Spectrally Defined Plant Functional Types Adequately Capture Multidimensional Trait Variation in Herbaceous Communities

Abstract

Our ability to measure plant characteristics across space and time is crucial for understanding and tracking the diversity and functioning of ecosystems. Ecological approaches to synthesize these characteristics have evolved from allocating species to predefined conventional plant functional types (cPFTs) to describing vegetation through delineating trait-based emergent plant functional types (ePFTs). At the same time, the remote sensing community has advanced in developing tools to measure functional traits, and defining plant optical types (POTs) from reflectance. However, the application of POTs is still underdeveloped and only few studies readily addressed the functional mechanisms underlying plant grouping. Especially at small ecological scales, relevant for studying community assembly processes and functional diversity, this is important because of scale-dependent sensitivity to drivers of trait variation. In this study, we therefore aimed to translate the ecological concept of ePFT delineation to spectrally measured data. We propose to cluster species based on traits estimated from reflectance (optical traits) to delineate emergent plant optical types (ePOTs). We conducted our study in herbaceous vegetation, where we examined four cPFT schemes, and measured the reflectance and six functional traits (four indicative of the leaf economics spectrum and two size-related) of 39 species across 11 sites (total of 73 site-specific species measurements). By means of PLSR models optical traits were retrieved with moderate to good accuracy (normalized RMSECV ranging between 4 and 22%, and R2CV ranging between 0.24 and 0.62). Subsequent agglomerative hierarchical clustering based on functional and optical traits resulted in 10 ePFTs and 9 ePOTs respectively, that well aligned (entanglement = 0.15). Whereas the four cPFT schemes poorly captured the multidimensional functional trait variation entailed in the dataset (R2 ranging between 0.05 and 0.43), the ePFTs and ePOTs much better represented functional trait (R2 = 0.70 and 0.52 respectively) and optical trait variation (R2 = 0.52 and 0.67 respectively). The presented ePOT delineation approach, combining spectrally measured data with standard ecological statistical approaches through optical traits, shows that spectral data can be used to delineate functional groups, while ensuring direct ecological interpretation. The method can easily be extended to different sets of traits to investigate specific responses and impacts of a plant community, and applied to larger spatial scales to study ecosystem functioning and biodiversity.