Dataset for the Environmental Research: Climate article by Im et al. (2025)

PAPER • OPEN ACCESS Future CH4 as modelled by a fully coupled Earth system model: prescribed GHG concentrations vs. interactive CH4 sources and sinks To cite this article: Ulas Im et al 2025 Environ. Res.: Climate 4 015008 View the article online for updates and enhancements. You may also like Broadening the scope of anthropogenic influence in extreme event attribution Aglaé Jézéquel, Ana Bastos, Davide Faranda et al. - Toward a process-oriented understanding of water in the climate system: recent insights from stable isotopes Adriana Bailey, David Noone, Sylvia Dee et al. - Analysing the development of the climate, land, energy, and water systems (CLEWs) modelling framework: a state-of-the-art review Kane Alexander, Naomi Tan, Francesco Gardumi et al. - This content was downloaded from IP address 130.225.21.26 on 04/11/2025 at 11:39 Environ. Res.: Climate 4(2025) 015008 https://doi.org/10.1088/2752-5295/adb3c0 OPEN ACCESS RECEIVED 15 November 2024 REVISED 31 January 2025 ACCEPTED FOR PUBLICATION 7 February 2025 PUBLISHED 25 February 2025 Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. PAPER Future CH4as modelled by a fully coupled Earth system model: prescribed GHG concentrations vs. interactive CH4sources and sinks Ulas Im1,∗, Kostas Tsigaridis2,3, Susanne Bauer3, Drew Shindell4, Dirk Olivié5, Simon Wilson6, Lise Lotte Sørensen1, Peter Langen1and Sabine Eckhardt7 1Department of Environmental Science/ Interdisciplinary Centre for Climate Change (iClimate), Aarhus University, Roskilde, Denmark 2Center for Climate Systems Research, Columbia University, New York, NY, United States of America 3NASA Goddard Institute for Space Studies, New York, NY, United States of America 4Nicholas School of the Environment, Duke University, Durham, NC, United States of America 5The Norwegian Meteorological Institute, Blindern, Oslo, Norway 6Arctic Monitoring and Assessment Programme (AMAP), Tromsø, Norway 7Norwegian Institute for Air Research (NILU), Kjeller, Norway ∗Author to whom any correspondence should be addressed. E-mail: [email protected] Keywords: methane, wetland emissions, Earth system modelling, future climate projections, anthropogenic methane emissions Abstract We have used the NASA Goddard Institute for Space Studies (GISS) Earth system model GISS-E2.1 to study the future budgets and trends of global and regional CH4under different emission scenarios, using both the prescribed GHG concentrations as well as the interactive CH4sources and sinks setup of the model, to quantify the model performance and its sensitivity to CH4sources and sinks. We have used the Current Legislation (CLE) and the maximum feasible reduction (MFR) emission scenarios from the ECLIPSE V6b emission database to simulate the future evolution of CH4sources, sinks, and levels from 2015 to 2050. Results show that the prescribed GHG version underestimates the observed surface CH4concentrations during the period between 1995 and 2023 by 1%, with the largest underestimations over the continental emission regions, while the interactive simulation underestimates the observations by 2%, with the biases largest over oceans and smaller over the continents. For the future, the MFR scenario simulates lower global surface CH4concentrations and burdens compared to the CLE scenario, however in both cases, global surface CH4and burden continue to increase through 2050 compared to present day. In addition, the interactive simulation calculates slightly larger O3and OH mixing ratios, in particular over the northern hemisphere, leading to slightly decreased CH4lifetime in the present day. The CH4forcing is projected to increase in both scenarios, in particular in the CLE scenario, from 0.53 W m−2in the present day to 0.73 W m−2in 2050. In addition, the interactive simulations estimate slightly higher tropospheric O3forcing compared to prescribed simulations, due to slightly higher O3mixing ratios simulated by the interactive models. While in the CLE, tropospheric O3forcing continues to increase, the MFR scenario leads to a decrease in tropospheric O3forcing, leading to a climate benefit. Our results highlight that in the interactive models, the response of concentrations are not necessarily linear with the changes in emissions as the chemistry is non-linear, and dependent on the oxidative capacity of the atmosphere. Therefore, it is important to have the CH4sources and chemical sinks to be represented comprehensively in climate models. 1. Introduction Atmospheric methane (CH4) is the second most important anthropogenic greenhouse gas with a global warming potential 27–30 times that of carbon dioxide (CO2) over a 100 year time horizon (Forster et al © 2025 The Author(s). Published by IOP Publishing Ltd Environ. Res.: Climate 4(2025) 015008 U Im et al 2021). CH4emissions mix through the troposphere on timescales shorter than the globally averaged atmospheric lifetime (AMAP 2015), thus the average trend in atmospheric concentration is roughly the same everywhere on Earth. While global concentrations have increased at varying rates, the mechanisms behind these variations are still not fully understood. Between approximately 2000–2005 there was no growth in global CH4mixing ratios but from about 2007, the growth increased and then accelerated from 2015 (Nisbet et al 2019) till today. CH4is emitted into the atmosphere from both anthropogenic activities (e.g. agriculture, energy, industry, transportation, waste management, and biomass burning) and natural processes (e.g. wetlands, termites, permafrost thawing and oceanic and geological processes), and it is removed from the atmosphere mainly by reaction with hydroxyl radical (OH) in the troposphere. The uncertainty in the chemical loss of CH4by OH is estimated around 15% (Saunois et al 2016). Depending on the approach used, total CH4emissions from natural and anthropogenic sources range between 575 and 669 Tg yr−1(Saunois et al 2024) and furthermore, top–down versus bottom-up estimates of CH4sources and sinks do not match, underscoring the incomplete knowledge of global CH4dynamics (Stavert et al 2022). Understanding and quantifying the global CH4budget is important for assessing realistic pathways to mitigate climate change. Recently, AMAP (2021) and von Salzen et al (2022) showed that ambitious global reductions of CH4, together with black carbon, could lead to Arctic climate benefits by 2050, similar to those from global CO2 reductions in a climate-focused mitigation strategy. A better understanding of the drivers of trends and variability in CH4abundance over the recent past is therefore critical for building confidence in projections of future CH4levels. In addition, a better understanding of CH4sinks is needed to better understand the potential to enhance the natural sinks for atmospheric CH4removal, and determine removal scales required (NASEM 2024). To date, important natural sources such as oceanic CH4(Weber et al 2019, Davies et al 2024), lakes (Zhuang et al 2023), and permafrost thawing (Turetsky et al 2020), and sinks such as tropospheric halogens (e.g. Basu et al 2022), and some VOCs such as isoprene (Guenther et al 2006, Arnold et al 2009)impact the atmospheric oxidative capacity, are either missing or outdated in the climate models participating in the recent IPCC (Forster et al 2021) and AMAP ( 2021) assessments. For example, halogens influence the atmosphere’s oxidative capacity by affecting the abundant hydroxyl radicals (OH) and thereby interfering with the CH4sink (Badia et al 2021). A significant atmospheric CH4sink is direct removal via chlorine (Cl) reactions (Basu et al 2022), and the indirect influence on OH via the catalytic destruction of O3by bromine (Br) and iodine (I) (Caram et al 2023). These processes are not accounted for in majority of the ESMs. Halogens can decrease tropospheric O3by 11%–15% and OH by 8% [9, 10] and increase global CH4lifetime and radiative forcing (Li et al 2022). To date, Earth system models (ESM), in most cases, have specified atmospheric CH4concentration to follow prescribed pathways, thereby disallowing important coupling mechanisms, such as coupling between global wetlands and atmospheric chemistry. For example, in all models contributing to the Coupled Model Intercomparison Project Phase 6 (CMIP6: Erying et al 2016), global mean CH4concentrations were prescribed at the surface, except for some future simulations with GISS-E2-1-G, the NASA Goddard Institute for Space Studies (GISS) chemistry–climate model version E2.1 (Kelley et al 2020), which used interactive online CH4emissions. Recently, Folberth et al (2022) implemented the interactive CH4chemistry in the UK-ESM1 ESM, where they have found a large negative bias between the 1920s and present-day global anthropogenic CH4. They proposed that this underestimation could be either a model bias in simulated processes (sinks) or an underestimate of historical emissions, therefore recommending better estimates and reconstructions of past CH4emissions. In addition, the emissions-driven interactive chemistry setup of the EMAC chemistry-climate model has been used by Stecher et al (2024) to perform idealized perturbation simulations driven either by increased carbon dioxide (CO2) mixing ratios, or by increased CH4emission fluxes. The CH4emission flux perturbation leads to a large increase of CH4mixing ratios due to increased atmospheric CH4lifetime. It is though important to account for sources and sinks of CH4 when assessing its impacts and future evolution (Shindell et al 2024). The aim of this study is to evaluate the performance of the GISS-E2.1 ESM in simulating observed atmospheric CH4concentrations in the near past using its interactive CH4chemistry, compare with the prescribed GHG simulations, and predict the future CH4concentrations under two emission scenarios in the frame of the AMAP SLCF activities. 2. Material and methods 2.1. GISS-E2.1 ESM GISS-E2.1 is the CMIP6 version of the GISS modelE ESM (Kelly et al 2020, Miller et al 2020). GISS-E2.1 has a horizontal resolution of 2◦in latitude by 2.5◦in longitude and 40 vertical layers extending from the surface to 0.1 hPa in the lower mesosphere. The tropospheric chemistry scheme used in GISS-E2.1 (Shindell et al 2 Environ. Res.: Climate 4(2025) 015008 U Im et al 2013) includes inorganic chemistry of Ox, NOx, HOx, CO, and organic chemistry of CH4and higher hydrocarbons using the CBM-IV scheme (Gery et al 1989), and the stratospheric chemistry scheme (Shindell et al 2013a), which includes chlorine and bromine chemistry together with polar stratospheric clouds. CH4is primarily oxidized by the hydroxyl radical (OH), which is sensitive to NOx, CO, and VOC levels. 2.2. Emissions 2.2.1. Anthropogenic emissions In this study, we used the ECLIPSE V6b (Höglund-Isaksson et al 2020) emissions, which has been developed with support of the EU-funded Action on Black Carbon in the Arctic (EUA-BCA) and used in the framework of the recent AMAP Assessment on Short Lived Climate Forcers (AMAP 2021, von Salzen et al 2022). The ECLIPSE V6b emissions dataset is a further evolution of the scenarios established in the EU funded ECLIPSE project (Stohl et al 2015, Klimont et al 2017). ECLIPSE V6b has a 0.5◦×0.5◦spatial resolution, and includes nine sectors: energy, industry, solvent use, transport, residential combustion, agriculture, open burning of agricultural waste, waste treatment, gas flaring and venting, and international shipping. A monthly pattern for each gridded layer was provided at a 0.5◦×0.5◦grid level. The ECLIPSE V6b dataset, used in this study, includes an estimate for 1990–2015 using statistical data and two scenarios extending to 2050 that rely on the same energy projections from the World Energy Outlook 2018 (IEA 2018) but have different assumptions about the implementation of air pollution reduction technologies, as described below. The future anthropogenic emissions (2015–2050) are defined in two scenarios. The Current Legislation (CLE) scenario assumes efficient implementation of the current air pollution legislation enacted before 2018, while the maximum feasible reduction (MFR) scenario assumes implementation of best available emission reduction technologies. The assumptions and the details for the CLE and MFR scenarios (as well as other scenarios developed within the ECLIPSE V6b family) can be found in Höglund-Isaksson et al (2020). 2.2.2. Natural emissions The natural emissions of sea salt, biogenic VOCs, dimethylsulfide (DMS), and dust are calculated interactively. CH4emissions from wetlands in GISS-E2.1 are calculated online using the model’s climate (Shindell et al 2003a, Shindell et al 2004). The emissions module calculates the response to GISS-E2.1 upper layer soil temperature and precipitation anomalies (Shindell et al 2004). The model uses the wetland locations from Fung et al (1991), which are then modified by the emissions model described above. Locations determined to contain wetlands which were not in the dataset were given the latitudinal mean emission value. In addition, the wetland emissions have been tuned by a tuning factor calculated to get a similar burden and lifetime as in the prescribed simulations during the historical period (1995–2023) in order to better follow the global observations. This tuning approach does not take into account the changes in changes in OH or regional differences in CH4emissions and concentrations but improves the model performance by getting closer to the observed global CH4concentrations. The calculated wetland tuning factor for the 1995–2014 period is then applied also in the future period 2015–2050. The prescribed simulations are based on the global CH4mixing ratios provided by Olivié (2021), which have been calculated using a simple box-model for the different scenarios that uses as input time series of globaland annual-mean emissions of CH4, NOx, CO and VOCs and of global-mean temperature, and calculates globaland annual-mean concentrations of CH4. The box model (Olivié 2021) starts in 2015 from observed CH4concentrations, leading to a good agreement of the modelled CH4concentrations with observations. The background CH4emissions over the period 1990–2015 have been estimated by an iterative method, where the background time-varying CH4emission were calculated that best reproduce the observed CH4concentrations, which leads to a good agreement with the historical prescribed simulation and the observations The average background emissions over the period 2005–2014 are then used as initial conditions for 2015 and onwards. The emissions used in the box model, along with the calculated global CH4 concentrations and lifetime are presented in figure 1. As seen in figure 1(a), the global CH4emissions continue to grow in the CLE scenario from 592 Tg in 2020–688 Tg in 2050, while in MFR, the emissions drop significantly from 592 Tg in 2020–496 Tg in 2025, then continues to decrease but slightly down to 480 Tg in 2050. The lifetime slightly increases from 9.22 years to 9.26 year in CLE (figure 1(b)) as the emissions continue to increase in CLE. On the other hand, in MFR, due to strong reductions in CO, NOx and VOC, the net effect is an increase in the lifetime until 2025 to 9.2 years, and then a slight decrease to 8.8 years in 2050. This lead to a continue increase in global CH4concentrations (figure 1(c)) in CLE from 1874 ppb to 2194 ppb in 2050. In MFR, CH4concentrations decrease to 1569 ppb in 2050. The tuning procedure assumes that the anthropogenic and other natural CH4emissions are correct and that the large difference in CH4burdens in the prescribed and interactive simulations are attributed to the wetland emissions, which are very uncertain and contribute largest among the natural emissions (Sauonois et al 2024). Other natural methane sources are fixed at values of 20 Tg yr−1for termites, close to the 3 Environ. Res.: Climate 4(2025) 015008 U Im et al Figure 1. Olivié (2021) box model global CH4(a) emissions, (b) lifetime, and (c) concentrations under CLE and MFR scenarios. 25 Tg yr−1from Basu et al (2022), and 27 Tg yr−1for combined geological +ocean +lake sources (Fung et al 1991). There are no changes in methane emissions due to thawing permafrost (Turetsky et al 2020) or increased release from hydrates over time. These emissions are also likely on the low side compared to newer literature, where global oceanic CH4emissions are estimated to be 6–12 Tg yr−1(Weber et al 2019) and lake emissions are estimated to be 24.0 ±8.4 Tg yr−1(Zhuang et al 2023). These suggest that the natural CH4 emissions in the GISS-E2.1 model should be updated based on recent findings. GISS-E2.1 includes only the chemical loss of CH4with OH (Shindell et al 2013b) and the soil sink is constant at 30 Tg yr−1(Fung et al 1991). Other known direct chemical sinks such as chlorine, or the indirect sink due to other halogens affecting OH are not included in the model. 2.2.3. Simulations Both the prescribed and interactive simulations are AMIP-type simulations, with prescribed SSTs and sea-ice concentrations calculated by the fully coupled GISS-E2.1 contribution to the AMAP (2021) assessment, described in detail in Im et al (2021) and von Salzen et al (2022). The prescribed simulation used methane mixing ratios calculated by Olivié (2021). We carried out two sets of simulations that used the CLE and MFR emission scenarios. Each simulation in these two sets of scenarios were initialized from a set of fully coupled recent past simulations (1990–2014) to ensure a smooth continuation from recent past to future. 3. Results and discussions 3.1. Anthropogenic and natural methane emissions The anthropogenic and natural CH4emissions in the present day (averaged over 2015–2020) and their spatial distribution are presented in figures 2and 3, respectively, while their temporal evolution in the historical (1995–2014) and future (2015–2050) are presented in figures 3and 6, respectively. The natural emissions (380 ±8 Tg yr−1) in the present day are simulated to be larger than the anthropogenic emissions (285 ±3 Tg yr−1), leading to a total CH4emission of 650 ±10 Tg yr−1in agreement with Saunois et al (2024). Both anthropogenic and natural emissions increase in the historical period, going from 222 and 265 Tg, respectively, to 278 and 321 Tg in 2015, respectively, with the total global CH4emission of 645 Tg (figure 3). In the present day, wetland CH4emissions are estimated to be 332 ±8 Tg yr−1, which is much larger than the estimate of 155–217 Tg yr−1in Saunois et al (2020). On the other hand, the global anthropogenic CH4emission we in the present study (285 Tg yr−1) is smaller than 369 Tg yr−1in Saunois et al (2024). Anthropogenic CH4emissions are largest in south and East Asia, while sources over Europe, North and South America, and sub-Saharan Africa stand out (figure 3). Wetland emissions are largest in South America and sub-Saharan Africa, as well as parts of Eastern Europe and Siberia. CH4from biomass burning stands out in the fire hotspots over the Amazons and Africa. CH4from termites are primarily over the southern hemisphere. Overall, CH4emissions are largest over South America, sub-Saharan Africa, and south and East Asia. 3.2. Model evaluation The global atmospheric CH4mixing ratios (ppm) as simulated by the prescribed and interactive simulations are evaluated against those observed by NOAA (Lan et al 2024) between 1995 and 2014 (figure 4). The interactive simulation underestimates the observed global levels by 5% (Mean bias (MB) =−0.09 ppm), while the prescribed simulation underestimates by 2% (MB =−0.04 ppm). The observed temporal evolution is well-captured by the prescribed simulation (r=0.99), while the interactive simulation has a much lower correlation (r=0.65). This is expected as the box-model that calculated CH4concentrations used in the prescribed simulations was tuned via the natural emissions over the period 1990–2014 to correspond as good as possible the observations CH4concentration. The deviation between the prescribed 4 Environ. Res.: Climate 4(2025) 015008 U Im et al Figure 2. (a) CH4emissions and their sources in the present day (2015–2020 mean), (b) the sources of natural CH4emissions, and (c) the evolution of CH4emissions in the 1995–2014 period (right) as modelled by the interactive simulation. Anthr stands for anthropogenic, BBurn for biomass burning, and term for termites. Soil stands for the soil sink for CH4. Figure 3. Global distribution of anthropogenic, wetland +tundra, biomass burning, termite, and other CH4sources averaged over 2015–2020. Figure 4. Observed and simulated global mean CH4concentrations in 1995–2014 period. and interactive simulations after 2007 is due to a larger increase in chemical sink compared to the sources after 2007, leading to a decrease in mixing ratios. We have also evaluated the surface CH4concentrations by the two simulations against observations at various stations under Global Atmospheric Watch (GAW), as shown in figure 5. The prescribed simulation underestimates the GAW surface CH4observations during the period between 1995 and 2023 by 1% [−8.4% 2.0%]. The largest underestimations are over the continental emission regions such as North America, Europe, and Asia, while biases are smallest over oceans. On the other hand, the simulation with interactive sources and sinks underestimates the GAW observations more than the prescribed simulation, by 2.3% [−8.4% 5.9%]. Opposite to the prescribed simulation, the biases are positive over oceans and more negative over the continents. The interactive simulation, with large sources exclusively over land and strong sink over oceans, has a land/ocean ratio larger than 1 while the prescribed simulation has this ratio equal to 1 as it 5 Environ. Res.: Climate 4(2025) 015008 U Im et al Figure 5. Normalized mean bias (MNB) in % in simulated surface CH4concentrations from (a) prescribed and (b) interactive (right panel) simulations in 1995–2014 period. Figure 6. Evolution of global CH4emissions in the interactive (a) CLE and (b) MFR scenarios, respectively, and (c) CH4chemical sink. distributes the global prescribed CH4concentration equally in longitude over a given latitude (figure 8). This clearly shows that using interactive sources and sinks has large impacts on relative surface values across geographies, so that there’s a potential to constrain sources and sinks’ distributions using observations. This also implies that the sink is regionally off due to the missing of some of the other OH-related emissions such as too low CO or too high NOx over land, or other chemistry mechanisms such as halogens over oceans. 3.3. Evolution of future CH4sources and sinks As shown in figure 6, CH4emissions continue to rise in the CLE scenario to 757 Tg in 2050, while in MFR, there is a rapid decline to around 600 Tg by 2025, which then stabilizes. The decline in the MFR scenario is due to the large reduction of anthropogenic emissions until 2025, which then continues to decrease very slightly, while the wetland emissions stay around 350 ±10 Tg yr−1throughout the simulation period. As shown in figure 6(c), the CH4chemical sinks in the CLE increase throughout the simulation period, while the sink in MFR decreases slightly in the same period, which leads to a slight increase in the simulated CH4 mixing ratios in the interactive simulation as opposed to the prescribed simulation. Under the MFR, less OH is simulated by the model (not shown) compared to the CLE scenario, however in both scenarios, OH continues to increase between 2020 and 2050. This also points to the non-linear chemistry in the model concerning oxidative capacity (e.g. OH) and CH4, which does not respond linearly to the changes in emissions. 3.4. Evolution of future CH4mixing ratios and lifetime The prescribed global mean CH4mixing ratios increase in the CLE scenario and decrease in the MFR scenario (figure 7). In the interactive simulations, as the future CH4emissions increase in the CLE scenario (figure 6), the simulated CH4mixing ratios also increase, following nicely the prescribed CH4mixing ratios (r=0.99, NMB =−2%), reaching to ∼2.2 ppm by 2050. The simulated lifetimes are also similar in the prescribed and interactive simulations in the present day (∼8.4 years). The lifetimes slightly increase towards 2050 to ∼8.6 years in all simulations, except for the interactive MFR simulation, where the lifetime increases to 9.2 years, and cannot capture the decrease in the prescribed global mixing ratios. As expected, the interactive MFR scenario simulates lower global surface CH4concentrations and burdens compared to the interactive CLE scenario, however in both cases, global surface CH4and burden continue to increase through 2050 compared to present day. Both the interactive CLE and MFR scenarios show increased concentrations over the whole globe (figure 8). In the CLE scenario, CH4concentrations increase largest over the northern hemisphere, in particular over India and East China, and the Pacific. On 6 Environ. Res.: Climate 4(2025) 015008 U Im et al Figure 7. (a) Evolution of simulated and observed (NOAA) global CH4concentrations and (b) lifetime in the prescribed and interactive simulations. Figure 8. Spatial distribution of (a) CH4concentrations in 2020 as simulated by the prescribed and (b) interactive simulations and the difference in 2050–2020 CH4concentrations as simulated by the (c) interactive CLE and (d) MFR scenarios. the other hand, the MFR scenario shows increases particularly over the whole Southern Hemisphere, however much smaller compared to CLE. While the evolution of global CH4concentrations in the future as simulated by the interactive model is consistent with the prescribed simulations under CLE scenario, as shown in figure 7(a), the interactive model under the MFR scenario simulates an increasing global CH4concentration opposite to the prescribed MFR simulation that show a decreasing trend. This is due to the different complexity of the chemistry mechanisms in the box model and GISS-E2.1, that results in different lifetimes; CH4lifetime decreases in MFR in the box model (figure 1(b)) while it increases in GISS-E2.1 (figure 7(b)). This highlights that the response of concentrations are not necessarily linear with the changes in emissions as the chemistry is non-linear, and dependent on the oxidative capacity of the atmosphere due to other species such as CO and VOCs. In addition, there are also differences in the CH4sinks; chlorine sink is taken into account in the box model while it is not in the GISS-E2.1, leading to less chemical removal and longer lifetime compared to the box model. Finally, the simple box model is based on global and annual mean emissions, which does not take into account the spatial and temporal (i.e. seasonal variation) in emissions and chemistry that is implemented in GISS-E2.1. These difference lead to difference responses in CH4lifetime and therefore concentrations in the box model and GISS-E2.1, even if the emissions are similar. 3.5. Tropospheric chemistry impacts We have investigated the difference between the surface ozone (O3) concentrations in the prescribed and interactive simulations both in its geographical distribution in year 2020 and evolution over time in the CLE and MFR scenarios (figure 9). The geographical distribution of surface O3 in the two simulations are very 7 Environ. Res.: Climate 4(2025) 015008 U Im et al Figure 9. Geographical distribution of 2020 surface O3in the (a) prescribed, and (b) interactive simulations, (c) their difference, and (d) temporal evolution of global surface O3concentrations in the interactive CLE and MFR scenarios. Figure 10. Geographical distribution of 2020 surface OH in the (a) prescribed, and (b) interactive simulations, and (c) their difference. similar (figures 9(a) and (b)) with highest levels over south and east Asia, as well as eastern U.S. and Eastern Mediterranean, and lowest concentrations over the Pacific and South America. The differences between the two simulations (figure 8(c)) are largest over southern hemisphere, with the interactive simulation giving larger O3concentrations over the ocean by up to 4 ppbs and over south Africa by up to 7 ppbs. Similar to surface O3, we have also calculated the hydroxyl radical (OH) surface mixing ratio in the two simulations (figures 10(a) and (b)). The patterns are very similar in the two simulations, with the interactive simulation calculating slightly higher concentrations globally compared to the prescribed simulation, by up to 0.06 ppt (45%), particularly over the Eastern Mediterranean and the Arabian Peninsula, as well as over the oceans (figure 10(c)). 3.6. Climate impacts We calculated the top of the atmosphere (TOA) instantaneous net (shortwave +longwave) radiative forcing due to CH4(RFCH4) in the CLE and MFR scenarios, using the formulation from Etminan et al (2016). As seen in figure 11(a), the global mean TOA RFCH4 increases from 0.42 W m−2in 2020 to 0.56 W m−2in 2050 8