IMPACT OF LAND-USE CHANGE ON GRASSLANDS OF THE WESTERN ITALIAN ALPS AND MANAGEMENT IMPLICATIONS: A MULTIDISCIPLINARY APPROACH

Semi-natural mountain grasslands are key components of European mountain landscapes, supporting high biodiversity and essential ecosystem services shaped by centuries of low-intensity agro-pastoral management. However, in the Alps, rural depopulation, land abandonment, and the decline of traditional practices have triggered widespread forest expansion and grassland loss, fragmentation, and degradation. These processes threaten both habitat conservation and forage provision, while also altering disturbance regimes and creating new challenges for the sustainable management of mountain landscapes. This thesis investigates the ecological consequences of land-use change on semi-natural grasslands in the western Italian Alps. By integrating landscape ecology, vegetation and forest ecology, remote sensing, and machine learning, the research is organised around three objectives: assessing the influence of historical and recent landscape structure on plant diversity and habitat conservation status of semi-natural dry grasslands and their post-disturbance responses; analysing how land-use legacies shape the structure and composition of post-abandonment forests; and developing a remote sensing and machine learning framework to estimate pastoral value and grassland carrying capacity at large spatial scales. The first study shows that the current plant diversity and conservation status of semi-natural dry grasslands cannot be interpreted solely through present-day landscape structure. Historical grassland continuity and extent contribute to explaining the persistence of better-preserved plant communities, highlighting the role of land-use legacies in shaping current grassland conditions. Moreover, grassland responses to disturbance depend strongly on pre-disturbance habitat conditions, indicating that the effects of events such as wildfire are filtered by the ecological state inherited from past land-use trajectories. The second study demonstrates that forest expansion after abandonment does not produce a homogeneous forest matrix, but rather a mosaic of post-abandonment forests shaped by previous land use. Forests established on former grasslands still retain structural and floristic signals of open habitats, whereas forests derived from former wood-pastures or sparse forests show more advanced successional characteristics. These differences indicate that land-use legacies influence both overstory and understory development, providing relevant criteria to distinguish areas where active restoration of open habitats may be more ecologically feasible. The third study addresses the need for operational tools to support sustainable grazing management in mountain regions. By combining georeferenced vegetation plots, remotely sensed predictors, and machine learning algorithms, the study developed a spatially explicit approach for predicting pastoral value, a proxy for forage quality and productivity and a key parameter for estimating grassland carrying capacity. This approach allows pastoral resources to be assessed at large scales beyond the limits of field-based surveys, supporting targeted management of grassland resources. Overall, this thesis demonstrates that the management of mountain grasslands requires a multi-scale and legacy-based approach. Historical land-use information, current habitat conditions, forest successional trajectories, and spatially explicit assessments of pastoral resources provide complementary knowledge to support grassland conservation, restoration, and management.