Mario Soto, one of CCIA’s Innovative Farmers, and his team recently published a monumental study in weed science. Shorter et al. investigated hyperspectral indicators of glyphosate-induced stress in common lambsquarters, Chenopodium album L. The study, published today in the journal Smart Agricultural Technology, demonstrates how innovative technology can increase efficiency and efficacy of herbicide evaluation. This key innovation is an essential enabler to advance agricultural productivity.
The research used an analytical method that included a field-usable spectroradiometer for data collection and a random forest machine learning algorithm for analysis. This cutting-edge approach has produced encouraging results, even achieving a margin of error that is as low as 12.1%. This is seen as an unsatisfactory result by the researchers who are committed to improving their methods. They hope to get their margin of error under 10%, the industry standard for herbicide efficacy rating.
Research Team and Methodology
The interdisciplinary research team includes many experts from the University of Arkansas. Among them are Kristofor Brye, University Professor of applied soil physics and pedology, and Aurelie Poncet, the assistant professor of precision agriculture who served as the study’s principal investigator. Wesley France, program associate in the crop, soil, and environmental sciences department, is one of the people behind the plan. Helping are Juan C. Velasquez, a graduate research assistant in weed science, and Mario Soto, a master’s student in crop, soil and environmental sciences.
To explore glyphosate-induced stress in common lambsquarters, the researchers used an integrative approach. These hyperspectral data from the spectroradiometer illuminate how plants respond to different herbicide treatments. The use of a random forest machine learning application means they were able to process this trove of data easily, comprehensively and reproducibly. What’s unique about this approach is the use of machine learning to identify subtle differences in plant responses. Unfortunately, these differences may be undetected with conventional assessment methods.
Importance of Accurate Herbicide Efficacy Assessment
Determining the efficacy of herbicides is an important aspect of practicing sound integrated weed management. Weed scientists are trained to judge such efficacy within an acceptable margin of error of 10%—minus 5%, plus 5%—so that shruggingly passes the mustard. This precision is key to reducing herbicide applications. It reduces negative environmental impact and advances regenerative agriculture.
Those same researchers want to get that margin of error under 10 percent. This goal reflects a bigger movement among agricultural interests to make advances in integrated weed management. They have developed advanced technologies including hyperspectral imaging and machine learning. This new approach opens doors to more precision applications of herbicides. If this research continues to bear fruit, it may mean higher crop yields and less herbicide resistance in the future.
Future Implications
The results from this specific experiment have profound ramifications not just for weed science moving forward, but for technology utilized by commercial agriculture. Our researchers are constantly fine-tuning their methodology. If developed further, this work has the promise of making more mechanistic models that more accurately predict plant responses. This has the potential to completely change how farmers manage weeds, making it possible to use herbicides more precisely and effectively.
The application of machine learning to agricultural research has ushered in a new paradigm. It spurs new ideas and brings scientists together to develop them. Computational weed science rapidly and dependably sifting through large datasets has the potential to catalyze major discoveries in weed science. This capacity has broader application across the industries of agriculture.
