University of the Aegean
Remote detection of floating marine litter using satellite imagery
Abstract
dc:descriptionFloating marine litter, and plastic litter in particular, represents a persistent and spatially dynamic pressure on marine and coastal environments. While traditional monitoring approaches provide valuable local observations, their limited spatial and temporal coverage, their inherent costs and their dependence on both human and non-human resources, constrains the systematic assessment of floating debris distribution and variability. Satellite remote sensing offers a complementary perspective, enabling repeated, large-scale observations of the sea surface; however, the systematic detection and monitoring of floating marine litter in coastal environments remain challenging. Floating marine debris detection using satellite remote sensing is constrained by a combination of spectral ambiguity, sub-pixel mixing, and strong environmental variability. Previous studies have demonstrated the general feasibility of detecting floating plastics, however significant gaps remain in the experimental characterisation of plastic spectral behaviour under realistic conditions, the quantification of detection limits at the pixel level, and the translation of these insights into operational monitoring systems. This dissertation investigates the potential and limitations of optical satellite remote sensing for the detection and monitoring of floating marine litter. The dissertation follows a progression from experimental observation and characterisation to operational framework development. The dissertation is composed around three main objectives. The first objective establishes an experimentally based understanding of the spectral characteristics of floating plastics and their separability from other surface features. The second objective quantifies detection limits and sensor capabilities at the sub-pixel level, providing empirical constraints for practical application. The final objective integrates these findings into an operational detection and monitoring framework, supporting the transition of floating matter remote sensing from exploratory studies toward systematic applications. Objective 1 – Determine whether floating plastics exhibit a distinct spectral signature under near-realistic marine conditions and assess its suitability for reliable spectral classification. (Chapter 2)Chapter 2 aims To experimentally establish the spectral signature of floating plastics under realistic marine conditions and to assess their spectral separability from surrounding water and other common floating matter types. This objective was addressed through a controlled longterm field experiment involving the deployment of large-area artificial floating plastic targets, designed to fully cover a satellite pixel. Multisensor observations, including Sentinel-2 multispectral imagery, UAV-based RGB and hyperspectral data, and in situ measurements, were acquired over several months. The analysis focused on characterising the spectral response of floating plastics, quantifying their variability, and examining the influence of environmental factors such as biofouling and partial submersion. Spectral comparison and separability analyses were conducted against natural floating materials, including woody debris, marine mucilage, and pollen, in order to identify fundamental limitations and ambiguities in purely spectral discrimination approaches.Objective 2 – Determine the minimum detectable plastic abundance fraction in a mixed pixel for the most widely used multispectral sensors. (Chapter 3)Chapter 3 aims to quantify the minimum detectable fraction of floating plastics within mixed satellite pixels and to evaluate the suitability of commonly used multispectral satellite sensors for operational detection applications. This objective was addressed through experimental analysis of mixed-pixel scenarios, combining controlled target deployments with satellite observations at varying spatial resolutions and band configurations. Sub-pixel detection capabilities were assessed using partial unmixing methodologies, allowing estimation of plastic abundance fractions within individual pixels. Detection thresholds were systematically evaluated under different environmental conditions, including varying degrees of biofouling and submersion. The analysis provided empirical constraints on minimum detectable fractions and highlighted sensor-specific limitations, thereby informing the realistic operational use of multispectral satellite data for floating plastic monitoring.Objective 3 – Develop a scalable operational detection framework using artificial intelligence and spectral information. (Chapter 4)Chapter 4 aims to develop and demonstrate an operational framework for the detection and monitoring of floating matter and debris using satellite remote sensing. Building on the experimental findings of the previous objectives, this objective focused on integrating machine learning–based image segmentation and spectral analysis into a scalable processing framework. The framework was designed to operate on large satellite datasets, accommodate diverse floating matter types, and operate under varying environmental and observational conditions. Its performance was evaluated across multiple study areas and event types, focusing on extreme environmental events, demonstrating its applicability for systematic monitoring rather than case-specific analysis. Overall, this thesis provides experimentally based insight into the spectral behaviour and detectability of floating marine litter and floating matter in general, and advances the transition of satellite-based marine litter detection from exploratory analyses toward more operationally mature applications.
Degree
thesis:*- Grantor dc:publisher
- University of the Aegean
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Papageorgiou, Dimitrios
- Παπαγεωργίου, Δημήτριος
Subjects
dc:subject × 14- Δορυφορική τηλεπισκόπηση
- Θαλάσσια ρύπανση
- Πλαστικά θαλάσσια απορρίμματα
- Βαθιά μηχανική μάθηση
- Sattelite remote sensing
- Marine pollution
- Marine plastic litter
- Deep learning
- Γεωεπιστήμες και Επιστήμες Περιβάλλοντος
- Φυσικές Επιστήμες
- Ωκεανογραφία
- Earth and Related Environmental Sciences
- Natural Sciences
- Oceanography
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/62101