Print this page

Digital Image Analysis of Video Footage of the May 19 2020 Edenville Dam Failure in Michigan

This page includes animations of the digital image analyses our research group conducted on the video footage collected on the Edenville dam failure (courtesy of Lynn Coleman) that occurred on May 19 2020. My research collaborator, John Manousakis of Elxis Group assisted me in the digital image analysis presented here.

 

This page is accompanying the report prepared by the ASCE Geo-Institute Technical Committee on Embankments and Dam and will be published by ASCE.

References to this work and report can be made as follows:

Pradel, D., Lobbestael, A., Athanasopoulos-Zekkos, A. Brooks, C., Champagne, C. Clark, M. Dobson, R., Edmonds, D., Esser, A., Gong, W., Hille, M., Manousakis, J., Marion, N., Martin, H., Midttun, N., Niemi, N., Oommen, T., Townsend, K., Yanites, B., Zekkos, D. (2020) May 19, 2020 Edenville and Sanford Dam Failures: Report on Field Reconnaissance Observations, ASCE Geo-Institute.

 

About the digital image analysis process

The conducted image analysis first involved image stabilization. This was necessary to minimize the influence of the movement of the camera on the video. Subsequently, pixel tracking was conducted on pixels that could be confidently tracked in the footage.  A total of five pixels, were selected along the main axis of the failure mass (i.e., not closer to or farther from the camera location). Five pixels were considered adequate to capture the key aspects of the deformation of the failed mass. Displacement vectors of the selected pixels are also shown in the footage.

Pixel P1 at the crest moves downwards first, followed shortly afterwards by P2 and P3 primarily in the horizontal direction. These three points are part of the initial failure, whereas P4 moves later as the failure surface expands. Pixel P5, the lowest pixel, selected at the toe of the dam, and on native ground, never moves, and becomes covered by the failing mass. These observations indicate that the failure was confined to the main body of the dam.  

Since the height of the dam at the location of the failure is known from LIDAR Digital Elevation Models, it is possible to calculate the displacements observed in the video in physical units, rather than just pixels. We determined that each pixel at the location of the failure was approximately equal to 4.5 cm. We used this information to calculate the displacement and velocity in the horizontal and vertical direction as well as the resultant displacement and velocities as a function of time, and also calculate the velocities in the horizontal and vertical axis.  The results of this analysis are shown in this video and allows for quantification of the displacements of the dam failure.  

The failing mass was found to gradually accelerate, reaching a velocity of up to 5 m/sec. Subsequently in the video, the mass was decelerating, but pixel tracking was no longer possible because of the disturbance in the failing mass.

More details and additional analysis are provided in the above referenced report.  

 

Video and Animations Results

The following four versions of the video processed video are provided here:

 

Video 1: Processed video illustrating the movement of the tracked pixels. 

 

Video 2: Zoomed-in version of the video illustrating the movement of the tracked pixels

 

Video 3: Zoomed-in and slowed-down version of the same video illustrating the movement of the tracked pixels

 

Video 4: Animation of the video along with charts illustrating calculated displacement and velocity time histories in the vertical and horizontal direction as well as the resultant vectors.