Drivers of biodiversity are often studied with specific taxonomic or geographical focuses, and while many such studies have significantly advanced ecology and conservation, until now, limited data availability has hindered truly global studies.
In a massive study analysing distribution data from GBIF and other sources of more than 67,000 vertebrate and invertebrate species across terrestrial and marine domains, researchers used artificial neural networks to predict global species richness and how it relates to environmental features in each domain.
Quantifying contributions of environmental drivers, the produced models identified sunlight and temperature as the most important factors shaping biodiversity across domains. In the ocean, depth and oxygen were also important, while terrestrial biodiversity was more influenced by precipitation and primary production.
The study also confirmed known latitudinal gradients, but pointed out that latitude in itself is not a mechanism but a proxy for, e.g. sunlight and temperature. Attempts to forecast impacts of climate change should, therefore, model these mechanisms directly.
Not all patterns of biodiversity could be explained by environmental drivers. The models under-predicted richness in several coral reefs and montane forests while tending to over-predict richness in regions with steep biogeographic boundaries and on isolated islands.