ozone along the U.S. west coast. However, we cannot rule out the possibility that surface deposition may have slightly reduced ozone in the lowest few hundred meters of the atmosphere as the marine air masses entered the turbulent continental boundary layer. The only way to address this potential continental effect in future studies is to launch the ozonesondes from ships anchored beyond the influence of the continental boundary layer.
3.4. Transport Pathways of Baseline Air Masses
 Tropospheric baseline ozone along the California coast is a complicated mixture of ozone produced over time scales ranging from the previous few minutes to over thirty days earlier, and from ozone precursors emitted from nearby ships or distant sources such as Asia or Europe. Some of the ozone may even have been produced in the North American boundary layer days earlier and subsequently circled the globe. Baseline ozone in this region is also strongly affected by transport from the stratosphere. Source attribution of baseline ozone along the California coast requires a detailed photochemical modeling analysis using a global chemical transport model, an effort beyond the scope of this study. However, we can use the FLEXPART particle dispersion model to describe the transport pathways of baseline air masses, and to identify the ozone precursor emission regions that are most likely to impact ozone mixing ratios along the coast.
 Figure 6 shows the average transport pathway that baseline air masses take on their way to each of the four coastal sites, over the previous 20 days. Figure 6a shows the average of the FLEXPART retroplumes released from every 200 m layer from every profile between 0 and 6 km a.s.l., with the retroplumes tracked regardless of their altitude. These panels differ from clustered plots of single backward trajectories because they account for the dispersion of the air masses as they are advected backward in time and place greater emphasis on the broad upwind source regions. Overall Figure 6a shows that the baseline air masses travel across the North Pacific Ocean, with the more northern sites having a greater influence from high latitude regions, and the more southern sites having more influence from low latitude regions. Figure 6b examines the same retroplumes but only shows where the retroplumes intersect the footprint layer, the lowest 300 m of the atmosphere which contains all Northern Hemisphere anthropogenic emission sources, except for aircraft. While most of the influence is from the marine boundary layer (MBL) of the North Pacific Ocean, all sites show transport from eastern Asia, with the southern sites having greater transport from Southeast Asia and southern China, and the northern sites having greater transport from Japan/Korea, northern China and Europe. All four sites also show some recent influence from western North America as all profiles were retained for this analysis, but as described in Section 3.3, these air masses had no significant impact on the average ozone values.
Figure 6. Maps show the average FLEXPART retroplumes released from each of the four coastal sites between 0–6 km a.s.l. Colors indicate the average residence time (weighted by the specific volume) of the retroplumes in each grid cell, integrated over both (a) the full atmospheric column and (b) the footprint layer (lowest 300 m of the atmosphere). Numbers (white) indicate the number of retroplumes used in each average plot. Magenta boxes in the footprint plots encompass the source regions of the anthropogenic NOx tracers shown in Figure 6d. (c) Median vertical profiles above each coastal site of ozone, sample size and percent of air that came from the stratosphere over the previous 20 days. (d) Median vertical profiles above each coastal site of FLEXPART 20-day NOx tracers.
Download figure to PowerPoint
 Figure 6c shows the median ozone profile above the four sites. The lowest ozone values occur in the MBL. Figure 6d shows profiles of passive anthropogenic NOx tracers (calculated with the FLEXPART retroplumes) indicating those source regions that make the greatest contribution to variations in atmospheric composition above a baseline site. Above 3 km China is the strongest source of NOx tracer suggesting that this region would have the greatest impact on pollution import into the free troposphere above the western USA. While international shipping is responsible for approximately 13% of global anthropogenic NOx emissions [van het Bolscher et al., 2007] it is the second largest source of NOx tracer above 3 km along the California coast because ship emissions in the North Pacific have a closer proximity to California than upwind regions such as India or Europe. The remaining source regions examined in this study (Japan/Korea, Southeast Asia, India and Europe) make relatively small contributions to the total amount of NOx tracer above 3 km.
 The FLEXPART biomass burning tracer (not plotted in Figure 6) indicates that fires in North America and Asia were not a major source of NOx above the California coast during May–June 2010. Above the four coastal sites and between 0 and 8 km the BB NOx tracer was typically 1–5% of the total anthropogenic NOx tracer, averaged over the IONS-2010 period. The maximum amount of BB NOx tracer in terms of percentage of total anthropogenic NOx tracer was 7% at 6700 m above San Nicolas Island, averaged over the IONS-2010 period. Biomass burning emissions within most regions immediately upwind of North America such as China, Japan/Korea, India, and Europe are small compared to anthropogenic emissions [van der Werf et al., 2006; van het Bolscher et al., 2007; Dentener et al., 2011]. However, southeast Asia has widespread BB in spring with NOx emissions comparable to the region's anthropogenic emissions. Accordingly, the greatest values of the BB NOx tracer were found in the mid- and upper troposphere above SN, the location above the California coast with the strongest transport from SE Asia (Figure 6a).
 Below 2 km, international shipping emissions make the greatest contribution to total NOx tracer, with the lower troposphere total NOx values greatly exceeding those in the mid-troposphere. Despite the high levels of NOx tracer in the lower troposphere, baseline ozone in this region is only about half that found in the mid-troposphere, perhaps due to the low ozone production efficiency associated with ship emissions. While ship emissions produce ozone and dominate total NOx emissions in the lower troposphere upwind of the western USA, their net contribution to summertime surface ozone in the eastern North Pacific Ocean is approximately 5–25% [Eyring et al., 2007]. Ship plume studies have found that OH concentrations within the plumes lead to rapid removal of NOx, which limits ozone production [Chen et al., 2005]. In general the MBL is an environment that can typically lead to net ozone destruction as pollution plumes cross the oceans [Parrish et al., 1992; Real et al., 2008]. Despite the increase in MBL ozone along the U.S. west coast over the past 20 years [Parrish et al., 2009] and despite the large ship NOx emissions into these air masses, the photochemical environment of the North Pacific Ocean MBL is not conducive to strong net ozone production.
 While Figure 6 indicates the anthropogenic emission regions that should have the most influence on ozone transported to the western USA it does not reveal any obvious link between ozone and NOx emission regions due to the complicated number of ozone sources and sinks, ozone's variable lifetime throughout the troposphere, and the vertical and latitudinal variation of ozone production efficiency across the North Pacific Ocean [Zhang et al., 2008]. But some insight is gained by viewing the ozone distribution within the framework of isentropic transport. Figure 7a shows the baseline ozone curtain for just those measurements made within the troposphere, with the most notable feature being the ozone enhancement that slopes from the mid- and upper troposphere in the north to the lower troposphere in the south. NCEP Reanalysis potential temperature surfaces along the coast of western North America have a similar orientation (Figure 7b), constraining air masses to descending transport as they travel from north to south. Several earlier studies have noted how air masses influenced by stratospheric ozone [Jaeglé et al., 2003; Cooper et al., 2004b] or air pollution [Stohl et al., 2002; Brock et al., 2004; Cooper et al., 2004a; Hudman et al., 2004; Liang et al., 2004; Holzer and Hall, 2007; Zhang et al., 2008] descend over the eastern North Pacific Ocean due to the subsidence associated with the upper level ridges and surface anticyclones common to this region. Curtain plots of the quantity of air that originates in the stratosphere (Figure 7c) and the China anthropogenic NOx tracer (Figure 7d) follow the same general slope as the ozone, suggesting that descending stratospheric intrusions and Asian pollution plumes influence the ozone distribution along the California coast. These influences are not mutually exclusive as most of the major Asian pollution plumes identified by FLEXPART were combined with air masses with varying degrees of stratospheric origin, a transport process noted by earlier studies [Jaeglé et al., 2003; Cooper et al., 2004b; Liang et al., 2007].
Figure 7. (a) Median ozone profiles above the four coastal sites using only tropospheric measurements. Layers with less than 20 measurements are not shown. (b) Potential temperature surfaces along a north-south transect at 125° W. The locations of the four coastal ozonesonde sites are indicated with black vertical lines. (c) Median profiles of the percentage of air above each site transported from the stratosphere over the previous 20 days, corresponding to the ozone profiles in Figure 7a. (d) Median profiles of the quantity of anthropogenic NOx tracer transported from China over the previous 20 days.
Download figure to PowerPoint
 The sloping isentropic structure along the U.S. west coast is a climatological feature that is present each year, but varies interannually in its latitudinal location over a range of 5°. Therefore the potential exists each year for polluted and stratospheric air masses to descend from north to south along the U.S. west coast, a conclusion supported by analyses of the average transport pathways of polluted Asian air masses [Stohl et al., 2002; Forster et al., 2004] and stratospheric intrusions [Sprenger and Wernli, 2003; Cooper et al., 2004b] from other years. Based on this evidence it seems likely that the latitudinal ozone variation measured in May–June 2010 is also a common feature, with some variation in gradient and latitudinal location.
3.5. Impact Regions of Baseline Ozone
 The next step in our analysis of tropospheric baseline ozone is to quantify its impact on the surface of midlatitude North America. Using FLEXPART forward plumes calculated with GFS 0.5° × 0.5° wind fields we are able to estimate the quantity of ozone in each 200 m layer of each profile that is transported into the lowest 300 m of the atmosphere above the following receptor regions (Figure 8a): the tropical North Pacific Ocean, the eastern midlatitude North Pacific Ocean, California, the western USA (including California, southwestern Canada and northern Mexico) and the eastern USA (including southeastern Canada). In this exercise we treat ozone as a passive tracer with no surface deposition. Ozone transport is only calculated at the times and locations of the ozonesonde measurements (source), therefore the quantity of ozone transported to the receptor sites is in Dobson Units which scales with the mass of ozone at the source. Quantifying the contribution of baseline ozone to surface ozone in units of ppbv requires a chemical transport model and we are working with colleagues to produce such estimates using a high resolution (half-degree) global chemical transport model (M. Lin et al., Transport of Asian pollution into surface air over the western United States in spring, submitted to Journal of Geophysical Research, 2011).
Figure 8. (a) Domains of the five receptor regions used to calculate the quantity of ozone that reaches the 300 m thick layer adjacent to the Earth's surface. (b–e) (left) The ozone distribution above each of the four coastal sites in units of ppbv (blue), showing the 5th and 95th percentiles (dashed lines), 33rd and 67th percentiles (thin solid lines), and the 50th percentile (thick solid line). Also shown is the median ozone profile in Dobson units (red line). (middle) The average quantity of ozone (Dobson units) from each 200 m layer that reaches the surface of the various receptor regions. For each forward plume the quantity of ozone in each receptor is calculated for each of five days of transport, but only the maximum value is recorded. (right) The total amount of ozone from all levels between 0 and 6 km a.s.l. that reaches the surface (Dobson units per grid cell), averaged over all ozone profiles above a site.
Download figure to PowerPoint
 Figures 8b (left)–8e (left) show the median ozone profiles above the four coastal sites in units of ppbv (every 200 m), and DU in 200 m layers. Figures 8b (middle)–8e (middle) show the quantity of ozone (DU) in each 200 m layer that is transported into the lowest 300 m of the atmosphere above the various receptor sites. On average 8–10% of the mass of ozone in the 0–6 km range of the four coastal sites impacts the 300 m surface layer of the combined western and eastern USA regions, with the peak contribution coming from the lowest 2 km. Most of the impact is in the western USA with very little ozone reaching the surface of the eastern USA. For example, the quantity of ozone transported from TH to the surface of the eastern USA is only 7% of the quantity transported to the surface of the western USA. Other modeling studies have shown that background or Asian ozone that reaches the surface of the USA is roughly a factor of two greater in the western USA than the eastern USA [Jacob et al., 1999; Wang et al., 2009; Brown-Steiner and Hess, 2011], although these studies account for ozone transport from all directions and are not just limited to the transport pathway across California. There are two reasons for the limited transport of California baseline ozone into the eastern USA. First, during May and June 2010 the average winds in the lower troposphere advected air masses from the Gulf of Mexico across the eastern USA limiting influence from westerly transport, similar to the summertime conditions analyzed by Fiore et al. . Second, Asian polluted air masses make their greatest impact on the eastern USA when they descend isentropically behind cold fronts [Brown-Steiner and Hess, 2011]. Isentropes that pass over California are far more likely to intersect the high terrain of the western USA than the low elevations of the eastern USA [Brown-Steiner and Hess, 2011]. Instead, descending air masses behind cold fronts are better positioned to impact the surface of the eastern USA if they originate to the north or northwest. Accordingly we find that Kelowna, British Columbia, the most northern IONS-2010 site, delivers twice as much ozone to the surface of the eastern USA as TH (not shown).
 Another notable feature of Figure 8 is the large quantity of ozone above the baseline sites that descends into the eastern and tropical North Pacific Ocean and never reaches North America. Zhang et al.  noted a similar transport pattern from a modeling study of Asian pollution plumes transported to North America during spring 2006.
 Ozone transport to seven receptor regions within California, representing a variety of urban and rural areas was also quantified (Figure 9a). The receptor sites and their average elevation a.s.l., are as follows. 1) The north Central Valley (554 m) includes the city of Redding and the Shasta ozonesonde site, 2) the mid-Central Valley (132 m) covers San Francisco and Sacramento, while 3) the south Central Valley (297 m) includes Bakersfield. 4) The LA Basin receptor region (494 m) covers all the urbanized areas of the basin as well as some of the surrounding elevated terrain. Elevated rural areas with CASTNET (The Clean Air Status and Trends Network) ozone monitors are represented by receptor regions that include, 5) Lassen Volcanic National Park (1397 m), 6) Yosemite-Sequoia-Kings Canyon National Parks (2080 m), and 7) Joshua Tree National Park (724 m).
Figure 9. (a) Domains of the seven receptor regions used to calculate the quantity of ozone that reaches the 300 m thick layer adjacent to the California surface. Gray circles indicate the locations of the IONS-2010 ozonesonde sites. (b–e) Average quantity of ozone (Dobson units) from each 200 m layer above each of the four coastal sites that is transported to the surface within the seven receptor regions shown in Figure 9a. For each forward plume the quantity of ozone in each receptor is calculated for each of five days of transport, but only the maximum value is recorded. Line colors correspond to the region colors in Figure 9a.
Download figure to PowerPoint
 The quantity of ozone transported from each ozonesonde site to the seven receptor regions is shown in Figures 9b–9e, calculated with FLEXPART using 12 × 12 km WRF winds. The Dobson unit values are very low compared to Figure 8 due to the much smaller area of the California receptor sites. Focusing first on the baseline ozone measurements above 2 km a.s.l., very little ozone is transported to the low elevation receptor regions, instead the major impact is on the high elevation National Park sites. For example, Yosemite-Sequoia-Kings Canyon National Parks are especially sensitive to ozone coming ashore in the 2–4 km range above PS and RY where baseline ozone mixing ratios are typically 50–60 ppbv.
 Baseline ozone coming ashore below 2 km has its strongest impact on the low elevation receptor regions. Due to the prevailing northwesterly flow along the coast the greatest impact of TH is on the mid-Central Valley. It has less influence on the nearby north Central Valley due to the blocking influence of the California Coast Range with mountain peaks between 1000 and 2200 m a.s.l. Parrish et al.  found that surface ozone measurements in the northern Central Valley were most highly correlated with baseline ozone in the 1–2.5 km range above TH. They concluded that this altitude range was a more important source of baseline ozone for the northern Central Valley than the MBL which is mostly blocked by the Coast Range. Our FLEXPART simulation generally agrees with these conclusions except that the layer above TH with the greatest impact on the north Central Valley is lower, at 1.0–1.2 km a.s.l. WRF also indicates that MBL ozone can reach the northern Central Valley but the quantities are 25–43% less.
 RY is the best overall site for monitoring California's baseline ozone. The much lower elevation of the California Coast Range east of RY provides little impediment to the transport of MBL air inland. Once these low altitude air masses reach the Central Valley they can travel to its northern and southern ends [Bao et al., 2008]. As a result RY lies on an important pathway of ozone into the mid- and southern Central Valley, and even delivers more ozone to the surface of the northern Central Valley than TH. This pathway is also a substantial source of ozone for the elevated National Park regions and the LA Basin.
 PS, despite its proximity to the south Central Valley, delivers less ozone to this region than RY due to blocking by the adjacent Santa Lucia Mountains. PS is however a good baseline site for the LA Basin and Joshua Tree National Park, and it is the best baseline site for Yosemite-Sequoia-Kings Canyon National Parks. SN is so far south that the prevailing northwesterly flow during the study period advected most of the ozone to the south and southeast, largely avoiding California. We conclude that SN is the least effective of the four coastal sites for monitoring California's baseline ozone, although ozone from SN does impact Joshua Tree National Park, and the island would be a suitable baseline site for San Diego.
3.6. Comparison of Baseline Ozone to Inland Measurements
 The ozone measurements above the baseline sites can be compared to inland sites to yield a simple quantification of net photochemical ozone production in the lower troposphere of California. To the best of our knowledge this is the first time that a purely measurement-based quantification of net ozone production has been possible for such a large portion of the state and for the entire depth of the lower troposphere. To make the calculation we cannot simply assume that baseline ozone is transported from the coast to an inland location in a laminar fashion with no vertical transport. Model analysis by Huang et al.  demonstrates that baseline air masses in the lower free troposphere above TH can be transported to the surface of the northern Central Valley by downward mixing within the daytime boundary layer, while two other recent studies show that air masses with enhanced ozone in the lower free troposphere can be mixed into the daytime boundary layer over the Los Angeles Basin (J. A. Neuman et al., Ozone transport from the free troposphere to the Los Angeles basin, submitted to Journal of Geophysical Research, 2011; A. O. Langford et al., Stratospheric influence on surface ozone in the Los Angeles area during late spring and early summer of 2010, submitted to Journal of Geophysical Research, 2011). Similarly, Section 3.5 of the present study shows that baseline ozone at 2–4 km a.s.l. can be transported to the surface of the Central Valley or mountainous regions by the same process. Therefore we determine the altitude range that encompasses the vertical transport between the surface and the top of the continental daytime boundary layer. We then compare average baseline ozone to average inland ozone in this altitude range only, quantifying ozone in units of DU 100hPa−1 as described in Section 3.1. If the average quantity of ozone is greater above the inland sites than the baseline sites then we can quantify the net photochemical ozone production within California due to emissions from anthropogenic sources and wildfires (although fire emissions during the study period were relatively low).
 An especially interesting site for this type of analysis is SH, at the northern edge of the northern Central Valley, a region that exceeds the NAAQS for ozone despite its relatively low population and emissions. The high surface ozone measured in this region could have a strong influence from pollution transport from the stronger emission regions to the south or from free tropospheric baseline ozone that is mixed down to the surface. As described above, Parrish et al.  concluded that free tropospheric baseline ozone transported to the surface of the northern Central Valley explains most of this region's ozone variability. If this is the case, the ozonesondes above SH should provide evidence that ozone-rich air from the lower free troposphere is mixed down to the surface, while moisture-rich air from the surface is transported upwards. Furthermore the SH ozonesondes should show that the total mass of ozone below 3 km (the altitude rarely exceeded by the daytime mixed layer above SH) is similar to the mass of ozone above the upwind baseline sites. The FLEXART forward plumes indicate that both TH and RY are upwind of SH. We have no profiles north of TH to characterize baseline air that crosses the coast of Oregon en route to SH, and we assume lower tropospheric ozone above this region is similar to TH and RY. The total mass of ozone in the lower troposphere (0–3 km a.s.l) above SH is 8% greater than TH, and 5% greater than RY, statistically insignificant differences.
 Water vapor measurements can be used to identify the extent of vertical mixing in the daytime boundary layer. The northern Central Valley is a source of water vapor due to the presence of several large reservoirs, rivers, crop irrigation and transpiration from the surrounding forests. Surface measurements from May–June 2010 at Redding Municipal Airport (data archived at NOAA National Climatic Data Center) show that the water vapor mixing ratio reaches a maximum of 6.4 g/kg at 06:00 local time when the air is coolest and the surface temperature inversion is strongest. By 15:00 local time, when the daytime boundary layer is deepest the average water vapor mixing ratio decreases by 30% to 4.6 g/kg, consistent with upwards transport of water vapor. Figure 10 shows that SH has significantly more water vapor near 2 km than TH indicating vertical transport of water vapor from the surface to the upper levels of the daytime boundary layer. This evidence for vertical mixing suggests that air aloft is likely mixed to the surface bringing ozone with it. Figure 10 does show that ozone below 1.1 km is significantly greater at SH than TH which could be due to downward transport of ozone or local photochemical ozone production. The fact that the total quantity of ozone in the lower troposphere (0–3 km) does not increase from the baseline sites to SH, while near-surface ozone increases, implies that ozone must decrease at higher altitudes to compensate for the near-surface ozone increase. Vertical mixing within the daytime boundary layer could account for this ozone redistribution.
Figure 10. (a) Ozone distributions above TH (black) and SH (white), showing from left to right: 5th, 33rd, 50th, 67th and 95th ozone percentiles. Red circles indicate the layers in which the ozone distribution at Shasta is significantly different from Trinidad Head, based on analysis of variance tests and p values less than 0.05. (b) Same as Figure 10a but for water vapor.
Download figure to PowerPoint
 These results do not quantify the amount of free tropospheric ozone transported to the surface of the northern Central Valley, but they and the FLEXPART simulations lend support to the hypothesis of Parrish et al.  that air and ozone from the lower free troposphere can be transported to the surface of the northern Central Valley, however there are two caveats to consider. First, the effect of upslope transport adds a degree of uncertainty to the analysis. Image loops of GOES-WEST 1 km visible satellite images (not shown) reveal convective clouds are more likely to form above the mountains surrounding the northern Central Valley than over the Central Valley itself due to the influence of daytime upslope winds [Bao et al., 2008]. Because SH is at the northwest edge of the Central Valley just east of the California Coast Range it is affected by the upslope winds. A few of the ozonesonde profiles (ozone, potential temperature, dewpoint depression) suggest that on some days the upslope flow above SH transports surface air upwards but may not bring lower free tropospheric air to the surface. Second, Parrish et al.  found no evidence that MBL air is transported to the surface of the northern Central Valley, in contrast to the WRF results from the present study. While the 12 km WRF resolution captures much of the terrain variability across the western USA, it misses fine-scale mountain and valley features, especially across the highly variable terrain of the northern California Coast Range between TH and SH. To account for the possibility that transport errors in WRF allow too much MBL air into the northern Central Valley we compared ozone between TH and RY to SH in the following layers: 300–3000 m a.s.l., 500–3000 m a.s.l and 1000–2500 m a.s.l. In all cases we found no statistically significant difference in ozone between SH and the coastal sites.
 With these caveats in mind, the ozonesonde profiles from SH have improved our understanding of the vertical ozone distribution between the coast and the northern Central Valley, although further research is needed as described in the Conclusions. For average conditions during IONS-2010 we found that there was no significant increase in ozone above SH compared to the baseline sites, and conclude that photochemistry did not have a strong influence on the northernmost Central Valley during the study period. In contrast photochemistry did have a strong impact on ozone in the daytime boundary layer across central and southern California as discussed below.
 Next we estimate lower tropospheric ozone enhancements above central and southern California. For comparison to the baseline ozone curtain in Figure 5a, Figure 11a shows ozone down the center of California, stretching from JT in the south through the Central Valley and up to SH in the north (transect 2 in Figure 4b). Ozone measurements above SH and JT were made by the daily ozonesondes while measurements above the Central Valley are from several flights of the NOAA WP-3D (sampling locations indicated in Figure 4b) and shown as a single composite profile. Figure 11b shows a west-east ozone curtain though southern California (transect 3 in Figure 4b) using ozonesonde measurements from SN and JT, and NOAA WP-3D measurements above the LA Basin. The quantity of ozone in the 0–3.0 km a.s.l. (1025–700 hPa) range above the Central Valley exceeds ozone above the baseline sites of TH, RY and PS by 23%, 20% and 12% respectively. These ozone enhancements imply net ozone production above the Central Valley, as expected above a region with strong ozone precursor emissions.
Figure 11. (a) Median ozone profiles above three inland locations in California along the south-north transect shown in Figure 4. (b) Median ozone profiles along the west-east transect shown in Figure 4. Ozone measurements above Shasta (SH), San Nicolas Island (SN) and Joshua Tree (JT) were made by daily ozonesondes. Ozone measurements above the Central Valley and the LA Basin were made by the NOAA P3 aircraft on several flights; see text for details.
Download figure to PowerPoint
 Even greater ozone enhancements were found by the NOAA WP-3D above the LA Basin, with 0–3 km a.s.l. ozone 41%, 32% and 33% greater than RY, PS and SN, respectively. These calculated enhancements may underestimate the impact of photochemistry within the LA Basin because the chosen boundary layer depth of 0–3 km a.s.l. has to be deep enough to encompass most of the upslope transport along the mountains that rim the LA Basin, which at certain times and locations can even reach 4 km a.s.l. [Langford et al., 2010]. However, this boundary layer depth greatly overestimates the height of the boundary layer nearer the coast. Analysis of the NOAA WP-3D measurements over the LA Basin during CalNex shows that on the days when the aircraft sampled the midday boundary layer, most of the local pollution remained below 1.5 km (Neuman et al., submitted manuscript, 2011). Using a boundary layer depth of 0–1.5 km a.s.l. (1025–850 hPa) yields average daytime ozone enhancements over the LA Basin that are 63%, 51% and 39% greater than RY, PS and SN, respectively.
 JT has a much deeper daytime boundary layer than the other sites considered in this study, and we use a boundary layer depth of 0–4 km a.s.l. (1025–625 hPa) for comparison to baseline sites. Accordingly, ozone enhancements above JT are 41%, 34% and 33% greater than RY, PS and SN, respectively. While pollution emissions within Joshua Tree National Park and the surrounding desert regions are relatively low, this region regularly experiences poor air quality in summer due to pollution exported from the LA Basin [Sullivan et al., 2001; Rosenthal et al., 2003; Langford et al., 2010]. As a result the ozone transect through southern California (Figures 4b and 11b) shows a large ozone increase from SN to the LA Basin, but little difference between LA and JT, which have a similar quantity of ozone in the 0–4 km a.s.l. layer.
 The percent increases in ozone above the inland sites in comparison to the baseline sites are summarized in Table 3. These relationships can be inverted to express the quantity of ozone above the baseline sites as a percentage of the ozone above the inland sites (Table 4). From this viewpoint, using average daytime data, baseline ozone is equal to more than 80% of the ozone measured above the Central Valley and SH. Across the polluted regions of southern California, baseline ozone is equal to 63–76% of the measured ozone above JT and the LA Basin.
Table 3. Percent Increase in Ozone Above the Inland Sites in Comparison to the Baseline Sites ((100*inland-baseline)/baseline)a
| ||Baseline Sites|
|SH (0–3 km)||8 (insignificant)||5 (insignificant)||-||-|
|Central Valley (0–3 km)||23||20||12||-|
|JT (0–4 km)||-||41||34||33|
|LA Basin (0–3 km)||-||41||32||33|
|LA Basin (0–1.5 km)||-||63||51||39|
Table 4. Ozone Above the Baseline Sites, Expressed as a Percentage of the Ozone Above the Inland Sites (100*baseline/inland)a
| ||Baseline Sites|
|SH (0–3 km)||93||95||-||-|
|Central Valley (0–3 km)||81||83||89||-|
|JT (0–4 km)||-||71||75||75|
|LA Basin (0–3 km)||-||71||76||75|
|LA Basin (0–1.5 km)||-||63||68||73|