Farmers in Sub-Saharan Africa use fertilizer well below recommended rates for optimal productivity. One reason is lack of trust in the products available from local agri-dealers, says Hope Michelson, professor in the Department of Agricultural and Consumer Economics (ACE) at the University of Illinois.
Michelson collaborated with the International Institute of Tropical Agriculture (IITA) to evaluate fertilizer quality and farmer perceptions in Tanzania. Their findings inspired the idea of creating a mobile app that farmers could use to quickly and easily test if the fertilizer they were about to purchase was good.
In the true spirit of hands-on learning, Michelson enlisted then-graduate student Anna Fairbairn to help with the first part of a project that ultimately aims to help struggling countries grow more food and gain prosperity.
“I was looking for a master’s thesis advisor, and Hope and I share a lot of the same interests,” Fairbairn says. “She pitched this idea she had been working on about fertilizer quality, and I was immediately interested. I was struck by how simple of a problem that was; I was able to tell people who are not agricultural economists about it and they would understand why it would be important to look into.”
Fairbairn did a year of fieldwork in Tanzania’s Morogoro region. First, she worked with a Tanzanian team to find, survey, and visit all 225 agri-dealers in the region to purchase urea fertilizer, which she had shipped to Kenya and the U.S. for testing. Then, she conducted surveys and experimental sessions with farmers to gauge their perceptions of fertilizer quality and to estimate their willingness to pay for quality-verified fertilizer.
Fairbairn and Michelson found the urea samples for sale in the region were agronomically fine; yet, farmers widely suspected the products were adulterated – mixed with other non-fertilizer substances. Fairbairn’s research suggested that if farmers knew fertilizer was pure, they would be more willing to purchase and use it.
After graduating from U of I in 2017, Fairbairn went to work for One Acre Fund (OAF), a non-governmental organization supporting farmers in East Africa. Now based with OAF Tanzania, she continues to stay in touch with Michelson and provided guidance and assistance to another graduate student, Ben Norton, when he took over working on the project.
Ben’s data work
In conjunction with research for his ACE master’s thesis, one of Norton’s main tasks working for Michelson was to plan and manage the development of a mobile app to test fertilizer quality.
“We chose to focus on urea, because it is widely used, and it’s incredibly hard to adulterate profitably,” Norton explains.
“You can perhaps use salt, but you can see if salt is in the fertilizer, though it’s harder to do with your eyes than with computer vision. So the idea behind the app was if someone is adulterating fertilizer, we’re going to be able to detect it.”
Norton visited markets throughout the Morogoro region to find substances that could look like urea fertilizer and were cheap enough to use as adulterants. Back in the U.S., he doctored thousands of batches of urea with those products to generate a data set for a machine-learning-based image classifier.
Norton and Michelson partnered with undergraduate and graduate students in statistics and computer science at U of I to help create the architecture for the application and to train a neural network to distinguish between images of pure and adulterated fertilizer.
“I’ve probably taken a couple of thousand pictures of fertilizer, adulterated with different substances. It was like a craft project; I taped together different colored construction paper for backgrounds,” Norton explains. “The other part of my project was helping to coordinate the team of computer science and statistics students.”
Norton says the researchers encountered some challenges for making the app work with local conditions. For example, most farmers in Tanzania do not have smartphones. However, the team will enlist the help of local extension agents and village leaders, who can assist farmers with getting immediate results. Norton also credits partners at IITA with extensive support and help for the project.
Norton received his master’s degree from U of I in 2020 and is now a doctoral student at Cornell University. He and Michelson continue to work on developing the app, which is undergoing beta testing in Tanzania. The plan is to expand to other types of fertilizer, including blends.
The app provides a simple way to help farmers test urea fertilizer quality at the point of purchase or in the farmer’s field. Ultimately, it could help to increase agricultural productivity in low-income countries where farmers currently underuse fertilizer, Michelson says.
For more information:
University of Illinois