SEBASTIEN - A smarter livestock breeding through advanced services tailoring innovative and multi-source data to users’ needs

This study presents SEBASTIEN, a data-driven Decision Support System (DSS) designed to support smart livestock management by combining satellite observations, IoT sensor data, climate reanalysis and projections within a unified and scalable data platform. The system integrates multi-source data streams into a Data Lake architecture and implements Machine Learning and statistical models, including Gradient Boosting Machines and linear mixed models, developed through an AutoML workflow. SEBASTIEN delivers four main operational services: (i) shortand long-term prediction of the Temperature–Humidity Index (THI) for animal welfare assessment; (ii) estimation of milk yield and quality variations under heat stress; (iii) pasture biomass evaluation using satellite data; and (iv) disease risk mapping based on climatic and environmental drivers. The models are trained on largescale datasets, suggesting robustness and applicability across real farming conditions. Predictive performance indicates high accuracy (e.g., THI prediction RMSE = 2.59, R² = 0.95), supporting reliable decision-making. Outputs are provided through interactive dashboards, geospatial maps, and interoperable APIs, enabling both farm-level management and regional-scale monitoring. The DSS supports key applications such as early warning of heat stress, optimization of feeding and grazing strategies, breed selection under climate scenarios, and proactive disease risk mitigation. These results indicate that SEBASTIEN represents a promising operational DSS for enhancing livestock resilience, improving animal welfare, and supporting climate adaptation strategies through integrated, data-driven decision-making.