On the relevance of intraspecific trait variability-A synthesis of 56 dry grassland sites across Europe

The intraspecific plant trait variability (ITV) is key for many ecological processes, but large-scale analysis of co-occurring species are rare. Here we studied ITV of 14 plant traits for five co-occurring species across 56 European dry grassland sites - large parts of the species distribution ranges. We evaluated variation in ITV relative to between species trait variability (BTV) with a particular emphasis on how within versus between population variability contributes to ITV. We performed this analysis trait by trait (univariate) and by considering hypervolumes in the multi-variate trait space. We also tested whether climatic variables can be used to predict between population variation in traits. For the trait by trait analysis, for 9 out of 14 traits ITV exceeded BTV, which suggests an extraordinary large role of ITV. However, when considering all traits jointly the hypervolume occupied by ITV was only 23% of overall trait variability. We found comparatively small effects of within population variability in the uni- as well as in the multivariate analysis decreasing rapidly with increasing number of considered traits. The dominance of between population variability suggests that ITV is mainly driven by environmental effects rather than local biotic interactions and microsite effects. ITV of leaf chemical traits was related to precipitation and growing degree days until sampling ITV can be substantial compared to BTV, in particular when considering single traits, such that ITV should be considered in trait-based research. Since the importance of ITV appears to decrease when considering multiple traits jointly, using species mean trait values is more appropriate for multi-variate trait analysis. Because ITV varied mainly between populations rather than within populations, and was related to climatic conditions, we suggest that ITV could be accounted for in future trait based research by including environmental covariates in hierarchical models of trait variation