Modeling the climate impact of road transport , maritime shipping and aviation over the period 1860 – 2100 with an AOGCM

For the period 1860–2100 (SRES scenario A1B for 2000–2100), the impact of road transport, maritime shipping and aviation on climate is studied using an Atmosphere Ocean General Circulation Model (AOGCM). In addition to carbon dioxide (CO2) emissions from these transport sectors, most of their non-CO 2 emissions are also taken into account, i.e. the forcing from ozone, methane, black carbon, organic carbon, sulfate, CFC-12 and HFC-134a from air conditioning systems in cars, and contrails. For the year 2000, the CO2 emissions from all sectors together induce a global annual-mean surface air temperature increase of around 0.1 K. In 2100, the CO 2 emissions from road transport induce a global mean warming of 0.3 K, while shipping and aviation each contribute 0.1 K. For road transport, the non-CO2 impact is largest between 2000 and 2050 (of the order of 0.1 K) becoming smaller at the end of the 21st century. The non-CO2 impact from shipping is negative, reaching −0.1 K between 2050 and 2100, while for aviation it is positive and its estimate varies between 0 and 0.15 K in 2100. The largest changes in sea-level from thermal expansion in 2000 are 1.6 mm for the CO 2 emissions from road transport, and around−3 mm from the non-CO2 effects of shipping. In 2100, sea-level rises by 18 mm due to the CO 2 emissions from road transport and by 4.6 mm due to shipping or aviation CO2 emissions. Non-CO2 changes are of the order of 1 mm for road transport, −6.6 mm for shipping, and the estimate for aviation varies between −1.2 and 4.3 mm. When focusing on the geographical distribution, the non-CO 2 impact from road transport and shipping on the surface air temperature is only slightly stronger in northern than in southern mid-latitudes, while the impact from aviation can be a factor of 5 stronger in the northern than in the southern hemisphere. Further it is observed that most of the impacts are more pronounced at high latitudes, and that the non-CO 2 emissions from aviation strongly impact the NAO index. The impacts on the oceanic meridional overturning circulation and the Niño3.4 index are also quantified.


Introduction
In recent years, the evidence for anthropogenic impacts on climate (IPCC, 1996(IPCC, , 2001(IPCC, , 2007) ) has increased.Where observational studies have shown that the global mean surface air temperature has risen by around 0.8 K over the 20th century, modeling studies have demonstrated that this increase, in particular since the mid-20th century, can be attributed mainly to anthropogenic influences.Each activity which Published by Copernicus Publications on behalf of the European Geosciences Union.
alters the radiative properties of the atmosphere (by emission or formation of greenhouse gases (GHGs) and aerosols) or modifies the properties of the Earth's surface (a change in land-use can affect the local hydrological cycle or the albedo) may have an impact on climate.A wide range of anthropogenic activities such as industrial production processes, agriculture, transport, power generation, or domestic heating therefore possibly contributes to climate change.
It is important to quantify the contribution from individual sectors such as road transport, shipping, or aviation to climate change, because this allows more informed assessments of the potential effects of mitigation of emissions from these sectors, given the high growth rate of transport emissions in comparison to other anthropogenic sources.Over the last two decades, many studies have been performed which addressed this question -three recent reviews presented current assessments of the radiative forcing due to road (and rail) transport (Uherek et al., 2010), shipping (Eyring et al., 2010), and aviation (Lee et al., 2010) -see also Fuglestvedt et al. (2008).From the transport sector, an important contribution to climate change is from carbon dioxide (CO 2 ) emissions, but the emissions of other species, including short lived ones, are also important.Several studies have assessed the impact of the emission of reactive gases (nitrogen oxides -NO x , carbon monoxide -CO or volatile organic compounds -VOCs) on tropospheric ozone (O 3 ) and the hydroxyl radical (OH) from road transport (Granier and Brasseur, 2003;Niemeier et al., 2006;Matthes et al., 2007), shipping (Granier et al., 2006;Eyring et al., 2007) and aviation (Brasseur et al., 1996;Kentarchos and Roelofs, 2002;Gauss et al., 2006), or from all three sectors (Hoor et al., 2009;Koffi et al., 2010;Dahlmann et al., 2011;Myhre et al., 2011).The present-day tropospheric O 3 changes lead to a radiative forcing between 10 and 30 mW m −2 for the different transport sectors individually.These O 3 perturbations have a typical lifetime of a few weeks (Stevenson et al., 2004).Through the impact of NO x emissions on the OH concentration, transport also affects the concentration of methane (CH 4 ).Generally, this is a reduction in the concentration of CH 4 , and therefore a negative radiative forcing; the CH 4 reduction also leads to an associated reduction in O 3 .Due to the long lifetime of CH 4 , these perturbations have a lifetime in the order of 1 to 2 decades (Stevenson et al., 2004).For the current emissions from the transport sectors the net combined effect (from O 3 and CH 4 ) for road transport and aviation is a positive radiative forcing while it is negative for maritime shipping (Myhre et al., 2011).
The emissions from the transport sectors also lead to increased concentrations of aerosols such as black carbon (BC), organic carbon (OC) and sulfate.Aerosols have a direct impact through their scattering and absorption of radiation.This impact has been assessed for the three transport sectors in Balkanski et al. (2010) and in Eyring et al. (2007) for shipping only.Aerosols also have an indirect effect by changing the albedo (first indirect effect) and lifetime (sec-ond indirect effect) of clouds as well as a semi-direct forcing due to BC -these forcings are much more uncertain (see Lee et al., 2010;Eyring et al., 2010;Uherek et al., 2010).
Several modeling and observational studies have tried to assess the global impact of contrails and aviation-induced cirrus (e.g.Bakan et al., 1994;Minnis et al., 2004;Marquart et al., 2003;Stordal et al., 2005;Burkhardt andKärcher, 2009, 2011;Kärcher et al., 2010).Best estimates for the current radiative forcing due to contrail and aviation-induced cirrus impacts vary between 10 and 80 mW m −2 (Lee et al., 2009).The most recent best estimates with the most detailed models of contrails and aviation-induced cirrus are at the lower end of this range, around 31 mW m −2 (Burkhardt and Kärcher, 2011).Detailed process studies show an increased understanding of the formation and properties of contrails (Paugam et al., 2010).
In addition to detailed studies focussing on one or a few aspects of the impact of transport, some studies have tried to calculate its total radiative forcing (Fuglestvedt et al., 2008;Penner et al., 1999).For aviation, multiple assessments have been made to assess its total impact (Brasseur et al., 1998;Penner et al., 1999;Sausen et al., 2005;Lee et al., 2009).As mentioned earlier, impacts on climate of all three modes of transport have recently been assessed (EU projects QUAN-TIFY and ATTICA) (Lee et al., 2010;Eyring et al., 2010;Uherek et al., 2010).
The impacts on climate in terms of changes in temperature and precipitation, are also of interest.Therefore, some studies use Atmosphere General Circulation Models (AGCMs) coupled to slab ocean models.Slab ocean models represent the ocean mixed layer but not the deep ocean, and they use prescribed local tendencies to represent large-scale ocean transport.These tendencies are assumed not to be affected by climate change, which might influence the quality of the climate predictions of this type of models.AGCMs coupled to a slab ocean model are used to study the climate change which would occur when the perturbation or anthropogenic influence is constant for a long period (e.g.Stuber et al., 2001Stuber et al., , 2005;;Joshi et al., 2003;Ponater et al., 2006), but are not able to describe so well the transient phase of climate change.
When the climate system is subject to a variable forcing, its thermal inertia will cause a delay in its response.Initially, there is a fast response of the atmosphere, the land surface and the ocean mixed layer with a characteristic time scale of 1 to 5 yr (e.g.Hasselmann et al., 1993;Boucher and Reddy, 2008;Olivié and Stuber, 2010).This fast response contributes around 60-80 % to the total long term response.In addition, the deep ocean is responsible for a slow response (which contributes 20-40 %) with an estimated time scale in the order of 100 to 500 yr.For the study of the impact of the transport sectors, this inertia has been taken into account mainly in studies on aviation (Sausen and Schumann, 2000;Ponater et al., 2005Ponater et al., , 2006;;Lim et al., 2007;Lee et al., 2009).Also slightly more complicated Simple Climate Models (SCMs) which contain a simple representation Atmos.Chem.Phys., 12, 1449-1480, 2012 www.atmos-chem-phys.net/12/1449/2012/ of ocean and sea-ice thermodynamics (Harvey et al., 1997;Skeie et al., 2009), have been used.They all simulate reasonably well the transient phase of climate change (Meinshausen et al., 2011a,b), but do not give information on the geographical distribution or seasonal variation of impacts.A more complete but computationally more expensive approach is to use an Atmosphere Ocean General Circulation Model (AOGCM) which contains a detailed description of the atmosphere, ocean and sea-ice.Within the perspective of climate change, oceans play an important role in redistributing heat, in mitigating atmospheric heating in the transient phase due to their large thermal inertia, and in the radiative balance due to the possible presence of sea-ice at high latitudes (impact on albedo).Over the last decade, AOGCMs have been used frequently to model the total climate impact from anthropogenic forcing (IPCC, 2001).These models are currently able to reproduce the temperature change observed in the 20th century, and confidence exists in the quality of their projections of future climate change (IPCC, 2007, Chap. 8).This confidence is stronger at continental than regional scales, and higher for variables as temperature than for precipitation.However, although the AOGCMs are the most advanced tools, there remain important uncertainties in the representation of some processes, for example those defining the strength of cloud-climate feedbacks (IPCC, 2007).In addition, AOGCMs, when used to make global climate studies over multiple decennia or centuries, cannot be expected to describe all physico-chemical processes in the atmosphere in great detail, e.g.emissions, formation and transformation of aerosols and chemical species, and possible removal mechanisms.Therefore, distributions of long-lived GHGs are often assumed to be homogeneously distributed in the atmosphere (CO 2 , CH 4 , nitrous oxide -N 2 O, and chlorofluorocarbons -CFCs), but their global concentration might evolve with time (according to a prescribed scenario).For aerosols and O 3 (or its precursors) which, due to their shorter lifetimes, are inhomogeneously distributed in the atmosphere, threedimensional (3-D) prescribed climatologies are often used based on calculations from more detailed models.Here we present one of the first studies where an AOGCM is used to assess the impact of the transport sectors on climate.Using an AOGCM allows the study of geographical distributions of changes in surface air temperature, precipitation and cloud cover, and to quantify impacts on typical ocean parameters such as ocean 3-D temperature, sea-level rise, and the meridional overturning circulation (MOC).We study the period 1860-2100 and perform full 240-yr long integrations.We focus on the impact of road transport, maritime shipping and aviation, and distinguish between the impact from CO 2 emissions and the impact from all other emissions (which we call, collectively, non-CO 2 ).
In Sect. 2 we describe the AOGCM which is used, the emissions from the transport sectors that are taken into account, and the simulations which are performed.In Sect. 3 we analyze the impacts of the transport sectors on the atmo-sphere, and in Sect. 4 their impacts on the ocean.In Sect. 5 we present our conclusions.
The atmosphere component ARPEGE-Climat4.6 is a spectral model with a T42 horizontal resolution (equivalent to about 2.8 • ×2.8 • ) and 31 hybrid sigma levels (model top at 10 hPa).Turbulent vertical fluxes and dry convection are described by Ricard and Royer (1993), and deep convection (Bougeault, 1985) is modeled using a mass-flux scheme with Kuo-type closure.The cloud microphysics is also described in Ricard and Royer (1993), and the model does not take into account ice supersaturation.The model uses the FMR15 radiation scheme (Morcrette, 1990(Morcrette, , 1991)), which considers two shortwave and six longwave spectral bands.The model takes into account the radiative impact of water vapour (H 2 O), of the well-mixed GHGs CO 2 , CH 4 , N 2 O, CFC-11, CFC-12, and of the O 3 and aerosol distributions, as well as clouds.Six types of aerosols (prescribed as monthly mean 3-D climatologies) are considered: continental (including BC and OC), marine, desert, volcanic, stratospheric, and sulfate aerosols.The model takes into account their direct effect, and the indirect effect of sulfate aerosols based on a parametrisation of Boucher and Rodhe (1994) and Boucher and Lohmann (1995) with a calibration from POLDER satellite data (Quaas and Boucher, 2005).The O 3 distribution is determined by a parametrisation of its homogeneous and heterogeneous chemistry (Cariolle and Déqué, 1986;Cariolle et al., 1990;Cariolle and Teyssèdre, 2007) (see further in Sect.2.2.4).The model contains a description for river routing from Tokyo University (Oki and Sud, 1998;Oki et al., 1999).Changes in land use are introduced through a modification of the fractions of crop and pasture types in the land-surface classification, and the resulting surface properties have been computed with an updated version (ECOCLIMAP-2) of the ECOCLIMAP vegetation map (Champeaux et al., 2005).Both the ocean model OPA8.0 and the sea-ice model GELATO2 (Salas-Mélia, 2002) are grid point models and share the same mesh of 182×152 points.OPA8.0 has 31 levels, including 10 levels in the upper 100 m of the ocean.
The principal changes w.r.t.CNRM-CM3.1 are a revision of the atmosphere-ocean coupling through OASIS2.2www.atmos-chem-phys.net/12/1449/2012/Atmos.Chem.Phys., 12, 1449-1480, 2012 to achieve a better conservation of the energy fluxes during interpolations between the atmosphere and the ocean grids.Moreover, in the ocean and sea-ice models minor corrections have been implemented to improve energy conservation.These improvements have led to reduced drift in the ocean temperature fields and in surface air temperature.Finally, the atmospheric component of the model was updated from version 3 to 4.6.

Forcing agents
To model the climate change over the period 1860-2100, CNRM-CM3.3takes into account changes in well-mixed GHGs, aerosols, total inorganic chlorine, and surface properties.Changes in the total solar irradiance and aerosols resulting from volcanic eruptions are not taken into account.For the period 2000-2100 we consider the SRES scenario A1B (Nakicenovic et al., 2000).The A1B scenario is based on the assumptions, for the 21st century, of rapid economic growth, further population increase until 2050 and a decline thereafter, global adoption of efficient technologies, and a balanced reliance on fossil-and non-fossil-fuel energy sources.In comparison to other SRES scenarios, this scenario results in a middle-of-the road increase in anthropogenic forcing.As the original SRES scenarios are not disaggregated enough, this general storyline was translated into specific emission scenarios for each transport sector individually (cf.Eyring et al., 2010;Lee et al., 2010;Uherek et al., 2010).The transport emission estimates for the 21st century are based on traffic demand estimates (which are assumed to be mainly driven by the gross domestic product (GDP) and population development), fuel efficiency estimates, and emission factor estimates (emission factors indicate how much NO x /SO 2 /BC/... is emitted per unit fuel burnt).A distinction is made between different world regions, different vehicle, ship and aircraft types, different engine types and different fuel types.In scenario A1B, fuel consumption is assumed to grow for all three sectors, with some stabilization for road transport in the second half of the 21st century.For road transport (Uherek et al., 2010) it is assumed that emission standards will be taken over in the next decades on a global scale such that it will ultimately lead to a net decrease in NO x emissions from 2030 on, despite the sustained increase in fuel consumption.For aviation (Lee et al., 2009(Lee et al., , 2010;;Owen et al., 2010), a sustained fuel efficiency improvement is projected over the whole period 2020-2100 (1 % yr −1 ), and the emission factor for NO x is assumed to decrease up to 2050 and remain constant thereafter.For shipping (Eyring et al., 2005(Eyring et al., , 2010;;Eide et al., 2007), it is assumed that gas usage will increase rapidly and strongly, and legislation for reduced emission factors will be implemented at a moderate pace.One assumes that around 2050 maritime shipping will have fulfilled the full potential for emission reduction for the given engine types and fuel types (Eide et al., 2007).The estimated use of biofuels is moderate for shipping and low for aviation.
From the transport sectors, we take into account 6 different forcings.This includes for all sectors, the forcing from CO 2 , CH 4 , O 3 , and aerosols.In addition we take into account the emissions of CFC-12 and HFC-134a from road transport (from air conditioning systems in cars), and contrails and aviation-induced cirrus.The perturbations from CO 2 , CH 4 , CFC-12 and HFC-134a are assumed to be homogeneous perturbations (both horizontally and vertically), while the other perturbations have a geographical and vertical distribution.In Sect.2.2.1 to 2.2.6, we describe the forcings in more detail.
Most of the emission data we use were developed during the QUANTIFY project, and can be found on http://www.ip-quantify.eu/(Borken et al., 2007).These data are spatially disaggregated with a 1 • ×1 • horizontal resolution for each transport sector for the years 2000, 2025, 2050, 2100, and also include time series of decadal global emission estimates for the complete period 1860-2100.In this study we have taken the point of view that we limit the transport activity to the tailpipe emission plus the direct energy consumption to generate the fuel.An overview of the emissions by road transport, maritime shipping and aviation can be found in Tables 1, 2, and 3.

Carbon dioxide
Of all the CO 2 emitted in the atmosphere, about 10 % disappears from the atmosphere on relatively short time scales of around 1 to 2 yr.A large proportion of the emitted CO 2 stays much longer in the atmosphere as it is taken-up only slowly by oceans during the centuries following its emission.A significant fraction (between 20 and 30 %) of the CO 2 remains in the atmosphere on time scales longer than 1000 yr.CO 2 has thus a long residence time and is rather well mixed in the troposphere and the stratosphere (IPCC, 2007).
Figure 1a shows the evolution of the CO 2 concentration from 1860 until 2100 assuming the A1B emission scenario as developed in QUANTIFY for the transport sectors (http://www.ip-quantify.eu/)and the remaining CO 2 emissions from the respective marker scenario (Jiang et al., 2000).The CO 2 concentration has risen from 286 ppmv in 1860 up to 360 ppmv in 2000.One sees a further increase up to 700 ppmv in 2100.Figure 2 . Lund, personal communication, 2009).These SCM simulations allow the derivation of the contributions from the different transport sectors to the total CO 2 concentration and it is this partitioning which is used in the CNRM-CM3.3model.In our simulation, we use observed CO 2 concentrations until 2000 and the SCM modeled CO 2 concentrations from 2000 onwards.To avoid a discontinuity due to a small difference of around 3.5 ppmv between the observed and modeled CO 2 concentrations around 2000, we phase out this transition over the period 1990-2010.
Once the CO 2 contribution from each sector is known, it is possible to estimate the corresponding radiative forcing F from CO 2 using the simple formula in IPCC (2001, Table 6.1), where r 0 CO 2 and r CO 2 refer to the reference and perturbed CO 2 mixing ratios respectively.(Note that the radiative forcing in the AOGCM simulations are computed using the model's radiation scheme, rather than this simple formula.) Figure 1c shows the radiative forcing from CO 2 by the different transport sectors.It is clear that road transport has the largest contribution.The evolution of the radiative forcing from shipping and aviation are very similar (this is specific for the A1B scenario), although the significant contribution from shipping starts somewhat earlier (around 1900) than the contribution from aviation (around 1940).The radiative forcing estimates compare reasonably well with the values in Fuglestvedt et al. (2008) for the year 2000 who obtained 150 mW m −2 for road transport and 35 and 21 mW m −2 for shipping and aviation, respectively.

CFC-12 and HFC-134a
Air conditioning systems in cars currently contain CFC-12 and HFC-134a, and these GHGs can end up in the atmosphere due to leakage or at the end of the car's lifetime (Clodic et al., 2005).Their radiative impact can be as large as 20 mW m −2 and it is therefore taken into account here (G.Rädel, personal communication, 2009).CFC-12 and HFC-134a are also used in cooling and freezing systems for the transport of goods, but we do not take this into account.Figure 1e shows the radiative forcing by these two gases.For the radiative efficiency we use 0.32 and 0.16 W m −2 ppb −1 for CFC-12 and HFC-134a, respectively (IPCC, 2001, Table 6.7).Due to the relatively long lifetimes (100 and 14 yr for CFC-12 and HFC-134a, respectively), their effect goes on well beyond the emission periods (CFC-12 emissions from cars span the 1970 to 2010 period and peak around 1990-1995, HFC-134a emissions span the 1993-2053 period and peak around 2015).We do not take into account the impact from these extra CFC-12 and HFC-134a emissions on stratospheric O 3 destruction.Uherek et al. (2010) estimated that this impact is smaller than −10 mW m −2 in the period 1980-2000.

Methane
The evolution of the CH 4 mixing ratio for the SRES scenario A1B is shown in Fig. 1b: the large increase during the 20th century continues in the first half of the 21st century followed by a decrease.The transport sectors also have an impact on the CH 4 concentration, not by direct CH 4 emissions which are assumed small for the transport sectors, but by an enhanced destruction of CH 4 by increased OH concentrations related to NO x emissions.This impact is often expressed as a reduction of the CH 4 lifetime (Prather, 1994).For the transport volumes of the year 2000, Hoor et al. (2009) calculated for each sector, the impact on the OH distribution.As the reaction with OH is the principal sink of CH 4 , one can use these results to obtain the reduction in CH 4 lifetime by each transport sector.For the year 2000, they found reductions of the CH 4 lifetime of 1.61, 4.12 and 1.04 % for road transport, shipping and aviation, respectively (including the feedback factor to include the long term equilibration to the new steady-state, Fuglestvedt et al., 1999).Their results were obtained using 6 different chemistry transport models (CTMs), driven by meteorological analyses for the year 2003 from the European Centre for Medium-range Weather Forecasts (ECMWF).
We use OH perturbation fields very similar to those of Hoor et al. (2009), derived from simulations by 3 CTMs (p-TOMCAT, LMDZ-INCA and Oslo-CTM2) which used updated emissions for road transport and aviation (Koffi et al., 2010;Hodnebrog et al., 2011).In addition to transport volumes for 2000, the OH impact was also calculated for the assumed transport volumes in 2025 and 2050, with the same 2003 meteorological analyses.
Using the average of the OH distributions obtained with these 3 CTMs, CH 4 lifetimes and lifetime changes for the years 2000, 2025 and 2050 are calculated.For intermediate years over the period 1860-2100 where no CTM results are available, we linearly interpolate or extrapolate the CH 4 lifetime, scaling the CH 4 lifetime change caused by a specific sector with the total global NO x emission from that sector.The time series of the total NO x emission for the different sectors are shown in Fig. 1g.Using these CH 4 lifetimes τ and lifetime changes τ and assuming that the transport sector does not contribute to the CH 4 emissions, one can calculate the impact of the reduced CH 4 lifetime on the CH 4 concentration from the differential equation     where one takes the background evolution of methane r 0 CH 4 from IPCC (2001, Appendix II).In this approach we disregard the stratospheric and soil sinks for CH 4 with lifetimes of 120 and 160 yr, respectively (IPCC, 2001).Also, one must be aware that the reference CH 4 lifetime is based on the mentioned CTM simulations and does not necessarily agree with the lifetime used in IPCC ( 2001), and therefore might hamper a coherent reconstruction.Figure 1d shows the reduction of radiative forcing due to CH 4 perturbations for the different sectors, calculated using IPCC (2001, Table 6.2).
Shipping has the strongest impact on the CH 4 lifetime (Hoor et al., 2009) and thus on its concentration and radiative forcing.Both shipping and aviation show an increasing impact of CH 4 in the 21st century due to increasing NO x emissions.At the end of the 21st century the impact from aviation on CH 4 even equals the shipping impact.We do not take into account the impact from CH 4 -induced O 3 changes, which might add an extra 42 % negative radiative forcing (Hoor et al., 2009).

Ozone
CNRM-CM3.3 contains a simplified O 3 chemistry description which was specifically developed to simulate stratospheric O 3 in AGCMs (Cariolle and Déqué, 1986;Cariolle et al., 1990;Cariolle and Teyssèdre, 2007).To be able to account for the tropospheric and lower stratospheric O 3 changes induced by the transport sectors, we additionally use information from the CTM simulations mentioned earlier.
From these simulations, the impacts of transport on O 3 , NO x , CO, and many other atmospheric components (e.g.OH, see Sect.2.2.3) were obtained for the years 2000, 2025 and 2050.
To take into account the O 3 impact from transport, we use two different methods.In a first approach (dynamical O 3 approach), we use an extended version of the linear O 3 scheme.As the tropospheric O 3 production is strongly dependent on the NO x and CO mixing ratio, the linear O 3 parametrization of the AOGCM (Cariolle and Teyssèdre, 2007) has been extended to take into account the impact of NO x , CO and H 2 O perturbations on the net O 3 production.The net O 3 production P is approximated by with T the temperature, O 3 the local overhead total O 3 column, φ the solar zenith angle, θ the unit step function, and r O 3 , r NO x , r CO , and r H 2 O the mixing ratios of the indexed species.The coefficients c 0 ,c 1 ,...,c 13 are zonal monthly climatologies derived with the 2-D stratospheric chemistry transport model MOBIDIC (Cariolle and Teyssèdre, 2007).Coefficient c 7 (≤ 0) represents stratospheric heterogeneous destruction with r Cl y the total inorganic chlorine mixing ratio.The factor θ (T s − T )θ(φ c − φ) with θ the unit step function assures that the heterogeneous destruction is only ac-tive below a threshold temperature T s = 195 K and for solar zenith angles smaller than φ c = 90 • .The coefficients c 8 ,...,c 13 describe the linearized impact of NO x , CO and H 2 O on the net O 3 production.In this setup the NO x , CO and H 2 O fields can be seen as external forcings, in the same way AOGCMs often use prescribed aerosol climatologies.The zonal-mean distributions of the NO x perturbations can be seen in Fig. 3a, representative for June-July-August (JJA) around the year 2000.In the current study, we do not use the dependency on the H 2 O concentration, and for CO we only use it for the impact from road transport.To obtain the net O 3 production for other years than 2000, 2025 or 2050, we scale the NO x and CO perturbation fields with the total global NO x and CO emissions (http://www.ip-quantify.eu/).In a second approach (fixed O 3 approach), we use the 3-D O 3 perturbations originating from the CTMs directly in the AOGCM.These perturbations can be seen in Fig. 3b and are added to the background O 3 described by Eq. ( 3), but now with c 8 = c 9 = c 10 = c 11 = c 12 = c 13 = 0. To obtain perturbation fields for years other than 2000, 2025 or 2050, we scale the O 3 perturbations with the NO x emissions, assuming that NO x is the predominant factor for tropospheric and lower stratospheric O 3 perturbations.

Contrails and aviation-induced cirrus
When describing the clouds generated by aircraft, one usually distinguishes between linear contrails (condensation trails) and aviation-induced cirrus.Contrails are line-shaped cirrus clouds produced in the wake of an aircraft when hot and moist air from the exhaust mixes with ambient air that is below a critical temperature (Schumann, 1996).Observational studies have been performed using satellite images to estimate the presence of contrails (Mannstein et al., 1998).Different modeling studies have estimated the cloud cover and radiative forcing associated with linear contrails (Marquart et al., 2003;Rädel and Shine, 2008;Rap et al., 2010), where the contrail coverage has been parameterized relying on observed values for the contrail formation frequency (Bakan et al., 1994).Lee et al. (2009) estimated the impact in 2005 of 12 mW m −2 .When linear contrails persist, they can give rise to aviation-induced cirrus which due to the loss of the linear shape become indistinguishable from natural cirrus clouds.Therefore aviation-induced cirrus is much more difficult to estimate.Stordal et al. (2005) have used 16 yr of observational cloud data (1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999) to estimate the increase or decrease in the presence of cirrus, and derived a radiative forcing of 30 mW m −2 that includes both linear contrails and aviation-induced cirrus.Recently, modeling studies have been performed which model the evolution and aging of contrails: Burkhardt andKärcher (2009, 2011) 2008).We impose this 80 % limit as the current version of the atmospheric model does not consider ice supersaturation.The relationship between the recent exhaust tracer rus is calibrated to have a top of the atmosphere (TOA) radiative forcing of 0.024 W m −2 for the year 2000 (slightly lower than the 31 mW m −2 in Burkhardt and Kärcher, 2011).This value of 0.024 W m −2 is based on the assumptions that the impact from linear contrails alone is 0.006 W m −2 in the year 2000 (Rädel and Shine, 2008) and that the impact of aviation-induced cirrus is 3 times larger than from linear contrails (Fuglestvedt et al., 2010).Using these distributions in CNRM-CM3.3,we find TOA net radiative forcings in 1980, 2000, 2025, 2050 and 2100 of 14.9, 24.1, 44.2,  101 and 211 mW m −2 , respectively.In tests where we im- aviation-induced cirrus in the regions with dense air traffic, we use a "recent exhaust" distribution obtained with a CTM (Teyssèdre et al., 2007).We model the distribution of this tracer for the years 2000, 2025, 2050 and 2100, taking as its source the fuel consumption, assuming a lifetime varying with height (i.e. 15 h at 250 hPa and 2 h at 850 hPa), and integrating the large scale transport with ECMWF meteoro-  and if the relative humidity is above 80 % (Rädel and Shine, 2008).We impose this 80 % limit as the current version of the atmospheric model does not consider ice supersaturation.The relationship between the recent exhaust tracer and the presence of linear contrails or aviation-induced cirrus is calibrated to have a top of the atmosphere (TOA) radiative forcing of 0.024 W m −2 for the year 2000 (slightly lower than the 31 mW m −2 in Burkhardt and Kärcher, 2011).This value of 0.024 W m −2 is based on the assumptions that the impact from linear contrails alone is 0.006 W m −2 in the year 2000 (Rädel and Shine, 2008) and that the impact of aviation-induced cirrus is 3 times larger than from linear contrails (Fuglestvedt et al., 2010).Using these distributions in CNRM-CM3.3,we find TOA net radiative forcings in 1980, 2000, 2025, 2050 and 2100 of 14.9, 24.1, 44.2, 101 and 211 mW m −2 , respectively.In tests where we impose a global uniform contrail coverage of 0.01, 0.1 and 1 with a contrail optical depth of 0.3, we find net TOA radiative forcings of 0.16, 1.63 and 15.1 W m −2 .This value of 0.16 W m −2 corresponds well with the values mentioned in Myhre et al. (2009) which is a comparative study among different line-by-line radiative transfer codes and codes used in AGCMs, including the one used in CNRM-CM3.3.For an experiment with a 0.01 global contrail cover at 0.3 opti-cal depth, where the grey body emissivity formulation had been replaced by a two-stream approximation in order to accommodate the prescribed optical properties of the contrail, a value of 0.19 W m −2 was reported for CNRM-CM3.3,close to the best estimate of 0.163 W m −2 .Myhre et al. (2009) further mention that for this experiment they found a strong similarity in the spatial pattern of the radiative forcing among the models, with rather low values at high latitudes.For years other than 2000, 2025, 2050 and 2100, we interpolate the 3-D recent exhaust tracer distributions scaling them with the global annual fuel consumption by aviation, which can be found in Fig. 1f.

Aerosols
The CNRM-CM3.3 model accounts for both direct and indirect effects of aerosols.The indirect effect is however limited to the sulfate component, which leads to an underestimation of the total indirect effect.As the transport sectors are a source of aerosols in the atmosphere, we take into account the BC, OC and sulfate aerosols from the different transport sectors, and use monthly mean 3-D distributions which are available from simulations with the INCA-AER model (Balkanski et al., 2010), using the emissions from http://www.ip-quantify.eu/.Figure 3c-e shows the zonal mean distribution of the perturbation in BC, OC and sulfate aerosols induced by the different transport sectors in JJA 2000.We use these 3-D aerosol distributions in the AOGCM in addition to the standard aerosol distributions in the model.
For the year 2000, we find optical depths at 550 nm similar to those of Balkanski et al. (2010, Table 2).However, changes in net incoming radiation at the TOA are quite different.BC induces changes at the TOA of 1.15, 0.027 and 0.00014 mW m −2 for road transport, shipping and aviation respectively, which is at least a factor 40 lower than mentioned in Balkanski et al. (2010).This will mainly have an impact on the road transport sector (see Fig. 1h).With correct forcings one would probably see a slight increase in the warming from the road sector, although this would depend on the poorly understood semi-direct effect of BC on clouds.
For OC we find equally small values.For sulfate we find changes at the TOA of −32.2, −95 and −4.5 mW m −2 for road transport, shipping and aviation, respectively.This is considerably more than the values in Balkanski et al. (2010), which only included the direct aerosol effect.As we also include the first indirect aerosol effect, part of the differences can be attributed to that.
To obtain the aerosol distribution for years other than 2000, we scale the 3-D distribution with the annual global BC, OC, and SO 2 emissions (see Fig. 1h-j).Road transport is the main contributor to BC and OC emissions, while the SO 2 emissions are strongest for shipping however showing a large reduction at the end of the 21st century.The SO 2 emissions from aviation, that peak around 2050, are much lower than from shipping.  .Schematic representation of the different simulations and impacts (a general impact is represented by X in the ordinate).The black curve (box) denotes the simulation which takes into account all the anthropogenic forcings.The other coloured curves (boxes) represent simulations where the CO2 or non-CO2 emissions of one of the transport sectors are modified.The blue curve (box) denotes a simulation without CO2 emissions of one of the transport sectors, and the red curve (box) denotes a simulation with 5 times the CO2 emissions of the same sector.The green curve (box) indicates a simulation with 5 times the non-CO2 emissions from one transport sector, using the dynamical O3 approach.The purple curve (box) indicates a simulation with the same emissions but using the fixed O3 approach (see Sect. 2.2.4).A simulation with doubled CO2 concentration (grey box) is used to derive the climate sensitivity.

Atmos
To study the impact of the different transport sectors, we perform a number of sensitivity simulations (also indicated in Fig. 4), making separate simulations to quantify the CO 2 and non-CO 2 impact.The non-CO 2 impact includes the effects from O 3 , CH 4 , CFC-12 and HFC-134a, aerosols, and contrails.To study the CO 2 impact, we do two types of simulations, represented in the upper right panel of Fig. 4: a first one without the CO 2 contribution from a certain sector (blue curve), and a second one with five times the CO 2 contribution from that sector (red curve).To study the non-CO 2 impact (see lower right panel in Fig. 4), we perform simulations where we add 5 times the non-CO 2 forcing from a certain sector w.r.t. the reference simulation.We perform simulations using the dynamical O 3 approach (which we will call non-CO 2 , green line or box in Fig. 4), and simulations using the fixed O 3 approach (which we call non-CO 2 *, purple line or box in Fig. 4).Each simulation is repeated 3 times, using different initial conditions for the ocean, sea-ice and atmosphere, resulting in small ensembles of 3 members.The initial conditions for the members of the ensembles are taken from a pre-industrial simulation with a 10 yr time interval.Figure 5 indicates how the use of ensembles and amplification of the forcings reduces the overlap between the uncertainty interval of the reference experiment and a perturbation experiment: using more members in an ensemble reduces the size of the uncertainty interval, and amplifying the forcing increases the spacing between these intervals.Note that due to non-linearity, the best estimate based on an amplification of the forcing (red dot, Fig. 5) might be different from the Note that due to non-linearity, the best estimate based on an amplification of the forcing (red dot) might be different from the actual impact (grey S1 dot).
As the emissions of the transport sectors are assumed negligible before 1890, our perturbation simulations do not show any impact before 1890.Because shipping and aviation have Fig. 4. Schematic representation of the different simulations and impacts (a general impact is represented by X in the ordinate).The black curve (box) denotes the simulation which takes into account all the anthropogenic forcings.The other coloured curves (boxes) represent simulations where the CO 2 or non-CO 2 emissions of one of the transport sectors are modified.The blue curve (box) denotes a simulation without CO 2 emissions of one of the transport sectors, and the red curve (box) denotes a simulation with 5 times the CO 2 emissions of the same sector.The green curve (box) indicates a simulation with 5 times the non-CO 2 emissions from one transport sector, using the dynamical O 3 approach.The purple curve (box) indicates a simulation with the same emissions but using the fixed O 3 approach (see Sect. 2.2.4).A simulation with doubled CO 2 concentration (grey box) is used to derive the climate sensitivity.

Experiments
We perform several simulations with the CNRM-CM3.3model over the period 1860-2100.A schematic picture of the different simulations can be found in Fig. 4. The reference simulation (see black curve and black box in Fig. 4) uses the standard forcings to model the evolution of the Earth's climate over the period 1860-2100 (scenario A1B from 2000 onwards).The CO 2 and CH 4 evolutions used are shown in Fig. 1a-b, but also N 2 O, CFC-11, CFC-12, the surface properties and the sulfate aerosol evolve with time.An overview of the time series of prescribed GHGs as used in the reference simulation can be found in Table 4. Comparing this simulation with a simulation under pre-industrial conditions (grey box in Fig. 4) allows the derivation of the "total anthropogenic impact".
To study the impact of the different transport sectors, we perform a number of sensitivity simulations (also indicated in Fig. 4), making separate simulations to quantify the CO 2 and non-CO 2 impact.The non-CO 2 impact includes the effects from O 3 , CH 4 , CFC-12 and HFC-134a, aerosols, and contrails.To study the CO 2 impact, we do two types of simulations, represented in the upper right panel of Fig. 4: a first one without the CO 2 contribution from a certain sector (blue curve), and a second one with five times the CO 2 contribution from that sector (red curve).To study the non-CO 2 impact (see lower right panel in Fig. 4), we perform simulations where we add 5 times the non-CO 2 forcing from a certain sector w.r.t. the reference simulation.We perform simula-tions using the dynamical O 3 approach (which we will call non-CO 2 , green line or box in Fig. 4), and simulations using the fixed O 3 approach (which we call non-CO 2 *, purple line or box in Fig. 4).Each simulation is repeated 3 times, using different initial conditions for the ocean, sea-ice and atmosphere, resulting in small ensembles of 3 members.The initial conditions for the members of the ensembles are taken from a pre-industrial simulation with a 10 yr time interval.Figure 5 indicates how the use of ensembles and amplification of the forcings reduces the overlap between the uncertainty interval of the reference experiment and a perturbation experiment: using more members in an ensemble reduces the size of the uncertainty interval, and amplifying the forcing increases the spacing between these intervals.Note that due to non-linearity, the best estimate based on an amplification of the forcing (red dot, Fig. 5) might be different from the actual impact (grey S1 dot).
As the emissions of the transport sectors are assumed negligible before 1890, our perturbation simulations do not show any impact before 1890.Because shipping and aviation have almost the same temporal evolution for their CO 2 contribution in scenario A1B (see Fig. 1c), we perform only one simulation that represents the CO 2 impact for both sectors.We also make a 100-yr long simulation where we double the CO 2 mixing ratio (grey box in Fig. 4) w.r.t. the pre-industrial value to estimate the climate sensitivity according to Gregory et al. (2004).
Finally, we want to mention that we do not consider certain forcings from the transport sectors such as the impact of BC  Note that due to non-linearity, the best estimate based on an amplification of the forcing (red dot) might be different from the actual impact (grey S1 dot).
As the emissions of the transport sectors are assumed negligible before 1890, our perturbation simulations do not show any impact before 1890.Because shipping and aviation have Note that due to non-linearity, the best estimate based on an amplification of the forcing (red dot) might be different from the actual impact (grey S1 dot).
on the formation of ice clouds (Penner et al., 2009;Liu et al., 2009), the impact of N 2 O emissions (conversion of NO x into N 2 O in catalytic converters in cars), the impact of water vapour emissions from aircraft in the upper troposphere or lower stratosphere, the direct impact of CH 4 emissions, the impact of CH 4 changes on stratospheric water vapour, or the impact of CH 4 changes on O 3 .We also do not take into account the indirect impact of increased O 3 concentrations (from NO x emissions) on the CO 2 uptake by vegetation (Sitch et al., 2007;Collins et al., 2010).Furthermore, in our approach resultant climate impacts do not feed back or affect the forcing mechanisms.E.g.OH, NO x , and O 3 distributions for the year 2050 have been calculated using 2050 emissions but using year 2000 (or 2003) meteorology, such that impacts of possibly warmer and wetter conditions on the presence of these species are not taken into account.Expected impacts of changes in precipitation on aerosol distributions are not taken into account either.Several studies (Brasseur et al., 2006;Wu et al., 2008;Hedegaard et al., 2008;Koffi et al., 2010) investigated the impact of climate change on tropospheric chemistry and aerosols.Over the period 2000-2050, Wu et al. (2008) ) found that OH changes from climate change prevail on changes from emissions, while O 3 changes are mainly driven by emission changes (Brasseur et al., 2006;Wu et al., 2008;Koffi et al., 2010).Hedegaard et al. (2008) found both regions with increasing and decreasing wet deposition of aerosols, but none of these changes were significant in their simulations.

Atmosphere
In this section, we describe the impact of the 3 transport sectors on some key aspects of the atmosphere over the period 1860-2100: O 3 , TOA forcing, surface air temperature, atmospheric temperature profiles, precipitation, cloud cover, and the NAO index.We show the separate impact of the transport sectors and distinguish between the CO 2 and non-CO 2 impacts, and, as a reference, we also show the total anthropogenic impact.We show time averages over 4 different periods, i.e. 1980-1999, 2011-2030, 2046-2065 and 2080-2099, which also have been studied in IPCC (2007).

Ozone
We use two different methods to take into account the O 3 perturbations.In the first method (dynamical O 3 approach) we impose 3-D NO x and CO perturbations which are used by the extended linear O 3 scheme to calculate the net O 3 production.The actual resulting O 3 mixing ratio might differ considerably from the prescribed O 3 perturbations of the second (fixed O 3 ) approach, where we impose 3-D O 3 perturbations directly.Figure 6a-d shows the impact from the transport sectors on the evolution of the O 3 mixing ratio at 850 and 250 hPa in the SH and NH.At 850 hPa, there is a reasonable agreement between the dynamical and fixed O 3 approaches.The correspondence is rather strong in the SH and for shipping in the NH, but the impacts from road transport and aviation in the NH differ by a factor of 2. In general, the dynamical approach leads to larger O 3 perturbations.For both approaches, we see stronger impacts in the NH than in the SH.Shipping has the smallest hemispheric difference, with an impact in 2100 of slightly more than 3 ppbv in the NH and slightly less than 2 ppbv in the SH.The impact of road transport is strongest between 1990 and 2020, and decreases rapidly after 2020.Finally one can see that at the end of the 21st century the impact of aviation and shipping are of similar magnitude.At 250 hPa, the differences between the dynamical and fixed O 3 approach become especially large for aviation.The impact at 250 hPa is generally dominated by aviation, whereas before 2020 the impact in the SH from road is dominant.The impact from aviation in the SH is almost a factor of 5 smaller than in the NH, a consequence of the much stronger emissions in the NH, the short tropospheric O 3 lifetime and the slow inter-hemispheric mixing.
Figure 6e-f shows the evolution of the total O 3 column in Dobson Units (DU), due to the total anthropogenic impact (left, absolute columns), and due to transport impacts (right, changes in O 3 columns).The O 3 column is around 315 DU in the first part of the 20th century, but at the end of the 21st century it reaches 325 DU.This increase is related to the impact of a colder stratosphere in a changing climate on O 3 concentrations.The Antarctic O 3 hole is manifest in the second half of the 20th and first half of the 21st century.To better see the evolution of the Antarctic O 3 hole, we show the mean O 3 almost the same temporal evolution for their CO 2 contribution in scenario A1B (see Fig. 1c), we perform only one simulation that represents the CO 2 impact for both sectors.We also make a 100-yr long simulation where we double the CO 2 mixing ratio (grey box in Fig. 4) w.r.t. the pre-industrial value to estimate the climate sensitivity according to Gregory et al. (2004).
Finally, we want to mention that we do not consider certain forcings from the transport sectors such as the impact of BC on the formation of ice clouds (Penner et al., 2009;Liu et al., 2009), the impact of N 2 O emissions (conversion of NO x into N 2 O in catalytic converters in cars), the impact of water vapour emissions from aircraft in the upper troposphere or lower stratosphere, the direct impact of CH 4 emissions, tions (from NO x emissions) on the CO 2 uptake by vegetation (Sitch et al., 2007;Collins et al., 2010).Furthermore, in our approach resultant climate impacts do not feed back or affect the forcing mechanisms.E.g.OH, NO x , and O 3 distributions for the year 2050 have been calculated using 2050 emissions but using year 2000 (or 2003) meteorology, such that impacts of possibly warmer and wetter conditions on the presence of these species are not taken into account.Expected impacts of changes in precipitation on aerosol distributions are not taken into account either.Several studies (Brasseur et al., 2006;Wu et al., 2008;Hedegaard et al., 2008;Koffi et al., 2010)   CO 2 emissions from transport also have some effect on the total O 3 column (Fig. 6f), probably due to a cooler stratosphere: there is a small increase at the end of the 21st century of around 1 DU due to road transport (light blue line), and less than 0.5 DU due to shipping or aviation (purple line).
These changes are in agreement with the total anthropogenic impact.For the non-CO 2 impact from aviation, the evolution of the O 3 column clearly shows again the strong difference between the perturbations due to the non-CO 2 (full line) and the non-CO 2 * (dashed line) approaches.

TOA forcing
For CO 2 , CH 4 , CFC-12 and HFC-134a, the radiative forcings from the transport sectors were presented in Fig. 1c-e found both regions with increasing and decreasing wet deposition of aerosols, but none of these changes were significant in their simulations.

Atmosphere
In this section, we describe the impact of the 3 transport sectors on some key aspects of the atmosphere over the period 1860-2100: O 3 , TOA forcing, surface air temperature, atmospheric temperature profiles, precipitation, cloud cover, and the NAO index.We show the separate impact of the transport sectors and distinguish between the CO 2 and non-CO 2 impacts, and, as a reference, we also show the total anthropogenic impact.We show time averages over 4 different periods, i.e. 1980-1999, 2011-2030, 2046-2065 and 2080-2099, which also have been studied in IPCC (2007).

Ozone
fer considerably from the prescribed O 3 perturbation second (fixed O 3 ) approach, where we impose 3-D O turbations directly.Figure 6a-d shows the impact fr transport sectors on the evolution of the O 3 mixing 850 and 250 hPa in the SH and NH.At 850 hPa, the reasonable agreement between the dynamical and fi approaches.The correspondence is rather strong in and for shipping in the NH, but the impacts from road port and aviation in the NH differ by a factor of 2. eral, the dynamical approach leads to larger O 3 perturb For both approaches, we see stronger impacts in the N in the SH.Shipping has the smallest hemispheric diff with an impact in 2100 of slightly more than 3 ppbv NH and slightly less than 2 ppbv in the SH.The im road transport is strongest between 1990 and 2020, a creases rapidly after 2020.Finally one can see that at of the 21st century the impact of aviation and shipping similar magnitude.At 250 hPa, the differences betw dynamical and fixed O 3 approach become especiall for aviation.The impact at 250 hPa is generally dom by aviation, whereas before 2020 the impact in the S road is dominant.The impact from aviation in the SH most a factor of 5 smaller than in the NH, a conse of the much stronger emissions in the NH, the short spheric O 3 lifetime and the slow inter-hemispheric m Figure 6e-f shows the evolution of the total O 3 col Dobson Units (DU), due to the total anthropogenic (left, absolute columns), and due to transport impacts changes in O 3 columns).The O 3 column is around 3 in the first part of the 20th century, but at the end of t century it reaches 325 DU.This increase is related to pact of a colder stratosphere in a changing climate on O centrations.The Antarctic O 3 hole is manifest in the half of the 20th and first half of the 21st century.To be the evolution of the Antarctic O 3 hole, we show the m concentration at 50 hPa averaged over 60-90 • S in S ber in Fig. 6g and over 60-90 • N in March in Fig. 6h 3 members of the reference simulations.One sees portant decrease in the 60-90 • S O 3 concentration in 2010, with an almost complete recovery up to pre-19 ues in 2070-2080.This agrees rather well with resul AGCMs with full stratospheric chemistry (WMO, 20 CO 2 emissions from transport also have some ef the total O 3 column (Fig. 6f), probably due to a cooler sphere: there is a small increase at the end of the 21st c of around 1 DU due to road transport (light blue lin for the period 1860-2100, we perform the radiative transfer calculation twice, once with the transport induced perturbation agents, and once without them (for the dynamical O 3 approach this is not possible).The difference gives the radiative imbalance induced by the forcing agent.With this method we obtain from the non-CO 2 simulations the summed impact from contrails and aerosols, and from the non-CO 2 * simulations the summed impact from O 3 , contrails and aerosols.By taking the difference between these approaches one can also derive the fixed O 3 impact.
Figure 7 shows the net TOA radiation impact from the O 3 forcing (dashed line) and from the combined aerosol and contrail forcing (full line).The O 3 impact in 2000 is 23.5, 16.8 and 13.5 mW m −2 for road transport, shipping and aviation, respectively, and evolves in 2100 to 0.7, 36.5 and 85.3 mW m −2 .Although the NO x emissions in 2100 are still considerably lower for aviation than for shipping, their radiative impact is considerably stronger -the radiative impact of both sectors is of similar size around 2020-2030.Surpris-ingly the values for the year 2000 are lower than the values from Hoor et al. (2009, their Table 8), especially when one takes into consideration their lower road transport and aviation emissions: they found 27.9, 27.3 and 16.3 mW m −2 for road transport, shipping and aviation, respectively.Our values are also smaller than the ones in Myhre et al. (2011) who found 31, 24 and 17 mW m −2 .The discrepancy can be partially caused by differences in the radiation scheme.For road transport, shipping and aviation, the impact from aerosols and contrails is found to be −28.4,−89.0 and 18.3 mW m −2 respectively in the year 2000.These values correspond reasonably well with the results for aerosols and contrails in Sects.2.2.5 and 2.2.6.As the forcing from BC is very small in our model, we find the total impact from road transport aerosols to be negative (this contrasts with Fuglestvedt et al., 2008;Balkanski et al., 2010;Uherek et al., 2010).The value for aviation is the sum of the positive impact from contrails and the negative impact from sulfate.In 2100 we find forcings of −0.7, −18.1 and 209 mW m −2 for road transport, shipping and aviation, respectively.The strong decrease in the SO 2 emissions of shipping in the last 50 yr of the 21st century leads to this strongly reduced forcing in 2100.
In Fig. 8 we show global maps of the net TOA radiative impact from O 3 perturbations (rows a, b and c) and from aerosol and contrail perturbations (rows d, e and f).The O 3 perturbations are more zonally homogeneous than the summed aerosol and contrail perturbations, and for aviation we see the strongest impact between 10 and 40 • N. When focussing on the aerosol and contrail impacts, we also see for road a negative impact caused by sulfate (due to underestimation of the BC impact).The impact is clearly stronger in the northern hemisphere (NH).For shipping, the strongest impact is over the NH oceans, and this perturbation is clearly greatest in the periods 1980-1999, 2011-2030 and 2046-2065, in agreement with the time series of SO 2 emissions.For aviation, the strongest positive impact is seen over the most frequently used aircraft routes in the NH.In regions of low air traffic however, we see some areas where the impact on TOA radiation is negative.This is clearest in the period 2046-2065 in the NH over the Arctic, and over the subtropical Atlantic and Pacific.As we do not find any of these signatures in simulations that included contrails only (not shown), we attribute the local negative forcing to sulfate.Sulfate has a clear signature in the Arctic (see the aviation sulfate distribution in Fig. 3c), is mainly confined to the NH and peaks around the year 2050.As the lifetime of contrails and SO 2 /sulfate are different, one can have different distributions for their radiative forcing.Moreover the contrail and sulfate distributions result from two distinct CTMs, driven by different ECMWF analyses.

TOA forcing
For CO 2 , CH 4 , CFC-12 and HFC-134a, the radiative forcings from the transport sectors were presented in Fig. 1ce.Here we present the TOA radiative forcing caused by O 3 (fixed O 3 approach), contrails and aerosols.To obtain these values for the period 1860-2100, we perform the radiative transfer calculation twice, once with the transport induced perturbation agents, and once without them (for the dynamical O 3 approach this is not possible).The difference gives the radiative imbalance induced by the forcing agent.With this method we obtain from the non-CO 2 simulations the summed impact from contrails and aerosols, and from the non-CO 2 * simulations the summed impact from O 3 , contrails and aerosols.By taking the difference between these approaches one can also derive the fixed O 3 impact.reference simulation.Around the year 2000, the temperature increase is around 0.8 K w.r.t.1860, increasing by almost 3.0 K in 2100.One notices that the three members of the ensemble show a very similar behavior.Using the results from the pre-industrial and doubled CO 2 simulation, we can determine the climate sensitivity using the method of Gregory et al. ( 2004) and find a value of around 0.8 K (W m −2 ) −1 , which corresponds with a 2.97 K temperature increase for a doubling of the CO 2 concentration.This is well within the interval mentioned in IPCC (2007) of 2.1 to 4.4 K with a mean value of 3.2 K.
Figure 9b shows time series of the impact of the transport sectors on the evolution of the global annual mean of the surface air temperature.The thin black lines show the annual global mean impact for the individual members of the ensemble, and the thick colored lines show the 11-yr running mean, averaged over the 3 ensemble members.For CO 2 the impact of road transport is strongest, showing a temperature increase of around 0.05 K in 2000, reaching 0.3 K in 2100.For aviation and shipping, the temperature impact until 2050 is small, reaching 0.1 K in 2100.The non-CO 2 impact from road is strongest between 2000 and 2050 (around Figure 7 shows the net TOA radiation impact from the O 3 forcing (dashed line) and from the combined aerosol and contrail forcing (full line).The O 3 impact in 2000 is 23.5, 16.8 and 13.5 mW m −2 for road transport, shipping and aviation, respectively, and evolves in 2100 to 0.7, 36.5 and 85.3 mW m −2 .Although the NO x emissions in 2100 are still considerably lower for aviation than for shipping, their radiative impact is considerably stronger -the radiative impact of both sectors is of similar size around 2020-2030.Surprisingly the values for the year 2000 are lower than the values from Hoor et al. (2009, their Table 8), especially when one takes into consideration their lower road transport and aviation emissions: they found 27.9, 27.3 and 16.3 mW m −2 for road transport, shipping and aviation, respectively.Our values are also smaller than the ones in Myhre et al. (2011) who found 31, 24 and 17 mW m −2 .The discrepancy can be partially caused by differences in the radiation scheme.For road transport, shipping and aviation, the impact from aerosols and contrails is found to be −28.4,−89.0 and 18.3 mW m −2 respectively in the year 2000.These values correspond reasonably well with the results for aerosols and contrails in Sects.2.2.5 and 2.2.6.As the forcing from BC is very small in our model, we find the total impact from road transport aerosols to be negative (this contrasts with Fuglestvedt et al., 2008;Balkanski et al., 2010;Uherek et al., 2010).The value for aviation is the sum of the positive impact from contrails and the negative impact from sulfate.In 2100 we find forcings of −0.7, −18.1 and 209 mW m −2 for road transport, shipping and aviation, respectively.The strong decrease in the SO 2 emissions of shipping in the last 50 yr of the 21st century leads to this strongly reduced forcing in 2100.
In Fig. 8 we show global maps of the net TOA radiative impact from O 3 perturbations (rows a, b and c) and from aerosol and contrail perturbations (rows d, e and f).The O 3 perturbations are more zonally homogeneous than the summed aerosol and contrail perturbations, and for aviation we see the strongest impact between 10 and 40 • N. When focussing on the aerosol and contrail impacts, we also see for road a negative impact caused by sulfate (due to underestimation of the BC impact).The impact is clearly stronger in the northern hemisphere (NH).For shipping, the strongest impact is over the NH oceans, and this perturbation is clearly greatest in the periods 1980-1999, 2011-2030 and 2046-2065, in agreement with the time series of SO 2 emissions.For aviation, the strongest positive impact is seen over the 0.05 K), and reduces thereafter.This is mainly caused by a strong reduction in the road transport emissions of NO x in the second half of the 21st century and of the earlier reductions in the emission of CFC-12 and HFC-134a.Taking into account a stronger impact of BC would probably strengthen this behaviour.The non-CO 2 emissions from shipping show a negative impact on the temperature of around −0.05 to −0.1 K over the period 2000-2100.This is caused by significant SO 2 emissions leading to the formation of sulfate aerosols, together with a strong impact of OH on CH 4 by, on the one hand, significant NO x emissions, and on the other hand, a characteristic strong impact of NO x shipping emissions on the CH 4 lifetime (Hoor et al., 2009).For the non-CO 2 impact from aviation we see a strong difference between the non-CO 2 and non-CO 2 * approaches, caused by the rather different O 3 perturbations (see Fig. 6).The non-CO 2 approach shows a positive impact reaching 0.15 K in 2100.This is caused by the strong increase in the NO x aviation emissions that are more than 2.5 times more efficient than the other transport emissions at producing O 3 (see Hoor et al., 2009), and by the impact from the linear contrails and aviation-induced cirrus.However, in the extended linear O 3 scheme the O 3 production in the upper troposphere seems to be overestimated (see Sect. 3.1).Using the non-CO 2 * approach leads to almost no temperature signal, except for a very small increase in the last part of the 21st century.Both approaches probably are affected by the too strong negative forcing from sulfate aerosols (see Sect. 2.2.6).Taking this into consideration together with the fact that the model is not very sensitive to O 3 perturbations (see Sect. 2.2.4), we assume that the actual impact from aviation will be somewhere in between the results for both approaches.Sausen and Schumann (2000) made projections of the impact of aviation on the global mean surface air temperature.They included the impact from CO 2 and O 3 changes due to NO x emissions but not the impact from the reduction in the CH 4 lifetime or the impact from contrails or aviationinduced cirrus.One should also note that the climate sensitivity of their SCM was rather low, i.e. 0.61 K (W m −2 ) −1 .They found a temperature increase of 0.006 K in 2000, and for their scenarios Fa1, Eab and Eah, respectively 0.025, 0.033 and 0.050 K in 2050 and 0.047, 0.086 and 0.146 K in Atmos.Chem.Phys., 12, 1449-1480, 2012 www.atmos-chem-phys.net/12/1449/2012/2100.We find in 2100 a total impact from aviation of around 0.25 K for the non-CO 2 approach and around 0.15 K using the non-CO 2 * approach.Skeie et al. (2009) performed simulations with a more evolved SCM, using the SRES scenarios A1B, A2, B1 and B2, and calculated the combined CO 2 and non-CO 2 impact on surface air temperature for the different sectors.With a climate sensitivity of 0.8 K (W m −2 ) −1 , they found for the A1B scenario, an annual global mean surface air temperature impact in 2100 of 0.38 K for road, 0.02 K for shipping, and 0.28 K for aviation, which is in rather good agreement with our results.They used similar emission scenarios, however they took into account more forcings (stratospheric water vapour feedback, methane ozone feedback, ...).Berntsen and Fuglestvedt (2008) did a similar study but only looked at the present-day impact.
We now focus on the geographical distribution of the surface air temperature impact.In Fig. 10 we present the annual mean changes averaged over the periods 1980-1999, 2011-2030, 2046-2065 and 2080-2099 from the total anthropogenic impact (row a) and from the separate transport impacts (rows b-g).Note the difference in the contour intervals between the first row and the other rows.For the total anthropogenic impact, a clear signal can already be seen in 1980-1999, which increases gradually until 2080-2099.Continents show a considerably stronger impact than oceans, and the temperature over the Labrador Sea shows an even stronger negative temperature response in 2046-2065 and 2080-2099.In the Southern Ocean, we also find a weak warming.A strong impact is also noticeable poleward of 65 • N.These results compare well with results shown in IPCC (2007, Fig. 10.8).
For the road sector, we find a significant CO 2 impact of 0.2 K over continents for 2011-2030, with some regions showing increases of more than 0.4 K in 2080-2099.Large similarities with the total anthropogenic impact exist, e.g. the stronger impact at high northern latitudes and the stronger impact over continents.For the CO 2 impact from shipping and aviation, increases of 0.2 K over continents are seen in 2080-2099.The non-CO 2 impact for road transport is slightly positive in 2011-2030 and 2046-2065, whereas that from shipping is clearly negative and rather constant over the 21st century, with some extremes over the continents and over the northern high latitude regions.Using the non-CO 2 approach, the impact from aviation is clearly positive by 2011-2030, and a steady increase is seen for the periods 2046-2065 and 2080-2099 especially in the NH.However, using the non-CO 2 * approach, we observe a clearly positive signal over the NH mid-latitudes only in 2080-2099.
In order to investigate the latitudinal dependence of the impacts, we show the annual zonal-mean surface air temperature in Fig. 11a.Results are shown for the periods 1980-1999, 2011-2030, 2046-2065 and 2080-2099, and are the mean of the 3 members over the 20-year periods.We also indicate the 95 % confidence intervals with thin lines.The zonal mean temperature response to the total anthropogenic forcings shows a smaller impact at mid-latitudes (50-60 • S, 40-50 • N), but an amplification at the poles which is most pronounced in the Arctic.The confidence intervals are larger closer to the poles and smaller in the extra-tropics to midlatitudes.Figure 11b-i shows the CO 2 and non-CO 2 impacts of the transport sectors on the annual zonal mean surface air temperature.The CO 2 impact from shipping and aviation only becomes clearly distinguishable in 2080-2099 (around 0.1 K in low and mid-latitudes), with a large amplification in the Arctic.The impact from road transport is stronger: at low latitudes, it is significant as from 2011-2030 (0.1 K), going up to 0.3 K in 2080-2099, and shows also a strong amplification in the Arctic.In contrast, the non-CO 2 impact from road in the tropics and at mid-latitudes is strongest in the periods 2011-2030 and 2046-2065 and has large confidence intervals in the Arctic and Antarctic.From shipping, the non-CO 2 impact is negative everywhere (except at southern high latitudes for the period 2011-2030 when it is weakly positive), with similar values in 2011-2030 and 2046-2065 in the tropics and extra tropics, and larger ones at northern high latitudes.The strongest non-CO 2 signal is from aviation with a very asymmetric response of up to 0.3 K in the region 20-60 • N in 2100.The non-CO 2 * approach however shows a much weaker signal, clearly positive between 20 and 50 • N, but negative between 60 and 90 • N.This local negative impact is possibly a consequence of the over-estimated sulfate (see the aviation sulfate distribution in Fig. 3c) and underestimated O 3 impacts.
To estimate the difference in geographical distribution of the impacts, we quantify the correlation between the patterns of climate change in the different periods shown in Fig. 10.For each impact in each period we calculate the correlation with the distribution that shows the strongest signal, i.e. the total anthropogenic impact in 2080-2099.Similar comparisons have been performed in IPCC (2007, Table 10.5), but using different measures (Watterson, 1996).Figure 12a shows these correlations for the surface air temperature distribution.One sees a strong correlation for the first 3 periods of the total anthropogenic impact, and for the CO 2 impact from road transport.Further, we see a negative correlation with the non-CO 2 impact from shipping.For the non-CO 2 impact from road, we see a positive correlation, which, however, disappears in 2080-2099.
In the sensitivity simulations, we amplified the forcings by a factor of five to derive the impacts from the different transport sectors.Whether such an amplification of the forcing is justified is determined by whether the perturbations are still small enough for the system to be in a linear regime (see Fig. 4).For the strongest perturbation (i.e.where we model the CO 2 impact from road) we can expect the strongest nonlinearity.In the year 2100, we find an increase in global mean surface air temperature of 0.3 K due to the CO 2 emissions from road transport.This means that the actual temperature difference between the simulations is around 1.5 K, which is large w.r.-1 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 -0.05 0.05 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 K Fig. 10.Global maps of the annual mean surface air temperature impact in the periods 1980-1999, 2011-2039, 2046-2065 and 2080-2099 caused by (a) the total anthropogenic forcing, the CO2 forcing from (b) road transport and (c) shipping or aviation, by the non-CO2 forcing from (d) road transport, (e) shipping and (f) aviation where we use the dynamical O3 approach to take into account the effect of O3, and (g) by the non-CO2* impact from aviation (using the fixed O3 approach).Notice that the contour intervals are 5 times larger for the total anthropogenic perturbation (a). 1. Annual zonal mean surface air temperature impact in 1980-1999 (green), 2011-2030 (blue), 2046-2065 (red) and 2080-2 ): (a) total anthropogenic impact, CO 2 impact from (b) road transport and (c) shipping or aviation, non-CO 2 impact using the dyna 3 approach for (d) road transport, (e) shipping and (f) aviation, and non-CO 2 * impact using fixed O 3 approach for (g) road transp ipping and (i) aviation.Thick lines are the mean over the 3 members and thin lines indicate the 95 % confidence interval.latitude mid-and upper-troposphere, up to 1, 1.5, 3.5 and 5 K for the respective 4 periods, which correspond well with those in IPCC (2007, Fig. 10.7).The impact at the surface is smaller and shows more variability.Figure 15b-c shows the impact on the annual zonal mean temperature profile by CO 2 from road transport and maritime shipping or aviation.The CO 2 impact from road transport is much stronger than that from shipping or aviation.In the mid-to upper-troposphere it varies around 0. 05, 0.15, 0.35 and 0.5 K in 1980-1999, 2011-2030, 2046-2065 and 2080-2099, respectively.For shipping and aviation, there is no clear signal in 1980-1999, 2011-2030 and 2046-2065 (although an indication of a lowlatitude mid-to upper-troposphere temperature increase ap-is a distinctive negative impact over the whole dep troposphere which is strongest in 2080-2099.The n impact from aviation is clearly positive in the trop and in contrast with that from road transport and shipping, it is very asymmetric being much strong NH with a maximum around 300 hPa at 30-60 • N. H temperature increases by more than 0.5 K in 2100 the non-CO 2 * approach the general pattern is sim though the heating is weaker and one observes a ened cooling around 200 hPa between 50 and 80 • N over, a negative signal appears at the surface betwee 90 • N. Using an AGCM coupled to a slab ocean m the emission scenario Fa, Ponater et al. (2005) foun imum zonal mean temperature impact from aviatio year 2050 of 0.35 K around 40-50 • N at 300-400 hP

Precipitation and cloud cover
Time series of cloud cover and precipitation can be Fig.9c-f.Taking into account the total anthropog pact, the precipitation rises by 0.2 mm day −1 in 210 the cloud cover decreases by 1.8 %, coinciding wit crease in global mean surface air temperature.A si lationship is seen for individual transport sectors, for cloud cover the evolution is rather noisy.Lar pacts can be observed for CO 2 from road transpor ing in 2100 a 0.015 mm day −1 increase in precipitat 0.15 % decrease in cloud cover.All other impacts a 0.01 mm day −1 for precipitation and 0.1 % for cloud To get an idea of the zonal mean distributions, the total anthropogenic impact in Fig. 16.The clo remains almost unaffected in the tropics, decreases c ably in the subtropics, and has a strong increase at high-latitudes.These characteristics correspond w those of IPCC (2007, Fig. 10.10).There is generally fidence in the changes in precipitation but increase cipitation at high latitudes are very consistent among and can also be observed in Fig. 16b.Our model als a decrease in subtropical regions.
In Fig. 12b-c, we show the correlation between terns of precipitation or cloud cover with the tota pogenic impact in 2080-2099 (as for the surface air ature).The results are similar to those for surface air ature but in general the correlations are weaker, e.g.CO 2 impact from shipping shows a strong anti-co only for the period 2080-2099.The behavior of the n impact from aviation is very different: there is no co neither for cloud cover nor for precipitation.obtain the CO 2 impact from road transport, we thus compare a simulation containing 5 times the CO 2 emissions from road transport (red line in Fig. 4) with a simulation with no CO 2 emissions from road transport (blue line).The CO 2 impact from road transport can also be analyzed using the difference between the reference simulation (black line) and the simulation with no CO 2 emissions from road (blue line).Figure 13 shows the evolution of the annual global mean surface air temperature for both approaches.The approach without amplification of the forcing gives a much noisier result, with a high inter-annual variability, while the approach with a 5 times stronger forcing shows a much more stable signal and a suppressed variability, but on average both approaches coincide.Figure 14 shows the impact on the zonal mean surface air temperature from both approaches for the periods 1980-1999, 2011-2030, 2046-2065 and 2080-2099.There, one sees that both approaches agree in the tropics up to the mid-latitudes rather well, but that at high latitudes the differences are larger.The 95 % confidence interval is considerably smaller for the simulations where we apply 5 times the actual forcing.

Atmospheric temperature profile
Figure 15a shows the total anthropogenic impact on the annual zonal mean temperature profile for the periods 1980-1999, 2011-2030, 2046-2065 and 2080-2099 w.r.t. the preindustrial simulation.Strongest values appear in the lowlatitude mid-and upper-troposphere, up to 1, 1.5, 3.5 and 5 K for the respective 4 periods, which correspond well with those in IPCC (2007, Fig. 10.7).The impact at the surface is smaller and shows more variability.Figure 15b-c shows the impact on the annual zonal mean temperature profile by CO 2 from road transport and maritime shipping or aviation.The CO 2 impact from road transport is much stronger than that from shipping or aviation.In the mid-to upper-troposphere it varies around 0. 05, 0.15, 0.35 and 0.5 K in 1980-1999, 2011-2030, 2046-2065 and 2080-2099, respectively.For shipping and aviation, there is no clear signal in 1980-1999, Fig. 13, Fig. 13.Time series of the annual global mean surface air temperature over the period 1860-2100: black (scaling 1), red (scaling 1, 11-yr running average), blue (scaling 5), and green (scaling 5, 11-yr running average).latitude mid-and upper-troposphere, up to 1, 1.5, 3.5 and 5 K for the respective 4 periods, which correspond well with those in IPCC (2007, Fig. 10.7).The impact at the surface is smaller and shows more variability.Figure 15b-c shows the impact on the annual zonal mean temperature profile by CO 2 from road transport and maritime shipping or aviation.The CO 2 impact from road transport is much stronger than that from shipping or aviation.In the mid-to upper-troposphere it varies around 0.05, 0.15, 0.35 and 0.5 K in 1980-1999, 2011-2030, 2046-2065 and 2080-2099, respectively.For shipping and aviation, there is no clear signal in 1980-1999, 2011-2030 and 2046-2065 (although an indication of a lowlatitude mid-to upper-troposphere temperature increase appears).In 2080-2099 this impact goes up to 0.15 K in the tropical upper troposphere.
The non-CO 2 impacts for the transport sectors are shown in Fig. 15d-g, with both non-CO 2 and non-CO 2 * approaches shown for aviation.We observe a positive signal in 2011-2030 and 2045-2065 for road transport, and a smaller impact at the end of the 21st century.For maritime shipping, there ened over, 90  2011-2030 and 2046-2065 (although an indication of a lowlatitude mid-to upper-troposphere temperature increase appears).In 2080-2099 this impact goes up to 0.15 K in the tropical upper troposphere.
The non-CO 2 impacts for the transport sectors are shown in Fig. 15d-g, with both non-CO 2 and non-CO 2 * approaches shown for aviation.We observe a positive signal in 2011-2030 and 2045-2065 for road transport, and a smaller impact at the end of the 21st century.For maritime shipping, there is a distinctive negative impact over the whole depth of the troposphere which is strongest in 2080-2099.The non-CO 2 impact from aviation is clearly positive in the troposphere, and in contrast with that from road transport and maritime shipping, it is very asymmetric being much stronger in the NH with a maximum around 300 hPa at 30-60 • N. Here, the temperature increases by more than 0.5 K in 2100.Using the non-CO 2 * approach the general pattern is similar, although the heating is weaker and one observes a strengthened cooling around 200 hPa between 50 and 80 • N.Moreover, a negative signal appears at the surface between 60 and 90 • N. Using an AGCM coupled to a slab ocean model and the emission scenario Fa, Ponater et al. (2005) found a maximum zonal mean temperature impact from aviation for the year 2050 of 0.35 K around 40-50 • N at 300-400 hPa.

Precipitation and cloud cover
Time series of cloud cover and precipitation can be found in Fig. 9c-f.Taking into account the total anthropogenic impact, the precipitation rises by 0.2 mm day −1 in 2100, while the cloud cover decreases by 1.8 %, coinciding with the increase in global mean surface air temperature.A similar relationship is seen for individual transport sectors, although  1980-1999, 2011-2030, 2046-2065 and 2080-2099.The order of the plots is as in Fig. 10.Notice that the contour intervals are 5 times larger for the total anthropogenic perturbation (a).(65.1 • N, 22.7 • W) normalized December-March average sea level pressure anomalies, with 1971-2000 as reference.One can observe a weak positive trend.Figure 17b shows the mean impact over the period 2050-2099 from the different transport sectors on the NAO index.The CO 2 impact is positive, road impact being around 20 % of the total anthropogenic one, and shipping or aviation being 5 %.For the non-CO 2 and non-CO 2 * cases, we find small impacts for shipping and road transport, but a relatively strong one for aviation, both for the non-CO 2 and non-CO 2 * approaches.

Ocean
In this section we describe the impact from the different transport sectors on ocean temperature, sea level, MOC, sea-ice extent and the Niño 3.4 index.

Ocean temperature
Figure 18 shows zonally averaged impacts on ocean temperature to a depth of 4000 m for the same periods as for the atmosphere, i.e. 1980-1999, 2011-2030, 2046-2065 and 2080-2099.Figure 18a shows the total anthropogenic impact.For all periods one observes a heating of the southern ocean, and the strongest impact can be seen in the southern mid-latitudes where the heating signal easily penetrates down to 3500-4000 m.At low latitudes, one sees initially (1980-1999 and 2011-2030) a cooling of the waters between 300 and 2000 m, probably caused by reduced mixing in response to more stable temperature profiles.Figure 18b-g shows the impacts from the transport perturbations: the CO 2 impact from road transport is the most important, while the non-CO 2 emissions from shipping have a cooling impact.The non-CO 2 effects of aviation have their greatest impact in the NH.
Figure 19 shows the zonally averaged increase in the heat content in the ocean integrated over the whole depth of the ocean (in 10 16 J m −1 ).The total anthropogenic impact (Fig. 19a) shows two characteristic features: most of the heat is stored in the southern ocean, while almost no heat is stored north of 45 • N, and the increase is very regular over the 4 periods.The CO 2 impacts from road transport, shipping or aviation present similar features.For the non-CO 2 impact, we notice the small values for road transport, the negative impact for shipping, and the large heating between 25 and 50 • N for aviation in the non-CO 2 approach, larger than that of the non-CO 2 * approach concentrated between 30 and 50 • N.
Time series of the global mean ocean temperature can be seen in Fig. 20a-b.One notices that these time series are rather smooth and show less inter-annual variability than the time series of the surface air temperature (see Fig. 9a-b).The non-CO 2 impact from shipping is negative, and prevails over the non-CO 2 impact from aviation and the CO 2 impact from shipping or aviation -a fact which is not clear for the surface air temperature evolution.Another interesting feature is the significant difference for shipping between the non-CO 2 and non-CO 2 * approaches in the second half of the 21st century.

Sea level
Sea level rise in a changing climate results from thermal expansion of the oceans as well as from melting of glaciers, ice caps and the Greenland and Antarctic ice sheets.shows the sea level rise (in mm) that includes only the thermal expansion of the oceans calculated from the standard expression for the density as a function of pressure, temperature and salinity in Millero and Poisson (1981).The total anthropogenic induced rise in sea level is 30 mm in 2000 and increases to 180 mm in 2100.The CO 2 impact from road is around 18 mm in 2100.The strongest non-CO 2 impact re- for cloud cover the evolution is rather noisy.Largest impacts can be observed for CO 2 from road transport, showing in 2100 a 0.015 mm day −1 increase in precipitation and a 0.15 % decrease in cloud cover.All other impacts are below 0.01 mm day −1 for precipitation and 0.1 % for cloud cover.
To get an idea of the zonal mean distributions, we show the total anthropogenic impact in Fig. 16.The cloud cover remains almost unaffected in the tropics, decreases considerably in the subtropics, and has a strong increase at northern high-latitudes.These characteristics correspond well with those of IPCC (2007, Fig. 10.10).There is generally less confidence in the changes in precipitation but increases in precipitation at high latitudes are very consistent among models, and can also be observed in Fig. 16b.Our model also shows a decrease in subtropical regions.
In Fig. 12b-c, we show the correlation between the patterns of precipitation or cloud cover with the total anthropogenic impact in 2080-2099 (as for the surface air temperature).The results are similar to those for surface air temperature but in general the correlations are weaker, e.g. the non-CO 2 impact from shipping shows a strong anti-correlation only for the period 2080-2099.The behavior of the non-CO 2 impact from aviation is very different: there is no correlation neither for cloud cover nor for precipitation.

North Atlantic Oscillation
Figure 17a shows time series of the North Atlantic Oscillation (NAO) index over the period 1860-2100 taking into account the total anthropogenic impact.NAO anomalies represent the Lisbon (38.7 • N, 9.1 • W) minus Reykjavik (65.1 • N, 22.7 • W) normalized December-March average sea level pressure anomalies, with 1971-2000 as reference.One can observe a weak positive trend.Figure 17b shows the mean impact over the period 2050-2099 from the different transport sectors on the NAO index.The CO 2 impact is positive, road impact being around 20 % of the total anthropogenic one, and shipping or aviation being 5 %.For the non-CO 2 and non-CO 2 * cases, we find small impacts for shipping and road transport, but a relatively strong one for aviation, both for the non-CO 2 and non-CO 2 * approaches.(65.1 • N, 22.7 • W) normalized December-March average sea level pressure anomalies, with 1971-2000 as reference.One can observe a weak positive trend.Figure 17b shows the mean impact over the period 2050-2099 from the different transport sectors on the NAO index.The CO 2 impact is positive, road impact being around 20 % of the total anthropogenic one, and shipping or aviation being 5 %.For the non-CO 2 and non-CO 2 * cases, we find small impacts for shipping and road transport, but a relatively strong one for aviation, both for the non-CO 2 and non-CO 2 * approaches.

Ocean
In this section we describe the impact from the different transport sectors on ocean temperature, sea level, MOC, sea-

Ocean
In this section we describe the impact from the different transport sectors on ocean temperature, sea level, MOC, seaice extent and the Niño 3.4 index.

Ocean temperature
Figure 18 shows zonally averaged impacts on ocean temperature to a depth of 4000 m for the same periods as for the atmosphere, i.e. 1980-1999, 2011-2030, 2046-2065 and 2080-2099.Figure 18a shows the total anthropogenic impact.For all periods one observes a heating of the southern ocean, and the strongest impact can be seen in the southern mid-latitudes where the heating signal easily penetrates down to 3500-4000 m.At low latitudes, one sees initially (1980-1999 and 2011-2030) a cooling of the waters between 300 and 2000 m, probably caused by reduced mixing in response to more stable temperature profiles.Figure 18b-g shows the impacts from the transport perturbations: the CO 2 impact from road transport is the most important, while the non-CO 2 emissions from shipping have a cooling impact.The non-CO 2 effects of aviation have their greatest impact in the NH.
Figure 19 shows the zonally averaged increase in the heat content in the ocean integrated over the whole depth of -1 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 K Fig. 18.Zonally averaged annual mean impact on the ocean temperature (in K) for the four different reference periods.The order of the plots is as in Fig. 10. .19. Zonally integrated heat uptaken by the ocean (in 10 16 J m −1 ) in 1980-1999 (green), 2011-2030 (blue), 2046-2065 (red) an 0-2099 (black).The order of the plots is as in Fig. 11. the ocean (in 10 16 J m −1 ).The total anthropogenic impact (Fig. 19a) shows two characteristic features: most of the heat is stored in the southern ocean, while almost no heat is stored north of 45 • N, and the increase is very regular over the 4 periods.The CO 2 impacts from road transport, shipping or aviation present similar features.For the non-CO 2 impact, we notice the small values for road transport, the negative impact for shipping, and the large heating between 25 and 50 • N for aviation in the non-CO 2 approach, larger than that of the non-CO 2 * approach concentrated between 30 and 50 • N.
Time series of the global mean ocean temperature can be seen in Fig. 20a-b.One notices that these time series are rather smooth and show less inter-annual variability than the time series of the surface air temperature (see Fig. 9a-b).The non-CO 2 impact from shipping is negative, and prevails over the non-CO 2 impact from aviation and the CO 2 impact from shipping or aviation -a fact which is not clear for the surface air temperature evolution.Another interesting feature is the significant difference for shipping between the non-CO 2 and non-CO 2 * approaches in the second half of the 21st century.

Sea level
Sea level rise in a changing climate results from thermal expansion of the oceans as well as from melting of glaciers, ice caps and the Greenland and Antarctic ice sheets.Figure 20c-d shows the sea level rise (in mm) that includes only the thermal expansion of the oceans calculated from the standard expression for the density as a function of pressure, temperature and salinity in Millero and Poisson (1981).The total anthropogenic induced rise in sea level is 30 mm in 2000 and increases to 180 mm in 2100.The CO 2 impact from road is around 18 mm in 2100.The strongest non-CO 2 impact results from shipping, causing a 3 mm decrease in 2000 and almost 8 mm in 2100.The CO 2 impact from shipping or aviation and the non-CO 2 impact from aviation are similar, leading to a rise in sea level of around 4 mm in 2100.The non-CO 2 * approach for aviation shows almost no impact.
IPCC ( 2007) report a sea level rise due to thermal expansion of between 130 and 320 mm (anthropogenic impact in 2090-2099 relative to 1980-1999, scenario A1B), and between 210 and 480 mm for the total sea level rise.This indicates that thermal expansion contributes to more than 50 % of the total rise in sea level, and shows that our results are at the low end of their range.Sausen and Schumann (2000) estimated the rise in sea level due to emissions from aviation.Their results were obtained using an SCM (see Sect. 3.3), whose calibration was based on an AOGCM that accounted for the rise in sea level due only to thermal expansion.They included the impact from melting ice-sheets by almost doubling the parameter responsible for the rise in sea level in the SCM.They found in 2100 increases in sea level by 6.5, 11.2 and 18.6 mm for their scenarios Fa1 (that best corresponds with our aviation emission scenario A1B), Eab and Eah, respectively, compared with our values of between 3.5 and 8.5 mm in 2100 due to thermal expansion only.

Meridional overturning circulation
Figure 20e shows the evolution of the MOC.It is calculated as the maximum of the depth integrated northward mass flux at 30 • N in the Atlantic Ocean.As our ocean model grid is not regular in the NH, the location of the integration varies slightly between around 33 and 27 • N. Compared with ocean temperature and sea level, there is a much stronger interannual variability.In our model, the MOC of the reference simulation between 1860 and 1970 varies mainly between 21 and 25 Sverdrup (Sv), values slightly higher than late 20th century observations.In 2100, the model predicts a decrease of the MOC down to 13-15 Sv for the total anthropogenic impact.Our modeled slowdown of the MOC corresponds well with results obtained in IPCC (2007).
Figure 20f shows the transport impact on the MOC.The CO 2 road emissions lead to a decrease of 0.8 Sv in 2100, corresponding roughly to 10 % of the total anthropogenic impact.The non-CO 2 impact from shipping is positive, while all the other impacts are rather small.

Sea-ice
Figure 21a-d shows the impact on the sea-ice extent in the Arctic (in 10 12 m 2 ) at the end of the winter (March) and at the end of the summer (September) from total anthropogenic influence (left) and from the transport sectors (right).It indicates that the maximum extent of the sea-ice cover in March is not affected very much by the changing climate, with quite good agreement among the different simulations.In September, however, the sea-ice cover decays strongly in the second  ults from shipping, causing a 3 mm decrease in 2000 and lmost 8 mm in 2100.The CO 2 impact from shipping or viation and the non-CO 2 impact from aviation are similar, eading to a rise in sea level of around 4 mm in 2100.The on-CO 2 * approach for aviation shows almost no impact.IPCC ( 2007) report a sea level rise due to thermal expanion of between 130 and 320 mm (anthropogenic impact in 090-2099 relative to 1980-1999, scenario A1B), and beween 210 and mm for the total sea level rise.This inicates that thermal expansion contributes to more than 50 % f the total rise in sea level, and shows that our results are at he low end of their range.Sausen and Schumann (2000) estimated the rise in sea evel due to emissions from aviation.Their results were ob-from melting ice-sheets by almost doubling the parameter responsible for the rise in sea level in the SCM.They found in 2100 increases in sea level by 6.5, 11.2 and 18.6 mm for their scenarios Fa1 (that best corresponds with our aviation emission scenario A1B), Eab and Eah, respectively, compared with our values of between 3.5 and 8.5 mm in 2100 due to thermal expansion only.

Meridional overturning circulation
Figure 20e shows the evolution of the MOC.It is calculated as the maximum of the depth integrated northward mass flux at 30 • N in the Atlantic Ocean.As our ocean model grid is not regular in the NH, the location of the integration varies slightly between around 33 and 27 • N. Compared with ocean From all transport sectors, we see small impacts in March and large impacts in September.In addition, the CO 2 impact in September from road transport is very different before 2050 compared with after.This is due to the fact that in the perturbed simulation there is no sea-ice in March from 2050 onwards, and the sensitivity is therefore reduced.The actual impact is therefore expected to be larger than indicated.Similar sudden changes can be seen around 2070 for the CO 2 impact from shipping and aviation, and for the non-CO 2 impact from aviation.Figure 21e-h shows the evolution of the sea-ice volume (in 10 12 m 3 ) in the Arctic.Due to the total anthropogenic impact, the sea-ice volume decreases considerably in March in the 21st century, which was not the case for the sea-ice extent.Further, all transport sectors show the same reduced sensitivities in September at the end of the 21st century.

El Ni ño
In Fig. 22a   and 25 Sverdrup (Sv), values slightly higher than late 20th century observations.In 2100, the model predicts a decrease of the MOC down to 13-15 Sv for the total anthropogenic impact.Our modeled slowdown of the MOC corresponds well with results obtained in IPCC (2007).Figure 20f shows the transport impact on the MOC.The CO 2 road emissions lead to a decrease of 0.8 Sv in 2100, corresponding roughly to 10 % of the total anthropogenic impact.The non-CO 2 impact from shipping is positive, while all the other impacts are rather small.

Sea-ice
Figure 21a-d shows the impact on the sea-ice extent in the Arctic (in 10 12 m 2 ) at the end of the winter (March) and at the end of the summer (September) from total anthropogenic influence (left) and from the transport sectors (right).It indicates that the maximum extent of the sea-ice cover in March is not affected very much by the changing climate, with quite good agreement among the different simulations.In September, however, the sea-ice cover decays strongly in the second half of the 21st century, showing large variability within and 2050 compared with after.This is due to the fact t perturbed simulation there is no sea-ice in March fr onwards, and the sensitivity is therefore reduced.tual impact is therefore expected to be larger than i Similar sudden changes can be seen around 2070 fo impact from shipping and aviation, and for the non pact from aviation.Figure 21e-h shows the evoluti sea-ice volume (in 10 12 m 3 ) in the Arctic.Due to anthropogenic impact, the sea-ice volume decrease erably in March in the 21st century, which was no for the sea-ice extent.Further, all transport sectors same reduced sensitivities in September at the end o century.

El Ni ño
In Fig. 22a we show the anomaly in the Niño 3.4 month running average sea surface air temperature domain 5 • S-5 • N, 120-170 • W, Trenberth, 1997) fo erence simulations with the total anthropogenic im reference period is 1971-2000.One sees that the in Niño 3.4 index increases in the 21st century.Fi shows the impacts from transport on the Niño 3.4 in the 2050-2099 period, and for reference the tota pogenic impact is also indicated.The strongest imp from CO 2 from road transport, which contributes t 15 % of the total.The non-CO 2 impact from sh negative, whilst for aviation, the non-CO 2 * appro a considerably smaller impact than the non-CO 2 ap

Conclusions
For the period 1860-2100 (SRES scenario A1B f 2100), we have studied the impact of road transp itime shipping and aviation on climate with the atm ocean general circulation model CNRM-CM3.3.T of the first studies with an AOGCM where the main of the transport sectors are explicitly taken into acc performed a reference simulation which represents anthropogenic impact and several sensitivity simu estimate the impact of the transport sectors.We s estimated the impact from CO 2 and from non-CO 2 As non-CO 2 forcings, we included the impact of aerosols, contrails, CFC-12 and HFC-134a.Mo emission inventories we used were the ones generat the QUANTIFY project (http://www.ip-quantify.eu As a principal indicator of climate change we hav domain 5 • S-5 • N, 120-170 • W, Trenberth, 1997) for the reference simulations with the total anthropogenic impact.The reference period is 1971-2000.One sees that the anomaly in Niño 3.4 index increases in the 21st century.Figure 22b shows the impacts from transport on the Niño 3.4 index over the 2050-2099 period, and for reference the total anthropogenic impact is also indicated.The strongest impact is that from CO 2 from road transport, which contributes to around 15 % of the total.The non-CO 2 impact from shipping is negative, whilst for aviation, the non-CO 2 * approach gives a considerably smaller impact than the non-CO 2 approach.

Conclusions
For the period 1860-2100 (SRES scenario A1B for 2000-2100), we have studied the impact of road transport, maritime shipping and aviation on climate with the atmosphere ocean general circulation model CNRM-CM3.3.This is one of the first studies with an AOGCM where the main forcings of the transport sectors are explicitly taken into account.We performed a reference simulation which represents the total anthropogenic impact and several sensitivity simulations to estimate the impact of the transport sectors.We separately estimated the impact from CO 2 and from non-CO 2 forcings.As non-CO 2 forcings, we included the impact of O 3 , CH 4 , aerosols, contrails, CFC-12 and HFC-134a.Most of the emission inventories we used were the ones generated within the QUANTIFY project (http://www.ip-quantify.eu).
As a principal indicator of climate change we have looked at global mean impacts.In the reference simulation, we found an increase in annual global mean surface air temperature of around 0.8 K in 2000, reaching 3.0 K in 2100.In 2000, the CO 2 impact from all transport sectors together is of the order of 0.1 K.The emission of CO 2 from road transport contributes to a global mean warming of 0.3 K in 2100, while shipping and aviation each contribute to 0.1 K in 2100.The contribution of CO 2 from the transport sectors to the total anthropogenic temperature change increases from 12.5 % in 2000 to 16.7 % in 2100.
The non-CO 2 impact differs strongly among the different sectors.For road, this impact is largest between 2000 and 2050 (order of 0.1 K) becoming smaller at the end of the 21st century.The non-CO 2 impact from shipping is clearly negative reaching −0.1 K between 2050 and 2100, while from aviation it is positive but depends strongly on the treatment of the O 3 perturbations, reaching possibly 0.15 K in 2100.This indicates that during the period 1900-2100, the net impact of road transport on climate is positive and dominated by its CO 2 impact, the net impact of maritime shipping is mainly negative only becoming neutral at the end of the 21st century, while for aviation it is clearly positive and presumably dominated by its non-CO 2 emissions, even in 2100.
The use of an AOGCM also allows us to obtain geographical distributions of impacts.We observe an amplification of the surface air temperature signal at the poles, especially in the Arctic, both in the reference simulation, as well as in the sensitivity simulations.However, while for road and shipping the non-CO 2 impact on the surface air temperature is only slightly stronger in northern than in southern midlatitudes, the impact for aviation can be up to a factor of 5 stronger in the northern hemisphere.The geographical pattern of the non-CO 2 climate impact for road transport and shipping coincides well with the total anthropogenic impact, while for aviation it is different.We also found a strong impact from the aviation non-CO 2 forcing on the NAO index.
Focussing on the ocean, we see that for the total anthropogenic impact and most of the transport impacts, the strongest deep ocean heating is observed in the southern midlatitudes, while for aviation there is a significant response in the northern mid-latitudes.Further we find a rise in sea level due to thermal expansion in 2100 of 17.6 mm for road transport and 4.6 mm for maritime shipping and aviation.The rise due to non-CO 2 emissions in 2100 is of the order of 1 mm for road transport, −6.6 mm for maritime shipping and probably between −1.2 and 4.3 mm for aviation.This can be compared with a total anthropogenic impact in 2100 of around 180 mm.An overview of our principal results on surface air temperature and sea level rise can be found in Table 5.
In order to obtain detectable impacts, we used ensembles of 3 members together with a five-fold amplification of the  transport induced forcings.If the forcings had not been amplified, a larger and computationally more expensive ensemble would have been required.We found that the amplification of the forcing did not excessively disturb variables such as the surface air temperature.However, for quantities like sea-ice extent or sea-ice volume in the NH in September (a period when the sea-ice is very sensitive to variations in a non-linear way), the method has shown some limitations.Further, we also observed shortcomings in the model.The impact from BC has probably been underestimated, which affects the results for road transport but has very little effect for the other two sectors.We also observed a significant negative temperature response from aviation at high latitudes, probably caused by sulfate aerosols.It is unclear whether this impact is very realistic.We found values for the O 3 radiative forcing which were considerably smaller than those found in other studies for similar perturbations.Although the O 3 impact in the lower troposphere was well described using the extended linear ozone scheme, there were important differences in the upper troposphere and lower stratosphere for the impact from aviation.
The results presented here are obtained using only one AOGCM, and therefore the results should be interpreted with care.Impacts on modeled temperature are closely related to the climate sensitivity of the model, and the climate sensitivity is known to vary significantly among models (IPCC, 2007, Table S8.1).However, relative impacts of the transport sectors w.r.t. the total anthropogenic climate impact are probably rather robust.The emissions that are the basis for this study are based on the SRES scenario A1B storyline for GDP development.This scenario is only one out of more GDP scenarios (Nakicenovic et al., 2000).Moreover alternative assumptions for the implementation of fuel efficiency and emission factors for the same A1B storyline have been suggested (Lee et al., 2009;Eyring et al., 2005).
The modeling of impacts from transport would benefit from the description of ice supersaturation in the model, and the possibility to advect tracers which might be beneficial for the contrail parametrization, the extended O 3 parametrization, or for the inclusion of small scale chemistry effects (Cariolle et al., 2009).In addition, a more extensive chemistry, an explicit description of aerosol processes, and an integrated carbon cycle would further increase the model's capability to describe impacts from transport.This would allow a more coherent modeling and limit the dependency on results from other models.
Although 2100 was the time horizon for our simulations based on the SRES scenario A1B, climate change is clearly not stabilized at that time: one can still see strong trends in the CO 2 forcings and in the response of the surface air temperature and ocean temperature.It is clear that in 2100 the CO 2 impacts (total anthropogenic impact and individual transport sectors) are still increasing considerably, suggesting further important changes in the 22nd century.

Fig. 1 .
Fig. 1.Time series of the forcings taken into account in the model integrations over the period 1860-2100.The first row shows the evolution of (a) CO 2 and (b) CH 4 mixing ratios in the reference simulation.Row 2 until 5 show forcings and emissions from the transport sectors: (c) radiative forcing from CO 2 , (d) radiative forcing from CH 4 , (e) radiative forcing from CFC-12 and HFC-134a from road transport, (f) global annual fuel consumption, and emissions from (g) NO x , (h) BC, (i) OC and (j) SO 2 .

Fig. 1 .Fig. 1 .
Fig. 1.Time series of the forcings taken into account in the model integrations over the period 1860-2100.The first row shows the evolution of (a) CO 2 and (b) CH 4 mixing ratios in the reference simulation.Row 2 until 5 show forcings and emissions from the transport sectors: (c) radiative forcing from CO 2 , (d) radiative forcing from CH 4 , (e) radiative forcing from CFC-12 and HFC-134a from road transport, (f) global annual fuel consumption, and emissions from (g) NO x , (h) BC, (i) OC and (j) SO 2 .

Fig. 3 .
Fig. 3. Contribution by road transport (left), shipping (middle) and aviation (right) to the distribution of (a) NO x (pptv), (b) O 3 (ppbv), (c) sulfate (ppt), (d) BC (ppt), and (e) OC (ppt) in JJA 2000.These contributions are obtained by off-line CTM simulations: O 3 and NO x are averages over the p-Tomcat, LMDZ-INCA and Oslo-CTM2 models, while the aerosol fields are those of the LMDZ-AER model.

Fig. 3 .
Fig. 3. Contribution by road transport (left), shipping (middle) and aviation (right) to the distribution of (a) NO x (pptv), (b) O 3 (ppbv), (c) sulfate (ppt), (d) BC (ppt), and (e) OC (ppt) in JJA 2000.These contributions are obtained by off-line CTM simulations: O 3 and NO x are averages over the p-Tomcat, LMDZ-INCA and Oslo-CTM2 models, while the aerosol fields are those of the LMDZ-AER model.
Fig.4.Schematic representation of the different simulations and impacts (a general impact is represented by X in the ordinate).The black curve (box) denotes the simulation which takes into account all the anthropogenic forcings.The other coloured curves (boxes) represent simulations where the CO2 or non-CO2 emissions of one of the transport sectors are modified.The blue curve (box) denotes a simulation without CO2 emissions of one of the transport sectors, and the red curve (box) denotes a simulation with 5 times the CO2 emissions of the same sector.The green curve (box) indicates a simulation with 5 times the non-CO2 emissions from one transport sector, using the dynamical O3 approach.The purple curve (box) indicates a simulation with the same emissions but using the fixed O3 approach (see Sect. 2.2.4).A simulation with doubled CO2 concentration (grey box) is used to derive the climate sensitivity.

Fig. 5 .
Fig. 5. Schematic representation of the impact of ensemble size and amplification of the forcing on the signal-to-noise ratio.The response (y-axis) is represented as a function of the size of the forcing (x-axis) in (R) a reference simulation, (S1) a simulation including once the forcing from a transport sector, and (S5) a simulation including 5 times the forcing from that transport sector.The black vertical bars indicate the uncertainty on the experiments best estimates.Increasing the ensemble size reduces these uncertainties.Note that due to non-linearity, the best estimate based on an amplification of the forcing (red dot) might be different from the actual impact (grey S1 dot).

Fig. 5 .
Fig. 5. Schematic representation of the impact of ensemble size and amplification of the forcing on the signal-to-noise ratio.The response (y-axis) is represented as a function of the size of the forcing (x-axis) in (R) a reference simulation, (S1) a simulation including once the forcing from a transport sector, and (S5) a simulation including 5 times the forcing from that transport sector.The black vertical bars indicate the uncertainty on the experiments best estimates.Increasing the ensemble size reduces these uncertainties.Note that due to non-linearity, the best estimate based on an amplification of the forcing (red dot) might be different from the actual impact (grey S1 dot).

Fig. 5 .
Fig. 5. Schematic representation of the impact of ensemble size and amplification of the forcing on the signal-to-noise ratio.The response (y-axis) is represented as a function of the size of the forcing (x-axis) in (R) a reference simulation, (S1) a simulation including once the forcing from a transport sector, and (S5) a simulation including 5 times the forcing from that transport sector.The black vertical bars indicate the uncertainty on the experiments best estimates.Increasing the ensemble size reduces these uncertainties.Note that due to non-linearity, the best estimate based on an amplification of the forcing (red dot) might be different from the actual impact (grey S1 dot).

Fig. 6 .
Fig. 6.Time series of the O 3 evolution: impact from the transport sectors on the O 3 mixing ratio at 850 hPa in the (a) SH and (b) NH, and at 250 hPa in the (c) SH and (d) NH; total O 3 column (e) in the reference simulation and (f) impact from the transport sectors; (g) O 3 mixing ratio in the reference simulation at 50 hPa averaged over the 60-90 • S region in September and (h) over the 60-90 • N region in March.In(a), (b), (c), (d) and (f), full lines indicate the dynamical O 3 approach and dashed lines the fixed O 3 approach.In (e), (g) and (h), the three different lines (black, red and blue) represent the three different members of the ensemble.
investigated the impact of climate change on tropospheric chemistry and aerosols.Over the period 2000-2050, Wu et al. (2008) found that OH changes from climate

Fig. 6 .
Fig. 6.Time series of the O 3 evolution: impact from the transport sectors on the O 3 mixing ratio at 850 hPa in the (a) SH and (b) NH, and at 250 hPa in the (c) SH and (d) NH; total O 3 column (e) in the reference simulation and (f) impact from the transport sectors; (g) O 3 mixing ratio in the reference simulation at 50 hPa averaged over the 60-90 • S region in September and (h) over the 60-90 • N region in March.In(a), (b), (c), (d) and (f), full lines indicate the dynamical O 3 approach and dashed lines the fixed O 3 approach.In (e), (g) and (h), the three different lines (black, red and blue) represent the three different members of the ensemble.

Fig. 7 .
Fig. 7. Time series of annual global mean impact of O 3 , aerosols and contrails on TOA (a) shortwave, (b) longwave and (c) net radiative fluxes from road transport (green), maritime shipping (blue) and aviation (red).The impact from O 3 (from the fixed O 3 approach) is indicated by the dashed lines, while the impact from aerosols and contrails together is indicated by the full lines.

Fig. 7 .
Fig. 7. Time series of annual global mean impact of O 3 , aerosols and contrails on TOA (a) shortwave, (b) longwave and (c) net radiative fluxes from road transport (green), maritime shipping (blue) and aviation (red).The impact from O 3 (from the fixed O 3 approach) is indicated by the dashed lines, while the impact from aerosols and contrails together is indicated by the full lines.

FigureFig. 8 .
Figure9ashows time series of the global annual mean of the surface air temperature for the three members of the

Fig. 8 .
Fig. 8. Global maps of the annual mean impact on the net TOA forcing in the periods 1980-1999, 2011-2030, 2046-2065 and 2080-2099 by O 3 perturbations from (a) road transport, (b) maritime shipping and (c) aviation, by aerosols from (d) road transport and (e) maritime shipping, and (f) by aerosols and contrails from aviation.

Fig. 9 .
Fig. 9. Left: time series of annual global mean (a) surface air temperature, (c) precipitation, and (e) cloud cover over the period 1860-2100 taking into account the total anthropogenic forcing.The three different lines (black, red and blue) represent the three different members of the ensemble.Right: time series of impact on (b) surface air temperature, (d) precipitation, and (f) cloud cover by road transport, shipping and aviation, separately for their CO 2 , non-CO 2 and non-CO 2 * impact (the non-CO 2 * impact is indicated by the dashed lines).The thin black lines indicate the individual impact from each of the three members of the simulation, and the thick lines indicate the 11-year running average of the ensemble mean.

Fig. 9 .
Fig. 9. Left: time series of annual global mean (a) surface air temperature, (c) precipitation, and (e) cloud cover over the period 1860-2100 taking into account the total anthropogenic forcing.The three different lines (black, red and blue) represent the three different members of the ensemble.Right: time series of impact on (b) surface air temperature, (d) precipitation, and (f) cloud cover by road transport, shipping and aviation, separately for their CO 2 , non-CO 2 and non-CO 2 * impact (the non-CO 2 * impact is indicated by the dashed lines).The thin black lines indicate the individual impact from each of the three members of the simulation, and the thick lines indicate the 11-year running average of the ensemble mean.
t. a total climate change of around 3 K.To www.atmos-chem-phys.net/12

Fig. 10 .
Fig.10.Global maps of the annual mean surface air temperature impact in the periods1980-1999, 2011-2030, 2046-2065 and 2080-2099   caused by (a) the total anthropogenic forcing, the CO 2 forcing from (b) road transport and (c) shipping or aviation, by the non-CO 2 forcing from (d) road transport, (e) shipping and (f) aviation where we use the dynamical O 3 approach to take into account the effect of O 3 , and (g) by the non-CO 2 * impact from aviation (using the fixed O 3 approach).Notice that the contour intervals are 5 times larger for the total anthropogenic perturbation (a).

Fig. 12 .Fig. 13 .
Fig. 12.First row: correlation between the impact on the geographical distributions of annual mean (a) surface air temperature, (b) cloud cover and (c) precipitation from a perturbation in a certain period with the total anthropogenic impact of the period 2080-2099.The dotted lines show the result for the fixed O 3 approach.Second row: correlation between the two different non-CO 2 approaches (fixed O 3 versus dynamical O 3 approach) for (d) surface air temperature, (e) cloud cover and (f) precipitation.

Fig. 15 .
Fig.15.Impact on annual zonal mean temperature profile in the periods1980-1999, 2011-2030, 2046-2065 and 2080-2099.The order of the plots is as in Fig.10.Notice that the contour intervals are 5 times larger for the total anthropogenic perturbation (a).

Fig. 15 .Fig. 16 .
Fig.15.Impact on annual zonal mean temperature profile in the periods1980-1999, 2011-2030, 2046-2065 and 2080-2099.The order of the plots is as in Fig.10.Notice that the contour intervals are 5 times larger for the total anthropogenic perturbation (a).

Fig. 17 .
Fig. 17.(a) Mean NAO anomalies calculated from sea level pressures taking into account the total anthropogenic impact (the different colors indicate the 3 simulations of the ensemble).Single dots indicate yearly values, and lines show 11-yr running averages.(b) Impact of the transport sectors on the NAO anomaly averaged over 2050-2099.The first three bars show the non-CO 2 (black) and non-CO 2 * (blue) impacts, the following two the CO 2 impacts, and the last one the total anthropogenic impact.
and 2080-2099 (black) from total anthropogenic forcing.The thick lines indicate the best estimate, the thin lines indicate the 95 % confidence interval.

Fig. 17 .
Fig. 17.(a) Mean NAO anomalies calculated from sea level pressures taking into account the total anthropogenic impact (the different colors indicate the 3 simulations of the ensemble).Single dots indicate yearly values, and lines show 11-yr running averages.(b) Impact of the transport sectors on the NAO anomaly averaged over 2050-2099.The first three bars show the non-CO 2 (black) and non-CO 2 * (blue) impacts, the following two the CO 2 impacts, and the last one the total anthropogenic impact.

Fig. 17 .
Fig. 17.(a) Mean NAO anomalies calculated from sea level pressures taking into account the total anthropogenic impact (the different colors indicate the 3 simulations of the ensemble).Single dots indicate yearly values, and lines show 11-yr running averages.(b) Impact of the transport sectors on the NAO anomaly averaged over 2050-2099.The first three bars show the non-CO 2 (black) and non-CO 2 * (blue) impacts, the following two the CO 2 impacts, and the last one the total anthropogenic impact.

Fig. 18 .
Fig. 18.Zonally averaged annual mean impact on the ocean temperature (in K) for the four different reference periods.The order of the plots is as in Fig.10.

Fig. 20 .
Fig. 20.Left: time series of total anthropogenic impact on (a) the ocean mean temperature, (c) sea level and (e) meridional overturning circulation.The three different lines (black, red and blue) represent the three different members of the ensemble.Right: the impact by road transport, shipping and aviation, separately for their CO 2 , non-CO 2 and non-CO 2 * impact (the non-CO 2 * impact is indicated by the dashed lines) on (b) the ocean mean temperature, (d) sea level and (f) meridional overturning circulation.The thin black lines indicate the individual impact from each of the three members of the simulation, and the thick lines indicate the 11-yr running average of the ensemble mean.
Left: time series in the reference simulation of (a) NH sea-ice coverage (in 10 12 m 2 ) in March, (c) NH sea-ice coverage in September, e) NH sea-ice volume (in 10 12 m 3 ) in March and (g) NH sea-ice volume in September.Right: the impact by road transport, shipping and viation, separately for their CO 2 and non-CO 2 impact on (b) NH sea-ice coverage in March, (d) NH sea-ice coverage in September, (f) NH ea-ice volume in March and (h) NH sea-ice volume in September.

Fig. 21 .
Fig. 21.Left: time series in the reference simulation of (a) NH sea-ice coverage (in 10 12 m 2 ) in March, (c) NH sea-ice coverage in September, (e) NH sea-ice volume (in 10 12 m 3 ) in March and (g) NH sea-ice volume in September.Right: the impact by road transport, shipping and aviation, separately for their CO 2 and non-CO 2 impact on (b) NH sea-ice coverage in March, (d) NH sea-ice coverage in September, (f) NH sea-ice volume in March and (h) NH sea-ice volume in September.

Fig. 22 .
Fig. 22. (a)Evolution of the Niño 3.4 index anomaly over the 1860-2100 period for the 3 simulations forced by the total anthropogenic forcing (reference value calculated as the ensemble mean over.(b) Mean impact on the Niño 3.4 index of the different transport sectors over 2050-2099: the first three bars indicate the non-CO 2 impact (black for the dynamic O 3 approach, blue for the fixed O 3 approach), the following two bars the CO 2 impact, and the last bar the total anthropogenic impact.

Fig. 22 .
Fig. 22. (a)Evolution of the Niño 3.4 index anomaly over the 1860-2100 period for the 3 simulations forced by the total anthropogenic forcing (reference value calculated as the ensemble mean over.(b) Mean impact on the Niño 3.4 index of the different transport sectors over 2050-2099: the first three bars indicate the non-CO 2 impact (black for the dynamic O 3 approach, blue for the fixed O 3 approach), the following two bars the CO 2 impact, and the last bar the total anthropogenic impact.

Table 4 .
Time series of mixing ratios of GHGs and total inorganic chlorine as used in the reference simulation.
ith the same emissions but using the fixed O 3 approach (see Sect. 2.2.4).A to derive the climate sensitivity.

Table 5 .
Global mean surface air temperature change and sea level rise for selected years.The values are 11-yr means around the year mentioned, and averaged over the different members of the ensemble.The non-CO 2 * forcing is as the non-CO 2 forcing except for O 3 where prescribed O 3 perturbation fields are used (see Sects.2.2.4 and 3.1).