Remote sensing from spaceborne, airborne, or proximal platforms has been widely used to assess biodiversity. A popular way to map biodiversity is spectral diversity, which links the spectral variation to biodiversity. This study explored the potential of airborne hyperspectral data (spatial resolution of 0.75 m) for mapping biodiversity in grassland habitats. The common challenge for mapping biodiversity in grasslands through airborne data is the coarse resolution of the data where the pixel size can be larger than the canopy size. In such cases, it is unclear what a spectral diversity metric sees. Furthermore, spectral diversity can be confounded by spectral indicators of productivity, which are known to be related to biodiversity. We tested the degree to which spectral diversity captures biodiversity and productivity. The predictive power of spectral diversity was assessed at seven different plot sizes: 3 × 3 m, 9 × 9 m, 15 × 15 m, 30 × 30 m, 45 × 45 m, 60 × 60 m, and 75 × 75 m. Our analysis was done using species richness and biomass data from plots under different management regimes: a manipulated experiment and a natural grassland, both located at the Cedar Creek Ecosystem Science Reserve in Minnesota, USA.
Results/Conclusions
The results of a causal mediation analysis showed that for all plot sizes in the manipulated experiment, the spectral diversity was driven by biomass while in the natural grassland, species richness was a more important factor affecting spectral diversity. The biomass effect in the manipulated experiment can be attributed to the design and maintenance of the experiment, where plots are heavily weeded every year and those plots with lower species richness are mostly covered with soil, leading to a strong biomass effect on the spectral variation. Our findings reflect how remote sensing of biodiversity in manipulated experiments can be affected by management practices. In case of such manipulations, “remote sensing of biodiversity” via medium- and coarse-resolution data might become less meaningful, and productivity can strongly influence estimated biodiversity. To achieve operational airborne and spaceborne biodiversity monitoring systems, we suggest establishing well-designed experiments from a remote sensing perspective. Such experiments should be designed to resemble natural landscapes by minimizing human-caused manipulations, include both biodiversity and biomass as experimental treatments, and consider the mismatch between the spatial resolution of the remote sensing sensor, plot size, and plant stature.