Observer-expertise bias in multiscale forest vegetation surveys: effects on species detection, richness and composition across spatial grains and forest ontexts

Observer-related uncertainty can affect vegetation surveys, but its magnitude and ecological consequences remain poorly quantified in forest understories. We assessed observer-expertise-related variation using independent presence–absence records collected by three observer categories differing in botanical experience, calibration history and familiarity with nested permanent plots in six Tuscan forest stands. The design combined three spatial grains (1, 10 and 100 m²), replicated stands within forest type and repeated observations of identical sampling units. We quantified effects on pooled taxon records, stratum-specific floristic record richness, three-observer concordance, pairwise pseudoturnover and species composition. Differences among observer-expertise categories were measurable but strongly context dependent. Effects on record-based richness were statistically significant in only two of eighteen site × grain combinations, indicating limited richness effects rather than a general expert–novice gradient. By contrast, concordance and pseudoturnover revealed broader disagreement in species detection: agreement was highest in beech forests and lower in species-rich, structurally complex oak stands. Plot size altered disagreement, but the direction and magnitude of this effect depended on forest context. PERMANOVA showed that site and plot size explained most compositional variation, whereas observer-expertise-related compositional differences were restricted to oak forests and were primarily site dependent. Thus, observer-expertise variation did not override the main ecological gradients, but it affected field outputs, reduced concordance, inflated apparent turnover and altered some compositional comparisons. Because each expertise category was represented by a single observer team, these effects should be interpreted as observer-expertise-related variation rather than as bias among observers with comparable expertise.