Using simulated grassland communities and radiative transfer models to test the Spectral Variation Hypothesis

Abstract

Increasing land use and climatic change lead to a global decline of biodiversity at alarming rates. To counteract this massive loss, global aims such as the convention of biological diversity’s Aichi targets and related conservation programs have been launched. These programs require a detailed monitoring of biodiversity across large areas, which raises high expectation with respect to biodiversity assessments from Earth Observation data. One frequently discussed approach is an application of the so called Spectral Variation Hypothesis (SVH). This approach aims to link the spectral variation of remotely-sensed image data to environmental heterogeneity as the main driver of species diversity in a given area. According to the SVH, diversity in leaf and canopy optical properties and habitat structures increase with increasing species diversity what in turn drives variations in the spectral signature of the plant communities. Various studies that explore these correlations in terrestrial ecosystems come to promising conclusions. However, the transferability of the proposed relations between spectral and taxonomic diversity to other ecosystem types and across different spatial resolutions remains unclear. Especially for grasslands where the mismatch between pixel and individual plant size is heavily pronounced, no comprehensive study has systematically tested the SVH yet. To fill this gap, we developed a theoretical framework that enables the simulation of realistic reflectance patterns of grassland vegetation. Thereby, we can mimic the spectral signal hypothetically reaching different sensor systems. Moreover, it allows us to test the relationships between spectral variation and taxonomic diversity for a high number of simulated plant communities. We created spatial distributions of artificial grassland units based on species inventories and trait data that we sampled in the field. We further simulated the spectral signature of these artificial communities using the leaf and canopy RTM PROSAIL. By including in-situ plant traits (species-, site- and season-specific, sampled on the individual plant level) we 1) simulate realistic reflection profiles which also incorporate seasonal dynamics, 2) modify species composition and species richness, and 3) use this as the basis to assess scaling effects. The modelling framework will be presented as well as the results of the spatial plant community simulations including the generated spectral patterns for three sites and seasons. Further, first results regarding the spectral-to-taxonomic diversity relationship will be discussed.