Skip to main content

In-Season Field-Scale Mapping of Crop Types and Delays with Machine Learning

satellite image gif
What We Do

We are developing systems that use machine learning analysis of remote sensing data to predict where specific crops are planted and when at different stages during the growing season.


United States

How Satellites Make This Work

Commodities markets that rely on in-season predictions of planting progress, crop conditions, and expected yields experience high uncertainty during the growing season, since currently the most accurate field-level planting information is released after the harvest period has ended (e.g., by the USDA). Information about planting progress for specific crops during the growing season could help stabilize markets, enable earlier condition and yield forecasts, and inform comparative analyses of farming practices. We are using satellite data from Landsat-8 and Sentinel-2 to predict at a sub-field scale where specific crop types are planted and the relative timeline of planting across a region.

Hannah Kerner, Arizona State University
Inbal Becker-Reshef, University of Maryland
Team Members
Estefania Puricelli, University of Maryland
Ritvik Sahajpal, University of Maryland
Brian Barker, University of Maryland
Mehdi Hosseini, University of Maryland