Uncertainties in ocean color remote sensing – How metrology can help
by Dr. FREDERIC MELIN

Abstract

Ocean color (OC) provides access to marine biological quantities, foremost the concentration of chlorophyll-a (Chl-a), a major and universal phytoplankton pigment. Other in-water quantities can be determined by OC (optical) remote sensing, e.g., in relation to sediment, dissolved organic matter or indicators of water quality. Ultimately all OC in-water products are derived from the spectrum of reflectance characterizing the water body, expressed as remote sensing reflectance RRS or water-leaving radiance Lw. This spectral quantity and Chl-a are recognized as Essential Climate Variables by the Global Climate Observing System. As any geophysical product, RRS data and derived products need to be accompanied by uncertainty estimates to allow an informed use of these data by the user community and a proper integration in climate science. Even though there is growing emphasis on the provision of uncertainties for satellite data, this practice is in its infancy in the field of OC, which can be partly explained by the fact that this is particularly challenging. OC remote sensing is affected by a large and complex ensemble of error sources, including errors associated with the top-of-atmosphere signal and the numerous assumptions and approximations needed to solve the remote sensing problem and find a solution for RRS, all this coupled with the complexity of optical properties in natural waters. Relevant issues are summarized, including the sources of uncertainties and the state of the art on the various approaches currently proposed to produce uncertainty estimates for OC data. The use of uncertainty tree diagrams and of ensemble data is illustrated. The propagation of uncertainty fields from pixel level to gridded and time-composited data is particularly relevant as these products are usually those of most interest to the user community. Uncertainty estimates for OC data traditionally have been estimated by comparison with field data, a process termed validation; its central role and limitations are also discussed. Finally, recommendations are introduced, ultimately aiming at promoting a metrologically sound processing and treatment of OC remote sensing products.