Artportalen record shares did not reproduce the 2005–2024 occupancy change of 138 understory plants in the Swedish forest inventory (r = −0.09), but gave their county distribution

Question: do opportunistic plant records give the same decadal change as a probability survey? Data: Swedish National Forest Inventory (NFI) presence/absence records of 242 listed taxa on permanent 100 m2 plots, 1995-2024 (main sample, plots paired by fixed coordinates), and Artportalen records from GBIF, counted by taxon, county (21) and period. Method: for three 10-year periods (1995-2004, 2005-2014, 2015-2024), we calculated county-weighted NFI plot occupancy on paired plots and the same county-weighted Artportalen record share (taxon records divided by all records of its plant group). We also used raw counts, the national share and a species-list reference. We compared log-ratio changes and latitudinal centroid shifts across species. NFI intervals come from a county-stratified plot bootstrap. Results: from 2005-2014 to 2015-2024 (138 species, 8,044 paired plots), the change in Artportalen share did not correlate with NFI occupancy change (r = -0.09, 95% interval -0.28 to 0.11; NFI reliability 0.81). For 68 species with a clear NFI change, the sign agreed for only 34%. The Artportalen change had a larger error than a 'no change' prediction (RMSE 0.41 vs 0.16 log units). Raw counts, the national share and the list reference also failed (r = -0.12 to -0.14). The agreement was weak for 1995-2004 to 2005-2014 (r = 0.20, 0.03 to 0.35) and absent for 1995-2004 to 2015-2024 (r = 0.08). Centroid shifts agreed weakly (r = 0.17, -0.03 to 0.35). Results did not change for date-matched plot pairs, all taxa, shade-tolerant species (EIVE light <= 5) or species x county cells (r = -0.04). Positive control: the same records reproduced the distribution across counties (median species Spearman rho 0.61). Artportalen changes, but not NFI changes, correlated with EIVE moisture and nitrogen values, which suggests changes in recording. Limits: we tested simple record-share indices, not occupancy-detection models; the NFI samples forest land and mires, while Artportalen covers all habitats; NFI detection was not modelled.

opportunistic records · citizen science · Artportalen · GBIF · national forest inventory · Sweden · understory plants · occupancy change · range shift · sampling effort · biodiversity monitoring · validation