DeltaCrop

Crop Growth Simulation Models (CGSM) are essential for predicting crop yields and understanding climate change impacts on agriculture. However, traditional CGSM face limitations in accurately representing reality and require complex parameterization. Recent advances in differentiable programming allows integrating AI into scientific models
2and offers a promising solution. By making CGSMs differentiable, we can better integrate them with machine learning which will enhance their predictive abilities and bridge the gap between data-driven and process-based modelling. Our project focuses on transforming the widely-used WOFOST crop growth model using differentiable programming frameworks. This will enable more efficient model training and deployment, especially for large-scale datasets and complex simulations. Partnering with the eScience Center, we'll leverage their expertise in high-performance computing and data assimilation for global-scale food security studies. This advancement holds potential for revolutionising agricultural research, facilitating hybrid models combining deep learning and traditional approaches, and improving food security worldwide.