OBSERVER: Sensing the currents – detecting marine plastic litter in the Mediterranean

Preserving our waters

Water is vital to life. With our marine resources increasingly at risk, the world’s attention is shifting to the protection of its assets from oceans, seas, rivers and lakes down to the smallest streams and ponds. In the way of this endeavour is the issue of pollution, and in particular, plastic pollution.

Due to its intrinsic properties – low cost and light weight – plastic is the favoured material for a number of everyday applications. However, there is no rose without thorns as it is these same  properties that threaten the wellbeing of water resources. Fragmenting quite easily, plastic breaks up into tiny pieces ranging from visible sizes to microplastics. When improperly disposed, these fragments can end up in the waters and are often carried away by river discharges through the sea and ocean currents. Due to the natural patterns in those currents, inanimate objects tend to aggregate in specific hotspots, forming ‘litter islands’ which present a danger to sea animals through ingestion or contact injuries.

Significant initiatives are thus being drawn to combat this problem on the global stage. The European Union (EU), for instance, is incorporating the phased decommissioning of plastic packaging into its flagship Green Deal package. At the global level, the United Nations, has proposed the draft resolution on marine litter and microplastics aiming to “prevent and significantly reduce marine pollution of all kinds, in particular from land-based activities, including marine debris and nutrient pollution” by 2025.

Tackling these issues requires appropriate tools. To that end, three primary areas of research are currently in the spotlight:

1. Detection of marine plastic litter islands;
2. Identification of flow and accumulation patterns, through the prediction of trajectories, and
3. Possible prevention, mitigation and/or removal techniques.

Copernicus unlocks important new avenues for marine plastic litter monitoring

On site imaging and observation methods – usually requiring field missions with crewed vessels and trained personnel – are neither cost-effective nor efficient for thorough monitoring. Moreover, sometimes they are not particularly successful. Following the detection of the location, size and makeup of litter aggregates, insights can then also be localised and augmented through unmanned aerial vehicles (UAVs). Though their use is somewhat limited by their range (typically on the order of a few kilometres), most drones remain useful for reconnaissance and special-purpose studies. Furthermore, they can be used in control experiments near shore for the creation of reference data.

As soon as location of pollution is detected, the Copernicus Marine Service’s ocean current models are used to provide input (initial and forcing conditions) for particle tracking models. Drift modelling or ocean particle tracking models, based on ocean currents and vertical mixing of water layers, have the potential to indicate the probable three-dimensional transport pathways (drifting and/or sinking) and the average time taken for plastics to travel from their source to their current location. This allows to monitor where plastics enter the ocean (identify the source of pollution), how they spread across the open ocean and estimate for how long they have been there.

These models can be used to track movement in the following ways: 

• “Forward”, which allows to follow the plastic particle from its source to its current location in the open ocean. The video below show the tracking of plastic particles on the ocean surface using the Copernicus Marine Service ocean current modelled products. Such tracking works for drifting plastics at surface but can also be performed to locate plastic particles at depths in the ocean. 
• “Backward or reverse”, which allows us to reconstruct the plastic pollution journey   from its current location back to its source in order to pinpoint the potential source the pollution may have originated from. 

Artificial Intelligence (AI) and algorithmics play a central role in the endeavour to combat marine litter pollution as they help sift through vast quantities of Copernicus data. From identification algorithms, to the extraction of currents and patterns in global marine data flows, the combined use of satellite imagery, in situ data, algorithmic processing and AI provides an invaluable tool. Never before have we been able to collect and, crucially, process quantities of information to formulate insights like we now can. With spectral unmixing (algorithm breaking down light into its components) and AI as the two primary alternatives for the identification of plastic litter, much work is going into developing different solutions worldwide.

Current pilot initiatives in the Mediterranean

In an effort to bridge this gap, the Department of Marine Sciences at the University of the Aegean, with the help of its undergraduate students, launched a series of exploratory research studies near the Aegean Island of Lesbos, named the “Plastic Litter Projects”. The first in the series, PLP2018, used an unmixing technique to process open-access Copernicus imaging data – specifically Level 1C and 2 data from Sentinel 2 – in conjunction with in-situ UAV imagery. The UAV images were used as a validation tool to verify plastic pixel coverage (what percentage of each pixel of data is indeed plastic litter). Leveraging a set of three 100 sqm plastic rafts (artificial stand-ins for plastic waste islands) composed of bags, nets, and over 3,600 bottles, PLP used the RGB (Red-Green-Blue) and NIR (Near-Infrared) imaging bands of Sentinel-2 to generate a reference identification case. The following studies, PLP2019 and 2020, worked to simulate more realistic sizes and deployment durations of plastic islands, setting out smaller and semi-permanent targets. PLP2021 evaluated the validity of the developed reference cases in the real world through two 600 sqm targets (one plastic, one wooden), taking an important step on the way to producing an identification mechanism ready for widespread use. PLP2022 sets out to use inflatable targets, paving the way for future experiments not just in the Mediterranean, but also in seas and oceans across the world.

A similar initiative is being undertaken by Planetek Italia in collaboration with the National Technical University of Athens and the University of the Aegean, using crowdsourcing, Sentinel-2 data and hyperspectral satellite data for marine plastic litter detection, quantification and tracking (REACT project). Hyperspectral imaging is an Earth Observation technique employing a range of electromagnetic frequencies to image an object or area, rather than assigning red, green and blue (primary colours) to each pixel of data. The analysis involved entails decomposing the incident light into several spectral bands. Each pixel thereby provides not only two-dimensional spatial imaging, but also the unique colour signature of the photographed object or area, revealing much more about what is being imaged. Such reference data would form the basis for identification methods and algorithms to be developed by researchers across different seas and oceans across different continents, based on the same fundamental principles. Planetek is employing a mixed approach to the problem of detecting marine plastic litter, combining spectral signature unmixing and AI. The project is designed to underpin the detection and monitoring of marine litter for the Regional Agency for the Environmental Prevention and Protection of Puglia (ARPA Puglia), which implements the EU Marine Strategy Framework Directive in this Italian autonomous region. Once again, cost efficiency, repeatability and flexibility are stressed as the core aims and advantages of satellite-based methods. The key actions involved include monitoring plastic litter over extensive areas regularly, understanding the spatial and temporal (migration and aggregation) dynamics of plastic litter, and supporting in situ activities through improved, actionable information. This particular project employs a combination of data from Sentinel-2 and from the Copernicus Marine Environment Monitoring Service (CMEMS).  The use of hyperspectral imagery is promising as a new application of the future expansion mission CHIME, an Hyperspectral imager to fly before the 2030’s to complement the Sentinels 1st generation of satellites.

The takeaways: envisioning global action

Despite the local scope of current projects, the monitoring tools currently available can be used over extensive areas and are scalable to larger areas. Establishing the foundations for the detection of marine plastic litter migration and aggregation patterns will enable the next generation of stakeholders and researchers to be better informed, and better-equipped for safeguarding our waters for the years to come.