Carbon cycle feedback dependence on imposed land-use forcing

Earth System Models used to quantify climate-carbon cycle feedbacks and remaining carbon budgets are initialised from pre-industrial land-use conditions, yet the real world has experienced centuries of land-use change. Whether the resulting land-surface state affects the diagnosed feedback parameters-concentration-carbon (β land ) and climate-carbon (γ land )-and the Transient Climate Response to cumulative Emissions (TCRE) remains poorly explored.Here we quantify how different initial land-use states-no land-use, pre-industrial, and present-day-affect carbon cycle feedback parameters in standard 1% CO 2 coupled and biogeochemically coupled simulations using HadCM3LC-Bris. Present-day land use reduces β land by 16% and makes γ land 8% more negative relative to no land-use, but the biophysical cooling from deforestation depresses the climate sensitivity parameter α, producing a net 8% reduction in TCRE at 2×CO 2 . This implies a systematic positive bias of ∼8% in TCRE diagnosed from pre-industrial-initialised experiments, corresponding to ∼40 GtCO 2 for a 1.5 • C carbon budget. At higher forcing, the land-use effect on TCRE diminishes as α differences converge while γ land continues to diverge. Our results highlight a baseline-dependent bias in current CMIP protocol design with direct implications for carbon budget estimates.