Ames, IA — Researchers at the Center for Agricultural and Rural Development at Iowa State University have created a new online tool to help Corn Belt farmers predict corn crop yields, harvested acreage, and total production. The PSI-CARD Corn Yield Prediction Project provides county-level predictions for South Dakota, Nebraska, Kansas, Minnesota, Iowa, Missouri, Illinois, Michigan, Indiana, and Ohio.
The daily corn yield predictions are calculated using a deep machine learning method, known as Long Short-Term Memory, which provides a high level of accuracy to predicted corn yields. The model provides predictions that are only .83 bushels per acre lower than actual yields, making it more accurate than the corresponding USDA predictions. Overall, about 80 percent of the county-level predictions fall within 20 bushels per acre of actual yields.
“In four years of out-of-sample testing, the model proved capable of giving accurate yield predictions at the Corn Belt level,” said Dermot Hayes, a professor of economics and one of the lead researchers on the project. “These yield predictions should prove valuable to anyone with a financial stake in national yields, including farmers and anyone trading corn futures.”
The project also includes a harvested acreage map, which uses USDA information to predict the difference in harvested acreage between the previous and current harvest years, and a total production map, which estimates the current year’s total corn production at the county level.
The PSI-CARD Corn Yield Prediction Project is free to use and available online at http://bit.ly/CARDCornYield.
The PSI-CARD Corn Yield Prediction Project was created in cooperation with the Plant Sciences Institute at Iowa State University (https://www.psi.iastate.edu/).