Yield Prediction Task Force
The Plant Sciences Institute established its Yield Prediction Task Force in 2015. The task force is pursuing three approaches to predict crop yields and total production (state-wide, nationally and ultimately globally).
One approach led by Dan Nettleton focuses on statistical approaches.
Dan Nettleton is a distinguished professor in the department of statistics
and the Laurence H. Baker Endowed Chair and director of the Laurence H. Baker Center for Bioinformatics and Biological Statistics. His research interests include statistical design and analysis of high-throughput experiments in biology and the development of statistical learning methods for prediction.
Dr. Nettleton has established a statistical research group consisting of faculty and graduate students in the department of statistics who are developing statistical methods for predicting phenotypes from genotypic and environmental data. Specific projects include the prediction of yield for a given variety in a given environment using historical yield trial results and weather information as training data, the prediction of multiple plant traits from DNA marker data and spatial locations of plants in a field, the prediction of plant phenotypes from DNA marker data combined with transcriptomic data, and improvement of the random forest methodology for general prediction problems.
Another approach led by Baskar Ganapathysubramanian and Sarkar Soumik, in collaboration with Dermot Hayes, focuses on using AI-based approaches.
Baskar Ganapathysubramanian is an associate professor in the department of mechanical engineering
. His research interests are in the areas of scientific computing and computational physics. His lab leverages advances in applied mathematics and high-performance computing to model, design and control real-world physical phenomena. From the application point-of-view, his lab is particularly interested in food, energy and environment-related phenomena. His group develops mathematical techniques and computational tools — model reduction, multi-scale frameworks, multi-physics simulators, control algorithms, data-driven methods, high-throughput computing pipelines — to efficiently model these systems. Most of these efforts are done in conjunction with experimental collaborators.
Dr. Ganapathysubramanian’s task force project focuses on high-throughput algorithms for image processing, data dimensionality reduction, as well as mechanistic models of plant growth. By leveraging machine learning tools with high performance computing, the project extracts large amount of time series phenotypic data from multiple sensors (field, flight and satellite). This is fed into the county level yield prediction project and used to infer conditional relationships among various agronomic, meteorological, and genotypic factors that affect yield.
Soumik Sarkar is an assistant professor in the department of mechanical engineering
. His research focuses on development of new data-driven learning and inference algorithms for autonomous perception and decision-making in complex cyber-physical systems with emphasis on health monitoring, decision frameworks, multi-agent systems, cyber-physical security and human-machine interaction for a large variety of application areas including building energy systems, plant science and agriculture, design and manufacturing and transportation systems. The nature of his research is inherently multi-disciplinary involving machine learning, information theory, theory of computation, dynamical systems, topology and statistical mechanics.
Dr. Sarkar’s task force project focuses on development data analytics and machine learning frameworks for county level corn yield prediction using long-term historical data involving weather and other environmental information, economic data as well as management practices. A key aspect of the project is to efficiently embed vast amount of domain knowledge into the data-driven models in order to perform periodic yield prediction throughout the growing season as well as to establish relationships among various factors involved in a complex agricultural system.
Dermot Hayes is a professor in the departments of economics
and finance and the Pioneer Hi-Bred International Chair in Agribusiness. In addition to his analysis of U.S. farm policy and international agricultural trade, his research interests include food safety, livestock modeling, demand analysis, and commodity markets.
Dr. Hayes’ task force project is using applied machine learning to predict crop production in each county in the US Corn Belt. It is a test run before doing the national data set. The results to date suggest that the model works well and the results compare favorably to classical statistical methods.
The third approach led by Michael Castellano and Sotirios Archontoulis focuses on process-based modeling.
Sotirios Archontoulis is an assistant professor in the department of agronomy
. His research goal is to deepen our understanding and model Genotype x Management x Environmental interactions towards improving production and environmental performance of various cropping systems. His approach combines the use of process-based models and field/lab experimentation of the soil-plant-atmosphere continuum.
Dr. Archontoulis’s task force project (in collaboration with Dr. Castellano) uses a process-based modeling platform coupled with field-based ground-truth measurements (FACTS: http://crops.extension.iastate.edu/facts/) to forecast crop yields and soil water and nitrogen status in real-time. The FACTS platform converts weather, soil, management, and genotype information into crop yields using mechanist science-based algorithms. Unique features of this approach are the abilities to determine yield gaps and conduct “what-if” scenario analyses that create strategies to close the gaps while improving economic and environmental performance of cropping systems.
Michael Castellano is an associate professor in the department of agronomy
and the William T. Frankenberger Professor of Soil Science. He uses expertise in soil science and ecosystem nitrogen dynamics to conduct systems-scale agronomic research that considers crop production and environmental outcomes.
Dr. Castellano’s task force project is to integrate soil processes and environmental nitrogen dynamics into the FACTS process-based cropping systems modeling platform http://crops.extension.iastate.edu/facts/ .