Suitability of atmospheric satellite sensors for ocean color applications

Abstract

Global warming is already impacting the modern ocean. To recognize changes and predict the future ocean, ocean color (OC) remote sensing aims at quantifying and monitoring optically active oceanic constituents such as phytoplankton, as well as the underwater light field, on synoptic scales. The underwater light field is typically characterized by the diffuse attenuation coefficient (Kd), which describes how fast the downwelling light diminishes with depth. Phytoplankton diversity is typically assessed by grouping phytoplankton with respect to their function in biogeochemical cycles, which are described as phytoplankton functional types (PFT). Typically, OC sensors record radiometric measurements at several bands, 10-20 nm wide, in the visible and near-infrared. However, direct information of OC variables in the ultraviolet (UV) spectral range is still widely lacking. Also, the coarse spectral resolution of traditional OC sensors hampers the retrieval of information on biodiversity, e.g., represented through PFT, and ecosystems. While OC sensors have weaknesses in spectral resolution, a suite of atmospheric satellite sensors measures radiation at high spectral resolution (about 0.5 nm) and in spectral ranges (UV) not provided by typical OC sensors. These sensors are designed for the retrieval of atmospheric trace gases using Differential Optical Absorption Spectroscopy (DOAS) and have been utilized for OC retrievals by an adaptation of this method to the ocean domain. PFT, light attenuation and availability, and chlorophyll-a fluorescence have been successfully derived from radiances recorded by the atmospheric sensor SCIAMACHY between 2002 and 2012. So far, only little experience has been gathered with other SCIAMACHY-like sensors, limited to fluorescence retrievals at red wavelengths. This thesis focuses on retrievals of OC variables in the UV to green spectral range from multiple atmospheric sensors, namely SCIAMACHY, GOME-2, OMI, and TROPOMI. By comparison of OC retrievals from multiple sensors, insight can be gained on retrieval robustness and sources of uncertainty. Merging of data sets from multiple sensors is desirable for creating long time series necessary for observing climate change impacts. Three hypotheses are investigated: (1) OC variables in the blue-green spectral range, such as PFT and the diffuse attenuation coefficient, can be successfully retrieved from multiple SCIAMACHY-like sensors; (2) OC variables from multiple SCIAMACHY-like sensors can be merged to create long time series; and (3) UV-visible bands of SCIAMACHY-like sensors can be exploited for the retrieval of novel OC products in the UV. The thesis comprises three studies. Two studies concentrate on the diffuse attenuation coefficient with the second study focusing on novel Kd products in the UV. The third study utilizes knowledge from the first two studies to retrieve two key PFT, diatoms and coccolithophores, in the Southern Ocean. Based on experience gathered with the SCIAMACHY sensor, the retrieval of OC variables was adapted to the other atmospheric sensors and further improved to provide complementary and novel OC information. The mean Kd in the blue spectral range, 390 to around 425 nm, was successfully derived from measurements taken by GOME-2, OMI, TROPOMI (as well as SCIAMACHY). Coccolithophores were successfully retrieved from GOME-2 data in the Southern Ocean, however, diatom retrievals were found to be more challenging and resulting data quality was deemed insufficient. Intersensor biases were found between the derived Kd data sets as well as for the derived PFT data in comparison to the established SCIAMACHY PFT time series prior to this work. Complex instrument-specific corrections would be necessary to remove these biases. Biases were successfully reduced, but unfortunately merging of the multiple data sets is still not possible at this stage. The UV-blue spectral range was exploited for deriving two novel Kd products in the UV, 312.5 to 338.5 nm and 356.5 to 390 nm. Comparison with field measurements and sensitivity analysis of the retrievals showed promising results. Findings from the third study on Southern Ocean PFT support the Great Calcite Belt hypothesis, which suggests that coccolithophores cause a wide band of enhanced water reflectivity around Antarctica, and indicate that total chlorophyll-a concentration is not an adequate variable for predicting coccolithophore abundance. Spectral-based PFT retrievals, as used in this thesis, are found to be more suitable. Future applications of the data sets derived and characterized in this thesis include: the estimation of the oceanic heat budget and the UV dose rate on marine organisms, the investigation of the sources of colored dissolved organic matter and of phytoplankton photoprotective pigments, as well as residence times of particulate inorganic carbon in the upper ocean.