In-season maps illustrate the state of crops in US Corn Belt using ML methods

In 2019, the midwestern United States experienced major flooding in what was the wettest January-to-May period on record. The coincidence of these floods with the corn and soybean planting season caused many farmers to wait longer than usual to plant their crops, making this year's corn planting the longest delayed in US history. Additionally, tariffs imposed on US exports of soybeans to China led to depressed soybean prices and futures, causing many farmers to plant corn or other crops instead of soybeans. Due in large part to these factors, there is a great amount of uncertainty in the markets about corn yields for 2019 that will not be clarified until after harvest this fall. 

Dr. Hannah Kerner, a new hire with the University of Maryland's NASA Harvest Hub, is working to leverage remote sensing data and machine learning methods to deliver insights about the state and yield forecast of major crops (especially corn) in the US Corn Belt during the growing season, rather than after the season has ended and the crop has been harvested.

As extreme weather events such as the 2019 midwestern floods become more frequent, our models for predicting crop area and yields need to become more flexible to account for variability in planting timelines and other factors. Using machine learning tools to uncover complex patterns in the remote sensing data, we can produce more robust in-season crop type maps and yield forecasts that will provide critical information to policymakers, farmers, researchers, and markets.