Wave kinematics-based bathymetry estimates from satellite optical video

Abstract

In this work, cBathyShortDwell, a version of the spectral depth inversion algorithm cBathy that has been adapted for shorter record lengths, was applied to four 29.9-60s satellite-based optical videos and the resulting bathymetry estimates were compared to bathymetry surveys collected within 13 days of each satellite collect. The Planet SkySat videos were collected at the U.S. Army Engineer Research and Development Center Field Research Facility between April and September 2023. The videos were collected at 30Hz with a ground sampling pixel resolution of 0.75-1 m and an image size of approximately 2.5km x 1km. RMSEs from post-processed cBathyShortDwell depth estimates over the 800 m x 1100 m analysis domain using the entire video duration at a 5 Hz frame rate ranged from 0.50 to 0.59 m with biases ranging from -0.19 to 0.16 m. Although some differences were seen in the RMSEs as a function of frame rate (2 to 30 Hz), larger differences were observed due to video length (17-60s), wave visibility in the imagery, and incident wave conditions. The highest overall RMSEs were observed in conditions with short video lengths (RMSE ranging from 0.76–1.32 m for all post-processed bathymetry estimates generated with 17-s videos) or conditions with low wave visibility (RMSE 0.80–0.99 m and $\left|\mathrm{bias}\right|$ 0.32-0.59 m for all post-processed bathymetry estimates using 30s videos with low wave visibility). These results are comparable to published RMSE values from established versions of the algorithm (Holman et al., 2013, Holman and Bergsma, 2021) and demonstrate the ability of cBathyShortDwell to provide good depth estimates from satellite videos in a range of environmental conditions.