Holistic optical water type classification for ocean, coastal, and inland waters

Abstract

Water constituents exhibit diverse optical properties across ocean, coastal, and inland waters, which alter their remote-sensing reflectance obtained via satellites. Optical water type (OWT) classifications utilized in satellite data processing aim to mitigate optical complexity by identifying fitting ocean color algorithms tailored to each water type. This facilitates comprehension of biogeochemical cycles ranging from local to global scales. Previous OWT frameworks have focused narrowly on either oceanic or inland waters and have relied too heavily on specific data collections. We propose a novel holistic OWT framework applicable to all natural waters, based on state-of-the-art bio-geo-optical modeling and radiative transfer simulations that encompass different phytoplankton groups. This framework employs a “knowledge-driven” paradigm, combining domain knowledge and insights from previous studies to simulate the reflectance spectrum from water constituent concentrations and inherent optical properties. Our method extracts optical variables to represent the full spectrum of reflectance, consolidating both spectral shape and magnitude. We apply the framework utilizing diverse in situ, synthetic, and satellite data (Sentinel-3 OLCI) and demonstrate its better classifiability than other frameworks. This framework lays the foundation for comprehensive global monitoring of natural waters.