2022 ESA Annual Meeting (August 14 - 19)

COS 170-4 Quantifying niche differentiation by soil condition in tropical forests

2:15 PM-2:30 PM
514A
Rafael D'Andrea, Stony Brook University;Mihir Umarani,Stony Brook University;Dianzhuo Wang,Harvard University;James O'Dwyer,University of Illinois at Urbana-Champaign;
Background/Question/Methods

The proverbially prodigal biodiversity of tropical forests tasks community ecologists with explaining how such diversity is structured and maintained. While trees must compete for soil nutrients and water, no species seems to win. This could be because different species specialize on different microhabitats—e.g., different local soil conditions such as high vs low nutrient, pH, and water levels. In forests where such microhabitats are temporally stable and spatially heterogeneous, this type of ecological niche differentiation should be reflected in the spatial distribution of trees. Namely, species with similar “soil niches” should occur in spatial proximity. Here, we compare spatial patterns in the location of trees with the spatial distribution of elemental nutrients in three censused forest dynamic plots: Barro Colorado Island (BCI) (Panama, 1000 m x 500 m), Yasuní (Ecuador, 500 m x 500 m), and La Planada (Colombia, 500 m x 500 m). We measure the degree to which a) certain species are spatially associated with each other, b) particular species groups associate with particular nutrient levels, thus reflecting the existence of discrete soil niches, c) species tend to recruit in their own soil niche rather than in other soil conditions.

Results/Conclusions

In all three plots, species fell into distinctive groups characterized by mutual spatial proximity (< 20 m). Both the number and spatial distribution of such groups were consistent across censuses, and consistent with the spatial distribution of soil nutrients in their respective forest plots. A decision-tree algorithm trained on a sample of the data correctly predicted the soil niche of the trees for 85% of out-of-sample data on BCI, and 70% on La Planada and Yasuní, an outcome which greatly exceeded null-model expectations. On BCI, where there was high covariation between soil nutrients, one of the four species groups predominantly occurred in sites with high levels of nutrients (Cu, Fe, Zn, Mg, Ca), whereas the remaining three differed by their respective associations with aluminum and phosphorus levels. Overall, trees were twice as likely to recruit in their own soil niche than in sites with different soil conditions.