Remotely sensed spectral variability predicts reef fish diversity

• Underwater spectral variability can be calculated from satellite and drone data. • This strategy applies across a global dataset, despite varying ecologies. • Image resolution, depth, fish feeding strategy affects spectral-fish correlation. • High-resolution fluid-lensed images gave the best results for this hypothesis. • High-resolution imagery does not correlate with fish diversity if not fluid-lensed. In terrestrial landscapes, the spectral variability hypothesis (SVH) enables estimation of species diversity from satellite data, thereby allowing biodiversity assessments to be upscaled. Whether the SVH works in the marine realm is an open question. To answer it, we tested the ability of this hypothesis to retrieve coral reef fish biodiversity from two remote sensing platforms on a global transect of reef sites. From orbit, we trialed the multispectral and panchromatic bands of WorldView-2 (WV-2) which have a spatial resolution of 2.5 and 0.5 m, respectively. At 100 times finer resolution, we repeated the experiment using unpiloted aerial vehicle (UAV) data. Encouragingly, the SVH evidently works as well in water as has been reported on land. Spectral variability was positively correlated with fish diversity for all sensors, though the area in which the variability was computed (window size) was important. The strongest relationship between spectral variability and fish biodiversity (R = 0.48) was returned using UAV imagery corrected for surface artifacts via fluid lensing. Splitting fish into herbivores, corallivores, and piscivores revealed that different feeding strategies correlate to spectral variability at different scales. Based on our results, we contend that remote sensing data are underutilized when used to simply map benthic habitat. Spectral variation can clearly serve as a proxy for in situ reef biodiversity.