Spatial covariance tracks grassland bird community occupancy in fragmented grasslands

Grasslands are an imperiled ecosystem, and grassland bird abundance is declining across North America. One of the strongest drivers for these declines is woody plant encroachment of grasslands. In the Great Plains and Sagebrush biomes of North America, spatial covariance—a remote sensing metric for tracking boundaries between vegetation types—is emerging as a new method to identify and strategize conservation of grassland cores in the face of woody plant encroachment. However, the relationship between spatial covariance and grassland bird community occupancy is unknown. Here, we used Bayesian multi-species occupancy models to understand how occupancy probability of six declining grassland species responded to spatial covariance at three scales (0.81ha, 7.29ha, and 65.61ha) in fragmented grasslands of Arkansas, USA. We found that the grassland bird community avoided tree-grass boundaries (negative spatial covariance) and was positively associated with more intact grassland cores (near-zero or slightly positive spatial covariance) at all scales, but model selection revealed smallest spatial scale (0.81ha) best explained grassland bird occupancy. The strength of tree-grass boundary avoidance was idiosyncratic across species. Our results indicate that spatial covariance can rapidly highlight grassland cores and serve as a proxy for grassland bird community occupancy, even in highly fragmented grasslands.