PS 23-41 - Evaluating support for alternative species distribution models with population genomic data and ABC

Tuesday, August 16, 2022

5:00 PM – 6:30 PM EDT

Located in: ESA Exhibit Hall

Presenting Author(s)

SN

Sarah R. Naughtin

Undergraduate Research Assistant
Michigan State University
Northville, Michigan, United States

Coauthor(s)

AC

Antonio R. Castilla

Michigan State University
East Lansing, MI, United States
AD

Andria Dawson

Mount Royal University, Canada
SH

Sean Hoban

Morton Arboretum, United States
AS

Allan E. Strand

College of Charleston, United States
AS

Adam B. Smith

Associate Scientist
Missouri Botanical Garden
St. Louis, MO, United States
JR

John D. Robinson

Michigan State University, Michigan, United States

Background/Question/Methods

Climate change poses a threat to biodiversity, and it is unclear whether species can adapt to changing conditions or shift their ranges to track suitable habitats. Reconstructions of range shifts over the past 20 Ka, including the use of hindcast species distribution models (SDMs), in response to environmental changes since the Last Glacial Maximum (LGM), can thus help to inform current conservation efforts. However, different SDM algorithms and climate reconstructions may produce contrasting patterns, highlighting uncertainty associated with this approach. Some recent studies have integrated SDMs and genetic data to model range shifts, but these studies often rely on a single SDM and rarely incorporate uncertainty. We modeled historically suitable habitat for Fraxinus pennsylvanica (Green Ash) using 24 SDMs built using different climate models, calibration regions (based on three different buffer extents), and modeling algorithms. We simulated genomic datasets using each of the 24 SDMs as habitat suitability in a spatially explicit model. Approximate Bayesian Computation (ABC) was then used to evaluate the support for alternative SDMs. Next, we estimated the speed of post-glacial range shifts ("biotic velocity") under the best supported models and compared inferences among alternative SDMs. Finally, the best-supported models were projected to future climate scenarios.

Results/Conclusions

Our 24 SDMs revealed substantial variation in the location and size of glacial refugia for F. pennsylvanica, but with general support for high suitability at the LGM across the southeastern United States. Clustering analysis showed that SDMs with similar climate models grouped at the LGM, whereas the buffer used to define the calibration region was the main driver of differences between contemporary suitability surfaces. Estimates of biotic velocity, when calculated directly from the 24 SDMs, highlight increased uncertainty during the Bølling-Allerød interstadial and Younger Dryas, ranging from less than 100 m/yr to more than 300 m/yr. Following integration of genomic data, support for the alternative SDMs varied and uncertainty around estimates of biotic velocity was reduced. Both climate model and SDM algorithm appeared to influence representation of the alternative models in simulations accepted during the ABC analysis. Our results highlight the different inferences that may result from application of alternative distribution modeling approaches and underscore the importance of propagating uncertainty between data types in integrative models of post-glacial movement.