End wind had no clear link to earlier bird counts in four Canada and US samples, 1985-2024

We tested whether survey-end wind predicts bird counts from the first ten stops after control for start wind. We reused 64,681 accepted Breeding Bird Survey runs in Canada and the US. We fit separate models for 1985-2004 and 2005-2024. The models control for route and observer, state and year, survey timing, temperature, sky and assistant status. All four first-block intervals include zero after correction for four tests. The first-block slopes range from -0.091 to -0.005 recorded taxa per end-wind code. Later-block slopes are negative in all four samples. This supports temporal specificity under the model. It does not prove a detection mechanism. End weather can proxy earlier unmeasured weather. Complete records, roadside sampling, changing observer skill, coarse weather codes and shared routes limit inference. We did not estimate pollution or climate-change effects. The package includes the derived panel, source lineage, code, all fits and checks.

bird counts · observation bias · wind · temporal specificity · negative control · Breeding Bird Survey