An interdisciplinary research team at the University of Illinois at Urbana-Champaign has developed a convolutional neural network (CNN) that generates crop yield predictions, incorporating information from topographic variables such as soil electroconductivity, nitrogen levels, and seed rate treatments.
The team worked with data captured in 2017 and 2018 from the Data Intensive Farm Management project, in which seeds and nitrogen fertilizer were applied at varying rates across 226 fields in the Midwest U.S., Brazil, Argentina, and South Africa.
In addition, on-ground measurements were combined with high-resolution satellite images from PlanetLab to predict crop yields.
Said Illinois's Nicolas Martin, while "we don’t really know what is causing differences in yield responses to inputs across a field … the CNN can pick up on hidden patterns that may be causing a response.”
From Illinois ACES
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA
No entries found