not-yet-known not-yet-known not-yet-known unknown 1 Innovations, limitations and suggestions for future approaches Significant progress has been made in recent years across a range of remote sensing technologies. The spectral, spatial and temporal resolution of satellite-based sensors are improving and much of the data now available free of charge. However, this does not appear to have translated into a significant increase in studies with a focus on grasslands. Indeed, it may be the case that greater temporal resolution, of instruments such as MODIS, does not make up for their moderate spatial resolution (Schmidtlein and Fassnacht, 2017). Even the higher resolution SPOT 5 and Sentinel 2 series have produced mixed results for biodiversity measurements, showing that much more work needs to be done to utilise this data source (Fauvel et al., 2020; Lopes et al., 2017). Furthermore, the lack of hyperspectral spaceborne instruments places a further limit on methods using the SV hypothesis. Future hyperspectral satellite launches, such as the CHIME (Copernicus Hyperspectral Imaging Mission for the Environment), from the European Space Agency, may help to address this issue. Meanwhile, UAV technology is rapidly developing. The ability of UAVs to carry both hyperspectral and multispectral instruments means that very high-resolution data, even down to the mm scale, can now be captured over increasingly large areas through UAV surveys. Simultaneously, optical images can be used to create point clouds through SfM, adding a 3D component capable of broadening the range of variables measured and classification accuracy. These factors have led to a rapidly growing number of studies utilising UAV technology, often alongside traditional field-based surveys. Conversely, as many of the methods involving UAVs are being actively developed and have not yet reached maturity, best practices are still being developed and the limitations and sources of error are still being discovered and analysed. UAV data may also prove useful for refining and truthing satellite-based imagery, potentially improving the relationship between SV measurements from space and grassland biodiversity on the ground, allowing for the development of national or global grassland biodiversity observations. To summarise, below are some of the key innovations, considerations and limitations present in the reviewed literature. Within remote sensing of biodiversity: The effects of bare soil, dead and live biomass and the vertical complexity of the vegetation structures need to be considered when analysing the data (Conti et al., 2021; Gholizadeh et al., 2018; Mansour et al., 2015; Rossi et al., 2022). Different data dimensionality reductions should be tested for hyperspectral data, and the SV compared with a range of species diversity indices, while keeping in mind that the relationship between spectral and species diversity may not be linear and can reach a saturation point (Gholizadeh et al., 2018; Wang et al., 2018; Zhao et al., 2021b). The highest spatial resolution data tends to work best and can even be used for species identification (Lopes et al., 2017; Wang et al., 2018). Seasonality can influence survey results, and the addition of spaceborne radar does not appear to offer additional benefits (Fauvel et al., 2020; Gholizadeh et al. 2020; Shoko et al., 2020). For remote sensing of functional traits: Remote sensing of functional traits typically captures community-weighted mean values, rather than species specific traits (Li et al., 2018; Zhang et al., 2022). Multispectral data works better with physical rather than chemical traits, while both hyperspectral and multispectral results may be improved by incorporating 3D data (Aasen et al., 2015; Grüner at al., 2021; Nasi et al., 2018; Tang et al., 2021). Regression and machine learning accuracy varies based on species type and diversity, so a range of methods need to be tested to ensure the most accurate results (Grüner at al., 2020; Wang et al., 2019; Wijesingha et al., 2020; Zhang et al., 2022). Spectral diversity may also be related to functional trait diversity, not just species diversity, and thus functional diversity may also be used to predict species diversity (Zhao et al., 2021b). 1 Innovations, limitations and suggestions for future approaches