MIND STEP Partners from the University of Bonn and IAMO have a new publication on Surrogate modeling of detailed farm-level models using state-of-the-art neural networks. The year of publication is 2022, but at the moment, it is in the pre-print stage as it is going through the peer-review process.
Abstract of the publication
Technological change co-determines agri-environmental performance and farm structural
transformation. Meaningful impact assessment of related policies requires farm-level models to be rich in technology details and environmental indicators, integrated with agent-based models capturing dynamic farm interaction. However, such integration faces considerable computational challenges affecting model development, debugging and application. Surrogate modelling using machine learning techniques may enable such integration for simulations with broad regional coverage. We develop surrogates of the farm model FarmDyn using state-of-the-art neural networks. All tested neural networks achieve a high fit but differ substantially in inference time. We develop evaluation metrics allowing practitioners to assess trade-offs among model fit, inference time and data requirements. The Multilayer Perceptron shows almost equal performance in all criteria but saves strongly on inference time compared to a Bi-directional Long Short Term Memory. If you would like to learn more about this scientific publication go to this link.