Using trait data improves correlation between environment and community data only if abundances are considered

A straightforward way to explore variation between communities is to calculate dissimilarity indices and relate them with environmental and spatial variables. Communities are most often represented by the (relative) abundances of taxa they comprise; however, more recently, the distribution of traits of organisms included in the communities has been shown more strongly related to ecosystem properties. In this study, we test whether taxon- or trait-based dissimilarity is correlated more tightly with environmental difference and geographical distance and how the abundance scale influences this correlation. Our study system is grassland vegetation in Hungary, where we sampled vegetation plots spanning a long productivity gradient from open dry grasslands to marshes in three sites. We considered three traits for vascular plants: canopy height, specific leaf area and seed mass. We obtained field estimates of normalized vegetation difference index (NDVI) as proxy of productivity (water availability) for each plot. We calculated between-community dissimilarities using a taxon-based and a trait-based index, using raw and square-root transformed abundances and presence/absence data. We fitted distance-based redundancy analysis models with NDVI difference and geographical distance on the dissimilarity matrices and evaluated them using variance partitioning. Then, using the pooled data, we calculated non-metric multidimensional scaling ordinations (NMDS) from all types of dissimilarity matrices and made pairwise comparisons using Procrustes analysis. Data analysis was done separately for the three sites. We found that taxonomical dissimilarity matches environmental and spatial variables better when presence/absence data is used instead of abundance. This pattern was mainly determined by the increasing variation explained by space at the presence/absence scale. In contrast to this trend, with trait-based dissimilarity, accounting for abundance increased explained variation significantly due to the higher explanatory power of NDVI. With abundance data, considering traits improved environmental matching to a great extent in comparison with taxonomical information. However, with presence/absence data, traits brought no advantage over taxon-based dissimilarity in any respect. Changing the abundance scale caused larger difference between ordinations in the case of trait-based dissimilarity than with taxonomical dissimilarity. We conclude that considering relevant traits improves environmental matching only if abundances are also accounted for. Supporting informationAdditional graphs supporting the results are presented as appendix. Open researchData used in this research are publicly available from Dryad ###link to be supplied upon acceptance###