Figure 1. Partitioning of individual traits variance across four nested scales through Partvar procedure, using data collected from 2017 to 2020. All data were normalized (log-transformation), except for specific root mass values which have been square-root transformed. The significance of the variance components (* p < 0.05; n.s. p > 0.05) was assessed by bootstrapping. N = 1430.
All studied traits exhibited larger variations within sites than among sites. Time (year) and space (site) had few influence (both additive effects explaining less than 8% of variance in all traits), even when they were significant. Conversely, phylogeny explained most of the variance within sites (Fig. 1). Most traits (with the exception of internode mass and LDMC) exhibited higher trait variance at the intraspecific scale as compared with the interspecific one. The intraspecific variance of the trait was even maximal in the internode length (79.48%) and the maximum root length (65.52%).

Variations of most traits were affected by both abiotic and biotic variables

Table 2. Traits responses to abiotic and biotic variables of pond habitat. Traits were considered by category (all, MANOVA tests) or one by one (ANOVA tests). MANOVA and ANOVA tests were both performed based on the most probable models (see Materials and Methods for details), using data recorded in 2020. Bolded predictors had a significant effect on trait values. Water and sediment nutrient dimensions were each reduced to the first axis of a PCA analysis (Supporting information 2 and 3), referred as ‘water nutrients’ and ‘sediment nutrients’. To take the influence of species abundance within ponds into account, we also added the first-three axes of a FCA performed on species abundance (Supporting information 4). Among the tested variables, functional diversity, Pielou evenness, dissolved oxygen, and pond area did not have effects on any trait responses.