Search

An effective way to reduce complexity in ecological modelling is by grouping species that share similar characteristics into functional groups or types. Often, the creation of plant functional groups (PFGs) is carried out for each case study in an ad-hoc way using a small set of traits. This limits the transferability of these PFGs to other geographical areas or study systems. We propose a novel generic framework to generate PFGs that considers the most important ecological dimensions, is applicable to case studies globally, and that emerges from patterns of functional redundancy across species. Based on most relevant and measured plant characteristics, we designed a multi-step process that includes: i) data harmonisation and missing values imputation; ii) species clustering based on multiple characteristics encompassing the main ecological dimensions featured in plant community ecological models (i.e., dispersal, competition, and demography) and iii) the combination of ecological dimension-specific groups into comprehensive PFGs. We demonstrate this framework by applying it to a global dataset of plant characteristics including a functional traits dataset and a plant-soil co-occurrence dataset for 19,102 species. Lastly, to test the ability of generated PFGs to summarise species’ functional variation within plant communities, we correlate taxonomical and functional diversity indices calculated at the species and at the PFGs level across a global dataset of plant communities (sPlotOpen). Our framework generated 465 global, robust data-driven PFGs with non-overlapping combinations of traits for each ecological dimension divided by growth form. The validation returned positive correlation values between PFGs and species-level diversity metrics, supporting the ability of the obtained PFGs to capture functional and taxonomic diversity patterns across a variety of plant communities worldwide. The framework allows for the easy integration of newly available species characteristics data. The obtained global PFGs, covering all main known ecological processes and environmental conditions at small resolution, can increase the predictive power and accuracy of process-based models and help furthering varying-scale ecological studies.