Item

Abstract

Although the world’s land digital elevation/surface model (DEM/DSM) has been well established through spaceborne and airborne sensors, high-resolution shallow clear water bathymetry remains poorly charted. Knowing the bathymetry in coastal zones is critical, especially for ocean navigation, environmental protection, mineral resources mining, and coastal management. However, surveys of bathymetry data in a traditional way relies heavily on humanpower and cost of equipment/vessels. National Aeronautics and Space Administration’s (NASA’s) Ice, Cloud, and land Elevation Satellite-2 (ICESat-2), launched in 2018, provides multi-beam georeferenced photon data every 70 centimeter along its ground tracks with a 91-day repeat orbit. The georeferenced photons can be utilized to derive water depth on coastal islands even without any on-site data. Here, we examine a synthesis of ICESat-2 ATL03 photon data and Sentinel-2 optical imagery to derive a bathymetry model based on Beer-Lambert law. Once the model is trained, the prediction of accuracy using goodness-of-fit (GoF) helps to select the most appropriate images. We select multiple islands in the South China Sea as testing sites, where the availability of coastal bathymetry models is scarce because it is costly and difficult to survey. The final bathymetry model is derived with a composite of multiple images, and the bathymetry accuracy via cross validation meets the requirement of category C in Zones of Confidence (ZOC) of the Electronic Navigational Chart (ENC) in 0-15m.