Network-informed analysis of a multivariate trait-space reveals optimal trait selection.

Abstract

Trait-based analyses have shown great potential to advance our understanding of terrestrial ecosystem processes and functions. However, challenges remain in adequately synthesising a multidimensional and covarying trait space. Reducing the number of studied traits while identifying the most informative ones is increasingly recognized as a priority in functional ecology. Here, we develop a trait reduction procedure based on network analysis of a global dataset comprising 27 traits in three steps. We first construct all possible reduced networks and identify optimal reduced networks that capture the structure of the full 27-trait network. Then we apply the constraints on trait consistency to identified optimal reduced networks and establish consistent network series across ecoregions. We find the best performing networks that capture the three main dimensions of the full network (hydrological safety, leaf economic strategy, and plant reproduction and competition) and the global variance of network metrics. Finally, we find a parsimonious representation of trait covariation strategies is achieved by a 10-trait network which preserves 60% of all the original information while costing only 20.1% of the full suite of traits. Our results show the network reduction approach can improve our understanding on the main plant strategies and facilitate the future trait-based research.