Comparing multiscale, presence-only habitat suitability models created with structured survey data and community science data for a rare warbler species at the southern range margin

Golden-winged Warblers (Vermivora chrysoptera, Parulidae) are declining migrant songbirds that breed in the Great Lakes and Appalachian regions of North America. Within their breeding range, Golden-winged Warblers are found in early successional habitats adjacent to mature hardwood forest, and previous work has found that Golden-winged Warbler habitat preferences are scale-dependent. Golden-winged Warbler Working Group management recommendations were written to apply to large regions of the breeding range, but there may be localized differences in both habitat availability and preferences. Rapid declines at the southernmost extent of their breeding range in Western North Carolina and Georgia necessitate investigation into landscape characteristics governing distribution in this subregion. We describe patterns of Golden-winged Warbler presence in Western North Carolina by examining habitat variables at multiple spatial scales using data from standardized Audubon North Carolina (NC) surveys and unstructured community scientist checklists submitted to eBird. We compared model performance and predictions between Audubon NC and eBird models and found that Golden-winged Warbler presence is associated with sites which, at a local scale (150m), have more young forest, less mature forest, and more road cover, and at a landscape scale (2.5km), have less road cover. eBird and Audubon models had similar parameter estimates for all of the land cover variables and similar overall performance. Based on parameter estimates, road density at the landscape scale (2.5km) is the primary variable driving the difference between Golden-winged Warbler breeding sites and random background sites in Western North Carolina. Our results show that eBird data can produce species distribution modeling results that are similar to results obtained from more standardized survey methods. ### Competing Interest Statement The authors have declared no competing interest.