European summer surface ozone 1990 – 2100

Abstract. The impact of climate change and changes in ozone precursor emission on summer surface ozone in Europe was studied using a regional CTM over the period 1990 to 2100. Two different climate simulations under the SRES A1B scenario together with ozone precursor emission changes from the RCP4.5 scenario were used as model input. In southern Europe regional climate change leads to increasing surface ozone concentrations during April–September, but projected emission reductions in Europe have a stronger effect, resulting in net reductions of surface ozone concentrations. In northern Europe regional climate change decreases surface O 3 and reduced European emissions acts to further strengthen this trend also when including increasing hemispheric background concentrations. The European O 3 precursor emission reductions in RCP4.5 are substantial and it remains to be seen if these reductions can be achieved. There is substantial decadal variability in the simulations forced by climate variability which is important to consider when looking at changes in surface O 3 concentrations, especially until the first half of the 21st century. In order to account for changes in background O 3 future regional model studies should couple global (hemispheric) and regional CTMs forced by a consistent set of meteorological and precursor emission data.


Introduction
Surface ozone (O 3 ) has negative effects on human health and vegetation.Background levels of O 3 in the troposphere over the northern mid-latitudes have increased by a factor of two to three since the end of the 19th century (Parrish et al., 2009).This is to a large extent related to increasing man made emissions of O 3 precursors such as methane (CH 4 ), carbon monoxide (CO), nitrogen oxides (NO x ) and non-methane volatile organic compounds (NMVOC) (Isaksen et al., 2009).O 3 in the lower troposphere is also regulated by climatic factors such as temperature and solar radiation as well as by surface deposition which in the case of vegetated surfaces is again regulated by temperature, solar radiation, humidity and soil water availability (Jacob and Winner, 2009;Andersson and Engardt, 2010).Given that projections of future climate in Europe indicate substantial changes in these variables there is concern about increasing concentrations of O 3 due to climate change (Royal Society, 2008).
Several studies with regional chemistry transport models have pointed out the importance of climate change for future surface O 3 in Europe (e.g.Langner et al., 2005;Meleux et al., 2007;Engardt et al., 2009;Andersson and Engardt, 2010;Katragkou et al., 2011).The previous studies based their analysis on simulations over time slices of 10-30 yr for current and future climate.In most cases climate projections based on one global climate model was used.The purpose of the present study is to extend the analysis of future changes in surface O 3 in Europe to include the combined transient effects of changes in climate, changes in O 3 precursor emissions and changes in background concentrations over the 110-yr period from 1990 to 2100.We use recent compilations of historic and future trends of emissions of air pollutants over Europe and climate projections from two different global climate models downscaled over Europe to show regional detail.
Section 2 provides details on the setup of the current study including general information on the utilized climate and chemistry transport models.Section 2 also contains an evaluation of the modelling systems' ability to reproduce surface O 3 concentrations in Europe.In Sect. 3 we present the results.We focus on summer (April-September) mean and mean of daily maximum (daily max) concentrations of O 3 .Summer is the main growing season and also when surface Published by Copernicus Publications on behalf of the European Geosciences Union.
O 3 is at its maximum in Europe and hence this period is of main interest for impacts on human health and vegetation.Finally we present conclusions in Sect. 4.

Methods and data
To simulate the evolution of surface ozone over Europe during the period 1990-2100 we make use of a chain of models starting with global climate models (GCMs) that feeds boundary conditions into a regional climate model (RCM) with higher spatial resolution over Europe.The resulting meteorological output from the RCM is then used to drive the simulation of emissions, transport, atmospheric chemistry and deposition of air pollutants in a chemistry transport model (CTM) over Europe.

Climate projection
In order to investigate the uncertainty related to the choice of GCM in projecting future climate change we used results downscaled by a RCM from two different GCMs; the ECHAM5 model (Roeckner at al., 2006) and the HadCM3 model (Collins et al., 2010).Both climate projections used in this study were derived using the SRES A1B scenario (Nakićenović et al., 2000).The RCM used was the Rossby Centre Regional Climate model version 3, RCA3.Model description and evaluation of both current and future climate simulated with RCA3 is given by Samuelsson et al. (2011) and Kjellström et al. (2011).They show that the simulation of past climate using ERA40 as forcing on the boundaries show improvements in the climate simulated by RCA3 compared to previous versions of the model.Remaining model biases include underestimate of the diurnal temperature range over northern Europe in summer and overestimates of precipitation and surface evaporation in northern Europe in the summer (Samuelsson et al., 2011).When using GCM input to drive the simulations for the control period  using boundaries from HadCM3 gives lowest bias compared to observations while the simulation using ECHAM5 is generally too cold and too wet in northern Europe in summer (Kjellström et al., 2011).The climate as downscaled by RCA3 carries on broad features of the climate simulated by the parent GCM.The projection downscaled from HadCM3 has an average temperature change of 2.9 • C between the periods 1961-1990 and 2040-2070 for the European domain while the corresponding value for the projection downscaled from ECHAM5 is 2.0 • C. The HadCM3 projection thereby has the second highest temperature change in an ensemble of 16 projections downscaled from different GCM runs by RCA3 over Europe by Kjellström et al. (2011) while the ECHAM5 projection has the fourth lowest.Both climate projections show an increase in summer precipitation in large parts of northern Europe extending also to south-east over Poland and Belarus between the periods 1990-2009 and 2040-2959 while precipitation generally decrease in southern and south-western Europe.Connected to this is also a summer total cloudiness increase in northern Europe and decrease in southern Europe with the same spatial pattern as for precipitation.The reduction of precipitation in summer is larger in south-western Europe in ECHAM5 and consequently also changes in cloud cover and soil moisture.These factors are important for ozone production and dry deposition to vegetation.Six-hourly meteorological output on 21 model levels, as well as a range of output variables at the surface, was stored from the RCA3 simulations to be used in the CTM modelling.

Regional chemistry transport model
We used the regional, off-line, Eulerian CTM, MATCH, to simulate the evolution of O 3 and a range of other air pollutants over Europe.The model structure, boundary layer parameterization, advection scheme and numerical treatment are given by Robertson et al. (1999).The chemical scheme in MATCH is based on Simpson et al. (1993), with extensions described by Andersson et al. (2007).The dry deposition of gases and aerosols is calculated using a resistance approach depending on land surface type and for gases the deposition to vegetated surfaces is coupled to soil moisture, temperature, vapour pressure deficit and photosynthetically active radiation (PAR) as described in Andersson and Engardt (2010).The wet scavenging is assumed to be proportional to the precipitation intensity for most gaseous and aerosol components.For O 3 , hydrogen peroxide and sulfur dioxide in-cloud scavenging is calculated assuming Henry's law equilibrium; sub-cloud scavenging is neglected for these species.Important model parameters, such as dry deposition velocities and scavenging coefficients are tabulated in Andersson et al. (2007).The model domain covers Europe and part of the North Atlantic using a rotated latitude/longitude grid with 85 by 95 cells and a horizontal resolution of 0.44 • by 0.44 • (ca.50 by 50 km).In the vertical direction, the CTM domain reaches 5-6 km above the surface using 15 model levels.The lowest model layer is ∼ 60 m thick, increasing to ∼ 700 m in layer 15.The temporal resolution of the meteorological input data is six hours, interpolated to one hour in MATCH; the overall model time step is ten minutes.

Emission data
Anthropogenic emissions of NO x , sulfur dioxide, ammonia, NMVOC and CO used in the MATCH simulations were taken from the RCP4.5 scenario (Thomson et al., 2011;Lamarque et al., 2010).The RCP scenarios have been developed as a replacement of the SRES scenarios and are now the basis for current modelling activities in the climate modelling community.There are four RCP scenarios corresponding to total anthropogenic radiative forcing estimated to 2.6, 4.5, 6 and 8.5 Wm −2 respectively in the year 2100.The RCP data base provides global gridded historic anthropogenic emissions and scenarios for future emissions for different sectors and pollutants geographically resolved on a 0.5 • by 0.5 • latitude-longitude grid.Emission estimates are available for every ten years and for use in the CTM modelling we linearly interpolated this data to yearly resolution.It should be noted that the RCP scenarios assume more aggressive air pollution abatement measures and correspondingly larger decreases in O 3 precursors both in Europe and globally compared to e.g. the SRES scenarios.The extent to which such decreases could be achieved is debated (Fiore et al., 2012).Aggregated over the European CTM domain NO x , NMVOC and CO emissions decrease by 53, 22 and 17 % respectively, from year 2000 to 2050.The decrease is however not uniform.Looking specifically at the changes between 2000 and 2020 we find that emissions decrease by 50-60 % for NO x in e.g.Germany, Italy, France, Spain and UK, while emissions decrease by more than 80 % in Russia and Ukraine.Wild fire emissions were not included in the simulation and the VOC speciation was constant with time.
Global emissions of CO 2 and methane for 2050 are 45 and 37 % larger respectively in SRES A1B, used in the climate projections, than in the RCP4.5 emission scenario used in the CTM modelling.Global emissions of NO x and NMVOC for 2050 are 65 and 46 % larger respectively in SRES A1B than in RCP4.5.SRES A1B therefore corresponds to a world with substantially larger emissions of both long-lived greenhouse gases and air pollutants.For the climate change projections this is not so important for the first half of the 21st century where already committed emissions of long-lived greenhouse gases dominate the radiative forcing while towards the end of the century the climate change signal under the RCP4.5 scenario would be expected to be smaller than for SRES A1B.For Europe emission reductions of air pollutants in the RCP4.5 until 2020 are comparable to e.g. the PRIMES base line scenario by IIASA (Amann et al., 2010) for NO x but smaller for VOC.
Emissions of biogenic isoprene are calculated online in MATCH following Simpson et al. (1995).The isoprene source strength is a function of vegetation type (i.e.tree species), local temperature and PAR.The calculated isoprene emissions increase with PAR and temperature, which are from the driving regional climate model.The relative proportion of the different tree species in each grid cell remains the same throughout all model scenarios.There is currently no treatment of natural sources of nitrogen in MATCH.

Boundary conditions
Boundary conditions of the chemical components simulated by MATCH were held at current levels (ca.year 2000) in the simulations, with a monthly, horizontal and vertical variation as described in Andersson et al. (2007).For O 3 , lateral boundary conditions are based on back-trajectory analyzed measurement data from the EMEP network for year 1999 and for the top boundary on ozone sonde data from Ireland, UK and Norway for the period 1996-2001 (Andersson et al., 2007).Concentrations of O 3 at background locations in Europe have been increasing during the last few decades (Lelieveld et al., 2004;Derwent et al., 2007) although the observed trend has weakened in the first years of the 21st century (Isaksen et al., 2009).A similar increase is also observed in the boundary layer air inflow at the west coast of North America (Parrish et al., 2009).To assess the importance of continued increase of background O 3 we performed one simulation where the background concentration of O 3 increase by 0.1 ppb(v) yr −1 from 1990 to 2100.This is a weaker trend than has been observed during recent decades at Mace Head (Derwent et al., 2007) but still implies a substantial increase of background O 3 of 5 ppb(v) over the period 2000-2050.In this experiment we also allow the concentration of CH 4 to vary according to the concentration pathway provided by the RCP4.5 scenario.This means that the global average concentration of CH 4 increases to a maximum of 1842 ppb(v) in 2040 and then drops to 1576 ppb(v) in 2100.

Model evaluation
To evaluate the performance of the MATCH-RCA3 modelling system, we compare our "Increasing boundary case" (see below) to measured O 3 concentrations for 1990-2009.
The observations were extracted from the EBAS data base (http://ebas.nilu.no).The results of the evaluation are displayed in Table 1.The O 3 measures compared are daily mean and daily maximum averages, for the summer half year for stations north and south of 50 • N respectively.In the evaluation we accounted for the fact that the emissions of ozone precursors change over time during 1990-2009 by forming and comparing station-averages only during those years with valid observations.The number of stations given in Table 1 refers to the total number of stations used in each part of the domain.The requirement for including a station apart from the elevation difference criteria was a 75 % data capture for each year.In the north and south, 82 and 69 % respectively of the included stations had more than 10 yr of valid data.The present setup of the modelling system shows comparable performance as earlier applications of MATCH using climate model output (e.g.Langner et al., 2005;Andersson and Engardt, 2010), but the spatial correlation is not as good.Part of the difference can be related to differences in climate simulated by the GCMs used in the different studies.Evaluation of MATCH driven by meteorological data constrained by observations shows better correlation.For example, evaluation of a 7-yr simulation with MATCH for the period 1997-2003 using ERA40 as meteorological input has been presented by Andersson et al. (2009).The correlation for the whole domain was in this case 0.61 for mean values and 0.87 for daily max values which are better than what we obtained in the present study which are 0.55 and 0.71 respectively.It is important to note that the comparison is sensitive to which simulated time period that is used in the comparison, due to long-term (decadal) variations in the simulated climate that do not necessarily coincide with the observed meteorology.
The spatial correlation for average ozone north of 50 • N is particularly low while it is higher to the south and also higher for average of daily maximum.The bias and RMSE for mean O 3 in north Europe is comparable to the values in south Europe and for mean of daily maximum.The low spatial correlation for average ozone in the north could partly be due to biases in the simulation of current climate in the global climate projection from ECHAM5.However, the scores using meteorological data downscaled from HadCM3 show similar low values for northern Europe, with a correlation of 0.02 and a bias of 12.4 % for summer mean concentrations of O 3 .

Results and discussion
In this section we discuss the results from four MATCH simulations covering the period 1990-2100: 1.  4. Increasing boundary case (ECH RCP4.5 BCtrend); including varying anthropogenic precursor emissions in Europe according to RCP4.5, increasing background ozone with 0.1 ppb(v) yr −1 and changing CH 4 concentrations following RCP4.5.The climate projection downscaled from ECHAM5 is used in this case.
Figure 1a and b show the temporal evolution of summer mean and summer mean of daily max O 3 for the four simulations averaged over the European land area north and south of 50 • N, respectively.The upper two lines in each four-line group correspond to the two climate sensitivity cases (ECH RCP2000 BC2000 and HAD RCP2000 BC2000).Here we kept the O 3 precursor emissions and boundary concentrations constant over time at their year 2000 values to isolate the effect of meteorological changes.For northern Europe we note a slight downward trend for daily max (−0.5 and −0.8 ppb(v) century −1 for ECH RCP2000 BC2000 and HAD RCP2000 BC2000 respectively) with small inter-annual variability.Trends over the whole period are given in Table 2.In southern Europe there are increases in surface O 3 over time for the climate sensitivity cases.The variability in the simulations are different for different time periods, but the trends in the two simulations are similar (+3.5 and +3.2 ppb(v) century −1 for daily max).Earlier studies using the MATCH model have shown that two main drivers for the increasing summer surface O 3 in southern Europe are decrease in dry deposition of O 3 due to reduced soil water and thereby reduced vegetation uptake, and increase in biogenic isoprene emissions, with decrease in dry deposition being the dominant effect in the southernmost parts of Europe and both being equally important in central Europe (Andersson and Engardt, 2010).The reasons for the reduced concentrations in northern Europe have not been addressed in detail here but both cloudiness and precipitation increase in northern Europe in the climate projections which should lead to increased scavenging of ozone precursors and less solar radiation for driving the photochemistry.It is interesting to note the substantial decadal variability in the trend of surface O 3 especially in southern Europe.For example, during the period 2020 to 2030 climate variability leads to decreasing surface O 3 in southern Europe in both the ECH RCP2000 BC2000 and HAD RCP2000 BC2000 simulation, with a more prominent decrease in the latter case.
When we include also the change in ozone precursor emissions in Europe following the RCP4.5 scenario (ECH RCP4.5 BC2000) we obtain decreasing surface O 3 concentrations across Europe.The impact of reduced emissions in Europe clearly dominates over climate change impacts and leads to reductions of daily max O 3 both in northern and southern Europe until 2100 (−9.2 and −12.8 ppb(v) century −1 , respectively).Applying a potential trend of background O 3 of +0.1 ppb(v) yr −1 (ECH RCP4.5 BCtrend) results in partial cancellation of the O 3 decrease due to emission changes in -especially northern -but also southern Europe.The resulting change in surface daily max O 3 from 1990 to 2100 in these regions now become −5.0 and −9.7 ppb(v) century −1 , respectively.The current model setup is less well suited to model ozone trends during 1990-2009 since the meteorological data used to drive the CTM simulations are taken from climate model output.Trends in meteorological variables in the simulation are therefore not necessarily in phase with observed changes.
Due to substantial year to year variability in the simulations it is important to study averages over sufficiently long time periods in order to find statistically significant changes.We use climate projections from two global climate models to account for some of the uncertainty in simulating future climate change.Although the long-term trend in the two simulations is similar differences for shorter periods can be large as a reflection of the inherent variability in the climate system.Table 2 summarizes 20-yr average concentrations and variability for northern and southern Europe over the periods 1990-2009 and 2040-2059 and 2080-2099.The absolute variability is larger in southern Europe than in northern and it increases with time in the climate sensitivity case based on ECHAM5.When using HadCM3 the variability first increases and then decreases.The opposite is seen in Emission reduction case and Increasing boundary case.The decrease in variability until mid-century is probably due to decreasing emissions the increasing variability in the ECHAM5 downscaling cause further variability beyond this point as can be seen throughout the century in the Increasing boundary case.
Figure 2 shows the current and future frequency distributions of O 3 concentrations in northern and southern Europe for the Increasing boundary case and for the Climate sensitivity case 1.For the Increasing boundary case we observe that there is a shift over time of the peak of the distribution to lower concentrations, the shift being slightly smaller for northern Europe than for southern Europe.For high concentrations we note larger changes in the frequency distribution both in northern and southern Europe.This shows that emission reductions within Europe are especially effective in reducing the highest O 3 concentrations connected to O 3 episodes also in northern Europe although the overall average is only reduced slightly due to increasing background concentrations.When considering only climate change (keeping emissions and lateral boundaries at their ca.2000 values), the trends in the frequency distributions are quite different.The peaks of the distributions remain unchanged in both Northern and Southern Europe.For southern Europe, the high end tail of the distribution is, however clearly increased during the period 2040-2059 compared to 1990-2009.This feature is also seen in northern Europe, but to a lesser extent.The implication is that climate change increases the peak values of surface ozone.
The spatial distributions of the changes in mean and daily max O 3 between the periods 1990-2009 and 2040-2059 for the four simulations are shown in Figs. 3 and 4 respectively.Climate change alone increases O 3 in southern Europe in both the climate sensitivity simulations.The increase is gen- erally in the range 0-3 ppb(v) for summer mean.Decrease of 0-3 ppb(v) in summer mean O 3 is simulated in the northern part of the model domain.Changes for summer mean of daily max are somewhat larger, 3-6 ppb(v), in regions of southern Europe, while the reductions in the northern part of the model domain are still in the range 0-3 ppb(v).Increases are somewhat larger when using the HadCM3 projection and also cover larger areas of north-western Europe.When including the trend in RCP4.5 anthropogenic precursor emissions reductions of O 3 are evident everywhere except in northern Africa were O 3 precursor emissions are increasing until 2040 in RCP4.5 and for a few grid points close to Helsinki and Moscow.In these urban areas the reductions in NO x emissions lead to a reduced loss of O 3 through reaction with NO and increasing O 3 concentrations locally.For mean O 3 the effect is visible also for a few other locations in the northern part of Europe, e.g.London and Amsterdam while the effect is not visible further south.This is related to differences in photochemical regimes in different parts of Europe with more NO x sensitive regimes in northern Europe (Beekman and Vautard, 2010).Reductions exceed 9 ppb(v) for summer mean and 12 ppb(v) for summer daily max in a region extending from northern Italy to Hungary.For the Increasing boundary case the assumed upward trend in background O 3 clearly impacts the concentration along the western and northern part of the model domain.The change over the North Atlantic and the British Isles is reversed from negative to positive compared to the scenario with constant boundaries, with net O 3 changes now of 0 to +3 ppb(v) on the British Isles for daily mean.However, most parts of the European land areas still experiences decreasing mean O 3 .Corresponding changes in daily max O 3 have a similar geographical pattern, but smaller magnitude.It is clear from the results presented above that the future evolution of background tropospheric O 3 in the northern hemisphere will be important for the resulting surface O 3 concentration in large parts of Europe.Wild et al. (2012) present estimates of regional changes of surface O 3 based on average source receptor calculations with 14 different global models for present (2001) meteorological and emission conditions.They estimate that changes in annual mean O 3 between 2000 and 2050 averaged over Europe would be negative for all RCP scenarios except for RCP 8.5 which has the largest increase of greenhouse gases.The changes are substantially different than estimated with the SRES A1B, A2 and B2 scenarios, reflecting the assumptions of more stringent O 3 precursor emission controls under the RCP scenarios.Domain and annual average changes in mean O 3 in our Increasing boundary case is −0.2 ppb(v) which can be compared to −2.7 and +0.3 ppb(v) estimated by Wild et al. (2012) for RCP4.5 and 8.5 respectively.The change in our Emission reduction case, which assumes that background O 3 is unchanged around Europe, is −2.5 ppb(v) which is similar to the total change given by Wild et al. (2012) for RCP 4.5 but it is somewhat larger than the change they estimate due to reduction in European precursor emissions for RCP 8.5 and 2.6, which are about −1.5 to −2 ppb(v).It should be noted that the European domains used in these calculations are not identical; the domain used by Wild et al. (2012) being larger and shifted to the east and south.For the summer period the decrease is much larger, −2.4 and −4.3 ppb(v) for our Increasing boundary case and Emission reduction case respectively, illustrating that the reduction in European O 3 precursor emissions are much more important in summer than in the annual mean.

Conclusions
We have studied the impact of climate change and changes in O 3 precursor emissions on summer time surface O 3 over Europe using two different projections of global climate from the SRES A1B scenario and the combined effect of also including ozone precursor emission changes under the RCP4.5 scenario over the period 1990 to 2100.The following conclusions can be drawn: 1.In southern Europe regional climate change gives a substantial positive contribution to future surface O 3 concentrations, but emission reductions are more important leading to reduced O 3 concentrations, also when including increasing hemispheric background concentrations.
2. In northern Europe regional climate change decreases surface O 3 and reduced emissions acts to further strengthen this trend also when including increasing hemispheric background concentrations.
3. The sensitivity to regional climate change using climate projections from two global climate models gave similar spatial changes in surface O 3 and similar long-term temporal trends although the trends differ between different periods.
4. There is substantial decadal variability in the simulations, forced by climate variability, which is important to consider when looking at changes in concentrations of surface O 3 , especially until the first half of the 21st century.
Conclusion 1 and 2 is dependent on our use of the RCP4.5 emission scenario for O 3 precursors in Europe which assumes substantial air pollution abatement measures.It remains to be seen if these reductions can be achieved.In this study we have used a CTM coupled to a regional climate model to simulate details in changes of surface O 3 over Europe over long time periods.A drawback with this model setup is that assumptions have to be made about trends in the concentrations of chemical components on the model boundaries.This means that possible future changes in stratosphere-troposphere exchange, lightning NO x and O 3 loss due to increasing water vapour on the global scale are only implicitly accounted for through an assumed trend on the model boundaries.In an ensemble analysis of 10 global GCM-CTMs the combined effect of these processes on background surface ozone in the northern hemisphere was estimated to −0.8 ± 0.7 ppb(v) for 2000-2030 climate change by Dentener et al. (2006).Additional potentially important missing processes/feedbacks include change in vegetation/land use and change in wild fire emissions.The study of the importance of land use changes is still in an early stage.
In a global model study Ganzeveld et al. (2010) simulated changes in boundary layer O 3 on the order of +2 % between 2000 and 2050 in southern Europe.Ultimately the spatial resolution of global CTMs will be adequate to simulate also regional details of surface O 3 .Until then coupling of global (hemispheric) and regional CTMs forced by a consistent set of meteorological data and precursor emissions is a useful way forward.

Fig. 1 .
Fig. 1.Simulated geographical-average evolution of summer (April-September) mean surface O 3 concentration and summer mean of daily maximum surface O 3 concentration from 1990 to 2100.Results are (a) for the European land area north of 50 • N and (b) south of 50 • N. Thick lines correspond to a 5-yr running mean and the shaded areas correspond to the interval between the 5-yr running mean of the minimum and maximum summer mean values.

Fig. 2 .
Fig. 2. Frequency distributions of daily max O 3 concentrations in North and South Europe during the periods 1990-2009 and 2040-2059 in the Increasing boundary case simulation (a) and Climate sensitivity case 1 (b).The distributions show the number of days per year one arbitrary model grid point in northern and southern Europe experiences daily maximum O 3 concentrations between X and X + 1 ppb(v) during the respective 20-yr period.

Fig. 3 .
Fig. 3. Change in summer (April-September) mean surface O 3 concentration from 1990-2009 to 2040-2059.(a) and (b) changes due to change in climate only; RCA3 downscaling of ECHAM5 and HadCM3, respectively.(c) ECHAM5 downscaling and change in European O 3 precursor emissions.(d) Increasing boundary case; the combined change due to change in climate, change in O 3 precursor emissions and an increasing hemispheric background of O 3 of 0.1 ppb(v) yr −1 .Non-significant changes at 95% confidence level are masked white.

Fig. 4 .
Fig. 4. Change in summer (April-September), daily maximum, surface O 3 concentration from 1990-2009 to 2040-2059.(a) and (b) changes due to change in climate only; RCA3 downscaling of ECHAM5 and HadCM3, respectively.(c) ECHAM5 downscaling and change in European O 3 precursor emissions.(d) Increasing boundary case; the change due to combined change in climate, change in O 3 precursor emissions and an increasing hemispheric background of O 3 of 0.1 ppb(v) yr −1 .Non-significant changes at 95 % confidence level are masked white.

Table 1 .
Modelled surface O 3 compared to observed values over the years 1990-2009 at EBAS stations in Europe a .Modelled (Increasing boundary case, ECH RCP4.5 BCtrend) and measured average values at EBAS sites were constructed for years with measurement data capture exceeding 75 % to account for trend in emissions and observed trend in hemispheric background.Sites at an elevation deviating more than 250 m from the model height were excluded from the evaluation.Model data were taken from the level 3 m above surface. a

Table 2 .
Model simulated 20-yr mean surface O 3 and variability a during different periods over the European land area.
a Variability is defined as the difference between the highest and lowest summer mean value in each 20-yr period.Trends are from linear regressions over the period 1990-2100.