Mapping woody plant community turnover with space‐borne hyperspectral data – a case study in the Cerrado

Abstract

Effective conservation measures require the knowledge on the spatial patterns of species communities and their turnover. This knowledge is, however, many times lacking, particularly so for complex systems. On the other hand, recent developments have resulted in tools that enable the mapping of these patterns from remote sensing data, such as Sparse Generalized Dissimilarity Modelling (SGDM). SGDM is a two-stage approach, which combines a Sparse Canonical Component analysis and a Generalized Dissimilarity Modelling (GDM), thus being designed to deal with high-dimensional data to predict community turnover in GDM. In this study, we use space-borne hyperspectral data to map woody plant community patterns collected in two study sites in the Cerrado (Brazilian savannah), namely, the Parque Estadual da Serra Azul (PESA) in Mato Grosso state and Parque Nacional da Chapada dos Veadeiros (PNCV) in Goiás state. Field data were collected in both study sites, following a systematic sampling scheme adapted for the Cerrado. The Cerrado is the most diverse of all the world’s savannahs, and while holding a high diversity and endemism of species, this biome is mostly unprotected and understudied. We used Hyperion data acquired over the two study sites, which were subject to data pre-processing (including radiometric and geometric corrections, as well as correction for sensor errors) and quality screening before analysis. Our models were used to map woody plant community patterns and turnover for the study areas. We also inspected the Hyperion spectral bands which most contributed in the SGDM, for each site. Furthermore, the modelled patterns were interpreted with respect to the ecological characteristics of the respective species, this way further enhancing our understanding of this complex system. This study has demonstrated that this approach is suitable for mapping woody plant communities in heterogeneous systems, based on combined field and space-borne hyperspectral data.