Abstract

In this work, the generation of Random Forest Regression models that allow the estimation of daily ET values using medium spatial resolution images without thermal band has been investigated. The models have been trained and tested with the spectral information obtained from the satellites images and the corresponding values of ET obtained by the METRIC model. The obtained results shown that the test and validation error of the models increase a magnitude order when the thermal information is not used; however, considering the agricultural application, this error is tolerable. As far as our knowledge, there are not similar studies to the presented here.