Speaker
Description
Trait abundance distributions can help identify processes that determine species coexistence in diverse communities. Niche differences enable species’ “stable” coexistence by reducing interspecific competition and competitively excluding non-optimal species. However, species are favored in their persistence and abundance the more similar they are to optimal niche strategies, producing “trait clusters” in contrast with standard expectations of trait overdispersion. Evidence for such trait clusters as a signal of emergent niches remains limited, requiring improved methods. Here, we modify the Gaussian Mixture Model (GMM) and associated Expectation Maximization algorithm by adding species abundances in the likelihood function. We ground-truth this method on trait abundance distributions from community assembly simulations, combining stochastic processes and niche differentiation mechanisms like resource partitioning, abiotic filtering, etc. We then use it on the Barro Colorado Island (BCI) forest data on functional traits like tree height, leaf mass area, seed mass, and wood density. GMM’s flexibility to detect more uneven cluster shapes leads to more significant clustering than detected by previous K-means approaches, indicating a stronger signature of niche-differentiating processes acting on these traits. Thus, our study provides a new toolkit for detecting clustering patterns indicative of niche differentiation that is more generalizable across complex empirical data.
