Background: Under ongoing climate and land-use change, biodiversity is continuously decreasing and monitoring biodiversity is becoming increasingly important. National Forest Inventory (NFI) programmes provide valuable time-series data on biodiversity and thus contribute to assessments of the state and trends in biodiversity, as well as ecosystem functioning. Data quality in this context is of paramount relevance, particularly for ensuring a meaningful interpretation of changes. The Swiss NFI revisits about 8%-10% of its sample plots regularly in repeat surveys to supervise the quality of fieldwork.
Methods: We analysed the relevance of observer bias with equivalence tests, examined data quality objectives defined by the Swiss NFI instructors, and calculated the pseudo-turnover (PT) of species composition, that is, the percentage of species not observed by both teams. Three attributes of woody species richness from the latest Swiss NFI cycles (3 and 4) were analysed: occurrence of small tree and shrub species (1) on the sample plot and (2) at the forest edge, and (3) main shrub and trees species in the upper storey.
Results: We found equivalent results between regular and repeat surveys for all attributes. Data quality, however, was significantly below expectations in all cases, that is, as much as 20%-30% below the expected data quality limit of 70%-80% (proportion of observations that should not deviate from a predefined threshold). PT values were about 10%-20%, and the PT of two out of three attributes decreased significantly in NFI4. This type of uncertainty - typically caused by a mixture of overlooking and misidentifying species - should be considered carefully when interpreting change figures on species richness estimates from NFI data.
Conclusions: Our results provide important information on the data quality achieved in Swiss NFIs in terms of the reproducibility of the collected data. The three applied approaches proved to be effective for evaluating the quality of plot-level species richness and composition data in forest inventories and other biodiversity monitoring programmes. As such, they could also be recommended for assessing the quality of biodiversity indices derived from monitoring data.