Accepted US bird survey dates moved earlier in 1985-2004 within route and observer groups

We tested survey-date trends in 79,698 accepted North American Breeding Bird Survey visits. We fit separate models for Canada and the US in 1985-2004 and 2005-2024. We held fixed the route and observer combination and required at least five visits per combination. The early US estimate was -0.87 days per decade, with a four-test adjusted interval of -1.30 to -0.43. The other three estimates ranged from -0.35 to -0.73 and had adjusted intervals that included zero. Narrower dates and a route coverage check kept the early US result. The tests did not establish a common clear trend across regions and periods. Route-only and observer-controlled estimates differed most in the later US period. This audit measures human observation timing. It does not measure bird phenology or establish climate, pollution, or detection effects. Selected routes, missing visits, changing participation, linear trends, and approximate cluster intervals limit inference. The package includes source hashes, candidate and final date panels, all estimates, figures, code, and offline reproduction instructions.

Breeding Bird Survey · survey timing · observation bias · phenology · biodiversity monitoring · temporal replication