Abstract. Potential evaporation (E_p) is a crucial variable for hydrological forecast and in drought monitoring systems. However, multiple interpretations of E_p exist, and these reflect a diverse range of methods to calculate E_p. As such, a comparison of the performance of these methods against field observations in different global ecosystems is badly needed. In this study, we used eddy-covariance measurements from 107 sites of the FLUXNET2015 database, covering 11 different biomes, to parameterize and compare the main E_p methods and uncover their relative performance. For each site, we extracted the days for which ecosystems are unstressed based on both an energy balance approach and on a soil water content approach. The evaporation measurements during these days were used as reference to validate the different methods to estimate E_p. Our results indicate that a simple radiation-driven method calibrated per biome consistently performed best, with a mean correlation of 0.93, an unbiased RMSE of 0.56 mm day⁻¹, and a bias of −0.02 mm day⁻¹ against in situ measurements of unstressed evaporation. A Priestley and Taylor method, calibrated per biome, performed just slightly worse, yet substantially and consistently better than more complex Penman, Penman-Monteith-based or temperature-based approaches. We show that the poor performance of Penman-Monteith based approaches relates largely to the fact that the unstressed stomatal conductance was assumed constant. Further analysis showed that the biome-specific parameters required for the simple radiation-driven methods are relatively constant per biome. This makes this simple radiation-driven method calibrated per biome a robust method that can be incorporated into models for improving our understanding of the impact of global warming on future global water use and demand, drought severity and ecosystem productivity.