News

Project partners researchers, librarians and AI to fight hunger

Experimental wheat varieties grow under severe drought stress near Njoro, Kenya. Ceres2030 aims to use machine learning, librarian expertise and cutting-edge research synthesis to improve the lives of the world's poorest farmers while preserving the environment.

For smallholder farmers in the developing world, adopting crops hardy enough to adapt to the droughts, floods, heat waves and frosts of a changing climate is increasingly essential.

But research sometimes points in different directions. So it can be hard for policymakers to decide where to dedicate limited funds, and how best to help farmers adopt the right crops.

Ceres2030 , a global effort led by International Programs in the College of Agriculture and Life Sciences ( IP-CALS), the International Food Policy Research Institute and the International Institute of Sustainable Development, is employing machine learning, librarian expertise and cutting-edge research analysis to use existing knowledge to help solve these and other challenges – all aimed at eliminating hunger by 2030.

“We have an opportunity to achieve higher food security, create a safety blanket to cope with climate shocks and improve overall livelihoods,” said Maricelis Acevedo , associate director of IP-CALS’ Delivering Genetic Gain in Wheat Project and lead author of a forthcoming Ceres2030 study. “Synthesizing the available scientific evidence can help scientists, policymakers and governments understand how to increase the utilization of climate-resilient crops.”

Climate-resilient crops are at the center of one of Ceres2030’s eight research questions, which seek to identify interventions that will improve the lives of the world’s poorest farmers while preserving the environment. The questions – built around the United Nations’ Sustainable Development Goals – range from reducing food loss to providing better agricultural skills training for young people in Latin America, Africa and Asia.

“We want to generate better evidence; we want to figure out what solutions we should be investing in right now,” said Jaron Porciello, primary investigator of Ceres2030 and associate director of research data engagement at IP-CALS. “For good public policy, we need to know what works. Science is always looking for the next horizon. It’s not a criticism of either side. But these are two adjacent systems that want to interact. They largely rely on the same evidence base for answers, but we don’t have a bridge to help them.”

The eight teams, including 77 researchers from 23 countries, are employing evidence-based synthesis – rigorous literature analyses that bring together all the studies on a particular issue to evaluate what they mean. These syntheses, which rely on transparency and rigor so other researchers can repeat the process and reach the same conclusions, are widely used in medicine but relatively new to agriculture.

“In this world of information overload, it’s important for researchers in all fields to understand that literature reviews that will inform decision-making must be performed in a methodical fashion,” said Kate Ghezzi-Kopel, health sciences and evidence synthesis librarian at Cornell University Library and a member of the climate-resilient crops team.

Librarians said the syntheses will evaluate how interventions that are successful in one geographic location will work elsewhere, too.

“Agriculture has geographic specificity – it isn’t the same all over the world or sometimes even 50 miles down the road,” said Mary Ochs ’79, director of Mann Library. “So you can’t just assume that one solution is going to work everywhere. But by pulling together all of the solutions through comprehensive systematic reviews, you can identify the approaches that have worked and consider whether they’re likely to work elsewhere.

In addition to embedding librarians on the teams, the Ceres2030 project employed machine learning to refine its research questions. With help from colleagues in Cornell’s Faculty of Computing and Information Science and around the world, Porciello created a machine learning model that accelerated the evidence synthesis. It also broadened the synthesis’ scope, making the process more inclusive.

“We wanted to find young, emerging researchers who are doing really excellent work on these types of interventions in Africa, Asia and Latin America,” she said. “We wanted to bring in voices of scientists we haven’t necessarily heard from before.”

Other Ceres2030 research questions relate to topics including livestock feed and water scarcity, food loss, how small-scale producers engage with food systems, farmers’ organizations and policies for sustainable practices.

Though she doesn’t yet know the results, Acevedo said Ceres2030’s method – including the close collaboration with librarians – is an exciting approach for agricultural science.

“Before joining the Ceres2030 team, I knew little about evidence synthesis in agriculture. Once I learned about it, I was hooked,” she said. “Working on this project allowed me to rediscover the power of collaborating with talented librarians who can help navigate and get the most out of the sea of literature that is accessible to us here at Cornell and other partner institutions.”

Funding support for Ceres2030 comes from Germany’s Federal Ministry of Economic Cooperation and Development, and the Bill & Melinda Gates Foundation.

This story was previously published in the Cornell Chronicle