Quantification of uncertainties in remotely derived optical properties of coastal and oceanic waters

Abstract

Optical properties of oceanic and coastal waters are not only important for describing subsurface light field, but also useful indexes of environmental status. To meet the demand of various users, optical data products of global waters are now generated from ocean color satellite sensors (e.g. SeaWiFS, MODIS, MERIS). These products, due to imperfect sensor technology and retrieval algorithms, inherently contain some degrees of uncertainties. Traditionally, an averaged difference (or so-called error) for a dataset is usually provided via comparing retrieved values with in situ measurements. This averaged "error" is good at providing an overall picture between the retrieved and measured properties, but cannot indicate uncertainties for a specific product or a pixel, because that uncertainties in these products are not spatially uniform. Here, using optical properties derived from the Quasi-Analytical Algorithm as an example, we present an approach to quantify pixel-wise uncertainties of remote-sensing derived properties. Further, we quantitatively evaluated the uncertainties of the derived inherent optical properties (IOPs) and water-clarity products with a simulated dataset, and found that the relative uncertainty is generally within 10% for total absorption coefficients of oceanic waters. This presentation shows the theoretical basis to evaluate and understand the impacts of the various components on the analytically derived optical properties, and that a practical means to quantify the uncertainties of inverted properties for each reflectance spectrum is now available. This effort lays the groundwork for generating quality maps of optical properties derived from satellite ocean color images.