Understanding the glacial methane cycle

Atmospheric methane (CH4) varied with climate during the Quaternary, rising from a concentration of 375 p.p.b.v. during the last glacial maximum (LGM) 21,000 years ago, to 680 p.p.b.v. at the beginning of the industrial revolution. However, the causes of this increase remain unclear; proposed hypotheses rely on fluctuations in either the magnitude of CH4 sources or CH4 atmospheric lifetime, or both. Here we use an Earth System model to provide a comprehensive assessment of these competing hypotheses, including estimates of uncertainty. We show that in this model, the global LGM CH4 source was reduced by 28–46%, and the lifetime increased by 2–8%, with a best-estimate LGM CH4 concentration of 463–480 p.p.b.v. Simulating the observed LGM concentration requires a 46–49% reduction in sources, indicating that we cannot reconcile the observed amplitude. This highlights the need for better understanding of the effects of low CO2 and cooler climate on wetlands and other natural CH4 sources.


Supplementary Note 2: Photolysis and stratospheric O 3
In HadGEM2-ES the photolysis rates are pre-computed (2) and so cannot respond to changes in the radiation balance in the model. Additional simulations with interactively calculated photolysis rates (2) were performed using a version of HadGEM2-ES that includes the Fast-J scheme (3). These simulations are otherwise identical to low-fire simulations described in the main text, including the same trace gas emissions and other boundary conditions. In these simulations the surface CH 4 concentration is prescribed with the global mean values for the pre-industrial or LGM.
Including the dynamic interactive photolysis scheme caused the present-day lifetime to reduce by 28% (2). The difference in lifetime between the PI and LGM simulations was also reduced to a negligible -0.2% compared to a 2.3% increase in the model configuration with pre-calculated photolysis rates. Thus, the impact of photolysis is smaller than the other factors considered.
A further consideration for the photolysis rates is the change in stratospheric ozone (4). HadGEM2-ES does not incorporate stratospheric halogen chemistry processes. For this reason, as noted in the Methods section of the main text, the O 3 field is overwritten by an observationallybased climatology. This is performed from 3 levels above the diagnosed tropopause. To test the role of stratospheric O 3 we additionally prescribed a 3% O 3 increase throughout the stratosphere in a LGM simulation consistent with past work (5; 6). This results in a 2% increase in the LGM CH 4 lifetime relative to the pre-industrial. Thus the lifetime change between the LGM and pre-industrial is 2% in the simulations with pre-computed photolysis and 2% when interactive photolysis rates with a 3% increase in stratospheric O 3 mixing ratios are included.
We performed a further LGM simulation in which the monthly-mean O 3 field at model levels from the tropopause and upwards from the pre-industrial simulation are prescribed in radiation code of the LGM simulation. This removes any difference in stratospheric O 3 . These sensitivity tests show little sensitivity to the level at which stratospheric O 3 is prescribed.
We then performed a final test in which the stratospheric O 3 is modified based on the troposphere-stratosphere chemistry simulations for the pre-industrial and LGM from ref. (5). The results show a similar 1.7% increase in lifetime, relative to the pre-industrial. We therefore use this final simulation to quantify the impact of changes in stratospheric O 3 during the LGM in figure 3 of the main text.

Supplementary Note 3: Alternative CH 4 source scenarios
In order to address uncertainty in the make-up of natural CH 4 sources two additional source scenarios are used in offline mass-balance calculations of the LGM CH 4 concentration. These alternative emission scenarios are based on published pre-industrial or modern emission categories as summarised in Supplementary Table 3. Harder07 is based on the Holocene base in ref. (7) and Kirschke13 is based on the average bottom-up estimates for the past three decades (8). Wetland and fire emissions are kept as in the default HadGEM2-ES because wildfire emissions in Harder07 and Kirschke13 are more appropriate for present day conditions, and are therefore likely to be substantially smaller than pre-industrial burning rates, because of anthropogenic fire-suppression activities.
In both cases, the non-wetland sources must be scaled down substantially to close the overall CH 4 budget. For the two scenarios, Harder07 and Kirschke13 bottom-up (BU) the scaling factors are 0.7 and 0.35 respectively, resulting in a global total source of 186TgCH 4 yr −1 in both cases (and including the sink term of 11.2TgCH 4 yr −1 ). LGM changes are applied to wetlands, fires, oceans, termites and the soil sink as in the standard HadGEM2-ES scenario. Freshwater emissions are scaled with the wetland flux, but no change is applied to the wild animal, permafrost, hydrate or geological source terms, reflecting incomplete knowledge of these systems. The resultant LGM predicted concentration is 479 and 495ppbv (Harder07 and Kirschke13BU, respectively) for a lifetime increase of 6.5% (and including peatland sources). This compares with a value of 464ppbv in the equivalent scenario from Supplementary Table 2. Thus an uncertainty of up to 30ppbv is introduced by considering these alternative representations of the make-up of pre-industrial CH 4 sources.

Supplementary Note 4: Example Offline CH 4 budget calculations
A decrease in emissions due to the LGM wetland change by -41.4 from 165TgCH 4 yr −1 , gives a fractional source reduction of δ=0.25. This causes a fractional concentration change of (1+δ) F (following ref.