Protecting biodiversity requires understanding what forces—natural and anthropogenic—shape it. We use a variety of methods to identify and predict how these forces have, are, and will affect species and their communities. Much of this work focuses on the effects of climate change, but even if climate were not changing, we would still be living in a biodiversity crisis.

Most species of conservation concern are “unmodelable” using traditional means: sample sizes are too small, samples are too inaccurate or old, taxonomic uncertainty complicates specificity, and there are extreme sampling biases. As a result, extremely rare species are often excluded from conservation assessments. We develop modeling methods and other techniques to make “unmodelable” species modellable. As an example, working with Alex Linan with funding from the National Science Foundation, we are conducting a taxon-wide assessment of Diospyros in Madagascar using phylogenetically-informed multi-species distribution models intended to locate aeras where discovery of new specimens of microendemic species would be most valuable for assessing conservation status.

Science often approaches species in a reductionist manner: we “segment” species into facets like occurrence, abundance, traits, genetics, interactions, evolutionary relationships, etc., then examine each as if the others did not matter. But real organisms and species are integrators of multiple facets, which challenges the reductionist approach to understanding how species interact with their environments. We used advanced statistical and mathematic modeling to combine different facets into coherent “gestalt” models to better understand how whole organisms and species react to their environments.

The first generation of biodiversity data mobilization occurred when major natural history museums and herbaria constructed and posted online databases of their holdings. The second—still ongoing—involves further mobilization, development of the “extended specimen” concept, and construction of data aggregators like GBIF for occurrences, TRY for plant traits, or GLoBI for biotic interactions. The result has ushered a renaissance in biodiversity data analytics, but also an extremely confusing array of thousands of databases and data sets that are difficult to find, challenging to compare, and time-consuming to integrate. We are developing the next generation of biodiversity infrastructure through Biodiversity on Demand (BOND), an AI-driver platform funded by the US National Science Foundation that will enable biodiversity data discovery, comparison, and integration.