Height layers. Visual representation of the forest on Barro Colorado Island, highlighting the four height-based groups identified by our metric. See paper for complete description. Illustration by John Megahan

Tropical forests challenge us to understand biodiversity, as numerous seemingly similar species persist on only a handful of shared resources. Biologists have puzzled over the forces allowing the coexistence of a surprisingly large number of tree species in tropical forests for many decades.  

In particular, they have debated over the extent to which coexistence is supported by species differences in resource needs and ecological strategies, known as niche differentiation, given the fact that coexistence can also, in theory, be supported by the opposite phenomenon, i.e. species ecological similarity.

It is thought that differences in strategy may allow trees to exploit different opportunities, enabling them to coexist stably, meaning that each species would be restored in the community if it becomes sparse due to some chance events. Alternatively, similarity in strategies between species may allow their long-term persistence together in the same habitat because none gets a competitive advantage over the other.

Rafael D’Andrea is first author of the paper published in Ecology, June 2, 2020. D’Andrea began this research in the lab of Professor Annette Ostling as a follow up on his doctoral work with her in the Department of Ecology and Evolutionary Biology at the University of Michigan. D’Andrea begins an assistant professorship this fall at Stony Brook University, New York.

The Ecology paper shows for the first time that trees in a tropical forest exhibit a clustering pattern. In a 50-hectare plot (the equivalent size of nearly 100 football fields) on Barro Colorado Island (BCI), Panama, species abundances exhibit clusters in two traits connected to a light capturing strategy, suggesting that competition for light structures community composition. Notably, the researchers found four groupings of species by maximum height, independently corroborating the classical qualitative grouping of Neotropical woody plants into shrubs, understory, midstory and canopy layers.

“By looking at a community large enough to encompass patches of different successional age, in a diverse forest, we are able to see the organization of species in the community based on traits such as height and wood density, a trait associated with shade tolerance,” said Ostling, senior author. “The pattern of groupings of species that we find suggests that both differences and similarities are playing a role in coexistence – differences are important to coexistence between groups, and similarities are important within groups.

“Though there is more work to do to show the pattern is more general than this forest, and to definitively show that the pattern we find is due to the role of these traits in shaping the competitive dynamics of trees, our result is an important early step,” said Ostling. “In the past, ecologists have focused on looking for patterns suggestive of the role of species' differences in the form of even spacing of species based on traits at small spatial scales, and had very limited results. Many of the ways that species can coexist through differences in strategy involve their interactions on a spatially heterogeneous landscape and hence their influence on the set of traits present in the community will not be visible at such small scales.”

Coauthors with D'Andrea and Ostling are EEB doctoral alumnus John Guittar, Hector Figueroa, a current EEB graduate student, and collaborators Joe Wright from the Smithsonian Tropical Research Institute, Richard Condit from the Field Museum of Natural History, and James O'Dwyer from the University of Illinois.

Currently, D’Andrea is working to better understand the conditions under which competition leads to the emergence of species trait clusters. D’Andrea and Ostling are planning to work collaboratively to investigate whether the pattern they found on BCI generalizes to other forests and how trait pattern depends on the spatial scale at which the community is examined. 

Ostling is also validating alternative cluster detection methods to the ones they used on the BCI data. She is exploring density-based methods, which are faster, and will better detect clusters involving multiple traits. "If we can develop faster and more robust methods, the next step will be to make them broadly available to ecologists.”

“We are also working to more definitively show that variation in traits like height and wood density in forests is linked with strategy differences that can actually enable competitive coexistence between tree species.”

For further information: The theoretical underpinnings of their Ecology paper were featured in a Science Trends piece by D'Andrea and can be found in a PLoS Computational Biology paper by D'Andrea, Ostling, and Maria Riolo, a former U-M doctoral student.  

Note: This work required intensive use of the U-M Flux computing cluster to implement the clustering algorithm on the data from Barro Colorado Island and its randomizations for computing statistical inference.