Proactive road maintenance starts far below the pavement’s surface. As transportation practitioners know, pavement isn’t static: it’s a living structure constantly reacting to temperature, moisture, and load.
For this reason, the Minnesota Department of Transportation (MnDOT) has invested in “thermocouple trees”—vertical arrays of sensors buried within the pavement layers. Although many agencies have this hardware, they often lack the translator to turn the vast amount of data they generate into actionable insights.
In a new MnDOT-sponsored study, a team of University of Minnesota researchers led by Ketson dos Santos, assistant professor of civil, environmental, and geo- engineering and a CTS scholar, set out to solve the persistent challenges of interpreting subsurface data. These challenges include missing information, faulty readings from the sensors, and the sheer volume of data created by frequent readings.
“The road is speaking to us through these thermocouple trees, but until now, we lacked a method for translating those signals into maintenance schedules,” Dos Santos explains.
The data for this study came from MnDOT’s MnROAD research facility, where vertical thermocouple trees have been recording long-term variations in subsurface pavement temperatures. This data is vital because the pavement layers act as a filter; if the filter’s behavior changes, it signals a physical change in the road, such as moisture, aging, or a loss of compaction.
To unlock these insights, researchers developed codes, using the Python computer language, to treat, process, and analyze the thermocouple data. This tool even uses compressed sampling to intelligently fill in the blanks when a sensor fails in the field.
“By modeling pavement layers as a cascade of filters, we can see how heat flow is sensitive to aging and moisture,” Dos Santos says. “This modular tool offers a scalable solution that doesn’t just record temperature—it monitors the structural integrity of the pavement layers.”
This study gives transportation agencies a valuable way to support the shift from reactive repairs to proactive data-driven management. For example, the tool can be used to detect “invisible” issues such as moisture infiltration before a pothole or crack ever appears on the surface.
“This research demonstrates that thermocouple-derived temperature data, when processed with advanced spectral tools, can reliably detect changes in pavement material properties,” Dos Santos adds. “It gives engineers a dependable indicator of degradation patterns, supporting long-term infrastructure management decisions before the damage becomes visible.”
Practitioners can put the study’s findings to work immediately using a “digital toolbox” created through this project. All the newly developed computational methods are available through a public GitHub repository. These “plug-and play” templates allow users to clean their own datasets, analyze thermal shifts, and perform probabilistic health checks on their road inventory. The project’s final report also describes ways in which the findings can support day-to-day pavement maintenance decision making.
—Megan Tsai, contributing writer
