Fig 1.
Schematic representation of niche models.
Panel A illustrates Lotka-Volterra competition, while panels B-D illustrate different niche mechanisms, with the niche axis represented on the abscissa. A. Lotka-Volterra: competition coefficients are tied to traits, here beak size. Curve shows competitive impact by each species on the focal species, increasing with trait similarity. B. Partitioning resources: We assume a continuum of resources (seeds), and species traits (beak size) determine their use of the different resources. Depending on the degree of specialization, resource depletion may occur (see text). C. Partitioning environment: species with different traits (seed size) are optimally adapted to different environments. This model is spatially explicit, and we implement it both with and without dispersal limitation; D. Competition-colonization tradeoff: species trade off fecundity (seed output) with ability to win sites against other species (mediated by seed size). Seed icons adapted from design by Brgfx—Freepik.com.
Fig 2.
Abundance-by-trait pattern across niche mechanisms.
For each scenario (A-H) we show one representative replicate. All communities with a niche mechanism (A-G) are clustered at the p < 0.05 level, while the neutral community (H) is not (p = 0.1). Alternating black and red colors highlight the clusters. We truncate the y-axis in the competition-colonization tradeoff scenario (G) to better show abundance structure among rare species. (Immigration rate m = 0.08, regional diversity c. 400 species. All communities had 21,000 individuals, except in the partitioning environment scenario, which had 1,000 individuals. Parameters of niche models were tuned so that without stochasticity and immigration they would produce about 13 transient clusters, and eventually 13 species stably coexisting at equilibrium).
Fig 3.
Clustering across niche mechanisms.
Bars show percentage of replicates that were significantly clustered at level p < 0.05 according to the k-means metric (blue) and the Ripley’s K metric (orange), out of 100 replicates. Numbers next to bars show average z-score. For comparison, 8% of neutral communities were clustered by the k-means metric (), and 7% by the Ripley’s K metric (
), thus close to the 5% background detection expected from a null model (z-scores and p-values were obtained by comparing each replicate against 100 null communities. m = 0.08, c. 400 regional species across scenarios).
Fig 4.
Impact of immigration and regional diversity.
Clustering increases with regional diversity (black), and has a modal relationship with immigration pressure (red). Bottom axis: number of species in the pool divided by the number of clusters in the community. Top axis: immigration rate. Points show the mean z-score of the Ripley’s K metric across 10 replicates. Error bars show 1.96 standard errors of the mean. (m = 0.08 when varying regional diversity, and the latter is fixed at c. 400 regional species when varying immigration).