Civil and environmental engineering assistant professor to use AI and satellite data to predict ground movement
Published: Oct 14, 2025 11:00 AM
By Dustin Duncan
Unseen shifts beneath Minnesota’s highways are quietly undermining the state’s road network.
Ali Khosravi, assistant professor of civil and environmental engineering, is leading a multi-phase research effort with the Minnesota Department of Transportation (MnDOT) to detect signs of ground movement before they cause visible damage.
Landslides, sinkholes and other ground shifts often go unnoticed until they crack pavement or close highways.
“These events usually aren’t noticed until cracks appear or lanes start closing,” Khosravi said. “Right now, monitoring is mostly reactive. Someone sees a problem, then crews investigate and install sensors. That approach doesn’t scale well and often misses early warning signs.”
Instead of reacting after damage appears, the new system uses satellite radar (InSAR), optical imagery and deep learning to scan for early warning signs of ground movement.
“InSAR has great potential for statewide monitoring because it offers broad coverage and access to historical data,” Khosravi said. “But it’s not perfect. Sentinel-1, which we’re starting with, struggles in areas with heavy vegetation or localized movement. That’s why this is a feasibility study — we’re testing how well it performs under Minnesota conditions.”
Khosravi’s team includes Jack Montgomery, associate professor of civil and environmental engineering at Auburn, and Anand Puppala, professor of civil and environmental engineering at Texas A&M. The group is using machine learning to refine radar results and confirm ground movement through field sensors and on-site data.
The project will roll out in three phases. Phase one focuses on analyzing freely available Sentinel-1 radar data using open-source tools.
In Phase two, the team will incorporate data from the upcoming InSAR mission, which offers improved accuracy in forested areas, and apply artificial intelligence techniques, including transfer learning, to mitigate radar noise from weather, vegetation and other sources. Satellite imagery will help confirm whether the movement is real.
Phase three involves testing the system at four to six MnDOT-selected sites, comparing satellite data with on-the-ground measurements like soil moisture and sensor readings. Some locations will include low-cost radar reflectors to improve precision. A web-based dashboard will deliver real-time alerts and visualizations to MnDOT staff.
Once deployed, the system will monitor ground deformation over time. If movement exceeds a site-specific threshold — often just a few millimeters per month — it flags the area for follow-up, based on trends, local conditions and ground-truth data.
To reduce false alarms from snow, moisture or vegetation, the system combines radar data with satellite imagery, weather data, vegetation maps and readings from ground-based sensors.
“False positives are a risk, especially in forested or snow-covered areas,” Montgomery said. “That’s why we’re cross-checking multiple data sources and validating results with field measurements.”
Alerts will follow a tiered structure, with high-confidence warnings triggering immediate inspections and lower-confidence ones logged for continued monitoring.
“We want this system to help MnDOT engineers prioritize what matters,” Montgomery continued.
By the end of the project, MnDOT will have a working prototype and a real-time monitoring portal. The team is also developing training materials and a roadmap for scaling the system statewide.
Media Contact: , dzd0065@auburn.edu, 334-844-2326
Ali Khosravi, assistant professor of civil and environmental engineering, is leading a research project with the Minnesota Department of Transportation to develop an AI and satellite-based system for monitoring ground movement.
