Species face uncertain futures under global climate change. Whereas some might adapt to changing climate with range shifts and local adaptation, others will experience range contractions, population declines, and even extinction. Although species will respond differently to climate change, they will not respond separately. Interactions among species (e.g., predator-prey, interspecific competition) strongly influence how species respond to their environment in space and time. The collective responses of species to climate change will result in species reshuffling into new communities for which there is no modern analog. Current models predicting the effects of climate change on species take a single-species approach that ignores species interactions.
Though species interactions largely occur at local scales, they may scale up to influence broader changes in species distributions and abundances as well as shifts in community composition. The major goal of this project is to determine if we can detect the influence of species interactions on changes in species distributions and abundances at different scales. Specifically, we will compare how known interacting species vs. co-occurring (non-interacting) species react as climate changes. Co-occurring species may respond similarly to changes in climate because they share the same environmental requirements. We hope to detect if and when species interactions are actually influencing species responses to changes in climate or whether the species responses are just a function of shared environmental requirements.
We will use the US-wide Breeding Bird Survey (BBS) data in the lower 48 to assess whether species interactions map to species distributions under recent changes in climate. From this knowledge, we hope to be able to improve predictions under climate change, and may be able to make some predictions using climate change scenarios.
This project will rely heavily on various computer programs including R and GIS, and potentially, Matlab. GIS/Remote Sensing tools will be used to stratify and extract the BBS data by habitat type (e.g., forest type from landcover datasets like Landfire, or FIA), and to extract additional environmental variables to each BBS point (e.g., 30-m DEM, PRISM temperature/precipitation rasters). R will be used to set up and run the statistical models. The modeling frameworks will include multivariate Bayesian spatial effects models, and multivariate autoregressive models, and will be implemented in R.