Training and intercalibration reduce observer-induced variability in forest vegetation surveys

The increasing availability of vegetation-plot data results from decades of field surveys conducted by numerous observers. This highlights observer error not as a marginal issue but as a structural component of uncertainty, with direct consequences for estimates of species turnover and temporal trends. Observer-related sampling bias can be investigated through variation in species richness, species composition and abundance. Specifically, pseudoturnover refers to changes in species composition caused by overlooking or misidentification of taxa between sampling events or among different observers surveying the same plot. Although the causes and implications of observer error have been widely discussed, it remains unclear whether observer-related pseudoturnover decreases within observer groups as a result of training. Using data from training and intercalibration sessions carried out in 2023 and 2025 within two forest monitoring programmes in Italy (the LIFE project ModerNEC and the Italian National Forest Inventory), we assessed whether targeted training and collective briefing reduce observer-induced pseudoturnover. We applied Bayesian multilevel models to estimate changes in inter-observer species richness variability and inter-observer dissimilarity. The former decreased across observers in both years, while the latter declined after training when using Jaccard and Euclidean distances in both years; Bray–Curtis dissimilarity decreased only in 2023 and increased in 2025. Overall, training and intercalibration are likely to reduce observer-induced pseudoturnover related to species presence, while variability in abundance estimation needs further study and remains a key challenge for future vegetation monitoring programmes.