Using ozone measurements from two sounding sites and two high-altitude stations in Central Europe, we show evidence for a dominant influence of changes in lowermost stratospheric ozone on the variability and overall upward trend of background ozone in the lower troposphere (3000–3500 m asl) during the 1992–2004 period. Numerical simulations with a stratospheric chemistry transport model suggest that changes in lower stratospheric ozone were driven by dynamics rather than by changes in stratospheric chlorine loading. In addition, Lagrangian model simulations indicate that changes in downward transport of ozone from the stratosphere into the troposphere were dominated by changes in lowermost stratospheric ozone concentrations rather than by variations of cross-tropopause air mass transport. This suggests that the positive ozone trends and concentration anomalies in the lower free troposphere over Europe during the 1990s were likely to a large extent due to enhanced stratospheric ozone contributions, particularly in winter–spring.
 Emission controls of ozone precursors throughout most Europe have resulted in a reduction of peak surface ozone levels during the 1990s over parts of Northwestern and Central Europe [Brönnimann et al., 2002; Derwent et al., 2003]. However, there is also evidence that background ozone – ozone that is only marginally influenced by European emissions – has continued to increase in the lower troposphere over Western and Central Europe during the 1990s [Brönnimann et al., 2002; Simmonds et al., 2004]. Several hypotheses have been put forward in order to explain these changes. Although NOx emissions have substantially increased in East Asia during the last years [Richter et al., 2005, and references therein], model studies indicate that changes in Asian emissions cannot explain the significant upward trends observed in tropospheric background ozone over Europe [Li et al., 2002; Derwent et al., 2004]. Other analyses suggest that the stronger than usual westerlies during a positive phase of the North Atlantic Oscillation (NAO) might lead to increased ozone in Western Europe as a consequence of the transport of pollution plumes from North America [Creilson et al., 2003]. In addition, measurements at Mace Head, Ireland, reveal that the rate of accumulation of ozone is similar to that of biomass burning related gases during years with intense global fires [Simmonds et al., 2005]. However, no general upward trend is found either in the NAO index http://www.cru.uea.ac.uk/cru/climon/data/nao/) or in carbon monoxide concentrations over the extratropical Northern Hemisphere during the 1990s [Wotawa et al., 2001].
 It is long known that transport from the stratosphere is a source of tropospheric ozone [Danielsen, 1968] but only recently it has been demonstrated from observations that there is a coupling between stratospheric and upper tropospheric ozone levels on multi-annual timescales [Tarasick et al., 2005; Thouret et al., 2006]. The possibility that the observed interannual variability in lower tropospheric ozone is influenced by changes in stratospheric ozone has been discussed previously [e.g., Lelieveld and Dentener, 2000; Fusco and Logan, 2003]. However, there is no consensus on the fraction of stratospheric ozone that contributes to the observed ozone levels in the troposphere. Different methods analyzing the downward cross-tropopause flux of ozone converge to estimate a global net flux in the range 400 – 600 Tg/year [e.g., Olsen et al., 2001; Hsu et al., 2005, and references therein]. However, global chemistry transport models (CTMs) have too coarse horizontal resolution to resolve small scale processes that contribute to stratosphere-troposphere exchange (STE). As an example, two surveys of global tropospheric CTM studies indicate a net influx of stratospheric ozone into the troposphere ranging from 400 to 1400 Tg/yr [Intergovernmental Panel on Climate Change, 2001] and from 150 to 900 Tg/yr [Stevenson et al., 2006]. In addition, chemical lifetimes and transport times from the upper troposphere to the lower troposphere might vary strongly among models, influencing the simulated contribution of stratospheric ozone to the lower troposphere.
 The aim of this study is to examine the impact of northern mid-latitude lowermost stratospheric ozone changes on background ozone in the proximity of the lower free troposphere over Western-Central Europe. For this purpose, surface ozone measurements at two elevated sites and ozonesonde measurements in the lowermost stratosphere, in addition to results from a stratospheric chemical transport model (CTM) and a Lagrangian model, are analyzed here.
2. Data and Methods
2.1. Ozone Measurements
 Time series of surface ozone measurements at two European high-altitude sites (Jungfraujoch, 3580 m asl, Switzerland; Zugspitze, 2962 m asl, Germany) as well as ozone measurements in the lowermost stratosphere (150 hPa, approx. 13.6 km altitude) at two sounding sites (Payerne, Switzerland; Hohenpeissenberg, Germany) are used in this paper to study the influence of stratospheric ozone changes on lower free tropospheric ozone during the period 1992–2004. Both datasets provide exceptionally good temporal coverage and are highly representative for northern mid-latitudes over Europe (see auxiliary material for details on the quality and selection of the measurements).
2.2. SLIMCAT Model
 To further investigate lower stratospheric ozone changes over the North Atlantic and Western-Central Europe, the 3D SLIMCAT stratospheric CTM is used with 17 isentropic levels in the vertical, from 340 K to 2400 K (approximately 11 to 55 km altitude) and 5.6° × 5.6° horizontal resolution. The model is forced by 6-hourly temperature and horizontal wind fields from the ERA-40 analyses [Uppala et al., 2005] until 2001 and by ECMWF operational analyses from 2002 to 2004. Vertical transport is calculated using a radiative transfer model and does not depend on ERA-40 vertical winds. The chemistry scheme used consists of a linear parameterization of gas-phase ozone photochemistry [Cariolle and Déqué, 1986] that is used with constant chlorine levels appropriate to the 1980s. The chemistry parameterization has no interannual variation so that any calculated interannual variability, or trend, must arise from changes in the forcing meteorology. For a more detailed description of the model setup see section 2 of Hadjinicolaou et al. .
2.3. Lagrangian Climatology of Deep STT
 To investigate the impact of changes in deep stratosphere-to-troposphere transport (STT) on lower free tropospheric ozone, the Lagrangian technique – introduced by Wernli and Bourqui  – was applied to global ERA-40 data for the 1990–2001 period. 1° × 1° gridded three-dimensional wind and temperature fields were used, with full vertical resolution and a temporal resolution of 6 hours. Backward and forward trajectories were calculated once per day with the LAGRANTO model starting on a regular grid spaced 80 km in the horizontal and 30 hPa in the vertical direction. The daily number of 8 × 105 trajectories for the Northern Hemisphere extratropics provides a robust statistical basis for the diagnosis of deep STT. Three criteria were applied to select deep STT events in this analysis: (1) crossing of the dynamical tropopause (as defined by the lower of the 2 PVU isosurface and the 380 K isentrope) from the stratosphere to the troposphere, (2) 96 hours residence time both in the stratosphere prior to and in the troposphere after crossing the tropopause, and (3) destinations at levels below 700 hPa (i.e. p > 700 hPa, approx. z < 3000 m) over the North Atlantic and Western-Central Europe (60°W to 20°E, 30°N to 70°N). The mass flux of these deep STT events was then integrated on a monthly basis. (See additional information in the auxiliary material.)
3.1. Impact of Stratospheric Ozone Changes on Lower Free Tropospheric Ozone
Figure 1 represents the 12-month running means of the normalized monthly anomalies (with respect to the 1992–2004 climatological monthly means) of both surface ozone at Jungfraujoch/Zugspitze (blue) and lowermost stratospheric ozone (150 hPa) averaged for Payerne and Hohenpeissenberg (red) (see original time series of the monthly anomalies in the auxiliary material). These datasets follow a strikingly similar evolution, with the exception of the year 2003 in which surface ozone concentrations were very high as a consequence of the anomalously dry and warm summer in Central Europe [e.g., Ordóñez et al., 2005]. The Pearson correlation coefficient between the 12-month running means is 0.77 (0.76) when Jungfraujoch (Zugspitze) is considered (N ≈ 144). Note the different scales used in Figure 1 because of the larger temporal variability in stratospheric ozone, which clearly argues for an influence of the stratosphere on the troposphere and not for the reverse (see further discussion on the propagation of stratospheric ozone changes down into the troposphere, including comparison between surface ozone and ozonesonde data at different pressure levels, in the auxiliary material). Both lowermost stratospheric ozone and lower free tropospheric ozone have increased over the period of analysis. Trends presented in this paper are expressed as percentage per decade, relative to the mean ozone mixing ratio of the period. The estimated trends (and 95% confidence intervals) of the yearly averaged ozone mixing ratios are 16.41 (±11.76) %/decade for 150 hPa, 10.26 (±4.80) %/decade for Jungfraujoch and 4.65 (±4.75) %/decade for Zugspitze during the 1992–2004 period.
 At mid-latitudes, ozone concentrations at 150 hPa peak in late winter/early spring (maximum of ∼600 ppb in April compared to ∼250 ppb in October at both sites). A similar late winter/early spring peak occurs in STT air mass fluxes [Appenzeller et al., 1996] and therefore in STT ozone fluxes [Lelieveld and Dentener, 2000]. In order to investigate the influence of stratospheric ozone on the lower free troposphere as a function of season, the correlations and slopes of the linear regression of the seasonal anomalies of near-surface ozone versus those of ozone at 150 hPa were calculated (Table 1). Compliant with a significant stratospheric influence on tropospheric ozone levels, the correlations are higher in winter–spring than in summer–autumn. In addition, the slopes are only significant in winter–spring and generally higher than in summer–autumn. The calculated correlations and slopes suggest an important stratospheric contribution to tropospheric ozone for the altitudes of Zugspitze and Jungfraujoch except in summer, when the influence of photochemical ozone production and transport of air masses from the polluted boundary layer is strongest for these sites.
Table 1. Pearson Correlation Coefficients and Slopes of the Linear Regressions of the Normalized Seasonal Anomalies of Surface Ozone at Jungfraujoch/Zugspitze Vs. Those of Lowermost Stratospheric Ozone Averaged for Payerne and Hohenpeissenberga
Lowermost stratospheric ozone is measured at 150 hPa. N = 12 in winter, N = 13 in the other seasons. Slopes are given together with their 95% confidence intervals. The seasons are defined as follows: spring, MAM; summer, JJA; autumn, SON; and winter, DJF. Winter 1993 corresponds to Dec 1992–Feb 1993.
0.25 ± 0.12
0.06 ± 0.31
0.21 ± 0.26
0.21 ± 0.18
0.15 ± 0.10
−0.04 ± 0.28
0.18 ± 0.25
0.20 ± 0.12
3.2. Modeled Changes in Lowermost Stratospheric Ozone
 The 12-month running means of the normalized monthly anomalies of the SLIMCAT output for ozone on the 380 K isentrope (∼150 hPa) averaged over an area covering the North Atlantic and Western-Central Europe are also shown in Figure 1. Both the overall upward trend and the year-to-year variability of lowermost stratospheric ozone are similar in model output (green dotted) and measurements (red), but with lower values from the model in 1998 and 1999. The smoothed variability in modeled ozone and the somewhat different timing in the peaks result as a consequence of averaging model output over a large area. The correlation coefficient between the 12-month running means of the corresponding anomalies is 0.68 and increases up to 0.76 if SLIMCAT output for a single grid point over Central Europe is considered (N = 144). The trend of the annual means of modeled ozone is 18.06 (±9.11) %/decade, very similar to that of measured ozone at 150 hPa. Since this version of the model has a simplified chemistry, the calculated interannual variability and overall upward trend are mainly dynamically driven. Other analyses [Tarasick et al., 2005; Yang et al., 2006] also suggest that the rebound in lower stratospheric ozone at northern mid-latitudes during the 1990s might be dominantly a result of changes in the atmospheric circulation, rather than a recovery of the ozone layer from halocarbon-induced depletion.
3.3. Impact of Changes in Deep STT on Lower Free Tropospheric Ozone
 The time series of normalized anomalies of the monthly integrated deep STT mass flux with destinations in an area covering the North Atlantic and Western-Central Europe is shown in Figure 2. The integrated deep STT mass fluxes have been calculated by adding up all the monthly mass fluxes fulfilling the three criteria of Section 2.3 (see time series in auxiliary material). It should be noted that an important fraction of the air masses contributing to the mass fluxes in Figure 2 penetrate the lower free troposphere far from the location of Jungfraujoch (7.98°E, 46.55°N) and Zugspitze (10.98°E, 47.42°N) because STT events are often associated with long-range horizontal transport. As a consequence, these air masses can influence ozone levels in the lower free troposphere over Western-Central Europe [e.g., Sprenger and Wernli, 2003], particularly in winter and early spring because of the coupling between elevated concentrations of lowermost stratospheric ozone, high activity of STT and long lifetime of tropospheric ozone during that time of the year.
 The flux of air from the stratosphere into the lower free troposphere has not significantly increased and has remained more stable than the lowermost stratospheric ozone levels during the 1990s. Lower year-to-year variability as well as a smaller magnitude of the normalized anomalies is observed in the case of deep STT mass fluxes. Given this moderate variability and the considerable contribution of stratospheric ozone to lower free tropospheric ozone levels, the high correlation between ozone anomalies at elevated surface sites and in the lower stratosphere can be well understood.
 This study demonstrates that not only the effect of hemispheric ozone precursor emissions but also stratospheric ozone changes have to be represented accurately in models in order to describe the evolution of the hemispheric background ozone. We show evidence for an influence of changes in lowermost stratospheric ozone concentrations on the variability of ozone levels in the lower free troposphere over Europe during the 1990s, particularly in winter–spring and to a lower extent in autumn. No significant relationship has been found for summer, when ozone at Jungfraujoch and Zugspitze is most strongly influenced by the polluted planetary boundary layer. The observed coupling, in combination with the moderate variability in STT mass fluxes at Northern mid-latitudes, indicates that the positive trends in background ozone over Europe during the 1990s were likely to a large extent due to the contribution of enhanced lowermost stratospheric ozone levels. This increase in background ozone led to a partial compensation of the decrease in summer ozone production resulting from the precursor emission reductions in Central Europe [Ordóñez et al., 2005; Jonson et al., 2006].
 Ozone concentrations in the lowermost stratosphere are highly variable and present levels appear to be relatively high due to dynamical effects [Brunner et al., 2006; Yang et al., 2006]. Nevertheless, the recovery of the ozone shield due to the phase-out of human-made ozone depleting substances is likely to contribute to an increase in lower stratospheric ozone in the coming decades. In addition, climate change might enhance the import of ozone from the stratosphere into the troposphere [Gauss et al., 2006; Stevenson et al., 2006]. As a consequence, a further increase in background ozone can be expected with potential implications not only for air quality but also for global warming.
 Ozonesonde data at Hohenpeissenberg and Payerne were provided by the German National Meteorological Service (DWD/MOHp) and MeteoSwiss, respectively, through the World Ozone and Ultraviolet Radiation Data Centre (WOUDC). Jungfraujoch measurements are conducted within the Swiss Air Quality Monitoring Network (NABEL), operated by EMPA and financed by the Swiss Federal Office for the Environment. We thank H. E. Scheel, Forschungszentrum Karlsruhe, for providing Zugspitze data. ERA-40 data for the Lagrangian calculations were made available by MeteoSwiss and ECMWF, and the calculations have been performed on the CSCS in Manno (Switzerland). We acknowledge financial support by ACCENT and SCOUT-O3.