Land surface temperature (LST) has been widely utilized to estimate surface energy fluxes, including evapotranspiration, at satellite spatial scales and repeat frequencies. Here we demonstrate the ability of in-situ proximal remote sensing of LST to allow for accurate estimation of surface energy fluxes by a biophysical land surface model at sub-hourly temporal resolution over growing seasons for multiple agricultural species. Machine learning is applied here to both reduce the input data requirements of the biophysical model and to guide the diagnosis of model error sources.