High-ozone days gave more three-day runs than fixed-count random order near selected US bird surveys in 2005-2019

We tested the order of high-ozone days in 1,120 complete exposure windows before selected bird surveys in four US states in 2005-2019. We compared observed overlapping three-day runs above 70 ppb with exact expectations from random date order at fixed high-day counts. Observed counts were 2.14-15.54 times the 30-day expectations in two regions and two periods. They were 1.37-3.19 times expectations when counts were fixed within five-day blocks. The excess stayed positive with second sites, complete paired sites, equal site weights, and one window per site-year. No single site or year deletion removed the excess. The later eastern sample had only three runs. At 80 ppb, both eastern main samples had no runs. Selected windows covered 20-39% of filtered surveys and had warmer survey starts. These are finite-sample exposure results. Repeated dates, selected monitors, rare events, and assumed random-order references limit inference. We do not estimate weather causes or bird effects. The package gives daily data, exact calculations, source hashes, all checks, and code.

ozone · bird surveys · exposure timing · three-day runs · conditional expectation · observation bias · temporal clustering