Winter feeder-bird communities in North America kept shifting to warm-range species in 2012–2024, but the shift did not follow year-to-year or local winter temperature (FeederWatch, 1989–2024)

Question: A 2015 study of Project FeederWatch data for 1989–2011 reported that winter bird communities in eastern North America changed to warm-adapted species, and that the community temperature index (CTI) tracked winter temperature at scales of 1 to 35 years. Did these results continue in new winters, and do they hold in a second region? Approach: We used the May 2024 FeederWatch release. The data were Dec–Feb counts at 67,842 eastern sites (243,683 site-winters, 54 species) and 16,530 western sites (86 species). We calculated CTI year effects with site fixed effects and GHCN-CAMS temperature. We kept 1989–2011 as calibration and 2012–2024 as the test period. We compared temporal responses with the spatial CTI–climate slope, tested trailing temperature windows out of sample, compared cross-site trends, and used a species-shuffle null and 14 robustness variants. Main findings: Within sites, the eastern CTI increased 0.16 ± 0.05 °C per decade in 1989–2011 and 0.24 ± 0.10 in 2012–2024. The western CTI increased 0.21 ± 0.02 °C per decade, five times the full-tracking expectation, because western winters did not warm. All variants gave an increase. The CTI did not follow temperature in time. Against a spatial slope of 0.33 per °C, the regional detrended response was −0.005 ± 0.016 per °C, the local anomaly response was 0.005 ± 0.003, and lags of 0–2 winters summed to 0.012 ± 0.007. In 1989–2011 the CTI increased while eastern winter temperature did not (−0.32 ± 0.44 °C per decade). Cells that warmed more had larger CTI increases in 1989–2011 (0.20 ± 0.04 per °C), but not in 2012–2024 (−0.06 ± 0.04 east, −0.17 ± 0.04 west). Windows of 1–3 years had no predictive skill. Windows of 30–35 years predicted 2012–2024 only as well as a linear time trend (RMSE 0.13–0.14 °C). The observed trend was near the upper edge of the species-shuffle null (p = 0.08 east, 0.14 west), and species trends did not increase with species temperature index. Limitations: This is a conceptual replication; the original study used minimum temperature and its exact species list was not available. With site effects, a linear site-tenure effect cannot be separated from the period trend; a cell-effect variant gave the same trend. The slow drivers (feeding, urban growth, non-climatic population trends, long-term climate) cannot be separated with these data.

community temperature index · Project FeederWatch · winter birds · climate tracking · replication · out-of-sample test · citizen science · species temperature index · North America · space-for-time