Methodology for Crop Rotation Mapping Using AI and Remote Sensing Imagery

A Support Vector Machine (SVM) classification model was developed using feature sets derived from multispectral Sentinel-2 Level-2A imagery, incorporating phenological descriptors extracted from NDVI time-series profiles—specifically, the date of maximum canopy development and the length of the growing period—along with spectral indices computed at the maximum canopy cover.

Following calibration, the model can be applied to historical Sentinel-2 archives to infer cropland trajectories and reconstruct crop-rotation sequences over multiple growing seasons. A similar methodology using a Random Forest classifier was reported in a previous study (DOI: doi.org/10.64898/2025.12.06.692759).