Earlier ozone offsets gave opposite changes in high-day disagreement for two transfer directions near US bird surveys in 2005-2019

We tested whether earlier mean differences between nearby ozone sites improved later labels for days above 70 ppb. We used published daily evidence near bird survey routes in California, Oregon, New York and Pennsylvania in 2005-2019. Four region and period samples had 12,860 test pair-days. In every sample, the change in disagreement had opposite signs for the two transfer directions. Changes ranged from -1.289 to +0.922 percentage points. Earlier work found lower concentration error in three samples, but each had one direction with worse label agreement. Neither site is a true exposure reference. The signs changed with cutoff, weights and data support. These are finite-sample method results, not population or causal effects. Repeated sites, rare high days, missing data and selected survey windows limit inference. We do not test climate or bird responses. The package gives code, daily evidence, exact provenance, all scores, a figure and independent checks.

ozone · exposure measurement · threshold agreement · bird surveys · temporal validation · observation bias