Using species and environmental data from an extensive grassland area in south-western Finland, we investigated the eVect of patch area and connectivity, management and local habitat variables on the occurrence of spring-Xowering vascular plants and their richness in boreal agricultural landscapes. Generalized linear models (GLM) and variation partitioning were used to study the explanatory power of the three groups of variables and their combined contributions on the richness and occurrence of six spring-Xowering plant species. Generalized additive models (GAMs) and associated cross-validation tests were used to evaluate the predictability of the species occurrence and richness patterns. Present-day grassland patch area and connectivity were important predictors for occurrence and richness of the studied plant species. In addition, local habitat factors, especially radiation, accounted for major fractions of occurrence patterns of the studied species. Hybrid models including variables from all three variable groups had higher explanatory power and predictive capability than partial models. However, performance of the separate singlespecies models varied considerably between the six study species. Exclusion of radiation or connectivity from the hybrid models decreased their predictive performance, suggesting that these factors are of particular importance for grassland plant species at their northern range margins. When developing conservation and management planning for grassland plant species in Northern Europe, attention should be paid to well-connected networks of grassland patches including large, steeply-sloped patches with a favorable microclimate.