LAUSR

Summary Atlas data provide biodiversity information at a relatively fine spatial grain over a broad spatial extent and, increasingly, at multiple points in time, which make them invaluable for understanding processes that affect species distributions over time. The effect of survey effort on species detection has long been appreciated and Atlases typically include survey standards and records of effort, but challenges remain in analysing Atlas data that have not been collected using a repeated sampling protocol designed to correct for imperfect detection. We developed a single-visit dynamic occupancy model to quantify the effects of climatic and land-use drivers on local species extinction and colonization while accounting for imperfect detection using repeat Atlas data. We evaluated model stability using data simulated under alternative scenarios and, ultimately, applied the model to empirical data for Canada warbler Cardellina canadensis, a wide-spread species exhibiting a long-term population decline. At sample sizes that are realistic for many Atlases (n = 1000–10 000 independent survey blocks), our models produced unbiased estimates of detection, occupancy, colonization and extinction parameters. Slope estimates for explanatory covariates were somewhat less stable than overall occupancy, colonization and extinction rates, with covariate effects being sensitive to the total number of, and relationships among, explanatory variables. In comparison to other analyses of Canada warbler distributions that indicated minor changes over time, our approach identified a widespread decline in occupancy probability across New York, consistent with the broader population trend, particularly in the areas where it was initially more likely to occur. Synthesis and applications. A single-visit dynamic occupancy model is a novel method for analysing common, ecologically valuable datasets, such as Atlases, that lack repeated sampling necessary to correct for imperfect detection using alternative multi-season occupancy modelling approaches. As a result, using this method can improve understanding of species distributions and factors that shape them over time, thereby providing more accurate information to guide conservation and management.