A Cluster-Based Method for Improving Water Depth Estimation from Satellite Derived Bathymetry

In this presented paper, an effort to improve the accuracy of the bathymetric model derived from optical satellite imagery is discussed. According to the previous study, the accuracy approximately measures in 2 m for depth up to 10 m by means of the Global Bathymetric Inversion Methods. Therefore, the improvement is done based on the so-called Cluster-Based Method applied to the analytical and log-ratio methods. The Cluster-Based Method partitions the working area on the by classifying the errors between the known depths and the predicted ones. The data for the known and predicted depths are acquired from the in-situ survey and the Sentinel-2 imagery. Improvement by one-third to the accuracy of the resulting bathymetry is obtained.