Introduction: Real-time grass field prediction is essential for optimizing harvesting operations and sustainable biomass management. However, current remote sensing approaches often provide deterministic predictions without explicit uncertainty quantification. Recent studies in precision viticulture demonstrated the potential of Bayesian Hierarchical Models (BHM) for integrating heterogeneous agricultural datasets and generating uncertainty-aware yield predictions. Building upon these developments, this project investigates probabilistic crop yield estimation for future real-time grass field applications. Methodologies: The methodology integrates multi-sensor remote sensing, meteorological, and on- farm data to generate high-resolution and uncertainty-aware yield predictions at field scale. The current work focuses on four years of winter wheat datasets collected from five fields in Denmark, using Bayesian Hierarchical Models to investigate uncertainty-aware yield mapping and the integration of relevant environmental covariates. The framework combines remote sensing observations with weather and historical field data to improve robustness and generalization under varying environmental conditions. In later stages, and with additional grass yield datasets collected within the BioReFarmeries project, the developed methodology and knowledge will be transferred to real-time grass field prediction and biomass management applications. Output: The project delivers high-resolution and uncertainty-aware field-scale yield predictions using Bayesian Hierarchical Models combined with remote sensing, meteorological, and on-farm data. The framework provides probabilistic yield maps together with uncertainty estimations and confidence intervals, supporting improved understanding of spatial and temporal crop variability. Impact: The proposed probabilistic modelling framework supports more reliable and interpretable agricultural predictions compared with deterministic approaches. By integrating uncertainty-aware estimation with multi-source environmental data, the project contributes to improved decision- making, better resource management, and reduced operational risks under varying climatic conditions. The developed methodology also supports future sustainable biomass harvesting and green biorefinery operations through more efficient real-time grass yield prediction and harvesting optimization. What I can offer • Expertise in remote sensing and agricultural AI What I need • Additional multi-year grass field datasets (yield and vegetation index measurements)

Source: Annual AgData Use Case Summit 2026 booklet.

• 3.2.2.2 Novel forecasting & prediction methodologies
• 3.2.2.4 Farm modelling for optimal practice