Ozone concentration was available at China National Environmental Monitoring Centre (http://www.pm25.in, last access: 7 September 2021). Gridded hourly ozone concentration, potential vorticity, and specific humidity are available from the fifth generation of ECMWF atmospheric reanalysis global climate data (ERA5, https://cds.climate.copernicus.eu/cdsapp#!/search?type=dataset, last access: 7 September 2021), with a horizontal resolution of 0.1°×0.1° and 13 vertical layers from 1,000 to 100 hPa.

In this study, a three-dimensional Weather Research and Forecasting (WRF) model version 4.1.1 coupled with the Community Multiscale Air Quality (CMAQ) model version 5.3 is used to conduct the simulations in May 2015. The lambert conformal conic projection is applied with a spatial resolution of 36 × 36 km and 34 vertical layers. The initial and boundary conditions are provided by Climate Forecast System Reanalysis (CFSR) version 2 (Saha et al., 2014). The parameterization schemes used are shown in Table 1. The meteorological field simulated by WRF is processed by Meteorological Chemistry Interface Processor (MCIP) to provide a meteorological input to the Community Multiscale Air Quality (CMAQ) model. For CMAQ, carbon-bond version 6 (CB6) (Luecken et al., 2019) and Aerosol Module Version 7 (AERO7) (Appel et al., 2021) are used to represent the gas chemistry and aerosol scheme. The anthropogenic emissions are available from the Multi-resolution Emission Inventory for China (MEIC) (Li et al., 2017). Biogenic emissions are generated by the Model of Emissions of Gases and Aerosols from Nature (MEGAN 2.1) (Guenther et al., 2012), similar to our previous study (Gao et al., 2022; Ma et al., 2022). The initial and boundary conditions for CMAQ are derived from the MOZART-4 (Emmons et al., 2010). A chemically inert tracer (O3_BC) with the same molecular weight as ozone has been added to the model, as well as the lower and upper boundary conditions, to track the impact of stratospheric intrusion on ozone concentration at a lower altitude.

Previous studies have shown that the potential vorticity on the isentropic surface of 300–350 K can be used to investigate the impact of STT on ozone concentrations (Wang et al., 2010; Xu et al., 2015). Similarly, 340 K isentropic potential vorticity is utilized by Ding and Wang (2006) to understand the mechanism modulating the transport of O₃-rich air from the upper level to the surface. Therefore, the spatial distribution of isentropic potential vorticity at 340 K on 9–12 May 2015 is shown in Figure 1. On 9 May, a high value of potential vorticity appeared in Mongolia and Northern China. The continuous development and movement of the weather system trigger the formation of a PV trough eastward, passing the Qinghai–Tibet Plateau of China, including the location of Waliguan (black triangle) and eastern China on 10–11 May, and eventually moving further eastward on 12 May.

The spatial distribution of specific humidity and wind vector at 320 K on 9–12 May 2015 is displayed in Figure 2 to further delineate the influence of stratosphere air on the troposphere through the downward transport. The selection of 320 K mainly considers that the pressure is approximately 300 hPa at Mount Waliguan. The water vapor over the free troposphere tends to decrease along with the increase in altitude, and a stratospheric air intrusion can be inferred when a substantially low value of water vapor appears, values smaller than 0.08 g kg⁻¹ based on Shou et al. (2014). Consistent with the spatial distributions of PV, low specific humidity is concomitant with the atmospheric circulation over the areas with high PV values (Figures 2 vs. Figure 1). In particular, the white areas with specific humidity lower than 0.08 g kg⁻¹ are implicative of the occurrence of stratospheric air intrusion.

In order to evaluate ozone exchanges between stratosphere and troposphere, the tropopause should be determined first. There are usually two methods used in the tropopause definition and calculation: the thermodynamic and dynamic tropopause (Hoerling et al., 1991; Wang et al., 2010). The dynamic tropopause defined based on potential vorticity can well reflect the conservative feature of the atmosphere in adiabatic and frictionless circumstances (Yang and Lü, 2004), which is widely used by the community in STT studies. In this study, the dynamical tropopause is applied to investigate the impact of STT on ozone concentration in Mount Waliguan and eastern China, and PV = 2 PVU (1 PVU = 10^–6 m² s⁻¹ K kg⁻¹) is used in the determination of dynamic tropopause based on previous studies (Hongyue Wang et al., 2020; Yiping Wang et al., 2020).

The vertical profile of PV and ozone concentration at Mount Waliguan is displayed in Figure 3 on 9–13 May 2015, from the surface of Mount Waliguan (∼ 600 hPa) to 100 hPa. The dynamic tropopause delineated by the 2 PVU isoline (white dashed line in Figure 3A) indicates a deep invasion of stratospheric air from 200 hPa to almost 400 hPa on 10 May. Accompanied by the clear PV invasion, the ozone-rich air in the stratosphere was transported downward into the upper troposphere, resulting in a substantial ozone enhancement therein (Figure 3B). A similar phenomenon was observed in other locations, such as Hong Kong, where the ozone concentrations reached more than 80 ppbv on 24 March 2010 over an altitude slightly above the planetary boundary layer height (Zhao et al., 2021).

WRF-CMAQ simulations were conducted with an ozone tracer added by including diffusion, advection, and dry and wet deposition without chemical reactions to quantify the contribution of stratospheric ozone intrusion to the surface ozone. We first evaluate how well WRF-CMAQ may reproduce the meteorological conditions and ozone concentrations. The spatial distributions of potential vorticity and specific humidity from WRF are displayed in Figure 4. Compared to the spatial patterns from ERA5 (shadings in Figures 1, 2), the simulated results from WRF well reproduced the spatial propagation and movement of PV and specific humidity, indicative of the capability of the model in capturing the process of STT.

To further evaluate how well CMAQ reproduces the evolution of ozone, the cross-section, over the same latitude of WLG, of simulated ozone tracer was used to compare with that from ERA5 (Figure 5), and the comparable spatial and temporal propagations of ozone warranted the confidence in interpreting the stratospheric ozone intrusion. In the early stage of STT, the high ozone concentration was concentrated at the height of 200–400 hPa (Figures 5A,B). When STT occurred on 10 May, the high value of ozone invaded significantly toward the location of Waliguan at a longitude of 100°54′E (Figures 5C,D). With the development of trough of PV and invasion of ozone-rich air (Figure 1), the high ozone concentrations display consistent eastward movement (Figures 5E–H).

It is important to understand the extent to which the STT event may affect the ozone concentration near the surface. To this end, Figure 6 shows the spatial distribution of surface ozone, based on the ozone tracer method, every 2 hours in northwestern China on 10 May 2015, overlaid by the observed data from the China National Environmental Monitoring Centre in dots. The simulated results are generally consistent with the observations, despite the existence of biases such as overestimation in the southern flank of Qinghai, likely attributable to the biases from the boundary conditions inherited from the global chemistry model MOZART [Figure 2F in Yan et al. (2021)]. At 0:00 UTC (Figure 6A), ozone concentrations were high in the northern flank of Qinghai province and gradually shifted southeastward, covering almost half of the entire province. The STT event lasts for about 12 h, yielding the highest ozone concentrations of 89 ppbv.

In order to determine the contribution of stratospheric ozone intrusion to ground-level ozone, hourly ozone anomalies were calculated relative to the mean value in May 2014–2018, and the results for every 2 hours are shown in Figure 7. It nicely delineates the spatial coverage expansion along with temporal evolution from the northwest flank of WLG (02:00 UTC, Figure 7B), propagating southeastward and invading close to half of the Qinghai province after 10:00 UTC (Figure 7F). The mean contributions based on ozone anomalies from STT are approximately 10–20 ppbv, with a maximal contribution of 32 ppbv during this event. Consistently, based on the chemical transport model GEOS-Chem, Lu et al. (2019) found an ozone enhancement of 10–20 ppbv in Qinghai due to STT during May and June in 2016 and 2017 [second panel in the third row of Figure 4 in Lu et al. (2019)]. In addition, the influence of STT may extend to even North China Plain. Although the overall scale is much smaller, the enhancement of 2–3 ppbv may deserve much attention in the future. It is necessary to further clarify and comprehensively study the regulatory effect of STT in ozone pollution-prone areas.