Quantifying how land surface height, such as that of the Tibetan region, has changed with time is crucial for understanding a range of Earth processes, including atmospheric dynamics, biotic evolution and tectonics. Elevation reconstructions are highly uncertain and controversial, in part because of assumptions used in their calculation. The largest uncertainties are in the choice of unconstrained thermal lapse rates. Thermal lapse rates are defined as a change in surface temperature with altitude and have long been used to estimate paleoelevation. If we know both the lapse rate and the temperature at two sites at different elevations, then in theory we can calculate their height difference. There are different types of lapse rates (Dry, Saturated, Environmental and Terrestrial), yet which is the most useful for paleoaltimetry is unknown. Previous paleoelevation studies have often used observed modern-day global annual mean free air or terrestrial thermal lapse rates to measure elevation change, with the assumption that observed modern-day lapse rates are similar to those of the past. Here, using the HadCM3L paleoclimate model we demonstrate that Eocene global mean free air and terrestrial thermal lapse rates are not only different from the modern, but also show little predictive skill in reproducing prescribed model topography. Free-air lapse rates are largely insensitive to increased pCO(2) (showing only a decrease of similar to 0.1-0.5. C/km), whereas lapse rates at Earth's surface, the most applicable for fossil-based paleoaltimetry, differed significantly locally and globally in the past compared to the Pre-industrial. This suggests that modern terrestrial lapse rate expressions are inappropriate for tracking altitude changes through geologic time. Moist processes and resultant moisture content of airmasses play a critical role in much of this uncertainty. The use of a wet-bulb temperature-derived lapse rate reduces this uncertainty significantly improving the predictive skill. Local terrestrial thermal lapse rates can be useful in paleoaltimetry, but only through climate model mediation where uncertainties can be reduced and quantified. Critically, paleoclimate models offer the opportunity to provide mean sea-level surface temperature to derive an elevation estimate where proxy-based values may not be available. (C) 2021 The Author(s). Published by Elsevier B.V.