Abstract ID: 3.14059 | Accepted as Talk | Talk/Oral | TBA | TBA

European mountain grasslands are complex agro-ecosystems shaped by environmental conditions and long-standing agro-silvo-pastoral activities. Climate change and land-use transformations are impacting their extent and provision of ecosystem services, such as landscape aesthetics, pollination, cultural heritage, and above all feed provision, highlighting the need for sustainable and adaptive management strategies. One key indicator for assessing grassland productivity and carrying capacity for livestock is the Pastoral Value (PV), which varies from 0 (low) to 100 (high), traditionally calculated through field-based time-consuming and specialistic methods. This study pioneers a spatial modeling framework to predict PV using machine learning (ML) models and remotely sensed data, reducing the need for labor-intensive field surveys. We combined field data from 390 vegetation surveys (2014–2019) and 40 remotely sensed variables describing vegetation phenology, climate, and topography. Several ML models were tested and the Forward Feature Selection algorithm, based on random forest, demonstrated the best performance (RMSE: 6.85, R²: 0.41, MAE: 5.23). To ensure reliable predictions, we implemented a cross-validation method designed to mitigate spatial autocorrelation. The model successfully predicted PV for central values but exhibited some difficulty with extreme values, likely due to input data limitations and a lack of extreme training samples. Despite challenges in predicting the highest and lowest PV values, the study provides valuable insights into the spatial distribution of grassland yield and quality in the Western Italian Alps. This framework demonstrates the potential of combining ML and remote sensing to improve the scalability and reproducibility of grassland assessments. The use of freely available, ready-to-use remote sensing products further enhances its applicability for designing sustainable management systems and supporting landscape-level conservation efforts.