Bayesian Integrated Species Distribution Models for Hierarchical Resource Selection by a Soaring Bird

Authors

Ryo Ogawa, Mississippi State University
Guiming Wang, Mississippi State UniversityFollow
L. Wes Burger, Mississippi State University
Bronson Strickland, Mississippi State UniversityFollow
J. Brian Davis, Mississippi State University
Fred L. Cunningham, United States Department of Agriculture, Wildlife Services, National Wildlife Research Center, Mississippi Field StationFollow

Document Type

Article

Date of this Version

2024

Citation

Ecological Informatics (2024) 82: 102787

doi: 10.1016/j.ecoinf.2024.102787

Comments

United States government work

Abstract

Migratory birds exhibit seasonal geographic range (hereafter, range) dynamics during the annual cycle. Few studies have examined how migratory birds select their habitats for range occupancy at the species level and space use at the individual level simultaneously. We hypothesized that environmental variables directly related to fitness components would affect the range occupancy probabilities of migrants, whereas environment variables related to movements and flights would affect the space use intensities of migrants. We built Bayesian integrated species distribution models (ISDMs) to evaluate the effects of climate conditions, wind conditions, and landcover compositions on the seasonal range dynamics of American white pelicans Pelecanus erythrorhynchos (hereafter, pelican) during summer and winter. The ISDMs estimated the summer range occupancy probabilities of pelicans with Breeding Bird Survey data, winter range occupancy probabilities with Christmas Bird Count data, and summer and winter space-use intensity rates with eBird data jointly. We evaluated the predictive performance of ISDMs using independent datasets of pelican GPS locations. Integrated species distribution models outperformed the occupancy-only models in the predictive performance of occupancy probabilities. Climate conditions had opposite effects on the range occupancy probabilities between the breeding and nonbreeding grounds, whereas landcovers had relatively consistent effects on range occupancy probabilities between the seasons.