Uniform five-species sampling reversed the similarity change at 3,527 Swedish forest plots over 20 years

We tested whether a rise in similarity between recorded forest species lists remained when each list supplied five species. We used open Swedish National Forest Inventory data for 3,527 fixed plots. Each survey covered 100 square metres of productive forest land. Visits were exactly 20 years apart, in 1999–2004 and 2019–2024. On the same plots, full-list Sørensen similarity rose from 0.46520 to 0.48484. Expected similarity after uniform sampling of five species per list fell from 0.16865 to 0.16290. The changes were +1.963 and −0.574 percentage points. Their 95% spatial block intervals were +1.561 to +2.343 and −0.817 to −0.363 points. We computed the sampling expectation exactly for 6.22 million plot pairs. All planned checks had negative point estimates; the fifteen-species interval included zero. An exact species account showed that two mosses partly offset the decrease. Several widespread species gained recorded plots but lost share within lists. The result shows dependence on the species-count rule. It does not establish detection-corrected ecological divergence or invalidate the full-list measure. We used equal plot weights and made no national area-weighted or causal claim. The package preserves source extracts, provenance, all code, results, figures, independent checks and offline repeat instructions. Every source status cell matched the original archive.

forest vegetation · Swedish National Forest Inventory · species similarity · species sampling · presence–absence · reproducibility