The USTC Ozone Lidar deployed at Yangbajing, Tibet (29° N, 99° E) was developed in-house by the University of Science and Technology of China. The system enables simultaneous profiling of ozone in the troposphere and stratosphere with high spatial and temporal resolution, providing a particularly valuable observational constraint on ozone variability over the Tibetan Plateau, where in situ data remain sparse. The lidar retrievals also allow us to identify clear signatures of downward ozone transport. Influenced by its unique high elevation topography and the Asian monsoon, the Tibetan Plateau is recognized as one of the most active regions for Stratosphere-Troposphere Exchange (STE) (Chen et al., 2006). Under these conditions, ozone transport from the stratosphere across the tropopause is likely to be an important source of background tropospheric ozone over Tibet, with potential implications for regional climate regulation and for the global coupling of chemistry and climate (Gettelman et al., 2002; Duan and Wu, 2005). In this section, we use ERA5 global reanalysis fields (0.25°×0.25° spatial resolution; 1 h temporal resolution) as supporting evidence to analyze the STE event captured by the ozone-profiling lidar and to provide an initial observational assessment of its evolution.

As shown in Fig. 4, the USTC Ozone Lidar captures pronounced temporal variations in the vertical distribution of ozone near the tropopause during the nights of 18 and 19 October 2017. To ensure an adequate signal-to-noise ratio, photon-counting returns were accumulated over 30 min and processed using a moving time-window method (Duchnowski and Wyszkowska, 2025), yielding a continuous ozone concentration distribution as a function of height and time. It is worth noting that ozone DIAL measurements are conducted only at night and are limited by weather conditions. As a result, the ozone results typically appear as banded segments separated by temporal gaps, shown in Fig. 4. Figure 4a shows the evolution observed during 16:00–21:00 UT on 18 October. An ozone-enhanced layer exhibits a pronounced downward displacement in the time–height cross-section, forming the tilted structure indicated by the blue arrow. Figure 4b shows the subsequent evolution during 15:00–21:00 UT on 19 October, when the ozone-enhanced structure exhibits an overall upward displacement, as indicated by the orange arrow. At this stage of the analysis, these features are interpreted descriptively as vertical displacements of ozone-rich structures near the tropopause; whether they represent irreversible cross-tropopause transport or primarily reflect tropopause displacement is examined using the dynamical and flux diagnostics presented below.

Within these broader variations, the lidar resolves several short-lived episodes of downward displacement of ozone-enhanced structures, marked T0–T6 in Fig. 4. Some episodes persist for 1–2 h, whereas others last for less than 1 h. Their short duration highlights the capability of the lidar to resolve rapid variability in ozone structure near the tropopause that may not be captured by satellite observations or lower-resolution meteorological datasets. Previous studies have given comparatively limited attention to such short-lived ozone structures near the tropopause. Importantly, these features should not necessarily be interpreted as individual cross-tropopause transport events, since rapid tropopause oscillations or undulations may also produce apparent vertical displacements of ozone-rich layers. Their repeated occurrence on sub-hourly to hourly timescales nevertheless provides a useful observational perspective on fine-scale ozone variability and tropopause dynamics. The relationship of these features to tropopause variability and cross-tropopause transport is examined further using the temperature, dynamical, and flux diagnostics presented below.

To corroborate the suspected STE-related ozone descent observed by the ozone lidar in autumn 2017, we further use ERA5 global reanalysis fields as an independent line of evidence. A prerequisite for diagnosing cross-tropopause exchange during STE events is a robust identification of the tropopause. For an initial ERA5-based screening of the event (interpolate the ERA5 data at uniform intervals of 0.5 hPa), the tropopause height is determined using the thermodynamic definition. Following the World Meteorological Organization (WMO) lapse-rate criterion proposed in 1986 (Wei, 1987), the thermodynamic tropopause is defined as the lowest level above the 500 hPa surface at which the atmospheric temperature lapse rate decreases to 2 °Ckm-1 or less, provided that the mean lapse rate within the 2 km layer above this level does not exceed 2 °Ckm-1. In this study, the thermodynamic tropopause was determined independently at each available ERA5 time step. Specifically, the temperature lapse rate was calculated from the corresponding ERA5 vertical temperature profile, and the WMO lapse-rate criterion was then applied to identify the tropopause. Thus, both the temperature lapse rate and the resulting tropopause height were allowed to vary with time. Analysis of ERA5 data during the periods covered by USTC Ozone Lidar observations reveals a pronounced cross-tropopause ozone descent event. We therefore focus on ERA5 data at the Yangbajing site for 16–19 October 2017. The continuous temporal coverage provided by ERA5 during this period allows the evolution of the event to be captured more completely, as illustrated in Fig. 5a. A strong STE-related ozone descent is evident during 18–19 October 2017, in close agreement with the USTC Ozone Lidar observations.

Both the USTC Ozone Lidar observations and ERA5 capture the onset and recovery phases of this ozone descent event. Differences between the lidar-retrieved ozone concentrations and those from ERA5 are further quantified. As shown in Fig. 5b, the ozone number density variations observed by the USTC Ozone Lidar during the STE event are compared with the corresponding ERA5 values. To ensure a consistent comparison, the analysis is performed at the tropopause level (16.6 km), and the 95 % confidence interval of the lidar retrievals is also shown. Notably, ERA5 provides hourly-resolution data, whereas the ozone lidar yields full-profile ozone measurements with 1 min update intervals. Figure 5b shows that the tropopause-level (16.6 km) ozone number densities measured by the USTC Ozone Lidar during 18–19 October 2017 agree closely with ERA5 in both absolute magnitude and temporal evolution. However, comparing Fig. 5a with Fig. 4b indicates that during the recovery phase on 19 October 2017, the ozone number densities retrieved by USTC Ozone Lidar below 15.5 km are systematically slightly lower than those in ERA5. Given the high SNR of the lidar returns during this period, the discrepancy is more likely to reflect a modest overestimation by ERA5 of the ozone number density associated with stratosphere-to-troposphere transport at this stage.

Mass and trace-constituent exchange between the stratosphere and troposphere is a central component of the coupled troposphere–stratosphere system. In 1987, Wei proposed a diagnostic framework for quantifying this exchange, in which all physical processes contributing to cross-tropopause transport can be expressed in an explicit mathematical form (Reiter, 1975; Wirth and Egger, 1999). The general form can be written as: 1F(f)=ρJηfdηdt-∂η∂tz0-U⋅∇z0ηηB, where F(f) is the air mass flux across the tropopause, ρ is the air density near the tropopause, η is a generalized vertical coordinate, Jη=∂z/∂η is the vertical coordinate-transformation factor between the generalized vertical coordinate η and geometric height z, z0 denotes the tropopause, dη/dt is the vertical velocity in η coordinates, U is the horizontal wind vector, and f is the constituent mixing ratio. In pressure coordinates, this expression becomes: 2F(ρ)=-g-1w-U⋅∇z0p-∂p∂tz0.

A key advantage of Wei's diagnostic formulation is that it allows the upward and downward cross-boundary fluxes to be computed separately and can be applied to any atmospheric constituent across an arbitrary reference surface. Given the large contrast in ozone mass mixing ratios between the stratosphere and troposphere, and the common assumption that an air mass becomes well mixed over a relatively short timescale after crossing the tropopause, quantitative assessments based on ozone profiles require the upward and downward fluxes to be treated separately. As noted above, we initially located the thermodynamic tropopause using changes in the vertical temperature lapse rate. In some situations, however, particularly near jet streams, this tropopause can be poorly defined. Because the following analysis examines lidar observations in detail and because the Yangbajing site lies in the low-to-mid latitudes, where large-scale dynamics and strong disturbances can substantially perturb the thermal structure, defining the tropopause solely from lapse-rate criteria may introduce appreciable uncertainty.

Holton noted that the tropopause is better regarded as a quasi-material surface (Holton et al., 1995). This perspective is well captured by the dynamical tropopause defined in terms of potential vorticity (PV), and cross-tropopause transport studies therefore commonly adopt a PV-based tropopause, typically using thresholds in the range of 1.5–4 PVU (where PVU denotes the potential vorticity unit). Moreover, Yang et al. showed that the choice of PV threshold can substantially affect estimates of cross-tropopause exchange, and advocated selecting an optimal threshold based on its correlation with observations or diagnostic quantities (Yang and Lü, 2003). It should be noted that the PV fields used here are derived from ERA5 and therefore cannot be expected to reproduce the fine-scale structures resolved by the lidar on a point-by-point or event-by-event basis. In this study, the ERA5-derived PV is used primarily to characterize the regional dynamical tropopause and the larger-scale dynamical environment associated with the STE event, rather than to reproduce the individual short-lived ozone features observed by the lidar. The two datasets are therefore regarded as complementary: the lidar provides high-spatiotemporal-resolution information on local ozone variability, whereas ERA5 provides the dynamically consistent three-dimensional meteorological fields required for diagnosing the tropopause structure and cross-tropopause transport.

Following the approach of Yang et al., we conduct a sensitivity analysis over a domain centered on the Yangbajing ozone lidar site (29° N, 99° E), spanning 25–32° N and 94–104° E, to assess how the choice of PV threshold influences the diagnosed cross-tropopause flux. As shown in Fig. 6, we quantify the changes in cross-tropopause transport on 18–19 October 2017 associated with different PV thresholds within this region. The results indicate that the estimated cross-tropopause flux is highly sensitive to the choice of threshold. For PV thresholds below 3 PVU, both upward and downward transports are relatively large, exhibiting more tropospheric characteristics; for thresholds above 3 PVU, both transports are markedly smaller, consistent with more stratospheric behavior. With increasing PV threshold, the total exchange in both troposphere-to-stratosphere transport (TST) and stratosphere-to-troposphere transport (STT) decreases. In addition, the magnitudes and temporal evolution of TST and STT suggest that, before and after the onset and recovery of the descent event on 18–19 October 2017, the total cross-tropopause transport over the study region is broadly consistent with mass conservation.

Furthermore, as shown in Fig. 7a and b, to ensure spatial continuity and robustness of the tropopause definition, we construct two PVU cross-sections from ERA5 in the vicinity of the lidar site: a zonal section along 90–110° E and a meridional section along 24–34° N. These sections show that, in some areas, tropopauses defined using lower PVU thresholds become discontinuous in pressure coordinates. By contrast, the continuity and stability improve markedly for thresholds of 2 PVU and above. Combining the sensitivity results with the cross-section structures, we consider 3 PVU (1PVU=10-6m2s-1kg-1) to be a reasonable dynamical tropopause threshold for this event (highlighted with an orange line in Fig. 7a and b). It should be noted that the 3 PVU surface is not assumed to represent an exact tropopause height with a prescribed absolute accuracy. Rather, uncertainty associated with the dynamical tropopause definition is evaluated through the PV-threshold sensitivity analysis described above. The 3 PVU threshold is therefore adopted as a physically reasonable regional-scale dynamical boundary for this event based on both the sensitivity of the diagnosed transport and the spatial continuity of the corresponding PV surface.

Figure 7c and d shows the time evolution of the domain mean transport components over 18–19 October 2017, diagnosed from ERA5. The wind fields used in this analysis are obtained from the ERA5 reanalysis, which assimilates a wide range of atmospheric observations and provides observation-constrained, dynamically consistent three-dimensional meteorological fields (Hersbach et al., 2020). Here, the ERA5 wind fields are used to characterize the regional-scale circulation and to provide the dynamical quantities required for the cross-tropopause flux diagnosis, rather than to reproduce the individual short-lived structures resolved by the lidar. The vertical wind contribution remains relatively steady over the two days. The horizontal wind contribution is of comparable magnitude but opposite sign, so that the net flux is controlled to a large extent by variations in the tropopause. It should be stressed that tropopause undulations mainly modify the geometric boundary used in the flux diagnosis and do not, by themselves, generate strong and persistent local transport.

Given the observed tropopause variations, a natural question is whether they were driven by temperature fluctuations in the tropopause region, and in particular whether such temperature changes corresponded to the short-lived, rapid ozone descent events shown in Fig. 4. As discussed above, ERA5 is used here to characterize the larger-scale dynamical evolution of the event, but its hourly output cannot fully resolve the short-lived variability observed by the lidar. Therefore, to examine whether the individual ozone-displacement episodes are accompanied by rapid temperature variations near the tropopause, we use the higher-temporal-resolution temperature profiles retrieved from the 387 nm N2 Raman lidar signal. As described in Sect. 2, the USTC Ozone Lidar also recorded the N2 vibrational Raman backscatter signals at 332 and 387 nm, corresponding to the 308 and 355 nm emission wavelengths, respectively, and these signals exhibit sufficient signal-to-noise ratio over the 14 to 30 km altitude range, as shown in Fig. 8a. Because 332 nm remains within the ozone absorption band, the 387 nm Raman signal was used to retrieve the temperature profiles. The temperature retrieval, shown in Fig. 8b, was initialized at 50 km using the temperature from the U.S. Standard Atmosphere 1976 as the upper-boundary reference (Standard, U.S., 1976), and the temperature profile was subsequently retrieved downward from this reference altitude. Comparison with ERA5 at the same times shows that the two temperature profiles are in close agreement. Figure 8c and d present the temperature variations retrieved from the 387 nm return over 14.5 to 17 km, where the temperature enhancements at the seven times T0 to T6 correspond closely to those identified in Fig. 4. These results indicate a strong association between the short-lived, rapid ozone descent events and temperature fluctuations in the tropopause region. Moreover, except for T2, all six of the other descent episodes are associated with temperature gradients exceeding 8 Kkm-1, while T2 still exhibits a pronounced gradient of 4.2 Kkm-1. They also represent the first detection of these descent features by the USTC Ozone Lidar, together with the first identification of their relationship to tropopause-region temperature variability using lidar-derived temperature profiles.

Turning back to Fig. 7c and d, the temporal evolution of the domain-mean flux reveals pronounced stratosphere-to-troposphere transport from 08:00 to 15:00 UT on 18 October and from 21:00 UT on 18 October to 03:00 UT on 19 October. As the analysis here focuses on the descent event, STT is defined as positive flux. The STE signal begins to recover around 10:00 UT on 19 October, consistent with the features in Fig. 5a, indicating that the dominant impact of this event is confined to within roughly 24 h. We also note that, in a domain-mean sense, the onset of the descent signal on 18 October precedes the STE timing inferred at the Yangbajing site (29° N, 99° E) from both the lidar observations and the ERA5-based diagnosis. This apparent lead suggests that the event may have first intensified in a core region away from the site, and then expanded to influence the vicinity of the observation station. To identify the main source region of the event and investigate its dynamical drivers, we further examine the spatial distribution of the cross-tropopause mass flux.

Using Wei's diagnostic framework together with ERA5 meteorological fields, we quantify the cross-tropopause mass exchange flux during 13:00–18:00 UT on 18 October. The flux is decomposed into contributions associated with tropopause motion, horizontal advection, and vertical motion, and we map the 6 h mean of the instantaneous mass flux over this period (Fig. 9). The red hexagon marks the location of the Yangbajing ozone lidar site. As shown in Fig. 9, the horizontal transport term exhibits a spatial distribution highly consistent with that of the total flux, with a comparable magnitude. This suggests that the spatial structure of the flux during this STE event was controlled mainly by horizontal transport. However, the flux estimates shown earlier in Fig. 7a and b indicate that the actual downward mass transport was induced by the tropopause motion term. Moreover, Fig. 9c further indicates that downward transport associated with the vertical transport term was also present at the Yangbajing lidar site. This behavior is consistent with the dynamical picture of tropopause folding (Sprenger et al., 2003; Škerlak et al., 2015). The regional PV structure indicates a tropopause-folding system over approximately 25–28° N, 95–99° E, characterized by a pronounced deformation and downward intrusion of stratospheric high-PV air toward the troposphere. The corresponding structure can also be identified in the PV cross-section near 26° N in Fig. 7a, and its spatial extent closely overlaps the region of enhanced cross-tropopause mass flux shown in Fig. 9. We note that the shorter-timescale oscillations of the tropopause observed during the evolution of the event do not contradict this regional folding interpretation. Tropopause folds developing in strongly sheared jet–front environments can be accompanied by inertia-gravity waves, wave breaking, and related instabilities, which further modulate and locally deform the tropopause and can contribute to irreversible mixing and STE (Cho et al., 1999; Bertin et al., 2001; Stohl et al., 2003). We therefore interpret the observed evolution as a regional tropopause-folding system accompanied and modulated by shorter-timescale wave-related tropopause variability.

As shown in Fig. 10, we use ERA5 meteorological fields together with Wei's diagnostic to derive the spatiotemporal evolution of cross-tropopause mass flux over the study region during 18–19 October 2017. Eulerian diagnostics formulated in pressure coordinates can be affected by numerical cancellation among individually large terms, potentially introducing substantial uncertainty into estimates of cross-tropopause fluxes (Wirth, 1995; Wirth and Egger, 1999). Lagrangian approaches, such as those developed and applied in previous studies (Wernli and Bourqui, 2002; Škerlak et al., 2014), provide an alternative that is generally less susceptible to this numerical-cancellation problem, particularly for global or large-domain analyses. However, because the present study focuses on the regional evolution of a specific STE event rather than on global-scale transport, we retained Wei's diagnostic framework but reformulated it in potential-vorticity (PV) coordinates. The corresponding results are presented in Fig. 10. Although this coordinate transformation does not eliminate all sources of uncertainty, it alleviates the numerical sensitivity and reduces part of the noise associated with cancellation among large terms. Each panel presents the 6 h mean of the instantaneous mass flux for the corresponding interval. From Fig. 10a–e, the tropopause-folding structure over 25–28° N, 95–99° E evolves markedly with time, and the associated mass flux field exhibits pronounced temporal non-stationarity and spatial heterogeneity. This sector lies along the southeastern flank of the Tibetan Plateau, within the transition zone between the entrance and exit regions of the upper-tropospheric jet, where strong shear and dynamical disturbances favor tropopause deformation, folding, and associated wave activity. Notably, the clearly identifiable folding band transitions from an approximately quasi-parallel, banded morphology in Fig. 10a and b to a meridionally offset, interleaving structure in Fig. 10g and h, suggesting a shift in the folding process from a stage modulated by near-linear wave perturbations towards a more strongly nonlinear breaking regime, possibly associated with inertial gravity waves induced by tropopause folding and their subsequent breaking. Given the fixed station location, the lidar measurements do not continuously sample the core of the tropopause fold, but more often reflect the downstream evolution and lateral modulation of the folding system (Shangguan et al., 2019; Zhang et al., 2024).

We offer the following conceptual interpretation of the dynamics. During the phase in Fig. 10a and b, PV anomalies near the tropopause are characterized by quasi-parallel PV contours, a relatively weak meridional gradient, and limited vertical extension. This structure suggests that folding is primarily modulated by disturbances, such as planetary waves and inertial gravity waves, superimposed on the shear environment of the upper-tropospheric jet (McIntyre and Palmer, 1983). In this stage, the lidar site (29° N, 99° E) lies on the northern flank of the folding band, within a wave-dominated sector that mainly reflects perturbations associated with tropopause undulations and PV field oscillations, while the net cross-tropopause transport remains limited. In Fig. 10c–f, as instability strengthens in the jet-exit region, the PV gradient across the folding band sharpens markedly. An ascending branch develops on the southern side of the fold, whereas a quasi-stationary descending branch forms on the northern side near the lidar site. The site then becomes intermittently aligned with the downstream transport pathway of the PV anomaly, indicating a transition from a fold evident in morphology to one that enables effective transport (Sprenger et al., 2003). In Fig. 10g and h, the fold exhibits clear PV tongue breakup and meridional displacement. The sawtooth-like offsets likely reflect wave breaking or nonlinear PV mixing (McIntyre and Palmer, 1983, 1984). At this stage, ascent-descent dipoles occur more frequently on smaller spatial scales, and the vertical circulation becomes shorter-lived and more localized. The previously coherent descending conduit weakens and becomes increasingly intermittent (Stohl et al., 2003). In the vicinity of the site, this fragmented circulation and enhanced mixing reduce sustained stratosphere-to-troposphere transport (STT), manifesting as a weakened or discontinuous downward flux. In addition, the ascending branch structure to the south of the site can further interrupt the local descent pathway.

Taken together, these results indicate that the short-lived, rapid ozone descent observed during this STE event was associated not only with temperature variations, but also that the onset and recovery of the event as a whole were modulated by tropopause folding and gravity waves over the region 25–28° N, 95–99° E.

Building on the mass-flux diagnostics, we further estimate the cross-tropopause ozone flux by incorporating ozone profiles from the USTC Ozone Lidar. Specifically, we use the lidar-retrieved vertical profiles of ozone number density to characterize ozone abundance in the troposphere and stratosphere and, in combination with the diagnosed cross-tropopause mass flux, to quantify ozone exchange across a layer extending several kilometers above and below the tropopause.

To ensure the robustness of the calculation, we assume that the ozone number densities measured at the lidar site (29° N, 99° E) are representative of the mean ozone level within 28.75–29.25° N and 98.75–99.25° E. Consistently, the mass flux is diagnosed from ERA5 gridded fields over the same spatial domain. Following Cui et al. (2004), the lidar-derived, vertically varying ozone mixing-ratio profiles are integrated over the 3 km layers immediately below and above the tropopause to obtain representative layer-mean mixing ratios. Consequently, the measured vertical ozone gradients within these layers are retained in the calculation rather than being replaced by a vertically uniform ozone concentration. 3Fq=Mup∫ptropopause-3ptropopauseq(p)dp∫ptropopause-3ptropopausedp+Mdown∫ptropopauseptropopause+3q(p)dp∫ptropopauseptropopause+3dp. In the equation, q(p) represents the uniform mixing ratio of ozone at pressure p, Mup denotes the upward flux of substances through the tropopause, and Mdown denotes the downward flux of substances. Positive and negative values indicate locations above and below the tropopause, respectively. Thus, the observed vertical variation of ozone within the 3 km layers above and below the tropopause is retained in the integration rather than assuming a vertically uniform ozone abundance. The two integral terms represent the pressure-weighted layer-mean ozone mixing ratios on the tropospheric and stratospheric sides of the tropopause, respectively.

The instantaneous cross-tropopause ozone flux diagnosed at the Yangbajing lidar site for 18–19 October 2017 is shown in Fig. 11. A clear STE-related ozone descent signal is evident from 13:00 UT on 18 October to 12:00 UT on 19 October, manifested as a positive ozone flux from the stratosphere to the troposphere. In contrast, Aura satellite sampling near the site is temporally sparse. MLS typically overpasses the vicinity of Yangbajing only twice per day, whereas OMI provides at most one pass per day, which largely explains the relatively low temporal resolution of the satellite products. For 18–19 October, MLS provides three nearby footprints at (29.66° N, 98.27° E), (25.22° N, 105.20° E), and (29.66° N, 95.52° E) (red hexagons in Fig. 11). The first and third footprints are closest to the Yangbajing site, and the ozone fluxes inferred from MLS there agree well with the ERA5 estimates. By contrast, the second footprint is displaced substantially in longitude and lies outside the Tibetan Plateau, and its inferred ozone flux deviates more strongly from both the ERA5 and lidar results. Overall, the ozone flux derived from the USTC Ozone Lidar measured profiles is broadly consistent with the ERA5-driven ozone flux. More specifically, on 18 October, the lidar observations mainly capture the intensification phase and yield slightly larger diagnosed fluxes than ERA5, whereas on 19 October, during the recovery phase, the two estimates closely agree.

In this study, we present an initial detection and analysis of a stratosphere–troposphere exchange (STE) event captured by ozone lidar observations. The ozone flux associated with stratosphere-to-troposphere transport derived from the USTC Ozone Lidar exhibits an instantaneous peak of approximately 3–4×10-10 kgs-1m-2, slightly exceeding the corresponding estimate derived from ERA5 for the same period. Despite this difference, the overall agreement between the lidar-derived and ERA5-derived ozone fluxes supports the robustness of the flux-calculation framework. Although ozone lidar provides ozone concentration measurements with high spatial and temporal resolution, the lack of time-continuous in situ wind observations necessitates the use of ERA5 wind fields at 1 h resolution in the flux diagnosis. Despite this limitation, our analysis demonstrates that ozone-profiling lidar can not only resolve the fine vertical structure of ozone but also, when combined with wind information, enable quantitative observation and flux estimation of STE-related downward ozone transport, thereby broadening the potential of lidar for studies of coupled atmospheric composition and dynamics.