Abstract
<b>Aim</b> Understanding how combinations of ecological traits at the community‐level vary with environmental conditions is crucial to anticipate and respond to the biodiversity crisis. While this topic is popular, most attempts to analyse and predict multiple traits in space and time ignore the inherent correlations between these traits. In doing so, the predicted traits in unobserved environments are likely to be flawed (i.e. unrealistic trait combinations). Here, we propose a framework that addresses this methodological question in functional biogeography. <br /><b>Innovation</b> Our framework relies on joint trait distribution models to quantify how combinations of mean trait values at the community level co‐vary as a function of environmental conditions. Making use of joint probabilities and the constraints imposed by trait correlations, our framework not only predicts the most suitable combination of traits in a given environment but also the envelope of possible alternatives. This innovation allows visualizing how the correlation structure between traits imposes a strong constraint (or not) on the expected combination of traits in a given environment. The stronger trait correlations, the smaller the envelope of likely alternative states. We provide an R package implementing our framework. <br /><b>Main Conclusions</b> Applied to plant communities in the French Alps, our framework quantifies the variation of trait combinations at the community‐level along environmental gradients that, otherwise, could not be identified. Strong correlations in community mean leaf traits led to a low diversity of alternative combinations of community means, which can only vary along the acquisition‐conservation spectrum. Instead, height varied partly independently of leaf traits, leading to a higher diversity in the combinations of mean community traits. Our framework allows for a more integrated understanding and prediction of the distribution of traits, thus moving a descriptive functional biogeography to a more predictive science.