Bathymetric Reconstruction Using Time-Series Satellite Images and Minimal In Situ Data
Abstract
This study proposes a method for reconstructing bathymetric terrain heights using time-series satellite images combined with a bathymetric interpolation technique. Conventional approaches of bathymetric surveys require LiDAR or echo sounders, which require significant cost and time. To overcome this limitation, we propose an algorithm to reconstruct an internal digital elevation model (DEM) of terrains under water bodies. We extract water boundaries from time-series satellite images. The extracted boundaries are treated as contour lines of bathymetric terrain, whose elevation values are not known. The contour lines are used as constraints during a hydrological bathymetric interpolation process. By assigning a true elevation to one water boundary and utilizing elevation values of nearby terrain, we reconstruct a bathymetric DEM. We can estimate a new water level from a new satellite image by overlaying a new water boundary on the bathymetric DEM reconstructed. We can validate the bathymetric DEM by comparing the estimated water level with the reference in situ water level. The proposed method was tested on three reservoirs using high-resolution KOMPSAT-3A images. With one reference water boundary and five nonreference water boundaries, the proposed method achieved an overall mean absolute error (MAE) of 0.61 m using six validation images for each reservoir, with MAEs of 0.61, 0.75, and 0.47 m for Nogok, Maegok, and Pungdo, respectively. These results show that time-series water boundaries can serve as effective constraints for bathymetric reconstruction and water-level estimation in small- and medium-sized reservoirs with limited in situ measurements.