Résumé
1. Investigating the functional facet of biodiversity provides ecologists with a deeper understanding of community assembly and ecosystem processes, from local to biogeographical scales. A central assumption in functional ecology is that interspecific trait variability is higher than intraspecific variability. The stable species hierarchy hypothesis states that for similar species found in different environmental conditions, their species trait ranking is conserved. In this study, we applied this trait hierarchy concept prevalent in plant ecology to the growing field of soil functional ecology, for which newly developed trait databases are being increasingly used. However, to date there have been few attempts to test for patterns of intraspecific trait variability in these databases. 2. We thus aimed to characterize how such patterns might influence (i) a species hierarchy based on trait values, and (ii) the conclusions of a trait-based analysis at a community level. To examine this, we used Collembola body size data (extracted from the BETSI database) as model trait. The source consisted of four regional trait datasets (Poland, Scandinavia, Spain and UK) and one dataset for which species traits are defined at a continental (European) scale. 3. We found that, firstly, species were consistently ranked in all the trait datasets, although slight differences were observed between continental and northern European (i.e. Scandinavia and UK) trait datasets. In the two northern datasets, body size was higher (ca. 10%), indicating an intraspecific body size gradient from temperate to colder northern regions that we assumed could be explained by latitudinal patterns. 4. Secondly, using selected published species abundance matrices (from experimental studies), we calculated the community-weighted mean body size using various trait datasets. The findings showed that the slight discrepancies observed between trait datasets can lead to different conclusions. 5. This work confirms that properly defining the extent of intraspecific trait variability in databases is of primary importance in order to ensure robust conclusions. This is particularly important for databases hosting large scale data, that might be influenced by biogeographical patterns as latitudinal gradients. We recommend using a local regional trait dataset when available or, if not, a continental trait dataset. As trait databases are now commonly used tools for performing trait-based analyses, it is crucial to carefully select the data used to make inferences.