Occurrence records recover moth flight timing but compress interannual phenological variation

Owen Russell Jones 

Phenological shifts are important indicators of climate-change responses, but large-scale inference requires coverage that structured monitoring rarely provides. Occurrence databases such as GBIF offer an alternative, although the reliability of different phenological metrics remains uncertain. I compared five metrics (Q10, Q50 and Q90 timing quantiles, weighted mean flight date and season width) derived from GBIF records for 34 UK moth species against the structured Rothamsted Insect Survey (RIS), using yearly and spatial analyses, and assessed agreement by absolute error, correlation and regression slope. Most GBIF records (97.1%) came from formal citizen-science recording schemes. In yearly analyses, timing quantiles and weighted mean flight date showed reasonable agreement with RIS, but substantial variation among species. GBIF consistently compressed interannual variation, recovering only about 60% of its amplitude even after correcting for benchmark measurement error. Season width tracked yearly patterns but with errors exceeding its natural variation, while peak timing was poorly recovered. Across latitude bands, central timing was most reliable, whereas Q90, season width and latitudinal contrasts showed weaker agreement. Temporal trends were generally poorly recovered; even Q50 trends, which showed moderate agreement, were strongly compressed. GBIF can provide moderately reliable estimates of seasonal timing, but season width, peak timing and temporal trends are less dependable. Compression of interannual variation also suggests that occurrence-derived estimates of phenological sensitivity to temperature will be biased towards zero. These findings provide a basis for metric selection, although their generality across regions and data sources requires further validation.