Effective irrigation of cotton can be improved by mapping the spatial distribution of evapotranspiration (ET) throughout the field. Using remotely sensed thermal infrared data and surface energy balance modeling, spatial ET can be estimated at daily time steps. However, the accuracy and consistency of  ET estimates is heavily dependent upon the quality of the land surface temperature estimates and the robustness of the selected energy balance model. To evaluate these, remotely sensed image data were collected and modeled for a 4.8 ha cotton experiment at Maricopa, Arizona in 2009 and 2011. Resulting ET estimates from two source (TSEB) and single source (METRIC) models were compared against ET derived from soil moisture depletion data via neutron probe data, eddy covariance observations at two sites, and large aperture scintillometry for two transects. ET modeling bias and accuracies for the two remote sensing models will be discussed.