Hierarchical models for sharing information between species when estimating nonlinear trends in community data

Population Ecology
14:55, Friday 20 Dec 2019 (15 minutes)
Room AB

As anthropogenic environmental impacts increase, we expect to see more cases of rapid change in the abundance and distribution of species. Effective ecological monitoring now requires characterizing how quickly whole communities are changing, rather than just the most common taxa. However, few monitoring programs or studies have the resources to estimate abundances with high precision for all species in a given study area; as such, detecting directional trends in species abundance can be difficult, especially for rare species which may be particularly vulnerable to environmental change. Hierarchical regression models, where parameter estimates for different groups are pooled toward one another, have become common in ecology, but have not seen widespread use for estimating nonlinear trends. In this talk, I show how one of the most popular approaches for fitting nonlinear trends, the Generalized Additive Model, can be adapted to develop hierarchical multi-species trend models which can share information between multiple rare species to detect common signals of environmental change. I illustrate the benefits of this approach by showing how this method is able to detect the start of two periods of rapid community change in the Newfoundland and Labrador Shelf benthic ecosystem, several years before the most abundant species themselves changed. I will also talk about new directions to extend this method to include functional and phylogenetic similarity between species when estimating trends.


My Schedule

Add to My Schedule as Favorite