Projections of Earth's Hottest Surface Temperatures in CMIP6

Abstract

Earth's hottest near‐surface air temperatures provide a benchmark for the upper tail of land heat extremes. Using CMIP6 simulations, we evaluate historical extremes against Berkeley Earth and project future changes under SSP2‐4.5 and abrupt‐4 × . Historical fidelity is assessed from the spatial pattern of the 99th percentile of daily maximum temperature over land using centered RMSE, mean bias, and spatial . We analyze three annual tropical land indices: tropical land‐mean temperature , daily‐mean temperature on the hottest day each year , and the maximum daily‐maximum temperature . Across models, and warm 30%–40% faster than in transient trends and equilibrium responses, consistent with amplified upper‐tail warming. Hotspot‐frequency maps of the annual global land maximum show persistence in subtropical arid and semi‐arid regions. We define a fixed‐threshold record‐like benchmark for hottest‐land temperature and relate exceedance likelihood to tropical land warming; pooled fits show exceedance odds rise ∼35% per C.