Prior and later ozone slopes were not clearly different in four-state bird counts in 2005-2019

We tested whether bird-count links to prior and later ozone differed in California, Oregon, New York, and Pennsylvania in 2005-2019. We used 1,175 consecutive-year survey pairs with the same route, observer, and ozone site. We fit joint models for mean daily maximum 8-hour ozone in the 30 days before and after each survey. We controlled for survey weather, timing, and state-year changes. We tested the prior slope minus the later slope in two regions and two periods. Estimates were 0.789, -0.764, -5.727, and -1.142 recorded taxa per 10 ppb. All four 95 percent and multiple-test intervals included zero. Planned distance, date, and coverage checks did not resolve a common timing contrast. One strict coverage check was not identified. These results do not establish equal slopes, absence of harm, or complete bias control. Later ozone can proxy earlier exposure, and coverage selects the sample. Seasonal weather, exposure error, imperfect detection, and small samples limit inference. The package supplies a temporal diagnostic, paired-window data, figures, provenance, code, and offline reproduction checks. A full raw ozone join matched the retained daily data, and independent numerical fits agreed.

ozone · biodiversity · bird counts · negative exposure control · temporal specificity · observation bias · Breeding Bird Survey · survey weather · null result