Nearby sites gave different three-day ozone labels in at least 88 of 1,599 US bird survey windows in 2005-2019

We tested whether nearby sites support the same three-day ozone episode label before a bird survey. We used 1,599 selected windows in California, Oregon, New York, and Pennsylvania in 2005-2019. Both sites were within 50 km of each route start. A high day was above 70 ppb. We gave exact per-site event bounds by setting missing daily readings low or high. At least 88 windows had different labels for every missing-data assignment. A further 132 had an unresolved label. Complete records retained 797 windows and 43 differences. A 25-km limit and equal site-pair or route weights retained differences in all four region-period groups. Positive event agreement had conservative outer bounds of 38.0-60.0%, 3.0-50.0%, 41.4-54.2%, and 57.3-73.5%. These are missing-data bounds, not confidence intervals. Selected routes, shared site-days, few positive events, and unknown true exposure limit inference. We do not estimate bird damage or a climate-pollution effect. The package includes the calendar-aware tool, daily evidence, source hashes, a raw rebuild, exhaustive checks, and figures.

ozone episodes · missing data · exposure agreement · bird surveys · observation bias · calendar timing · reproducible tool