Crop classification using multi-temporal MODIS vegetation indices

The ability to map agricultural lands by crop type is crucial to understanding the geography and dynamics of land use/land cover change and global production. Existing remote sensing methods that can differentiate crops by type require high spatial resolution data, high spectral resolution data, or extensive ground truth information to develop training sites, none of which are freely available for much of the world. As an alternative, I propose a new method of crop classification using multi-temporal MODIS vegetation indices as a base image from which to extract crops using their phenologies. I test and refine this method in Kansas, USA using the USDA Cropland Data Layer as reference. I discuss the numerous factors that effect the application and accuracy of the method, the method’s current limitations, and how the method might be further tested and refined.