End wind code 2 versus 0 was linked to lower bird taxon overlap in four models of US surveys from 1985 to 2004

We tested whether end wind code 2 versus 0 was linked to lower overlap of bird taxa between the first and last ten stops of a survey. We used 64,681 accepted Canada and US surveys in four country-period samples from 1985 to 2024. Models controlled observer-route and state-year effects, start wind, temperature, sky and survey timing. The four overlap slopes were -0.394, -0.052, -0.704, -0.504 percentage points. Both US four-test intervals were below zero. Both Canada intervals included zero. The early US link remained negative with intervals below zero in all three planned checks. All late US check intervals included zero. This does not establish a country difference. Fixed block locations, missing stop-level weather, selected mild-weather runs and unmeasured bird activity limit inference. This is an audit of recorded taxon overlap. It does not identify detection probability, population loss, or a climate or pollution effect. The package gives the raw-data builder, source hashes, derived evidence, all fits and numerical checks.

biodiversity · observation bias · bird surveys · wind · taxon overlap · regional replication