This review introduces a novel method for modeling evapotranspiration and surface heat fluxes built on the theory of Maximum Entropy Production (MEP) as an application of the maximum entropy principle to non-equilibrium thermodynamic systems. The formulation of the MEP model uses the Bayesian probability theory, information theory through the concept of information entropy, and the similarity theory of the atmospheric boundary-layer turbulence. The MPE model provides simultaneous solution of latent, sensible and surface medium heat fluxes using only three input variables: net radiation, surface temperature and specific humidity. A unique feature of the MEP model is that the surface energy balance is closed at a range of space and time scales. The model does not require data of temperature and water vapor gradient, wind speed and surface roughness. It does not include empirical tunable parameters such as atmospheric and stomatal conductance. The MEP model is a promising new approach for the study of water and energy cycles of the Earth system across space-time scales.