An early eastern ozone model overpredicted high-day fractions in selected western survey windows in 2005-2019

We tested whether a 30-day mean ozone value gave the same relation to high-ozone days across bird survey windows. An early New York and Pennsylvania model was tested in later years and in California and Oregon. In 363 later western windows, the observed fraction above 70 ppb was 2.23%, compared with 5.67% predicted. The error was -3.44 percentage points. Its corrected site interval excluded zero. A wider check that also resampled years included zero after correction for three tests. Early western errors were smaller. Threshold, distance, and observed-bin checks gave the same sign in later West. These selected records show a calibration limit for this fitted model, but do not establish a stable regional difference. Survey dates, monitor placement, device averaging, few years, and model form limit inference. We do not estimate bird harm or climate-change effects. The package includes daily evidence, provenance, code, all estimates, independent checks, and an offline reproduction recipe.

ozone · exposure measurement · bird surveys · high-ozone days · model transfer · observation bias · uncertainty