An ozone and warmth product did not lower time-transfer error for annual US bird richness change in 1995-2024

We tested whether a prior ozone and warmth product improved predictions of annual recorded bird richness change across time periods. We used 23,357 consecutive same-observer survey pairs from a published US Breeding Bird Survey panel. We fit separate eastern and western models in 1995-2009 and tested them in 2010-2024. We also reversed the periods as a retrospective check. The product raised held-out mean squared error by 0.010-0.165 percent in the four tests. All four state and year resampling intervals included zero. Added ozone curvature, first-year exclusions and closer current monitors gave small error increases too. Thus, this fixed linear product gave no clear prediction gain. This does not establish that the joint biological effect is zero. Sparse western coverage, coarse weather, changing ozone monitors, remaining detection bias and uncertainty conditional on the training fits limit use. The package includes the signed source panel, survey pairs, predictions, code, source hashes and checks.

ozone · bird richness · temporal transfer · prediction · observation bias · seasonal temperature · null result