There exists now a number of satellite-based applications capable of providing estimates of evapotranspiration (RS-E) globally at high spatial resolution. Efforts to validate these products through observation has been impeded by errors in eddy covariance flux measurements and complicated by the mismatch in representative scale. Here, we provide a comprehensive assessment of the uncertainties in a Penman-Monteith RS-E attributable to input errors, algorithmic assumptions, and RS data coverage limitations. We quantify and rank the impact of these errors for seven major U.S. basins for the period 2003-6. During this period, the RS-E is also evaluated for bias via intercomparisons with VIC-model and GRACE and NARR inferred E. Ultimately, it is shown that the RS-E is comparable in skill to expected output from a land surface model forced with satellite-based precipitation.