Identifying the processes that maintain biodiversity is crucial for predicting how ecological communities will respond to environmental change. Unfortunately, mechanistic dynamical models are ill-suited to this task, requiring overly simplistic assumptions that fail in real-world systems. Conversely, correlative approaches struggle to separate cause and effect, leading to erroneous predictions in the face of novel conditions. My work focuses on the development of tools to bridge this divide between mechanistic and statistical approaches, borrowing ideas from computer science, statistics, and graph theory to improve out-of-fit predictions in complex ecosystems. In this talk, I will discuss some of the statistical and computational tools we've developed along with their application to various organisms and ecosystems, illustrating how a focus on emergent properties can allow for more robust predictions. Lastly, I will discuss how this macro-statistical view of mechanism is critical for addressing ecological challenges over short time scales and broad spatial scales.
Daniel Maynard (UCL)
