Abstract
Questions: One central assumption of trait screening approaches in comparative plant ecology, i.e. simultaneous measurement of traits on a large number of species or populations, is that the species level captures a major part of trait variation. The current development of large databases has led to a new screening approach that relies on the extraction of trait values from databases, rather than on measurement of traits in the field. We tested this assumption with the following questions: (1) is the magnitude of intra-specific variability of co-occurring species lower than inter-specific variability for a given trait, in comparisons at different spatial scales; (2) is species hierarchy based on trait values conserved across different spatial scales and data sets (stable species hierarchy hypothesis); and (3) when we compare different traits, what is the more stable trait that is conserved across different spatial scales and data sets? Methods: We combined approaches commonly used in functional ecology, i.e. experimental data, field observations and extraction of data from a global database, and analysed the magnitude of intra-specific and inter-specific trait variations for a large number of traits across contrasting environmental conditions for 18-39 (mostly) herbaceous species, according to the data set used. Results: For most traits, inter-specific variability was higher than intra-specific variability, and species ranking was conserved across different data sets and spatial scales. However, we also detected important differential responses in terms of intra-specific trait variability, depending on the trait examined: SLA, LDMC, SM, seed N concentration and onset of flowering were more stable, whereas leaf chemical traits and RH were more flexible traits. Conclusions: Our study validated, for the species studied, the stable species hierarchy hypothesis in the case of several, but not all, widely used traits. The main conclusion is that the strength of the species signal is strong enough for some traits to allow values to be used from different data sets (experiments, databases) to characterize local populations of species: for SM, seed N concentration, RH, SLA and LDMC.