Back

A JSDM with zero-inflation to increase the ecological relevance of analyses of species distribution data.

Tous, J.; Chiquet, J.; Deacon, A. E.; Fontrodona-Eslava, A.; Fraser, D. F.; Magurran, A. E.

2025-07-31 ecology
10.1101/2025.07.24.666553 bioRxiv
Show abstract

O_LIDeciphering the mechanisms that explain species distributions has been a long-term goal in ecology and is critical to predicting the responses of ecological communities to environmental change. To this end, Joint Species Distribution Models (JSDMs) assess the effects of abiotic factors on these communities and produce species association networks that describe statistical dependencies between species abundances. Species distribution data often contain numerous zeros. Not accounting for structural zeros (as opposed to zeros owing to species rarity or abiotic effects, modelled by JSDMs) can impair inferences of species association networks. C_LIO_LIWe propose a novel JSDM, the ZIPLN-network model, which directly models count data while including zero-inflation, thereby allowing it to account for structural zeros. Using simulated data, we compare the results obtained by this model to existing JSDMs in terms of association network inference from abundance data containing structural zeros. We illustrate the ZIPLN-network approach using data from tropical freshwater fish communities. C_LIO_LIBy integrating zero-inflation, our approach overcomes a previous shortcoming in analyses of species distribution data, and makes it possible to extract more of the ecological signal. A further advantage of this method is that it increases our understanding of ecological mechanisms that structure species distributions at the landscape scale; this helps generate testable hypotheses. When the ZIPLN-network results are compared with those of its non-zero-inflated counterpart, a finer analysis of a species distribution dataset results. C_LIO_LIThe ZIPLN-network approach can be widely applied to species distribution datasets to produce more informed analyses and help distinguish different mechanisms so as to better understand community assembly rules. We provide guidance for getting started with our approach. C_LI The data and code used for this paper can be found in the public github repository: jeannetous/JSDM_with_zero_inflation_for_increased_ecological_relevance_in_species_distri bution_analysis

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.