Emission data are a key contribution to the Coupled Model Intercomparison Project Phase 6 CMIP6;. This framework provides both historical emissions and emissions for future scenarios . Apart from their usage within the CMIP itself, several studies have also used the aviation emissions provided within the framework of CMIP6 for other purposes, as they present an available data set with future projections that are consistent with those of other sectors e.g.,. The geographical and annual distributions of aviation emissions are identical throughout all historical and scenario data sets in CMIP6, leaving only the total annual emission amounts as variables that are different for each year and each scenario. According to the documentation , the geographical distribution of the CMIP6 aviation emissions is based on that of the CMIP5 aviation emissions , which in turn are derived from the Future Aviation Scenario Tool FAST; for the European QUANTIFY project , and is not affected by the regridding performed within CMIP6 . Based on this information we would expect an identical geographical distribution of the aviation emissions in CMIP5 and CMIP6.

Here, we report on an inconsistency in the spatial pattern of aviation emissions between CMIP5 and CMIP6 . The latitudinal emission distribution differs by an approximate factor of 1.344cos⁡ϕ for historic emissions provided in the Community Emissions Data System CEDS; and by an approximate factor of 1.912cos⁡2ϕ for the Shared Socioeconomic Pathway (SSP) scenarios , where ϕ is the latitude (Fig. ). This difference is particularly noticeable in the northern polar region, where emissions are several times larger in the CMIP6 data sets, but in terms of total amount of emissions, the difference is largest in the regions from ∼50 to ∼65∘ N and from ∼25 to ∼40∘ N, where most emissions occur (see also Fig. ). A comparison with an independent aviation emission inventory derived in the REACT4C project gives a very good match with the CMIP5 data set. The difference is observed for aviation emissions of nitrogen oxides (NOx), black carbon (BC), and CO2. Other emitted species (CO; NH3; non-methane volatile organic compounds, NMVOCs; SO2; organic carbon) have an identical geographic distribution to that of NOx and BC in CMIP6 but were not provided in the CMIP5 data. The factors by which the latitudinal emission distributions differ are independent of the year, as the geographical pattern of emissions is also constant over time in CMIP6. To obtain aviation emissions that have the same total amount of emissions as in CMIP6 but exhibit approximately the same latitudinal distribution of emissions as the CMIP5 emissions, one has to multiply the CMIP6 CEDS historic emissions (until the year 2014) of all species by 1.344cos⁡ϕ and the CMIP6 SSP scenario emissions (from the year 2015) of all species by 1.912cos⁡2ϕ. The parameters 1.344 and 1.912 originate from the latitudinal distribution of aviation emissions and ensure that the total amount of emissions is not modified.

The aim of this paper is to investigate the impacts of the differences in the latitudinal distribution of emissions on aviation-induced ozone and aerosols and on their radiative forcing (RF). We do not consider emissions of CO2, as it is a well-mixed greenhouse gas with a long lifetime, implying that the spatial distribution of the emissions has a minor effect on the CO2-induced climate effect. We also do not consider the potential differences in the contrail climate effect because the CMIP data do not contain data on flight distance per area, which would be required for their computation .

In Sect. 2, we introduce the used earth system model and simulation setup, and in Sect. 3, we present results on the aviation-induced atmospheric ozone concentration and aerosol distributions and differences in radiative fluxes.

To investigate the effect of the difference in latitudinal distribution of aviation emissions on ozone, aerosols, and the related radiative forcings, we perform simulations with the chemistry–climate model ECHAM/MESSy Atmospheric Chemistry (EMAC), using each of the geographical distributions, but identical total amounts of emissions. For the investigation of aviation-induced ozone changes, we perform a set of 5-year simulations, where the model is configured as a quasi-chemical transport model (QCTM) and uses a source apportionment (tagging) method. For the investigation of the aviation-induced aerosol effect, we perform a separate set of 13-year simulations, where the model is configured as a chemistry–climate model with nudged meteorology.

The EMAC model is a numerical chemistry and climate simulation system that includes sub-models describing tropospheric and middle-atmospheric processes and their interaction with oceans, land, and human influences . It uses the second version of the Modular Earth Submodel System (MESSy2) to link multi-institutional computer codes. The core atmospheric model is the fifth-generation European Centre Hamburg general circulation model ECHAM5;. The physics subroutines of the original ECHAM code have been modularized and reimplemented as MESSy sub-models and have been continuously further developed. Only the spectral transform dynamical core, the flux-form semi-Lagrangian large-scale advection scheme, and the nudging routines for Newtonian relaxation remain from ECHAM.

For the simulations of aviation-induced ozone changes in the present study, we applied EMAC (MESSy version 2.54.0.3) in T42L90MA resolution, i.e., with a spherical truncation of T42 (corresponding to a quadratic Gaussian grid of approximately 2.8∘×2.8∘ in latitude and longitude) with 90 vertical hybrid pressure levels up to 0.01 hPa. The applied model setup comprised RF calculations based on the sub-model RAD and the sub-model TAGGING version 1.1; for the attribution of RF to emissions from the aviation sector . The simulations use specified dynamics, and the setup is very similar to the one of the simulation RC1SD-base-10a described in detail by . The gas-phase mechanism is implemented using the sub-model Module Efficiently Calculating the Chemistry of the Atmosphere MECCA; and incorporates the chemistry of ozone, methane, and odd nitrogen. Alkanes and alkenes are considered up to C4, while the oxidation of C5H8 and some non-methane hydrocarbons (NMHCs) are described with the Mainz Isopren Mechanism version 1 . Further, heterogeneous reactions in the stratosphere sub-model MSBM; as well as aqueous-phase chemistry and scavenging SCAV; are included. Emissions of methane (CH4) are not considered explicitly. Instead, CH4 mixing ratios are relaxed towards observations using Newtonian relaxation with the sub-model TNUDGE . Using the sub-model LNOX, lightning NOx is parameterized after with total global emissions of ∼4.5 Tg(N)a-1, which is within the range given by . Emissions of NOx from soil and biogenic C5H8 emissions were calculated using the MESSy sub-model ONEMIS , using parameterizations based on for soil NOx and for biogenic C5H8.

For one simulation we use the unaltered CMIP6 aviation emissions of the SSP2 4.5 scenario for the year 2015, whereas for a second simulation we use the CMIP6 aviation emissions with their latitudinal distribution changed to be equal to that of the CMIP5 emissions. Other simulation settings are identical. The presented results are obtained as a 5-year mean after a spin-up period of 6 months in QCTM mode , where feedback between chemistry and dynamics is suppressed, and using meteorology data spanning from 2013 to 2017 and specified dynamics by Newtonian relaxation towards ECMWF ERA-Interim reanalysis data . For the spin-up period, we use meteorology data from the second half of 2012. The simulations were initialized from a previous 1.5-year simulation including TAGGING for the spin-up of the TAGGING tracers. This spin-up simulation itself was initialized from the long-term (since 1950) SC1SD-base-01 simulation, which is similar to the RC1SD-base-10a simulation .

For the simulations of the aviation-induced aerosol effect, we used EMAC with the aerosol sub-model MADE3 Modular Aerosol Dynamics model for Europe, adapted for global applications, third generation; in the configuration described by . With respect to the version adopted for the ozone changes, the EMAC–MADE3 setup for aerosol uses a lower vertical resolution with 41 layers, mostly covering the troposphere and the lower stratosphere, and a simplified chemistry scheme, only including the reactions relevant for the aerosol processes. The aerosol simulations cover a period of 13 years with nudged meteorology using the ECMWF ERA-Interim reanalysis data from 2006 to 2018 and again using emissions for the year 2015. The QCTM mode and the tagging method cannot be applied for investigating the aerosol effects due to the role of the cloud feedback and the complexity of the liquid-phase chemistry for sulfate, respectively. Hence, the statistical significance of the changes in the aerosol RF between the original and the corrected emission data set is evaluated using a paired sample t test at the 95 % confidence level.

Our analysis for the SSP scenarios shows that regional emission amounts from aviation differ substantially in the northern mid-latitudes (Fig. ). Emissions of NOx north of 45∘ N are 36.8 % lower, and emissions south of 45∘ N are 31.9 % higher when using the CMIP5 latitudinal distribution of emissions compared to when using the unaltered CMIP6 emissions. The mean emission latitude shifts from 41.3∘ N for the CMIP6 latitudinal distribution to 34.3∘ N for the CMIP5 latitudinal distribution.

The difference in the ozone distribution between the two QCTM simulations reflects the latitudinal difference in aviation emissions (dashed and solid lines in Fig. , respectively). However, atmospheric dynamics and the larger chemical activity in tropical latitudes lead to a southward shift in the ozone burden difference with respect to the emission difference. The increased NOx emissions southwards of 45∘ N cause a positive ozone burden whose value (2.13 Tg) is larger than the absolute value of the negative ozone burden caused by the decreased NOx emissions northwards of 45∘ N (-1.18 Tg). In total, using the CMIP5 latitudinal pattern of emissions increases the atmospheric ozone burden by 0.95 Tg, corresponding to an increase of 3.4 % in the total ozone burden attributed to aviation.

We also compute the stratospherically adjusted radiative flux at the tropopause resulting from these differences in the ozone concentration distribution. The pattern of the radiative flux difference between the two simulations closely follows the pattern of the ozone burden difference, but the radiative flux decrease at high northern latitudes is more pronounced than the corresponding ozone decrease (Fig. ). We show radiative flux instead of radiative forcing to keep all quantities in Fig.  independent of the area for better comparability. The radiative forcing attributed to aviation emissions is 30.82 mWm-2 in the simulation with unaltered CMIP6 emissions and 33.16 mWm-2 in the simulation using the CMIP5 emission pattern, corresponding to a difference of 2.34 mWm-2 or 7.6 %. The difference in total RF between the two simulations is 2.08 mWm-2. The total difference is smaller due to the non-linearity between nitrogen oxide emissions and ozone changes. Emissions from other sectors cause weaker radiative effects in a more polluted atmosphere, partly compensating for a larger aviation RF.

Transport emissions also influence the lifetime of methane, with aviation emissions generally leading to a lifetime decrease . In the simulation using the CMIP5 emission pattern, we found a 5.7 % larger decrease in lifetime for aviation. All these changes are statistically significant because they are 5 to 6 times larger than their standard deviation over the 5-year simulation period. For context, aviation emissions have consistently increased over time, with a decadal increase ranging from 10 %–25 %, depending on the time period . This points to the importance of accurately quantifying not only the magnitude and spatial distribution of aviation emissions but also their changes over time.

Finally, we investigate the impact of the corrected emissions on aviation-induced changes in aerosol concentrations. This is shown in Fig.  for BC and reveals that adopting the CMIP5 latitudinal distribution of emissions results in a lower aviation-induced BC concentration compared to the unaltered CMIP6 inventory. Differences around 0.1–0.2 ngm-3 are found throughout the troposphere at high latitudes (>40∘ N). Even larger differences, up to 0.5 ngm-3, can be seen at and below the typical cruise altitude (∼200–300 hPa). Analogous differences are found for aviation-induced aerosol number concentration (not shown). These changes can be relevant for the quantification of the impacts of aviation aerosol on climate, through their interactions with both warm clouds and cirrus e.g., as well as ice surface albedo . The large variability in the climate system associated with aerosol–cloud interactions, however, hampers a robust quantification of the aerosol RF from aviation. Here, we quantify a RF of -46.0 and -54.7 mWm-2 for the simulations with the CMIP5 and unaltered CMIP6 emissions, respectively. This means that the climate impact of aviation due to aerosol is reduced (in absolute terms) by 8.7 mWm-2, i.e., 16 %. This reduction is consistent with the aforementioned differences in aviation-induced particle concentrations, but its statistical significance is low (79.8 %) for the reasons outlined above.

The effect of the difference in emissions on air quality is small. For example, the differences in the surface mixing ratio of ozone and nitrogen oxides between the two QCTM simulations are smaller than 0.5 % and 2.1 % at all locations, respectively. For nitrogen oxides, 98 % of the 2.8∘×2.8∘ grid cells in the model exhibit differences smaller than 0.2 %.

In this study, we do not consider the effect that a different latitudinal distribution of aviation soot emissions would have on contrail RF. According to , a 50 % reduction in soot emissions could lead to a 14 % reduction in contrail RF. This implies a lower contrail RF north of 45∘ N and a higher contrail RF south of 45∘ N when using the CMIP5 latitudinal distribution of emissions compared to when using the unaltered CMIP6 emissions. The net effect may be a lower contrail RF due to the rarer occurrence of persistent contrails in tropical areas, but this effect is likely small due to the relatively small effect of soot emissions on contrail RF.

In summary, the inconsistency in the latitudinal distribution of aviation emissions between CMIP5 and CMIP6 leads to differences in not only the latitudinal distributions and regional emission amounts, but also the total amounts of resulting ozone changes, methane lifetime changes, and RF attributed to aviation. The usage of the CMIP6 latitudinal distribution of emissions leads to an overall lower climate effect of aviation emissions, even though the same total global amount of emissions was assumed in the simulations. The difference of 2.34 mWm-2 reported in this study for the SSP2 4.5 scenario is small in the context of anthropogenic climate change but constitutes 7.6 % of the RF attributed to aviation ozone in our model. We therefore recommend that scholars studying the effects of aviation emissions on ozone and climate consider the inconsistency in the latitudinal distribution of aviation emissions reported here. We also investigated the effect of the inconsistency on aerosol RF but could not detect a significant difference.

The impact of the inconsistency in the latitudinal distribution of aviation emissions on the RF and climate also depends on the background chemical composition of the atmosphere, which is a function of future global emission and pollution pathways. In a warmer and more polluted atmosphere, the chemical activity would be generally larger, particularly at high latitudes . Therefore, the negative ozone burden change at northern mid-latitudes and northern high latitudes would likely be closer to the positive change at tropical and southern latitudes, leading to a smaller net relative effect of the inconsistency in terms of ozone burden and RF. The opposite would be expected for a less polluted atmosphere.

Furthermore, the results emphasize the importance of a correct and realistic geographic distribution of emissions when studying their effects on atmospheric composition and climate. Future aviation emission data sets should also consider temporal changes in the spatial distribution of emissions. No spatial changes over time were incorporated in either the CMIP5 or CMIP6 aviation data sets because such changes have not been estimated by the research community. The spatial distribution of aviation emissions has certainly changed over time, however . For example, from 1990 to 2017 the share of estimated NOx emissions from flights originating in (roughly) the Northern Hemisphere (here former Soviet Union, Europe, China, and North America) declined from 73 % to 62 %, implying a shift in aviation emissions away from the northern mid-latitudes . Such shifts in the mean aviation emission latitude in the past and future have an impact on the climate effects of aviation NOx emissions, which many climate studies neglect.

We note in closing that the difference between CMIP5 and CEDS was found to be caused by an error in data pre-processing in CEDS and will be corrected in the next data release . This type of error can occur during conversion between masses and mixing ratios.