UMN researchers to use machine learning to observe global change

The University of Minnesota was granted $1.43 million to conduct the project.

Nikki Pederson

The University of Minnesota received a three-year, $1.43 million grant earlier this month from the National Science Foundation to further advance machine learning techniques to better monitor global agriculture and environmental changes.

Machine learning is when computers can “learn” from data, without needing direct human programming. Through this project, it can be used to help society address climate change issues, manage land use and natural resources and sustainably feed a growing population.

The NSF grant funds three groups of researchers from the University, consisting of members of the College of Science and Engineering, College of Food, Agricultural and Natural Resource Sciences and the Minnesota Supercomputing Institute. Each group will focus on a different aspect of the project.

Vipin Kumar, the project’s principal investigator, will lead the CSE portion of the project involving the actual algorithms.

“The end goal is to advance state-of-the-art machine learning with automated algorithms,” Kumar said. “We can identify what crops are being grown at what locations, and [be] able to figure out what the yield of the crop would be as soon as possible.”

James Wilgenbusch, the co-principal investigator for the project, and MSI are tasked with cyber infrastructure development.

“How do we implement this in a way that people can [use] what algorithms are being designed, and that there are interfaces that people can use to access the data discoveries that we make?” Wilgenbusch said.

The third leg of the project involves CFANS and Phil Pardey, a professor in the Department of Applied Economics.

“[Pardey has] a good sense of what potential value there will be to the methods we’re developing, when it comes to farmers and other people with a vested interest,” Wilgenbusch said.

The triad of knowledge can be summarized as “good algorithms, good infrastructure and good actual application support around agriculture and economics,” he said.

By analyzing agricultural cropping data and urban landscapes, researchers plan to advance machine learning techniques that can monitor changes such as the conversion of forest to farmland, the loss of farmland due to urbanization and soil and water degradation, according to the grant brief.

The reach of these advancements could extend past Minnesota. The project is collaborating with the Nature Conservancy and DC Water to evaluate the effectiveness of the project’s techniques and ways to implement them into further sectors.

Agricultural and land changes that have happened in the past, such as the ones being monitored by the University, are still occurring now and can have a direct impact on water quality, said Matt Ries, the chief of water quality and watershed management at DC Water.

“One of the things we pride ourselves on is what we call a digital utility,” Ries said. “That involves operations systems and transparency of the data we have. We have an open water data portal here where you can find information about our construction and information about our infrastructure.”

Ries said that it’s somewhat unique that a water municipality has high-tech approaches such as theirs, but it reflects a general trend in society about open data and information.

“Ultimately, we are looking to provide the best value for our [customers] for the services we provide,” he said. “If we can use new technologies to do that, we are interested in exploring them.”