Detecting differences in species abundance patterns in niche and neutral communities

Background/Question/Methods

Species abundance distributions may reflect the dynamic processes that influence population growth rates, but how they do so is not yet well understood. Recent studies focus on the relative importance of two types of mechanisms: niche and neutral dynamics. Neutral dynamics are based solely on demographic stochasticity and immigration and niche dynamics are generated by trait differences that affect the fitnesses of competing species and can enhance coexistence. One recent study showed that abundance patterns produced under niche and neutral dynamics are very similar, especially when diversity is high relative to the number of niches. However, for simplicity, that study assumed species in separate niches are essentially non-interacting. Here we consider the distributions arising from a stochastic Lotka-Volterra competition model in which the strength of competition depends on differences in a trait. This model produces a more complex niche structure in which species in separate niches interact with one another. We reported at a recent meeting that with this model we find more substantial differences between niche and neutral species abundance distributions than was found in the prior study, even when diversity is high. Here we focus on how detectable the differences from neutral predictions would be in data.

Results/Conclusions

We show how the detectability depends on whether knowledge of the neutral parameters is established a priori from ecosystem characteristics, or whether the parameters are fit to species abundance data using a likelihood sampling approach. We also show how the detectability is altered by the number of niches and by increasing the number of trait dimensions from one to two. With a priori parameters, moderate to reasonable detection levels can be achieved even with a modest number of niches in a data set comparable to the 50 hectare Barro Colorado Island Forest plot. Detection is much more challenging under parameter fitting. The qualitative nature of the departure from the neutral case, as characterized by evenness, varies as the number of niches increases. This may be problematic for ascribing detected differences to niche differentiation specifically, as opposed to other potential influences such as habitat filtering.