Identifying main crop classes in an irrigated area using high resolution image time series

We present a robust and reproducible method to map the main land cover classes in an irrigated area near Marrakech (bare soil, herbaceous crops, trees on bare soil, trees on herbaceous understory), using a time series of 9 SPOT high resolution images. The images were first radiometrically corrected in reflectance using a physical model (SMAC), and then applying an adjustment based on invariant objects located on the scene. The 9 images were then used to generate a NDVI profile for each pixel, which was analyzed using simple decision rules to assign it to a landcover class, considering simple criteria like the minimum, maximum, and dynamics of the NDVI.