Stratospheric contribution to the summertime high surface ozone events over the western united states

We analyzed daily surface $\mathrm{O_3}$ and stratospheric $\mathrm{O_3}$ tracer (denoted by $\mathrm{O_3}$S) from 1995 to 2014 summer months (June–August, JJA) using the Whole Atmosphere Community Climate Model version 6 (WACCM6) of the Community Earth System Model version 2 (CESM2). It is the high-top version of the Community Atmosphere Model version 6 (CAM6), integrating the atmospheric physics and chemistry from the surface to nearly 140 km. The WACCM6 uses the same atmospheric physics as CAM6. The chemical mechanism includes comprehensive troposphere, stratosphere, mesosphere and lower thermosphere chemistry, described by Emmons et al (2020). The standard emissions are based on anthropogenic and biomass burning inventories specified for the Coupled Model Intercomparison Project 6 (CMIP6). WACCM6 is coupled to the interactive Community Land Model version 5 (CLM5), which handles dry deposition. The simulations shown here are fully coupled ocean-atmosphere experiment, and feature $0.95^\circ\times1.25^\circ$ (latitude× longitude) horizontal resolution and 70 layers, with ∼1.2 km vertical resolution above the boundary layer to the lower stratosphere. We consider WACCM6 is well suited for studying the transport during stratospheric intrusion because that 1. the atmospheric chemistry and deposition scheme for $\mathrm{O_3}$ are well represented and tropospheric $\mathrm{O_3}$ simulations are improved in comparison to observations (Emmons et al 2020); 2. WACCM6 is able to effectively reproduce the observed wind and temperature climatologies as well as stratospheric variability (Gettelman et al 2019); 3. the twenty-year simulation with daily output of $\mathrm{O_3}$ and $\mathrm{O_3}$S provides us with large samples to study. Detailed model formulations, descriptions, and evaluations can be found in Gettelman et al (2019) and Emmons et al (2020).

To quantify the stratospheric contribution to high surface $\mathrm{O_3}$ events, we study an artificial tracer, $\mathrm{O_3}$S, for $\mathrm{O_3}$ originating in the stratosphere, which is implemented in a manner as described in Tilmes et al (2016). $\mathrm{O_3}$S experiences the same loss rate as $\mathrm{O_3}$ in the troposphere but is not affected by $\mathrm{NO_x}$ photolysis as defined by the Chemistry-Climate Model Initiative (CCMI). Earlier studies have shown that the diagnostics of deep stratospheric intrusions are insensitive to the choice of tropopause definition (Yang et al 2016) or $\mathrm{O_3}$S tagging methods (Lin et al 2012).

Maximum Covariance Analysis (MCA), known as Singular Value Decomposition (SVD) analysis, is a useful tool for detecting coherent patterns between two different geophysical fields (e.g. (Bretherton et al 1992, Hurrell 1995, Dai 2013)). In this study, we isolate pairs of spatial patterns and corresponding time series by performing the eigenanalysis on the temporal covariance matrix between $\mathrm{O_3}$ anomalies and $\mathrm{O_3}$S anomalies at the lowest model level (detailed calculations are discussed in (Bretherton 2015)). Daily anomalies of $\mathrm{O_3}$ and $\mathrm{O_3}$S at the lowest model level are derived with respect to the twenty-year (1995–2014) mean of that day. The considered domain in this study is 20^°–50^°N, 70^°–140^°W.

Here we first evaluate the WACCM6 simulations against observations from the Tropospheric Ozo-ne Assessment Report (TOAR) (Schultz et al 2017). Figure 1(a) is the 1995–2014 mean of 'all$_mean' variables from the TOAR $5^\circ\times 5^{\circ}$ (latitude× longitude) monthly median products. High $\mathrm{O_3}$ values ($\sim45\,\mathrm{nmol\ mol^{-1}}$) are seen over the western U.S. in TOAR measurements. For WACCM6, we have gridded the simulations onto the horizontal grid of TOAR data and calculated median $\mathrm{O_3}$ concentrations using the same metrics to guarantee an apple-to-apple comparison (see figure 1(b). Generally a good agreement is found between the model and observations over the Central U.S. and most of the West Region, with percentage differences within 15% (see figure S1 (https://stacks.iop.org/ERL/15/1040a6/mmedia)). However, WACCM6 overestimates the surface median $\mathrm{O_3}$ over the southeast U.S. by 55% ($\sim15\, \mathrm{nmol\ mol^{-1}}$), which is consistent with the evaluation in Emmons et al (2020). The biased O₃ over the eastern U.S. is a long existing problem that can have various reasons starting with still insufficient complexity of the chemistry, but also the model resolution, deposition scheme, etc (Schwantes et al 2020).

Zoom In Zoom Out Reset image size

Figure 1(c) and (d) shows the 1995–2014 mean of daily $\mathrm{O_3}$ and $\mathrm{O_3}$S simulations in $0.95^\circ\times1.25^\circ$ (latitude× longitude) resolution at the lowest model level, respectively. Figure 1(c) is similar to figure 1(b) and shows that high $\mathrm{O_3}$ are concentrated over the western U.S. As shown in 1(d), strong stratospheric impact, ranging from 6 to 10 nmol/mol, is found over the Canada-U.S. border and the Western States, including southern British Columbia, Washington, Oregon, Idaho, Montana, Wyoming, California, Nevada, Utah, Arizona, and Colorado. Our model simulation is in agreement with previous observational studies which reported that deep stratospheric intrusions preferentially occur in the West and the Intermountain West (Brioude et al 2007, Bourqui and Trépanier 2010, Ambrose et al 2011, Lefohn et al 2011, Lefohn et al 2012, Lin et al 2012, Langford et al 2012, Langford et al 2015, Langford et al 2017, Clark and Chiao 2019). A minimum stratospheric impact occurs in the Southeast, in good agreement with results in Lin et al (2012). These features are also consistent with simulations by the Geophysical Fluid Dynamics Laboratory global chemistry-climate model (Clifton et al 2014). Additionally, the simulated stratospheric impact in late spring (figure S2) agrees well with that shown in figure 11(c) and (d) from Lin et al (2012).

We now estimate the importance of $\mathrm{O_3}$S on surface $\mathrm{O_3}$. We calculate the percentage of variance (r²) of surface $\mathrm{O_3}$ explained by surface $\mathrm{O_3}$S when surface $\mathrm{O_3}$ exceeds the 95th percentile (p95, hereinafter) value at each grid point during 1995–2014 summer months using daily average $\mathrm{O_3}$ and $\mathrm{O_3}$S (see figure 1(c) and (d)). The linear trend has been removed before the calculation. As suggested in figure 1(e), high peaks are located over Wyoming, southern Texas, and the Four Corners area (Colorado, Utah, Arizona, and New Mexico). We see in Arizona, for example, $\mathrm{O_3}$S can explain as high as 54% of surface $\mathrm{O_3}$ variance during days in which surface $\mathrm{O_3}$ exceeds p95. We compare our results with those from prior observational studies. Lefohn et al (2012) studied stratospheric intrusion events associated with daily maximum hourly $\mathrm{O_3}$ in exceedance of $50\, \mathrm{nmol\ mol^{-1}}$ at both rural and urban monitoring sites of the U.S. Statistically significant relations are found over the southern Texas, Arizona, Colorado, Utah, Nevada, California, and Wyoming in summer months during 2006–2008. Overall, our model simulations support observational findings that stratospheric intrusions coincident with $\mathrm{O_3}$ preferentially take place in the West and Intermountain West during summertime. We also quantify the stratospheric impact when surface O₃ exceeds p95 at each grid cell and find that in regional average over the western U.S. ($30^\circ \mathrm{N}-50^\circ$N, 100^°W–125^°W), the stratospheric contribution is 18.4% (not shown).

We employ MCA to investigate how surface $\mathrm{O_3}$ is related to $\mathrm{O_3}$S reaching the surface on daily basis. The first two leading modes are summarized in figure 2. Together, these two modes explain 60% of the covariance pattern over the domain of interest. The first mode shows large positive covariances over the eastern Pacific, western and northern tier of the U.S. (figure 2(a). The second mode shows dipole structure with $\mathrm{O_3}$ and $\mathrm{O_3}$S surplus over the western U.S. and deficit over the east (figure 2(b). Surface level $\mathrm{O_3}$S anomalies are found to lag the 200–500 hPa $\mathrm{O_3}$S anomalies by two days (not shown), suggesting the downward influence. No significant lead-lag relationships are found either between the expansion coefficient time series (loadings) of $\mathrm{O_3}$ and $\mathrm{O_3}$S at the surface or between the first two SVD modes.

Zoom In Zoom Out Reset image size

We identify ∼20 days per month (1404 days in twenty years for the first two leading modes) when both expansion coefficient time series are greater than 0, indicating that surface O₃ concentration increase is coincident with O₃S reaching the surface. Lefohn et al Lefohn et al (2011), Lefohn et al (2012) investigated the frequency of surface O₃ enhancements that are associated with stratosphere-to-troposphere transport down to the surface across the U.S. and reported that the average number of days per month ranges from 16 days to 23 days at monitoring sites in the West and Intermountain West in summer months. Our results of intrusion frequency are in remarkable agreement with their results.

Next, we define high surface O₃ events caused by stratospheric intrusions when both loadings exceed 1.5 standard deviations, as marked by black circles in figure 2. Each event lasts for a few days. Based on this definition, composite analyses according to high surface $\mathrm{O_3}$ events are carried out using $\mathrm{O_3}$S daily data. The composited patterns are not sensitive to the thresholds we chose for the analyses. Schematics in figure 3 show the large amplitude composited $\mathrm{O_3}$S structures and near-surface circulations corresponding to both modes. We use neon yellow color to highlight the area with $\mathrm{O_3}$S larger than $120\, \mathrm{nmol\ mol^{-1}}$, which can be treated approximately as the tropopause (Yang et al 2016). During Mode 1, depressed tropopause height followed by enhanced $\mathrm{O_3}$S can be found around 50^°N, 120^°W (figure 3(a). Since stratospheric air contains higher values of potential vorticity (PV) and $\mathrm{O_3}$, the intrusion of the tropopause tends to replace the tropospheric air by ozone-rich stratospheric air with large PV (Danielsen 1968, Mote et al 1991, Wimmers et al 2003). Transport of $\mathrm{O_3}$S to low levels is tied to stratospheric intrusions and strong subsidence in the troposphere. Figure 4(d) shows a map of composited means of 500 hPa vertical velocity (ω) anomalies on the day of the events corresponding to Mode 1. The intrusion is associated with intensified subsidence on the U.S.-Canada border. Coherently, enhanced $\mathrm{O_3}$S on the northern tier can be seen throughout the troposphere, while the maximum over the Pacific appears below 700 hPa (figure 3(a)). We conduct composited analyses on anomalies of 860 hPa geopotential height as well as surface winds according to events of Mode 1. A quadrupole pattern of low-level geopotential height anomalies is seen over the North American continent. In contrast to the positive correlation between the upper-level cyclonic vorticity and $\mathrm{O_3}$S (figure S3a), low-level geopotential height and $\mathrm{O_3}$S reaching the surface are strongly anti-correlated (figure 3(b)). Dry air (figure S3b) with high PV value (figure S3a) descends on the northern tier of the U.S. as a result of stratosphere-to-troposphere transport by intensified subsidence. Warm and moist tropospheric air is seen (figure S3b–c) downstream (east) of the surface low. In addition to the anomalous descent, we also see a strengthening of northeasterly along the west coast near the surface (figure 3(b)). This intensified easterly associated with stronger anticyclone facilitates the horizontal transport towards the subtropical Pacific in the mid-to-lower troposphere (figure 3(a)).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

During Mode 2 (figure 3(c)), the intrusion occurs around 40^°N, 115^°W. Surface $\mathrm{O_3}$S maximum is concentrated over the Intermountain West, such as Colorado, Arizona, Utah, Nevada. Different from Mode 1, a dipole pattern of the low-level geopotential height anomalies is seen over the continental U.S. Large values of negative geopotential height anomalies are observed above the intermountain regions (figure 3(d) aligned with cold dry air and positive PV anomalies (figure S4a–c) subsiding west of the surface low pressure while rising warm and moist air occur on the east (figure 4(h) and figure S4b–c).

During both modes, the $\mathrm{O_3}$S reaching the surface can be 10–$18\, \mathrm{nmol\ mol^{-1}}$ (see surface contours of figure 3(a) and 3(c), respectively). In regional average over the western U.S. (30^°N–50^°N, 100^°W–125^°W, the stratospheric contribution is 11 nmol/mol. The amplitude agrees well with the observational record from the California Baseline Ozone Transport Study (CABOTS), in which they found that stratospheric contribution is 10–$20\,\mathrm{nmol\ mol^{-1}}$ to the surface over Northern California during an intrusion case in August 2016 (Clark and Chiao 2019). We also see ∼10 days per summer season when high surface O₃ events associated with the first two SVD modes occur, with variation across years.

Next we examine the dynamical mechanism underlying the stratospheric intrusion events associated with Mode 1 and Mode 2. Figure 4 shows the time evolution of 500 hPa vertical velocity (ω) anomalies and 200 hPa zonal winds for the composites based on high $\mathrm{O_3}$ and $\mathrm{O_3}$S events in Mode 1 (figures 4(a)–(d) and Mode 2 (figure 4(e)–(h), respectively. The positions of the upper-level jet axis are marked by dashed lines. The downstream development (Chang 1993), characterized by eastward propagation of the wave systems, can be seen in both modes.

More specifically, for Mode 1, the large descending anomaly is centered at 50^°N, 160^°W on day −6 (figure 4(a). After 2 days, it shifts eastward following the continuous jet (figure 4(b). On day −2, the descending center manifests near the west coast (figure 4(c), consistent with the result that relationship of 500 hPa $\mathrm{O_3}$S with the principal component of Mode 1 is maximized about two days ahead of surface $\mathrm{O_3}$S peaks (not shown). During day −2 to day 0, the subsidence stays close to the U.S.–Canada border, embedded within the jet stream (figure 4(d). On day 0, the strong descent on the northern tier of the U.S. continuously facilitates the downward transport of $\mathrm{O_3}$S toward the surface (figure 3(a).

Regarding Mode 2, large disturbance center can be seen on day −6 near the eastern Pacific (figure 4(e)). From day −4 to day −2, the descending center continues to grow near 135^°W (figure 4(f)–(g)). The downstream side strengthens remarkably near 125^°W where the jet breaks. On day −2, strong anomalous subsidence reaches the western U.S. whereas ascending anomalies are triggered on the east. It is the enhanced subsidence that induces the stratospheric intrusion and downward transport of stratospheric $\mathrm{O_3}$ (figures 4(h) and 3(c)).

During boreal summer, the dominant upper-level circulation consists of westerly jet north of 40^°N, together with frequent eastward traveling baroclinic waves (White 1982). The baroclinic waves commonly exhibit life cycles of baroclinic growth and barotropic decay along the storm track regions (Simmons and Hoskins 1978, Blackmon et al 1984). Observational and modeling studies have revealed that the baroclinic disturbances leave imprint on atmospheric composition. Aircraft experiments provided evidence that high amounts of stratospheric radioactive debris and ozone were drawn into the troposphere by midlatitude storms (Danielsen 1968). Schoeberl and Krueger (1983) and Mote et al Mote et al (1991) identified the coherent fluctuations in total $\mathrm{O_3}$ and medium scale waves along the wintertime oceanic storm track regions using satellite data. Stone et al (1996) found baroclinic wave features using upper tropospheric water vapor measurement. General circulation model was capable of representing stratosphere–troposphere exchange associated with baroclinic waves in midlatitudes (Mote et al 1994). In our study, baroclinic wave system and its resulting circulation is also found important for the occurrence of deep stratospheric intrusions over the western U.S. in summertime (Sprenger and Wernli 2003).

In both modes, we see that the wave trains originate over the baroclinically unstable central Pacific, grow by baroclinic conversion, radiate energy downstream through ageostrophic geopotential fluxes, and dissipate over the jet exit near the western U.S., as discussed in previous studies (Chang and Orlanski 1993, Chang 1993). The upper-level jet stream acts as a wave guide, constraining the baroclinic system tightly along its core. From day −2 to day 0 in both modes, the intensified descents at 500 hPa, which corresponded to the downward transport of $\mathrm{O_3}$S, are evident near the western U.S. (115^°–120^°W). These two modes mainly differ by the location of the jet core. For Mode 1, the jet is almost continuous across the North Pacific and the averaged jet axis is located at 46^°N. The passage of the wave trains takes about 8 days, starting from central North Pacific (40^°N, 170^°W, see figure 5(a), propagating eastward to the U.S.–Canada border (45^°N, 120^°W). For Mode 2, the polar jet and the subtropical jet are well separated. The successive baroclinic waves develop rapidly, beginning with a large descending anomaly at the jet exit (40^°N, 140^°E) from day −6, ending with a trough on the west-and-a ridge on the east over the continental U.S. (figure 3(d)).

Zoom In Zoom Out Reset image size

Finally, Hovmöller diagrams are constructed to summarize the eastward propagating baroclinic systems that occur prior to high surface $\mathrm{O_3}$ and $\mathrm{O_3}$S events. The longitude-time plots of 500 hPa ω anomalies along the latitudinal band of the jet axis are depicted for both modes (figure 5). More specifically, figure 5(a) is the ω anomalies along the jet axis as marked in figure 4(a)–(d) while figure 5(b) is plotted along 38^°N for simplicity. We can see a sequence of downstream developing wave train originating from the North Pacific and propagating along the upper-level jet. Eastward propagation of Mode 2 is less clear than that of Mode 1. The descending anomalies begin to amplify near the jet exit two days before the deep intrusions over the western U.S. Similar conclusions are also found using upper-level meridional velocity (not shown). The results are consistent with evolving of baroclinic waves, and are in good agreement with the wave train signature diagnosed in Lim and Wallace (1991) and Chang (1993), in terms of their structure, magnitude, and traveling speed. To sum up, the large $\mathrm{O_3}$ events caused by deep intrusions are associated with eastward propagating baroclinic systems tied closely to the North Pacific storm tracks, with enhanced wave amplitudes and descents over the jet exit region near the western U.S.