Conclusions

This review has examined the remote sensing of grassland biodiversity and six functional traits with a focus on recent technological and methodological developments. Advances in UAV technology have accelerated the increase in grasslands surveys featuring very high spatial resolutions, with 3D components and increasingly employing multispectral and hyperspectral sensors. At the same time, machine learning methods are becoming prevalent within the research community, requiring a strong understanding of the many parameters needed to construct a useful statistical or predictive model. While these developments open the door to new approaches and new discoveries, they also present new sources of error and uncertainty. This requires a more structured and systematic approach to investigating, documenting and addressing these issues. Utilising UAV surveys as a bridge between point-based groundwork and satellite remote sensing, helping to integrate measurements across spatial and temporal scales, is one step in this this process that can be implemented in many locations across the planet. This could be further enhanced by making better use of currently existing and future hyperspectral satellite platforms, such as EnMAP, PRISMA and the Firefly constellation. Finally, we stress the importance of a global database, similar to TRY (Kattge et al., 2011) of traits, species and related spectra from multi- and hyper-spectral devices, that could enhance the development of more robust, scalable and generalisable remote sensing models. This could then contribute to more accurate monitoring of grassland species diversity as well as functional traits.