Scaling species richness: When vegetation surveys don't match in plot area

Abstract Differences in study plot area remain a central obstacle to the analysis of biodiversity data. Plot area (i.e. the surface assessed for a single observation) in vegetation surveys is rarely standardized and often varies across studies, complicating comparisons. The species–area relationship (SAR), often modelled as a power law, provides a simple means of scaling species richness across different plot areas. Yet, more than a century after its introduction, applying the power law to compare vegetation surveys conducted at different plot areas still lacks thorough evaluation and quantification of uncertainty. Using a unique dataset with over 1000 nested‐plot vegetation surveys from Swiss grasslands, with subplot sizes ranging from 1 to 100 m 2 , we assessed the scaling of species richness across plot areas. To do so, we compared SAR models with machine learning approaches with area parameters as covariates. Additionally, we examined how ecological indicator values (EIVs) vary with plot area, evaluating their potential integration into studies based on scaled species richness. We found that species richness can be reliably scaled across surveys differing in plot area. SAR models using a constant accumulation coefficient offered a simple and effective approach, particularly for small differences in plot area between surveys. In contrast, machine learning models, including area parameters as covariates showed greater accuracy when extrapolating across surveys with widely differing plot areas. EIVs differed only little across plot areas, with differences remaining well below the overall variability observed between surveys—even when comparing widely differing plot areas. This suggests that indicator values are largely independent of spatial grain and can therefore be used as stable covariates for characterizing vegetation, even when data are derived from differing plot areas. Synthesis . This study shows that species richness in grasslands can be reliably harmonized across surveys differing in plot area. It also highlights that EIVs remain largely stable across different plot areas, making them suitable for the use in data‐rich ecological analyses. Together, these findings provide practical tools for harmonizing grassland datasets and offer new insights into how we can compare biodiversity assessments across space and time.