Optical remote sensing of sea water quality through a multi-sensor data-driven approach

This thesis explores advanced methodologies for coastal water quality assessment using remote sensing techniques, with a particular focus on chlorophyll-a (Chl-a) concentrations and the detection of Escherichia coli (E. coli) pollution. With respect to Chl-a detection, this study employs Sentinel-2 multispectral data, in-situ observations, and neural networks to enhance the accuracy of Chl-a estimation in coastal regions. The results reinforced the significance of satellite data for large-scale environmental monitoring, despite challenges in data validation. In addition to that, indeed, a comparison between available in-situ datasets (ISPRA and ARPA) has been realised. In parallel, the research pioneers the development of a novel algorithm to detect E. coli pollution from satellite-derived parameters, an area largely unexplored in existing literature. By analysing bio-optical properties, sea surface temperature, and additional satellite-based indicators such as turbidity and suspended particulate matter, a neural network model was designed to classify coastal waters into categories of pollution, ranging from not polluted to highly polluted. Validation using in-situ data demonstrated promising results, achieving 95% accuracy in detecting highly polluted waters. This research highlights the potential of satellite remote sensing as a non-invasive, cost-effective tool for environmental monitoring, particularly for coastal waters. Future work should focus on expanding the in-situ dataset to further refine the model and strengthen its applicability across diverse geographical areas.