Files
Culot_38531700_2023.pdf
Closed access - Adobe PDF
- 81.54 MB
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- Environmental legislation aimed at addressing marine pollution does not always provide sufficient measures to prevent oil discharges into the marine environment. To mitigate the consequences of oil pollution, it is crucial to promptly detect oil spills and forecast their potential impact on different areas. Synthetic-aperture radar (SAR) imagery has proven effective in monitoring oil spills over large regions. Once an oil spill is identified, it can be used as input for an oil spill dispersion model, which enables the simulation of its future drift (trajectory prediction) and past drift (origin determination). In order to develop such an integrated system, this master's thesis focused on the performance evaluation of oil spill detection using Sentinel-1 images. As a result, the classifier demonstrates performance close to that of an ideal classifier within the training dataset. However, when applied to new data outside the training dataset, the results are more varied and depend on the specific images. Nevertheless, these results suggest the potential for spatial and temporal portability, although the classification appears to be somewhat dependent on the image acquisition conditions. This methodology was subsequently employed across the entire Persian Gulf region for the year 2017, revealing two phenomena that could potentially compromise the accuracy of oil spill detection. Firstly, the incidence angle, which affects backscattering, generates differences in detection between the start and end of the swaths. Secondly, extensive areas of the Gulf are affected by radio-frequency interferences, producing artifacts that may be erroneously identified as oil spills. This study represents a first step towards the development of an integrated detection and modelling system. Although this method is constrained by the availability of data representing oil spills, it presents an initial methodology that, while still requiring further improvement, has the potential to yield satisfactory results in the future.