DETECTION OF TREE PHENOSTAGES USING DJI PHANTOM DRONE ORTHOMOSAIC

Evergreen tropical forest leaf demography drives seasonality of photosynthesis and may also drive other critical seasonal processes, such as wood growth, herbivore avoidance and water use efficiency. Here we assess the separability of fine-scale canopy elements that comprise different crown phenostages, which must be successfully classified as part of the workflow for inferring landscape scale leaf demography. We used a drone-derived orthomosaic covering 4.5 ha from which training and validation sets of pixel classes were extracted. Pixels classes were chosen to represent three foliar phenostages at the crown scale, needed to derived leaf demography -- recently flushed crowns, bare crowns, and crowns with mature to old leaves. Linear Discriminant Analysis (LDA) identified the corresponding pixel classes with high user´s accuracy (> 99%).