Normalizing functional diversity metrics across heterogeneous data streams: from field to satellite data

Fighting the current biodiversity crisis requires monitoring systems able to determine ecosystems’ biodiversity at the global scale systematically. Resource-demanding field surveys are fundamental, but they are unable to provide continuous coverage in space and time. Recent studies have shown that remote sensing theoretically has the potential to overcome this limitation and is thus becoming a promising tool for biodiversity monitoring.

The present and upcoming fleet of Earth observation satellites offer wide variability of resolutions and spectral information that could be jointly exploited to map plant functional diversity robustly. However, this heterogeneity hampers the comparability of functional diversity metrics inferred from different sensors because their values depend on the trait-space dimensionality (e.g., the number of spectral bands). This problem is also inherent to comparing metrics computed from satellite imagery and field data or field surveys sampling different traits. Such dependency hides the actual information contained in the metrics and may mislead interpretation. Here we present a global normalization approach that removes the effect of dimensionality from functional diversity metrics such as Rao’s quadratic entropy index (Rao Q), allowing the computation of its equivalent number from independently processed imagery.

The method outperforms image-based normalization and set metrics computed from the heterogeneous field and remote sensing datasets at the same scale. This enhanced comparability reveals the differences in diversity information related to trait selection and spatial resolution between the different data streams. We expect this new method to become broadly used in remote sensing, facilitating the integration of multiple missions and the validation of functional diversity products with field data.