We use a chemical transport model (GEOS-CHEM) to quantify the contribution of long-range transported pollution to the European ozone (O3) budget for the year 1997. The model reproduces the main features observed over Europe for O3, carbon monoxide and nitrogen dioxides, as well as two events of enhanced O3 of North American origin over the eastern North Atlantic and over Europe. North American O3 fluxes into Europe experience a maximum in spring and summer, reflecting the seasonal variation in photochemical activity and in export pathways. In summer, North American O3 enters Europe at higher altitudes and lower latitudes because of deep convection, and because the flow over the North Atlantic is mostly zonal in that season. The low-level inflow is only important in spring, when loss rates in the boundary layer over the North Atlantic are weaker. Asian O3 arrives mainly via the westerlies, and usually at higher altitudes than North American O3 because of stronger deep convection over Asia. In addition, Asian O3 fluxes are at a maximum in summer during the monsoon period because of enhanced convection over Asia, increased nitrogen oxides sources from lightning and direct transport towards Europe via the monsoon easterlies. Over Europe, total background accounts for 30 ppbv at the surface. North American and Asian O3 contribute substantially to the annual O3 budget over Europe, accounting for 10.9% and 7.7%, respectively, while the European contribution only accounts for 9.4%. We find that in summer, at the surface, O3 decreases over Europe from 1980 to 1997, reflecting the reduction of European O3 precursor emissions. In the free troposphere, this decrease is compensated by the increase in O3 due to increasing Asian emissions. This may explain the lack of trends observed over most of the European region, especially at mountain sites.
 Tropospheric ozone (O3) results from a complex interaction between transport from the stratosphere [e.g., Junge, 1962; Danielsen, 1968] and in situ photochemical production taking place through the oxidation of volatile organic carbons (VOCs) and carbon monoxide (CO) in the presence of nitrogen oxides (NOx = NO + NO2) [e.g., Levy, 1971; Chameides and Walker, 1973; Crutzen, 1974]. The lifetime of O3 in the free troposphere ranges from a few days to several months [e.g., Liu et al., 1987], which allows its transport over distances of intercontinental and hemispheric scales. Transport of O3 and related species may thus have impacts on O3 concentrations found downwind of industrialised regions of the Northern Hemisphere, where most of the O3 precursors (VOCs and NOx) are emitted [Jacob et al., 1999; Berntsen et al., 1999; Jonson et al., 2001; Wild and Akimoto, 2001; Stohl et al., 2002; Wild et al., 2004]. Li et al.  reported for example that anthropogenic emissions from North America led to an additional 20% of violations of the European Council O3 standard in the summer 1997 over Europe. There is a crucial need to quantify the relative contributions of the O3 produced by the precursors emitted within a region and that of the O3 transported from regions upwind (the so-called background O3) to provide better information to policy makers [Holloway et al., 2003]. In this paper we examine how tropospheric composition and O3 distributions, in particular over Europe, are affected by long-range transport of pollution.
 National reports indicate that European emissions of O3 precursors have decreased from 1990 to 2000 due to control strategies in most of the western and northern European countries. In eastern Europe the decrease is due to modernization the shut down of air-polluting industrial branches [Vestreng and Klein, 2002]. In the mean time, O3 precursor emissions have more than doubled since the 1990s in Asia because of rapid industrialization [Van Aardenne et al., 1999], while North American emissions have remained fairly constant or have slightly decreased [EPA, 1997, 2003]. The decrease in European emissions has likely induced a decrease in O3 peaks in summer [Derwent et al., 2003; Brönnimann et al., 2002] and an increase in mean O3 in winter over Europe due to less O3 titration by NOx [Lindskog et al., 2001]. For example, Derwent et al.  showed a large decline in the annual maximum eight-hour mean O3 concentrations and in O3 peaks observed between 1990 and 2000 over the United Kingdom. A decrease in O3 peaks was also found at several stations in Switzerland from 1991 to 1999 [Brönnimann et al., 2002].
 However, recent studies indicate either trends in monthly mean O3 close to zero or slightly increasing over the last decade for a number of sites over Europe. Kuebler et al.  reported, for example, trends close to zero for the 90th percentile summer O3 concentrations from 1988 to 1998 at some urban and rural sites in Switzerland, as well as at the mountain site of the Jungfraujoch (Switzerland, 3580 m above sea level) despite a negative trend in NOx, VOCs, and CO concentrations. Brönnimann et al. [2000, 2002] reported that the deseasonalised daily and monthly mean values over the periods 1992–1998 and 1991–1999 increase in a number of rural or urban stations in Switzerland. Logan et al.  found trends close to zero in the whole troposphere for two O3 soundings over Europe (Uccle and Hohenpeissenberg) for the period 1980–1996 and Pochanart et al.  reported no change in the average monthly mean O3 between 1996–1997 and 1989–1991 at Arosa (Switzerland, 1840 m above sea level). The small or close-to-zero trends in O3 observed at various sites in Europe, despite the reduction of local anthropogenic emissions, could indicate an increase in the background O3 over Europe [Collins et al., 2000; Brönnimann et al., 2002].
 In this paper, we use the three-dimensional (3-D) chemical transport model (CTM) GEOS-CHEM to investigate the processes that contribute to long-range transport of O3 into Europe. We focus our analysis on the contributions from North America (because of its direct influence on Europe [e.g., Wild and Akimoto, 2001; Stohl et al., 2002]), and from Asia, which is expected to contribute the most significantly to global atmospheric changes in the coming decades [IPCC, 2001]. We use the model to establish a European O3 budget and to examine its perturbations by long-range transport. We quantify how changes in O3 precursor emissions contribute to changes in background and total O3 over Europe. We focus on the period 1980–1997 over which Asian anthropogenic emissions significantly increased, and we examine to what extent the divergent trends observed over Europe for the last decades could reflect the superposition of various processes, including the decrease in local O3 precursor emissions and the increase in O3 transported from other continents.
2. Model Description and Simulations
2.1. GEOS-CHEM Model
 The GEOS-CHEM model (http://www-as.harvard.edu/chemistry/trop/geos/) [Bey et al., 2001a] is a global 3-D CTM driven by assimilated meteorological observations provided by the Goddard Earth Observing System (GEOS) of the NASA Global Modeling and Assimilation Office (GMAO). The assimilated meteorological fields include winds, surface pressure, temperature, water content, cloud information, convective mass flux, and other surface properties, with a 3- or 6-hour temporal resolution, depending on the variable. The fields are provided with a horizontal resolution of 2° of latitude by 2.5° of longitude, and 48 sigma levels (up to 0.01 hPa) for the GEOS-STRAT version (December 1995 to 1997). They are degraded in this study to a 4° of latitude by 5° of longitude and to 26 levels for computational expediency. Here we used the 5-02 version of the GEOS-CHEM model with some improvements as described below.
 The chemical mechanism is based upon that of Horowitz et al.  and includes 80 species and over 300 reactions with detailed photooxidation schemes for major anthropogenic hydrocarbons and isoprene. Heterogeneous reactions on aerosols (sulfate, black carbon, organic carbon, sea salt and dusts) are included, following recommendations from Jacob . The aerosol fields are provided by the Global Ozone Chemistry Aerosol Radiation and Transport (GOCART) model [Chin et al., 2002] and are coupled to the GEOS-CHEM model as described by Martin et al. . Photolysis frequencies in the troposphere are calculated with the Fast-J algorithm of Wild et al.  which accounts for aerosols and clouds. The dry deposition is computed using a resistance-in-series model [Wesely, 1989]. The wet deposition, applied to HNO3 and H2O2 follows Liu et al. . Transport of O3 from the stratosphere is simulated using the Synoz method (synthetic ozone) proposed by McLinden et al. , in which stratospheric O3 is represented as a passive tracer that is released uniformly in the tropical lower stratosphere at a rate constrained to match a prescribed global mean tropopause O3 flux. On average, we obtain a net stratosphere troposphere exchange of 450 Tg.yr−1. The spatial and temporal (year-to-year) variability of the stratospheric O3 column is accounting for by using the Total Ozone Mapping Spectrometer/Solar Backscatter Ultraviolet (TOMS/SBUV) merged total O3 data sets. The year-to-year varying methane (CH4) distributions are specified for each latitudinal band using the Climate Monitoring and Diagnostics Laboratory (CMDL) observations.
 Anthropogenic emission inventories in the GEOS-CHEM model are based on the NOx emissions from the Global Emission Inventory Activity (GEIA) [Benkovitz et al., 1996], the nonmethane hydrocarbon emission inventory from Piccot et al. , and CO emissions from Wang et al. [1998a]. We adjusted this base emission inventory of 1985 to specific years using scaling factors as described in Bey et al. [2001a]. In the present work, we used the year-specific European Monitoring and Evaluation Program (EMEP) database inventory for NOx, CO and VOC over Europe, which significantly improves our simulation of O3 precursors over Europe (not shown). Biomass burning emissions are from a climatological inventory described by Wang et al. [1998a] and its interannual variability is estimated using Along Track Scanning Radiometer (ATSR) fire-counts [Duncan et al., 2003].
 To investigate the ability of the model to reproduce observations of key atmospheric compounds over Europe, we first conducted a simulation using assimilated meteorology for 1997 (hereafter called the standard simulation). We then conducted a number of sensitivity simulations to distinguish between different components of total O3, including the hemispheric background and the anthropogenic contributions from Europe, Asia, North America and the rest of the world. In our study, the hemispheric background includes the stratospheric input and the contemporary (1997) O3 produced by precursors from natural sources (e.g., soils, lightning, biosphere), from biomass burning, and from CH4 oxidation. The hemispheric background is obtained from a simulation with all anthropogenic sources shut off, including NOx, CO, and NMHC emissions from fossil fuel and biofuel combustion as well as NOx from fertilizer and aircraft. The contributions of the main industrialised regions, i.e., North America (northward of 12°N, 132.5°W–62.5°W), Asia (northward of 12°S, 57.5°E–147.5°E), extended Europe (northward of 36°N, 12.5°W–57.5°E) and the rest of the world are calculated by subtracting the results of the sensitivity simulations (with anthropogenic emissions of a given region turned off) from that of the standard simulation. The total background for Europe is thus the sum of the hemispheric background plus the different anthropogenic contributions except Europe. Each sensitivity simulation is initialized by a one-year run. We also conducted sensitivity simulations with anthropogenic emissions scaled for the years 1980 (but with contemporary hemispheric background), to quantify changes in total O3 that could be attributed to changes in anthropogenic emissions.
 The contribution of stratospheric O3 is obtained by a tagged O3 simulation following the methods proposed in Wang et al. [1998b] and Fiore et al. . Production and loss frequencies are archived from the standard simulation and are further used to drive an off-line simulation in which total O3 is divided into individual tracers produced in different regions of the atmosphere, including the stratosphere.
3. Model Evaluation of Ozone and Related Species Over Europe
 The model has been previously the subject of global [Bey et al., 2001a] and more regionally focused evaluations, e.g., over the Pacific Rim [Liu et al., 2002], the United States [Fiore et al., 2002; Li et al., 2002], the North Atlantic Ocean [Li et al., 2002, 2004], and Europe [Duncan and Bey, 2004]. No obvious global bias was found for a number of species including O3, CO, and nitrogen dioxide (NO2), except that monthly mean O3 show a too low seasonal amplitude especially in the middle troposphere. In most of the cases, monthly mean CO were too low by 10 to 20 ppbv. Global NO2 distributions in the model were also found to be in good agreement with the NO2 columns retrieved from Global Ozone Monitoring Experiment (GOME) except that the model underestimates the observations in some regions of the world (South Africa, U.S. and industrial region of Europe) [Martin et al., 2002]. Li et al.  evaluated the model over the western North Atlantic using observations from the North Atlantic Regional Experiment (NARE) campaign, and found that observed CO were usually underestimated by 10–20 ppbv in the boundary layer, while O3 concentrations were well reproduced (within 5 ppbv at Sable Island). It was also reported that the nighttime depletion of surface O3 (due to deposition and chemical loss in shallow boundary layer) is poorly reproduced in the model [Fiore et al., 2002; Li et al., 2002]. In this paper, we focus the model evaluation over Europe using observations for CO, NO2 and O3. Tables 1–3 present a number of statistical quantities computed between the model results and atmospheric observations, which allow us to better quantify the evaluation. Correlation coefficient (r2) indicates the ability of the model to reproduce the seasonal cycle, while mean bias and absolute mean bias indicate whether the model tends to overestimate or underestimate the observed concentrations. We further evaluate the model in section 4.1 using observations available for two specific events of O3 transport over the North Atlantic ocean and Europe.
Table 1. Statistical Quantities for CO and NO2 Model Evaluationa
Statistical quantities between monthly mean NOAA/CMDL unfiltered CO data and monthly mean NO2 sampled at EMEP stations and simulated monthly mean CO and NO2 for 1997 at the surface.
Standard deviation, ppbv
Model Mean Bias, ppbv
Model Absolute Mean Bias, ppbv
Correlation Coefficient (r2)
Table 2. Statistical Quantities for O3 Model Evaluation at Three Different Levelsa
Statistical quantities between observed (ozonesonde data and MOZAIC data) and simulated monthly mean O3 concentrations for the seven stations in Europe which have observations for 1997. Model results and observations are used at 800, 500 and 300 hPa for ozonesonde and at 850, 550 and 350 hPa for MOZAIC.
Standard deviation, ppbv
Model Mean Bias, ppbv
Model Absolute Mean Bias
Correlation Coefficient (r2)
Table 3. Statistical Quantities for O3 Model Evaluation at the Surfacea
Monthly Afternoon Average O3
Daily Afternoon Average O3
Statistical quantities between EMEP data and model results for 1997 for monthly afternoon average O3 concentrations and for daily afternoon average O3 over June, July and August.
Standard deviation, ppbv
Model Mean Bias, ppbv
Model Absolute Mean Bias
Correlation Coefficient (r2)
3.1. Carbon Monoxide
Figure 1 compares monthly mean CO simulated by the model and observed from the cooperative National Oceanic and Atmospheric Administration (NOAA)/Climate Monitoring and Diagnostics Laboratory (CMDL) flask sampling program [Novelli et al., 1992; J. Logan, Harvard University, personal communication]. For the Negev Desert and Assekrem stations, only one observation is available per month. The model reproduces much of the observed seasonal variation at different sites (r2 = 0.63) but the model CO concentrations tend to be too low in winter and spring by about 20 ppbv and too high in summer and fall by about 20 ppbv, resulting in a global absolute mean bias of 15% (Table 1). The simulated CO at the Baltic sea station is too high by 200 ppbv in January in comparison with the observed value. Duncan and Bey  found however a much better agreement between the model and the observations at that site when the model is sampled in an adjacent box which appears to be more representative of the measurements taken at sea.
3.2. Nitrogen Dioxide
 Monthly mean NO2 simulated by the model and provided by EMEP are compared in Figure 2. At most of the sites, the model reproduces reasonably well the strong seasonal variation (r2 = 0.52) but the low summer concentrations are usually severely underestimated by the model, resulting in a high mean bias of 43% (Table 1). Evaluation of NO2 concentrations is made difficult by their great spatial and temporal variability. In addition, data quality issues should not be discarded, as NO2 measurements appear to be too high for the low concentrations [Barrett et al., 2000], especially for the German sites. This could partly explain the disagreements seen at the Brotjacklriegel and Schauinsland sites during the summer months.
 O3 distributions over Europe are evaluated using ozonesonde data [Logan et al., 1999], observations from the Measurement of Ozone by Airbus In-service Aircraft (MOZAIC) program [Marenco et al., 1998; Thouret et al., 1998] as well as surface measurements provided by EMEP. Figure 3 compares simulated and observed monthly mean O3 concentrations at 8 sites over Europe at three different altitudes. The model reproduces poorly O3 concentrations at the 300 hPa level (close to the tropopause) for the more northern stations (e.g., Ny Alesund and Sodankyla). The agreement between observed and model values improves in the upper troposphere for the sites further south, but the amplitude of the seasonal cycle is too weak. These discrepancies are likely due to a vertical resolution that is too coarse around the tropopause and a seasonal cycle that is too weak in the cross-tropopause flux in our model, as already noted by Bey et al. [2001a] and Fusco and Logan . The annual cross-tropopause flux (451 Tg O3.yr−1) is however in the range of other studies [IPCC, 2001]. Statistical quantities for the middle troposphere are relatively good (r2 = 0.59 and absolute mean bias less than 5 ppbv, Table 2) but the model shows however a lack of seasonal cycle, and in some cases, underestimates the summer concentrations by 5 to 10 ppbv. The model reproduces well the seasonal cycle in the lower troposphere, including the low values observed in the summer in the northern sites and the large summer increase (especially in August) at the mid-latitude sites (r2 = 0.49 and absolute mean bias less than 3.5 ppbv, Table 2).
 Simulated O3 is also evaluated at a mountain site (Jungfraujoch) with a higher temporal resolution. Figure 4 compares the hourly time series obtained at the Jungfraujoch from the National Air Pollution Monitoring Network (NABEL) with model results. The mean value is simulated correctly (49 ppbv for the model and 51 ppbv for the observations), but the maximum and minimum concentrations are usually not too well represented. This may be related to the too coarse resolution of the model (4° × 5°) which makes it difficult to capture the specific meteorology at a mountain site: the Jungfraujoch is often located within the boundary layer in summer, while model results are sampled in the free troposphere. That results in a relatively poor correlation of 0.34, and a mean absolute bias of 6.2 ppbv (12.7%).
 Simulated O3 concentrations are further evaluated at the surface using observations from EMEP (Figure 5). We used monthly afternoon averages for both model and observed data because the model does not resolve correctly the nighttime depletion as mentioned previously. The seasonal variations are well reproduced by the model at most of the sites (r2 = 0.84, Table 3). The model however overestimates the observed concentrations at some sites in summer (by up to 10 ppbv). This is seen particularly at some coastal sites (e.g., Yarner-Wood) where NOx dilution within the model grid box would be critical. The absolute mean bias remains however only of about 4 ppbv (13%).
 We then proceed to a comparison of the daily afternoon average during the summer period (June-July-August) to determine the abilities of the model to reproduce O3 during photochemical episodes. Figure 6 presents comparison of daily afternoon averages of surface O3 at different EMEP stations over Europe for the year 1997 from June to August, obtained by the model simulation and provided by the EMEP network. Most of the photochemical events are captured by the model (r2 = 0.56), but the model does not capture the lowest O3 concentrations. This leads to a mean bias of about 5 ppbv (20%) (Table 3).
4. Long-Range Transport From North America to Europe
4.1. Export From the North American Boundary Layer and Transport Across the Atlantic
 The transport associated with mid-latitude cyclones is now recognized as a major pathway for pollution export from the North American boundary layer [Stohl and Trickl, 1999; Cooper et al., 2001, 2002; Stohl, 2001; Trickl et al., 2003]. In particular, two specific airstreams that compose the mid-latitude cyclones, the warm conveyor belt (WCB) (a rising air mass ahead of the cold front) and the post cold front (PCF) airstream (a low level airstream running behind the cold front) have been shown to contribute to pollution transport [Cooper et al., 2002]. The other important process that ventilates the boundary layer is deep convection [Thompson et al., 1994; Horowitz et al., 1998; Q. B. Li et al., Outflow pathways for North American pollution in summer: A global 3-D model analysis of MODIS and MOPITT observations, submitted to Journal of Geophysical Research, 2005] (hereinafter referred to as Li et al., submitted manuscript, 2005) which injects air directly and quickly from the surface to the free troposphere.
 Once out of the North American boundary layer, the pollution follows the general circulation over the North Atlantic area which is mainly determined by the strength and position of the Azores High and the Icelandic Low (Figure 7). In general, in winter and spring, the westerly flow is mostly zonal out of North America, but it turns anticyclonically before reaching the continent (Figures 7a and 7c). The pollution path presents thus a curved shape over the North Atlantic ocean, as already mentioned in previous studies [e.g., Trickl et al., 2003]. By late spring-early summer (Figure 7e), the Icelandic low weakens while the Azores High expands and move northwards. The westerlies decrease in strength throughout the column in summer (Figures 7e and 7f). Part of the flow actually turns anticyclonically around the Azores High and goes back towards North America before reaching Europe, especially at lower levels (Figure 7e). In the upper levels (350 hPa), the westerlies still run quickly across the Atlantic, but they can reach Europe at lower latitudes than in winter and spring (Figures 7b and 7d).
4.2. Simulation of Episodic Pollution Transport Events
4.2.1. Ozone Transport Over the Eastern North Atlantic
 A first examination of the ability of the model to reproduce episodic transport events over the North Atlantic is conducted using observations from the Atmospheric Chemistry Studies in the Oceanic Environment (ACSOE) campaign. The ACSOE campaign took place in April and in September 1997 over the eastern North Atlantic. The C-130 UK Meteorological Office aircraft (based at the Azores) captured several transatlantic transport events [Reeves et al., 2002; Penkett et al., 2004]. On one of these flights (September 14), the aircraft flew south of the Azores, then turned west to intercept the outflow of hurricane Erica from 16 UTC onward, before flying back to the Azores (Figure 8b). Enhanced O3 concentrations (up to 90 ppbv) were observed between 5 and 6 km (Figure 8b) and back trajectory analysis indicated that these polluted air masses were uplifted by a frontal system over North America 3 to 6 days earlier [Penkett et al., 2004].
 The model underestimates O3 by about 10 ppbv (Figures 8c and 8d), but reproduces quite well the variations along the flight track (r2 = 0.75). The model total O3 is enhanced by 20 ppbv between 6 and 8 km from 16 UTC onward, reaching up to 65 ppbv. We find that the simulated hemispheric contribution remains constant at this altitude and at this time, while the North American contribution increases up to 20 ppbv. The model O3 enhancement is thus mostly due to an increase in North American O3, consistently with the back trajectory analysis.
4.2.2. Ozone Import Into the European Upper Troposphere
 We examine here an episode of elevated O3 levels observed in the free troposphere over Europe [Stohl and Trickl, 1999]. The 1997 May 28–29 episode is illustrated in Figure 9a that shows a 26-hour record of O3 concentrations taken by a high-resolution lidar at Garmisch-Partenkirchen in Germany. The lidar observations show the presence of a thin O3 tongue of about 65 to 110 ppbv of stratospheric origin that penetrated deeply into the troposphere as far down as 3 km over the course of May 28–29 [Stohl and Trickl, 1999]. This stratospheric intrusion leads to the O3 peak of about 80 ppbv observed at the Jungfraujoch on May 28 and 29 (Figure 4). Figure 9a also shows the presence of pockets of high O3 concentrations (up to 80 ppbv) at 00 UTC between 6 and 8 km and between 07 and 12 UTC at 8 km on May 29. Air trajectory calculations performed by Stohl and Trickl  revealed that these elevated O3 concentrations were produced in the North American boundary layer and were subsequently lifted to the free troposphere through a WCB ahead of a cold front and transported to the European free troposphere.
Figure 9b shows that the model reproduces the enhanced O3 concentrations on May 28–29 in the upper and middle troposphere. Even though the presence of high O3 pockets is not evident in the simulated total O3 concentrations at high altitudes, tagged O3 and sensitivity simulations clearly reveal that both stratospheric and North American contributions are also found in our model. Figure 9c shows the concentration of stratospheric O3 as simulated by the model over the period May 27–31, and an O3 tongue is clearly seen from May 28 to May 29 that descends down to 3 km, although it is thicker and presents lower concentrations than in the observed data. On May 29, the model shows a pocket of O3 concentrations originated from North America (Figure 9d), with concentrations as high as 20 ppbv and located above the stratospheric tongue between 6 and 8 km, similar to that described in Stohl and Trickl . One should note that the arrival of the high North American concentration is delayed by 16 hours in the model, in comparison with the observed data.
4.3. Seasonal Variation of North American Ozone Entering Europe
 We further extend our analysis of long-range transport to a one-year period by examining the O3 mass fluxes from North America entering Europe through the western, northern and southern European boundaries, as diagnosed by the model sensitivity simulations (Figure 10a; see Figure 7a for the definition of the European domain). North American O3 enters Europe mainly through its western side, but also, to some extent, through the northern boundary because of the anticyclonic curvature of the flow reaching Europe (see section 4.1). The May episode (section 4.2.2) is shown with a black arrow. The September episode (section 4.2.1) is not visible in these fluxes as the pollution was entrained around the Azores high towards North Africa, avoiding continental Europe.
 North American O3 enters Europe all year round but with a strong seasonality, reflecting the seasonality of both the photochemical activity and the transport pathways. Stohl  reported that WCB originate from the eastern seaboard of North America all year round, but their frequency is higher in winter, spring and fall than in summer. The smaller O3 fluxes in winter and early spring are due to the lower O3 production and thus reflect predominantly the strength of the winds.
 North American O3 enters Europe throughout the column in spring while it is restricted to higher altitudes in summer because of a greater chemical loss in the marine boundary layer. In general, low-level transport only contributes to a small extent to the import of O3 into Europe, as winds are weaker and that O3 is efficiently destroyed in the marine boundary layer due to higher water vapor and dry deposition [Fehsenfeld et al., 1996].
 The highest O3 fluxes are seen in the middle and upper troposphere between 8 and 12 km in late spring and summer (Figure 10a). These high altitude fluxes are due to deep convection which lifts pollutants higher than the WCB, as diagnosed by a sensitivity run. Deep convection is especially important in summer over the central and south-eastern United States (Li et al., submitted manuscript, 2005). In addition to this shift in altitude between spring and summer, a seasonal shift in the latitudinal distribution of the O3 fluxes entering the western side of Europe is also observed in our model (Figure 11). In spring, O3 fluxes are at a maximum around 50 to 60°N (because of the anticyclonic curvature of the flow reaching Europe), while the summer O3 fluxes are at a maximum around 40°N, thus having a stronger effect on the Mediterranean basin.
 In addition, the Mediterranean Intensive Oxidant Study (MINOS) aircraft experiment recently pointed out another export process that directly affects Europe [Lelieveld et al., 2002]. From the end of May to the end of August, a warm upper tropospheric anticyclone forms at around 450 hPa over the Tibetan plateau due to its heating after the winter period [Barry and Chorley, 2003]. This leads to an upper easterly current (Figure 12a), delimited by the Intertropical Convergence Zone (ITCZ) at the south end of the Tibetan plateau. These strong easterlies extend westward across South Arabia and Africa and overlay the southwest monsoon winds of the lower troposphere. During MINOS, pollution with an Asian origin was frequently found as far as the Mediterranean Basin in the upper troposphere and even in the lower stratosphere [Lawrence et al., 2003; Scheeren et al., 2003; Roelofs et al., 2003; Traub et al., 2003].
Figure 10b shows the time series of daily Asian O3 mass fluxes entering Europe through its west, south and north sides. Asian O3 fluxes are imported into Europe all year round. The Asian O3 fluxes present a strong seasonal variation with a maximum in summer, especially at high altitude (8 to 12 km). Those high-altitude fluxes reach Europe through its southern boundary in the Mediterranean basin and are associated with the monsoon easterly winds as described previously (Figure 12b). The enhanced convection over Asia injects the pollution in the free troposphere [Scheeren et al., 2003], which explains the high altitude (between 8 km and 12 km) of O3 Asian fluxes simulated in the summer period (Figure 10b).
 Convection during the monsoon season is accompanied by thunderstorms and leads to a high source of NOx from lightning, particularly along the ITCZ [Christian and Latham, 1998; Nesbitt et al., 2000; Li et al., 2001]. By conducting a sensitivity run in which we isolate the role of lightning emissions, we find several events during the summer period in which O3 due to lightning is clearly linked to the enhanced convection and winds system associated with the monsoon. Figure 12d presents those concentrations and fluxes averaged for August. O3 produced by lightning emissions is at a maximum along the ITCZ and at the east of the Tibetan plateau, and is then transported towards Europe by the easterly upper tropospheric winds.
6. Tropospheric O3 Budget Over Europe
 In this section, we define O3 as the extended odd oxygen family (Ox = O3 + NO2 + 2NO3 + PANs + HNO4 + HNO3 + 3N2O5). We will refer to Ox as O3, since O3 usually accounts for 99% of Ox. The European O3 budget calculated with the model is given in Table 4. Figure 13 shows the seasonal variations of the different processes contributing to the European budget for the boundary layer and for the whole troposphere. Positive values correspond to a source for the given region. We also examine the O3 budget over Europe by dividing total O3 into different components (Figure 14), as diagnosed with model sensitivity simulations. Interpretation of the different contributions to total O3 requires caution because of the nonlinear chemistry involved in O3 production. We find that, averaged over the whole European troposphere, the sum of the O3 components does not differ from the total O3 burden (obtained from the standard simulation) by more than 7% in summer and by 4% in winter. The difference is maximal in the boundary layer in summer and can reach up to 10% in June. The nonlinearity in O3 burden is smaller when averaged globally, as the sum of the components does not differ annually by more than 2%. The one-year averaged sums of production and loss rates of individual contributions do not differ from the total O3 production and loss rates by more than 9% and 4%, respectively for the European troposphere, and by more than 2% and 1.6%, respectively on a global scale.
Table 4. European Budget for Tropospheric Ox in the GEOS-CHEM Model
The budget is for the extended odd oxygen family defined as O3 + NO2 + 2NO3 + PANs + HNO4 + HNO3 + 3N2O5 and is calculated over the European region (36°N–72°N, 12.5°W–57.5°E). Values are annual mean for 1997.
The boundary layer is taken from the surface up to 600 hPa (4.1 km).
The troposphere is taken from the surface to the tropopause.
Vertical transport corresponds here to nonconvective transport associated with subsiding air for the boundary layer and with transport from the stratosphere for the whole troposphere.
Vertical transport corresponds here to convective downdraft from the free troposphere to the boundary layer.
The deposition term includes both dry deposition on the ground and wet deposition of O3-related species (e.g., HNO3), therefore it is slightly higher in the whole troposphere than in the boundary layer.
 Net O3 production occurs over Europe all year round (Figure 13) accounting for a total of about 105 Tg O3.yr−1 (Table 4), and is mainly due to European precursors. Net O3 production is more efficient in the boundary layer (due to the proximity of European precursors) and reaches a maximum in summer. This is consistent with the expected increase in summertime photochemical production and vertical extent of the boundary layer. Deposition accounts for a total of 100 Tg.yr−1 and is also at a maximum in summer, following the seasonal variation of chemical production. About half of this process is due to deposition of European O3 and the other half is due to deposition of total background O3.
 Horizontal fluxes of total O3 into the European domain, in the whole troposphere, tend to act as a source (sink) of O3 for Europe in winter (summer) (Figure 13). They present a high variability, also reported by Laurila  in Northern Europe, due to the high variability of the horizontal transport of hemispheric background O3. As expected, the fluxes through the western and northern boundaries act as a source for total O3, while those through the eastern and southern boundaries act as a sink (Table 4). Horizontal advection has a larger impact in the free troposphere as winds are stronger. Throughout the tropospheric column, European O3 is exported all year round, while North American and Asian O3 are imported all year round. North American horizontal O3 fluxes are greater than the Asian ones because North American sources are closer to Europe. We can note that, as discussed in the previous section, transport of North American and Asian O3 through the southern boundary acts as a source, especially in summer. Total O3 mass horizontal fluxes act as a net sink in the boundary layer; although North American and Asian O3 are imported all year round, they do not compensate the export of European O3. European horizontal fluxes are higher in the boundary layer than in the free troposphere, as already noticed by Wild et al.  and present a maximum in summer. Duncan and Bey  showed that the major pathways for pollution export in summer is towards Russia and the Mediteranean Basin/North Africa.
 Vertical transport includes convective processes (when O3-rich air is transported in the convective downdrafts from the free troposphere to the boundary layer) and nonconvective processes (when air subsides from the free troposphere to the boundary layer).
 Convective downdraft of total O3 is a small net source for the boundary layer, as already reported by Langmann et al. , especially in summer (Table 4 and Figure 13). However, convective lifting of European O3 leads to a net export of O3 from the boundary layer, especially during the summer months. This is consistent with previous findings of Duncan and Bey , who reported that areas over Germany and the Ural Mountains in Russia frequently experience deep convection, and participate to pollution export from Europe. Conversely, convective downdraft of North American, Asian and hemispheric background (not shown) fluxes act as a net source for the boundary layer, and are at a maximum in summer. During that season, convective downdraft and horizontal transport of North American and Asian O3 to the boundary layer contribute to the same extent.
 Nonconvective subsidence, which is a source for both the boundary layer and the whole troposphere should be either associated with large-scale synoptic systems [Cooper et al., 2004; Heald et al., 2003] or with dry air intrusions [Stohl and Trickl, 1999; Trickl et al., 2003; Jaeglé et al., 2003]. North American, Asian and hemispheric background O3 subside into the European boundary layer, acting as a net source. Subsidence of Asian O3 in the boundary layer is constant all year round, while subsidence of North American O3 peaks in July. The nonconvective mass flux entering at the top of the tropospheric column corresponds to input from the stratosphere (56 Tg.yr−1, Table 4) and shows two maxima, in late winter-early spring and mid-summer. James et al.  reported that the contribution of stratospheric O3 to tropospheric O3 throughout the whole troposphere in the Northern Hemisphere is likely to be higher from February to May and lower during the fall months, in agreement with our findings. Moreover, this seasonal cycle is somewhat consistent with the findings of Sprenger and Wernli , who reported that the net stratosphere-troposphere exchange in the extratropical Northern Hemisphere is at a maximum in winter, but also noted a significant downward flux in summer over the continents.
6.3. Burden Over Europe
 The total O3 burden integrated over the European tropospheric column ranges from 12.5 Tg (January) to 15.7 Tg (July) (Figure 14). Integrated over the whole troposphere, the hemispheric background O3 (9.0 Tg) accounts for 64% of total tropospheric O3 burden (including 10% due to lightning and 20% due to stratospheric O3) and shows only little seasonal variation. The hemispheric background O3 reaches a maximum over the Mediterranean basin, as well as the lightning and stratospheric contributions (not shown). A lot of the seasonal variation of the total O3 burden in the whole troposphere is driven by the summer increase in the European contribution, which accounts for 16% of the total O3 in summer and less than 3% in winter. At the surface, we find that European contribution is negative for some winter months due to the O3 titration by NOx (see Figure 15). As also reported by Wild et al. , the contribution of European O3 is greater in the boundary layer (e.g., 27% in July) than in the free troposphere (e.g., 8% in July), due to the proximity of local sources and to a relatively low venting of the boundary layer. The European O3 is at a maximum over the Mediterranean basin at the surface (35 ppbv in summer), over southern central Europe in the boundary layer and over eastern central Europe in the free troposphere (not shown).
 North American and Asian O3 contribute to about 11.0% and 7.7%, respectively, to the total tropospheric column over Europe on annual average. Those contributions present two maxima in spring and fall in the boundary layer (as well as at the surface, see Figure 15), due to stronger low-altitude fluxes during these seasons (see section 4), and a maximum in summer in the free troposphere due to enhanced O3 production over the source regions. In fact, we find that the sum of the North American and Asian contributions is higher than the European contribution all year round both in the free troposphere and in the whole troposphere, indicating the significance of intercontinental transport of pollution to the total O3 budget over Europe. Even taken individually, both contributions of North America and Asia are higher than that of Europe in the free troposphere. North American and Asian contributions experience a maximum at the surface in spring over Western, Northern Europe and at the north of the Mediterranean basin (7.5 ppbv). These contributions in the free troposphere go through a maximum in summer over the Mediterranean basin, the European region with the highest O3 concentration [Lelieveld and Dentener, 2000; Kourtidis et al., 2002].
 We find that total background O3 averaged over continental Europe reaches a maximum at the surface (30 ppbv) in early spring and in the free troposphere (52 ppbv at 550 hPa) in summer (not shown). The highest concentrations are found in the Mediterranean basin. At the surface, this appears to be slighty lower than observations (28–42 ppbv [Simmonds et al., 1997; Derwent et al., 1998; Brönnimann et al., 2000; Pochanart et al., 2001; Naja et al., 2003]). However, the spring maximum is well reproduced by the model and appears to be due to several factors, including a maximum in North American and Asian contributions, as well as stratospheric input. The background O3 concentrations increase with altitude and are observed to range from 46 to 55 ppbv in locations representative of free tropospheric conditions [Brönnimann et al., 2000; Bonasoni et al., 2000; Naja et al., 2003], in agreement with our findings.
7. Impact of Anthropogenic Emission Changes for the Period 1980–1997
 As already pointed out, it is difficult to draw a clear picture of the trends in O3 over Europe for the last two decades, partly because different trends are found in the lower troposphere in comparison with those observed at higher altitudes (e.g., at the mountain sites) and also because they depend on the quantities examined (e.g., average versus maximum). The major parameters contributing to long-term changes in tropospheric O3 concentrations include O3 precursor emissions, stratospheric O3 input, CH4 levels, ultraviolet (UV) radiation, meteorology, temperature, and humidity [e.g., Vukovich et al., 1977; Guicherit and Van Dop, 1977; Beekmann et al., 1994; Fuglestvedt et al., 1994; Sillman and Samson, 1995; Lelieveld and Dentener, 2000; Fusco and Logan, 2003]. Fusco and Logan  have investigated the changes in those different processes from 1970 to 1995. They find that the major factors influencing tropospheric O3 are the decline of the stratospheric O3 flux in the free troposphere, the global increase in NOx surface emissions and the increase in the lower troposphere temperature.
 Previous sections of this paper have also clearly highlighted that long-range transport of O3 significantly affects the levels of O3 over Europe, especially in the middle and upper troposphere, while the effect of regional (European) emissions is preferentially seen in the lower troposphere. We suggest that the variation with altitude of the O3 trends observed over Europe may also reflect, to some extent, the impact of the long-range transport of O3 pollution. We further investigate this point by examining the variations in O3 concentrations over Europe at various altitudes due only to the changes in anthropogenic emissions in Europe, North America and Asia over the last two decades. For that purpose, we conducted a number of sensitivity studies where anthropogenic emissions from North America, Asia and Europe were successively fixed to 1980 levels, while keeping the 1997 meteorology, and the CH4 concentrations, UV radiation, stratospheric input at their 1997 levels.
 The O3 precursor emissions used in our model for these two years are presented in Table 5. NOx decreased by 15%, CO by 37% and VOC by 29% over the period 1980–1997 in Europe, following the EMEP recommendations. This decrease is maximal over Central Europe, but there is a local increase in NOx emissions in Iceland, Ireland and in the southwest Europe (Spain and Portugal) and the southeast Europe (Croatia, Cyprus, Greece and Turkey) [Vestreng and Klein, 2002]. We used data from EPA  to estimate the 1980 emissions over North America, as described by Fusco and Logan . In our model, NOx emissions increase by 6%, while CO and VOC emissions decrease respectively by 12.5% and 13%. However, it should be noted that a new emission database recently provided by EPA  indicates that both NOx and CO emissions have declined by 8.5% and 36%, respectively over the period 1980–1997. These newer numbers appear to be closer to the reality, at least for CO, as discussed by Parrish et al. , who found a larger decrease in CO emissions due to on-road vehicles than previously reported in the EPA . However, the authors also suggest an increase in vehicle NOx emissions over North America of 2–3% per year due to the use of heavy-duty diesel-powered trucks, which is more in agreement with EPA . For Japan, we use data from the Organization for Economic Cooperation and Development (OECD) , while for the rest of Asia, we used emissions scaled to annual fossil fuel CO2 emission statistics provided by the Carbon Dioxide Information Analysis Center (CDIAC) [Marland et al., 2001]. For the period 1980–1997, we found a large increase in NOx (x2.2), CO (x2.6) and VOC (x1.3) anthropogenic emissions over Asia. Asian emissions are still largely uncertain, however, previous modelling studies have shown that O3 simulated over that region with the GEOS-CHEM model are in good agreement with observations [e.g., Bey et al., 2001b; Liu et al., 2003]. Van Aardenne et al.  reported an increase by a factor of 1.7 for NOx over Asia from 1990 to 2000, in good agreement with our number.
Table 5. NOx, CO, and VOC Emissions in the Different Geopolitical Regions in the GEOS-CHEM Model
NOx, Tg N.yr−1
CO, Tg CO.yr−1
VOC, Tg C.yr−1
Figure 15 shows the seasonal variation in 1980 and 1997 of total O3 and of the European, Asian and North American O3 contributions (all quantities are averaged over continental Europe). Figure 16 shows the spatial distribution at the surface in winter and summer of the change between 1980 and 1997 in total O3, European O3 and total background O3. As hemispheric background does not vary in our simulations, total background O3 represents only changes in Asian O3 and in North American O3. Because of the nonlinearity in the O3 chemistry, more emphasis should be given to the tendencies rather than the actual concentrations or changes in concentrations.
 Between 1980 and 1997, we find that the North American O3 contribution remains fairly constant, while the Asian O3 contribution increases by about 1 ppbv at the surface (Figure 15). This results in a net increase in the total background seen in all seasons at the surface over the whole continental Europe (Figure 16). The annual mean continental-averaged European O3 decreases by 0.6 ppbv, between 1980 and 1997 and up to 1.5 ppbv in July (Figure 15) due to the reduction of local O3 precursor emissions. As the O3 distributions at the surface over Europe are mostly driven by the European emissions (see section 6 and Wild et al. ), the change in emissions between 1980 and 1997 leads to a decrease in total O3 at the surface in summer. One should note however that, in summer, at the surface, total O3 and European O3 decrease by up to 5 ppbv over Central Europe, while they increase by up to 2–3 ppbv over southwest and southeast Europe (Figure 16). In winter, European O3 increases in central Europe due to the reduced O3 titration by NOx, and decreases over the other parts of Europe (Figure 16). However, as total background increases, total surface O3 also increases over all Europe in winter.
 At higher altitude, we find that the decrease in European O3 due to the reduction of local emissions is compensated by the increase in Asian O3, resulting in a small increase in total O3 concentrations (Figure 15). Even though it should be recognized that only changes in emissions are accounting for in our simulations, this compensating effect may explain the small (or lack of) trends observed in O3 over Europe for the last two decades, especially at the mountain sites.
 We also examine the sensitivity of O3 to the change in CH4 levels from 1980 to 1997. CH4 levels are specified in the model using the set of CMDL observations. These observations are available for 1983–2001, and were scaled for the year 1980 using a polynomial regression for each latitude band. CH4 increases globally from 1560 to 1730 ppbv between 1980 and 1997. We find that, over Europe, the changes in O3 concentrations due to the changes in CH4 levels are smaller that those induced by the changes of other anthropogenic emissions: hemispheric background O3 increases up to 1.2% (0.3 ppbv), while Asian O3 contribution increases up to 1.6% (1.4 ppbv) in the whole troposphere (at the surface).
Figure 16 presents the change in O3 peaks (computed as a 0.98th percentile) for the summer period, between 1980 and 1997, for total O3, European O3 and total background O3. In summer, total O3 peaks decrease in central Europe by up to 12 ppbv. This is in accordance with previous studies over England [Derwent et al., 2003] and Switzerland [Brönnimann et al., 2002]. However, in southwest Europe and southeast Europe O3 peaks increased by up to 4 ppbv.
Figure 17 shows the change between 1980 and 1997 in hourly mean for total O3, European, Asian and North American O3 for different locations representative of western Europe (Mace Head), central Europe (Waldhof), and the alpine site of the Jungfraujoch. For each station, the change in total O3 has to be interpreted differently, although change in North American O3 is systematically close to zero while Asian O3 increases at all sites. Mace Head is a coastal station where European emissions have only a small impact (Figure 16, 2nd row, 2nd column), and thus presents only a small decrease in European O3 in summer (<2 ppbv) but an episodic increase in the European O3, particularly in fall and winter (up to 6 ppbv) due to a weaker O3 titration by NOx. Asian O3 increases all year round by less than 2 ppbv with a minimum impact in summer, leading to an increase in total O3 in fall, winter and spring. In summer, total O3 is consequently driven by European O3.
 The Waldhof station is at the heart of the area impacted by the changes due to European emissions (Figure 16), and presents an increase in O3 in winter and a decrease in summer. Asian O3 increases (of 0.8 ppbv averaged on the year), while North American O3 presents a small increase in winter. At that station, the change in total O3 is thus driven by the change due to European emissions all year round.
 The Jungfraujoch station is less sensitive to the local emissions as it is usually typical of free tropospheric conditions (see section 6). European O3 decreases during photochemical episodes in summer and Asian O3 increases all year round by 1.5 ppbv, while North American O3 remains more or less constant. Thus, total O3 increases all year round (+1.4 ppbv) due to the change in Asian O3 except during strong photochemical episodes in summer. This last section illustrates the difficulty of assessing O3 trends over Europe, as each station appears to be differently impacted by various contributions.
8. Summary and Conclusion
 We use a global chemical transport model to investigate the processes associated with pollution import into Europe and to assess its impact on the European O3 budget. Sensitivity simulations enable us to separate O3 fluxes and concentrations into different components, including hemispheric background (which includes stratospheric input and O3 produced by precursors from natural sources, biomass burning, and by CH4 oxidation) and the contributions due to anthropogenic sources over Europe, North America and Asia. We also examine the change in different O3 contributions over Europe due to the change in local and continental emissions over the period 1980–1997.
 The model reproduces generally well O3 concentrations over Europe throughout the column with the exception that O3 concentrations are overestimated in summer at the surface at some sites, underestimated in the middle troposphere in summer, and that the seasonal cycle of O3 is too flat in the upper troposphere. We also examined two events of O3 enhancements of North American origin over the eastern North Atlantic and over Europe, and showed that the model reproduces well the features associated with long-range transport of pollution.
 The altitudinal and latitudinal distributions of North American O3 fluxes entering Europe show a clear seasonal signal, reflecting the seasonal variation in export pathways and photochemical activity. North American O3 fluxes into Europe are at a maximum in spring and summer because of the stronger photochemical activity. The pollution usually follows a curved shape over the North Atlantic, and therefore enters Europe on both its western and northern sides. In summer, North American O3 enters Europe at higher altitudes and lower latitudes because deep convection lifts pollution to higher altitudes than warm conveyor belts in general, and the flow over the North Atlantic is mostly zonal in that season. The low-level inflow is only important in spring, when loss rates in the boundary layer over the North Atlantic are at a minimum.
 Asian O3 reaches Europe all year round, but usually at higher altitudes than North American O3. Over Asia, the pollution is rapidly lifted to the free troposphere through convection or mid-latitude cyclones and is then transported around the world in the westerly flow. In addition, during the monsoon season (summer), the anticyclone centred over the Tibetan plateau extends far to the west and induces an easterly flow by which Asian pollution is directly transported towards Europe, especially in the Mediterranean area. We note both a strong contribution of lightning and Asian pollution in this O3 import.
 Total O3 burden integrated throughout the tropospheric column over Europe ranges between 12.4 Tg in winter and 15.8 Tg in summer. Hemispheric background accounts annually for 64.4% of total O3 burden, while North American, Asian and European contributions account for 10.9%, 7.7% and 9.4%, respectively. Most of the seasonal variation of the European O3 burden is driven by the European emissions. In the free troposphere, Asian and North American contributions are highest in summer due to the greatest chemical O3 production over the source regions. This is also the case in the low troposphere in spring and fall, when net chemical O3 loss are weaker over the North Atlantic. Total background (the sum of the hemispheric background plus the Asian and North American contributions) at the surface presents a maximum in spring of 30 ppbv.
 From 1980 to 1997, we find that O3 concentrations at the surface decrease in summer because of the decrease in European O3 precursor emissions and increase in winter because of i) the weaker O3 titration by NOx and ii) the increase in Asian contribution. In the free troposphere, the change in O3 is driven by the increase in Asian contribution, which compensates for the decrease due to the local emission reductions in all seasons. We find that these changes are larger than those due to the global CH4 increase. This result has political implications as the effect of the European emissions reductions may be offset, to some extent, by the increase in foreign emissions. The need for international treaties regulating the emissions is thus reinforced [e.g., Holloway et al., 2003].
 It should be noted that the results presented here are for 1997. It has been suggested that the distributions of different chemical tracers over the North Atlantic is related to the North Atlantic Oscillation (NAO) [Li et al., 2002; Creilson et al., 2003; Duncan and Bey, 2004], which has a strong year-to-year variability. Further work is required to investigate the sensitivity of our results to interannual variability of NAO.
 Finally, uncertainties still remain about the model's quantitative ability to reproduce events of long-range transport. In particular, only limited data are available to constrain the evaluation of the model's chemical loss and production rates. In future work, we will examine the O3 production terms during long-rang transport events with a special focus on the individual chemical reactions driving those chemical terms. The Intercontinental Transport of Ozone and Precursors (ITOP) campaign, which took place over the North Atlantic in summer 2004, will provide a valuable database to better characterize the chemical environment in the polluted air masses travelling over the North Atlantic and their impact on Europe.
 The GEOS-CHEM model is managed by the Atmospheric Chemistry Modeling Group at Harvard University with support from the NASA Atmospheric Chemistry Modeling and Analysis Program. We are very grateful to Anne-Gunn Hjellbrekke for providing the EMEP ozone observations, the Swiss Agency for Environment Forests and Landscape (SAEFL) for providing NABEL data at the Jungfraujoch, and Thomas Trickl for the lidar observations at Garmisch-Partenkirchen. We would like to express our appreciation to Philippe Thunis for his assistance with the EMEP emission inventories. Discussions with Bryan Duncan, Arlene Fiore, Mathew Evans, David Parrish, and Johannes Staehelin were very helpful.