A Novel Multitemporal Approach for Satellite-Derived Bathymetry for Coastal Waters of Palau

Abstract: Shallow water bathymetry is important for understanding biogeophysical and socioeconomic processes in coastal areas. In recent years, satellite-derived bathymetry (SDB) methods have been increasingly used to provide high-resolution bathymetry estimation in different regions using single remote sensing images.

To tackle the common issues of single-image SDB, such as data gaps due to cloud coverage and false bathymetry due to water turbidity, this study applied a novel multitemporal workflow for SDB estimation in Palau, a Pacific Island nation. This workflow implements the typical empirical SDB steps to calculate relative water depth (i.e. log ratio of the blue and green band) of 20 Landsat 8 images that were composited and subsequently related to in situ depth measurements to estimate true depth. Before composition, a histogram equalization approach was employed to normalize the images and identify clear water areas of each image pair by applying a 1% difference threshold.

To achieve better performance, different methodological options at three key steps were evaluated, including temporal composition (mean vs. median), point data extraction (direct vs. bilinear interpolation), and regression (linear vs. piecewise vs. polynomial). Among 12 models, the polynomial model built upon bilinearly interpolated mean composition data performed the best, accurately estimating water depth up to the extinction depth of 13.7 m (45 ft), with a root mean square error of 1.76 m (5.77 ft). This multitemporal approach, with proper methodological choices according to local circumstances, could be applied to other regions to derive gap-free and accurate bathymetry estimations.