Early-period temperature models forecast later spring timing of Swiss plants and UK butterflies, but gave no forecast gain for UK butterfly or eastern US bird abundance (1951-2025)

Question: for which ecological responses does a temperature model fitted to an early period forecast a later period better than a model without climate? Method: one pre-specified protocol for 314 annual series in three monitoring systems. These are MeteoSwiss plant phenology (1951-2025), UK Butterfly Monitoring Scheme flight dates and abundance (England, 1976-2024) and Breeding Bird Survey abundance (eastern US, 1966-2025). Each series has fixed-effect year effects. We fitted simple monthly temperature models on early years and forecast later years (and the reverse). The score is skill against a no-climate baseline, with paired block-bootstrap intervals. We also simulated the skill expected if each model were stable. Results: median skill was 0.83 (95% CI 0.76 to 0.87) for Swiss spring phenology, which was equal to its stable expectation. The predicted shift (-7.8 days) agreed with the observed shift (-8.5 days). Skill was 0.43 (0.21 to 0.60) for UK butterfly flight dates, but this was no better than a linear trend, and the model explained about half of the observed advance. Skill was -0.67 (-1.23 to -0.29) for 4 long-record Swiss autumn phases, because the model predicted later dates and the dates became earlier. Skill was 0.00 (-0.10 to 0.07) for butterfly growth and -0.10 to 0.01 for resident, partial-migrant and migrant bird growth. The reversed split gave the same order. Training |t| predicted test skill for 67 timing series (Spearman 0.69). It did not predict test skill for 247 abundance series (-0.12). Only 48% of abundance series with a training p < 0.05 gave positive skill. Butterfly climate slopes decreased by about 70% in the test period. Bird winter effects were mostly too small to forecast. Carolina Wren's slope decreased from 0.114 to 0.063 per deg C. Limitations: simple linear monthly models, no detection models, only 4 long autumn series, brood mixing in flight dates. We added the bird system and a short Swiss split after the protocol was written.