 This research proposes a novel methodology for mapping annual crop lands at high spatial resolution without the need for field data. It uses temporal features specific to crops as well as cost-effective classifiers to reduce mislabeled pixels and produce accurate maps. The methodology was tested on eight sites around the world and the results show that it can provide accurate maps at the end of the season with an overall accuracy of more than 85% even when the baseline land cover is coarse. Additionally, early crop land maps can be obtained at three-month intervals during the growing season with an overall accuracy of more than 70%. This suggests that the proposed approach could be used for operational agriculture monitoring programs. This article was authored by Nicholas Matten, Guadalupe Sepulcarcanto, Francois Waldner, and others.