|Aim: Mapping the distribution and diversity of plant functional traits is critical for projecting future changes to vegetation under global change. Maps of plant functional traits, however, are scarce due very sparse global trait data matrices. A potential solution to this data limitation is to utilize the known levels of phylogenetic signal in trait data to predict missing values. Here we aim to test existing phylogenetic comparative methods for imputing missing trait data for the purpose of producing continental-scale maps of plant functional traits.
Location: North America and Europe.
Methods: Phylogenetic imputation models and trait data from one continent were used to predict the trait values for tree species on the other continent and to produce trait maps. Predicted maps of trait means, variances and functional diversity were compared with known maps to quantify the degree to which predicted trait values could estimate spatial patterns of trait distributions and diversity.
Results: We show that the phylogenetic signal in plant functional trait data can be used to provide robust predictions of the geographical distribution of tree functional diversity. However, predictions for traits with little phylogenetic signal, such as maximum height, are error prone. Lastly, trait imputation methods based on phylogenetic generalized least squares tended to outperform those based on phylogenetic eigenvectors.
Main conclusions: It is possible to predict patterns of functional diversity across continental settings with novel species assemblages for most of the traits studied for which we have no direct trait information, thereby offering an effective method for overcoming a key data limitation in global change biology, macroecology and ecosystem modelling.|