Model-based inference of environmental niches and niche evolution: fundamental and practical issues for discussion

Species Distribution Models (SDMs) typically rely on machine-learning approaches to estimate the niche of a species, where the "niche model" is effectively a correlational model built on physical environmental variables such as temperature and precipitation.

This approach is widely used, but has been criticised on a number of grounds, including (1) the problematic assumption that physical environmental variables are the most relevant ones to build models on; (2) the complex models that are fit by sophisticated algorithms, with the potential for overfitting and poor extrapolation ability; (3) the problem of confounding variables (dispersal limitation, evolutionary history, competition, etc.). 

Some of these issues (not all!) may be ameliorated by including evolutionary information in the inference, and jointly modeling the niches of living species, and an evolutionary model describing how the niche models evolve on the evolutionary tree. This idea raises a host of theoretical issues (What is evolving? How do we model it?) and practical issues (The number of parameters rapidly increases with more species; Models that are evolutionarily conservative may have better extrapolation ability, but poorer fit to training data than machine-learning approaches.).

The topic has interesting connections to areas such as morphometrics in phylogenetics (see previous TEA talk by Sherratt and Brennan) and model-based inference in both ecology and evolution. I will present my current thinking and progress, but I am most interested in a free-flowing discussion of all aspects of niche evolution, both from a theoretical perspective and practical perspective (i.e., what do biologists need that they don't currently get from standard SDM approaches).

People from all academic backgrounds and levels are welcome.

We appreciate you taking the time to register (it's free) - it gives us an idea of who’s attending our TEA Talks, and the presenters some background on their audience.

Hosted by the Centre for Biodiversity Analysis, TEA Talks (Techniques in Evolutionary Analysis) are a monthly series of short workshops that introduce a range of current methods and analytical approaches in phylogenetics, bioinformatics and macroevolution.