Interchangeability of multi-decade skin and surface air temperature trends over land in models

Satellite land surface temperature ( T _s ) records have now reached 20+ year length, but their trends may differ from historical records built from in-situ measurements of near-surface air temperature ( T _as ). In the ERA5 reanalysis, 60° S–60° N land T _s and T _as trends can differ by up to ±0.06 °C decade ^−1 over 20 years, depending on the period, or more on smaller spatial scales. Here I use 1979–1998 outputs from ACCESS1-0 climate model simulations with prescribed land T _s to understand changes in T _s and T _as . CO _2 ’s effective radiative forcing causes adjustments that warm T _as relative to T _s . In ACCESS1-0, vegetation enhances the adjustments to CO _2 over land. Meanwhile, feedbacks in ACCESS1-0 oppose the adjustments, resulting in small long-term net effects on global temperature estimates. In coupled simulations from other models, there is no agreement on whether T _s or T _as warms more and the most extreme case shows global long-term differences of just 5% between land T _s or land T _as trends. The results contrast with over-ocean behavior where adjustments and feedbacks reinforce each other, and drive larger long-term T _as warming relative to T _s across all models.