Spatial patterns from a dispersal limitation perspective: revealing biotic interactions

Abstract

Local spatial distributions of populations are often studied in comparison to Complete Spatial Randomness (CSR) and are found to be ubiquitously aggregated, likely due to dispersal limitation. Here we theoretically examine the advantages of comparing observed distributions to simulated populations subject only to drift and Dispersal Limitation (DL). Compared to this DL null, local competition produces overdispersion out to surprisingly large scales, much larger than the scale of competitive interactions. Furthermore, strong overdispersion provides a hallmark that a key requirement of stable coexistence is met, as it can only be observed if intraspecific competition is substantially stronger than interspecific competition. Dispersion compared to CSR is insensitive to competition and as a result unreflective of its presence. Hence, we suggest DL as a complement to CSR since the former focuses on biologically relevant spatial scales and has the potential to detect biotic interactions and habitat specificity.

Pervasive species-specific repulsion among adult tropical trees

“…While we expect overdispersion to occur primarily at the short distances at which CNDD operates 3,5,8 , species are typically overdispersed at distances up to ~100 meters, having ~ 45% fewer neighbors 75 – 125 meters away from a typical tree (under the Standard null, but this is not the case under all null variants, Figure 2, Extended Data Table 1). Our recent simulation study shows that such long-distance overdispersion can be observed in a model with short-distance CNDD 42 . Intuitively, while each tree “repels” only its neighbors, they in turn repel their neighbors and so forth, creating large-scale patterns from short-distance interactions 43 .…”

“…Intuitively, when HNDD = CNDD the competitive effect of neighbors is equal for all neighbors regardless of their species, and so competition cannot generate a strong repulsion from conspecifics. Moreover, in a theoretical study of simulations across a wide range of parameter space of the strengths of CNDD and HNDD 42 , we found that large values of HNDD "erode" the species-specific overdispersion created by CNDD even when CNDD > HNDD. Hence, only the scenario CNDD >> HNDD is compatible with patterns of overdispersion observed here.…”

“…What underlies this widespread, strong and robust (under different dispersal kernels and alternative assumptions of the null models) overdispersion? Theoretical studies show that while aggregation can be caused by dispersal limitation 14,25,39 , habitat specificity 40 and facilitation 41 , overdispersion is driven by competition, causing density dependence 19,39,42 . Importantly, we show here using a simulation model that the observed overdispersion cannot be created by competition if CNDD and HNDD are similar in magnitude.…”

“…Our work is the first, to our knowledge, to test this classic hypothesis in multiple species and bring evidence that density dependence has a substantial effect on adult distribution and is much stronger among conspecifics than among heterospecifics. Since spatial overdispersion is the mark of CNDD 19,39,42 , and given the substantial evidence for strong CNDD acting on early life stages of trees in BCI 3,8,9 , we infer that the observed overdispersion is highly likely a result of CNDD acting on juveniles that propagates to the adult stage. A similar goal was attempted with a null model that considers the change in aggregation from the sapling to the adult life stages 12,24 .…”

“…More broadly, patterns of species distributions at local scales have been investigated and compared to CSR for over one hundred years 18 , often with the aim of detecting the signals of habitat specificity and biotic interactions 12,13,15,37,51 . However, theoretical models predict that dispersal limitation alone could override and obscure the signal of biotic interactions and dominate spatial patterns, creating ubiquitous aggregated distributions 39,42 . Indeed, spatial patterns have been found to be predominantly aggregated with respect to the CSR null 12,13,51 .…”