It has long been known that aand bdiversity are driven by different ecological processes. Recently, several theoretical papers have suggested that b-diversity can be expressed in many different ways, and that these measures might have different meanings and behaviours. In terms of ecological understanding, it is important to test how these diversity measures behave in different settings, yet this has been rarely attempted. We investigated this question using forest openings in the Trascȃu Mountains, Romania, which contain species from grasslands of high nature value. The sampling was conducted in 40 openings, using edge-to-interior transects composed of 1-m 2 plots (a-diversity), from which we calculated five b-diversity measures at transect level. As predictor variables, we used canopy openness (from hemispherical photos), tree litter cover, heat load index, altitude, and bedrock type for a-diversity, and the means and ranges of these for b-diversity. Generalized linear mixed models showed that adiversity was mostly explained by the first two variables. Amongst the b-diversity measures, the classical additive and multiplicative measures differed, the latter being similar to the Sørensen-based multiple-site dissimilarity. None of the predictors explained the slope of distance decay or Simpsonbased multiple-site dissimilarity, except when considering ecological subsets of the species. We discuss the possible ecological processes underlying the different results, and the implication of our findings for nature conservation in the region. In conclusion, we support the joint application of different measures of aand bdiversity, as long as their particular properties are taken into account.