Researchers propose new data system to quickly spot, stop the spread of crop plant diseases

Researchers propose new data system to quickly spot, stop the spread of crop plant diseases

06/15/2017

A plant pathologist spots disease in a soybean plot – but it could be too late.

Researchers at Iowa State University are trying to find a way to detect crop plant diseases earlier. They’re building a data analytics and machine learning framework that will allow farmers to act quickly to mitigate the disease’s spread. The data will come from ground robots, unmanned aerial systems and satellites.

Their research is one product of a broader effort that began two years ago through Data-Driven Discovery for Agricultural Innovation (D3AI), an interdisciplinary team led by Danny Singh, associate professor of agronomy; Baskar Ganapathysubramanian, associate professor of mechanical engineering; and Carolyn Lawrence-Dill, associate professor of genetics, development and cell biology.

D3AI is making strides in the collection, management, interpretation and use of data related to agriculture. It was funded by an ISU Presidential Interdisciplinary Research Initiative (PIRI) award, a program that is sponsored by the Office of the President and overseen by the Office of the Vice President for Research.

Soumik Sarkar, assistant professor of mechanical engineering, leads the crop plant disease data analytics project that came out of D3AI. Sarkar is working alongside Ganapathysubramanian, Danny Singh and Arti Singh, adjunct assistant professor of agronomy.

Photo: Iowa State researchers – from left, Danny Singh (agronomy), Soumik Sarkar (mechanical engineering), Arti Singh (agronomy) and Baskar Ganapathysubramanian (mechanical engineering) – are building a data analytics and machine learning framework for early detection and mitigation of crop plant diseases using data from ground robots, unmanned aerial systems (such as the device Sarkar is displaying here), and satellites. Photo provided by Soumik Sarkar

These researchers landed external funding in March through the National Science Foundation’s Cyber-Physical Systems program, with support from U.S. Department of Agriculture’s National Institute of Food and Agriculture (NIFA).

“With federal agency awards, if you just say, ‘We think that it should work’ – that’s not going to cut it,” Sarkar said. “That’s why the PIRI program was so important. It helped us build a tight collaboration and preliminary data to give them a credible story.”

Setting the stage for data collection

A plant pathologist finds diseases by crop scouting.

“It’s a slow process, and even an experienced crop scout or pathologist can’t thoroughly scout acres of a field in a day,” Arti Singh said. “So that’s where the need arises.”

Another challenge is in the variability of data collection.

“One person is fresh in the morning, and you evaluate the plants in the field more accurately. But late in the afternoon your ability to observe symptoms in the field decreases significantly,” Sarkar said. “In addition, we have a problem with inter-rater variability, for example, when you have five pathologists in the same field – they may come up with different observations.”

That’s where smart tools come into play: ground-based robotics, aerial vehicles such as UASs and fixed-wing aircrafts, and satellite imagery to detect signatures of disease.

Satellites will collect low-resolution imagery covering acres of land, and information from those images can be used to tell the UASs to zoom in on one area in finer resolution, and finally local sensors will collect images of a plant in that field. The team is collaborating with the NASA Jet Propulsion Laboratory to collect and understand satellite imagery.

The framework this NIFA team is developing will pull from these layers of data – plant-level digital images, multispectral and hyperspectral images, weather conditions, soil conditions, and computational model predictions – to identify and suggest disease mitigation strategies.

“By using larger images to detect disease symptoms, you are identifying the location of diseases and controlling them in a very efficient way,” Ganapathysubramanian said.

The ground-based platforms are being developed through a separate project where the team is collaborating with roboticist and assistant professor of mechanical engineering Sourabh Bhattacharya, via a grant awarded in December from USDA-NSF’s National Robotics Initiative program.

After collecting large amounts of data, “learning-based decision systems will remove a lot of these human biases and enable extremely high throughput,” Sarkar said.

The idea isn’t to remove humans from the equation, however.

“The approach actually improves human capability and helps the experts make more efficient and quick decisions, and eventually improve the productivity and profitability of the entire farm,” he said.

Last year, farmers learned how the project will directly impact them – and gave the researchers feedback on what will and won’t work.

“This summer there will be a lot of activity because we’ll get the opportunity to do field experiments and start developing the data analytics framework,” Sarkar said. “We really hope that we’ll be able to demonstrate some tools that we can off-ramp and actually market in the near future.”

What Sarkar’s NIFA team needs to figure out is how to make sense of the massive amount of data the system will collect.

“How do you combine all these data sources and make meaningful decisions? That’s the core of this problem,” Ganapathysubramanian said.

Collaboration key to success

The PIRI grant set the stage to eventually apply for the funding the team received this spring. About 30 faculty came together starting in 2015 to brainstorm about big data and agriculture.

This funding is also helping plant scientists and engineers overcome the barriers of collaborations in diverse disciplines and tackle current and future problems in food and environment.

“With advanced data analytics, I’m learning … a new language,” Danny Singh said. “The PIRI collaboration was essential for us to get to a stage where we could have more meaningful discussions in order to propose a big picture agriculture-related problem and solutions.”

Arti Singh pointed out that the project will have an impact beyond the fields.

“It is catering to the needs of farmers, researchers and the scientific community,” she said.

This tool – a data analysis framework that pulls together information on the crops and what plant diseases may be present – would be a huge benefit to farmers, Arti Singh said.

“The end product we are going to generate from this project is going to apply right in the farmers’ fields,” Arti Singh said. “This is what I think is the most beautiful part.”

Learn more about PIRI here.

By Chelsea Davis

Contacts:

Chelsea Davis, Office of the Vice President for Research, 515-294-0672

Soumik Sarkar, Mechanical Engineering, 515-294-5212

Copyright © 2017 Iowa State University.