Mitigation of CH4 emissions would significantly increase the feasibility of stabilising global warming below 1.5 °C, alongside having co-benefits for human and ecosystem health

To understand the potential additional benefits of CH₄ reductions on allowable cumulative carbon emissions consistent with the Paris targets, we use the Joint UK Land-Environment Simulator (JULES) (Clark et al 2011) coupled with the intermediate complexity climate model IMOGEN ‘Integrated Model Of Global Effects of climatic aNomalies’ (Huntingford et al 2010). The combined IMOGEN-JULES framework thus provides an intermediate complexity climate-carbon modelling system. IMOGEN utilises ‘pattern-scaling’ to capture the main features of expected local and monthly meteorological changes interpolated to alternative future levels of global warming. This is connected to a gridded version of the land surface model JULES (version 4.8) (Clark et al 2011) to understand the impacts of any transition to different stable warming levels.

IMOGEN comprises a global energy balance model (EBM) whose global climate response characteristics (climate sensitivity for land and ocean, ocean diffusivity etc.) can be chosen to represent any global climate model (GCM). It is driven by time-series of CO₂ concentrations and non-CO₂ radiative forcing. IMOGEN generates gridded outputs of monthly anomaly fields of surface temperature, precipitation, humidity, wind-speed, surface shortwave and longwave radiation and pressure. These anomalies are derived by scaling the patterns from the output from each GCM, assuming these are linear in global surface temperature change. Here the data from the 34 CMIP5 GCMs running the RCP8.5 scenario (Taylor et al 2013) are used to derive both the global climate characteristics and climate patterns. Although the greenhouse gas forcings used in this study will be closer to the RCP2.6 scenario, the RCP8.5 scenario was used to get the clearest signal to determine the climate patterns.

 Zoom In Zoom Out Reset image size

The JULES configuration also includes modelled O₃ damage to photosynthesis, affecting land-atmosphere CO₂ exchange (Sitch et al 2007). This O₃ damage parameterisation can be set to ‘low’ or ‘high’ sensitivity to span the uncertainty in our knowledge of the sensitivity of plants globally. We also include a ‘no’ sensitivity to allow the separation of the ozone effect. In this study we use the low sensitivity parameterisation as the standard configuration, with separate tests of the effects of using the ‘no’ and ‘high’ sensitivities. Surface O₃ concentrations are parameterised as two-dimensional fields as a function of the global average CH₄ concentration. These are previously derived from global chemistry-climate simulations using the HadGEM3 model for global mean atmospheric CH₄ mixing ratios of 1285 ppb and 2062 ppb (Stohl et al 2015). Within IMOGEN-JULES, the O₃ concentration is calculated at each grid point as a function of CH₄ using a linear interpolation between O₃concentrations at the above mixing ratios.

To set the initial (2015) conditions for the land carbon stores, the IMOGEN-JULES model is spun up for 1000 years at 1850 conditions and then run to 2015 with prescribed historical CO₂ mixing ratios, land use, and global surface temperatures from Morice et al (2012) (reaching 0.89 °C by 2015). The spin up and historical simulation are carried out for each climate model realisation. For this study we invert the IMOGEN-JULES configuration, running forward from 2015 with specified global temperature profiles, and specified non-CO₂ radiative forcing changes from 2015. IMOGEN-JULES derives the CO₂ concentrations in each year from the EBM calculations and thence the uptake by the land biosphere; the global carbon cycle is closed with a simple description of global oceanic draw-down of CO₂ (Joos et al 1996). A control simulation is also run maintaining 1850 forcings and temperatures until 2100. Further details of the IMOGEN-JULES setup and the inversion procedure can be found in Comyn-Platt et al (2018).

We determine the carbon budgets consistent with three specified temperature trajectories that stabilise at 1.5 °C (with and without overshoot) and 2.0 °C above pre-industrial levels as shown in figure 1(a). These profiles are generated according to the algorithm in Huntingford et al (2017) as in Comyn-Platt et al (2018). The results are found not to be sensitive to the exact form of the temperature trajectories.

The future non-CO₂, non-CH₄ radiative forcings are taken from one of the Shared Socio-economic Pathways (SSPs) SSP2–2.6 (O’Neill et al 2017, Riahi et al 2017) by subtracting the CO₂ and CH₄ (and associated O₃ and stratospheric water vapour) contributions from the total SSP2–2.6 radiative forcing. We follow the prescription of these terms in the MAGICC climate model (Meinshausen et al2011). After 2015, land-use is fixed at 2015 levels. Here, the IMOGEN physical parameters are varied to represent the climate characteristics, such as the different climate sensitivities, of 34 CMIP5 models.

There is a wide range in the CH₄ emissions in the SSPs that achieve a forcing of 2.6 W m⁻² in 2100, suggesting that the options for mitigation are not exhausted (Gernaat et al 2015). We construct four different anthropogenic CH₄ mitigation scenarios (figure 1 (b)). The first three are ‘High’ CH₄ and ‘Medium’ CH₄ which span the highest and lowest of the SSP2–2.6, and ‘Low’ CH₄ which we parameterise as following the Medium scenario to 2020 then decaying faster to 62 Tg CH₄ yr⁻¹ by 2100. For the High CH₄ scenario, CH₄ concentrations increase following the an upper bound of SSP4-2.6 and SSP5-2.6 CH₄ concentration projections from the GCAM integrated assessment model (IAM) (Calvin et al 2017). For the Medium CH₄ scenario, concentrations follow SSP2–2.6 as generated by the IMAGE 3.0 IAM (van Vuuren et al 2017). For the Low CH₄ scenario, we assume extra reductions are possible by removing the restriction on cost minimisation. To generate a smooth curve we parameterise emissions (in Tg CH₄ yr⁻¹) as , where x is the number of years after 2020. This projects a lower CH₄ projection curve than the strongest mitigation SSP storyline (SSP1–2.6 variants). The High, Medium and Low scenarios lead to year 2100 atmospheric CH₄concentrations of 1839, 1275 and 1008 ppb, respectively. We also consider a fourth scenario ‘Late’, to test whether the timing of the CH₄ mitigation matters, where emissions are maintained at current (2015) levels until 2050 and then apply the same rate of mitigation for the Low CH₄ profile post-2015, but extended to ensure that the 2100 concentration matches Low CH₄. Note that we are not assuming specific methane mitigation measures in these scenarios, or possible effects on co-emitted species such as N₂O.

Emissions are converted into concentrations using the formulation of the MAGICC model (which includes natural emissions of 250 Tg CH₄ yr⁻¹). Radiative forcings for the CH₄ scenarios are calculated using formulae including the short-wave absorption (Etminan et al 2016), and the overlap with N₂O using the N₂O concentrations in SSP2–2.6. The contributions from O₃ and stratospheric water vapour are added in as linear functions of CH₄ mixing ratio. From IPCC AR5 (Myhre et al2013) these amount to 2.36 × 10⁻⁴ ± 1.09 × 10⁻⁴ Wm⁻² per ppb CH₄ (0.65 ± 0.3 times the CH₄radiative efficiency).

This spread in possible CH₄ trajectories is wider than typically projected in integrated assessment models (IAMs) (Rogelj et al 2015a). However, the IAM outputs are unlikely to span the full range of CH₄ measures that are available. This is partly due to their cost minimisation approaches which exclude the more expensive measures and neglect the social costs of methane (Shindell et al 2017), and their lack of diversity in treatment of non-CO₂ mitigation measures. These IAMs also have limited representation of the specific processes responsible for methane production and of the technologies available for methane mitigation. It is therefore difficult to estimate how deep (or not) reductions can go. Achieving our most stringent scenario would be expected to draw on specific sectoral measures to address CH₄. These could include increasing agricultural efficiency, decreased food waste and decreased beef consumption (van Vuuren et al 2017). The Low and Late scenarios should therefore be seen as illustrative examples.

For the High CH₄ scenario (no CH₄ mitigation) the allowable carbon emissions from 2015–2100 span from 149 ± 51 GtC for 1.5 °C (no overshoot), 143 ± 56 GtC for 1.5° with overshoot, to 403 ± 94 GtC for the 2° temperature pathway. The uncertainty is due to the range of climate sensitivities of the CMIP5 models emulated by the IMOGEN framework. Rather than these absolute budgets we focus on the differences in the cumulative carbon emissions from the inversions for the different CH₄scenarios. These show almost no dependence on the climate model realisation and little dependence on the temperature profile. The benefit of the Medium vs the High CH₄ scenario is approximately 155 GtC over the period 2015–2100 (figure 2(a)). Stronger CH₄ mitigation down to the Low scenario gains another 80 GtC if it is done early. The loss in benefit from delaying CH₄ mitigation according to the Late CH₄ scenario is 40 GtC. These values are similar to a study comparing no mitigation with stringent mitigation (Rogelj et al 2015b) which calculated an increase of 130 GtC in the carbon budget, with a 30 GtC penalty for late mitigation.

 Zoom In Zoom Out Reset image size

The relationship between the allowable carbon emissions from 2015–2100 and CH₄ concentrations at 2100 is almost linear (excluding the Late CH₄ scenario) with very little difference between the climate model realisations (figure 2(b)). The slopes are −0.269 ± 0.001 GtC ppb⁻¹ for the 1.5° and 1.5° overshoot profile and −0.277 ± 0.002 GtC ppb⁻¹ for the 2 °C profile. Compared to the CH₄ forcing at 2100 (including the O₃ and stratospheric water vapour effects), this is equivalent to 350 or 360 GtC (Wm⁻²)⁻¹. There are uncertainties in these relationships due to the uncertainty in the total radiative efficiency of methane. As these relationships are based on the methane concentrations, rather than emissions, uncertainties in the methane lifetime do not affect the result. The uncertainty in the direct methane radiative efficiency is taken to be 9% of the total (Etminan et al 2016). When combined with the 16% uncertainty from the ozone and water vapour contributions this leads to an overall uncertainty of 18%, (0.048 GtC ppb⁻¹). This uncertainty includes within its span the relationship (−0.236 GtC ppb⁻¹) expected using the Myhre et al (1998) forcing instead of Etminan et al (2016).

The change in carbon budgets (high methane vs low methane) can be broken down in to the different carbon stores: atmosphere, land (soil and vegetation) and ocean (figure 3(a)). We define the airborne fraction α_F = ΔCO₂/ΔE_CO2, where ΔCO₂ is the change in the atmospheric CO₂ burden and Δ_ECO2 is the change in cumulative CO₂ emissions, both in GtC. We find that the α_F of the extra carbon allowed through CH₄ mitigation is independent of the climate sensitivity of each climate model. α_F is also the same when comparing Low-High and Medium-High CH₄ mitigation (not shown). There is a slight dependence of α_F on temperature profile with the 1.5 °C profiles having an α_F of 0.44 vs 0.49 for the 2 °C profile. The Late CH₄ mitigation does not follow the same linear relationship as the Low or Medium scenarios, falling well below the line of proportionality in figure 2(b). With late CH₄mitigation, the comparative increase in allowable atmospheric CO₂ concentrations (compared to High CH₄) does not occur until late in the century. The increase in the atmospheric carbon is the same as for the early mitigation, but the ocean and the land have not had time to take up this extra carbon and the α_F of the extra CO₂ is thus higher (0.53).

 Zoom In Zoom Out Reset image size

 Zoom In Zoom Out Reset image size

Since surface O₃ decreases vegetation productivity, mitigation of CH₄ leads to additional climate benefits than might be expected simply through the radiative forcing. Decreasing atmospheric CH₄concentrations reduces O₃ levels and increases the uptake of carbon into vegetation and soils. In terms of equation (1), reducing O₃ reduces a_F. We test this through further inversions assuming no and high sensitivity of vegetation to O₃, compared with the baseline parameterisation in the previous results of lower plant-O₃ sensitivity. We find that by increasing the impacts on the land carbon uptake, O₃ damage adds 9–28 GtC (4%–12%) to the benefit of the Low vs High CH₄ scenarios depending on the assumed sensitivity of vegetation to O₃ (figure 3(b)).

To maintain the radiative balance in the inverse model the change in atmospheric CO₂ is entirely determined by the change in the non-CO₂ forcing. Since we invert IMOGEN to derive the radiation balance consistent with the specified temperature profiles, the greenhouse gas forcing must be the same at any given time, such as at 2100, (assuming the climate sensitivities to radiative forcing from CH₄ and CO₂ are equal). So

where ΔCO₂ and ΔCH₄ are the CO₂ and CH₄ burdens in GtC and GtCH₄, and  and  are the average radiative efficiencies for increases in CH₄ (including its indirect effects) and CO₂ in Wm⁻² GtCH₄⁻¹ or Wm⁻²GtC⁻¹. So combining these with the airborne fraction α_F defined previously gives the ratio of extra cumulative carbon emissions (ΔE_CO2) to change in CH₄ abundance:

This equation is exact and simply follows from the way we have defined ,  and α_F. The linear relationship between the change in the allowable emissions and the change in 2100 forcing therefore implies a constant ratio between the cumulative emissions to 2100 and the 2100 atmospheric CO₂burden, i.e. a constant airborne fraction for the extra allowable emissions as found in figure 3(a). Although  and  are not constant, but functions of the atmospheric CO₂ levels and the magnitudes of the changes ΔCO₂ and ΔCH₄, the deviations from linearity are small for the methane mitigation scenarios used here. The slightly higher α_F for the 2.0° temperature profile is due to the lower radiative efficiency () at higher absolute CO₂ levels.

The equation also holds in the more realistic case where the extra allowed CO₂ is not emitted with the time profile required to precisely follow the prescribed temperature curve, although in this case the α_F may be slightly different from found in this study. The energy balance has little dependence on the shape of the temperature curve before 2100 (or any specific time), and is dominated by the absolute temperature and its time derivative at 2100. This relationship has no dependence on climate sensitivity. However the α_F will be affected by the sensitivity of the carbon cycle to changes in atmospheric CO₂, surface temperature and precipitation (Arora et al 2013).

Allen et al (2016) have derived a variant of the Global Warming Potential metric (GWP⁎) that relates the change in cumulative emissions of CO₂ to the change in instantaneous emissions of a short-lived species (here CH₄).

where Δe_CH4 is the change in the instanteous CH₄ emission rate (in GtCH₄ yr⁻¹), H is a chosen timeframe, and AGWP^H_X are the absolute GWPs for CH₄ and CO₂. The absolute GWPs can be expanded to give:

where α_F(H) is the airborne fraction of a pulse of CO₂ averaged over H years. This is similar to equation (1), but only equal to it if the CH₄ has reached equilibrium (i.e. ΔCH₄ can be replaced by Δe_CH4 × τ_CH4) and the airborne fraction of CO₂ in the AGWP_CO2^H (α_F(H)) is equal to the α_F of the extra allowed CO₂.

In terms of GWP^⁎, the results from our experiments give a ratio  at the end of the century of 2900 (low-medium mitigation) to 3300 (low-high mitigation) in GtCO₂ GtCH₄⁻¹ yr⁻¹, which compares well with a GWP⁎ (100 years) of 2800 yr, given that implicit in the GWP^⁎ approximation are the assumptions that the CH₄ concentrations have equilibrated and that the CO₂ airborne fraction is constant.

We find that CH₄ mitigation has non-climate benefits in terms of air quality and vegetation productivity (by allowing greater atmospheric CO₂ levels, and by reducing the damage from O₃). West et al (2012) found that a strong methane mitigation scenario (emission decrease of 180 Tg CH₄ yr⁻¹) resulted in a decrease in global ozone concentrations of around 2 ppb and avoided mortalities of around 90 000 per year. In this study, mitigation by 260 TgCH₄ yr⁻¹ (Low vs High scenario) achieves a decrease in surface O₃ concentration of 3 ppb as a global average, with the largest impact in the tropics (see figure 4(a)). Therefore a rough scaling of West et al (2012) would suggest a benefit of around 130 000 avoided mortalities per year.

The increased allowable CO₂ levels lead to increased net primary plant productivity (NPP) in JULES by 4% as a global average (figure 4(b). If we assume the high sensitivity of plants to ozone the effects of O₃ reduction add up to another 2% increase in NPP globally. In places where the changes in ozone overlap with areas of high productivity (Eastern US, northern Europe) the reductions in ozone could increase total NPP by 4%–6% in the high sensitivity case (figure 4(c)).