Ocean Colour Variability across the Southern Atlantic and Southeast Pacific

• Ano de defesa: 2013
• Autor(a) principal: Natália de Moraes Rudorff
• Orientador(a): Milton Kampel, Robert Frouin
• Banca de defesa: Evlyn Márcia Leão de Moraes Novo, Aurea Maria Ciotti, Vivian Lutz
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Instituto Nacional de Pesquisas Espaciais (INPE)
• Programa de Pós-Graduação: Programa de Pós-Graduação do INPE em Sensoriamento Remoto
• País: BR
• Link de acesso: http://urlib.net/sid.inpe.br/mtc-m19/2013/09.17.14.26

Ocean colour radiometry (OCR) provides essential information for studies of primary productivity, heat fluxes, and biogeochemical cycles in the upper ocean. Generalized OCR models relating satellite radiometric data to biogeochemical variables are developed using global in situ data sets. However, when applied to specific regions these models commonly give results with significant deviation from in situ measurements, mainly due to field and satellite measurement uncertainties and model underrepresentation of ocean colour variability. Hence to improve OCR products further understanding of the sources of measurement uncertainty and bio-optical variability across different oceanic regions is needed. This work was focused on the Southern Atlantic and Southeast Pacific Oceans encompassing important biogeochemical provinces with highly distinct optical waters. In situ data was collected during a summer campaign on board the research vessel Melville (MV1102 cruise). The first part of the investigation was an uncertainty analysis of the radiometric and bio-optical data with three main objectives: (i) test different radiometric techniques with above and in-water approaches (ii) apply closure analyses with forward modeling of remote sensing reflectance (Rrs); and (iii) analyze the impacts of the uncertainties on operational OCR models. The uncertainty analysis revealed moderate to high levels associated with the various techniques, with 12 to 26\% relative differences (RD) for the ocean-colour bands (412- 555 nm) and 3-12\% for the reflectance ratios (412-510/555). The use of a merged Rrs (Instruments, INS) reduced uncertainties since each individual technique was subject to different instrumental and environmental biases. Complete closure was not obtained, especially for the stations with more adverse environmental conditions (with winds, waves and clouds), with 18-34\% RD compared to modeled Rrs bands. Nonetheless, the impact of INS uncertainties on retrieved OCR products for empirical and semi-analytical (SA) models was still generally smaller than the intrinsic errors of the inversion schemes. Hence, the approaches applied to obtain more accurate measurements were effective in reducing the main sources of uncertainties. Significant sources of deviations of the OCR models were related to the optical variability of the study region and intrinsic model errors. The second part of the investigation analyzed the sources of bio-optical variability and their relations to biogeochemical variables across distinct provinces. The bulk inherent optical properties (IOPs) were in first order associated with the chlorophyll a concentration (Chla) gradient. Second order variations were explained by specific IOPs linked to the phytoplankton community structure, composition and size distribution of the particle assemblage and variability of the coloured dissolved and particulate organic matter (CDM). To synthetize the first and second order optical variations across the study region, a Regional Specific Optical Water Type (R-SOWT) classification was proposed by defining 5 classes that integrate the specific IOPs and bio-optical indices, i.e., a phytoplankton Size Index, CDM index, the specific backscattering coefficient (bbp/Chla) and spectral slope ($\eta$). The R-SOWT significantly improved the performance of SA models by using class-specific parameterizations, especially for the bbp retrieved by the GSM (Garver-Siegel-Maritorena) model, reducing from 35 to 9\% RP, and the CDM absorption coefficient of the QAA (Quasi-Analytical Algorithm) model, reducing from 30 to 23\% RD. For more optically complex waters the improvements of the retrievals were much more significant. Further analysis of spatiotemporal variations of the optical relations and applicability of the R-SOWT for different seasons (and regions) are recommended for future studies. This approach has potential to improve OCR satellite products and be used as a new product that integrates relevant information for biogeochemical studies.