AbstractAimLeaf and air temperatures are seldom equal, but many vegetation models assume that they are. Land‐surface models calculate canopy temperatures, but how well they do so is unknown. We encourage consideration of the leaf‐ and canopy‐to‐air temperature difference (ΔΤ) as a benchmark for land‐surface modelling and an important feature of plant and ecosystem function.LocationTropical SW China.Time period2013.Major Taxa studiesTropical trees.MethodsWe illustrate diurnal cycles of leaf‐ and canopy‐to‐air temperature difference (ΔΤ) with field measurements in a tropical dry woodland and with continuous monitoring data in a tropical seasonal forest. The Priestley–Taylor (PT) and Penman–Monteith (PM) approaches to evapotranspiration are used to provide insights into the interpretation and prediction of ΔT. Field measurements are also compared with land‐surface model results obtained with the Joint U.K. Land Environment Simulator (JULES) set up for the conditions of the site.ResultsThe ΔT followed a consistent diurnal cycle, with negative values at night (attributable to negative net radiation) becoming positive in the morning, reaching a plateau and becoming negative again when air temperature exceeded a ‘crossover’ in the 24–29 °C range. Daily time courses of ΔT could be approximated by either the PT or the PM model, but JULES tended to underestimate the magnitude of negative ΔT.Main conclusionsLeaves with adequate water supply are partly buffered against air‐temperature variations, through a passive biophysical mechanism. This is likely to be important for optimal leaf function, and land‐surface and vegetation models should aim to reproduce it.