Plant spectral diversity from high-resolution multispectral imagery detects functional diversity patterns in coastal dune communities

Abstract

Remote sensing is a fundamental tool to monitor biodiversity over large spatial extents. However, it is still not clear whether spectral diversity (SD - variation of spectral response across a set of pixels) may represent a fast and reliable proxy for different biodiversity facets such as taxonomic (TD) and functional diversity (FD) across different spatial scales. We used fine resolution (3 cm) multispectral imagery on coastal dune communities in Italy to explore SD patterns across spatial scales and assess SD relationships with TD and FD along the environmental gradient. We measured TD as species richness, while SD and FD were computed using probability densities functions based on pixels and species position in multivariate spaces based on pixel values and traits, respectively. We assessed how SD is related to TD and FD, we compared SD and FD patterns in multivariate space occupation, and we explored diversity patterns across spatial scales using additive partitioning (i.e., plot, transect, and study area). We found a strong correspondence between the patterns of occupation of the functional and spectral spaces and significant relationships were found along the environmental gradient. TD showed no significant relationships with SD. However, TD and SD showed higher variation at broader scale while most of FD variation occurred at plot level. By measuring FD and SD with a common methodological framework, we demonstrate the potential of SD in approximating functional patterns in plant communities. We show that SD can retrieve information about FD at very small scale, which would otherwise require very intensive sampling efforts. Overall, we show that SD retrieved using high-resolution images is able to capture different aspects of FD, so that the occupation of the spectral space is analogous to the occupation of the functional space. Studying the occupation of both spectral and functional space brings a more comprehensive understanding of the factors that influence the distribution and abundance of plant species across environmental gradients.