Re-parameterisation of a three-band semi-empirical algorithm for c-phycocyanin-pigment estimation in an optically complex tropical reservoir

• Ano de defesa: 2016
• Autor(a) principal: Diogo de Jesus Amore
• Orientador(a): José Luiz Stech
• Banca de defesa: Milton Kampel, Cláudio Clemente Faria Barbosa, Enner Herenio de Alcântara
• Tipo de documento: Dissertação
• 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-m21b/2016/02.05.11.51

The eutrophication of aquatic systems is a worldwide environmental problem. A major aftermath is health-inflicting toxic algal bloom, which can affect humans. Therefore, aquatic systems, mostly near urban environments require environmental monitoring. The use of remote sensing for monitoring algal blooms via bio-optical modelling is based on the spectral behaviour of the optically active components (OACs) in the water to estimate their concentrations. The detection of cyanobacteria, one of the main phyla of harmful algae, takes place via the identification of a unique pigment in inland waters cyanobacteria, the phycocyanin (PC). Remote sensing techniques, such as semi-empirical algorithms - a sort of bio-optical model - have been used to estimate PC concentration in aquatic systems using in situ hyperspectral data and satellite multispectral data. However, there is a lack in scientific works tackling PC prediction in tropical inland waters bearing PC in low concentration such as in city-supplying Guarapiranga reservoir at the southwestern region of São Paulo city. This is mostly true because scientific studies attempt to generate models based on bloom events. However, much uncertainty is associated with models results at the low concentration ranges. Therefore the goal of this research was to evaluate the re-parameterization of a semi-empirical algorithm for a tropical oligo-mesotrophic inland water. Radiometric, fluorometric, limnological, and multi-parameter sonde data were collected in Guarapiranga Reservoir, located. This thesis presents the findings which led to the algorithm re-parameterisation. Results showed that the calibration dataset (n=15) improved PC prediction R$^{2}$ by 15.3\% after the re-parameterisation; and for the validation dataset (n=19), PC prediction R$^{2}$ was improved by 4.79\%. NRMSE for the calibration dataset was bettered by 1.76\%; and it was almost equalised for the validation dataset (differed by 0.19\%). The new re-parameterisation correlation coefficient developed in this study presented a better R$^{2}$ (68\%) than that of the original algorithm (46\%). These correlations linked the band ratios used as enhancing coefficients to known PC spectral features. The bio-optical, radiometric, and water quality characterisation of Guarapiranga reservoir, and the evaluation of signal processing techniques of radiometric data yielded results that supported the generation of the new re-parameterisation coefficient. Such results were related to features in the blue-to-green spectral region capable of improving PC prediction. Uncertainties in the estimations are mainly due to the lack of in situ data. The re-parameterization was also considered for a synthetic dataset of the Ocean \& Land Colour Imager (OLCI) sensor/Sentinel 3. The simulation of OLCI data was conducted using its spectral response function, and it was important because of its potential use in environmental monitoring. Overall results were encouraging, however, further studies are suggested to further validate this new algorithm. Nevertheless, the development of a semi-empirical algorithm for low-concentration PC prediction in tropical inland waters is an important step for the development of an ever-improving robust tool for water quality monitoring.