European semi-natural grasslands host numerous and diverse living organisms within a relatively small area; therefore, conservation of these habitats is crucial to maintain high biodiversity in agricultural areas. The aim of this study was to examine the effects of environment, habitat continuity and spatial structure on semi-natural grassland species richness (SR) at the regional extent. SR and characteristic SR (CHSR) were calculated in 709 georeferenced vegetation plots obtained from the Polish Vegetation Database. For each vegetation plot, (1) environmental variables representing climate, topography, soil characteristics (e.g. pH and physical soil properties), landscape and anthropogenic pressure; (2) habitat continuity in time and (3) variables reflecting the spatial structure of vegetation plots using principal coordinates of neighbour matrices were measured. Environmental data were measured at four different spatial scales (buffers of 0.05-, 0.5-, 2.5- and 5-km radii surrounding each vegetation plot). Boosted regression trees were generated to explore the relationships of SR and CHSR with predictors. Variables associated with anthropogenic pressure, landscape structure and soil characteristics operating in the 0.5- and 5-km buffers significantly improved SR and CHSR model. However, effects of individual predictors on SR and CHSR were not predominant. Habitat continuity and topographic heterogeneity operating in the 0.05-km buffer exclusively improved CHSR model. The results suggest that neutral processes (dispersion limitation, local management history and/or stochastic events) reflected by spatial variables are strongly related to SR and CHSR. Patterns observed herein may also exist in other agriculture-dominated regions in temperate Europe. This knowledge will contribute to effective conservation of valuable habitats over large spatial areas.