The main purpose of this study was to describe the relative importance of hay meadows and other types of semi-natural grasslands for species diversity and to focus on the impact of the fragmentation of hay meadows on species diversity. The study area was the rural districts of Hjartdal and Seljord in the county of Telemark, southern Norway. Interrelationships between the landscape and vegetation were revealed by combining research at three hierarchical levels, i.e. the landscape, biotope and species levels. A descriptive and analytical technique based on CIR aerial photo interpretation was developed for the landscape mesoscale. The historical background of the landscape and former biotope patterns were described, since traditional and sustained management is an important parameter in the landscape ecological analysis. The vegetation of different grassland biotopes was classified and numerically analysed. The results of the aerial photo interpretation revealed a landscape with considerable complexity and high biodiversity both at landscape level and at biotope and species level, due to the many traditional hay meadows and pastures still left in this area. The results also showed that the remaining fragments of traditional hay meadows are of vital importance for biological diversity. They had the highest average number of species per subplot and furthermore contained species that did not occur in any other grassland biotopes. In addition, the largest populations of several grassland species were found in the traditionally managed hay meadows. Unfertilised pastures and verges along small roads were also species-rich but were not in this connection substitutes for the hay meadows, nor were the `modern' hay meadows or the verges along the main roads. The spatial distribution and configuration of the biotopes and the abundance of meadow plant species within the study area suggest that in order to maintain species diversity it may be important to preserve and manage all the remaining fragments of traditional hay meadows. The surrounding environment, the matrix, should also be managed appropriately.