UMD Researchers Help Kenya Build Its Agricultural Data Infrastructure

Despite advances in remote sensing and the growing capability of artificial intelligence for analysis, the pipeline that carries important agricultural data to farmers is still too slow.

Steered by Kenya’s Ministry of Agriculture and other government bodies in the East African nation, researchers with the University of Maryland-led NASA Harvest program are working to help end that time lag. They’re supported by a recent $4.9 million grant from the Gates Foundation and $2.5 million in in-kind support from the Microsoft AI for Good Lab.

The team, led by NASA Harvest Director Inbal Becker-Reshef, is guided by Kenya’s requirements for a system through which it can use satellite images and AI to consistently produce its own, in-season agricultural data products that government officials, farmers, agribusinesses and insurers can depend and act on in a more timely way.

Called AI‑enabled Data for Agriculture Products & Transformation for Kenya (ADAPT-Kenya), the initiative brings together working groups focused on cropland and crop-type mapping, area estimates and crop statistics, yield estimation, early monitoring, insurance services, and market and trade intelligence—and has already supported the government’s response to the current drought.

“We’re really aiming to develop something that will be sustained into the future; a geospatial agricultural data backbone that is owned and run by the government, and, if we are successful, scale this out to other countries,” explained Becker-Reshef, a research professor in the Department of Geographical Sciences and managing director of the Microsoft AI for Good Lab.

The researchers decided to start with Kenya given their work on other projects in the area—like the GEOGLAM Crop Monitors—and the country’s agricultural diversity, which could help scale up the system in other countries in the future. In addition, the Microsoft AI for Good Lab has a team of AI scientists situated in Nairobi, Kenya’s capital, and has frequently worked with domestic and international partners.

Ultimately, the project will provide national cropland and crop-type maps, yield forecasts and crop-condition monitoring reports, and field boundary datasets that are to be made openly available to the public when there is still time to adjust. The team also plans to create systems that can provide early warnings about droughts, floods or other natural disasters.

Most countries around the world lack a system that regularly produces timely agricultural data products like NASA Harvest and Kenya are setting out to establish, Becker-Reshef said. In the United States, for example, the Cropland Data Layer—which shows what crops are growing in the country’s fields—is released to the public several months after the growing season.

“This is a very ambitious project; we have no illusions about that. But we feel like this is something that could really substantially support a lot of data and information products that are really critical across the agricultural space,” said Becker-Reshef. “This will be a real test of that.”