On Uncertainty in Global Terrestrial Evapotranspiration Estimates from Choice of Input Forcing Datasets

Evapotranspiration ET is a critical water, energy, and climate variable, and recent work has been published comparing different global products. These comparisons have been difficult to interpret, however, because in most studies the evapotranspiration products were derived from models forced by different input data. Some studies have analyzed the uncertainty in regional evapotranspiration estimates from choice of forcings. Still others have analyzed how multiple models vary with choice of net radiation forcing data. However, no analysis has been conducted to determine the uncertainty in global evapotranspiration estimates attributable to each class of input forcing datasets. Here, one of these models [Priestly-Taylor JPL (PT-JPL)] is run with 19 different combinations of forcing data. These data include three net radiation products (SRB, CERES, and ISCCP), three meteorological datasets [CRU, Atmospheric Infrared Sounder (AIRS) Aqua, and MERRA], and three vegetation index products [MODIS; Global Inventory Modeling and Mapping Studies (GIMMS); and Fourier-Adjusted, Sensor and Solar Zenith Angle Corrected, Interpolated, Reconstructed (FASIR)]. The choice in forcing data produces an average range in global monthly evapotranspiration of 10.6 W m(-2) (similar to 20% of global mean evapotranspiration), with net radiation driving the majority of the difference. Annual average terrestrial ET varied by an average of 8 W m(-2), depending on choice of forcings. The analysis shows that the greatest disagreement between input forcings arises from choice of net radiation dataset. In particular, ISCCP data, which are frequently used in global studies, differed widely from the other radiation products examined and resulted in dramatically different estimates of global terrestrial ET.