WK 27

Demography in a Continuous World: New Advances in Integral Projection Models (IPMs)

Structured population models have become a central tool to explore evolutionary and ecological processes. With recent advances in theoretical ecology and computing power, these models have further evolved. Over the last decade, Integral Projection Models (IPMs) have gained popularity because of their simplicity, robustness and flexibility (e.g., no arbitrary discretization of state variables, few parameters, high resolution, incorporate demographic and environmental variability).

 Although the same set of biologically meaningful outcomes can be obtained from IPMs as those obtained from matrix models, IPMs come with a different set of quantitative and computational challenges: model selection and formal tests of which approaches are best for different questions and organisms/growth forms.

 Participants will learn to (i) organize and analyze their own data based on examples provided by the organizers, (ii) construct basic and advanced IPMs (e.g. using Bayesian approaches to incorporate uncertainty), and (iii) perform basic (population growth rate, stable stage distribution, lifespan, passage time…) and advanced model construction, demographic projections and perturbations (complex state models including seedbank, dormancy, multiple state classes, stochastic models, latent variables). Modeling will be performed in R using the library IPMpack. Following a general introduction, participants will break into groups where, assisted by the organizers, they will learn about various aspects of IPMs according to expertise and interests.

Additionally, we will also pay attention to ways to speed-up computation in R. By improving computational efficiency, participants can feasibly solve more complex tasks, ask more ambitious questions, and include more sophisticated analyses in their research. We will review various solutions to common computational efficiency problems in ecological and evolutionary research. This part of the workshop is aimed at those who have at least an introductory knowledge of R, but aren't familiar with code profiling, parallel computing or refactoring in a lower-level languages.