Introduction

Compelling evidence has suggested that high ozone (O3) concentrations in the surface layer can be harmful to human health and vegetation growth1. As known, tropospheric O3 originates from two sources: photochemical production within the troposphere and dynamical injection from the stratosphere. Though the injected stratospheric O3 is estimated to account for only 5-10% of the tropospheric O3 sources2, case studies of stratospheric intrusions (SI) to the troposphere have documented how SI occurred and impacted tropospheric and even surface O33,4,5. Events of deep and fast stratospheric intrusions to the surface (SITS) can lead to high-O3 episodes, inducing severe O3 pollution3b), hindering the formation of deep PBL for downward transport of stratospheric air to the surface15,16. The stratospheric O3 injected to the troposphere averaged over China, deriving from the Trajectory-mapped Ozonesonde dataset for the Stratosphere and Troposphere (TOST)51 data, also indicates a declining tendency over 2015-2022 (Supplementary Fig. 4). Supplementary Fig. 5 presents the estimated amounts of stratospheric O3 in the surface layer from the MERRA-2 GMI global atmospheric chemistry model52, which also shows a decreasing trend of stratospheric contributions to surface O3.

Discussion

Estimates of the stratospheric influences on surface O3 are necessary for making effective mitigation policies, since these inputs of natural stratospheric O3 can substantially enhance the risk of O3 pollution episodes and partially determine the floor value for air quality managements. In a short term, SITS-induced O3 has non-negligible significance for transient high-O3 episodes, given its large fractions of surface O3 budget (30-45%) and high risks of O3 exposure to harmful-level concentrations during SITS periods in affected areas. While the absolute stratospheric influences are highest in March and October, special attention to O3 pollution control should be paid in spring and summer when extra SITS-induced O3 inputs, plus the high background O3, promote possibility to exacerbate O3 pollution beyond the WHO and national standards. In 10% of the SITS cases, surface O3 can be elevated by over 40 ppbv, setting alarms for possible severe O3 pollution in affected areas. On the other hand, SITS could synchronously reduce concentrations of other air pollutants, including CO, NO2, and SO2. Such stratospheric perturbation can also substantially enhance the oxidation capacity of tropospheric air. On the annual basis, detectable O3 with stratospheric origins consists of 1.6–2.2% of surface O3 in China, implying that O3 pollution mitigation over long terms in China should mainly focus on surface O3 variations through photochemical reactions under the influence of meteorology and anthropogenic emissions. In this study, we conservatively estimate the stratospheric influences by only including the dynamically injected stratospheric O3 associated with deep, direct, and fast SITS events. The influences of aged stratospheric air injected into the troposphere from the stratosphere are spared, since such influences could not be easily identified from surface measurements. A combination of satellite remote sensing technology and deep machine learning methods in future work can help solve these issues. The chemically-induced O3 production due to stratospheric perturbation may also contribute to surface O3 variations in the presence of nonlinear chemical reactions during SITS53,54. The above factors can amplify the influences of the stratosphere especially in transient surface O3 pollution events, and hence enhance the impact of O3 on human health and crop yield.

Methods

Screening SITS based on comprehensive surface observations

Taking advantage of surface gaseous pollutant measurements, e.g., O3, CO, NO2, and SO2, with a high spatial and temporal resolution, here we develop a methodology of detecting SITS events over large areas and for long periods, based on and further refined from our previous studies9,14. This method is effective in detecting deep, direct, and fast SI retaining stratospheric properties, such as “O3-rich and CO-poor”3,4,15. In this study, we focus on such SITS events, while aged stratospheric air that has reached the surface but lost its stratospheric properties is not considered. Relying on the characteristics of stratospheric air reaching the surface (richer O3 and poorer CO relative to tropospheric air), we identify a SITS event based on hourly concurrent O3 and CO measurements at the surface based on the following points.

  1. 1.

    Distinct upward and downward spikes of O3 and CO indicators, respectively. The hourly measurements of stratospheric indicators (O3 and CO) are screened to filter out their distinct spikes, e.g., the sharp increase in O3 and decrease in CO simultaneously, a unique indication for air with recent stratospheric origins. Hourly variations of O3 and CO concentrations throughout the year are calculated site by site and year by year. The 95th percentile of the O3 rising rate and 5th percentile of the CO decline rate in each year are chosen to identify those sudden and sharp spikes when stratospheric air initially reaches the surface55. The synchronous appearance of extreme O3 increase and CO decrease could help isolate the sudden surface O3 change due to SI from that due to O3 transport or photochemical processes. As shown in Supplementary Fig. 6, the two parameters simultaneously determine the start timing of a SITS event (SITS_start).

  2. 2.

    The large departures of O3 and CO from their normal values. The intruded stratospheric air contains higher O3 than that in the troposphere; therefore, surface O3 with additional inputs in SITS events is supposed to increase from its normal values. To minimize the blurring of photochemically produced tropospheric O3, we consider that the O3 concentrations at the SITS_start hour should exceed the seasonal mean value during noontime (\({\bar{{O}_{3}}}^{{noon}}\), 1st O3 criterion) when photochemical reactions are active. Simultaneously, the CO concentrations during the SITS should decline below their seasonal mean value (CO criterion). These criteria could also help remove the occasions that could be falsely identified when O3 is transported downward from the residual layer, an O3-rich “reservoir” containing photochemically produced O3 in the preceding day56,57. Due to the mixing with tropospheric air and chemical sinks of O3, the properties of the intruded stratospheric air subside over the time58. When the O3 concentrations fall back to their seasonal mean values (\({\bar{{O}_{3}}}^{{season}}\), 2nd O3 criterion) or the CO concentrations rebound over the CO criterion, stratospheric air is not distinguishable from the tropospheric air and hence the SITS events end (SITS_end; referred to the case illustrated in Supplementary Fig. 6).

Provided with the start and end timing of SITS events, we estimate the amounts of injected stratospheric O3 reaching the surface by integrating the excess of O3 concentrations above their reference baselines (the seasonal means at the corresponding hour) during the SITS events (see details in the next section). At a given station for a period, such as a month, the number of SITS occurrences, the length of time between SITS_start and SITS_end averaged over all SITS events in the period, and the averaged excess of O3 concentrations above the baselines are regarded as the frequency, duration, and intensity of the SITS at that station for that period.

Similar to the definition of a chemical tropopause59, we rely on the variations in atmospheric chemical constituents O3 and CO, rather than some dynamic indicators, to define the frequency, duration, and intensity of SITS events. These definitions are referred throughout this manuscript. Both O3 abundance in the lower stratosphere and frequency of deep stratosphere-to-troposphere processes primarily determine the injected amounts of stratospheric O3 into the troposphere60. When intruded into the troposphere, stratospheric O3 can be strongly mixed with tropospheric air and be chemically destroyed, responding to the complicated dynamical and chemical processes in the troposphere, especially in the PBL. Therefore, assessing the stratospheric contribution to surface O3 depends on not only the detailed information of SITS (e.g., their frequencies and magnitudes), but also the varying tropospheric environments that control the fate of injected stratospheric O3 (SITS duration).

Although stratospheric air is also characterized with low RH, RH is not selected as an indicator in our detect algorithm. This is because RH is inversely related to temperature. Low RH may also appear when air parcels descend from higher altitudes to the lower troposphere experiencing adiabatic warming. The air parcels can also experience various atmospheric moisture conditions on their way to the surface, so RH of the air parcels is less conservative than O3 and CO3,8. In addition, concurrent RH measurements are usually unavailable in air quality monitoring stations in China.

We have developed this SITS methodology with a goal of being objective, robust, and accurate, i.e., reducing the commission and omission errors as much as possible. We have inclined to be conservative and set the detecting criteria rather strictly. For example, we assure that SITS would enhance surface O3 concentrations, i.e., as long as surface O3 concentrations are not above the background value, SITS stops. In this way, the detected SITS events are highly likely to be the cases, while some weak SITS events may be omitted.

Estimation of contributions of SITS to surface O3

The contributions of SITS to surface O3 are estimated event by event and station by station. The hourly mean surface O3 concentrations (\({\bar{O}}_{3}^{h}\); where h = 1, 2, 3,…24) are calculated by averaging O3 observations at each of the 24 hours in each season based on station-level observations in each year. The \({\bar{O}}_{3}^{h}\) values are taken as reference baselines to measure the departure of O3 concentrations from their baselines during SITS periods. Provided with the start and end timing of SITS events, we integrate the excess of O3 above the reference baselines during SITS periods (unit: ppbv*hour), and take it as the amount of injected stratospheric O3 (\({O}_{3}^{{strat}}\)) in each SITS event14,28,31,48:

$${O}_{3}^{{strat}}={{\int }}_{{SITS}{{\_}}{start}}^{{{SITS}{{\_}}{end}}} ({O}_{3}^{h}-\,{\bar{O}}_{3}^{h}){dt}$$
(1)

where \({O}_{3}^{h}\) denotes the in situ hourly O3 observations at hour h, and the \({\bar{O}}_{3}^{h}\) represents the baseline O3 concentrations at the same hour. The differences between \({O}_{3}^{h}\) and \({\bar{O}}_{3}^{h}\) are summed over the SITS period with a temporal resolution of 1 hour (i.e., dt = 1 hour).

The sum of O3 concentrations during each SITS event and its corresponding month (unit: ppbv*hour) are calculated following Eqs. (2) and (3), respectively:

$${{O}_{3}}_{{SITS}}^{{sum}}={{\int }}_{{SITS}{{\_}}{start}}^{{{SITS}{{\_}}{end}}}{O}_{3}^{h}{dt}$$
(2)
$${{O}_{3}}_{{month}}^{{sum}}={{\int }}_{{month}{{\_}}{start}}^{{{month}{{\_}}{end}}} {O}_{3}^{h}{dt}$$
(3)

The ratio of stratospheric O3 (\({O}_{3}^{{strat}}\)) to the sum of O3 concentrations during each SITS event (\({{Ratio}}_{{SITS}}\)) is given by Eq. (4):

$$R{{atio}}_{{SITS}}=\frac{{O}_{3}^{{strat}}}{{{O}_{3}}_{{SITS}}^{{sum}}} * 100\%$$
(4)

The sum of stratospheric O3 inputs during all SITS events in a month (\({{O}_{3}}_{{month}}^{{strat}}\), unit: ppbv*hour) is calculated by Eq. (5), given the number of SITS events in the month being N:

$${{O}_{3}}_{{month}}^{{strat}}={\sum }_{i=1}^{N}{O}_{3,i}^{{strat}}$$
(5)

Finally, the ratio of stratospheric O3 to the sum of O3 concentrations in the corresponding month (\({{Ratio}}_{{month}}\)) is given by Eq. (6):

$$R{{atio}}_{{month}}=\frac{{{O}_{3}}_{{month}}^{{strat}}}{{{O}_{3}}_{{month}}^{{sum}}} * 100\%$$
(6)

The time series (2015–2022) of SITS-induced O3 and its ratio to overall surface O3 concentrations during SITS periods and the entire month are showed in Supplementary Fig. 1.

Validations of the SITS detection algorithm

As SITS appears as rare events in local areas, it is important to verify the reliability of our detection algorithm in order to address the impact of deep SI on surface O3. We previously published a detailed analysis of two SITS events that occurred in China, which were selected from the large samples of SITS events9,14. The stratospheric origins and transport pathways of the two SITS cases were revealed by means of surface air pollutant observations, vertical profiles of RH and O3, PV evolution, and backward trajectory simulations. The general characteristics of SITS detected with our algorithm, such as their seasonality and contribution to surface O3, are in good agreement with the observed deep SI climatology by Stohl et al.24, Elbern et al.28 and Cristofanelli et al.31. The detected frequency of SITS is further compared with published observational studies. For example, based on multiple stratospheric tracers including RH, CO, and cosmogenic radionuclide 7Be, Lin et al.48 identified 14 SI days during a 13-month campaign in a high-elevation station (3380 m asl) located in Mt. Hehuan of Taiwan (Fig. 1a). The annual frequencies of SITS in Panzhihua (16 per year) and Chuxiong (14 per year) detected in the present study agree reasonably with those at Mt. Hehuan which is with a similar latitudes and altitudes (Fig. 1a). Using a combination of stratospheric tracers including RH, potential vorticity (PV), 7Be and the tropopause height anomaly, Cristofanelli et al.31 reported that there were 33 days (average 5.5 days per year) of deep and direct SI which were characterized by distinct stratospheric properties at Mt. Cimone (2165 m asl) in Italy over 1998-2003. We apply the SITS detection algorithm to the O3 and CO measurements collected in Mt. Cimone during 2013–2016, and find a total of 26 direct SI events (average 6.5 days per year). The results from our detection algorithm are in line with these SI studies shown above and indicate the feasibility of using sudden and sharp spikes of O3 and CO to identify SI reaching the surface.

The origins of SITS events are investigated with backward trajectory simulations of 10 days over selected cities (Supplementary Fig. 7; see details of the backward trajectory simulations in the following section). We select Panzhihua as an example where the highest SITS frequency is detected. The cities Bei**g and Fuzhou are also selected to examine SITS occurred in northern and southern China, respectively. The trajectory analysis indicates that the majority of air parcels at the surface during the detected SITS events originated in the upper troposphere and lower stratosphere (UTLS; above 400 hPa), i.e., 96% in Panzhihua, and 100% in Fuzhou and Bei**g. As another piece of evidence, the ensemble RH profiles during the 33 SITS events over Bei**g show substantial dryness characterized by RH < 30% in the PBL and near the ground level15, suggesting the dry stratospheric air has descended into the surface. In addition to these selected cities, we further analyze the ensemble of backward trajectories associated with the detected 27,616 SITS events (Supplementary Fig. 8). We evenly divide every trajectory from the beginning to the end into three travel stages. The height, PV, and O3 concentrations in each stage are extracted from MERRA-2 reanalysis data. The air parcels of SITS initially reside in 300-200 hPa, where PV values exceed 2 PVU (an iso-surface representing the dynamical tropopause) and O3 concentrations are larger than 250 ppbv61,62, showing prominent stratospheric origins.

Backward trajectory simulations and MERRA-2 reanalysis data

Backward trajectories are simulated to check the origins of airmass of detected SITS events (Supplementary Fig. 7 and Supplementary Fig. 8) using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model. HYSPLIT is developed by National Oceanic Atmospheric Administration’s (NOAA)63 (https://www.arl.noaa.gov/hysplit). The 10-day backward trajectories are driven by the meteorological data from the Global Forecast System (GFS) with a resolution of 0.25°. The PV values and O3 concentrations along the trajectory are extracted from the Modern-Era Retrospective analysis for Research and Applications version 2 (MERRA-2) reanalysis data. The MERRA-264 reanalysis data have a spatial resolution of 0.5° latitude × 0.625° longitude with 72 model levels (https://gmao.gsfc.nasa.gov/reanalysis/MERRA-2; DOI: 10.5067/WWQSXQ8IVFW8).

Surface observational data and stratospheric O3 tracer data

The present study is mainly based on analysis of hourly ground-based measurements of O3, CO, SO2, and NO2 concentrations at more than 1,600 stations in Chinese cities. For O3 and CO, they are measured with a CO analyzer (Thermo Fisher Model 48i) and an O3 analyzer (Thermo Fisher Model 49i). The detection limit (precision) for Model 48i and Model 49i are 0.04 ppmv (±0.1 ppmv) and 0.5 ppbv (±1 ppbv), respectively. The data are from the public website of the Chinese Ministry of Ecology and Environment (MEE) (https://english.mee.gov.cn/).

To explore the variations in stratospheric O3 during the study period (Supplementary Fig. 3a), the stratospheric O3 profile observations are acquired from the Stratospheric Water and OzOne Satellite Homogenized (SWOOSH) dataset65 (https://csl.noaa.gov/groups/csl8/swoosh). SWOOSH provides a merged record of stratospheric O3 on the basis of a number of limb sounding and solar occultation satellites from 1984 to the present.

To investigate the thermal inversion variations over China during 2015-2022 (Supplementary Fig. 3b), radiosonde observations in China are processed, which are available from https://www.ncei.noaa.gov/products/weather-balloon/ integrated-global-radiosonde-archive.

The Trajectory-mapped Ozonesonde dataset for the Stratosphere and Troposphere (TOST)49,66 is a 3-dimensional O3 dataset derived from ozonesondes at over 100 stations using a trajectory-based map** methodology with the HYSPLIT model. The thermal tropopause height is determined for each O3 profile, and the stratospheric O3 distribution is mapped for the O3 with stratospheric origination along the trajectory paths. All O3 values along the trajectory paths are binned into grids of 5° × 5° × 1 km (latitude, longitude, and altitude) from sea level to 26 km in each month. TOST has been validated against independent ozonesondes and widely used in global O3 climatology studies67. In the present study, we further extend the TOST data by conducting 10-day forward trajectories simulations over 2015-2021 (Supplementary Fig. 4).

Simulations from the MERRA-2 GMI global chemical transport model are analyzed for long-term variations in stratospheric O3 inputs to the surface-layer during the period (Supplementary Fig. 5). The MERRA-2 GMI (Global Modeling Initiative’s) model with stratosphere-troposphere chemical mechanisms is driven by MERRA-2 meteorology including winds, temperature and pressure52 (https://acd-ext.gsfc.nasa.gov/Projects/GEOSCCM/MERRA2GMI/). This model is run at a MERRA-2 native horizontal resolution of ∼50 km with 72 vertical levels. The model applies a stratospheric O3 tracer to diagnose the stratospheric O3 influence on the troposphere.