What happens beneath the soil surface shapes everything above it — crop productivity, water quality, nutrient efficiency and climate resilience. Yet for decades, understanding subsurface soils has depended on slow laboratory methods. Besides timeliness, these methods provide only limited snapshots of what is happening underground. Now, new advances in AI soil imaging are providing unprecedented insight into the underground world.
Researchers in the UF/IFAS Department of Soil, Water, and Ecosystem Sciences (SWES) are working to improve the process. A recent publication demonstrates how AI and advanced imaging technology can rapidly analyze microscopic soil structures directly in the field. Using a multi-sensor probe called the Digital Soil Core (DSC)TM developed by Landscan.AI, they captured high-resolution images of soils 120 centimeters below the surface. Then they trained a deep-learning model to identify soil particles and pore spaces that control water, nutrients and air movement underground.
A Research Partnership
“Our work shows AI can help us characterize subsurface soils quickly and accurately,” said Perseverança Mungofa, SWES doctoral candidate and the study’s lead author. “Doing this in real-world field conditions creates new opportunities for smarter and more sustainable soil management.”
The project also highlights a growing collaboration between academia and industry. LandScan, the company that developed the DSCTM technology, partnered with the SWES team. Mungofa first connected with LandScan during a summer internship in 2023. Since then, she has continued collaborating with the company while receiving mentorship and professional development support.
“It gave me the opportunity to work with emerging technologies and see how research can move from theory to application,” Mungofa said.
Sabine Grunwald, SWES professor and co-chair of Mungofa’s dissertation committee, previously collaborated with LandScan on related research. She also serves as a scientific adviser to the company.
“This shows how university research and industry innovation can come together to accelerate advances in sustainable agriculture,” Grunwald said. “We are moving toward real-time soil intelligence systems that can support decision-making in the field.”
How AI Soil Imaging Is Transforming Soil Science
Traditional soil analysis often relies on disturbed samples collected from shallow depths and processed in laboratories. By contrast, the DSCTM system preserves natural soil structure while continuously imaging soils at fine depth intervals. That matters because many critical agricultural and environmental processes happen below the surface. These include root growth, water infiltration, nutrient movement, carbon storage, and microbial activity.
“Instead of relying only on lab samples, we can now observe soil structure directly in the field and at much finer resolution,” said Arnold Schumann. The SWES professor is Mungofa’s dissertation committee chair.
To process the massive digital image datasets, the researchers used a deep-learning neural network. It identified pore structures with more than 90% accuracy across several evaluation metrics, even under highly variable field conditions. The researchers also found that AI-based image segmentation significantly improved estimates of important soil properties.
“Knowing characteristics such as porosity, pore connectivity and fractal structure and how they influence parameters such as soil tortuosity are important,” Schumann explained. “These directly influence water retention, drainage, aeration and nutrient cycling.”
From Research Tool to Practical Farm Technology
The system processed images in less than 100 milliseconds per frame. That makes this technology increasingly practical for real-time field sensing, robotic soil assessment, and precision agriculture systems. The implications for sustainable agriculture could be substantial.
“More accurate subsurface information can help growers improve irrigation scheduling, reduce fertilizer losses and better manage root-zone conditions,” Mungofa said. “It can also strengthen soil health monitoring with better measures of compaction, degradation, salinity and structural change.”
Beyond agriculture, the AI soil imaging system could support wetland restoration, watershed management, carbon sequestration monitoring, soil health assessments, and climate resilience research.
“This is where soil science is heading,” Grunwald said. “We are moving toward scalable, real-time soil intelligence systems that can support sustainable land management at multiple scales.”
The Impact on Florida Agriculture
Much of Florida’s agriculture depends on sandy, highly permeable soils that are vulnerable to nutrient leaching and water loss. With a better understanding of subsurface pore networks, producers could improve management of the state’s major crops. At the same time, they would be protecting springs, rivers and groundwater from nitrate contamination.
Florida citrus production could benefit from improved monitoring of root-zone water dynamics and nutrient movement. That would help growers who continue adapting to challenges associated with the citrus greening disease, HLB.
“With Florida’s sandy soils and increasing pressure on water resources, technologies like this could play an important role in improving both agricultural efficiency and environmental protection,” Schumann said.
The researchers note that additional testing will still be needed across a broader range of soil types. This includes Florida’s sandy and organic soils, but the study offers an important look into the future of AI-driven agriculture.
The research article, “Integrating deep learning with digital soil core sensing for subsurface soil image segmentation,” was recently published in Frontiers in Soil Science. The co-authors from LandScan are Daniel Rooney, Stephen Farrington, Woody Wallace, and Nicolas Guries.
Read the full research article at Frontiers.
Feature image from Perseverança Mungofa, UF/IFAS Department of Soil, Water, and Ecosystem Sciences.
2
