Linking multivariate trait variation to the environment: the advantages of double constrained correspondence analysis with the R package douconca

Trait-based ecology aims to predict changes in species composition and traits by understanding how species distribution and success are influenced by their traits and environmental conditions. It extends joint species distribution modelling and species-environment relationship studies by incorporating trait data. Traditional methods include RLQ ordination, fourth-corner correlation analysis, and community-weighted means (CWM) regression, with generalized linear mixed models (GLMM) also being used. Through three case studies, we demonstrate that double-constrained correspondence analysis (dc-CA), an ordination method with biplot visualization, is a competitive alternative. Unlike RLQ, dc-CA is regression-based, allowing for model comparison and prediction. It extends CWM-regression and fourth-corner analysis by constructing trait and environmental composites that best explain abundance data and maximizes fourth-corner correlation. In the first case study using Dune Meadow data, dc-CA showed that ecological traits (Ellenberg indicator values) are more closely related to environmental characteristics than functional traits, a conclusion difficult to draw with RLQ. The second case study re-analyses trait-environment relationships in Wisconsin forest understorey vegetation, previously examined by GLMM and fourth-corner correlation analysis. Here, dc-CA identified the leaf economics spectrum, which was less apparent in GLMM. In the third case study on Amazonian and Atlantic forest trees in relation to landscape configuration across six regions, dc-CA summarized multiple CWM-regressions in a single diagram. All three case studies were analysed with the R package douconca, which simplifies the use of dc-CA and enhances trait-environment analysis. Two of the three cases revealed multi-dimensional species-environment relationships, with functional traits explaining only a single dimension, posing the challenge of identifying traits that account for other dimensions. • We analyse functional β-diversity using double constrained-Correspondence Analysis. • Trait composites from dc-CA formalise trait trade-offs and syndromes. • Case studies demonstrate dc-CA's effectiveness in creating trait composites. • In two case studies, functional traits defined only one dimension, not the others. • R package douconca simplifies dc-CA use and enhances trait-environment analysis.