The precise determination of evapotranspiration rate is challenging because it is a quantity that is difficult to measure and to parametrize. Direct estimates include the determination of the change of mass of a volume of soil and vegetation that evapotranspirates using lysimeters, or direct measurements of turbulent water vapour fluxes by eddy-covariance systems. Parametrized estimates that make use of the Monin–Obukhov similarity theory use vertical gradient measurements of temperature and moisture at one point, and line or area averages by means of scintillometers operating at high frequency. A relation for the evapotranspiration from well-watered surfaces was initially developed by Penman and later expanded for vegetated surfaces and for heterogeneous croplands. A popular simplified expression was obtained by Priestley and Taylor. The current challenge is to find expressions for the evapotranspiration in non-saturated conditions, which are common in arid and semi-arid climates, and for heterogeneous terrain. In numerical models, the estimated actual evapotranspiration over land is obtained as the result of the explicit representation of the different involved sub-processes taking place in the soil and the canopy, using so-called land-surface models. Usually these mechanisms are described in a simplified manner and rely on a number of adjustable parameters. The improvement of such descriptions relies in the availability of experimental measurements to make the physical models more complete and robust.