This paper outlines a new methodology to derive evaporation from satellite observations. The approach uses a variety of satellite-sensor products to estimate daily evaporation at a global scale, with a 0.25 degree spatial resolution. Central to this approach is the use of the Priestley and Taylor (PT) evaporation model. Because the PT equation is driven by net radiation, this strategy avoids the need to specify surface fields of variables, such as the surface conductance, which cannot be detected directly from space. Key distinguishing features are the use of microwave-derived soil moisture, land surface temperature and vegetation density, as well as the use of a detailed rainfall interception module. The modelled evaporation is validated against one year of eddy covariance measurements from 43 stations. The estimated annual totals correlate well with the stations' annual cumulative evaporation (R = 0.84, N = 43) and show a negligible bias (−1.5%). The validation of the daily time series at each individual station shows good model performance in all vegetation types and climate conditions with an average correlation coefficient of R = 0.84, still lower than the R = 0.91 found in the validation of the monthly time series. The first global map of annual evaporation developed through this methodology is also presented.