Supporting Information accompanying M. R. Moura & W. Jetz: “Shortfalls and opportunities in terrestrial vertebrate species discovery”. In press. Nature Ecology and Evolution


View the new interactive map!


Downloads

Mapping opportunities for species discovery

Our knowledge on Earth’s diversity is not static and since the beginning of modern taxonomy in 1758, more than 1,8 million of species have been described. But our planet likely has more than 10 million of species. For centuries explorers and taxonomists have worked hard to discover and describe species of terrestrial vertebrates:


Animations of the change in species richness change over time for Amphibia as more species were discovered and described. See background for other taxa

But many still remain undiscovered.

Maps of discovery potential

In the research represented on these pages we extrapolated the signal of past patterns of discovery into the future and developed a map of likely future discovery of new species! Check them out here: mol.org/patterns/discovery. The maps show the portion of total yet to be discovered species of a particular group of vertebrates that our models predict to be found in a particular region.

Our approach

In the paper published in Nature Ecology and Evolution (here) we describe the approach. Our research took advantage of an unprecedent dataset including 11 biological, geographical, and sociological attributes computed for 32,172 species of amphibians, reptiles, mammals, and birds. Since the chances of being discovered and described early are not equal among species, we were able to use these species-level attributes to model the discovery probability of all known terrestrial vertebrate species and use those probabilities to construct metrics of discovery potential across different grid cells, taxa, countries, and biomes and realms.

Limits to interpretation

Please note that we do not expect our geographic discovery projections to hold up in exact form. They are estimates that are a direct reflection of past description processes and their correlates, and any forward interpretation therefore needs to recognize intrinsic limitations. Notably, species represent scientific hypotheses that are sometimes revisited, refuted or revalidated. Our models therefore are not able to distinguish operational definitions of valid species and the potential heterogeneous associations arising from variable practices around, for example, recognizing cryptic species or splits. There may also be parts of the multivariate predictor space that lack data to inform the model and thus miss actual discovery opportunities.

Acknowledgments

We gratefully acknowledge support from the National Geographic Society and E.O. Wilson Biodiversity Foundation for this work. The research is facilitated through the National Science Foundation VertLife project (http://vertlife.org).