Environment

Bayesian inference on beta diversity via feature allocation models with imperfect detection

Federica Stolf, Tommaso Rigon, David B. Dunson

Featured August 25, 2026

AI-generated analysis — This is SciGrove's AI interpretation of the paper, not peer-reviewed content. Always refer to the original paper.

Simply

A new statistical model helps ecologists better understand how species vary across different places by separating true presence from observation errors, even predicting unseen species and accounting for site differences.

In depth

The paper introduces a novel Bayesian framework, MOSAIC, that simultaneously addresses imperfect species detection and systematic variation in species composition across ecological communities. It achieves this by separating the true occupancy probability of a species at a site from its detection probability, and by using partially exchangeable feature allocation models, enabling robust inference on beta diversity and prediction of unseen species.

Key Takeaways

1
The MOSAIC model provides a principled Bayesian framework for quantifying beta diversity by explicitly accounting for imperfect species detection and partial exchangeability across sites.
2
The framework offers analytical tractability for posterior and predictive distributions of species sharing and diversity, facilitating scalable inference in high-dimensional biodiversity studies.
3
A new model-based index for beta diversity is proposed, which is a cosine dissimilarity between latent occupancy profiles, providing robust and interpretable measures of compositional heterogeneity.

Conceptual Flow

HIGH LEVEL

Methodology: Modeling True Presence and Detection

The model figures out if a species is truly living somewhere and if we actually saw it, treating these as two separate steps.

Observed Species Data

Separate True Presence from Detection

Real Species Patterns

Detection Likelihood

Results: Accurate Diversity and Unseen Species

By doing this, the model gives a much clearer picture of how different places share species and how many species are still hidden.

Biased Diversity Estimates

Correct for Hidden Species and Errors

Accurate Beta Diversity

Predicted Unseen Species

This breakdown was generated by SciGrove. Get the same analysis — intuition, storyboard, peer review, a runnable prototype and a glossary — on any paper you upload or paste a DOI for.