Mean-corrected climate forecasts had positive plant timing skill and negative butterfly timing and growth skill in 2015-2024

We tested whether simple updates repair old climate forecasts. We used signed annual forecast statistics for Swiss plant dates and British butterfly dates and growth. Original models used data through 2004. We estimated corrections in 2005-2014 and tested 2015-2024. The comparison model used the same recent observations to update its mean, with no climate term. Mean skill after a mean-bias correction was 0.691 for 19 spring plant phases, 0.369 for 8 later plant phases, -0.156 for 43 butterfly timing species and -0.173 for 36 butterfly growth species. Two-year block intervals excluded zero for both plant classes and growth. Independent-year intervals included zero for both plant classes. A fixed half-strength climate signal gave small uncertain positive butterfly gains. A recent mean reduced butterfly timing error more than the frozen climate model when both were compared with the old baseline: 28% versus 8%. Thus, the timing-versus-growth claim needs a limit: butterfly timing gain depends on the comparison model. Mean drift alone did not explain the growth forecast failure. This is a retrospective reuse of published statistics, not new field evidence. Short test periods, changing source coverage, observed weather inputs and conditional uncertainty limit inference.