A Bayesian GLMM for presence-only data on species richness of Ross Sea Mollusca

What is it about?

In many research fields great efforts have been made to integrate data from different sources to create databases that are as complete as possible. Unlocking the information contained in presence‐only data, i.e. data collected by non standardized means, with no sampling design and no standardized protocols, presents a challenge for statistical modeling because of insidious observational errors due to the opportunistic nature of the data‐gathering process. For the first time we adopt a Bayesian semiparametric generalized linear mixed model (GLMM) with Dirichlet process random effects and apply it to historical and newly collected presence‐only data on species richness of the Ross Sea mollusca.

Featured Image

Why is it important?

The so-called primary biodiversity data (presence-only data) are continuously increasing and currently constitute the main component of the biodiversity information stored in large-scale aggregators like the GBIF or the BioCASE. Statistical methods that try to deal with observational errors occurring with presence-only data usually rely on parametric assumptions and auxiliary information. To the best of our knowdledge this is the first time that, without using auxiliary information, the potential of the Dirichlet process is exploited in modeling sampling bias, detection errors, and, more generally, the very complicated errors resulting from opportunistic data collation.

Perspectives