We examine the influence of daily variations in baseline ozone (O3) on urban air quality in the U.S. Pacific Northwest (PNW) during 2004 to 2010 through two analyses: (1) transport of free tropospheric (FT) O3 from Mount Bachelor Observatory (MBO) to Boise, Idaho; and (2) transport of marine boundary layer (MBL) O3 from Cheeka Peak (CP) to Enumclaw, Washington. Both Boise and Enumclaw experience days with maximum daily 8 hour averages of O3 (MDA8) exceeding U.S. standards. Backward trajectory cluster analyses identify days when FT and MBL O3 strongly influence MDA8 in Boise and Enumclaw. On these days, MBO and CP O3 observations explain 40% and 69% of the variations in Boise and Enumclaw MDA8, respectively. Bivariate regressions for Boise/MBO and Enumclaw/CP have slopes of 0.52 ± 0.16 and 1.04 ± 0.08, respectively, representing the differing interplay of O3 dilution, production, and loss during FT to boundary layer transport (Boise/MBO) and fast boundary layer transport (Enumclaw/CP). AIRPACT-3/CMAQ (Air Indicator Report for Public Access and Community Tracking version 3/Community Multi-scale Air Quality model) high-resolution air-quality simulation results demonstrate how transport of O3 from the FT above MBO contributes to elevated O3 at Boise. Average MDA8 O3 in Boise is higher than in Enumclaw due to site elevation and greater entrainment of FT air masses, a finding likely applicable to other PNW sites. Days with high baseline influence at Boise and Enumclaw have lower average MDA8 O3 than other days; however, some of these days would still exceed the U.S. standard if it is substantially tightened in 2013, highlighting the increasing importance of FT O3 influence on urban MDA8.
 Ground-level ozone (O3) is a regulated air pollutant with known impacts on human health and the environment. Both short-term and chronic exposure to O3 have been linked to reduced lung function, aggravation of respiratory conditions, and increased mortality [Bell et al., 2004; Lippmann, 1991]. Baseline O3 is defined by the Task Force on Hemispheric Transport of Air Pollution as the observed mixing ratio of tropospheric O3 at a site that is not influenced by locally emitted or produced pollution [Dentener et al., 2011]. There are large spatial and temporal variations in baseline O3, which are influenced by changes in upwind emission of O3 precursors, elevation, meteorological conditions, and photochemistry [National Research Council (NRC), 2009]. Studies at marine boundary layer (MBL) sites along the U.S. West Coast show a seasonal cycle in mean baseline O3, with a spring maximum and summer minimum [Oltmans et al., 2008; Parrish et al., 2009]. It is well documented that episodic events, including long-range transport of industrial and biomass-burning emissions and subsidence of air masses originating in the upper troposphere/lower stratosphere (UT/LS) can have a strong influence on day-to-day variations in baseline O3 in western North America [Ambrose et al., 2011; Cooper et al., 2010; Jaffe, 2011; Jaffe et al., 2004; Lefohn et al., 2001, 2011; Lin et al., 2012a, 2012b; Macdonald et al., 2011; Oltmans et al., 2010; Weiss-Penzias et al., 2006].
 The U.S. Environmental Protection Agency (EPA) regulates the allowable surface O3 mixing ratios with primary and secondary National Ambient Air Quality Standards (NAAQS), which focus on human health and public welfare, respectively. Both NAAQS are based on the maximum daily 8 hour average (MDA8) of O3, and currently set the design value as 75 parts per billion by volume (ppbv) for a 3 year running average of the annual fourth highest MDA8 [U.S. EPA, 2006]. In 2010, the EPA proposed to lower the allowable mixing ratio in the primary O3 NAAQS to 60 to 70 ppbv [U.S. EPA, 2010], but the rule revision was later withdrawn [The White House Office of the Press Secretary, 2011]. The next scheduled revision of the O3 NAAQS occurs in 2013, at which time the EPA may lower the O3 standard in line with the Clean Air Scientific Advisory Committee recommendations [Henderson, 2006].
 To facilitate analyses of background versus regional influences on surface O3, the EPA defines policy-relevant background (PRB) O3 in the United States as the surface mixing ratios in the absence of North American anthropogenic emissions [U.S. EPA, 2006], a definition that necessitates the use of atmospheric models [McDonald-Buller et al., 2011]. Modeling studies of U.S. PRB O3 demonstrate that seasonal average mixing ratios range between 15 and 50 ppbv [Fiore et al., 2002, 2003; Lin et al., 2012b; Wang et al., 2009; Zhang et al., 2011], with larger mixing ratios at higher elevation sites [Fiore et al., 2003; Zhang et al., 2011].
 Unlike PRB O3, which is a model construct, baseline O3 is the measurable quantity of O3 in air flowing into a region [NRC, 2009]. Studies have investigated spatial variations of O3 in the western United States by analyzing O3 transport from sites measuring baseline air masses to downwind surface sites. Baseline air is regularly observed at a number of western U.S. sites including the Mount Bachelor Observatory (MBO) mountaintop site in central Oregon [Ambrose et al., 2011; Fischer et al., 2010; Reidmiller et al., 2010; Weiss-Penzias et al., 2006]; ozonesonde launch sites at Trinidad Head (TH), California [Cooper et al., 2011; Parrish et al., 2010], and Boulder, Colorado [Jaffe, 2011]; and the MBL sites at TH [Parrish et al., 2010] and Cheeka Peak (CP), Washington [Jaffe et al., 2001; Weiss-Penzias et al., 2004]. Using ozonesondes from a network of sites along the California coast, Cooper et al.  found that baseline O3 observed within 0 to 6 km above ground level (agl) impacted U.S. O3 mixing ratios in the lowest 300 m of the atmosphere. The authors also found that within California, baseline O3 from above (below) 2 km generally influenced high-elevation (low-elevation) sites. Parrish et al.  identified a correlation between TH ozonesonde measurements and downwind low- and higher elevation surface sites, and found that the contribution from baseline O3 was high on days exceeding the 75 ppbv U.S. standard.
 Variations in free tropospheric (FT) O3 impact O3 mixing ratios in the boundary layer. The lifetime of O3 against chemical loss in the FT is on the order of weeks to months, facilitating intercontinental transport [Dentener et al., 2011]. Using in situ observations at MBO during 2004 to 2009, Ambrose et al.  identified 25 high O3 events (8 hour average O3 > 70 ppbv) in the Pacific Northwest (PNW) FT. The authors found that eight of these events were caused by subsidence of UT/LS air masses, four by Asian industrial pollution, and six by a combination of these sources. During daytime growth of the boundary layer, FT and mixed layer air masses are combined in the entrainment zone, which is approximately 10% to 40% the depth of the mixed layer [Crum and Stull, 1987]. The FT O3 entrained into the mixed layer adds to the O3 mixing ratio already in that layer [Berkowitz et al., 2000]. Therefore, the afternoon mixing layer combines the O3 originating in the entrained and local air masses [Cooper et al., 2011; Hudman et al., 2004; Zhang and Rao, 1999]. Jaffe  determined that surface O3 mixing ratios in the western United States were statistically correlated with FT ozonesonde measurements in spring, which the author attributed to flux from the FT to the surface. In addition, model studies have shown that long-range transport of Asian pollution plumes causes O3 enhancements at western U.S. surface sites in spring [Brown-Steiner and Hess, 2011; Hudman et al., 2004; Lin et al., 2012a; Zhang et al., 2008].
 Variations in FT O3 can contribute to surface mixing ratios that exceed the current U.S. standard of 75 ppbv. Using a high-resolution global model, Lin et al. [2012a] found that in the southwestern United States, more than half of the MDA8 exceeding the current O3 standard would not have occurred without some influence from Asian emissions. Similarly, observations of MDA8 greater than 75 ppbv in Colorado and California [Langford et al., 2009, 2012] and enhanced O3 in the northern United States [Lefohn et al., 2001] have been linked to subsidence of UT/LS air masses. Lin et al. [2012b] found that stratospheric intrusions from 13 events elevated background MDA8 O3 mixing ratios to 60 to 75 ppbv.
 Although a number of studies have focused on baseline O3 in the western United States, there is still uncertainty about the impact of daily variations in baseline O3 on surface O3 mixing ratios, particularly in urban areas. To build on the existing body of research, we analyze the transport of FT and MBL O3 observed at the MBO and CP sites, respectively. Our goal is to understand the range of O3 mixing ratios at urban, surface sites on days when there is a strong influence from air masses originating in the FT or the MBL, and whether the O3 mixing ratios on these days exceed current or potential future U.S. O3 standards. Section 3.1 focuses on transport of FT air masses observed at MBO to the boundary layer at Boise, Idaho. Section 3.2 focuses on transport of MBL air masses from the Washington coast to Enumclaw, Washington. Section 3.3 discusses the correlation between the baseline and urban O3 mixing ratios, and section 4.1 compares these correlations to another study. Section 4.2 describes the range of MDA8 O3 mixing ratios in Boise and Enumclaw on days with a strong baseline influence and their relationship to current and potential future U.S. O3 standards. The Boise and Enumclaw urban areas were chosen as the focus of this analysis because they are representative of two types of urban areas in the PNW. Boise is a higher-elevation inland site that generally has more afternoon surface warming during the warmer months than Enumclaw, which is a lower-elevation, coastal site.
2.1 Ozone Data Averaging
 This study analyzes baseline O3 data from two sites in the U.S. PNW: MBO, a mountaintop site in central Oregon; and CP, a remote coastal site in western Washington. Surface O3 mixing ratios from Boise, Idaho, and Enumclaw, Washington, are also analyzed. Eight-hour averages are used to compute daily O3 mixing ratios, and only averages calculated with at least five hourly data points are used for statistical analyses. Information on these sites and the O3 data analyzed are provided in Table 1, and a terrain map of the region surrounding the sites is provided in Figure 1.
2004–2005: Jaffe Group data archive; 2006–2010: PSCAAc
43.6°N, 116.2°W, 825 m asl
Same dates as MBO
Rick Hardy at Idaho DEQ
47.14°N, 121.94°W, 402 m asl
Same dates as CP
 Since 2004, MBO has been the site of near-continuous measurements of meteorological parameters, aerosol scattering, CO, O3, and other chemical species, in addition to intensive campaigns with a broader array of observations [Ambrose et al., 2011; Weiss-Penzias et al., 2006]. MBO is well positioned to measure FT air masses due to the site elevation, lack of local pollution sources, and tendency to measure air masses that were recently transported over the North Pacific Ocean [Weiss-Penzias et al., 2006]. Like other mountaintop sites, MBO experiences a diurnal cycle in most species due to upslope flow of mixed boundary layer air in the daytime and subsidence/downslope flow of FT air at night [Reidmiller et al., 2010; Weiss-Penzias et al., 2006]. Baseline O3 is best measured at MBO during the hours when downslope flow prevails. From soundings taken at MBO during 2008 using temperature, relative humidity, and pressure sensors attached to ski resort chairlifts, Reidmiller et al.  determined that upslope flow typically occurs between 5 hours after sunrise through 1 hour after sunset during spring.
 The CP monitoring station has been operated by the Olympic Region Clean Air Agency since 2004, and continuous O3 observations are available starting in 2006. In addition, O3 was measured at the site during research conducted by the University of Washington between 1997 and 2005 [Jaffe et al., 2001; Weiss-Penzias et al., 2004]. CP mainly observes clean MBL air, although episodes of pollution from ship traffic [Jaffe et al., 2001] and easterly flow of continental air masses [Weiss-Penzias et al., 2004] have been detected. Like other MBL sites in the western United States, baseline O3 is best measured at CP during the afternoon hours, when westerly air masses from the Pacific Ocean are entrained into the growing mixed layer, rather than during the night when stable atmospheric conditions inhibit mixing [Parrish et al., 2009].
 The Idaho Department of Environmental Quality operated several air-monitoring stations in Boise during 2004 to 2010, and hourly data at four stations are available starting in 2006. The correlation between the hourly O3 observations at these four monitoring stations ranges between r = 0.85 and 0.93 (p ≤ 0.01). Therefore, we averaged the available data to create a single hourly O3 data set. The Enumclaw air-monitoring station is operated by the Washington Department of Ecology. This station regularly measures high O3 mixing ratios due to pollution from the Seattle-Tacoma metropolitan corridor, and was in noncompliance with the O3 NAAQS for the 2006 to 2008 averaging period. Boise is a larger urban area than Enumclaw, and in 2010, it had a population of 205,671 compared to 10,699 in Enumclaw. Summing the nitrogen oxide (NOx) emissions from the 2008 Emissions Database for Global Atmospheric Research (EDGAR) v4.2 emissions inventory (http://edgar.jrc.ec.europa.eu), the Boise and Enumclaw grid cells had NOx emissions of 4.14e-11 and 4.83e-12 kg m−2 sec−1, respectively.
2.2 Hybrid Single-Particle Lagrangian Integrated Trajectory Model and Cluster Analyses
 Forward and backward air mass trajectories are used to examine transport between sites. Air mass trajectories were calculated using the National Oceanic and Atmospheric Administration Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model, version 4 [Draxler, 1999; Draxler and Hess, 1997, 1998] with Eta Data Assimilation System (EDAS) meteorological data. EDAS has a 3 hour time resolution, 40 km horizontal resolution, and a 50 millibar (mb) vertical resolution between the 50 and 1000 mb pressure levels. Most trajectory models are documented to have errors in horizontal distance of up to 30% [Stohl, 1998], and the HYSPLIT model is also sensitive to the vertical structure of wind data, particularly in the boundary layer [Draxler and Hess, 1998].
 A cluster analysis of trajectories, an objective methodology used in numerous air-quality studies (e.g., Brankov et al., 1998; Cape et al., 2000; Moody et al., 1995, and references therein), was used to identify days when there is air mass transport between paired baseline and urban sites. This study uses the cluster analysis method from the HYSPLIT model, version 4 [Draxler, 1999]. Each cluster analysis is based on a set of 24 hour backward trajectories initiated from one of the urban monitoring sites, using every fourth hourly trajectory point. The analysis iteratively compares the set of backward trajectories, clustering the two most similar trajectories on each iteration based on the sum of the squared Euclidian distances between the trajectories' hourly points, referred to as the total spatial variance. Following Draxler , we used the criteria of total spatial variance ≥ 30% for selecting the appropriate number of clusters for a given set of trajectories.
2.3 AIRPACT-3/CMAQ (Air Indicator Report for Public Access and Community Tracking Version 3/Community Multi-Scale Air Quality Model)
 The Air Indicator Report for Public Access and Community Tracking version 3 (AIRPACT-3) is a numerical air-quality forecast system for the PNW reporting to the public daily via the web [Herron-Thorpe et al., 2012]. The AIRPACT system relies on the Community Multi-scale Air Quality Model (CMAQ) v4.6 using Weather Research and Forecasting [WRF; Skamarock et al., 2005] meteorological forecast fields provided by colleagues at the University of Washington. The modeling domain is a grid of 95 columns (west to east) by 95 rows (south to north) of 12 × 12 km cells with 21 vertical layers of increasing thickness from the surface to 100 mb. The boundary conditions are derived from daily Model for OZone and Related Chemical Tracers (MOZART-4) forecasts, with assimilated Measurements of Pollution in The Troposphere (MOPITT) carbon monoxide, simulated at the National Center for Atmospheric Research (NCAR). The MOZART-4 global chemical forecasts are driven by National Center for Environmental Prediction Global Forecasting System forecast meteorology. MOZART-4 simulations at NCAR are based on the Precursors of Ozone and Their Effects in the Troposphere (POET) inventory for anthropogenic emissions [Emmons et al., 2010]. AIRPACT uses the Sparse Matrix Operator Kernel Emissions [SMOKE; Houyoux et al., 2012] emission processing system. Area and nonroad mobile emissions are based on the 2002 EPA National Emission Inventory adjusted to 2005 using the EPA's Economic Growth Analysis System software. On-road mobile emissions are generated using emission factors from the EPA MOBILE v6.2 model and state-specific activity data, and are adjusted for WRF forecast temperature. For this study, two AIRPACT/CMAQ simulations were conducted: (1) a full-chemistry simulation as described earlier, and (2) a simulation using MOZART boundary conditions and a nonreactive, tracer-only version of CMAQ 4.6.
2.4 Statistical Techniques
 All statistics were calculated using degrees of freedom corrected for autocorrelation, following the degrees of freedom calculation of Bretherton et al. . Linear correlations (Pearson's r) were computed using two-tailed hypotheses, and error bars were calculated using the Fisher's z-transformation. Bivariate least squares regressions were computed following the method of York et al.  using the spreadsheet prepared as supplementary material by Cantrell . Unless otherwise noted, all averages are reported as the mean ± one standard deviation.
3.1 Mount Bachelor Observatory/Boise Analysis
3.1.1 Forward Trajectory Analysis
 A forward trajectory analysis from MBO was conducted to identify a surface monitoring site located in a region consistently downwind of MBO. Two 48 hour forward trajectories initiated at 22:00 and 4:00 PST from MBO were calculated for each day during 2004 to 2010. A density plot of these trajectories (Figure 2) clearly illustrates westerly flow from MBO. Boise, Idaho, was chosen because it is the closest urban monitoring site to the east of MBO.
3.1.2 Boise Backward Trajectory and Cluster Analysis
 A cluster analysis of backward trajectories initiated at Boise was used to identify the days when air masses contributing to Boise MDA8 were also observed at MBO. Daily 24 hour HYSPLIT backward trajectories from 1500 m agl were initiated at Boise at 9:00, 12:00, 15:00, and 18:00 PST, hours generally within the Boise MDA8. Using the EDAS meteorological output from these trajectories, the average maximum daily mixed layer height in Boise was 1654 ± 729 m agl. This supports initiation of the trajectories at 1500 m agl because air masses at this height are at the upper limit of the boundary layer or within the entrainment zone and, therefore, mixed into the air masses measured at the surface. As a check, we also completed this cluster analysis with trajectories initiated at 900 m agl; however, these trajectories generally remained within the boundary layer for longer time frames than would be expected for FT to boundary layer transport.
 A cluster analysis of the calculated backward trajectories resulted in five clusters (Figure 3). Clusters 2 and 3 show horizontal transport between MBO and Boise, and the average transport times between Boise and the longitude of MBO for trajectories in clusters 2 and 3 are 23.1 ± 7.9 and 13.3 ± 5.7 hours, respectively. There is a natural offset between the daily MBO and Boise O3 data sets, as the 8 hour FT O3 average at MBO starts at 21:00 PST, approximately 12 to 14 hours before the average Boise MDA8 start time of 9:00 to 11:00 PST. Therefore, we expect cluster 3 to better approximate intersite transport over this time scale.
 For each of the five clusters, Figure 4a shows the correlation between the daily baseline O3 observed at MBO and the subsequent Boise MDA8. As hypothesized, cluster 3 demonstrates the best correlation between the MBO and Boise O3 observations (r = 0.63, p ≤ 0.01). Further investigation of cluster 3 trajectory elevation between Boise and MBO shows that, on average, the air masses represented by these trajectories descend continuously as they are transported from MBO to Boise. At the longitude of MBO, the average pressure level of the trajectories is 710 ± 49 mb, which encompasses the average pressure at MBO during 2006 to 2010, 733 ± 5 mb.
 Figure 4b shows the 95% confidence intervals for each of the cluster correlations. First, Figure 4b shows that there is a significant difference between the correlation for the cluster best representing transport of FT air masses from MBO to Boise (cluster 3) and the correlation of one of the other clusters (cluster 5). Therefore, the cluster analysis technique is a useful method for separating trajectories with different transport patterns. Second, Figure 4b shows that there is not a significant difference between the correlations of many of the clusters. This is expected because there is often little variation in baseline O3 within small spatial areas [NRC, 2009], so it is likely that many of the clusters represent transport of similar air masses to Boise. The objective of this analysis is not to identify every trajectory representing transport of FT O3 from MBO to Boise, but rather to identify a set of trajectories that is strongly representative of this transport pattern. Therefore, further analysis in section 4 will focus on cluster 3.
3.2 Cheeka Peak/Enumclaw Analysis
3.2.1 Forward Trajectory Analysis
 A forward trajectory analysis from CP was conducted to identify an urban monitoring site consistently downwind of CP. Two daily 48 hour forward trajectories initiated at 12:00 and 15:00 PST were calculated from 1000 m agl at CP during 2004 to 2010. A density plot of these forward trajectories (Figure 5) demonstrates that there is a wide range of transport pathways from CP. Enumclaw was chosen because it is an urban site along one of the trajectory paths and was in noncompliance with the O3 NAAQS for the 2006 to 2008 averaging period.
3.2.2 Enumclaw Backward Trajectory and Cluster Analysis
 A set of 24 hour backward trajectories were initiated from 500 m agl at Enumclaw at 10:00, 13:00, 16:00, and 19:00 PST, hours generally within the MDA8. Using the EDAS meteorological output from these four daily trajectories, the average maximum daily mixed layer height in Enumclaw was 960 ± 359 m agl, showing that the initiation height for the trajectories is within the boundary layer.
 A cluster analysis of the backward trajectories resulted in eight clusters (Figure 6). There is sometimes easterly flow from the continent toward CP [Weiss-Penzias et al., 2004], which could cause a correlation between the CP and Enumclaw data that is due to transport of continental air toward CP rather than transport of MBL O3 from the Pacific Ocean toward Enumclaw. The percentage of days in the cluster analysis with easterly flow from the continent (defined as a wind direction measurement at CP between 45° and 135° of due North) was calculated. Only 5.4% of days in the clusters had easterly flow, and none of these days are in the cluster discussed in sections 3.3 and 4 (cluster 5).
 The daily 8 hour average for the CP observations starts at 11:00 PST, and the MDA8 at Enumclaw generally starts between 10:00 and 12:00 PST. Therefore, if air masses are transported very quickly between the coast and Enumclaw, we would expect the daily averages from the two sites to be well correlated, but if the air masses are transported more slowly, the CP data on a given day would likely correlate best with the Enumclaw data from the following day. An autocorrelation analysis for each of the 3 years where there are continuous daily MBL O3 observations at CP during May to September (2004, 2008, and 2010) shows that for a 1 day lag, the autocorrelation is always greater than 0.6. Because of this low day-to-day variability, the Enumclaw MDA8 were correlated with a 2 day running average of the CP MBL O3 mixing ratios for each of the eight clusters (Figure 7a). Specifically, each Enumclaw MDA8 is compared with the average of the MBL O3 from the same day as the MDA8 and the MBL O3 from the previous day. As a check, correlations for each cluster were also calculated separately for: (1) the Enumclaw MDA8 and CP observations on the same day, and (2) the Enumclaw MDA8 with the CP observations from the previous day. These calculations showed the same general pattern as in Figure 7a, supporting the decision to use a 2 day running average.
 The low correlation for clusters 1, 2, and 8 is expected because these clusters do not represent transport between CP and Enumclaw, and clusters 1 and 2 represent transport near major metropolitan areas (see Figure 6). Two factors can help explain the correlations for clusters 3 to 7 [r = 0.58, 0.36, 0.83, 0.43, and 0.83 (p ≤ 0.01), respectively], which represent transport between the Washington coast and Enumclaw:
 Low-elevation sites are not as likely to observe strong variations in O3 due to subsidence of UT/LS air masses and long-range transport of pollution plumes [Lin et al., 2012a], and there is likely little variation in O3 along the PNW coast (based on the average standard deviation of CP O3 data, which during a month is 6.4 ppbv). We infer a low spatial variation in O3 along the coast because of this low temporal variation at CP despite a large range in wind direction observations at the site. Therefore, we find significant correlations between the CP and Enumclaw data for any days with westerly transport from the coast.
 An analysis of the average elevation of the trajectories in each of these clusters between Enumclaw and the coast shows that they vary between 286 and 617 m agl, demonstrating boundary layer transport. There are anthropogenic emissions present throughout the PNW (see Figure 6), so the amount of time spent in transport over the continent is likely related to the amount of local pollutants present in the air masses.
 The trajectories in clusters 3 to 7 have average transport times between Enumclaw and the longitude of CP of 13.0 ± 3.5, 18.1 ± 3.8, 10.3 ± 3.0, 18.0 ± 4.3, and 15.1 ± 4.1 hours, respectively. Clusters 5 and 7 have the best correlations (r = 0.83 and p ≤ 0.01 for both) and are also characterized by fast average transport between the coast and Enumclaw. However, we find that the correlation for cluster 7 is largely driven by the seasonal variations in MBL O3, whereas the correlation for cluster 5 holds true for daily variations in MBL O3 within one season. The relatively lower correlations for clusters 3, 4, and 6 can be explained by the greater influence of local pollution during continental transport. Figure 7b shows the 95% confidence intervals for each of the cluster correlations, which again shows the utility of using the cluster analysis technique for selecting the set of days to analyze. Further analysis in section 4 will focus only on cluster 5.
3.3 Urban/Baseline O3 Regression Slopes
 Figure 8 shows a scatterplot of the Boise and MBO O3 mixing ratios for days in cluster 3 (r = 0.63; R2 = 0.40; p ≤ 0.01), which have a bivariate least squares regression of Boise_O3 = 0.52 * MBO_O3 + 19. The slope depends on production and loss of O3 in the FT, dilution during mixing into the boundary layer, and O3 production and loss—from both deposition and chemical reactions—within the boundary layer. To test whether the regression is driven by the difference in high-value summer/spring points and low-value winter/fall points, the regression was run with just the spring/summer points, which make up the majority of the data. This regression was significant (r = 0.58; p ≤ 0.01), indicating that the correlation is largely driven by daily variations in baseline air mass composition. The lowest MBO FT O3 mixing ratios correspond with relatively higher Boise O3 mixing ratios, likely reflecting some contribution from net boundary layer production of O3 from local and upwind sources. The highest MBO FT O3 mixing ratios correspond with relatively lower Boise O3 mixing ratios, suggesting that some of the FT O3 is destroyed or that the FT air mass is entrained into a boundary layer with a very low O3 mixing ratio.
 Figure 9 shows AIRPACT/CMAQ results for the transport of a nonreactive O3 tracer from the area above MBO to the surface near Boise on 6 April 2010, one of the high O3 days in cluster 3 (see point indicated in Figure 8). This plot results from tracking 21 tracers into the model domain; these tracers were set, one tracer per layer, to match the MOZART-4 O3 mixing ratios within each layer along the western boundary throughout 5 April 2010, but were zero otherwise. The average O3 mixing ratio at the western boundary for model layers 13 to 16 (approximately 875 to 700 mb) between 0:00 and 20:00 PST on 5 April 2010 was 48.8 ± 2.4 ppbv. Advection of these boundary condition O3 tracers was modeled for 5 April 2010 and then a subset of the result was used as the initial conditions for simulating 6 April 2010; the boundary conditions for 6 April 2010 were everywhere zero. Figure 9a shows a vertical cross section of the modeled volume of O3 tracer remaining in model layers 13 to 16 above MBO on 6 April 2010 at 4:00 PST. This volume, which is an air mass approximately 180 × 180 km, has an average mixing ratio of 43.3 ± 20.9 ppbv. MBO observed an 8 hour average near the top of this range (61.5 ppbv). Figure 9b shows the modeled surface mixing ratio of the nonreactive O3 tracer at 14:00 PST. In Boise, for that time, the modeled contribution from this FT tracer was 22.4 ppbv. The measured MDA8 in Boise on this day was 47.7 ppbv. A second AIRPACT/CMAQ simulation using full chemistry for the same time period and initial boundary conditions resulted in an O3 mixing ratio of 44.5 ppbv in Boise. In addition to validating that there is transport of an O3-rich air mass from the FT above MBO to Boise on one of the days selected for this analysis, the nonreactive tracer plots in Figure 9 also suggest that FT O3 is diluted by about 50% during mixing into the boundary layer on this day. In combination with the full-chemistry AIRPACT/CMAQ simulation results for 6 April, the nonreactive tracer results suggest that the final O3 mixing ratio in Boise can be explained by the addition of the diluted FT O3 (advected from the western boundary) to photochemical O3 production and loss, and dry deposition. It is of some interest that a stratospheric intrusion event documented by Lin et al. [2012b] may have contributed to the high baseline O3 at MBO.
 In an analysis of high O3 events at MBO, Ambrose et al.  determined that two of the highest baseline O3 days in cluster 3, 14–15 June 2008, were influenced by subsidence of UT/LS air masses and a combination of UT/LS subsidence and long-range transport of Asian pollution, respectively. The Boise MDA8 corresponding to these high-O3 events are 58.3 and 54.0 ppbv, respectively. Both of these values are above the 90th percentile for Boise MDA8 in cluster 3 and above the 80th and the 60th percentiles, respectively, of all Boise MDA8 during 2006 to 2010, demonstrating that episodic events contribute to high surface O3 observations.
 Figure 10 shows a scatterplot of the CP and Enumclaw O3 mixing ratios for days in cluster 5 (r = 0.83; R2 = 0.69; p ≤ 0.01), which have a bivariate regression of Enumclaw_O3 = 1.04 * CP_O3 − 0.73. The regression slope suggests that there is either minimal or balanced production, dilution, and loss during the rapid transport of MBL O3 from the coast to Enumclaw. However, Enumclaw cluster 5 represents transport over areas with relatively large anthropogenic emissions (see Figure 6), so the MDA8 likely include some boundary layer production of O3 from upwind emissions. Because there is little day-to-day variability in the CP data due to its low elevation, the Enumclaw/CP correlation is largely driven by the seasonal cycle in MBL O3; however, there are still significant correlations if each season is analyzed separately (spring: r = 0.81; summer: r = 0.66; autumn: r = 0.79; p ≤ 0.01), indicating that daily variations in baseline O3 also drive the slope of this regression.
4.1 Regression Slope Comparison
 A correlation has been identified between surface and FT O3 in the western United States [Jaffe, 2011; Parrish et al., 2010], but only Parrish et al.  quantify the relationship of baseline and in situ surface O3 observations. Parrish et al.  compared MDA8 from four low-elevation surface monitoring sites in California's Northern Sacramento Valley with average baseline O3 observations from TH ozonesondes, and quantified the relationship between baseline and downwind surface mixing ratios using bivariate least squares regression slopes. However, the authors used all available data to quantify this relationship, whereas this study uses a cluster analysis to select the days with a strong influence from FT or MBL O3. Parrish et al.  found that the regression slopes for the four sites varied between 0.65 and 1.10, with a weighted average of 0.88 ± 0.13 (mean ± 95% confidence interval). Likewise, using a 95% confidence interval, the slopes calculated in this analysis are 0.52 ± 0.16 for transport from the FT to the boundary layer and 1.04 ± 0.08 for transport within the boundary layer. We attribute the lack of significant difference between the Parrish et al.  slope and our within-boundary layer transport slope primarily to the elevation of the baseline O3 observations in each analysis. Parrish et al.  define baseline as the average O3 between 1 and 2.2 km asl (above sea level), a range that is consistently above the MBL but is lower than the FT O3 observations at MBO (2.7 km asl).
4.2 Range of Maximum Daily 8 Hour Average Ozone Mixing Ratios in Boise and Enumclaw
 The mean Boise cluster 3 MDA8 O3 in May to September are 48.3 ± 8.4, 48.0 ± 10.1, 48.2 ± 5.8, 41.5 ± 5.7, and 35.8 ± 6.9, respectively. The mean Enumclaw cluster 5 MDA8 in May to September are 38.4 ± 8.1, 32.4 ± 2.6, 20.3 ± 0.9, 25.3 ± 2.5, and 31.0 ± 4.2 ppbv, respectively. The maximum MDA8 in these Boise and Enumclaw clusters are lower than 75 ppbv, indicating that local, boundary layer O3 production is necessary to exceed the current U.S. standard at both sites. Parrish et al.  found that the average baseline contribution at four low-elevation sites in the Northern Sacramento Valley is 49 ppbv, and that at one of these sites the baseline contribution can range from 34 to 90 ppbv.
 Mean MDA8 O3 in Boise cluster 3 is significantly higher than in Enumclaw cluster 5 (Table 2). The inland Boise site is located at 825 m asl, whereas the Enumclaw site is located in the Puget Sound region at 402 m asl. This difference in elevation could partly account for the differing mean O3 at the sites, as O3 mixing ratios tend to increase with increasing elevation [Cooper and Peterson, 2000; Jaffe, 2011]. However, a second reason for this difference is likely the greater entrainment of FT air masses in Boise. This is shown by the relatively higher average maximum daily mixed layer height at Boise (1654 ± 729 m agl) compared to Enumclaw (960 ± 359 m agl), which is likely a function of the greater surface warming in Boise during the months analyzed.
Table 2. Maximum Daily 8 Hour Average O3 Descriptive Statistics for the Boise and Enumclaw Clusters
Mean ± 95% confidence intervals (ppbv)
Total No. of Days
No. of Days with MDA8 > 60 ppbv
No. of Days with MDA8 > 65 ppbv
No. of Days with MDA8 > 70 ppbv
No. of Days with MDA8 > 75 ppbv
Clusters that best represent transport from baseline sites (Boise cluster 3 and Enumclaw cluster 5; see ;sections 3.1.2 and ;3.2.2).
 Table 2 presents descriptive statistics for the MDA8 in each of the Boise and Enumclaw clusters, with Boise cluster 3 and Enumclaw cluster 5 representing 9% and 4% of the data analyzed for these cities, respectively. Many factors could influence the range of O3 mixing ratios in the clusters, including varying boundary layer O3 production from local and upwind emissions, O3 production from lightning NOx, O3 titration by urban NOx emissions, dry deposition rates, and the degree of boundary layer growth and mixing. However, as shown by the NOx emissions in Figures 3 and 6, all of the clusters are transported near areas with anthropogenic emissions, indicating that O3 production from regional pollution is a key factor enhancing O3 above baseline values on many days. The days in Boise cluster 3 and Enumclaw cluster 5 have lower average O3 mixing ratios than days in many other clusters, likely reflecting less local/regional photochemical O3 production on these days.
 No days in Boise cluster 3 or Enumclaw cluster 5 exceed the current U.S. standard of 75 ppbv; however, if the standards are tightened to 60 to 70 ppbv, the Boise cluster would include 1 to 3 exceedance days during the time period studied, whereas the Enumclaw cluster would continue to have no exceedance days due to baseline O3 alone. The U.S. O3 NAAQS is based on a 3 year running average of the annual fourth highest MDA8 in a given location. Therefore, even one additional MDA8 exceeding the standard can impact compliance. Our results show that the entrainment of FT air masses transported from the west results in days that have O3 concentrations ≥ 60 to 70 ppbv.
 This study presents the first multiyear analysis of the impact of FT and MBL O3 on urban surface O3 in the U.S. PNW. We analyzed the transport of FT O3 from MBO, a central Oregon mountaintop site, to Boise, Idaho, and the transport of MBL O3 from CP, a low-elevation coastal site in Washington, to Enumclaw, Washington. Using a cluster analysis of backward trajectories, we identified a set of days when FT and MBL air masses have a strong influence on MDA8 O3 in Boise (9% of data) and Enumclaw (4% of data), respectively. On the identified days, MBO FT O3 observations explain 40% of the variation in Boise MDA8 and a bivariate least squares regression of the Boise MDA8 versus MBO O3 observations results in a slope of 0.52 ± 0.16, indicating that FT O3 is diluted during transport to the surface. AIRPACT/CMAQ results support the conclusion that on 6 April 2010, one of the high O3 days analyzed, there was transport of an O3-enriched FT air mass, which was diluted by approximately 50% during mixing into the boundary layer, and its entrainment contributed to the boundary layer O3 at Boise. The CP observations explain 69% of the variation in Enumclaw MDA8 on the identified days, and the slope of the bivariate least squares regression of Enumclaw MDA8 versus CP MBL observations is 1.04 ± 0.08. This slope indicates that dilution, production, and loss are minimal or balanced during rapid transport of MBL O3.
 Days with rapid westerly transport of FT O3 to Boise and MBL O3 to Enumclaw are characterized by relatively low MDA8, indicating that local, boundary layer O3 production is low on these days. On the set of days analyzed, MDA8 O3 is often higher at the inland Boise site than at the Enumclaw site, which is located in the Puget Sound region. One reason for this is the higher elevation at the Boise site. A second explanation is the greater entrainment of FT air at Boise due to a higher average daily maximum mixed layer height (1654 m agl, compared to 960 m agl in Enumclaw). This finding is likely applicable to other PNW sites, as it describes an important difference between lower-elevation coastal sites like Enumclaw, which during the warmer months tend to have less surface warming than the higher-elevation, inland sites such as Boise. More surface warming can lead to greater mixed layer growth and, therefore, greater entrainment of air masses that may include O3 resulting from stratospheric intrusions and long-range transport of industrial pollution.
 The average MDA8 in Boise and Enumclaw show that O3 produced locally in the boundary layer is necessary to exceed the current U.S. standard of 75 ppbv. However, if the NAAQS is tightened in 2013 to between 60 and 70 ppbv, our results show that entrainment of FT O3 will play an increasing role in exceedances in Boise.
 The authors thank Rick Hardy at the Idaho DEQ for access to the Boise O3 data, the PSCAA for access to the CP and Enumclaw O3 data, both the European Commission Joint Research Centre and Netherlands Environmental Assessment Agency for access to the EDGAR v4.2 NOx emissions data, Louisa Emmons at NCAR for access to MOZART results for use as boundary conditions, and Zhang Rui at Washington State University for help with the CMAQ tracer code. Funding for research at MBO was provided by National Science Foundation (NSF) grant AGS-1066032. Funding for integrating MOZART boundary conditions into AIRPACT and for CMAQ tracer modeling was provided under NASA grant NNX11AE57G. Some of this research is based on work supported by the NSF Graduate Research Fellowship under grant DGE-0718124.