Improving remote sensing retrieval of global ocean transparency with optical water classification

Abstract

• Remote sensing empirical inversion expressions of ocean water transparency (Z sd ) for different optical water types and water trophic states. • A class-based hybrid Z sd empirical model for global oceans. • Long-term spatiotemporal climatology patterns of global ocean Z sd . Ocean water transparency is an important indicator to measure the productivity of marine ecosystems, and it has great impact on phytoplankton biomass and community structure. Secchi disk depth (Z sd ) is a traditional measure of water transparency. With the development of satellite remote sensing technology, inversion through empirical model or semi-analytical model has become an effective and efficient means to obtain ocean Z sd . In recent years, ocean parameter inversions based on water classification become an ongoing trend. Jia et al. (2021) constructed a fuzzy-logic optical water type scheme (i.e., U-OWT) for the global oceans and multi satellite sensors, which was a potential water classification template for remote sensing inversion of ocean Z sd . Using U-OWT as the water classification basis, and by collecting 613 global ocean distributed SeaWiFS (Sea-viewing Wide Field-of-view Sensor) R rs and Z sd (with the range of 0.3–44.3 m) in-situ matchups, this study explored the optimal Z sd empirical models for each optical water type and each water trophic state respectively. As a result, a class-based Z sd hybrid empirical model (abbreviated as the Z sd-OWT model) was developed, which had relative high inversion accuracy over the entire Z sd range (MAPE = 8.25 %, RMSE = 0.65 m, bias = – 0.03 m). Specifically, the Z sd-OWT model conducted the global empirical inversion for the oligotrophic water classes and the class-based hard fusion empirical inversions for the mesotrophic and the eutrophic water classes. Although the Z sd-OWT model was designed for SeaWiFS sensor, it still has the potential to be migrated to other multispectral sensors with similar band settings.