Quantifying uncertainty in a remote sensing-based estimate of evapotranspiration over continental USA

We calculate evapotranspiration (E) from remote sensing (RS) data using the Penman-Monteith model over continental USA for four years (2003-2006) and explore, through an ensemble generation framework, the impact of input dataset (meteorological, radiation and vegetation) selection on performance (uncertainty) at the monthly time-scale. The impact of failed or missed RS retrievals and algorithmic assumptions are also quantified. To evaluate bias, we inter-compare RS-E with three independent sources of E: Variable Infiltration Capacity (VIC)-model simulated, North American Regional Reanalysis (NARR) inferred, and Gravity Recovery and Climate Experiment (GRACE) inferred. Overall, we find that the choice of vegetation parameterization, followed by surface temperature, has the greatest impact on RS-E uncertainty. Additional uncertainty (4-18%) is linked to sources of net radiation-used to scale instantaneous RS-E under the assumption of constant daytime evaporative fraction-including the Surface Radiation Budget (SRB), International Satellite Cloud Climatology Project (ISCCP), and North American Land Data Assimilation System (NLDAS)-VIC. The ensemble median agrees to within 21% of VIC-modelled E, except for the Colorado and Great Basins for which the need for a soil moisture constraint on RS-E becomes evident.