AI identifies Senegal’s smallholder crops 84% of the time using limited training data

Food in Senegal is mostly grown on small-scale, rain-dependent farms, leaving much of the population vulnerable to climate shocks, according to the World Food Programme. However, satellite crop-mapping technologies designed to monitor the impact of climate on farming are largely beyond the reach of the West African country.

A new analysis of Senegalese croplands using an artificial intelligence model developed at the University of Cambridge shows how technological benefits enjoyed by industrial agriculture could be brought to smallholder farms in Senegal and other countries in the Global South.

Researchers used Cambridge’s Tessera, an open-source AI model trained on satellite images, to map crops in Senegal’s groundnut basin. They found that Tessera was more accurate than current methods, getting the right result 84% of the time in tests while using only a fraction of the computational resources and prelabeled data. In one scenario, it performed 28% better than the next-best model.

This suggests that Tessera could become an important tool for governments and food security organizations, particularly in areas in the Global South where ground data is difficult or expensive to gather, said lead author Madeline Lisaius, who helped develop Tessera when she was a doctoral student at Cambridge’s Department of Computer Science and Technology.

“Accurate and up-to-date crop statistics can guide food security planning and help decide where best to target support. But most local governments and bodies can only afford to collect ground data every few years,” Lisaius said. “With Tessera, you can train on the data you already have and extend it into the years in between, with more accurate crop information than baseline methods have ever been able to provide.”

Governments, NGOs and other food security organizations can begin using the technology to produce their own crop statistics now, she added.

The research is given added urgency by this year’s El Niño, which scientists say is the strongest ever recorded. The climate phenomenon is known to disrupt rainfall patterns in West Africa, and past strong events have brought prolonged drought to the region.

“Reliable agricultural data is essential to anticipate food security and climate-related risks. In Senegal, WFP is working with national partners to explore how geospatial data and artificial intelligence can strengthen food security monitoring systems and support faster, more informed decision-making,” said Pierre Lucas, representative and country director of the United Nations World Food Programme in Senegal.

The study, “Embedding-based Crop Type Classification in the Groundnut Basin of Senegal,” was published Sept. 29 in the journal Environmental Research: Food Systems.

Like much of West Africa, Senegal’s agriculture is dominated by smallholder producers, who grow crops on plots not much larger than a football field. As the paper makes clear, knowing what’s grown where allows for “informed decision-making at regional, national and global scales that can mean survival for vulnerable people.”

Currently, when governments and food security organizations estimate crop production, they base their calculations on information either collected on the ground, which can be painstakingly slow and hard to gather, or derived from expensive computational analyses of satellite imagery.

Tessera offers an alternative way of mapping crops, a key step in estimating production. Its underlying model analyzes a year’s worth of satellite images to compress each 10-meter (33-foot) point of land into a string of numbers, known as an embedding. That embedding carries information about how the land (and what grows on it) changes over time. A simple algorithm with a few calibration data points can then turn the embedding into a large-scale crop map.

In the study, the researchers compared Tessera with two satellite mapping methods widely used for monitoring agriculture, as well as Google DeepMind’s AlphaEarth, which is similar to Tessera but whose underlying model is not public. Each method was used to map crops for 2018, 2019 and 2021, then assessed according to accuracy, reliability, reusability and computational cost.

Besides outperforming the other methods, Tessera held up best when trained on one year’s data and applied to another. This suggests it can be used in later years without the need for new ground surveys every year.

The study has some limitations, in particular a drop in accuracy between 2018 and 2021 that the researchers think is linked to the quality of the ground survey data. It also did not test for secondary crops in fields where more than one crop is grown, which could change the overall accuracy in the most diverse locations.

Lisaius is keen to stress that Tessera’s biggest advantage is that it brings the power of sophisticated satellite analysis to those who don’t typically have access to it.

“One of the great contributions of this technology is not that it’s perfect, but that it’s incredibly accessible,” she said. “It’s a step toward greater geospatial data democratization.”

The current study builds on previously published research that demonstrated Tessera’s ability to map small fields in Austria.

Who’s behind this story?

Swati Mestri

Swati Mestri holds a bachelor’s degree in Electronics Engineering and has worked as a content editor since 2019. She has experience editing research documents across technology, health care, and materials science, and has a particular interest in technology and space.

Full profile →

Andrew Zinin

Master’s in physics with research experience. Long-time science news enthusiast. Plays key role in Science X’s editorial success.

Full profile →