Satellite data contain a wealth of information that are indirectly related to evapotranspiration (ET). Measurements in the microwave, infrared, and visible spectra have complementary capabilities for ET estimation. These capabilities have been demonstrated by using neural networks to estimate monthly-average ET from satellite-derived parameters: microwave emissivities, day-night difference in land surface temperature (from microwave), downward solar and infrared fluxes from ISCCP cloud analysis, and visible and near-infrared surface reflectances. To obtain global coverage, the ET reference for the training is from a land surface model, pending availability of more widespread in-situ reference measurements. The satellite-derived ET products are evaluated at sites where additional satellite and in-situ measurements are available, with a particular focus on anomalous areas. An objective is to identify potential improvements in the use of satellite data in land surface data assimilation.