Linking AI with Crop Simulation Models to Understand and Predict Agricultural Systems Dynamics

Newswise — October 26, 2020 – Accelerated crop improvement is needed to meet both global population growth and climate change generated stresses on crops. The “Linking AI with Crop Simulation Models to Understand and Predict Agricultural Systems Dynamics” symposium at the Translating Visionary Science to Practice ASA, CSSA, SSSA International Annual Meeting will address these topics.

The virtual meeting will be hosted Nov. 9-13, 2020 by the American Society of Agronomy, Crop Science Society of America and Soil Science Society of America. Media are invited; preregistration is required.

The presentations are:

1. “Synergies between AI and Crop Models for Long-Term Agricultural Sustainability,” presented by Bruno Basso, Michigan State University. Artificial Intelligence and Machine Learning have become important tools in agriculture because of the vast amount of data generated from different types of sensors and platforms. The quality of AI model performance in predicting outcomes is highly dependent on the quality and quantity of data fed into the models. In this presentation, Basso will describe how AI and ML are coupled with crop models to better understand spatial and temporal variability of crop yields and environmental impacts across the Midwest.
2. “Integrating AI with Process Understanding of Agroecosystem Dynamics,” presented by Kaiyu Guan, University of Illinois at Urbana-Champaign. In this presentation, Dr. Guan will share his lab's recent progress in integrating AI with process understanding of agroecosystem dynamics for predicting crop productivity, crop needs of nutrient, and water, and soil carbon sequestration.
3. “Metamodels for Crop Growth Models,” presented by Bernardo Maestrini, Wageningen University. To test the potential uses of metamodels, a research team created metamodels of Tipstar, a potato growth model, using different data-driven predictive tools, including neural networks and regression trees. The metamodels were trained on a synthetic dataset composed of simulations that differed in weather, soil and management typical of the cultivation of potato in the Netherlands. Different combinations of inputs and outputs where tested. In this presentation, Maestrini will report about a literature survey along with the results on the creation of metamodels to be used in the context of decision support systems.
4. “Coupling Machine Learning and Crop Modeling Improves Crop Yield Prediction in the US Corn Belt,” presented by Guiping Hu, Iowa State University. Technological advances with remarkable capabilities have provided farmers with precise and reliable predictions. Many of these tools, like simulation crop models, machine learning (ML) models, and remote sensing predictions, currently work separately. This presentation will discuss the results of research that merged prediction tools in hopes of increasing yield prediction, which suggest that adding APSIM forecasts as input features to ML models can make a significant improvement (up to 37%) in the prediction performance.

Presentations may be watched asynchronously, and there will be a scheduled Q&A time to speak with presenters during the meeting. Presentations will be available for online viewing for 90 days after the meeting for all registrants. For more information about the Translating Visionary Science to Practice 2020 meeting, visit https://www.acsmeetings.org/.

Media are invited to attend the conference. Pre-registration by Nov. 2, 2020 is required. Visit https://www.acsmeetings.org/media for registration information.