Using bayesian inference to validate plant community assemblages and determine indicator species

• Book Chapters
• Jan, 2017

Summary:

Recently, we described changes in plant community composition along gradients of anthropogenic disturbance, using a multinomial distribution in a Bayesian framework. Species were organized into categories (e.g. endemic, native, naturalized, invasive) and the proportions of each category in each community were represented by a multinomial vector. We now extend the use of the multinomial distribution to represent all the species in a community, individually, and use this approach to (i) validate plant community assemblages according to their specific composition, and (ii) determine indicator species for each community assemblage. Communities were assembled according to different models: null (all together); saturated (all separated); semi-saturated (only community replicates together); random (random assemblages); gradient (communities assembled in types along an ecological gradient). The models were calculated by using WinBugs and model fit was evaluated using Deviance Information Criterion (DIC). After the best community assemblage was found, we used Bayes rule to estimate the probability of a community, given the presence of a species, and compared the resulting indicator species with those determined by using conventional indicator values (IndVal). Both community assemblage and indicator species analysis gave good results when using two comprehensive plant community data sets for the Azores, i.e., a gradient of anthropogenic disturbance and an altitude gradient. Our method allows to (i) statistically validate plant community assemblages; and (ii) incorporate the prevalence of a plant community in the calculations pertaining to indicator species analysis.

---

https://link.springer.com/chapter/10.1007978-3-319-55236-1_21