Challenge background
The Mediterranean is the sixth greatest accumulation zone for marine litter, holding only 1% of the world's waters but concentrating 7% of all global microplastics. Plastics account for over 85% of marine litter in the sea, with most of it accumulating on the seafloor, particularly in deep-sea sediments. Plastic pollution has long-term negative impacts on terrestrial ecosystems, biodiversity, and human health. It affects marine species through ingestion, suffocation, and entanglement, and poses risks to human health through the presence of microplastics in seafood.
The problem
Plastic pollution monitoring is crucial for effective mitigation, but current strategies face significant challenges. Traditional methods like field surveys and manual sampling are labor-intensive and time-consuming, covering only small, localized areas. This results in inconsistent and infrequent data collection, making it difficult to track trends and respond quickly to changes in pollution levels. Additionally, many remote areas are hard to access, leading to underrepresented data for critical habitats. The high costs associated with these traditional methods further limit the frequency and extent of monitoring efforts. These challenges highlight the need for more efficient and comprehensive monitoring approaches, especially for superficial plastic debris.
Goal of the project
To overcome the challenges of traditional plastic pollution monitoring, we propose using Sentinel-2 satellite imagery. This approach provides extensive coverage, frequent updates, and high-resolution multispectral images to monitor large areas and identify pollution hotspots.
Sentinel-2 satellites can cover vast regions and detect large plastic debris patches on the surface with up to 10-meter resolution. With a revisit time of approximately 5 days, they offer continuous monitoring and timely detection of pollution changes. The accessibility of free Sentinel-2 data also promotes widespread use and collaboration.
By leveraging Sentinel-2 imagery, we aim to create a reliable tool for tracking superficial plastic debris. This will help assess pollution patterns and inform mitigation strategies, improving plastic pollution monitoring in the Italian and Mediterranean seas and preserving marine biodiversity.
Project timeline
- 1
Week 1
Domain learning:
- Study satellite imagery basics and preprocessing techniques.
- Conduct a literature review on plastic debris detection methods.
- 2
Week 2
Data Collection and Preprocessing: Leveraging knowledge from the previous chapter project on Seagrass Mapping, download satellite images for areas of interest in the Italian and Mediterranean seas.
- 3
Week 3
Exploratory Data Analysis and Preprocessing:
- Perform statistical analysis of the collected dataset.
- Explore and apply preprocessing techniques.
- 4
Week 4
Modelling: Experiment with different deep learning architectures for plastic debris detection.
- 5
Week 5
Modelling: Train the selected model(s) and evaluate their performance.
- 6
Week 6
Web deployment and report:
- Deploy the trained model via a web application.
- Prepare and finalize a report summarizing the project, findings, and recommendations.
What you'll learn
- Collection and preprocessing of satellite imagery.
- Geospatial analysis: detecting and mapping environmental features.
- Building, training, and validating deep learning models for vision.
- Model deployment through web application.
- Scientific reporting: analyzing results and communicating findings effectively.