Biodiversity from the Sky: Testing the Spectral Variation Hypothesis in the Brazilian Atlantic Forest

Abstract

Tropical forests have high species richness, being considered the most diverse and complex ecosystems in the world. Research on the variation and maintenance of biodiversity in these ecosystems is important for establishing conservation strategies. The main objective of this study was to test the Spectral Variation Hypothesis through associations between species diversity and richness measured in the field and hyperspectral data collected by a Remotely Piloted Aircraft (RPA) in areas with secondary tropical forest in the Brazilian Atlantic Forest biome. Specific objectives were to determine which dispersion measurements, standard deviation (SD) or coefficient of variation (CV), estimated for the n pixels occurring within each sampling unit, better explains species diversity; the effects of pixel size on the direction and intensity of this relationship; and the effects of shaded pixels within each sampling unit. The spectral variability hypothesis was confirmed for the Atlantic Forest biome, with R2 of 0.83 for species richness and 0.76 and 0.69 for the Shannon and Simpson diversity indices, respectively, using 1.0 m illuminated pixels. The dispersion (CV and SD) of hyperspectral bands were most strongly correlated with taxonomic diversity and richness in the red-edge and near-infrared (NIR) regions of the electromagnetic spectrum. Pixel size affected R2 values, which were higher for 1.0 m pixels (0.83) and lower for 10.0 m pixels (0.71). Additionally, illuminated pixels had higher R2 values than those under shadow effects. The main dispersion variables selected as metrics for regression models were mean CV, CV for the 726.7 nm band, and SD for the 742.3 and 933.4 nm bands. Our results suggest that spectral diversity can serve as a proxy for species diversity in the Atlantic Forest. However, factors that can affect this relationship, such as taxonomic and spectral diversity metrics used, pixel size, and shadow effects in images, should be considered.