The FAST-LVOS experiment was designed to further our understanding of the impacts of STT, wildfires, long-range transport from Asia, and regional pollution on air quality in the Las Vegas Valley. The field campaign was carried out between 17 May and 30 June 2017 in Clark County (NV), which includes the greater Las Vegas area (Fig. 1). The measurement campaign consisted of daily lidar and in situ measurements supplemented by aircraft and ozonesonde profiling during the four IOPs (23–25 May, 31 May–2 June, 10–14 June, and 28–30 June). The daily measurements included chemical composition (e.g., CO and O3) and meteorological parameters (e.g., air temperature and water vapor) recorded with high temporal resolution by instruments installed in a mobile laboratory (Wild et al., 2017) parked on the summit of Angel Peak (36.32∘ N, 115.57∘ W; 2682 m above sea level, a.s.l.), the site of the 2013 LVOS field campaign. This mountain-top site, located ∼45 km northwest of Las Vegas City (see Fig. 1), is far from anthropogenic emission sources and mostly receives free-tropospheric air at night but is frequently influenced during the day by air transported from the Las Vegas Valley through upslope flow in late spring and summer (Langford et al., 2015). The Tunable Optical Profiler for Aerosols and oZone (TOPAZ) three-wavelength mobile differential absorption lidar (DIAL) system, which was previously deployed to Angel Peak during LVOS, was relocated to North Las Vegas Airport (NLVA; Fig. 1), where it measured 8 min averaged vertical profiles of O3 and aerosol backscatter from 27.5 to ∼8 km above ground level (a.g.l.) with an effective vertical resolution (for O3) ranging from ∼10 m near the surface to ∼150 m at 500 m a.g.l. and ∼900 m at 6 km a.g.l. The aerosol backscatter profiles were retrieved at 7.5 m resolution. TOPAZ was operated daily, but not continuously, throughout the campaign. NOAA also deployed a continuously operating micro-Doppler lidar at NLVA to measure vertical velocities and relative aerosol backscatter throughout the campaign. Boundary layer heights were inferred from the micro-Doppler measurements following the method in Bonin et al. (2018).

The routine in situ and lidar measurements described above were augmented during the four IOPs by ozonesondes launched up to four times per day (30 launches total during the entire campaign) from the Clark County Department of Air Quality Joe Neal monitoring site located ∼8 km north–northwest of the NLVA. Aircraft measurements were also conducted by Scientific Aviation to sample O3, methane (CH4), water vapor (H2O), and nitrogen dioxide (NO2) between NLVA and Big Bear, CA, during the IOPs. Readers can refer to our previous studies (Langford et al., 2010, 2015, 2017, 2019; Alvarez II et al., 2011) for detailed descriptions and configurations of TOPAZ and the other measurement instruments. The FAST-LVOS field campaign is also described in more detail elsewhere (Langford et al., 2020).

The FAST-LVOS measurements were augmented by hourly surface O3 measurements from Joe Neal and other regulatory air quality monitoring sites operated by the Clark County Department of Air Quality (Table S1 in the Supplement). Surface observations of O3 from these and other mostly urban sites were obtained from the U.S. Environmental Protection Agency (EPA) Air Quality System (AQS; https://www.epa.gov/aqs, last access: 19 March 2019). We average the AQS measurements into 0.5∘×0.625∘ grids for a direct comparison with model results (as in Lin et al., 2012a, b). Surface observations from rural sites and more representative of background air were obtained from the EPA Clean Air Status and Trends Network (CASTNet; https://www.epa.gov/castnet, last access: 19 March 2019).

Comparisons of key model configurations are shown in Table S2. AM4 is the new generation of the Geophysical Fluid Dynamics Laboratory chemistry–climate model contributing to the Coupled Model Intercomparison Project, Phase 6 (CMIP6). The model employed in this study, a prototype version of AM4.1 (Horowitz et al., 2020), differs from the AM4 configuration described in Zhao et al. (2018a, 2018b) by including 49 vertical levels extending up to 1 Pa (∼80 km) and interactive stratosphere–troposphere chemistry and aerosols. Major physical improvements in GFDL-AM4, compared to its predecessor GFDL-AM3 (Donner et al., 2011), include a new double-plume convection scheme with improved representation of convective scavenging of soluble tracers, new mountain drag parametrization, and the updated hydrostatic finite-volume cubed-sphere dynamical core (Zhao et al., 2016, 2018a, b). For tropospheric chemistry, GFDL-AM4 includes improved treatment of photooxidation of biogenic volatile organic compounds (VOCs), photolysis rates, heterogeneous chemistry, sulfate and nitrate chemistry, and deposition processes (Mao et al., 2013a, 2013b; Paulot et al., 2016, 2017; Li et al., 2016), as described in more detail by Schnell et al. (2018). We implement a stratospheric O3 tracer (O3Strat) in GFDL-AM4 to track O3 originating from the stratosphere. The O3Strat is defined relative to a dynamically varying e90 tropopause (Prather et al., 2011) and is subject to tropospheric chemical loss (in the same manner as odd oxygen of tropospheric origin) and deposition to the surface (Lin et al., 2012a, 2015a). The model is nudged to NCEP reanalysis winds using a height-dependent nudging technique (Lin et al., 2012b). The nudging minimizes the influences of chemistry–climate feedbacks and ensures that the large-scale meteorological conditions are similar to those observed across the sensitivity simulations. We conduct a suite of AM4 simulations at C192 (∼50×50 km2) horizontal resolution for January–June 2017: (1) a base simulation (BASE) with all emissions included; (2) a sensitivity simulation without anthropogenic emissions over North America (15–90∘ N, 165–50∘ W; NAB); (3) a sensitivity simulation without anthropogenic emissions over the US (USB); (4) a sensitivity simulation without Asian anthropogenic emissions, and (5) a sensitivity simulation without wildfire emissions (see Table S3). The high-resolution BASE and sensitivity simulations for January–June 2017 are initialized from the corresponding nudged C96 (∼100×100 km2) simulations spanning from 2009 to 2016 (8 years). Compared to the NAB simulation, the USB simulation includes additional contributions from Canadian and Mexican anthropogenic emissions. The USB estimates are now generically defined as “background O3” and used by the U.S. EPA. Over the western US (WUS), the vertical model resolution ranges from ∼50 to 200 m near the surface to ∼1–1.5 km near the tropopause and ∼2–3 km in much of the stratosphere.

The Goddard Earth Observing System coupled with Chemistry (GEOS-Chem; http://geos-chem.org, last access: 28 October 2019) is a widely used global chemical transport model (CTM) for simulating atmospheric composition and air quality (Bey et al., 2001; Zhang et al., 2011), driven by assimilated meteorological fields from the NASA Global Modeling and Assimilation Office (GMAO). We conduct high-resolution simulations over North America (10∘–70∘ N, 140∘–40∘ W), with 0.25∘ (latitude) ×0.3125∘ (longitude) horizontal resolution, using a one-way nested-grid version of GEOS-Chem (v11.01) (Wang et al., 2004; Chen et al., 2009) driven by the Goddard Earth Observing System – Forward Processing (GEOS-FP) assimilated meteorological data. The model uses a fully coupled NOx-OX–hydrocarbon–aerosol–bromine chemistry mechanism in the troposphere (“Tropchem”), whereas a simplified linearized chemistry mechanism (Linoz) is used in the stratosphere to simulate stratospheric ozone and cross-tropopause ozone fluxes (McLinden et al., 2000). Although GEOS-Chem can also be run with the Universal tropospheric–stratospheric Chemistry eXtension (UCX) mechanism that simulates interactive stratosphere–troposphere chemistry and aerosols (Eastham et al., 2014), this option was not used in the simulations presented in this study due to computational constraints. To further save computational resources, we used a reduced vertical resolution of 47 hybrid eta levels, by combining vertical layers above ∼80 hPa from the native 72 levels of GEOS-FP. The thickness of model vertical layers over the WUS ranges from ∼15 to 100 m near the surface to ∼1 km near the tropopause and in the lower stratosphere. Similar GEOS-Chem simulations with simplified treatments of stratospheric chemistry and dynamics have been previously used to estimate background O3 for U.S. EPA policy assessments (Zhang et al., 2011, 2014; Fiore et al., 2014; Guo et al., 2018). Thus, it is important to assess the ability of this model to represent high-background-O3 events from stratospheric intrusions. We conduct two nested high-resolution simulations with GEOS-Chem for February–June 2017: BASE and a USB simulation with anthropogenic emissions zeroed out in the US (Table S3). Initial and boundary conditions for chemical fields in the nested-grid simulations were provided by the corresponding BASE and USB GEOS-Chem global simulations at 2∘×2.5∘ resolution for January–June 2017. Only results for April–June from the nested simulations are analyzed in this study. The 3-month spin-up period (January–March) used for GEOS-Chem is relatively short compared to the multi-year GFDL-AM4 simulations, although it should be sufficient given that the lifetime of ozone in the free troposphere is approximately 3 weeks (e.g., Young et al., 2018).

The anthropogenic emissions used in GFDL-AM4 are modified from the CMIP6 historical emission inventory (Hoesly et al., 2018). The CMIP6 emission inventory does not capture the decreasing trend in anthropogenic NOx emissions over China after 2011 as inferred from satellite-measured tropospheric NO2 columns (Liu et al., 2016; Fig. S1 in the Supplement). We thus scale CMIP6 NOx emissions over China after 2011 based on a regional emission inventory developed by Tsinghua University (Qiang Zhang at Tsinghua University, personal communications, 13 March 2018; Fig. S1). The adjusted NOx emission trend over China agrees well with the NO2 trend derived from satellite retrievals. We also reduce NOx emissions over the EUS (25∘–50∘ N, 94.5∘–75∘ W) by 50 % following Travis et al. (2016), who suggested that excessive NOx emissions may be responsible for the common model biases in simulating O3 over the southeastern US. These emission adjustments reduce mean MDA8 O3 biases in GFDL-AM4 by ∼5 ppbv in spring and ∼10 ppbv in summer over the EUS (Fig. S2). The model applies the latest daily resolving global fire emission inventory from NCAR (FINN) (Wiedinmyer et al., 2011), vertically distributed over six ecosystem-dependent altitude layers from the ground surface to 6 km (Dentener et al., 2006; Lin et al., 2012b). Biogenic isoprene emissions (based on MEGAN; Guenther et al., 2006; Rasmussen et al., 2012), lightning NOx emissions, dimethyl sulfide, and sea salt emissions are tied to model meteorological fields (Donner et al., 2011; Naik et al., 2013).

For GEOS-Chem, anthropogenic emissions over the United States are scaled from the 2011 US NEI to reflect the conditions in 2017 (https://www.epa.gov/air-emissions-inventories/air-pollutant-emissions-trends-data, last access: 10 October 2019). Similar to AM4, we reduce EUS anthropogenic NOx emissions in GEOS-Chem by 50 % to improve simulated O3 distributions. Anthropogenic emissions over China are based on the 2010 MIX emission inventory (Li et al., 2017), with NOx emissions scaled after 2010 using the same trend as in GFDL-AM4. Biogenic VOC emissions are calculated online with MEGAN (Guenther et al., 2006). Biomass burning emissions are from the FINN inventory but implemented in the lowest model layer. The model calculates lightning NOx emissions using a monthly climatology of satellite lightning observations coupled to parameterized deep convection (Murray et al., 2012). The calculation of lightning NOx in this study differs from that in Zhang et al. (2014), who used the U.S. National Lightning Detection Network (NLDN) data to constrain model flash rates.

We first compare O3 simulations in AM4 with those from its predecessor, AM3, which has been extensively used in previous studies to estimate background O3 (Lin et al., 2012a, 2012b, 2015a; Fiore et al., 2014). Figure 2 shows the comparisons of simulated and observed March mean O3 vertical profiles and mid-tropospheric O3 seasonal cycles at the Trinidad Head and Boulder ozonesonde sites. Free-tropospheric O3 measured at both sites in March is representative of background conditions, with little influence from US anthropogenic emissions. Thus, we also show O3 from the NAB simulations with North American anthropogenic emissions zeroed out. As constrained by the availability of AM3 simulations from previous studies, we focus on the 2010–2014 period and compare the NAB estimates as opposed to the USB estimates used in the rest of the paper. Compared with AM3, simulations of free-tropospheric O3 are much improved in AM4. Mean O3 biases are reduced by 10–25 ppbv in the middle troposphere and 20–65 ppbv in the upper troposphere in AM4, reflecting mostly an improved simulation of background O3 (Fig. 2a). These improvements are mainly credited to the changes in dynamics or convection schemes in AM4 (Zhao et al., 2018a), according to our sensitivity simulations (not shown). The difference in emission inventories contributes to some of the O3 differences but is not the major cause because the largest differences between the two models in simulated free-tropospheric O3 occur during the cold months (November–April) when photochemistry is weak (Fig. 2b).

Next, we examine how GFDL-AM4 compares with GEOS-Chem in simulating the mean distribution and the day-to-day variability of total and USB O3 in the free troposphere (Fig. 3) and at the surface (Figs. 4 and S3) during FAST-LVOS. Comparisons with ozonesondes at Joe Neal show that the total O3 concentrations below 700 hPa simulated by the two models often bracket the observed values (Fig. 3a). Between 700 and 300 hPa, GFDL-AM4 better captures the observed mean and day-to-day variability of O3, as evaluated with the standard deviation. Further comparison with lidar measurements averaged over 3–6 km altitude above Las Vegas shows that total and USB O3 in GFDL-AM4 exhibits larger day-to-day variability than in GEOS-Chem (σ=8.1 ppbv in observations, 8.1 ppbv in AM4, and 6.7 ppbv in GEOS-Chem; Fig. 3c). For mean O3 levels in the free troposphere, AM4 estimates a 7 ppbv contribution from US anthropogenic emissions (total minus USB), while GEOS-Chem suggests only 3.5 ppbv. The largest discrepancies between the two models occurred on 11–13 June (the blue shaded period in Fig. 3c), which we later attribute to a stratospheric intrusion event (Sect. 4). During this period, AM4 simulates elevated O3 (70–75 ppbv) broadly consistent with the lidar and sonde measurements, while GEOS-Chem considerably underestimates the observations by 20 ppbv. Consistent with total O3, USB O3 in GFDL-AM4 is much higher than GEOS-Chem during this event.

Figure 4 shows the time series of observed and simulated surface MDA8 O3 at four high-elevation sites and one low-elevation site in the region during the study period. Statistics comparing the results at all sites are shown in Table S1. The two models show large differences in simulated total and USB O3 on days when the O3Strat tracer in AM4 indicates stratospheric influence (highlighted in blue shading). AM4 O3Strat indicates frequent STT events during April–June, with observed MDA8 O3 exceeding or approaching the current NAAQS of 70 ppbv. Compared with observations, GFDL-AM4 captures the spikes of MDA8 O3 and elevated USB O3 during these STT events (e.g., 23 April, 13 May, and 11 June). On these days, GEOS-Chem underestimates observed O3 by 10–25 ppbv and simulates much lower USB O3 levels than GFDL-AM4. For some days, GFDL-AM4 overestimates total MDA8 O3 due to excessive STT influence (e.g., 7 May at Spring Mountain Youth Camp). The two models also differ substantially in total and USB O3 (14–18 ppbv) on 22 June (yellow shading), with GEOS-Chem overestimating observations at high-elevation sites, while GFDL-AM4 underestimates observations at both high- and low-elevation sites. We provide more in-depth analysis of these events in Sect. 4 and identify the possible causes of the model biases.

Relationships between concurrently measured O3 and CO are useful to identify the possible origins of elevated surface O3 (Parrish et al., 1998; Herman et al., 1999; Langford et al., 2015). During FAST-LVOS, in situ 1 min measurements at Angel Peak show differences in ΔO3/ΔCO and water vapor content between air plumes during a variety of events (Figs. 5, 6, and S5). Notably, on 11 June, O3 was negatively correlated with CO (ΔO3/ΔCO = -3.79). This anti-correlation is distinctly different from the O3/CO relationships during other periods (e.g., ΔO3/ΔCO = 0.68–0.70 on 16 June or ΔO3/ΔCO = 1.08 on 2 June). The negative correlation (high O3 together with low CO) serves as strong evidence of a stratospheric origin of the air masses on 11 June, since O3 is much more abundant in the stratosphere than in the troposphere, whereas CO is mostly concentrated within the troposphere where it is directly emitted or chemically formed (Langford et al., 2015). By contrast, simultaneously elevated O3 and CO suggest influences by wildfires (e.g., 22 June) or anthropogenic (e.g., 16 June) pollution (Figs. 6b–d and S4). In particular, exceptionally high CO levels (∼100–440 ppbv) on 22 June (Fig. 6e) suggest influences from wildfires. Ozone enhancements were measured by the TOPAZ ozone lidar on 22 June (Sect. 4.3), although the correlation between CO and O3 at Angel Peak is not strong. The net production of O3 by wildfires is highly variable, with many contradictory observations reported in the literature (Jaffe and Wigder, 2012). The amount of O3 within a given smoke plume varies with distance from the fire and depends on the plume injection height, smoke density, and cloud cover (Faloona et al., 2020).

We gain further insights by examining water vapor concurrently measured at Angel Peak. Air masses from the lower stratosphere are generally dry, whereas wildfire or urban plumes from the boundary layer are relatively moist (Langford et al., 2015). Thus, the dry conditions of the air masses on 11 June support our conclusion that the plume was transported downward from the upper troposphere and lower stratosphere (Fig. 6a). These conditions are in contrast to those of the urban or wildfire plumes transported from the Las Vegas Valley (Fig. 6c–d). Additionally, we separate the anthropogenic plumes on 16 June into daytime and nighttime conditions because of a diurnal variation in air conditions (relatively dry at night versus wet during daytime; Fig. 6c–d). This analysis further demonstrates that the anthropogenic pollution plume during nighttime is wetter than the stratospheric air on 11 June. On 14 June (Fig. 6b), measured O3 was positively correlated with CO, indicating regional or local pollution influence, but the lower levels of water vapor than those in regional pollution and wildfire plumes suggest that the stratospheric air which reached Angel Peak earlier may have been mixed with local pollution. On 28 June (Fig. 6f), O3 was positively correlated with CO and the air masses were relatively dry, indicating that the plume was likely from aged pollution transported from Asia or southern California as opposed to from fresh pollution from the Las Vegas Valley. Identifying the primary source of the high-O3 events solely based on observations is challenging; additional insights from models are thus needed as we demonstrate below.

Analysis of the 250 hPa potential vorticity and the AM4 model stratospheric O3 tracer shows significant stratospheric influence on surface O3 in the SWUS on 22–23 April (Fig. S6), 13–14 May (Fig. S6), and 11–14 June (Figs. 7–8). During these events, surface MDA8 O3Strat in AM4 was 20–40 ppbv higher than the mean baseline level (15–20 ppbv; see dashed purple lines Fig. 4). Below, we focus on the 11–14 June event, which was the subject of a 4 d FAST-LVOS IOP with 60 h of continuous O3 lidar profiling and 13 ozonesonde launches, in addition to continuous in situ measurements at Angel Peak.

Significant enhancements in aerosol backscatter were observed at 3–6 km altitude above NLVA on 21–22 June, indicating the presence of wildfire smoke (Fig. 10a). Under the influence of the wildfire plume, mobile lab measurements at Angel Peak (∼3 km altitude) detected elevated CO as high as 440 ppbv in warm, moist air masses (Fig. 6e). The lidar measurements at NLVA on 22 June showed broad O3 enhancements (80–100 ppb) from the surface to 4 km altitude (Fig. 11a). After 12:00 PDT (19:00 UTC), a deep PBL (3–4 km) developed and O3 within the PBL was substantially enhanced (>80 ppbv), likely due to strong O3 production through reactions between abundant VOCs in the wildfire plumes and NOx in urban environments (Singh et al., 2012; Gong et al., 2017). Surface MDA8 O3 exceeded 70 ppbv at multiple sites in the Las Vegas Valley during the event (Table 1). Unfortunately, the synoptic conditions did not trigger an IOP, so there were no aircraft or ozonesonde measurements during this event.

GFDL-AM4 has difficulty simulating the O3-rich plumes above Clark County on 22 June (Fig. 11a). GEOS-Chem captures the observed high-O3 layers within the PBL but overestimates O3 above 4 km altitude (Fig. 11a). GEOS-Chem overestimates of free-tropospheric ozone seem to be common for the non-STT events during late spring through summer (Figs. 3b, 8b, and 11b and comparisons with lidar data for 24 May and 16 June shown in Sect. 4.4–4.6), likely due to excessive O3 produced from lightning NOx over the southern US (Zhang et al., 2011, 2014). At the surface, total MDA8 O3 concentrations simulated by the two models bracket the observed values at sites in the Las Vegas area (see yellow shading in Fig. 4) and across the Intermountain West (Fig. 12a). AM4 does not simulate elevated O3 during this event, while GEOS-Chem simulates elevated total and USB O3 levels across the entire southwest region. GEOS-Chem simulations during this wildfire event agree better with the observed MDA8 O3 enhancements (>70 ppbv) at Joe Neal (Fig. 4). At the high-elevation sites Angel Peak and Spring Mountain Youth Camp, however, GEOS-Chem overestimates the observed MDA8 O3 by 10–15 ppbv. Overall, GEOS-Chem seems to be more consistent with observations than GFDL-AM4 during this wildfire event. However, we cannot rule out the possibility that the better agreement between observations and GEOS-Chem simulations during this event may reflect excessive O3 from lightning NOx in the model (Zhang et al., 2014).

Meteorological conditions (e.g., temperature and wind fields) on 22 June in the reanalysis data used by GFDL-AM4 and GEOS-Chem are similar over the WUS (not shown). The two models use the same wildfire emissions (FINN) but with different vertical distributions. Fire emissions are distributed between the surface and 6 km altitude in GFDL-AM4 but are placed at the surface level in GEOS-Chem. We conduct several sensitivity simulations with GFDL-AM4 to investigate the causes of the model biases. Placing all fire emissions at the surface in GFDL-AM4 results in ±5 ppbv differences in modeled MDA8 O3 on 22 June (Fig. S8). Observations suggested that 40 % of NOx can be converted rapidly to peroxyacyl nitrate (PAN) and 20 % to HNO3 in fresh boreal fire plumes over North America (Alvarado et al., 2010). Both models currently treat 100 % of wildfire NOx emissions as NO. We conduct an additional AM4 sensitivity simulation, in which 40 % of the wildfire NOx emissions are released as PAN and 20 % as HNO3. This treatment results in ±2 ppbv differences in simulated monthly mean MDA8 O3 during an active wildfire season (August 2012; Fig. S9). Overall, these changes do not substantially improve simulated O3 on 22 June. Future efforts are needed to investigate the ability of current models to simulate O3 formations in fire plumes (Jaffe et al., 2018).

Regional and local anthropogenic emissions were important sources of elevated O3 in Clark County during FAST-LVOS, contributing to 3 out of 10 observed high-O3 events above 65 ppbv during April–June 2017 (Table 1). Below, we focus on the 16 June event when severe O3 pollution with MDA8 O3 exceeding 70 ppbv occurred over California, Arizona, parts of Nevada, and New Mexico. Analysis for the 2 and 29–30 June pollution events are shown in the Supplement (Figs. S5, S10, and S11). The TOPAZ lidar measurements on 16 June show elevated O3 of 55–90 ppbv in the 4 km deep PBL (Fig. 11b). However, this event did not trigger an IOP, so ozonesonde and aircraft measurements are unavailable. Both GFDL-AM4 and GEOS-Chem capture the buildup of O3 pollution in the PBL on 16 June (Fig. 11b). Both models show boundary layer enhancements of total O3 but not of USB O3 (Fig. 11b), indicating that regional or local anthropogenic emissions are the primary source of observed O3 enhancements. Similar to 16 June, GEOS-Chem clearly shows enhancements in total O3 in the PBL but not in USB O3 on 2 and 29–30 June (Fig. S10). The model attribution to US anthropogenic emissions is consistent with the positive correlation between O3 and CO measured at Angel Peak on 16 (Fig. 6c–d), 2, and 29–30 June (Fig. S5). It is noteworthy that, with its higher horizontal resolution, GEOS-Chem better resolves the structure of the O3 plumes as observed by the TOPAZ lidar for all the three pollution events. At the surface, both models capture the large-scale MDA8 O3 enhancements across the SWUS on 16 June (Fig. 12b). The surface O3 enhancements on 2 and 29–30 June are relatively localized in southern California and the Las Vegas area (Fig. S11), and both models have difficulty simulating the observed peak MDA8 values (Fig. 4).

During 20–24 May, long-range transport of Asian pollution toward the WUS was observed via large-scale CO column observations with the Atmospheric Infrared Sounder (AIRS) on NASA's Aqua satellite (Fig. 13a). These Asian plumes traveled eastward across the Pacific for several days, reaching the west coast of the US on 23 May during the first FAST-LVOS IOP (23–25 May). The lidar measurements at NLVA on 24 May clearly showed high-O3 plumes (>70 ppbv) concentrated within layers of 1–4 km and 6–8 km altitude above the Las Vegas Valley throughout the day (Fig. 14a). Both GFDL-AM4 and GEOS-Chem capture the observed O3-rich plumes at surface–4 km and 6–8 km altitude above Clark County during this event. Elevated O3 at 6–8 km altitude reflects the long-range transport from Asia, as supported by concurrent enhancements in total and USB O3 in both models and by the large difference in O3 between the AM4 BASE simulation and the sensitivity simulation with Asian anthropogenic emissions zeroed out. Elevated O3 at 1–4 km altitude appears to be influenced by a residual pollution layer from the previous day; this plume was later mixed into the growing PBL (up to 4 km altitude), elevating MDA8 O3 in surface air on 24 May. Further supporting the impact from regional or local pollution below 4 km altitude, both models simulate much larger enhancements in total O3 (70–90 ppbv) than in USB O3 (∼50 ppbv).

On 24 May, MDA8 O3 approached or exceeded the 70 ppbv NAAQS at multiple sites in California, Idaho, Wyoming, and Nevada (Fig. 15a), likely reflecting the combined influence of regional pollution and long-range transport of Asian pollution. MDA8 O3 at four surface sites in Clark County was above 65 ppbv. More exceedances would have occurred if the level for the NAAQS were lowered to 65 ppbv. In parts of Idaho, Wyoming, and California where observed MDA8 O3 was higher than 60 ppbv, the contribution of Asian anthropogenic emissions as estimated by GFDL-AM4 was 8–15 ppbv (Fig. 15a), much higher than the springtime average contribution of ∼5 ppbv estimated by previous studies (e.g., Lin et al., 2012b), supporting the episodic influence from Asian pollution during this event. At several high-elevation sites in California such as Arden Peak (72 ppbv) and Yosemite National Park (70 ppbv), where observed MDA8 O3 exceeds the NAAQS level, the contribution of Asian pollution is approximately 9 ppbv. Ozone produced from regional and local anthropogenic emissions dominates the observed MDA8 O3 above 70 ppbv in the Central Valley of California.

The lidar measurements from 28 June show a fine-scale structure with a narrow O3 layer exceeding 100 ppbv at 3–4 km altitude during 08:00–14:00 PDT (15:00–21:00 UTC shown in Fig. 14b). An ozonesonde launched at 12:00 PDT also detected a high-O3 layer (∼115 ppbv) between 3.5 and 4 km altitude (not shown). This high-O3 filament appears to descend and mix into the PBL after 14:00 PDT (21:00 UTC), contributing to elevated O3 within the PBL in the afternoon. Both models are unable to represent this fine-scale transport event, possibly due to diffusive mixing of the narrow layer (Fig. 14b). We, therefore, focus on available airborne and in situ measurements to investigate the origin of this fine-scale O3 filament.

Our examinations of large-scale satellite CO column measurements reveal a migration during 23–27 June of high-CO plumes from Asia that arrived at the west coast of the US on 27 June (Fig. 13b). GFDL-AM4 estimates 5–6 ppbv contributions from Asian pollution over the WUS on 28 June (Fig. 15b), which do not represent a significant enhancement above the mean Asian contribution. Aircraft measurements above the Las Vegas Valley in the late morning showed collocated enhancements in CH4 and O3 coincident with low free-tropospheric water vapor values at 3–4 km altitude (Fig. 10b). In situ measurements at Angel Peak show concurrent increases in CO and O3 coincident with relatively dry conditions that are consistent with transported Asian pollution, but these increases did not appear until several hours after the fine-scale filament was entrained by the mixed layer (Fig. 6f). These observations indicate that the O3-rich plume appears to be unrelated to stratospheric intrusions. Aerosol backscatter measurements at NLVA show only a slight enhancement in backscatter within the elevated O3 layer on 28 June, in contrast to the thick smoke observed on 22 June when the Las Vegas Valley was influenced by fresh wildfires (Fig. 10). HYSPLIT and FLEXPART analyses presented in Langford et al. (2020) suggest a possible connection to the Schaeffer Fire (https://en.wikipedia.org/wiki/Schaeffer_Fire, last access: 20 September 2019) in the Sequoia National Forest in California. Another possible source is the fine-scale lofting of pollution from southern California followed by transport into the free troposphere over Las Vegas (Langford et al., 2010). This event further demonstrates the complexity of O3 sources in the SWUS. We recommend measurements of atmospheric compounds like acetonitrile (CH3CN, abundant in fire plumes) and methyl chloride (CH3Cl, abundant in Asian pollution) (Holzinger et al., 1999; Barletta et al., 2009) via aircraft and in situ platforms in future field campaigns in the region to help identify the sources of such high-O3 filaments.