Cory Merow
University of Connecticut
Department of Ecology and Evolutionary Biology
I am an ecologist interested in modeling strategies for community and population level processes. I have a background in physics and mathematics, but switched to ecology because the problems are so much more interesting and I’m surrounded by people who can teach me amazing things about the natural world. I’m really interested in how we obtain the conclusions we extract from data, and how confident we are in those conclusions. I try to balance mathematical or statistically rigorous analyses with biological and ecological sensibilities to produce models that accurately reflect our assumptions and data, while highlighting which parts of our inference derive from each. Thus far, I’ve focused on maximum entropy methods, Bayesian methods, and simulations with cellular automata. I conduct my field research in the fynbos vegetation of South Africa, where I focus on explaining abundance patterns along ecological gradients based on traits. I am fortunate to work in the lab of John Silander, where we focus on quantitative methods for understanding spatial patterns in ecology.
Current Projects
Suggestions for modeling species distributions with Maxent
with John Silander
I’m developing a suite of methods to better explore distribution models built with Maxent. In my opinion, many models built with default settings in Maxent are overly complex and therefore compromise their transferability and interpretation. I’m outlining a modeling strategy to produce simpler models based on a set of new techniques for significance testing of predictors and model comparison diagnostics. These methods will help users explore the consequences of Maxent’s assumptions and better understand their models to produce more robust predictions. Methods will be implemented in R, to use with any data set.
Bayesian integral projection models for testing demographic hypotheses
with John Silander and Mark Brand
Integral projection models represent the next generation of demographic models. They offer all of the advantages of matrix models in addition to many new features. I’m outlining a Bayesian framework to test demographic hypotheses, given incomplete vital rate data, that rigorously accounts for uncertainty in model parameters. The outcome (hopefully) will be a set of R tools to build IPMs for an arbitrary data set to rigorously capture the demographic information (e.g. population growth rates, stable stage distributions, etc.) contained in incomplete data while properly accounting for uncertainty in the estimates. As a case study, I’m applying this protocol to evaluate the invasive threat of 17 cultivars of Japanese barberry (Berberis thunbergii) in the northeastern United States.
Community abundance patterns in South African fynbos
with John Silander, Adam Wilson, Jasper Slingsby, Andrew Latimer, Kent Holsinger, Cynthia Jones and Carl Schlichting
This work is part of a larger collaborative effort focused on understanding the different aspects of biodiversity in plant communities in the Greater Cape Floristic Region (GCFR) of South Africa. The larger project seeks to understand the relationship between functional traits and patterns of genetic, functional and taxonomic diversity. Our goal is to integrate knowledge collected at these different scales to predict species and community responses, enhanced by understanding their evolutionary past. I’m focusing on determining how community-level distributions of functional traits vary along ecological gradients and how this variation can predict species’ abundance patterns. Here’s a brief article about the project: link.
Novel features for entropy maximization models to predict community abundance patterns from traits
with John Silander
Entropy maximization has recently been applied to the problem of predicting community abundance patterns based on species’ traits along ecological gradients (see work by Bill Shipley and my paper). These methods are only recently developed for ecology and have proven useful in predicting abundance patterns in a handful of systems. I am developing new extensions for these methods that better reflect ecological processes and generalize the method to more complex problems.
Protea demography using Bayesian integral projection models
with John Silander, Adam Wilson and Andrew Latimer
To understand a species’ response to a changing environment, it is critical to relate demographic heterogeneity back to its environmental drivers. Typically climate variables that represent average weather over a 30 or 50-year interval are used to characterize the environment, however these are only proxies for more physiologically relevant predictors (e.g. maximum drought length). We’re developing a daily weather data set from which to construct these predictors that includes uncertainty. I’m developing demographic models that relate vital rates back to these environmental drivers. This will allow us to map predicted population viability, including uncertainty, based on environmental conditions across the species’ range.
