Larger wind rises were linked to lower late-stop taxon counts in three of four US bird samples in 1997-2025

We tested weather-related observation bias in public bird survey records. We used 6,748 consecutive-year pairs with the same route and observer in California and Oregon, and New York and Pennsylvania. We split the tests into 1997-2010 and 2011-2025. The outcome was the change in mean recorded taxa per stop at the last ten stops relative to the first ten stops. All four slopes for end-minus-start wind were negative, from -0.154 to -0.114 taxa per stop per wind code. Three intervals were below zero after correction for four tests. The later eastern interval included zero. Controls covered mean wind, temperature, sky, survey timing, assistant status and state-year effects. Four sensitivity checks kept negative signs but changed some interval decisions. This extends a prior block-count study with stop-level means and matched annual changes. Data overlap with the prior work. The result is a weather association, not a measure of detection probability or climate damage. Endpoint weather, sample selection, habitat change and shared survey methods limit inference. The package gives code, source hashes, derived runs, paired records, model designs, checks and a figure.

biodiversity · observation bias · wind · Breeding Bird Survey · stop-level counts · annual differences · regional replication