Crop yield forecasting is increasingly critical for addressing global food security in the face of climate change and extreme weather events. In Denmark, accurate yield estimates are also essential for regulatory compliance, as farmers must report predicted yields when planning fertilizer use under national nitrogen quotas. Crop yield prediction and simulation models are valuable not only for forecasting, but also for evaluating alternative management scenarios and reconstructing historical yield records where data are missing.

Despite the availability of detailed agricultural records and high-resolution satellite imagery, Denmark currently lacks a comprehensive, field-level crop yield prediction framework. The Danish National Inventory currently reports yield at aggregated regional levels, which obscures the fine-scale variability between individual fields and limits the potential for precision agriculture and targeted interventions. Existing field-level yield data is limited to a few small catchments, leaving major spatial gaps across the country. This project addresses that gap by developing and systematically comparing four modelling approaches: machine learning prediction, process-based simulation, and two separate methods that integrate remote sensing data into process-based simulation. The resulting models will be used to create a historical field-level yield database for Denmark, providing a clearer understanding of spatial and temporal yield variability across Danish croplands.