Radiative forcing due to changes in ozone and methane caused by the transport sector

-induced


Introduction
The climate impact of the transport sector occurs through emissions of CO 2 , aerosols (and their precursors), water vapour, and species that affect ozone and the oxidative state of the atmosphere such as NO x , CO and non-methane hydrocarbons (NHMC), with direct emissions of CH 4 being negligible for these sectors.This paper examines the impact of this latter class of oxidant emissions from three transport sub-sectors e land transport (ROAD), maritime shipping (SHIP) and aviation (AIR).We calculate the radiative forcing (RF) by considering both short-term and long-term changes in atmospheric composition.The estimated RF due to all emissions is positive for ROAD and AIR, but negative for SHIP (Fuglestvedt et al., 2008).This switch for SHIP is partly due to the strong direct and indirect aerosol effect from SHIP (Balkanski et al., 2010;Fuglestvedt et al., 2008) but also a result of the high NO x emissions which reduce the CH 4 lifetime.This coupling in the atmospheric chemistry between NO x , CH 4 , and O 3 is well established (Fuglestvedt et al., 1999;Lelieveld et al., 1998;Naik et al., 2005;Shindell et al., 2005Shindell et al., , 2009;;Wild and Prather, 2000).Emissions of the short-lived trace gas species NO x , non-methane hydrocarbons (NMHC), and CO lead to production of tropospheric O 3 , and we denote this as the short-term O 3 RF.The reduction in CH 4 also leads to additional changes in O 3 , but due to the long lifetime of CH 4 compared to the other O 3 precursors, this change in O 3 occurs on a longer time-scale than the short-term O 3 RF (Prather, 1994;Wild et al., 2001).We denote the change in O 3 from CH 4 changes as the CH 4 -induced O 3 change.
The RF from O 3 changes since pre-industrial time is estimated to be 0.35 W m À2 (Forster et al., 2007;Gauss et al., 2006) and the transport sectors have been estimated to contribute as much as a third of this value (Fuglestvedt et al., 2008).In terms of RF, the reduction in the CH 4 lifetime from NO x emissions acts on a global scale in opposition to the positive RF due to O 3 production from NO x , leading to a smaller net effect of NO x (Forster et al., 2007;Shindell et al., 2005;Wild et al., 2001).
Sector-specific analyses of RF are important, particularly when mitigation measures are being considered because each sector's impact is unique.This is especially relevant for the transport sectors where the emissions are introduced into quite different environments e road emissions are predominantly released into the polluted boundary layer, ship emissions are mostly released into the clean maritime boundary layer, and aircraft emissions are mostly released into the upper troposphere and lower stratosphere.Using several global atmospheric chemistry models and radiative transfer schemes, we quantify the RF resulting from changes in O 3 and the CH 4 lifetime due to emissions of NO x , CO and NMHCs from the different transport sectors.We focus on the sign of the net RF and its uncertainty for the three transport sectors.The results can be used, together with calculations of the forcing due to other emissions (e.g.CO 2 and black carbon) from the transport sector, to assess and quantify the overall climate impact of each sector (e.g.Skeie et al., 2009).
This paper presents results from the European Union project QUANTIFY (Quantifying the Climate Impact of Global and European Emission Systems).Hoor et al. (2009) reported an analysis of preliminary QUANTIFY simulations, focusing on the behaviour of the different chemical models; they also included a brief discussion of the RF from the different transport sectors.The simulations reported here use the final revised QUANTIFY emission inventories, which were developed during the course of the project (Uherek et al., 2010).We concentrate on a more detailed analysis of the resulting RF and its associated uncertainty and on the computation of climate emissions metrics (Global Warming Potential (GWP) and Global Temperature change Potential (GTP)) for aviation NO x emissions.An estimate of current climate importance of the accumulated emissions from the transport sector up to year 2000 can be made using RF.To assess the future importance of current emissions and for considerations for mitigation purposes, forward looking emission metrics such as the GWP and the GTP metrics are more useful as they account for the persistence of atmospheric perturbations and thus the integrated impacts (Fuglestvedt et al., 2010).

Methods and models
Five global chemistry models consisting of four Chemistry Transport Models (CTMs) and one Climate Chemistry Model (CCM) have been used to simulate changes in O 3 and OH due to precursor emissions (NO x , CO, and NMHC) from the different transport sectors with year 2000 emissions (see Table 1 for model descriptions).Each model ran with fixed CH 4 abundances and then adopted a spin-up period of a year in order to attain chemical steady state with respect to O 3 .The imbalance in the CH 4 budget is diagnosed; this allows the calculation of the change in CH 4 abundance that would have occurred over decades if the emissions were held constant.The QUANTIFY methodology (see Hoor et al., 2009 andGrewe et al., 2010 for details) is to include all natural and anthropogenic emissions and then to compute the impact of an individual transport sector by reducing each respective sector's emissions by 5%, so as to ensure that perturbations act on an atmosphere close to present-day composition (Grewe et al., 2010).The 5% results are scaled by a factor 20 to represent the total change attributable to the transport sectors.The difference between a 20 Â 5% reduction and a complete 100% reduction are discussed in Section 3.1.Fig. 1 compares the improved emissions used in this study compared to those of Fuglestvedt et al. (2008) and Hoor et al. (2009).ROAD emissions of NO x , NMHC and CO are highest in Fuglestvedt et al. (2008), whereas the new QUANTIFY NO x emissions for AIR are greater than those in Fuglestvedt et al. (2008) and Hoor et al. (2009).Only Fuglestvedt et al. (2008) included CO and NMHC emissions for the AIR sector, but these emissions are negligible in terms of global anthropogenic sources.Note the much higher ratio of NO x to CO 2 in SHIP and AIR emissions compared to ROAD emissions.The gridded emission data can be downloaded from www.ip-quantify.eu.
The RF calculations for the O 3 changes from the chemical models are performed using two sets of radiative transfer models, University of Oslo (UiO) and University of Reading (UoR).Except where stated, the simulations here use the UiO schemes, which consist of a broad-band scheme for thermal infrared radiation and a scheme using the multi-stream DISORT code for shortwave radiation (Myhre et al., 2000).UoR have performed O 3 RF calculations using the O 3 changes for each sector averaged over 4 of the global chemistry models (Oslo CTM2, TM4, LMDz-INCA and p-TOMCAT; the UCI results are a recent addition) using the Edwards Table 1 A short description of the CTMs and CCM used in this study.A more detailed overview of the models is given in Hoor et al. (2009).

Model acronym Institution Short description Reference
Oslo CTM2 University of Oslo, Norway CTM driven by ECMWF meteorology in T42 horizontal resolution and 60 vertical layers (Berglen et al., 2004;Gauss et al., 2006;Isaksen et al., 2005 (Folberth et al., 2006;Hauglustaine et al., 2004) UCI University of California, Irvine, USA CTM driven by ECMWF meteorology in T42 horizontal resolution and 37 vertical layers (Hsu et al., 2005;Wild et al., 2003) and Slingo (1996) two-stream radiation code, with 6 bands in the shortwave and 8 bands in the longwave.
The methodology for calculating the forcing due to CH 4 changes follows the simple approach described by Berntsen et al. (2005) and used in Hoor et al. (2009), with two important extensions.The chemistry models do not explicitly calculate the change in CH 4 concentrations and in any case the simulations are not long enough for the CH 4 to come into equilibrium with the changed OH field.Instead, the change in OH is used to estimate the fractional change in CH 4 lifetime.This is then multiplied by the present-day concentration of methane and a model-average feedback factor of 1.4 (Prather et al., 2001), to account for the impact of changes in CH 4 concentration on its own lifetime, to yield the fractional change in CH 4 concentration for steady-state conditions (Fuglestvedt et al., 1999).The QUANTIFY study did not directly calculate this feedback factor, nor the increase in tropospheric O 3 per change in CH 4 , so we resort to the published, model-average values here.
The RF is calculated assuming a specific CH 4 RF of 0.37 mW m À2 ppbv À1 , which assumes a background concentration of 1740 ppb.The CH 4 -induced O 3 RF is then computed, following Berntsen et al. (2005) and using results from Prather et al. (2001), whereby a 10% increase in CH 4 leads to a 0.64 DU increase in O 3 , and this O 3 has a specific RF of 42 mW m À2 DU À1 (Ramaswamy et al., 2001); this RF factor is more applicable to the global-scale change in tropospheric O 3 resulting from the methane change, than it is to the more regional short-term ozone change resulting from transport sector emissions.
The first extension to the Hoor et al. (2009) methodology is to account for the impact of CH 4 changes on stratospheric water vapour.Based on Myhre et al. (2007), we take the stratospheric water vapour RF to be 0.15 times that of the CH 4 RF.The second extension to the Hoor et al. (2009) methodology is to relax the assumption that the CH 4 concentration in 2000 is in steady state with that year's change in OH.The actual degree of imbalance depends on the history of change in OH, which is not accounted for in the chemical model calculations which used year 2000 emissions.The degree of imbalance will be greatest for AIR, for which the emissions have been growing most rapidly in recent years, and least for ROAD.The method and assumptions about the historical emissions are described in Grewe and Stenke (2008).The factor to correct this transient response in year 2000 is taken to be 0.85 for ROAD, 0.8 for SHIP and 0.65 for AIR.These factors are then applied to the CH 4 RF, the CH 4 -induced O 3 RF and the stratospheric water vapour RF from each of the CTMs.These corrections to the instantaneous year 2000 RF are a result of the historical emissions and do not apply to the calculation of GWP and GTP for AIR in Section 4.
Other consequences of the O 3 precursor emissions have been identified but are not considered here.For example, changes in O 3 and OH may alter the sulphate burden (e.g.Unger et al., 2006;Shindell et al., 2009) and surface O 3 interaction with vegetation may impact the carbon cycle (Sitch et al., 2007).NO x emissions have a small direct RF due to the absorption of solar radiation by NO 2 (Kvalevåg and Myhre, 2007) and formation of nitrate aerosols from the NO x enhances the overall aerosol negative RF (Forster et al., 2007).This paper focuses on the primary, well-established impacts of the ozone-precursor emissions on O 3 and CH 4 and their consequential effects.

Short-term O 3
Table 2 compares the global and annual mean total column O 3 change (in DU) for the five global chemistry models and the three transport sectors for the QUANTIFY preliminary emissions (Hoor et al., 2009) and QUANTIFY final emissions used here.The largest changes in the O 3 column between the simulations using the preliminary and final emissions are seen for ROAD and AIR.In addition to changes in the emissions inventories, there have also been model improvements and updates by many of the modelling groups over the two-year period in which the QUANTIFY emissions were updated.p-TOMCAT is the only model with a reduction in the O 3 column change for all three transport sectors between , respectively) for two previous studies (Fuglestvedt et al., 2008;Hoor et al., 2009) and in this study for ROAD (top), SHIP (middle) and AIR (bottom).For comparison, the CO 2 emissions (as CO 2 ) in year 2000 are 4200 Tg, 622 Tg and 677 Tg for ROAD, SHIP, and AIR; respectively.preliminary and final QUANTIFY simulations in spite of increased NO x emissions.
Fig. 2 shows the global and annual mean RF for short-term O 3 , using the UiO radiation model, for the three transport sectors calculated by the five global chemistry models.Amongst the transport sectors, ROAD is largest, followed by SHIP and AIR sectors in four of the models.The five model average for the ROAD, SHIP and AIR sectors are 31, 24 and 17 mW m À2 , respectively.A composite, four model (without UCI) mean of the O 3 change is used to compare the radiation models.For this subset the UiO RF values are 30, 21 and 16 mW m À2 for the three sub-sectors, and the UoR calculations yield corresponding forcings of 28, 19 and 17 mW m À2 indicating agreement to within about 10% in the RF calculations.
The relative spread in the RF among the models is within a factor of 2 for AIR, (13e21 mW m À2 ) and SHIP (17e32 mW m À2 ); by contrast for ROAD the spread in the RF is almost a factor of 3 (15e42 mW m À2 ).The model range in the sum of RF across all three sectors is less and ranges from 57 to 90 mW m À2 .
The zonal-mean RFs of short-term O 3 for the three transport sectors for all five models are shown in Fig. 3.The patterns of RF are similar for the four models Oslo CTM2, TM4, LMDz-INCA, and UCI, as governed by the global distribution of emissions for each sector, with ROAD dominating in the Northern Hemisphere, ROAD and SHIP being quite comparable in the Southern Hemisphere, and AIR being very small in the Southern Hemisphere.By contrast, the p-TOMCAT RF shows AIR as dominating in the Northern Hemisphere and comparably large also in the Southern Hemisphere.
The RF results are based on radiation calculations using the 100% perturbation in O 3 that is derived by multiplying the ozone change resulting from the 5% perturbation in emissions by 20 (Grewe et al., 2010;Hoor et al., 2009).Hoor et al. (2009) showed that the sum of O 3 changes from 5% perturbation in the individual ROAD, AIR and SHIP emissions equals that from a single calculation of the combined emissions from all three sectors.Here, we investigate the difference between a scaled 5% reduction and a full 100% reduction of ROAD emissions with the Oslo CTM2: the 100% reduction gives column O 3 changes that are 1.07 times greater than 20 Â 5%; however, the non-linearity in the longwave RF reduces this to a factor of 1.03.Compared to the inter-model differences in the RF, the non-linear response to the magnitude of the perturbations is relatively small, and hence contributes little to the overall uncertainty.
Normalized radiative forcing from O 3 (NRF) (RF divided by the change in the O 3 column expressed in Dobson Units (DU)) is dependent on both the region and altitude of the resulting O 3 change (Berntsen et al., 2000;Gauss et al., 2003).In all models, except LMDz-INCA, the AIR sector has a higher NRF than ROAD and SHIP with a mean of 36 mW m À2 DU À1 with a range from 33 to 42 mW m À2 DU À1 .The values for ROAD and SHIP are 32 mW m À2 DU À1 (range from 30 to 36 mW m À2 DU À1 ) and 30 mW m À2 DU À1 (range from 29 to 32 mW m À2 DU À1 ), respectively.For the sum of RF across the three transport sectors, p-TOMCAT yields the highest NRF (34 mW m À2 DU À1 ) and Oslo CTM2 the lowest (31 mW m À2 DU À1 ).These NRF numbers show the largest spread for AIR.In some instances the agreement in the RF between models is a coincidence resulting from compensating differences in the NRF and the ozone change (as is the case for p-TOMCAT and UCI for AIR).Nevertheless, these comparisons in NRF show much less relative spread compared to the RF calculations, indicating that it is inter-model differences in the total O 3 change that are mainly responsible for the differences in their RF, rather than inter-model differences in the distribution of the O 3 change.

CH 4
The O 3 precursor emissions change the atmospheric lifetime of CH 4 due to changes in OH concentrations.The percentage changes in the CH 4 lifetime due to destruction by OH (integrated between the surface and 50 hPa for each model) are given in Table 3.For most models and most transport sectors, increased emissions lead to an increase in global OH, a decrease in CH 4 lifetime and hence a decrease in CH 4 concentrations.The changes are largest for SHIP because of the different geographical pattern of emissions and background conditions as well as the mix of emitted components compared to the other sectors (Fuglestvedt et al., 2008;Hoor et al., 2009).Although AIR emissions have the largest impact on CH 4 on a per unit NO x emission basis, SHIP and ROAD each have nearly ten times larger NO x emissions.Unlike the results in Hoor et al. (2009), where all models (including p-TOMCAT) and all sectors showed a decrease in CH 4 lifetime, the current p-TOMCAT version using the final QUANTIFY emissions calculates a small increase in CH 4 lifetime for ROAD.Our analysis finds that p-TOMCAT has a much stronger response of the OH concentrations to emissions of CO and NMHC than the other models and this overwhelms the effect of the NO x emissions.This model has lower CO concentrations than the four other models, which can partly explain the different impact of the ROAD emissions.

Table 2
The short-term global and annual mean O 3 column change (in DU) for the five models and the year 2000 emissions from the three transport sectors for both the QUANTIFY preliminary emissions (Hoor et al., 2009)  The resulting RF for year 2000 due to CH 4 change from the transport sector (including the transient effect but not the CH 4induced changes in O 3 and stratospheric water vapour) is shown in Fig. 4.This CH 4 RF is larger for SHIP than ROAD and AIR.The intermodel difference in the CH 4 RF is rather small for SHIP, with a spread of only 8 mW m À2 for an average RF of À27 mW m À2 .For AIR, all models show a negative RF, with a mean of À7.3 mW m À2 and a relatively higher spread of 4.4 mW m À2 .The ROAD RF from CH 4 varies little among four global models, À12 to À14 mW m À2 , but p-TOMCAT gives þ1 mW m À2 , reducing the magnitude of the mean RF for ROAD for the CH 4 lifetime to À10 mW m À2 .

RF due to short-term O 3 and CH 4 changes combined
The net RF of short-term O 3 and long-term CH 4 combined (and now including the CH 4 impacts on O 3 and stratospheric water vapour) are shown in Fig. 5.All five models have a positive net RF for ROAD and AIR and a negative RF for SHIP.For ROAD, the net forcing varies from 9.3 to 21 mW m À2 , with a mean of 16 mW m À2 .For SHIP, the range is À12 to À25 mW m À2 with a mean of À18 mW m À2 .Although the general agreement in Fig. 5 is quite encouraging in terms of absolute RF, the degree of agreement is less    For AIR, the spread is from 4.1 to 8.7 mW m À2 with a mean of 6.0 mW m À2 .Compared to the multi-model mean, Oslo CTM2 is always more positive, by several 10's of percent, TM4 is always more negative by several tens of percent, while deviations from the mean for LMDz-INCA, UCI and p-TOMCAT are generally smaller and all are within 10% of the multi-model mean for two of the three sectors.
To derive the uncertainties in the RF, we use uncertainties in the emissions (Fuglestvedt et al., 2008), and simulations of distribution changes and radiative forcing performed in this study combined with Monte Carlo simulations (Boucher and Haywood, 2001).For O 3 and the CH 4 lifetime effect we use the difference in the global chemistry models as one standard deviation representative for the uncertainty in distributions for each of the transport sectors.The uncertainty in the CH 4 impact on O 3 takes into account uncertainty in the model change in CH 4 and O 3 in addition to the uncertainty in emissions.For stratospheric water vapour uncertainties in the emissions, CH 4 change and the radiative transfer are considered.The latter is taken from an intercomparison study with an uncertainty of 30% (Myhre et al., 2009).The derived uncertainties (one standard deviation) are 13 mW m À2 (81%), 10 mW m À2 (55%), and 5 mW m À2 (81%) for ROAD, SHIP, and AIR, respectively.The dominating contributor to the uncertainty is the short-term O 3 .et al. (2010) reported values (as well as a number of important caveats) for the GWP and GTP for NO x emissions from aviation using available results in the literature.They found substantial differences in the derived GWPs and GTPs, which included a difference in sign for some time horizons (H).The results derived here allow a cleaner comparison of GWPs and GTPs between global chemistry models, as they are derived using the same emissions and emission perturbations, the same experimental design and use the same radiative transfer scheme to calculate the RF.Thus the spread in results gives us a measure of the uncertainty in the global chemical modelling of aviation's impact on atmospheric chemistry.We do not present values for the other sectors, as the ozone changes for these are influenced by CO and NMHC emissions as well as NO x .

Fuglestvedt
The Fuglestvedt et al. (2010) methodology is adopted here for the calculation of GWPs and GTPs, although we also include the effect of CH 4 changes on stratospheric water vapour.In line with the conventional definitions of GWP and GTP, CO 2 is used as the reference gas.The GTP values are somewhat sensitive to the model used to calculate the temperature change, and the assumed climate sensitivity.Appendix 2 in Fuglestvedt et al. (2010) describes the method used here, which is based on fits to coupled ocean-atmosphere general circulation model experiments, which have an equilibrium climate sensitivity to a doubling of CO 2 of 3.9 K.The CH 4 lifetimes given in Table 3 are multiplied by 1.4 to give the CH 4 perturbation lifetimes (which accounts for the effect of a change in CH 4 on its own lifetime).Table 4 shows the specific RFs (in W m À2 (kgN yr À1 )) for the short-lived ozone, methane-induced ozone and methane for the 5 models, as well as the multi-model mean, which are required as input to the GWP and GTP calculations.
The resulting GWP values for three time horizons (H ¼ 20, 100 and 500 years) are shown in Table 5a, and for the GTPs (H ¼ 20, 50 and 100 years), in Table 5b.In both tables, the values are split into three components e the short-term O 3 effect resulting directly from the NO x emissions, the decadal CH 4 -induced O 3 change and the CH 4 change itself, which is scaled to include the stratospheric water vapour effect (which is 0.15 of the direct effect).The range of results given in Fuglestvedt et al. (2010) is also shown.Considering the GWP, the results indicate at first sight that the range from the present models e an indicator of the uncertainty in the GWP e is no smaller than that derived in Fuglestvedt et al. (2010).For H ¼ 20 years, there is a spread of a factor of 3; for H ¼ 100 and 500 years, the different models do not even agree in sign.For GTP, the situation is similar: for H ¼ 20 and 50 years, the spread is no less than in Fuglestvedt et al. (2010), while for H ¼ 100 years, there is disagreement in sign.
However, Table 5 reveals important patterns that give hope that the reasons for the model differences can be resolved.For all metrics and all time horizons, values from the Oslo CTM2 are clearly the most positive (as is true also for the net forcings for all sectors shown in Fig. 5); for GWP (H ¼ 20), GWP (H ¼ 100) and GTP (H ¼ 100) it is the only model generating a positive value.Removing the Oslo CTM2 model, markedly reduces the range in both the GWPs and GTPs.For example GWP (H ¼ 100) changes from À21 to 67, to À21 to À6.3; and GTP (H ¼ 100) changes from À5.8 to 7.9 to À5.8 to À4.6.Much less markedly, TM4 is the most negative for all GWP time horizons, and p-TOMCAT is generally the most negative for GTP (H ¼ 50) and GTP(H ¼ 100).
This behaviour seems largely a result of one characteristic of the models e the ratio of the percentage change in CH 4 lifetime to the O 3 column change.Using Tables 2 and 3, the values for AIR range from À1.7% DU À1 for Oslo CTM2, À2.8% DU À1 for TM4, À2.5% DU À1 for UCI, À2.7% DU À1 for LMDz-INCA to À3.0% DU À1 for p-TOMCAT.Hence, for the Oslo CTM2 model the compensating effect of the negative forcing due to CH 4 and its resulting effect on stratospheric water vapour and O 3 is smaller than for the other models.The differences in CH 4 lifetime between the models (Table 4) have only a small influence on the range of results.
If the reason why the models differ in ratio of CH 4 change to O 3 change can be understood, there is hope for a marked reduction in the inter-model range in the estimated net RF from transport and in metrics such as the GWP and GTP.However, as is clear in Table 5, even when models agree in the net value of a metric, the individual components contributing to this net value can be quite different e for example, UCI and p-LMDz-INCA agree well in the net for both metrics and most time horizons, but disagree significantly for the three components; this effect is traceable to the larger change in O 3 in UCI compared to LMDz-INCA (see Table 2).

Conclusions
We have investigated the RF for the year 2000 due to changes in O 3 and CH 4 caused by the transport sector, using five global chemistry models and two radiation models.We find the difference between ROAD, SHIP and AIR to be robust across all models.For the year 2000, this study reduces the CH 4 and CH 4 -induced O 3 impacts to account for the slower response of CH 4 perturbations to changes in OH.It also includes the effect of CH 4 changes on stratospheric water vapour.The results are also used to present values of GWP and GTP for AIR NO x emissions which are based on a range of global chemistry models adopting the same experimental design.Fuglestvedt et al. (2008) found, based on one global chemistry transport model and one radiative transfer model, a year 2000 RF for the combined effect of O 3 and CH 4 amounting to 42 mW m À2 for ROAD, À11 mW m À2 for SHIP and 12 mW m À2 for AIR, based on a different set of emissions (see Fig. 1).The multi-model means obtained here are 16 mW m À2 , À18 mW m À2 and 6.0 mW m À2 respectively.The ROAD and AIR are significantly smaller in the present analysis than in (Fuglestvedt et al., 2008), while SHIP is significantly more negative.Various factors explain the differences.First, the emissions are different.As a consequence, a lower ozone production (at least for ROAD with lower CO and NHMC emissions) and hence less net positive radiative forcing is calculated.Second, the secondary consequences of methane changes included here (i.e.responses in O 3 and stratospheric H 2 O) increase the negative radiative forcings.Third, the Oslo CTM2, which was used in the Fuglestvedt et al. (2008) study is seen here to produce results at the upper end of the spectrum, compared with the multi-model mean.All three factors act together to lower the calculated radiative forcing from the transport sectors compared to the values in Fuglestvedt et al. (2008).The RF for the combined effect of O 3 and CH 4 can be compared to the RF due to CO 2 which has previously been estimated to be 150 mW m À2 , 35 mW m À2 , 21 mW m À2 for ROAD, SHIP, and AIR, respectively (Fuglestvedt et al., 2008).
The results reported here with the QUANTIFY inventories can be compared with the previous Hoor et al. (2009) results using preliminary QUANTIFY inventories, where the multi-model means were 7.3 mW m À2 , À26 mW m À2 and 2.9 mW m À2 for ROAD, SHIP and AIR, respectively.For SHIP and AIR part of the difference is due to account being taken here of the lack of steady state of the CH 4 field with the changed OH field, which reduces the size of the CH 4 offset to the positive short-term O 3 forcing.The factor of more than two change for ROAD has a number of identified reasons (emissions, model updates, and method for CH 4 RF calculations); the short-term O 3 forcing is only changed by about 10% while the offset resulting from the changed OH field is now smaller.Based on the much higher emissions of CO and NMHC in Fuglestvedt et al. (2008) it is expected that the O 3 RF for ROAD was higher in that study compared to Hoor et al. (2009) and this study (see Fig. 1).On robustness and uncertainties for road transport, the O 3 RF is highly dependent on the background NO x emissions used in the model, including those from power generation, agriculture, lightning and biomass burning.
The inter-model absolute differences are smaller when the combined effects of O 3 and CH 4 RFs are calculated than they are for the, short-term O 3 RF alone.Nevertheless, the combined O 3 and CH 4 RF from individual models can deviate by many tens of percent from the multi-model mean.One significant factor in these differences, and the difference in the aviation climate emission metrics, is the ratio of the percentage change in CH 4 lifetime to the column O 3 change.If the underlying reasons for this ratio could be understood, there is the possibility of markedly decreasing the intermodel differences.4) and lifetimes (Table 3).All numbers are rounded, so that the net values may not be the sum of numbers as they are presented here.All values are on a per kg N basis and are relative to CO 2 .The GTP values are specific to a given value of climate sensitivity e see text for details.The RF resulting from the effect of transport-related emissions on ozone and methane reported here must be combined with estimates of the transport-related RF from aerosols and CO 2 (e.g.Balkanski et al., 2010) to improve understanding of the overall impact of the transport sector on climate.

Fig. 2 .
Fig. 2. Global and annual mean radiative forcing due to short-term O 3 (mW m À2 ) for year 2000 emissions from three transport sectors and five global chemistry models, as calculated using the UiO radiation schemes.

Fig. 3 .
Fig. 3. Zonal and annual mean radiative forcing (mW m À2 ) from short-term O 3 for five global chemistry models.Each panel shows results for year 2000 emissions from three transport sectors.
(yr) due to destruction by OH (between the surface and 50 hPa) for the base case and the relative changes due to year 2000 emissions from three transport sectors e the values are derived from a 5% change in emissions for each sector and then multiplied by 20.The feedback effect of changes in CH 4 on its own lifetime is not included in this

Fig. 4 .
Fig. 4. Global and annual mean radiative forcing for the year 2000 due to CH 4 changes (mW m À2 ) accounting for the time-history of the emissions (see text for details) for three transport sectors and five global chemistry models.

Fig. 5 .
Fig. 5. Net RF of O 3 and CH 4 (including stratospheric water vapour) in year 2000 accounting for the time-history of the emissions (see text for details) for three transport sectors and five global chemistry models.
(GWP) and Global Temperature change Potentials (GTP) for year 2000 AIR NO x emissions (a) GWP values for one-year pulse emissions of NO x for a 20, 100 and 500 year time horizons and (b) GTP values for 20, 50 and 100 years.The first three numbers show the individual contributions from the short-lived O 3 , the CH 4induced O 3 and the CH 4 (which includes stratospheric water vapour changes), respectively; the net GWP and GTP are shown in bold.The mean values use the multi-model means of the specific forcings (Table

(
and the QUANTIFY final emissions (this study).

Table 4
Steady-state radiative forcings (in W m À2 (kgN yr À1 )) for sustained AIR NO x emissions for each of the 5 models and the mean of the 5 models.In this table, for example, 1.94E-11 denotes 1.94 Â 10 À11 etc.