Figure 2.

Box 1 Figure ‐ The spectral species algorithm phases. The original image was acquired with the CAO AToMS imaging spectrometer during an airborne campaign over the CICRA experimental site (Amazonian Peru) (https://www.amazonconservation.org/about/mission-vision/cicra-station/). The first image (a) corresponds to the RGB representation of an imaging spectroscopy subset. A standardized PCA is applied on (a) and a reduced set of components is selected (b) to maximize signal corresponding to biological patterns on forested areas and discard noisy components. Spectral species are defined for each pixel by applying an unsupervised k‐means clustering on the spectral space defined by selected components (c). In this phase, a field survey recognition based on in situ data is crucial to define the number of singular spectral signatures (spectral species) expected. The spectral species map is divided into elementary spatial units and the spectral species inventory is performed for each spatial unit, by further calculating Shannon's H and Bray‐Curtis metrics to derive (d) alpha‐ (ranging here from minima to maxima from black to blue, green and red) and (e) beta‐diversity (in which colors represent differences among spectral species) maps, respectively.