Future impact of non-land based trafﬁc emissions on atmospheric ozone and OH – an optimistic scenario and a possible mitigation strategy

. The impact of future emissions from aviation and shipping on the atmospheric chemical composition has been estimated using an ensemble of six different atmospheric chemistry models. This study considers an optimistic emission scenario (B1) taking into account e.g. rapid introduction of clean and resource-efﬁcient technologies, and a mitigation option for the aircraft sector (B1 ACARE), assuming further technological improvements. Results from sensitivity simulations, where emissions from each of the transport sectors were reduced by 5 %, show that emissions from both aircraft and shipping will have a larger impact on atmospheric ozone and OH in near future (2025; B1) and for longer time hori-zons (2050; B1) compared to recent time (2000). However, the ozone and OH impact from aircraft can be reduced sub-stantially in 2050 if the technological improvements considered in the B1 ACARE will be achieved. North ozone, methane (including stratospheric water vapour), and methane-induced ozone changes. For the B1 scenario, shipping leads to a net cooling with radiative forcings of − 28.0 ( ± 5.1) and − 30.8 ( ± 4.8) mW m − 2 in 2025 and 2050, respectively, due to the large impact on OH and, thereby, methane lifetime reductions. Corresponding values for the aviation sector shows a net warming effect with 3.8 ( ± 6.1) and 1.9 ( ± 6.3) mW m − 2 , respectively, but with a small net cooling of -0.6 ( ± 4.6) mW m − 2 for B1 ACARE in 2050.


Introduction
Increasing population and economic turnover will lead to increasing transport demand notably for aviation (AIR) and maritime shipping (SHIP). This will outpace technological improvements and lead to increasing emissions of various air pollutants, affecting air quality and climate through a complex system of chemical reactions and aerosol interactions. From a climate perspective, the present (2000) impact on radiative forcing (RF) is positive for AIR and negative for SHIP (Fuglestvedt et al., 2008;Balkanski et al., 2010), mostly due to formation of contrail cirrus in the first and sulphate in the latter. A first estimate of aircraft-induced cloudiness based on a climate model study was just recently given (Burkhardt and Karcher, 2011), however, the magnitude of positive RF from contrail cirrus is under debate . Substantial contributions also stem from the emissions of nitrogen oxides (NO x ), carbon monoxide (CO) and non-methane hydrocarbons (NMHCs). These relatively short-lived gases change the oxidative state of the atmosphere and tend to give a positive radiative forcing through the increase in ozone (e.g. Ramanathan and Dickinson, 1979;Berntsen et al., 1997), while enhanced OH levels are known to reduce the lifetime of methane (CH 4 ) and thereby cause negative RF (e.g. Crutzen, 1987;Shindell et al., 2005). Among the various transport sectors, previous studies have shown that the warming effect is most efficient (relative to the number of NO x molecules emitted) for AIR (e.g. Fuglestvedt et al., 2008) because changes in ozone have their largest impact on climate when they occur in the UTLS (upper troposphere/lower stratosphere) region, due to the low temperatures found near the tropopause (Wang and Sze, 1980;Lacis et al., 1990;Hansen et al., 1997). On the other hand, the cooling effect caused by changes in methane lifetimes dominates the SHIP impact (e.g. Myhre et al., 2011) because of the large amounts of NO x emitted into the clean maritime boundary layer. Due to the different effects on the atmospheric composition, it is important to study the AIR and SHIP sectors individually, especially when it comes to initiating mitigation measures. Another important aspect regarding non-land based traffic emissions, as opposed to land based traffic emissions, is the fact that the background conditions are usually relatively clean and this is known to increase the enhancement efficiencies of ozone and OH (e.g. Hoor et al., 2009). IPCC (1999) published an assessment of the impact of aviation on climate, which was later updated by , and recently by Lee et al. (2009). Several other studies have also investigated how aircraft NO x emissions alter the chemical composition of the atmosphere (e.g. Hidalgo and Crutzen, 1977;Johnson et al., 1992;Brasseur et al., 1996;Schumann, 1997;Grewe et al., 1999Grewe et al., , 2007Schumann et al., 2000;Kraabøl et al., 2002;Stevenson et al., 2004;Gauss et al., 2006;Søvde et al., 2007). The studies dealing with impact from future subsonic aircraft NO x emissions project an increase in aircraft-induced ozone in 2050 compared to the present day atmosphere, but the effect depends on the emission scenario used. Søvde et al. (2007) estimated a maximum zonal mean aircraft-induced ozone increase of about 10 ppbv in the UTLS region for 2050 (annual mean) (aircraft NO x emissions of 2.18 TgN yr −1 ). In another study, Grewe et al. (1999) calculated 7 to 10 % increased ozone mixing ratios due to aviation for the same year and region, but for two different emission scenarios (aircraft NO x emissions of 2.15 and 3.42 TgN yr −1 , respectively).
Among the studies on impacts from ship emissions (e.g. Lawrence and Crutzen, 1999;Corbett and Koehler, 2003;Endresen et al., 2003Endresen et al., , 2007Eyring et al., 2005Eyring et al., , 2007Dalsøren and Isaksen, 2006;Dalsøren et al., 2009Dalsøren et al., , 2010, only one has made future projections with atmospheric chemistry models (Eyring et al., 2007). They estimate maximum near-surface ozone contributions from shipping of 5-6 ppbv (annual mean) in the North Atlantic for the year 2000, increasing to 8 ppbv in one of the 2030 scenarios (ship NO x emissions of 3.10 and 5.95 TgN yr −1 , respectively). In a second 2030 scenario the ship emissions were assumed to stabilize at 2000 levels, but the higher background NO x levels caused a slight decrease in the ozone impact from shipping.
The present study is performed within the EU project QUANTIFY (Quantifying the Climate Impact of Global and European Transport Systems), which was the first attempt of investigating the global scale impact on the atmospheric composition due to emissions from each of the transport sectors. Fuglestvedt et al. (2008) investigated the climate forcing from the transport sectors for year 2000, and later Skeie et al. (2009) estimated transport-induced RF for future scenarios. A multi-model study of the year 2000 impact of transport emissions on the atmospheric chemical composition was performed by Hoor et al. (2009) using the preliminary QUANTIFY emissions. They found a maximum ozone increase from aircraft of 3.69 ppbv in the upper troposphere between 30 and 60 • N, and they also found that shipping emissions contributed the most to ozone perturbations in the lower troposphere with around 50 % of the total traffic induced perturbation. The RF results from Hoor et al. (2009) were recently updated by Myhre et al. (2011) who used the final version of the QUANTIFY emissions data to estimate the year 2000 impact. The future impacts of both the emissions and climate changes on transport-induced ozone have so far only been studied by Koffi et al. (2010), using the climate-chemistry model LMDz-INCA (also used in this study). Cariolle et al. (2009) investigated the effect of including an aircraft plume parameterization in large-scale atmospheric models, and similarly the inclusion of a ship plume parameterization was studied by Huszar et al. (2010). Both studies suggested significant reductions in non-land based traffic-induced ozone when plume processes were taken into account. Main results obtained in QUANTIFY are summarized in Lee et al. (2010)  The objective of this study is to investigate how emissions from the non-land based traffic sectors (AIR and SHIP) impact atmospheric ozone, OH, and the resulting RF, if precursor emissions from all sectors show reductions due to environmental concerns leading to improved technology. The effects of possible future high emission scenarios will be dealt with in a follow-up study, using the same set of atmospheric chemistry models. The scenarios selected for this purpose are the SRES (Special Report on Emission Scenarios) optimistic B1 and pessimistic A1B (Nakicenovic et al., 2000), respectively, whereas results from the first are presented in this study. As the B1 scenario is considered to be far more optimistic than the A1B scenario (both for transport and nontransport emissions), due to assumptions of e.g. rapid introduction of clean and resource-efficient technologies in the first, the combined results represent possible low or high developments, respectively, of the transport-induced impact on the atmospheric chemical composition, taking into account uncertainties related to the evolution of e.g. economy and technology. More specifically, the B1 scenario is characterized by environmental concerns leading to improved NO x technology and a relatively smooth transition to alternative energy systems, but without assumptions of climate policies (in accordance with the SRES terms of reference) (Nakicenovic et al., 2000). We have also studied the effect of utilizing additional technological improvements to the aircraft through the B1 ACARE mitigation option (Owen et al., 2010). The fuel efficiency improvements in B1 ACARE probably require the technology to be driven by concerns over climate change and is thus considered a mitigation scenario. In the following we describe the simulation setup and the emission scenarios (Sect. 2), before giving a short presentation of each of the models in the ensemble (Sect. 3). The impacts on ozone and OH are dealt with in Sects. 4 and 5, respectively, while global radiative forcing calculations are presented in Sect. 6. Finally, our conclusions are given in Sect. 7.

Emissions and simulation setup
Traffic emissions of the ozone precursors NO x , CO and NMHC have been developed for year 2000 and for future scenarios through QUANTIFY (data can be downloaded from www.ip-quantify.eu). Anthropogenic emissions from non-traffic sources were taken from the EDGAR3.2-FT2000 inventory  for year 2000, while the future non-traffic emissions evolve according to the IPCC (Intergovernmental Panel on Climate Change) SRES (Nakicenovic et al., 2000) B1 scenario. Table 1 lists the global annual emissions used in this study, and Fig. 1 shows the time development of NO x emissions from different transport sectors and for several scenarios. Although the future increase in the global NO x emissions from AIR and SHIP is relatively small for the B1 scenario, important changes in the regional distribution of emissions can be seen for these transport sectors .
New aircraft emissions scenarios have been developed using the FAST model and are described in Owen et al. (2010). B1 ACARE is a mitigation scenario for aviation and contains additional emission reductions on top of the reductions that are assumed in the B1 scenario. B1 ACARE can be seen as a very optimistic, but feasible, scenario due to concern over climate change. The traffic demand is the same in both scenarios, but excellent fuel efficiency and NO x improvements are assumed in the mitigation scenario, in accordance with the targets set by the Advisory Council for Aeronautical Research in Europe (ACARE, 2002). Consequently, the B1 ACARE scenario has 4 and 24 % lower aircraft NO x emissions than B1 in 2025 and 2050, respectively, and the emissions for 2050 are even lower than the year 2000 estimate. Figure 2 shows that the increase in Europe and Asia in 2025 is smaller for B1 ACARE compared to the B1 scenario, and the emission reductions over the US are stronger in the first scenario. When comparing 2050 with 2025, only a few flight routes have increased emissions in B1 ACARE, and major emission reductions can be seen in Europe, the US, and Japan. The dependency of aircraft NO x emissions on different latitude intervals can be seen in Fig. 4.
Shipping emissions are based on Endresen et al. (2007Endresen et al. ( ) (year 2000 and Eide et al. (2007Eide et al. ( ) (years 2025Eide et al. ( and 2050, and are characterized by increased NO x emissions in the future, even for the optimistic B1 scenario. Already in 2025 the shipping sector could have become the largest emitter of NO x among the three transport sectors (Fig. 1). Most of the increase for the B1 scenario is distributed among 12 new shipping routes (Fig. 3) that are predicted by Eide et al. (2007). It is worth noting the Northern Sea Route which has increasing emissions from 2000 to 2025, and then shows a slight decline from 2025 to 2050.
Emissions from road transport are based on Borken et al. (2007). Future emissions are documented in Uherek et al. (2010). While road traffic was the dominating source of NO x emissions among the transport sectors in 2000, assumptions of stricter vehicle emission standards and improvements in technology will lead to a substantial decrease of NO x emissions in the future scenarios ( Fig. 1), particularly for B1. Technology with low NO x emissions is already available for the road sector, and is faster to implement than for aircraft and shipping as the lifetime of vehicles is shorter. Koffi et al. (2010) showed that the rapid decline in NO x emissions from road transport will lead to a drastic decrease in the ozone impact of road emissions in 2050. For this reason the impact of road traffic on ozone and OH has not been dealt with in this study, but model results from a policy failure scenario (A1B HIGH) for the road transport sector are subject of a follow-up study.
Additional emissions used in this study include biogenic emissions of isoprene and NO from soils (Jöckel et al., 2006), lightning NO x emissions specified at 5 TgN yr −1 (Schumann and Huntrieser, 2007), and biomass burning emissions based on monthly mean Global Fire Emissions Database (GFED) estimates for 2000 (van der Werf et al., 2006(van der Werf et al., ) with multiyear (1997(van der Werf et al., -2002 averaged activity data using emission factors from Andreae and Merlet (2001). The biogenic and soil emissions are fixed using the climatology described in Hoor et al. (2009).
Regarding methane, all models used prescribed surface boundary conditions with global CH 4 abundances taken from IPCC (2001), but with a hemispheric scaling such that the mixing ratios were approximately 5 % higher in the Northern than the Southern Hemisphere. Two of the The global models included in the ensemble have been run with year 2000 emissions and with the future emission scenarios B1 and B1 ACARE for the years 2025 and 2050. For the year 2000 and for the future B1 simulations, a reference run (BASE) and perturbation runs, one for aircraft (AIR) and one for ship (SHIP), have been performed by each model for each year. For the future B1 ACARE simulations, new reference runs and new aircraft perturbation runs were required in order to study the atmospheric chemistry impact of aircraft emissions under this scenario. In the perturbation simulations, a 5 % reduction has been applied to all emitted species of the respective traffic sector. The reasons not to switch off the emissions in the various transport sectors completely are both to reduce non-linearities in chemistry, and because the unscaled response (i.e. impacts due to 5 % emission perturbation) of the chemical system is expected to be closer to the effect of realistic emission changes than a total removal of the emissions (Hoor et al., 2009). The 5 % reduction approach was used to derive the sensitivity of the atmospheric chemical composition, e.g. ozone concentration, to an emission category with an appropriate accuracy (Hoor et al., 2009). The effect of, e.g. road traffic emissions is obtained by multiplying this sensitivity, e.g. change in ozone concentration per kg emission from road traffic, with the total road traffic  emissions. When comparing results from this study with other studies, it is important to note that the 5 % perturbation approach is very different from removing an emission source by 100 %. Non-linearities in chemistry can lead to large differences between the two approaches as described in Hoor et al. (2009) and Koffi et al. (2010). For further discussions on the small perturbation approach, the reader is referred to Grewe et al. (2010) who state that the method of perturbing an emission source by a small amount (e.g. 5 %), in order to derive its sensitivity, is well suited to address impacts of e.g. future emission policies. In order to simplify the comparison with Hoor et al. (2009) and Myhre et al. (2011), the ozone and OH results (Sects. 4 and 5.1) are shown unscaled (i.e. the direct difference between BASE-AIR and BASE-SHIP), while calculations of methane lifetime changes and radiative forcings (Sects. 5.2 and 6) have been scaled to 100 % by multiplying the impacts caused by the 5 % emission perturbation with 20 (i.e. 20 × (BASE-AIR) and 20 × (BASE-SHIP)).  Table 2, and short descriptions are given below.
The models have been evaluated by Schnadt et al. (2010) who compared simulated CO to aircraft measurements from 2003, and found underestimation of tropospheric CO at northern hemispheric middle and subtropical latitudes. The discrepancies were possibly related to the biomass burning emissions which were abnormally high in 2003 (e.g. Yurganov et al., 2005), while biomass burning emissions from 2000 were used in the simulations. Comparisons of the annual cycle of ozone from multi-year sonde observations and from the six models are shown in Appendix A (Fig. A1). Overall, the ensemble mean of the model results agrees relatively well with the observations.

TM4
TM4 is a global chemistry transport model with a horizontal resolution of 3 • × 2 • and 34 vertical layers up to 0.1 hPa. The version used here has been comprehensively described in Williams et al. (2010). The chemical scheme is the modified CBM4 mechanism decribed by Houweling et al. (1998), supplemented with sulphur chemistry and with the chemical reaction data being updated by Williams and van Noije (2008). The advection scheme is the slopes method developed by Russell and Lerner (1981), while convection is based on the Tiedtke mass flux scheme (Tiedtke, 1989).

p-TOMCAT
The global offline chemistry transport model p-TOMCAT is an updated version (see O'Connor et al., 2005) of a model previously used for a range of tropospheric chemistry studies (Law et al., 1998(Law et al., , 2000Savage et al., 2004). The model is used here with a horizontal resolution of 5.6 • × 5.6 • (T21) and extends from the surface to 10 hPa in 31 vertical levels. The chemical mechanism includes the reactions of methane, ethane and propane plus their oxidation products and of sulphur species. The model chemistry uses the atmospheric chemistry integration package ASAD (Carver et al., 1997) and is integrated with the IMPACT scheme of Carver and Stott (2000). The chemical rate coefficients used by p-TOMCAT are taken from the IUPAC summary of March 2005. The Prather (1986) scheme is used for advection while convective transport is based on the mass flux parameterization of Tiedtke (1989).

OsloCTM2
The OsloCTM2 chemistry transport model is used here with both tropospheric and stratospheric chemistry (Gauss et al., 2003;Søvde et al., 2008). It extends from the surface to 0.1 hPa in 60 vertical layers and a horizontal resolution of Gaussian T42 (2.8 • × 2.8 • ) is used. Advection in OsloCTM2 is done using the second order moment scheme (Prather, 1986), while convection is based on the Tiedtke mass flux scheme (Tiedtke, 1989). The Quasi Steady-State Approximation (Hesstvedt et al., 1978) is used for the numerical solution in the chemistry scheme, and photodissociation is done on-line using the FAST-J2 method Bian and Prather, 2002).

LMDz-INCA
The LMDz-INCA model consists of the LMDz General Circulation Model (Le Treut et al., 1998), coupled on-line with the chemistry and aerosol model INCA (Folberth et al., 2006). The version 4.0 of the LMDz model has 19 hybrid levels on the vertical from the ground to 3 hPa and a horizontal resolution of 2.5 • in latitude and 3.75 • in longitude. The large-scale advection of tracers is performed using the finite volume transport scheme of Van Leer (1977), as described in Hourdin and Armengaud (1999). The turbulent mixing in the planetary boundary layer is based on a second-order closure model. The INCA model considers the surface and 3-D emissions, calculates dry deposition and wet scavenging rates, and integrates in time the concentration of atmospheric species with a time step of 30 min. The CH 4 -NO x -CO-O 3 photochemistry, as well as the oxidation pathways of non-methane hydrocarbons and non-methane volatile organic compounds are taken into account. The winds and temperature predicted by LMDz have been nudged to 6-hourly ECMWF data over the whole model domain with a relaxation time of 2.5 h .

UCI CTM
In this study, the University of California, Irvine (UCI) chemistry transport model extends from the surface to 2 hPa with 40 vertical layers and is run at T42 horizontal resolution. The UCI CTM contains separate tropospheric and stratospheric chemistry. The tropospheric chemistry is simulated by the ASAD (A Self-contained Atmospheric chemistry coDe) software package (Carver et al., 1997) with UCI updates (Tang and Prather, 2010), which include the chemical kinetics and photochemical coefficients from the JPL publication 06-2 , and quasi steady-state initial guesses for radicals, and O( 1 D) included with O 3 . The stratospheric chemistry uses the linearized ozone scheme (Linoz), which can include up to four independent species (e.g. O 3 , N 2 O, NO y , CH 4 ), but in this case consists only ozone (Prather and Hsu, 2010). The tropopause, the boundary between the troposphere and the stratosphere, is determined by an artificial tracer (e90) Tang et al., 2011). The advection and convection use the same schemes as OsloCTM2.

MOCAGE
The MOCAGE chemistry transport model is used here with a horizontal resolution of 5.6 • × 5.6 • (T21) and 60 levels up to 0.1 hPa (Teyssèdre et al., 2007). The chemical scheme is a combination of the tropospheric RELACS scheme (Crassier et al., 2000) (a reduced version of the RACM scheme, Stockwell et al., 1997) and of the stratospheric REPROBUS scheme (Lefèvre et al., 1994) including the heterogeneous chemistry described in Luo et al. (1995). The large-scale transport of the tracers is done using a semi-Lagrangian transport scheme (Williamson and Rasch, 1989), while the convective transport is as described in Bechtold et al. (2001).

Ozone
The 2050 B1 annual mean ozone column response to a 5 % perturbation in emissions is shown in Fig. 5 for the aircraft and shipping sectors. The ozone impact from AIR is zonally well mixed, and mostly confined to the Northern Hemisphere (NH) in accordance with the latitudinal distribution of emissions shown in Fig. 4. A method to estimate the contribution of a source is to estimate the sensitivity (ozone change per emission) with a 5 %-perturbation and then to multiply it with the total emission of the source category (e.g. AIR), which effectively equals a scaling of impacts to 100 % by multiplying the response of the 5 % emission perturbation by 20 . The results show a maximum ozone column response of 1.6 DU. The corresponding perturbation for SHIP is also 1.6 DU, but much less homogeneously distributed with maximum values occurring over the North Atlantic Ocean and at the coastal areas of South East Asia. These two areas were also identified as peaks in the 2030 model simulations performed by Eyring et al. (2007), who attributed the increased ship-induced tropospheric ozone column over the Indian Ocean to the higher tropopause and more effective vertical transport found there. Figure 5 also shows the relative standard deviation which represents the spread in results between the different models (see Appendix B for individual model results). Only results for 2050 B1 are shown in Fig. 5, but the relative standard deviations for the other years and scenarios also have similar distributions. The robustness of the models is quite good for the aircraft perturbation case, with a relative standard deviation less than 20 % in most of the areas where the perturbation effect is strong. However, model differences in interhemispheric transport result in a larger relative standard deviation in the Southern Hemisphere (SH). The relative standard deviation for shipping demonstrates larger deviations between the models compared to the aircraft case, and is mainly in the range 30-40 %. One of the models (OsloCTM2) gives large impacts from shipping with a scaled maximum value of 2.5 DU, while another model (MOCAGE) has a corresponding value of only 0.9 DU. However, as model intercomparison is beyond the scope of this study, the reader is referred to e.g. Danilin et al. (1998) and Rogers et al. (2002) for thorough discussions of differences between CTMs. Figure 6 shows the impact of aircraft emissions on ozone in the UTLS region for January and July along with the difference between the B1 and B1 ACARE scenarios (represented by the red lines). When focusing on the B1 scenario, the results indicate an increase in ozone impact from aircraft between 2000 (black dotted line) and 2025 (black dashed line), while the difference between 2025 and 2050 (black solid line) depends on the season and on the latitude. Because of the strong net decrease in NO x emissions from aircraft at northern mid-and high latitudes from 2025 to 2050 (Fig. 4), the resulting ozone effect is a decrease in zonal mean values north of 30 • N during summer (Fig. 6, bottom right) when the photochemistry is more intense. This signal is to a large extent consistent between the models, but it depends strongly on the model whether or not there is an ozone decrease in the NH during winter. In the SH, the ozone impact from aircraft emissions is likely to increase in 2025 and further to 2050, if emissions evolve according to the B1 scenario. The zonal mean local maximum of 0.066 ppbv (or 1.3 ppbv scaled to 100 %) at about 25 • S in 2050 B1 (July) is caused by a combination of increased emissions in the tropics (Fig. 4) and transport across the hemispheres.

Effects of aircraft emissions
The local maximum can also be seen in Fig. 7, which shows the zonal mean ozone impact for 2050 B1 (see Fig. B1 for individual model results), together with the average vertical profile of the NH impact on ozone for each year and scenario. The results show that the maximum absolute impact in the UTLS region is larger in July compared to January, because of the difference in the length of the day between both months. However, the maximum ozone impact relative to the reference simulation (BASE) in 2050 B1 is larger in January with a peak value of 0.21 % (or 4.2 % scaled) located in the middle to upper troposphere at 20-30 • N. This peak in the tropics during winter arises because aircraftinduced NO x and O 3 are transported from the northern midlatitude UTLS to the low-latitude middle troposphere where the background levels of NO x and O 3 are lower. For comparison, Grewe et al. (1999) calculated annual maximum relative ozone changes due to aircraft emissions of 7 % in a 2050 scenario, but the aircraft NO x emissions used in their study were about twice as large as the 2050 B1 emissions that are used here (2.15 TgN yr −1 compared to 1.05 TgN yr −1 ), and the surface NO x emissions were 60 % larger (69.7 TgN yr −1 compared to 43.8 TgN yr −1 ). The comparison is also influenced by several other factors such as the difference in year of meteorology used and potential model developments (e.g.   -chemical reaction rates, deposition parameterizations, improved resolution) since the Grewe et al. (1999) study. Also worth noting is the fact that ozone perturbations in the lower troposphere are weaker during summer than in winter, presumably because the surface deposition to plants is faster and the photochemical lifetime of ozone is shorter (because increased water vapour gives more HO x and thereby ozone loss) at lower altitudes during summer.
If emissions evolve according to the B1 ACARE mitigation option rather than the B1 scenario, Figs. 6 and 7 show that ozone will only be reduced by a small amount in 2025 (red dashed line), while in 2050 (red solid line) the mitigation option will have a substantial effect on reducing ozone levels. A direct comparison between B1 and B1 ACARE yields mean ozone differences poleward of 30 • N in the UTLS of 0.18-0.30 ppbv and 0.94-1.40 ppbv during summers of 2025 and 2050, respectively (Fig. 6, red lines). Notably in the NH, the average ozone perturbation from aircraft in 2050 for B1 ACARE (blue solid line) is reduced to considerably lower values than the estimated impact in year 2000 (black dotted line) for altitudes below 200 hPa (Fig. 7). The fact that the mitigation option only has a minor effect in 2025 could be associated with the long lifetime of aircraft. Hoor et al. (2009) found that the change in ozone burden per NO x -emission was highest for aircraft when comparing with the road and ship transport sectors. In Table 3 we have presented the ozone enhancement efficiency in order to investigate how the sensitivity changes with different years and scenarios. Previous studies (e.g. Grooß et al., 1998;Grewe et al., 1999) have found that the non-linearity in the ozone production normally leads to a smaller positive ozone perturbation per aircraft emitted NO x -molecule when the emissions are higher. This is also the case when we compare the ozone change per aircraft NO x -emission in 2000 with 2025 (Table 3), when the emissions are expected to increase. However, when looking at the 2050 B1 scenario, all models show a higher ozone enhancement efficiency compared to year 2000, although the aircraft NO x emissions also are higher. This unexpected effect can be explained by the change in the location of the emissions. In 2050, the aircraft NO x emissions are shifted further south compared to year 2025 (Fig. 4), and will then take place in more pristine regions where the background NO x levels are lower. Additionally, Hoor et al. (2009) emphasized the role of road traffic for the chemical state of the UTLS, and as NO x emissions from this transport sector are assumed to decrease rapidly in the future, the NO x background levels at cruise altitude are affected. Not surprisingly, the highest ozone enhancement efficiency is found in 2050 B1 ACARE, which is the scenario with the lowest aircraft emissions. For this case, the models estimate that the ozone abundance increases by 2.31 molecules for every NO x molecule emitted from aircraft. Figures 8 and 9 show the effects of shipping emissions on atmospheric ozone in January and July for different years and scenarios (see Fig. B2 for individual model results). Even if emissions evolve according to the optimistic B1 scenario, model results show that the shipping sector will increase its effect on ozone in the future. Focusing on the estimated ozone impact in the lower troposphere in 2050 B1 (Fig. 8), the largest effect from shipping can be found in the North Atlantic Ocean with ozone values of about 0.2 ppbv (or 4 ppbv scaled) during summer when the photochemical activity reaches a maximum. Notably, the impact in the Arctic region is expected to increase in the future due to the expected introduction of new ship tracks associated with melting of the polar ice cap. This area is especially sensitive to emission perturbations because of the low background NO x levels which lead to higher ozone enhancement efficiencies. Consequently, the maximum relative effects in July are found in this region, showing zonal mean impacts exceeding 0.6 % (or 12 % scaled) near the surface (Fig. 9).

Effects of ship emissions
Interestingly, Fig. 8 shows that the ozone impact from shipping at northern mid-and high latitudes will increase from 2025 (dashed line) to 2050 (solid line), especially during winter, although the emissions in these regions are expected to decrease (Fig. 4). Except for transport from lower latitudes, where the emissions increase, this feature can be explained by lower ambient levels of NO x which act to increase the change in ozone burden per NO x emitted from ships. The increase in ozone enhancement efficiency can also be seen in Table 3, where global annual average values are given. In the future B1 scenario, large reductions in anthropogenic non-traffic and road emissions of NO x are assumed over the Eastern US and Europe. This significantly increases the ozone enhancement efficiency from shipping, as transport from these polluted continental areas normally leads to higher levels of NO x over the Atlantic Ocean and the North Sea.

Global OH
Changes in the concentration and distribution of the hydroxyl radical (OH) are important for air pollution and the selfcleaning capacity of the atmosphere (Lelieveld et al., 2002) as OH is the main oxidant in the troposphere. Validation of modelled OH is difficult, however, particularly because of its short lifetime (less than one second) which makes it almost impossible to measure directly. In a recent study by Montzka et al. (2011), indirect measurements of the interannual variability of global OH are consistent with past model studies (Dentener et al., 2003;Dalsøren and Isaksen, 2006;Lelieveld et al., 2006;Duncan and Logan, 2008), and this suggest that larger confidence should be given to models than previously assumed (Isaksen and Dalsøren, 2011).
The impact of aircraft and shipping emissions on tropospheric OH is shown in Figs. 10 and 11 for July, when the effect is largest due to enhanced photochemistry in the NH. For aircraft emissions, there is an indication of a future increase in the impact on tropospheric OH. Figure 10 shows that enhanced OH levels in the UTLS are expected at all latitudes in 2025 (black and blue dashed lines), with an exception near the Arctic region where increased aircraft emissions seem to cause a slight decrease in OH levels. However, when comparing the impacts between 2025 and 2050, the OH response depends strongly on both scenario and latitude. In 2050 (black and blue solid lines), a decrease is seen between approximately 30 and 60 • N, associated with a strong decrease in aircraft NO x emissions (Fig. 4), while a zonal mean increase is expected elsewhere, in particular near the equator where aircraft NO x emissions are assumed to increase. In 2025 B1, the zonal mean average in the UTLS region for July peaks at 40 • N with a value of 8.6 × 10 3 molecules cm −3 (unscaled). When averaging the aircraft-induced OH vertical profile over the entire NH, 2025 B1 and 2050 B1 both show a maximum of 5.4 × 10 3 molecules cm −3 (unscaled) near 250 hPa (Fig. 11).
The largest effects from aircraft emissions during northern summer in the future (July 2050; B1) can be found east and southeast of Asia, but also with significant impacts close to the North Atlantic flight corridor (Fig. 10, top left). The zonal mean OH impact peaks above 300 hPa, and this is also the region of maximum relative impact with a value of 0.6 % (or 12 % scaled) (Fig. 11, top left). In the SH, the absolute values are lower, but due to low background levels the relative impact is fairly high with a value of almost 0.5 % (or 10 % scaled).
As was the case with ozone, the technological improvements that are assumed in the B1 ACARE mitigation scenario have significant effects on OH. If the ACARE targets will be met in 2050, the northern summer OH levels in the UTLS region (red solid line) could be reduced by up to 5.7 × 10 4 molecules cm −3 (Fig. 10, bottom left). The reductions are substantial also in the NH as a whole, showing average differences between B1 and B1 ACARE of 3.9 × 10 4 molecules cm −3 near 250 hPa (Fig. 11, bottom left), and with significant effects also in the middle troposphere. In the near future (2025), the gain of fulfilling the ACARE targets is much lower, but still the maximum difference between B1 and B1 ACARE is 7.6 × 10 3 molecules cm −3 (red dashed line). Hoor et al. (2009) emphasized the large impact of ship emissions on the boundary layer OH levels, and concluded that the effect of ship emissions is more important for the global OH budget than road and aircraft emissions. Figure 10 shows that the impact from ship emissions on OH in the boundary layer is expected to increase in the future if emissions evolve according to the B1 scenario. Between 2025 and 2050, OH will continue to increase at all latitudes except north of 60 • N where a small reduction in ship NO x emissions is expected.
Focusing on northern summer in the future (July 2050; B1), the largest impact from ship emissions can be found in the North Atlantic Ocean with maximum values reaching 3.0 × 10 4 molecules cm −3 (unscaled) (Fig. 10, top right).
As discussed in Sect. 4.2, ozone production in this area is largely sensitive to an increase in NO x emissions, and because of high humidity and strong incoming solar radiation, additional OH is produced when O( 1 D) reacts with H 2 O. Additionally, the background levels of CO and NMHCs, which act to deplete OH, are relatively low in these pristine regions. In the zonal mean, maximum absolute values are found near the surface at 45 • N (Fig. 11, top right), while the relative impact reaches a maximum of about 1.5 % (or 30 % scaled) near 75 • N. The reason is the low background values of OH, as shipping is a dominant source of air pollutants at these high latitudes, and because the production of additional OH caused by ship emissions is more effective in the summer month of July.

Methane lifetime
Emissions from the aircraft and shipping sectors greatly affect the OH concentration, and this leads to changes in the methane lifetime. Methane lifetimes due to reaction with OH have been calculated for each model and for each year and scenario. As in Hoor et al. (2009), monthly mean 3-Dfields of methane and OH were used, and the resulting lifetime changes were then scaled from a 5 % perturbation in emissions to a 100 % perturbation (to get a stronger signal in the RF calculations) by multiplying with a factor of 20 Table 4. Relative changes (%) in methane lifetimes (integrated up to 50 hPa) due to a 5 % decrease in traffic emissions. Values are given relative to the BASE case, and are scaled to 100 % by multiplying with 20. Both the mean of the six models and the standard deviations (indicating the spread of the models) are given. Note that this table does not include the feedback effect of methane changes on its own lifetime.  . The resulting model mean and standard deviation of the methane lifetime changes are given in Table 4 (see Tables B1-B2 for  The rather low relative standard deviation of 12 % is similar for the BASE simulations of the other scenarios, and indicates that the ensemble mean of the models is relatively robust when calculating methane lifetimes. As discussed in Sect. 5.1, SHIP exhibits the largest impact on OH levels and consequently the largest impact on methane lifetime. The model ensemble predicts that the shipping sector contributed to a methane lifetime reduction of 3.68 % in year 2000, and that this number will increase to 3.92 % in 2050, if emissions evolve according to the B1 scenario (Table 4). Eyring et al. (2007) calculated methane lifetime changes from shipping in 2030, and their estimates range from 1.14 % to 1.81 % for a low and high emission scenario, respectively. Their result is much lower than both 2025 B1 and 2050 B1 from this study, and that was also the case for the year 2000 results discussed in Hoor et al. (2009). According to Hoor et al. (2009), part of the differences could be attributed to the very different distribution of ship emissions. The same reasoning applies here as the future ship emissions in Eyring et al. (2007) are more concentrated along the major shipping routes in contrast to the QUANTIFY future ship emissions, which are spread out over larger areas.
The methane lifetime changes for SHIP do not follow the trend in ship NO x emissions, which increased a lot more between 2000 and 2025 B1 than between 2025 B1 and 2050 B1 (Fig. 1). The reason is the assumption of large reductions in land based NO x emissions, particularly from road traffic, in Europe and the US between 2025 B1 and 2050 B1 (not shown). This effect exceeds the impact of increased ship emissions, as a decrease in background NO x levels leads to a more efficient OH production from shipping.
The future methane lifetime changes for AIR are to a certain degree in accordance with the evolution of the aircraft B1 emissions; an increased impact in 2025 followed by stabilization in 2050 (Table 4). The model ensemble predicts a much lower impact of aircraft on methane lifetimes for the B1 ACARE scenario, especially in 2050 when the relative methane lifetime change is lower than the year 2000 value.

Radiative forcings
Radiative forcings have been calculated using the same method as in Myhre et al. (2011). The Oslo radiative transfer model (Myhre et al., 2000) was used to calculate ozone radiative forcings based on monthly mean ozone fields from each model simulation. In order to obtain a robust signal the fully scaled perturbations have been used in the RF calculations, i.e. the ozone change resulting from the 5 % perturbation has been multiplied by 20 (Hoor et al., 2009;Grewe et al., 2010). According to Myhre et al. (2011), the non-linearities arising from the perturbation magnitude are of little importance compared to the inter-model differences in the RF. Figure 12 shows the yearly averaged ozone RF for the 2050 B1 scenario as means of all models and with absolute standard deviations. The aircraft sector has a larger impact than ship emissions in the NH and the changes in RF are relatively homogeneous throughout different latitude zones. The impacts are large throughout the NH and have maximum values reaching 76 mW m −2 near 30 • N. In the SH the ozone RF from aircraft is low, except in the region 0-30 • S. This latitude band also shows a large spread of the models with standard deviations up to 21 mW m −2 , indicating possible model uncertainties related to convection and transport between the hemispheres. For SHIP, the ozone RF is stronger in the SH compared to AIR, but the RF in the NH is much weaker. Maximum impact from SHIP takes place between 30 • N and 30 • S, and peaks at 50 mW m −2 .
Global average ozone RF for all scenarios is given in Table 5 (see Tables B3-B4 for individual model results). For the B1 scenario there is a small increase in the ozone RF from AIR between 2025 and 2050, although the NO x emissions from aircraft are assumed to decrease slightly during this time span. As explained in Sect. 4.1, this is probably caused by the latitudinal shift in the location of the aircraft emissions leading to higher ozone enhancement efficiencies in 2050.  (Table 5) is consistent with the aircraft NO x emissions, which also are assumed to increase (Fig. 1). For the 2050 B1 scenario, Skeie et al. (2009) estimated ozone RF from AIR to be 38 mW m −2 using a simple climate model. The relatively large difference to our study can be partly explained by a lower normalized radiative forcing in this study (36.3 compared to 42.9 mW m −2 DU −1 ), and partly by lower ozone enhancement efficiencies (2.16 compared to 2.69 ozone molecules enhanced per NO x molecule emitted from AIR). In Lee et al. (2009), a scaling approach Table 5. Radiative forcings (mW m −2 ) from changes in ozone, methane (including stratospheric water vapour), and methane-induced ozone for different transport sectors and years/scenarios given as ensemble means and standard deviations. Note that the history of emissions has been taken into account, and that the fully scaled perturbations were used.  was used to estimate the aviation RF of ozone in 2050 to be in the range 59.4-109.8 mW m −2 , depending on the scenario. Their estimates are much higher than in this study, but this is expected because even their lowest emission scenario (SRES B2 with technology 2, see IPCC, 1999) had larger aircraft NO x emissions than the 2050 B1 scenario used here. The effect of the B1 ACARE mitigation strategy is evident already in 2025, with a 1.6 mW m −2 lower ozone RF than in the B1 scenario (Table 5). However, in 2050 the corresponding number has increased to 7.3 mW m −2 , showing the large potential impact of technological improvements in the aircraft sector. The ozone RF from SHIP is calculated to increase between 2025 and 2050, this time due to a combination of the slight increase in ozone precursor emissions for this transport sector, and the stronger ozone enhancement efficiency in 2050 (see Sect. 4.2). Despite the assumed increase in ship emissions between 2000 and 2025 B1, Myhre et al. (2011) have a larger estimate for 2000 (24 mW m −2 ) than our results for 2025 B1. The reason is that the MOCAGE model is included in our study, and this model is at the lower end of the spectrum when calculating ozone RF from SHIP (Table B4). Again, Eyring et al. (2007) calculated lower impacts than in this study, with ozone RF from shipping ranging from 7.9 to 13.6 mW m −2 in 2030. On the other hand, Skeie et al. (2009) estimated a larger impact from shipping with shortlived ozone RF at 36 mW m −2 for the 2050 B1 scenario, also this time due to lower ozone enhancement efficiencies in this study (0.535 compared to 0.859 ozone molecules enhanced per NO x molecule emitted from SHIP).
Emissions of NO x from aircraft and shipping sectors normally lead to an increase in OH concentrations, resulting in a reduction of methane lifetime. The forcing due to the changes in methane is given in the second column in Table 5, and has been calculated mainly using the method described by Berntsen et al. (2005). For each year and scenario, the changes in methane lifetime were multiplied by the estimated methane concentrations reported in IPCC (2001). Further, to account for the impact of methane changes on its own lifetime, a feedback factor of 1.4 was used (IPCC, 2001). The linearized methane specific forcing of 0.35 mW m −2 ppbv −1 was then applied, assuming a background methane mixing ratio of 1909 ppbv and 1881 ppbv for 2025 B1 and 2050 B1, respectively. In Table 5, the CH 4 RF term includes the impact of methane changes on stratospheric water vapour (SWV), and the RF of SWV is assumed to be 0.15 times that of the methane RF (Myhre et al., 2007).
RF from changes in methane-induced ozone is calculated assuming that a 10 % increase in methane leads to a 0.64 DU increase in ozone (IPCC, 2001), and that this ozone has a specific forcing of 42 mW m −2 DU −1 (IPCC, 2001). As explained in Myhre et al. (2011), the chemical model calculations are one year simulations only, hence the methane concentration may not be in steady state with the change in OH during that year, but depends on the time-history of the emissions. In order to correct for this transient response, factors have been applied based on the method described in Grewe and Stenke (2008). The factors for 2025 are 0.88, 0.91 and 0.94 for AIR B1, AIR B1ACARE and SHIP B1, respectively. Corresponding factors for 2050 are 1.00, 1.15 and 0.99. Note that the factor for 2050 B1 ACARE is larger than 1 because the aircraft emissions are assumed to decrease in the preceding years for this scenario.
The sum of the three RF components in Table 5 shows a negative RF from SHIP, and this is consistent with what has been found in previous studies (Fuglestvedt et al., 2008;Skeie et al., 2009;Myhre et al., 2011). For AIR, however, the total RF is slightly positive in three out of four cases. Interestingly, a slightly negative RF is predicted for the 2050 B1 ACARE scenario. It is important to note that the net effect for AIR is the sum of a fairly large positive number (O 3 RF) and two smaller negative numbers (CH 4 plus CH 4 -induced O 3 RF), each associated with uncertainties. The subtle difference between these effects results in a small net value, and considering the uncertainties it is difficult to say for sure whether or not the net effect is positive or negative. However, the general tendency of an increasing importance of methane RF relative to the ozone RF for future air traffic emissions is consistent for both B1 scenarios. It reflects the decreasing rate of ozone RF increase, whereas the CH 4 RF decreases are effective with a time-lag associated with the methane lifetime. For the B1 scenario, a reduction of 1.9 mW m −2 is predicted between 2025 and 2050, and for B1 ACARE the reduction is 3.5 mW m −2 . Compared to Myhre et al. (2011), our estimates of total RF may seem low considering that the aircraft NO x emissions are higher for the B1 scenario in 2025 and 2050 than they were in 2000. The reason is related to the factors used to correct for the time-history of the emissions, which has to be kept in mind when interpreting RF from methane plus induced ozone changes. The aircraft NO x emissions increased rapidly prior to year 2000, while the in-crease levelled off towards 2025, and further turned to a reduction towards 2050 in the B1 and B1 ACARE scenarios. As a consequence, the cooling effect caused by changes in methane and methane-induced ozone RF may compensate the warming from short-term ozone RF (which is unaffected by the time-history of emissions), if emissions evolve according to the optimistic B1 or B1 ACARE scenarios.

Conclusions
Six atmospheric chemistry models have been applied in order to investigate how emissions from the non-land based traffic sectors (AIR and SHIP) impact the distributions of ozone and OH according to the B1 and B1 ACARE emission scenarios. Although the B1 scenario is considered optimistic, model results show that the impacts of both AIR and SHIP emissions on ozone and OH will increase in the future (2025 and 2050) compared to recent time (2000). We used the perturbation approach to calculate both, the contributions of the individual sectors to ozone and OH, and their changes over time, knowing that this methodology has principle limitations in the calculation of contributions . The choice of −5 % emission perturbations guarantees a consistent calculation of the atmospheric sensitivity with respect to aircraft and ship emissions (Hoor et al., 2009). Our contribution calculation highlights the large impact of ship emissions on the chemistry in the lower troposphere, and indicates that ship-induced ozone could exceed 4 ppbv over the North Atlantic Ocean during future summer (July 2050; B1). At the same time, aircraft emissions dominate in the UTLS region with a maximum zonal mean ozone impact that could reach 5 ppbv polewards of 30 • N.
Model simulations with the B1 ACARE mitigation scenario for aviation show modest reductions in ozone levels in 2025, while substantial reductions can be expected in 2050. Zonal means of the UTLS region at northern mid-and high latitudes show that B1 ACARE yields 0.9-1.4 ppbv lower ozone values than the already optimistic B1 scenario during future summer (July 2050), and this is even lower than for recent time (2000). However, all of our future simulations predict an increase in aircraft-induced ozone in the SH compared to year 2000, and this is mainly a response to the assumed increase in aircraft NO x emissions in this region. Additionally, the shift in emission location between 2025 and 2050, from the already polluted mid-and high northern latitudes to the more pristine regions in the south, leads to an increase in the ozone enhancement efficiency with an increase in the ozone concentration of 2.31 molecules per emitted aircraft NO x molecule for the 2050 B1 ACARE scenario.
Emissions from SHIP have important effects on the OH concentrations, particularly in the marine boundary layer, and this impact will become increasingly important in the future. As a consequence, the models estimate a relative methane lifetime reduction of 3.9 % (scaled) due to SHIP in  Fig. 2 in the multi-model study by Stevenson et al. (2006). 2050 B1. The corresponding value for AIR is 1.7 %, but if the ACARE targets will be achieved, this number is reduced to 1.2 %.
The large impact of SHIP on OH is reflected in the radiative forcing calculations. When considering RF from changes in short-term ozone, methane (including stratospheric water vapour), and methane-induced ozone, our results suggest that SHIP will have a net cooling effect in 2025 and 2050 of −28.0 (±5.1) and −30.8 (±4.8) mW m −2 , respectively, for the B1 scenario. The uncertainties relative to net RF are larger for AIR, but positive RF from short-term ozone normally dominates. The resulting RF for AIR in the B1 scenario is 3.8 (±6.1) and 1.9 (±6.3) mW m −2 in 2025 and 2050, respectively. Interestingly, a small cooling effect of -0.6 (±4.6) mW m −2 is estimated for 2050 B1 ACARE, but it is important to note that the time-history of emissions has been taken into account, and this leads to a dominance of RF caused by changes in methane and methane-induced ozone, as the larger aviation emissions prior to 2050 have no impact on the 2050 short-term ozone RF (due to shorter lifetime of ozone compared to methane). In order to obtain knowledge of the total impact from AIR and SHIP on future climate, the RF from CO 2 , contrails (including contrailcirrus) and aerosols must be considered in addition to the RF from ozone and methane presented here.
To summarize, emissions from the two transport sectors aviation and shipping will have an increased impact on atmospheric ozone and OH in the future, even if emissions evolve according to the optimistic B1 scenario. However, the aviation impact through ozone formation can be reduced significantly by initiating the ACARE mitigation option, which is purely based on technological improvements. The long operating time of aircraft suggests that mitigation measures for this traffic sector should be considered at an early stage.

Comparison with ozone observations
Results from the BASE simulation of each model have been compared to ozonesonde observations from Logan (1999) and are shown in Fig. A1. In general, the model results agree relatively well with the observations both regarding magnitude and annual cycle. However, a few exceptions can be found, particularly in the tropics where the OsloCTM2 model is biased high. This bias is well-known from previous studies and recent model development has shown that the inclusion of an HNO 3 forming branch of the HO 2 + NO reaction reduced tropical tropospheric O 3 modelled by OsloCTM2 significantly and improved the agreement with ozonesonde  Table B3. Radiative forcings (mW m −2 ) from changes in ozone, methane (including stratospheric water vapour), and methane-induced ozone for different years/scenarios, due to emissions from aircraft. Note that the history of emissions has been taken into account, and that the fully scaled perturbations were used.  (Søvde et al., 2011). Furthermore, modelling of tropospheric ozone and its precursors is particularly difficult in the tropics due to uncertainties related to convective mixing and lightning parameterizations (e.g. Doherty et al., 2005). At mid-and high latitudes, the UCI CTM model overestimates ozone while there is an underestimation of ozone by p-TOMCAT in the lower troposphere in these regions. As model validation is beyond the scope of this study, interested readers are refered to e.g. van Noije et al. (2006), Shindell et al. (2006) and Dentener et al. (2006) for evaluations of NO 2 columns, CO distributions and deposition budgets, respectively. When interpreting the comparison in Fig. A1, one should be aware that the Logan data are mainly from sondes launched at northern midlatitudes, hence the observations may not be as representative in the tropics and in the Southern Hemisphere. Additionally, the observations are collected from the period 1980-1993, while the models have used emissions from year 2000 and meteorological data from 2003. Nevertheless, Stevenson et al. (2006) found that there have only been minor ozone trends between the Logan data period and year 2000, suggesting that comparisons with Logan data are still meaningful. All in all, we can conclude that the six models are capable of representing atmospheric ozone, and the ensemble mean provides a robust result as individual model errors tend to counterbalance.

Results from individual models
Individual model results are shown for the 2050 B1 scenario as zonal mean ozone perturbations for AIR and SHIP in Fig. B1 and Fig. B2, respectively. Relative changes in methane lifetimes are listed in Tables B1-B2, and RF calculations are given in Tables B3-B4.