March 2025 Webinar: Levee Health Monitoring: A Data-Driven Framework for Flood Resiliency

Levees are complex earthen structures used for flood control management. Over 24,000 miles of levees in the United States protect over 23 million people, therefore their performance is critical. The American Society of Civil Engineers has rated levees based on their current performance, consistently assigning them a grade of “D,” indicating significant deterioration and a strong risk for failure.  

One major contributing factor to the poor performance of levees is the current inspection practices in the United States, which primarily rely on visual assessments of the surface with sparse, discrete instrumentation providing a data point in the subsurface at the point of installation. Another challenge in inspection is the lack of standardization among different organizations, leading to poor system-wide assessment. Furthermore, due to the spatial extent of levees, the temporal frequency of inspection is not sufficient to ensure the health of these systems.  

With these challenges in mind, a monitoring framework for levee health maintenance that leverages a combination of non-invasive and continuous UAV-enabled sensors and geophysical techniques to incorporate surface and subsurface health information has been developed. The work leveraged data collected along three stretches of the Grand Island levee system in the Sacramento-San Joaquin Delta during the 2022 dry season and the 2023 wet season. UAV-enabled optical imagery, thermal imagery, and LIDAR were collected at all three sites, along with two geophysical techniques: the Multichannel Analysis of Surface Waves (MASW) to capture the shear wave velocity distribution of the subsurface and Electromagnetic Induction (EMI) to capture the electrical resistivity distribution of the subsurface. 
 
This framework utilizes remote sensing change detection techniques on the surface data to understand how the health of the system changes over time using image differencing and classification methods. Change detection can then be applied to the subsurface data, along with unsupervised clustering and anomaly detection algorithms, to identify potentially problematic zones in the subsurface that could be attributed to internal failure mechanisms. The framework fuses the surface and subsurface health information into hazard maps, revealing increased areas of hazard on the levee. This information can potentially identify locations where more detailed investigation is warranted and therefore enables better and more informed decision-making in future rehabilitation and new developments.