Remote sensing of biodiversity: Soil correction and data dimension reduction methods improve assessment of α-diversity (species richness) in prairie ecosystems

Abstract

Hyperspectral data, with their detailed spectral information at different wavelengths, offer multiple ways to assess biodiversity. One approach, known as the “spectral variation hypothesis” (SVH), proposes that biodiversity is linked to spectral diversity. However, SVH-based approaches, which we refer to as “spectral diversity metrics”, can be confounded by soil exposure and are sensitive to the spatial resolution of the data. To address these issues, we 1) investigated the impact of soil exposure on spectral diversity, 2) identified optimal bands for mapping biodiversity using a spectral diversity metric based on dimension reduction, and 3) assessed the impact of spatial resolution on spectral diversity metrics. In this study, α-diversity (species richness) was used as a measure of plant biodiversity. The study was based on two imaging spectrometry data sets from the Cedar Creek Ecosystem Science Reserve in Central Minnesota, USA, at two levels: proximal and airborne. The data sets included varying degrees of soil background sampled at two different spatial resolutions (1mm and 0.75m). We explored five spectral diversity metrics, including the coefficient of variation, convex hull volume, spectral angle mapper, spectral information divergence, and a newly proposed dimension reduction-based metric called “convex hull area.” For the proximal data set (pixel size of 1mm), filtering soil pixels by applying a normalized difference vegetation index (NDVI) threshold improved the performance of all spectral diversity metrics significantly, with the coefficient of variation showing the highest correlation with species richness. In the airborne data set (pixel size of 0.75m), the convex hull area outperformed other metrics. These findings demonstrate promising approaches for remote sensing of biodiversity, illustrate a confounding effect of soil background on remote diversity measurement, and indicate that the most informative regions of the electromagnetic spectrum for estimating species richness can vary with spatial scale.