Applying the spectral variability hypothesis to arid shrublands using multispectral drone imagery

Arid biodiversity monitoring is required to assess ecosystem function and inform land management. The spectral variability hypothesis (SVH) posits that spectral diversity captured by remote sensors is related to plant biodiversity, offering a method for scaling biodiversity assessments at lower costs. However, the SVH has not been applied in arid shrublands, meaning optimal implementation choices are unclear.We applied both traditional, on-the-ground sampling and drone-based spectral analysis to test the SVH and associated method choices for arid shrublands. We considered continuous and categorical spectral metrics and evaluated whether metric performance was improved when applied to specific spectral band combinations.Spectral diversity metrics with less sensitivity to outlier pixels, namely co-efficient of variance (CV), were most successful at predicting taxonomic diversity using high spatial resolution multispectral drone imagery. Performance of the CV metric was greatest when applied to red edge and near infrared bands compared to the five-band dataset (R2 = 0.43).Multispectral drones offer a low-cost and accessible methodology for environmental monitoring in vast and poorly monitored arid regions. Our results show promise for arid biodiversity monitoring using drone technology and offer directions for implementation. To facilitate further work in arid systems and beyond, we introduce the saltbush R package.