Marine environmental data management and post-processing via ICT tools

The Mediterranean Sea is increasingly impacted by shipborne pollution, making the development of reliable operational oil spill forecasting systems a priority for environmental protection and maritime safety. The aim of this thesis is to study the behaviour of oil spills and to create a user-friendly WebGIS platform that can simulate the dispersion of pollutants in the Mediterranean Sea. The study presents a detailed Lagrangian modelling framework to investigate oil spill dynamics using multiple oceanographic forcing datasets. First, oil spill simulations are performed using only sea surface currents derived from three operational ocean models: Copernicus Marine Environment Monitoring Service (CMEMS), the Naval Hydrographic and Oceanographic Service (SHOM), and the French Research Institute for the Exploitation of the Sea (IFREMER). Comparing the models reveals that differences in ocean circulation fields are the main source of variability in the predicted trajectories. Subsequently, the simulations are extended to include additional physical processes affecting surface transport, such as Stokes drift and wind drag. These forcings are obtained from Copernicus ERA5 reanalysis and the MeteOcean-UniGe atmospheric and wave modelling system. The contribution of each forcing component is then assessed quantitatively. Additionally, the influence of both absolute and relative wind formulations on dispersion patterns is evaluated. For each simulation, the centroid of the oil slick is computed to track its trajectory and to quantify its spatial evolution and deformation over time. A Bayesian optimisation approach is applied to identify the optimal coefficients for weighting the contributions of Stokes drift and wind drag, providing a calibration of the transport parameters based on the data. The modelling approach is further strengthened by incorporating oil weathering processes and transitioning from inertial passive particles to particles that are characterised by mass and chemical behaviour. This allows for a more realistic representation of the fate and transformation of oil in the sea. All simulations are conducted using the Parcels Lagrangian framework, which enables flexible implementation of custom kernels to account for various physical and chemical processes. The methodology is validated through a real-life case study, which involves comparing simulated trajectories with Sentinel-1 SAR satellite imagery acquired in the days following the accident, as well as documented reports. The results demonstrate that including Stokes drift and wind forcing significantly improves trajectory prediction compared to simulations driven only by surface currents. Additionally, substantial variability between ocean models is observed, emphasising the importance of model selection in operational forecasting and supporting the adoption of a multi-model probabilistic approach. Bayesian calibration reduces trajectory errors further and improves agreement with observations. The incorporation of oil weathering processes serves to enhance the simulations’ physical realism, thereby enabling the slick to evolve dynamically in accordance with the type of oil, resulting in modifications to the mass, viscosity, spreading behaviour, and the influence of droplet entrainment under wind and wave action. Finally, the validated Lagrangian modelling framework is implemented within a dedicated backend and frontend architecture, leading to the development of a user-friendly WebGIS forecasting tool capable of simulating the dispersion of particles involving combined surface currents, Stokes drift and wind forcing.