Carbon Cycle Feedbacks to Add 20-30% Extra Warming by 2100 in Coupled Models

The New Scientist piece summarizes results from CMIP6-era ensembles that allow interactive biogeochemistry rather than prescribed carbon fluxes. These runs show an extra 50-120 GtC released from terrestrial and wetland sources under SSP5-8.5, driven mainly by temperature-sensitive respiration and thaw. The magnitude matches independent constraints from the Global Carbon Project’s 2023 budget update, which already records an 8% decline in the land sink fraction since 2000.

Earlier generations of models underestimated these loops because respiration and methane modules were not fully coupled; the resulting bias kept 2100 temperature estimates 0.2-0.3 °C too low. This matters for policy because national determined contributions and remaining carbon budgets were calibrated against those older numbers, implying faster emissions cuts are now required to stay below 2 °C.

Observational gaps remain large: FLUXNET and satellite methane records cover only a fraction of high-latitude and tropical wetlands, leaving the upper tail of feedback strength poorly constrained. CMIP7 plans to assimilate new satellite and eddy-covariance data streams, which should narrow the 20-30% range by the 2028 assessment cycle.

If the high-end feedback strength materializes, committed warming already exceeds 2 °C even under net-zero pathways that ignore these loops, underscoring the urgency of protecting intact carbon stores rather than relying solely on emission reductions.