Observer retention gave no clear extra temperature report persistence in four Canada and US bird survey samples in 1985-2024

We tested whether a five-unit temperature report pattern had greater annual persistence with the same observer. We used 70,789 adjacent-year pairs from accepted bird surveys in Canada and the US in 1985-2024. Four models controlled route, state-year, temperature-bin pairs and survey dates. Extra persistence estimates ranged from 0.019 to 0.145 score units per prior score unit. All four adjusted intervals included zero. Start-report and central-range checks also had intervals that included zero. Observer changes were scarce, especially in Canada. Shared years, a prior-outcome predictor, missing instruments and selected fair-weather visits limit inference. This result does not show that observer habits have no effect. The package gives code, exact source hashes, the selected records, model evidence and an independent numerical check.

observation bias · temperature precision · bird surveys · observer continuity · null result · weather · reporting persistence