Determining where species are likely to occur and understanding the mechanisms leading to patterns of species occurrence form a fundamental problem in ecology. Species occurrence is often treated as primarily determined by environment (the fundamental niche) and secondarily limited by the presence of other species (the realized niche). It has been proposed that the relative importance of these factors depends on spatial scale, with environmental filtering more important at larger spatial scales and biotic interaction increasingly important at local scales. The development of joint species distribution models provides a potential way to quantify the relative influence of environmental factors, species traits, and species co-occurrence on species presence and abundance. Here we used the Hierarchical Modeling of Species Communities approach (developed by Ovaskainen et al. 2017) to study the structure of the diverse crustacean zooplankton community of Carolina Bays, intermittent freshwater wetlands in the south-east U.S. We used 2 years of bi-weekly samples from 14 wetlands on the Savannah River Site, with a total community richness of 81 species. We collected environmental covariates from each wetland and compiled a trait database (size, feeding guild, reproductive mode, etc.) from published literature. We then used MCMC as implemented in the HMSC package in R to fit models to the presence/absence and abundance data of these species.
Results/Conclusions
Explanatory power of the models was relatively high, with an average R-squared of 0.73 across the 81 species. We found that species presence was best predicted by environmental variables and community structure on the previous sampling date, while the spatial random effects that include the influence of species co-occurrence accounted for less than 10% of the variance explained. When we repeated the analysis on species abundance the impact of random effects doubled, suggesting differential effects of environment and species co-occurrence on species presence vs. abundance. In addition, we found a large variation in how sensitive species were to spatial effects, with a majority of species responding almost only to the environment, while others had a 50% dependence on the spatial random effects. We then looked at individual species relationships to see if species presence and abundance respond similarly to the environmental factors. Applying these new computational tools to a set of highly variable, discrete habitats can advance our understanding of community structure and the relative importance of the factors creating that structure.