Influence of El Niño–Southern Oscillation on the interannual variability of tropospheric ozone in the northern midlatitudes



[1] We use the Goddard Earth Observing System Chem (GEOS-Chem) model to interpret long-term measurements of tropospheric ozone (O3) and carbon monoxide (CO) and to investigate the factors that contribute to their interannual variation (IAV) during the period from 1987 to 2005. The model reproduces relatively well the observed IAV of CO. The simulation of O3 IAV is not as successful. In particular, the negative anomalies in 1991–1993 and the following upward trend in 1993–1996 observed at several sites in the northern midlatitudes are not reproduced by the model, which may result from a poor representation of stratospheric chemistry and dynamics. We examine in detail the period of 1998–1999 when a large anomaly in tropospheric ozone column is observed and simulated over Europe (maximum of +4.9 Dobson units). Three consecutive periods can be distinguished from January 1998 to April 1999, during which different processes affected the O3 burden over Europe. Spring 1998 is largely influenced by the preceding 1997 El Niño that affects (1) stratosphere-troposphere exchange and (2) Asian pollution export and transport toward Europe by a change in convective activity in East Asia and a strengthening of the subtropical jet stream. An enhanced pollution export from North America is also noticed for this period. The second period (summer-fall 1998) shows a mixed influence from both boreal wildfires and Asian pollution. The third period is influenced by enhanced wildfires in Southeast Asia. Throughout the period from 1987 to 2005, positive anomalies in tropospheric O3 column and in surface O3 are found over Europe in the spring following an El Niño year.

1. Introduction

[2] Tropospheric ozone (O3) is a key atmospheric compound that contributes to several important environmental issues. O3 is a direct greenhouse gas [Intergovernmental Panel on Climate Change, 2001] and a dangerous air pollutant that causes serious health impacts and damages ecosystems, agricultural crops, and materials [European Environment Agency, 2003]. In addition, photolysis of O3 leads to the production of hydroxyl radicals (OH) which, in turn, control the oxidizing capacity of the atmosphere and hence the rates at which many natural and anthropogenic compounds are eliminated from the atmosphere. Research conducted over the last decades provided evidence that the spatial and temporal distributions of tropospheric O3 are determined by a complex interaction between photochemistry and transport processes. The two sources of O3 in the troposphere are transport from the stratosphere and in situ photochemical production through oxidation of hydrocarbons in presence of nitrogen oxides (NOx = NO + NO2); these two sources result in about 552 ± 168 Tg(O3) a−1 and 5110 ± 606 Tg(O3) a−1, respectively, for the present day conditions according to a recent multimodel analysis [Stevenson et al., 2006].

[3] While the processes that control the distributions of tropospheric O3 have been relatively well documented in general, less attention has been paid to the interannual variability (IAV) of O3. Fusco and Logan [2003] found that O3 trends from 1970 to 1994 are most significantly influenced by the increase in surface NOx emissions from fossil fuel combustion and by the stratospheric O3 source. Karlsdottir et al. [2000] found that surface emission variations are better reflected in CO than in O3. They also found that changes in stratosphere-troposphere exchange (STE) significantly impact tropospheric O3 concentrations and should be accounted for to reproduce the trends in the high latitudes of the Northern Hemisphere. Other factors that have been suggested to contribute to O3 IAV include tropospheric temperatures, solar radiation, and dynamics (e.g., patterns of climate variability, like the El Niño Southern Oscillation (ENSO) or the North Atlantic Oscillation) [Kim and Newchurch, 1996; Ziemke and Chandra, 1999; Karlsdottir et al., 2000; Lelieveld and Dentener, 2000; Thompson et al., 2001; Peters et al., 2001; Creilson et al., 2003; Fishman et al., 2005; Fusco and Logan, 2003; Stohl et al., 2003; Zhou et al., 2003; Doherty et al., 2006; Thouret et al., 2006]. Nevertheless, the relative contribution of individual factors to the O3 IAV remains poorly quantified. In the present study, we interpret long-term measurements of O3 using multiyear O3 simulations performed with the three-dimensional (3-D) global Chemical Transport Model (CTM) GEOS-Chem over the period 1987–2005. We also document the IAV of carbon monoxide (CO) that is one of the major O3 precursors. The main objectives of our work are to provide quantitative insights into the processes that drive tropospheric O3 and CO IAV. In particular, we examine the influence of year-to-year changes in O3 precursor emissions (anthropogenic, biomass burning, and lightning), methane (CH4) concentrations, total O3 column, and meteorology. We focus on the midlatitudes of the Northern Hemisphere where most of the long-term ground-based data are available.

[4] The paper is organized as follows. A description of the model and simulation set up is given in section 2. In section 3 we discuss the IAV of O3 and CO in the northern midlatitudes using several long-term data sets and we show the model's capabilities to reproduce those temporal variations. Section 4 is dedicated to the analysis of the specifically large 1998–1999 O3 anomaly in Europe. Sections 5 and 6 describe the possible influence of ENSO on this anomaly and section 7 expands on this analysis to examine the influence of ENSO on the European O3 IAV over the past 2 decades. Section 8 provides a summary and conclusions.

2. Modeling Setup

2.1. GEOS-Chem Model

[5] We use the GEOS-Chem model, a global CTM driven by assimilated meteorological data from the NASA Global Modeling and Assimilation Office (GMAO) [Bey et al., 2001a]. The work shown here employed the version 7-02-04 of GEOS-Chem ( driven by the GEOS-4 meteorological fields that are available from 1987 to 2005. Meteorological fields include winds, surface pressure, temperature, cloud properties, heat flux, precipitation, and other quantities that are provided with a 1° latitude by 1.25° longitude horizontal resolution and 55 layers in the vertical from the surface up to 0.01 hPa. The resolution was degraded to 4° × 5° in the horizontal and to 30 layers in the vertical so that several multiyear simulations can be conducted. The model includes a detailed description of tropospheric O3-NOx-hydrocarbon chemistry. It solves the chemical evolution of about 120 species with a Gear solver [Jacobson and Turco, 1994] and transports 24 tracers. Photolysis rates are computed interactively in the model using the Fast-J radiative transfer algorithm [Wild et al., 2000] which includes Rayleigh and Mie scattering by clouds and aerosols. The variability in total O3 column needed as input to Fast-J is accounted for in our model by using the TOMS O3 data set ( Missing data during our study period are replaced by climatological values. The model uses aerosol fields from the GOCART model [Chin et al., 2002; Ginoux et al., 2004; Martin et al., 2003] as input to Fast-J and for heterogeneous chemistry. Heterogeneous reactions of HO2, NO2, NO3 and N2O5 on sulphate aerosols, mineral dust, black carbon, organic carbon and sea salt are accounted for as described by Martin et al. [2003]. Note that aerosol distributions do not vary interannually in our simulation. The model uses the TPCORE scheme from Lin and Rood [1996] for the advection. The deep and shallow convection are described with the Zhang and McFarlane [1995] and the Hack [1994] parameterizations, respectively. The cross-tropopause flux of O3 is specified using a flux boundary condition to obtain a more realistic annual cross-tropopause O3 flux. The model uses the SYNOZ (synthetic ozone) method developed by McLinden et al. [2000]. With this method, an O3 production rate of 450 Tg a−1 is applied in the tropical stratospheric region (from 30°S to 30°N and from 70 hPa to 10 hPa) that leads to a global annual cross-tropopause flux between 420 to 470 Tg(O3) a−1 during our study period.

[6] Year-to-year variations in anthropogenic emissions of trace gases are accounted for using a base emission inventory for 1985 described by Wang et al. [1998b], that includes NOx emissions from the Global Emission Inventory Activity (GEIA) [Benkovitz et al., 1996], nonmethane hydrocarbon (NMHC) emissions from Piccot et al. [1992], and CO emissions from Duncan et al. [2007], scaled for specific years as described by Bey et al. [2001a]. For United States and for Europe, emission inventories such as those provided by Environmental Protection Agency [1997] and Berge [1997], respectively, are used for the scaling factors [Bey et al., 2001a]. In addition, for the United States we use the Environmental Protection Agency (EPA) National Emissions Inventory (NEI) 1999 v.1 inventory (NEI99,, with some modifications as described by Hudman et al. [2007], for the year 1999. For the rest of the world, NOx emissions are scaled using trends in CO2 emissions from fossil fuel combustion, and CO and NMHC emissions are scaled using trends in CO2 emissions from liquid fuels [Bey et al., 2001a]. Yearly CO2 emissions are provided by Marland et al. [1999], who primarily used energy statistics published by the United Nations [1998]. In that version of the model, anthropogenic NOx and CO emissions for the period 1998 to 2005 are set to the values of 1998 (except for U.S. emissions in 1999). However, as our analysis mainly focuses on the period from 1987 to 2000, we argue that this strong assumption does not affect significantly our conclusions.

[7] Monthly biomass burning emissions are derived from an inventory of the total annual biomass burned described by Lobert et al. [1999] and Duncan et al. [2003b]. Annual biomass burned is converted to CO emissions by applying emission factors [Andreae and Merlet, 2001; Cofer et al., 1998; Yevich and Logan, 2003], thus providing a climatological inventory for biomass burning emissions. Interannual variations are further accounted for using the TOMS Aerosol Index product [Hsu et al., 1996; Herman et al., 1997; Torres et al., 1998] from 1987 to July 1996 following Duncan et al. [2003b] and the Advanced Along Track Scanning Radiometer (AATSR) [Arino and Melinotte, 1995] active fire data set from August 1996 to 2005 following Generoso et al. [2003] with slight improvements as described in the following. The latest available AATSR data (up to 2005) were included to compute scaling factors for various regions [see Generoso et al., 2003, Figure 1] which are used to scale the climatological inventory. Missing days of AATSR detection within a given month were filled using the daily average of fires detected during that month. In addition, since the presence of clouds prevents the detection of underlying fires, we applied a correction by weighting the number of fires by (1-C)−1 where C is the monthly cloud cover fraction provided by the International Satellite Cloud Climatology Project (ISCCP) [Rossow and Schiffer, 1999].

[8] Ship emission estimates are from Benkovitz et al. [1996] and are constant throughout the period (1.6 TgN a−1). We implemented a +3% annual increment of total NOx aircraft emissions in GEOS-Chem (0.5 TgN in 1992) following the Intergovernmental Panel on Climate Change [1999] and Environmental Protection Agency [2000]. Emissions of NOx from lightning (5.7 TgN a−1 on average, ranging from 5.1 to 6.3 TgN a−1 during the 19-year period) are linked to deep convection following the parametrization of Price and Rind [1992] as described by Wang et al. [1998b]. The year-to-year varying CH4 concentrations are specified for each latitudinal band using the Climate Monitoring and Diagnostics Laboratory observations (CMDL) up to 2004.

2.2. Model Simulations

[9] We first performed a 19-year control simulation from January 1987 to December 2005 (labeled “S0”) using the model configuration described above. We then carried out six sensitivity simulations with one parameter kept fixed (Table 1). We successively examined the role of changing biomass burning emissions (set to climatological values in the “S1” simulation), meteorology (set to 1988 in the “S2” simulation), anthropogenic emissions (set to the 1988 values in the “S3” simulation), total O3 column (set to the 1988 values in the “S4” simulation), CH4 concentrations (set to the 1988 values in the “S5” simulation) and lightning NOx emissions (set at 6 TgN/a in the “S6” simulation). Results from the sensitivity simulations are useful to diagnose the relative contribution of individual processes to the O3 IAV although one has to keep in mind the strong nonlinearity that characterizes the O3-NOx-VOC system to quantitatively interpret the simulations. Note that the “fixed meteorology” simulation accounts for a number of processes, including the changes in transport, water vapor, and lightning emissions.

Table 1. Simulations Performed With the GEOS-Chem Model Over the Period 1987–2005 in the Present Work
  • a

    Using production rates from S0.

  • b

    Using production rates from S1.

Control Simulation
S0all parameters vary interannually-
Sensitivity Simulations
S1biomass burning emissionsclimatology
S3anthropogenic emissions1988
S4total O3 column1988
S5CH4 concentrations1988
S6lightning NOx emissions6 TgN
“Tagged” Simulations
T0aall parameters vary interannually-
T1bbiomass burning emissionsclimatology

[10] Two additional so-called “tag” O3 simulations were also conducted. Tagging O3 involves archiving 3-D fields of daily mean production and loss frequencies of the extended odd oxygen family (Ox = O3 + O + NO2 + 2NO3 + PANs + HNO4 + HNO3 + 3N2O5) from a standard simulation. Ox will be hereafter referred to as O3 since O3 usually accounts for more than 95% of Ox. The archived production and loss frequencies are then used to drive an off-line tag O3 simulation, in which total O3 is divided into different tagged tracers, each of them being produced in a specific region of the atmosphere following the methods proposed by Wang et al. [1998a]. We used several source regions including the lower, middle, and upper troposphere of the major continental (e.g., Asia, Europe, and North America) and oceanic (the Pacific and Atlantic oceans) regions as well as the stratosphere. In order to conduct additional analysis, we further decomposed the Asian region into four characteristic regions, including Indonesia, Southeast Asia/India, China and Siberia, and the North American region into two subregions, including United States and Canada. We conducted two tagged O3 simulations, a first one (T0) that uses the production rates archived from the standard S0 simulation, and a second one (T1) that uses the production rates archived from the sensitivity simulation S1 (i.e., with fixed biomass burning emissions). The interpretation of results for tagged ozone simulation requires some caution due to the nonlinearity of the chemistry involved in O3 production and the possible transport of O3 precursors from another region into the region of production [Li et al., 2002].

3. Interannual Variability in the Northern Midlatitudes

3.1. CO

[11] In this section, we discuss the CO IAV using the National Oceanic and Atmospheric Administration/Earth System Research Laboratory (NOAA/ESRL) ground-based CO measurements (Table S1 in the auxiliary material) conjointly with a series of sensitivity simulations. Note that a brief discussion on the ability of the model to reproduce the seasonal variation in monthly mean CO concentrations is provided in the online auxiliary material (Figure S1 and text therein). Observed and simulated CO anomalies are computed as the difference between the monthly mean concentration for a given year and the climatological monthly mean value calculated over the 19-year period for the model or over the available period of observations (Figure 1).

Figure 1.

Time series of observed (black) and simulated (red) CO anomalies. Solid fine line shows monthly mean; circles show 12-month running averages. The correlation coefficient (r2) between the observed and the simulated running averages is indicated for each site along with the length of data record (number of months in parentheses). Observations are from the National Oceanic and Atmospheric Administration/Earth System Research Laboratory (NOAA/ESRL) (

[12] The model captures the main features of observed CO anomalies at most of the sites (Figure 1). Three sites show correlation coefficients (r2) below 0.3, including Mace Head (0.22), Mauna Loa (0.27), and Tae Ahn (0.09). The small correlation at Tae Ahn may be related to local sources that are not well accounted for in the model [Duncan et al., 2007]. The model reproduces well the downward trend from the early 1990s until 1997 [Intergovernmental Panel on Climate Change, 2001; Duncan et al., 2007, and references therein] in the northern sites (e.g., Barrow, Alert, Mace Head, Ulaan Uul; Figure 1). This trend is related to the decline of European anthropogenic emissions and is seen at almost all the sites in the subtropical and higher latitudes, although it is more evident in the high latitudes and midlatitudes, as further discussed by Duncan et al. [2007].

[13] After 1997, a larger IAV is seen at the midlatitude and high-latitude sites. Results from the sensitivity simulations indicate that this is largely due to biomass burning events during the period from 1998 to 2005, although the year-to-year changes in meteorology also influence the CO surface levels to some extent (e.g., in Mace Head in 2002) (Figure 2). Large enhancements in CO are seen in 1998, 2002, and 2003 when severe burning events occur in boreal regions and primarily in Siberia [Yurganov et al., 2005; Duncan et al., 2007]. Although the simulated anomalies are always less pronounced than in the observations, the model reproduces rather well these anomalies in most of the northern stations (e.g., Barrow, Alert, Niwot Ridge, Ulaan Uul). This indicates that the multiyear CO inventory for biomass burning inventory is likely reasonable although the simulation of continental outflow from the boreal regions could benefit from several improvements (especially in terms of injection height and temporal resolution of the inventory) [Generoso et al., 2007].

Figure 2.

CO anomalies in the standard simulation (black) at Barrow, Mace Head, Niwot Ridge, Tutor Hill, Mauna Loa, and Guam. The green line shows the difference between the CO concentrations in the standard simulation (S0) and those from the simulation with climatological biomass burning emissions (S1). The blue line shows the difference between the CO concentrations in the standard simulation (S0) and those from the simulation with fixed meteorology (S2). The circles represent the 12-month running averages.

[14] The observed downward trend in CO anomalies in the midlatitude and high-latitude sites levels off almost completely after 1998 (Figure 1). It is possible that the large amounts of CO emitted from the biomass burning events of 1998, 2002 and 2003 influenced the overall background CO concentrations in the Northern Hemisphere. There are also some indications that the downward trend in European anthropogenic CO emissions slowed down in the later part of the study period according to data from the expert emissions developed by the European Monitoring and Evaluation Program (EMEP) [Vestreng et al., 2006]. The decrease in the EMEP total European CO emissions during the period from 2000 to 2005 is −1.1% a−1, against −3% a−1 in the 1990–2000 decade (as shown in Table S2 and Figure S2 in the auxiliary material). This change is most largely due to an increase in CO emissions in eastern Europe (+1.5% a−1), which are currently largely unregulated and because road transport emissions are increasing again (+1.9% a−1) after the economic stagnation of the 1990s [United Nations Environment Programme, 2004]. A more modest slowdown in the CO emissions is also seen in the western Europe (from 3.3% a−1 in 1990–1999 to 2.9% a−1 in 2000–2005). Overall, this resulted in a slowdown of the CO anthropogenic emission downward trend over Europe. The CO emissions over the USA [Environmental Protection Agency, 2007] (data from EPA/NEI: also showed a similar slowdown between the period 1990–1999 (−2.9% a−1) and the period 2000–2005 (−2.1% a−1), and probably also contributed to the leveling off of the downward trend in CO anomalies. As the model includes constant anthropogenic emissions after 1998, it is not possible to further analyze this issue.

[15] Similarly to other studies [e.g., Szopa et al., 2007], we find that the CO IAV in the model is largely controlled by variations in meteorology and biomass burning emissions (Figure 2). In order to measure the influence of individual factors, we compute the global median of the absolute differences between monthly mean CO column from the control simulation and that from each sensitivity simulation. We find that the simulation with fixed meteorology and fixed biomass burning emissions show the highest deviations (4.6% and 2.3%, respectively) from the CO column in the control simulation during the study period. Fixed anthropogenic emissions and fixed methane concentrations lead to deviations in CO of 2% and 1.5%, respectively, in comparison with the control run. Results from the other sensitivity simulations (with fixed total ozone column and fixed lightning emissions) are not shown because the deviations from the reference simulation are even smaller (below 1%).

[16] IAV in the tropics and subtropics is largely linked to year-to-year changes in biomass burning emissions [Duncan et al., 2003b]. Nevertheless, the decline of European fossil fuel emissions is also reflected in the tropical northern Pacific sites (e.g., Guam, Mauna Loa) which show a downward trend in the 1990s in both model and observations [Duncan et al., 2007]. A large broad peak is seen in 1998–1999 at the tropical sites (e.g., Tutor Hill, Mauna Loa, Guam, Ascension). An exception is found for the elevated site of Mauna Loa that shows two distinctive peaks, in 1998 and a second smaller one, in 1999. Figure 2 shows that the 1998 peak in CO is driven by both meteorology and biomass burning emissions while the peak in 1999 is entirely due to biomass burning emissions. This large anomaly in 1998–1999 is related with the strong El Niño event that induced large fires in Indonesia in August to November 1997 [Duncan et al., 2003a]. These fires resulted in large enhancement of CO levels over this region and also throughout the Indian Ocean. In the Seychelles a large peak is seen in fall 1997 (Figure 1 and Duncan et al. [2003a]). The model captures very well the anomalies in the Antarctic station of Syowa. Those are largely driven by the variability in biomass burning emissions and in transport (not shown) which gives confidence in the representation of these processes in the model.

3.2. O3

[17] The performance of the model in terms of O3 IAV is evaluated using vertical profiles from the World Ozone and Ultraviolet Radiation Data Center (WOUDC) and ground-based measurements from the World Data Center for Greenhouse Gases (WDCGG) and the EMEP programs. We show a comparison between simulated and observed O3 anomalies at six surface O3 stations and six ozonesonde sites at 500 hPa (with long enough records) that cover different regions including the Pacific Ocean (Mauna Loa), North America (Barrow, Alert, Churchill), Europe (Mace Head, Jungfraujoch, Rigi, Harwell), and Japan (Kagoshima, Tateno) (Figures 3 and 4, and Figure S3 in the auxiliary material for additional sites). As for CO, a short comparison between observed and simulated O3 concentrations (rather than anomalies) is provided in the auxiliary material (Figure S4 and text therein).

Figure 3.

Time series of observed (black) and simulated (red) O3 anomalies (in ppb). Solid fine line shows monthly mean; circles denote 12-month running averages. Blue circles correspond to the 12-month running average of the simulated stratospheric O3 anomalies. Observations are from the WOUDC ozonesondes at 500 hPa. At some locations, as only few profiles are available each month, the model was sampled at the exact time and date of the observations.

Figure 4.

Same as Figure 3 but for selected WDCGG and EMEP surface stations.

[18] The agreement between the observed and simulated O3 anomalies is variable over the 19-year period and depends on the region considered. The correlation coefficients (r2) range between 0.3 and 0.47 at the surface sites (Figure 4), but it drops dramatically for the sites at high latitudes and at high altitudes as discussed in the following. A relatively good agreement (r2 = 0.45) is found for the pacific station of Mauna Loa (Hawaii). Most of the largest negative (e.g., 1989, 1992–1993, 1996–1997) as well as positive (e.g., 1998–1999, 2003–2004) anomalies are reproduced by the model although, in some cases, the model exhibits larger anomalies than in the observations, such as in 1998. On the other hand, the observed 1999 peak is not reproduced by the model.

[19] At higher latitudes in the Northern Hemisphere (e.g., Barrow, Alert and Churchill), and especially in the middle and upper troposphere, the model fails to capture the decline observed from 1991 to 1993 (up to a negative anomaly of 10 ppbv in the middle troposphere) as well as the following upward trend in 1993–1996. A similar behavior is also seen at the Jungfraujoch and Sonnblick mountain sites and also to some extent over Hohenpeissenberg at 500 hPa (Figure S3 in the auxiliary material). Recent studies have linked these changes to the coupling between the lowermost stratosphere and the troposphere, and in particular to changes in O3 concentrations in the lowermost stratosphere [Oltmans et al., 1998; Tarasick et al., 2005; Thouret et al., 2006; Ordóñez et al., 2007]. The ozonesonde observations at the above mentioned sites show larger anomalies at higher altitudes (not shown), which implies a plausible stratospheric influence. Oltmans et al. [1998] argued that the decline in the early 1990s is possibly linked to reduced lower stratospheric O3 concentrations following the eruption of Mt. Pinatubo. Observations show that, in early 1993, after the eruption of Mt. Pinatubo, the total O3 column reached record low values in part due to the effects of volcanic aerosols [Herman and Larko, 1994; Gleason et al., 1993; Randel et al., 1995]. Global total O3 concentrations averaged between 60°S to 60°N showed an almost immediate decrease which peaked in mid-1992 and began to slowly recover over the next 3 years [Chartrand et al., 2000].

[20] Observed O3, however, continued to increase after 1995 at several sites (e.g., Alert, Barrow, Jungfraujoch as shown in Figures 3 and 4), even though stratospheric aerosol loading had dropped to normal levels [World Meteorological Organization, 2003], suggesting that additional processes may be involved. Tarasick et al. [2005] reported that these positive anomalies after 1993, which were also found at several sites in Canada, may be partially a result of small changes in the atmospheric circulation, notably wave-driven dynamics in the stratosphere, which resulted in higher stratospheric O3 amounts. Ordóñez et al. [2007] also observed an upward trend in the 1990s at four European sites (including the mountain sites of Jungfraujoch and Zugspitze, and the sounding sites of Payerne and Hohenpeissenberg). They argued that it is likely associated with enhanced stratospheric O3 contributions, driven by dynamics rather than by changes in stratospheric chlorine loading. As mentioned in section 2.1, O3 production in the stratosphere remains constant throughout our study period, therefore we do not expect the model to be able to reproduce the decline in stratospheric O3 associated with the Mt. Pinatubo event. In addition, our model does not seem to capture the changes in dynamics in the stratosphere that have been proposed as a possible explanation for the upward trend in the mid-1990s. This could be possibly due to a too low vertical resolution and a too low model top. Note that an O3 peak (with a strong stratospheric signature) is also seen in 2001 in our simulations over most of the midlatitude and high-latitude sites. This peak is not present in the observations presented here (Figures 3 and 4) and is restricted to the lower stratosphere in the MOZAIC observations (not shown) (Marenco et al. [1998] and

[21] From 1995 onward, the model reproduces better the observed IAV in O3 than during the earlier period over the European stations. In particular, the 1998–1999 anomaly that is seen in O3 measurements at several sites (but not all of them) is relatively well reproduced by the model, although the magnitude is somewhat underestimated in some locations. The 1998–1999 anomaly is apparent as a broad peak from late 1998 to early 1999 in the observations over the low-altitude sites of Mace Head (Figure 4) and Ekdalemuir (Figure S3 in the auxiliary material) and is well reproduced by the model. At the elevated European sites (Jungfraujoch, Figure 4 and Sonnblick, Figure S3 in the auxiliary material) there is an early 1998 peak in the observations that is captured by the model although the observed 1999 peak is somewhat missing in the model. The 1998–1999 anomaly is also seen in the observations collected at the Swiss ground-level stations of Rigi (Figure 4) and Payerne (Figure S3 in the auxiliary material) but it is not too well reproduced by the model. On the contrary, the simulated O3 anomalies at 500 hPa altitude over Uccle and Hohenpeissenberg shows a small peak in 1998 which is not seen in the observations. The 1998–1999 O3 anomaly has also been found in the MOZAIC observations, that show a significant O3 increase in the upper troposphere over eastern United States, Europe, and Iceland [Thouret et al., 2006] and also further down in the troposphere [Zbinden et al., 2006]. This anomaly is discussed in details in the following sections.

[22] The O3 anomaly seen in summer 2003 over Europe (e.g., Jungfraujoch, Rigi and Harwell in Figure 4) is related to the anomalous high temperatures [Tressol et al., 2008; Cristofanelli et al., 2007] that began in Europe in June 2003 and continued through July until mid-August, inducing summer temperatures 5 K higher than the seasonal average over a large portion of the continent [United Nations Environment Programme, 2004]. GEOS-Chem captures well the observed anomaly, as also demonstrated by Guerova and Jones [2007]. However, we find that it systematically underestimates the double peak in O3 during the heat wave in June and August 2003. One reason for this bias might be the coarse grid of the model used in the simulations. The strong influence of meteorology (e.g., temperature) to the anomaly is seen at Jungfraujoch (Figure 5).

Figure 5.

Anomalies in O3 in the standard simulation (black) at Barrow, Mace Head, Harwell, Jungfraujoch, Rigi, and Mauna Loa. The green line shows the difference between the O3 concentrations in the standard simulation (S0) and those from the simulation with climatological biomass burning emissions (S1). The blue line shows the difference between the O3 concentrations in the standard simulation (S0) and those from the simulation with fixed meteorology (S2). The circles represent the 12-month running averages.

[23] The O3 IAV over the Japanese stations is fairly well reproduced. Over Tateno and Kagoshima (at 500 hPa altitude) both model and observations show a sharp increase in spring 1998. The observations show a second peak in 1999 that is followed by a large decrease. The model captures better the 1999 peak at the ground station of Ryori (Figure S3 in the auxiliary material) while underestimating it above the two other sites. The O3 decrease that follows is well captured at all stations.

4. Processes Influencing the 1998–1999 O3 Anomaly Over Europe

[24] In this section, we focus on the large anomaly in tropospheric O3 anomalies in Europe, from January 1998 to April 1999. Figure 6 shows monthly Tropospheric Ozone Column (TOC) anomalies averaged over a region representative of continental Europe (35°N–75°N, 10°W–30°E, see Figure 7). Monthly TOC are calculated from the surface up to the climatological annual mean tropopause, as diagnosed by the WMO temperature gradient [World Meteorological Organization, 2003] in the GEOS-4 meteorological data. An unprecedented enhancement in O3 is seen throughout the entire column in 1998 (with a maximum enhancement of 4.9 DU or 14% above the 19-year February mean in February 1998) and also at the surface (maximum enhancement of 4.9 ppbv or 11% above the 19-year February mean in February 1998). The anomaly extends up to 1999 (see discussion in section 3.2 for the anomaly in 2001) and persists throughout 1999 in some specific regions of Europe (e.g., in the southern part of Europe and over the western Atlantic Ocean (0°–30°W, 35°N–45°N); Figure S5 in the auxiliary material). Thouret et al. [2006] suggested that this anomaly may be related to the strong El Niño of 1997–1998. Simmonds et al. [2004] reported enhancements in 1998–1999 in background concentrations of O3 and CO but also additional compounds such as CO2, H2, CH4 and CH3Cl over Mace Head, suggesting that the most likely explanation of the anomaly is large-scale biomass burning events in tropical and boreal regions during 1997–1999 that were associated with the 1997 El Niño event. We should mention that in general the model reproduces well the observed correlations between TOC and the Southern Oscillation Index (SOI) in the tropics for our 19-year simulation (see Figure S6 in the auxiliary material), which provides some confidence in the model's abilities to reproduce the processes associated with ENSO.

Figure 6.

Simulated TOC anomalies (in DU) for the standard simulation (S0) averaged (top) over eastern United States (35°N–50°N,60°W–90°W) and (bottom) over Europe (35°N–75°N, 10°E–30°W).

Figure 7.

Map of regions and slices used in the manuscript. See text for details.

[25] Sensitivity simulations indicate that year-to-year changes in biomass burning and meteorology most largely contribute to the anomalies in TOC and Tropospheric CO Column (TCOC) (Figure 8). Figure 9 shows the time series of the tagged O3 tracers that most largely contribute to the O3 burden over Europe, that is, “North American,” “Asian,” “European” and stratospheric tracers as previously reported by Auvray and Bey [2005]. The stratospheric, “Asian,” and “North American” tracers show positive anomaly in 1998–1999 (Figure 9) indicating that several processes contribute conjointly to the 1998–1999 anomaly. A detailed analysis shows that three different periods can be distinguished during the 1998–1999 anomaly, including the first half of 1998, the second half of 1998, and the beginning of 1999, as discussed in the following.

Figure 8.

(top) Solid line shows TOC anomalies (relative to the 1987–2005 period) in the standard simulation (S0); dotted line shows TOC difference between S0 and S1 (climatological biomass burning); dashed line shows TOC difference between S0 and S2 (fixed meteorology); TOC are averaged over Europe (in DU). (bottom) Same but for tropospheric CO column. For convenience, similar units (DU) are used for O3 and CO columns.

Figure 9.

(top) Monthly time series of TOC originating from North America (red line), Atlantic Ocean (red dotted line), Asia (blue line), Europe (green line), and the stratosphere (black line) for the period 1997–1999 and averaged over Europe (35°N–75°N, 10°E–30°W) from the surface to the tropopause. (bottom) Same but for the percent contribution of each tracer to the TOC.

4.1. Winter and Spring 1998

[26] The anomaly in TOC starts to rise in late 1997–early 1998 to reach almost 5 DU in February. On average, the increase from January to April is 3.6 DU and the anomaly is found to be driven by meteorology (Figure 8, top). During this period and especially in January–February 1998, the contribution of stratospheric O3 is enhanced and amounts to 34% of the total O3 (Figure 9), that is +1.9 DU on average and corresponds to a 16% increase above the 19-year January–April average value. We suggest that this increase in the stratospheric O3 results from the 6-month lagged relationship between ENSO and STE in the north midlatitudes that has been reported previously by Langford et al. [1998] and Zeng and Pyle [2005]. The primary influence of ENSO is to change the location and amplitude of convective activity via changes in the Walker circulations over the Pacific [James et al., 2003]. Furthermore, one of the most robust signals of ENSO is its modulation of the intensity and position of the subtropical jet stream over the Eastern North Pacific [Shapiro et al., 2001], which in turn, influences STE [Zeng and Pyle, 2005; Baray et al., 2003; Langford et al., 1998].

[27] The second most important tagged tracer during the same period (January–April 1998) is O3 originating from Asia which accounts for ∼14% of the total O3 (Figures 9 and 10). “Asian” O3 contributes to +0.6 DU to the anomaly and mainly originates from Southeast Asia (+0.4 DU, not shown). Positive anomalies are seen for the “North American” (+0.3 DU, 13% of the TOC, Figures 9 and 10) and “Atlantic Ocean” (+0.4 DU, 9% of the TOC) tracers, as well. The discussion in section 5 presents evidence that the positive anomaly in the “Asian” contribution is related to ENSO that enhances export from Asia. The discussion in section 6 describes the mechanisms that lead to the enhanced contribution of the “North American” tracer. In contrast, the “European” tracer shows only little anomalies during the same period (−0.1 DU, 8% of the TOC, on average over January–April 1998), except for April 1998 (+0.3 DU, 13% of the TOC).

Figure 10.

Anomalies in O3 tagged tracer originating from (a) Asia, (b) North America, and (c) the stratosphere averaged over the tropospheric column in March 1998.

[28] In contrast with O3, the TCOC anomalies are rather low and even negative during the same period (Figure 8, bottom). The processes responsible for those low anomalies are discussed in section 5.

4.2. Summer and Early Fall 1998

[29] During the second period (summer–early fall of 1998), we find positive anomalies in both O3 and CO columns. TOC increases by +1.7 DU and TCOC by +2.2 DU on average during July–October 1998. During this period, “Asian,” “North American” and “European” tracers contribute most (∼20% each) to the TOC over Europe (Figure 9, bottom). Highest positive anomalies are found for the “Asian” (+0.6 DU mainly coming from Siberia) and “North American” (+0.4 DU mainly coming from Canada) tracers and most largely affect northern Europe. Furthermore, a comparison between the results from the control tag simulation (T0) and that from the tag simulation with fixed biomass burning emissions (T1) indicates that biomass burning emissions are responsible for approximately half of the anomalies of the Siberian and Canadian tracers (+0.3 DU). Intense boreal fires were reported during that period by Leung et al. [2007] and Yurganov et al. [2004]. The hot and dry conditions in the boreal regions in northern and central Canada and eastern Siberia in 1998 that induced these large fires are possibly linked to the strong El Niño event of 1997/1998 [Novelli et al., 2003; Duncan et al., 2003b; Spichtinger et al., 2004]. Spichtinger et al. [2004] argued that El Niño conditions have also enhanced the transport from Siberia to Canada in 1998. Forster et al. [2001] have attributed part of the enhanced CO concentrations at Mace Head in August 1998 to transport from Canadian and Russian forest fires. This is confirmed by our model results that show large anomalies in surface CO at Mace Head in September and October 1998 as a result of both meteorology and biomass burning emissions (Figure 2). Further discussion of the Mace Head anomaly in surface O3 and CO concentrations is provided by Figure S7 in the auxiliary material.

4.3. Spring 1999

[30] The third period of the 1998–1999 anomaly corresponds to spring 1999. This anomaly is especially seen for CO in April 1999 (Figure 8), and is mostly due to enhanced biomass burning emissions. The anomaly is not as clearly seen in O3. Only a small positive anomaly is found in TOC (+1.0 DU) which is limited to the southern part of Europe (Figure S5 in the auxiliary material). O3 that originates from Southeast Asia/India (+0.5 DU) and China (+0.5 DU) is found to be responsible for this increase. Our tag simulations indicate that biomass burning emissions are responsible for these anomalies, contributing to around 0.7 DU to the total increase (+1.0 DU). Intense biomass burning emissions were documented in India and Southeast Asia in spring 1999 [Staudt et al., 2001; Duncan et al., 2003b] and China [Dong, 1999].

5. Influence of ENSO on the Transport of Asian Pollution Toward Europe

[31] In the following, we investigate the possible relation between ENSO and transport of pollution toward Europe. We focus on the spring season, when the Asian pollution outflow is maximum [Jacob et al., 2003], and in the free troposphere where trans-Pacific transport usually occurs to a greater extent [Liang et al., 2005]. Both transport ahead of cold front and deep convection over Southeast Asia contribute to the export of pollution from Asia and to its eastward transport to the northern Pacific Ocean [Bey et al., 2001b; Liu et al., 2003]. We investigate the Asian pollution outflow using CO which is good proxy for long-range transport of pollution [Bey et al., 2001b; Liu et al., 2003]. Figure 11 shows anomalies in CO concentrations taken at the cross section located at 150°E longitude from 0° to 60°N latitude for March 1992 and 1998 (El Niño) and March 1999, 2000, and 2001 (La Niña). CO anomalies are largely positive during El Niño conditions, reflecting an enhanced Asian outflow. In particular, in March 1998, the suppression of convection in Southeast Asia induced by El Niño results in a larger northeastward flux in the planetary boundary layer toward the convergence region located near 25°–30°N where lifting into the lower free troposphere subsequently takes place as reported by Liu et al. [2003]. As a result, pollution from Southeast Asia adds more significantly to the pollution export from Asia, resulting in a significantly higher export in the lower free troposphere during El Niño conditions [Liu et al., 2003].

Figure 11.

Vertical CO anomalies at a slice located at 150°E and between 0°N and 60°N (see Figure 7) for El Niño years (1992 and 1998) and La Niña years (1999, 2000, and 2001) for (left) the control run simulation and (right) the simulation with climatological biomass burning emissions.

[32] Positive anomalies in CO concentrations are also seen for March 1999 (La Niña) over the Pacific Rim (Figure 11). The large reduction in CO seen in the simulation with fixed biomass burning emissions indicates that the enhanced outflow in 1999 is largely due to enhanced biomass burning emissions that occurred in India, Southeast Asia and China [Dong, 1999; Staudt et al., 2001; Duncan et al., 2003b] rather than changes in export pathways. Impact of biomass burning emissions is also seen in March 1998 but to a much smaller extent.

[33] Several studies have reported an eastward extension of the subtropical jet stream during El Niño winters [e.g., Chen and Van den Dool, 1999; Bell et al., 1999; Kitoh et al., 1996]. In particular, the amplified subtropical jet stream in 1998 affected both the upper and the middle troposphere and extended across northern Mexico and southern United States [Bell et al., 1999]. The GEOS-4 reanalyzed winds that are used to drive the model show similar results. Figure 12 shows the zonal wind pattern at 200 hPa associated with the subtropical jet averaged over La Niña and El Niño years. The subtropical jet extends eastward during the El Niño years. We suggest that the extension of the subtropical jet stream results in an enhanced transport of Asian pollution toward the North Atlantic area, as illustrated in Figure 10. In addition, the model indicates enhanced O3 production rates into the free troposphere where the pollution transport takes place (3–9 km) in March 1998 (Figure S8 in the auxiliary material).

Figure 12.

Zonal wind speed (m s−1) at 200 hPa averaged for the March months of (top) La Niña years (1999, 2000, and 2001) and (bottom) El Niño years (1992, 1995, and 1998).

[34] As a result, the O3 mass fluxes over those regions showed significantly larger values than the 19-year mean, especially in March 1998. Those enhanced fluxes are found around the subtropical jet and extend eastward to North America (Figures 13 and 14).

Figure 13.

Total O3 mass fluxes (kg m−2 month −1) averaged over the tropospheric column (arrows) and averaged over the month of March of (a) La Niña years (1999, 2000, and 2001) and (b) El Niño years (1992, 1995, and 1998). The colors denote the magnitude of the fluxes.

Figure 14.

Total O3 mass fluxes (kg m−2 month −1) averaged over the tropospheric column in March 1998 (arrows). The colors denote the anomalies in the amplitude of the fluxes.

[35] We next examine results of the model in the eastern Pacific Ocean after the pollution is exported from Asia and before it enters the North American continent. Figure 15 depicts the anomalies in O3, NOx, PAN, and “Asian” O3 tropospheric columns and in O3 production rates from January 1997 to December 2000 obtained from the control simulation and from the simulation with fixed biomass burning emissions. As expected, high positive anomalies are found in 1998–1999 mostly owing to Asian pollution outflow, as already discussed. There are three distinctive peaks in the “Asian” O3 over this region, in April–May 1998, August 1998, and March–April 1999 (Figure 15d). As mentioned before, the influence of biomass burning emissions is most clearly seen in spring 1999. The PAN anomalies are also large in 1998–1999 (Figure 15c) and show the same three positive peaks, suggesting that they are also due to export from Asia. It is well known that PAN provides an important storage reservoir for NOx species. High NOx anomalies are also simulated by our model (Figure 15b) during almost all the 1998 year. While a fraction of the NOx may be directly transported from Asia, it is also likely that a fraction results from thermal decomposition of PAN. We suggest that these enhanced NOx levels induced higher than average O3 production during spring and summer of 1998 (Figures 15e). Hudman et al. [2004], for example, argued that PAN decomposition represents a major and possibly dominant component of the O3 enhancement in trans-Pacific Asian pollution plumes. Using a photochemical box model constrained by measurements from the Photochemical Ozone Budget of the Eastern North Pacific Atmosphere (PHOBEA) experiment, Kotchenruther et al. [2001] found that PAN decomposition increases the production of O3 from 11 to 30% in the springtime northeastern Pacific troposphere. Note that the impact of ENSO on the lightning NOx source is found to be small outside the tropics in the model, and therefore has only a minor role in the increase in NOx.

Figure 15.

Monthly anomalies (solid line) for (a) TOC (DU), (b) PAN (TgN), (c) NOx (TgN), (d) TOC of O3 originating from Asia, and (e) Ox production rates (Tg O3 month−1) averaged over a slice in the Pacific Ocean (30°N–50°N, 150°W; see Figure 7) from January 1997 to December 1999. The dotted line shows results from the S1 simulation (with fixed biomass burning emissions).

[36] As mentioned in section 3.2, the model reproduces well the increased CO and O3 concentrations that are observed in 1998–1999 at Mauna Loa, a station that samples the remote marine environment but is also under the influence of the long-range transported pollution from Asia. In addition, observations at the Cheeka Peak Observatory, in the western United States (124.6°W, 48.3°N, 480m asl), as part of the PHOBEA project [Jaffe et al., 2001], showed elevated concentrations of CO in the spring of 1998 relative to spring of 1997, consistent with enhanced biomass burning and enhanced transport of industrial emissions originating from Asia.

[37] Finally, as mentioned in section 4, the TCOC anomalies are rather low and even negative during the same period (Figure 8, bottom) in contrast with the anomalies in O3, especially those in “Asian” O3. The difference in CO between the standard simulation and the simulation with fixed biomass burning is positive, which confirms the influence of the biomass burning from Asia on CO over Europe (Figure 8, bottom). In the mean time, the difference in CO between the standard simulation and the simulation with fixed meteorology is negative while one would expect to see an increase in CO concomitantly to the increase in the tagged “Asian” O3. In 1998, the model simulates enhanced OH concentrations that are mainly driven by enhanced water vapor concentrations (I. Bey et al., Interannual variability in hydroxyl radicals over the last decades: Processes involved and model intercomparison, manuscript in preparation, 2008). We suggest that these enhanced OH concentrations strongly limit the CO lifetime in 1998 (see the enhanced CO chemical loss in 1998 on Figure S9 in the auxiliary material), and therefore its transport toward Europe. Even though El Niño induced quite particular conditions in 1998, this suggests that caution may be needed when using CO as a proxy to examine long-range transport of pollution.

6. Trans-Atlantic Transport

[38] In section 5, we suggest that the O3 outflow from Asia in spring 1998 extends eastward following the amplified subtropical jet stream over the northern Mexico and the Southern United States. Then the O3 flow turns anticyclonically over the North Atlantic and enters the European continent from the northwest (Figure 13). A recent study by Stohl et al. [2007] has demonstrated the possibility of long-range transport of air pollution plume from Asia across the North Pacific, North America (through southern United States) and up to Europe.

[39] Figure 16 shows the CO anomalies at a cross section from 20° to 70°N latitude and at 15°W longitude (see Figure 7) for March 1992 and 1998 (El Niño) and 1999, 2000, and 2001 (La Niña) in the standard simulation and in the simulation with climatological biomass burning emissions. Higher CO concentrations are found during El Niño conditions (e.g., in 1992 and especially in 1998) in the upper troposphere (above 8 km), which suggests that they could be directly transported from Asia. Similarly to what we find over the Pacific Ocean, positive CO anomalies in 1999 are found to be mostly due to enhanced biomass burning emissions in Southeast Asia/India and China. In 2001 (during La Niña conditions), we also note a positive anomaly originating from North America at 40°N–50°N in the middle and lower troposphere. Similar positive anomalies are also seen in the lower troposphere in several non-ENSO years (e.g., 1988, 1989, 1990, 1991, not shown). These anomalies are likely related to different processes involved in trans-Atlantic transport occurring at lower levels [Li et al., 2002; Stohl and Trickl, 2001; Guerova et al., 2006]. Li et al. [2002] found for example that low-level trans-Atlantic transport of North America to Europe correlates with the North Atlantic Oscillation.

Figure 16.

Vertical CO anomalies at a slice located at 15°W and between 20°N and 70°N (see Figure 7) for El Niño years (1992 and 1998) and La Niña years (1999, 2000, and 2001) for (left) the control run simulation and (right) the simulation with climatological biomass burning emissions.

[40] Finally, in section 4, we also mentioned that a positive anomaly in the “North American” O3 contribution is simulated over Europe in spring 1998. The possibility of O3 transport from North America to Europe has been demonstrated by several studies [e.g., Li et al., 2002; Stohl and Trickl, 1999], and export fluxes from North America are strongest in spring and summer [Li et al., 2002]. We suggest that the enhanced export from North America resulted from an enhanced photochemical activity in the eastern United States during spring 1998. O3 concentrations in the eastern United States are found to have increased throughout the period 1998 extending through 1999 (Figure 6, top) and this appears to be related mainly to an increase in O3 produced locally and to some extent to an increase in O3 transported from Asia (Figure 10). Large positive anomalies in PAN, NOx and net O3 production are seen over the eastern U.S. region (Figure 17). The model indicates an increase in temperature that is more pronounced near the surface (e.g., +3.7 K maximum anomaly in the surface of the eastern United States in February 1998) but extends to higher altitudes (Figure 17e). The photochemical activity was most probably favored by the enhanced temperatures during spring 1998. It has been suggested by several studies that the year 1998 was one of the warmest years of the century which was probably linked with the 1997 El Niño event [Mann et al., 1999; Hansen et al., 1999], although the warmth of 1998 was too large to be fully accounted for by the El Niño [Hansen et al., 1999].

Figure 17.

Monthly time series of vertical profiles of anomalies in (a) O3 (ppb), (b) PAN (ppb), (c) NOx (ppb), (d) Ox production rates (Tg/month), (e) temperature (K), and (f) O3 originated from Asia (ppb) averaged over the eastern United States (see Figure 7) from 1994 to 2001.

7. Generalization of the ENSO Influence on European O3

[41] In this section, we examine the influence of El Niño on European tropospheric O3 column throughout the period from 1987 to 2005. We also examine the effect of El Niño on the European surface O3 levels. We use the Southern Oscillation Index (SOI) to measure the ENSO activity and we examine the correlation between the SOI averaged over the fall months (September–October–November) and the European O3 (tropospheric column or surface) averaged over the early springtime season (February–March–April) of the following year (Figure 18, top and middle, respectively). This 6-month time lag corresponds to the response time of STE and Asian export to ENSO events. We find high anticorrelation of −0.65 (n = 18, significant at the 0.05 level). The anticorrelation remains significant even if the strong 1997 El Niño year is not accounted for (r = −0.58, n = 17) or even if both the strong 1997 El Niño and 1988 La Niña are not accounted for (r = −0.73, n = 16). Figure 19 shows the scatter plot of SOI versus TOC and demonstrates that increasing TOC can be seen over Europe the following spring of an El Niño year. The TOC averaged over the years with an absolute value of SOI above 1 (i.e., the years that are characterized by a clear El Niño or La Niña signal) amounts to 38.1 and 35.2 DU for El Niño and La Niña years, respectively. The difference between the two values is larger than the standard deviation (1.6 DU) of the spring TOC computed over the 19-year period, which indicates that the El Niño induces a variability in the European TOC that is larger than the mean natural variability over the 19-year period. On the basis of our analysis of the 1998–1999 episode, we suggest that the correlation results from a combination of several El Niño–induced effects, i.e., increase in stratospheric input of O3, increase in pollution transport from Asia, and enhanced biomass burning events. The influence of El Niño is however not restricted to the spring season, as El Niño can trigger biomass burning events that also affect European O3 in summer and fall for some years.

Figure 18.

Interannual variation in (top) SOI averaged over the SON months, (middle) TOC (in DU) over Europe, and (bottom) surface O3 (in ppbv) over Europe, averaged over the FMA months of the year that follows. The anticorrelations between SOI and TOC (r = −0.65) and SOI and surface O3 (r = −0.50) are significant at the 0.05 level (n = 18).

Figure 19.

Scatterplot of the early spring TOC (in DU) over Europe versus the fall SOI. The years indicated close to the symbols correspond to the year of the El Niño/La Niña event. Squares show El Niño year; triangles show La Niña year; grey circles show neutral year. Note that an ENSO year is defined by its SOI in the September–October–November months. For example, the year 1998 is defined as a La Niña year here, whereas in Figures 11, 12, and 16 the 1998 year is indicated as an El Niño year in order to illustrate the influence of the 1997 El Niño event on the 1998 spring.

[42] We also find high anticorrelation between SOI (September–October–November) and the surface O3 (February–March–April of the following year) over Europe (Figure 18, bottom). The anticorrelation is lower than the one calculated for TOC, but still significant (−0.50, n = 18; see also Figure S10 in the auxiliary material). Surface O3 over Europe amounts to 45.2 ppb for the El Niño years and 42.1 ppb for the La Niña years. As previously the difference between the two values is larger than the standard deviation (2.0 ppb) of the spring surface O3 calculated over the 19-year period.

8. Conclusion

[43] We used the GEOS-Chem model to interpret long-term measurements of tropospheric O3 and CO and to investigate the factors that contribute to their interannual variation (IAV) during the period from 1987 to 2005 with a focus on the northern midlatitudes. We find that the model reproduces relatively well the observed IAV of CO but shows some problems in reproducing the IAV of O3. The clear decreasing CO trend in the Northern Hemisphere in the 1990s as well as many of the large CO anomalies due to biomass burning emissions are fairly well reproduced by the model. The model however misses the negative anomaly in tropospheric O3 in 1991–1993 as well as the following upward trend in 1993–1996 seen over several sites in the northern midlatitudes, most probably because of a misrepresentation of the year-to-year changes in stratospheric O3 concentrations and dynamics.

[44] We examined in detail the large 1998–1999 anomaly using a variety of sensitivity and tag simulations. We find that changes in both meteorology and biomass burning emissions in tropical and boreal regions played a role. Three different periods can be distinguished during this long-lasting anomaly, during which different processes affected the O3 burden over Europe.

[45] During the first half of 1998, the influence of meteorology is associated with the preceding 1997 El Niño, which was one of the most extreme El Niño events of the century. El Niño is found to affect significantly the stratosphere-troposphere exchange and the Asian pollution export and transport toward Europe.

[46] The effect of El Niño on stratospheric input is mainly seen during the first 6 months of 1998 and reaches a maximum in January and February. Note that, although the model fails to reproduce the changes in STE over the 1990s, we argue that it reproduces relatively well the enhanced STE induced by ENSO with a 6-month time lag as previously reported by Langford et al. [1998] and Zeng and Pyle [2005]. This is probably because the ENSO forcing on STE originates from the troposphere [Bronnimann et al., 2006], while the processes leading to the variation in O3 in the 1990s appear to be initiated in the stratosphere [Tarasick et al., 2005; Ordóñez et al., 2007].

[47] In spring 1998, we find enhanced contribution in O3 originating from Asia throughout the northern midlatitudes. We suggest that this is related to an enhanced pollution export from Asia due to change in convection activity in East Asia that favors the export of Asian pollution and an extension of the subtropical jet stream during El Niño conditions that favors the transport of pollution further east toward Europe. A significant amount of Asian pollution is thus transported directly to Europe and contributes to the O3 anomaly in spring 1998. In addition, we also find an increased export from North America during the same period. We suggest that this is related to a significative O3 anomaly over the eastern United States. This, in turn, results from enhanced photochemical activity in spring 1998 associated with larger than normal temperature over the eastern United States.

[48] During summer and fall 1998, boreal forest regions such as Canada and Siberia experienced large fires that influenced O3 and CO concentrations over Europe, and also contributed to the anomaly in O3 [Simmonds et al., 2004; Spichtinger et al., 2004]. These fires may be related to ENSO as suggested by previous studies [van der Werf et al., 2004]. Finally, the increased European O3 and CO concentrations in spring 1999 are found to be linked with long-range transported pollution from particularly large biomass burning emissions that occurred in China, Southeast Asia and India in spring 1999.

[49] We also examined the influence of ENSO over Europe throughout the 19-year period. We correlate the fall average (September–October–November) SOI with the early-spring average (February–March–April) TOC over Europe considering a 6-month time lag and find a significant correlation of −0.65. The 6-month time lag is related to both the response time of changes in stratospheric O3 influx [Langford et al., 1998; Zeng and Pyle, 2005] and the maximal export of Asian pollution after the low ENSO Index in September–October–November. The impact of ENSO is important for O3 at the surface, as well. Significant anticorrelation is found between SOI (September–October–November) and the surface O3 (February–March–April) over Europe (−0.50).

[50] Additional steps should be taken to refine our quantitative understanding of the processes that affect the temporal variation of O3 and CO. In particular, our study highlights again the strong need to use fully coupled stratosphere-troposphere models especially if one wants to reproduce the O3 IAV in the midlatitudes of the Northern Hemisphere. Although long-term ground-based point measurements provide critical information for trends assessments, it is difficult to analyze O3 IAV from single-point measurements because of the influence of local phenomena. Long-term and consistent measurements from instruments that can provide information for larger areas (such as satellites or programs like MOZAIC) are essential to identify large-scale anomalies in O3 and CO and quantitatively assess the model representation of those variabilities.


[51] This work was supported by funding from the Swiss National Science Foundation under grants 2100-067979 and 200020-112231. The GEOS-Chem model is managed by the Atmospheric Chemistry Modeling group at Harvard University with support of the NASA Atmospheric Chemistry Modeling and Analysis Program. The authors wish to thank Johannes Stähelin and Alexis Berne for useful conversations. In addition, we are grateful to the World Ozone and Ultraviolet Radiation Data Center, the World Data Center for Greenhouse Gases, the European Monitoring and Evaluation Program, and the National Oceanic and Atmospheric Administration/Earth System Research Laboratory for providing the different set of data used in that study. We also wish to thank the reviewers who all contributed to improve the paper.