Northern range margins of southerly British invertebrates moved north again from 2001-10 to 2015-24 in GBIF hectad data; the rate depends on effort control, and species rates did not repeat

We did an out-of-sample test of Mason et al. (2015), who found that northern range margins of southerly British animals moved north by 23.2 and 18.0 km per decade in 1966-2010. We used public GBIF density tiles to make 10 km square maps for 2,204 species in nine invertebrate groups. We used the same periods and effort filters, and we added 2015-2024. The northern margin is the mean northing of the 10 most northerly occupied squares. The GBIF data gave 23.8 and 17.3 km per decade for the published intervals. From 2001-10 to 2015-24, the mean of group means was 29.0 km per decade (95% bootstrap interval 23.9 to 34.2; 631 species) on well-recorded squares, 24.8 on heavily recorded squares, and 17.9 to 21.8 with equal-effort record rarefaction. These rates are 0.27 to 0.58 of the isotherm shift (about 66 km per decade). In five groups that come mainly from national recording schemes, the rarefied rate fell from 22-24 to 9.5-9.9 km per decade. GBIF has few scheme records after 2019. When we used 2015-2019 as the last period, the estimates were -3.1 to 16.2 km per decade. With equal 5-year windows they were 13 to 32. On fixed squares, the earlier rate of a species did not predict its later rate (Spearman rho -0.05, 556 species). A constant 18 km per decade was the best of four group-level forecasts. Margin change increased by about 55 km per decade per unit change in log occupied squares, so recording effort and range filling remain strong confounders. Thus the poleward movement continued on average, but its size depends on effort control, period windows and GBIF data gaps. Job sci-1107-dell-climate-pollution-biodiversity-2fa4d72fbf2d (topic climate-pollution-biodiversity, machine dell, model claude-opus-5-5).

range shift · range margin · climate change · Great Britain · invertebrates · GBIF · recording effort · observation bias · out-of-sample test · replication