Warm springs act like a 1.5-day later survey per 1 C in Breeding Bird Survey counts of 317 species (1966-2025); this phenology-detection artefact is real but too small to explain weak climate tracking

Question: the Breeding Bird Survey (BBS) counts each route once per year on a calendar date. In a warm spring the breeding cycle is earlier, so a warm spring must act like a later survey date. Does this phenology-detection artefact explain part of the weak climate tracking of bird communities? Data: 130,303 standard BBS runs on 5,319 US and Canadian routes (2026 release, 1966-2025), GHCN_CAMS April-May temperature anomalies, 317 species; UKBMS 2024 flight periods and site indices of 48 butterfly species with E-OBS temperature. Method: per-species models with route x observer and year fixed effects gave the within-route survey-date slope and the spring-anomaly response; a cross-fitted (odd/even years) random-effects meta-regression across species gave the implied phenological advance k. We wrote the plan before the outcomes. Results: species whose counts fall through the season had lower counts after warm springs. k was 1.45 days per 1 C (species bootstrap 1.19-1.72; permutation p = 0.001), 1.52 with migration, temperature-index and abundance covariates, 1.40 with family effects, and 1.24 from the within-species anomaly x date term. The previous spring (-0.07) and the next spring (-0.02) gave no relation. Residents gave 2.2 and migrants 1.3 days per 1 C (exploratory). The artefact explains about 20% of between-species variance in spring responses, but its effect on the community temperature index trend is +0.0003 C per decade, and survey-date drift adds -0.0011, against an observed 0.024 C per decade. In butterflies, flight dates advanced 3.9 days per 1 C, and an emulated single visit changed same-summer responses by 5-13% per 1 C, as much as the true response; the emulation showed that the count-based estimator recovers about 77% of k. Conclusion: the artefact is real and must be corrected in single-species weather responses, but it does not explain the tracking deficit. Limitations: correlational design, no breeding-date data, check terms after the survey gave small slopes (about one sixth of the main value), coarse temperature, Gaussian flight curves.