Mean flight date gave no clear forecast gain beyond abundance and trend for 12 UK summer butterflies in 2001–2024

We tested whether mean adult flight date improved next-year butterfly abundance forecasts beyond two years of abundance, a site baseline and a linear time trend. We used the fixed UKBMS 2024 source release and fitted each model with earlier targets only. The main test had 11,521 forecasts for 12 summer species at 455 sites in 2001–2024. The target was ln(1 + abundance index). The equal-species relative mean squared error score fell by 0.66% with flight date. The nominal 95% interval from 4,999 shared three-year blocks of test years was −0.44% to +1.73%. The 2015–2024 gain was 0.10%, with an interval of −0.75% to +0.87%. First and peak date checks also had intervals across zero. Without a time trend in either model, the full-period gain was 1.48%, with an interval of 0.30% to 2.66%. Thus, the gain depends on the baseline. These results do not show that timing has no ecological effect. They apply to a linear predictor, selected established populations and finalized annual indices. The intervals condition on fitted forecasts and the fixed species and sites. The package has source data, hashes, provenance, predictions, figures and code. A separate repeat matched all 47 numerical result files exactly.

butterflies · phenology · population forecasts · UKBMS · temporal validation · open data