Remote sensing of shallow-water bathymetry: leveraging multispectral ocean color observations

Water depth or bathymetry is a critical geophysical parameter for shallow-water navigation and recreation, and is indispensable for mapping, discriminating, and characterizing the diversity of benthic habitats (coral, algae, sand, etc). Earlier studies have demonstrated that accurate bathymetry can be retrieved from hyperspectral remote sensing reflectance spectra (R_rs(λ)). Until recently, however, the operational ocean color satellites have only generated multispectral R_rs(λ) products. For instance, the Visible Infrared Imaging Radiometer Suite (VIIRS) on the Suomi National Polar-orbiting Partnership (SNPP) measures R_rs(λ) at six visible bands, the Operational Land Imager (OLI) on the Landsat 8 operates at four visible bands, and the Ocean and Land Colour Instrument (OLCI) on the Sentinel-3A may produce data up to ~10 visible bands. The limited number of wavelengths has largely prevented these multispectral sensors for routinely generating reliable bathymetry products. In this study, we review the existing approaches for shallow-water bathymetric retrieval and then propose a simplistic optimization approach. Based on synthetic data analyses, we show that our new approach can substantially improve the bathymetric estimation over shallow waters covered with sand, coral, or algae and over a range of spatial scales. Further, we apply this approach to satellite multispectral images obtained from SNPP-VIIRS, Landsat 8-OLI, and Sentinel 3A-OLCI in Florida Keys, and validate the retrievals with in situ bathymetric measurements. We demonstrate that the shallow-water bathymetry can be produced as a reliable ocean color product from operational multispectral ocean color images. It is anticipated that accurate bathymetry will facilitate subsequent retrievals of water-column and benthic properties.