Advances in Remote Sensing of Plastic Waste

Advances in remote sensing
of plastic waste

Acknowledgements
GIZ thanks the IOCCG Task Force on Remote Sensing, Marine Litter and Debris;
Nina Gnann (BfG), Dimitris Papageorgiou (University of the Aegean), Prof. Dr. Konstantinos
Topouzelis (University of the Aegean), Bijeesh Kozhikkodan Veettil (Thu Dau Mot University,
Vietnam), Dr. Jonas Franke (Remote Sensing Solutions), Jennifer Mathis (University of Geor-
gia) and Dr. Stephan Ziegler (WWF Germany) for their expertise and contributions
to this publication.
As a federally owned enterprise, GIZ supports the German Government in achieving its
objectives in the field of international cooperation for sustainable development.
Published by:
Deutsche Gesellschaft für
Internationale Zusammenarbeit (GIZ) GmbH
Registered offices:
Bonn and Eschborn, Germany
Dag-Hammarskjöld-Weg 1
565760 Eschborn, Germany
T +49 6196 79-0
F +49 6196 79-11 15
I https://www.giz.de/en/worldwide/93799.html
Programme:
Global sector project to support the BMUV in implementing the Marine Debris Framework –
Regional hubs around the globe
Authors: (in alphabetic order of last names)

Steffen Blume (GIZ); Jonas Franke (Remote Sensing Solutions, Germany);
Shungu Garaba (University of Oldenburg); Phong Giang (GIZ);
Jennifer Mathis (University of Georgia, USA); Nadine Ortwig (GIZ);
Stephan Ziegler (WWF Germany)
Reviewers and Editors:

Phong Giang, Nadine Ortwig (GIZ)
Responsible:
Elisabeth Duerr
E [email protected]
Design/layout:
kippconcept GmbH, Bonn, Germany
On behalf of
Federal Ministry for the Environment, Nature Conservation,
Nuclear Safety and Consumer Protection (BMUV)
Photo credits:
lavizzara/shutterstock (Title), Naeblys/shutterstock (p. 7),
AdobeStock (p. 16), M. Jurzyk/AdobeStock (p. 37), ink drop/AdobeStock (p. 40),
3dsculptor/AdobeStock (p. 52), Parilov/AdobeStock (p. 88)
Disclaimer:
This study was commissioned by and completed exclusively for the Deutsche Gesellschaft
für Internationale Zusammenarbeit (GIZ) GmbH. The opinions expressed in this study are
those of the authors and reviewers and do not necessarily represent the opinions or
positions of the GIZ. GIZ does not, however, guarantee the accuracy or completeness of
information in this document, and cannot be held responsible for any errors, omissions or
losses which result from its use. Responsibility for the content of external websites linked
in this publication always lies with their respective publishers.
Bonn 2023
// 2

Table of Contents
Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Table of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Executive summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.1 Plastic leakage pathways and monitoring challenges . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Remote sensing in the context of environmental monitoring . . . . . . . . . . . . . . . . . . . . 11
1.4 History on Remote Sensing of Plastics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 Basics of Remote Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2 Maturity and Advances in Remote Sensing of Plastics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.1 Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3 Quantification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.4 Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.5 Online Platforms and Web-based Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.5.1 Ocean Scan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5.2 Litterbase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.5.2 Global Partnership on Marine Litter (GPML) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.5.3 Global Plastic Watch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.6 Stakeholder Community Initiatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1 Prototype Satellite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2 Baselines and Standardized Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3 Stakeholder Community Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Annex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
A1. Scientific Research Projects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
A2. Application of Remote Sensing in Monitoring Plastic Waste . . . . . . . . . . . . . . . . . . . . . 70
A2.1 Platforms for Remote Sensing of Plastic Waste . . . . . . . . . . . . . . . . . . . . . . . . . 74
A3. Machine Learning Application in Remote Sensing of Plastic Waste . . . . . . . . . . . . . . . 75
A4. Conferences, Proceedings and Workshops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
A4.1 Conference Proceedings 7th Marine Debris Conference, Busan, South Korea
(7IMDC, 18 – 23 September 2022) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
A4.2 ESA Living Planet Symposium, Bonn, Germany
(25 - 26 May 2022) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
// 3

Acronyms
API Application programming interface
ASEAN Association of Southeast Asian Nations
BMUV Federal Ministry for the Environment, Nature Conservation and Nuclear Safety
and Consumer Protection
CYGNSS Cyclone Global Navigation Satellite System
DOI Digital object identifier
EO earth observation
FML Floating marine litter
GEO Group on Earth Observation
GESAMP Group of Experts on the Scientific Aspects of Marine Environmental Protection
GIZ Deutsche Gesellschaft für Internationale Zusammenarbeit GmbH
GOCI Geostationary Ocean Color Imager
GOOS Global Ocean Observing System
GPML Global Partnership on Marine Litter
IOCCG International Ocean Colour Coordinating Group
IR Infrared
LIDAR Light Detection and Ranging
MWIR Midwave Infrared
NASA National Aeronautics and Space Administration
NIR Near Infrared
NKUA National and Kapodistrian University of Athens
NTUA National Technical University of Athens
PO.DAAC Physical Oceanography Distributed Active Archive Center
RGB Red-Green-Blue
SWIR Shortwave Infrared
TSM Total suspended matter
UAV Unmanned aerial vehicle
// 4

Table of Figures
Figure 1: Pathways and fluxes of plastics into the oceans . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Figure 2: Process of monitoring marine litter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Figure 3: Spatial distribution of the “MODIS burned areas” reference products
and Copernicus EMS rapid mapping activations . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Figure 4: Examples of plastic samples of varying sizes
that have been observed in remote sensing experiments . . . . . . . . . . . . . . . . . . . . 17
Figure 5: Overview of remote sensing platforms relevant for plastic waste . . . . . . . . . . . . . 18
Figure 6: A simplified pipeline of the approach from image capture
to deriving essential parameters from remote sensing imagery . . . . . . . . . . . . . . . 19
Figure 7: Linking Earth Observation data with multi-source in-situ data
for modelling debris pathways from source to sink . . . . . . . . . . . . . . . . . . . . . . . . 21
Figure 8: Camera-based macroplastic detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Figure 9: Orthomosaic (composed of hundreds of single pictures) of Barricata beach . . . . . 23
Figure 10: NASA’s CYGNSS mission maps the concentration of ocean microplastic
with data from eight microsatellites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Figure 11: A cropped image (RGB) from Planet 3-m data for Mumbai
showing the potential waste-positive localities after applying
heuristic feature space reduction and clustering . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Figure 12: Ocean Scan graphic interface showing the dataset catalogue
from various geographic locations and additional options related
to matching satellite to in-situ data, observation platform
for validation, time, type and size of litter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Figure 13: Examples of the features and products available in Litterbase graphic interface . 32
Figure 14: GPML Risk and Warning System for Macroplastic Litter user interface . . . . . . . . . 33
Figure 15: Example showing Global Plastic Watch web interface . . . . . . . . . . . . . . . . . . . . . . 34
Figure 16: Potential illegal dumpsite close to a river.
The site has changed significantly over time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
// 5

Executive summary
Executive summary
Remote sensing technology is widely used in environmental monitoring for various purposes, such as
assessing climate change impacts, monitoring forest fires, and tracking wildlife movement. In recent
years, remote sensing has shown great potential for monitoring (marine) plastic litter, with promising
results from the use of simple surveillance cameras, drones, and satellite imagery. The development of
methods to support traditional monitoring concepts using satellite imagery has been in particularly
driven by space agencies and national governments.
High-resolution satellites with multispectral wavebands and specific data models developed by multi-
disciplinary teams have shown promise in detecting suspected plastic accumulation zones. However,
standardized monitoring methods for plastics are not well established, and coordinated efforts for in-situ
sampling matching satellite observation for validation and verification tasks are necessary for algorithm
development. There are currently no market-ready applications available for (plastic) litter monitoring.
Previous research has mainly focused on developing algorithms to improve the detection of plastic
waste accumulation. The areas of quantifying the scale of pollution, identifying the type of plastic, and
monitoring the pathways from source to sea are even less developed, and further research is necessary to
develop a methodology that meets stakeholders’ needs.
Preliminary scientific evidence-based studies have reported recommendations on possibilities to scale up

from controlled experiments using artificial plastic targets and small area coverage studies. A prospec-
tive approach is the design and launching of a prototype satellite sensor aimed at bridging the gaps and
improving the monitoring of plastic waste. A team of experts conducted a community user needs survey
complemented by the current literature on remote sensing of plastic litter to define specifications, capa-
bilities, and characteristics for such a prototype satellite. A prototype satellite could also be an opportu-
nity for the community to define and set standardized methods for plastics assessments using well-de-
fined baselines. Remote sensing experts believe that the invaluable scientific evidence-based knowledge
derived from the proposed prototype satellite will allow for a more refined assessment of the capabilities
of space-based monitoring of plastics.
A coordinated interdisciplinary, holistic approach is critical for a harmonized monitoring tool for plastics
in the terrestrial and marine environment. One example is the International Ocean Colour Coordinat-
ing Group (IOCCG) Task Force on Remote Sensing of Marine Debris and Litter, composed of experts
from academia, industry, civil service, non-profit organizations, governments, and space agencies. The
Task Force coordinates and supports the advancement of the topic, especially through thematic work-
shops and networking events. One priority has been to engage stakeholders to better understand their
user needs while also communicating the capabilities and limitations of remote sensing as a complemen-
tary monitoring tool for plastics in the aquatic environment. At present, efforts are currently underway,
so this report provides a status on remote sensing of marine litter. The findings and scientific research
gaps identified in recently completed projects will be bridged or supplemented by ongoing or future
studies.
// 6

Introduction

Introduction
Today, one of the most pressing global threats The detailed objectives of this report are:
to the oceans is plastic pollution. Plastics and
microplastics have become ubiquitous in land- › To introduce remote sensing technologies and
scapes as well as in marine and freshwater their modes of operation for environmental
environments 1. They disperse in and enter the monitoring.
sea through rivers, sink to the river and sea floor
and accumulate in sediments and biota. While › To provide the reader with the context of
monitoring of marine litter and its impacts is still application regarding marine litter detection
limited, research has already identified marine and monitoring through remote sensing tech-
litter as a growing issue affecting various species of nologies.
fish, cetaceans, reptiles and invertebrates 3. Effects
at population and ecosystem level still need to be › To present concrete examples of remote
better understood but could be significant 2. A sensing projects in practice to combat marine
sound database derived from marine litter mon- plastic litter.
itoring can help to demonstrate the dimension
of the problem, to identify key sources of marine › To support policy makers in making informed
pollution, to initiate preventive measures and to decisions on marine litter monitoring strate-
show the effect of various planned or implemented gies.
measures 3. Currently, there are different tools to
assess and monitor the leakage of (plastic) waste To this end, the report is structured as follows:
into the environment, each of them with specific Chapter one briefly summarizes the status quo of
requirements of data collection and knowledge marine litter monitoring methods alongside their
on data interpretation. However, for all tools the main challenges and introduces the emergence
existing data collection systems are limited and and foundations of remote sensing as a comple-
thus unable to answer fundamental questions, e.g. mentary technology for marine plastic litter detec-
regarding the concentration of plastic waste in the tion and monitoring. The second chapter reviews
sea and its spatial and temporal dynamics. Efforts in depth the existing advances of remote sensing
to fill the current gaps of marine litter monitor- in detecting, identifying, quantifying and track-
ing have led to a growing interest in imaging or ing marine litter by illustrating existing use cases
remote sensing technologies using boats, aircrafts and presents web-based tools and stakeholder
or satellites to monitor physical characteristics of community initiatives in this context. Combining
the earths environmental components 13 as it has the insights from the two previous chapters, the
the potential to complement existing monitoring report concludes with the (current) limitations of
approaches with high quality data in a cost-effec- the applications and an outlook on the potential
tive way. applications and development of remote sensing
for marine litter monitoring. This study is part of
The purpose of this report is to provide an over- the GIZ Global Marine Litter Project, which sup-
view of the emergence and current state of the art ports the Federal Ministry for the Environment,
in remote sensing applications for plastic waste Nature Conservation and Nuclear Safety and
monitoring and as well as examples of current Consumer Protection (BMUV) in the implemen-
research projects. tation of the funding program “Marine Debris
Framework – regional hubs around the globe”
(Marine: DeFRAG). The aim of the funding
program is to support developing and emerging
countries in improving their waste management
system and to create incentives to prevent marine
litter pollution.
// 8

Introduction
1.1 Plastic leakage pathways

and monitoring challenges
Marine plastic litter can originate from multiple Especially in low- and middle-income countries,
sources, both land- and sea-based. A rather small due to a lack of centralised systems for separating
fraction of it stems from sea-based sources, e.g. and treating waste, plastics are often dumped in
waste that ships discard directly at sea. Instead, landfills and might end up in adjacent rivers, lakes
most of the marine plastic litter comes from land- and the sea 4. The location and fluxes of plastic
based sources, such as households, municipalities, debris depend on various factors, such as pollution
(touristic) activities at the beach/coast or har- sources, local weather, site and water conditions
bours/ports, agricultural activities and landfills. (e.g. currents), as well as the type of plastic 5.
Figure 1: Pathways and fluxes of plastics into the oceans 6.
To identify the sources and assess the extent of sound database derived from marine litter moni-
plastic pollution, different monitoring approaches toring and waste management.
have been developed, which require in all cases a
// 9

Introduction
Figure 2: Process of monitoring marine litter (own delineation based on 5).

Assessment
Goal Zoning for Observation Observation (e.g. material,
determination Site selection observation timing execution Documentation
type, size)
Figure 2 describes the generic steps of a marine lit- toring only provides a snapshot at a place at a
ter monitoring process. In advance of selecting an given time, so magnifying the sample data and
appropriate monitoring method, the monitoring estimating the total amount of litter may not
objectives should be defined to enable the alloca- be accurate if not conducted regularly 5.
tion of required resources. Potential monitoring
objectives may be, for example, to generate com- 2) Floating booms and nets serving as floating litter
parable results based on which targeted preventive barriers are typically used to measure floating
measures could be initiated and/or to raise public litter on the surface of river basins and in the
awareness on the issue of marine litter. After the first metres of the rivers’ water column enabling
determination of the goal, a monitoring approach its collection, quantification, categorisation,
typically starts with the selection and description and weighting. The method provides a snapshot
of suitable sites prone to marine litter accumula- of the litter in the part of the river that was
tion. Such sites are usually 7: monitored at a given time. Thus, high-quality
data showing the variation of waste can only be
› Areas near land-based sources of pollution derived if the method is applied frequently 5.
(e.g. rivers, sewage);
3) Net trawling to monitor benthic debris can
› Coastal areas; cause serious environmental damage to the eco-
system and should therefore only be considered
› Depressions on the seafloor; when existing regular bottom trawling pro-
grammes, for example in assessing fish stocks,
› Sites with slow hydrodynamics (e.g. weak are implemented. In addition to avoiding
circulation, low currents) 7. harmful consequences, monitoring costs can be
reduced in this way. The main advantage of this
Based on the monitoring goal and the selected site, method is its accuracy as it can provide exten-
resources are being allocated, the selected loca- sive multi-parameter data for macro-waste 5.
tion is divided into zones, the observation period
is determined, and finally the data is collected, 4) Scuba-diving is used to monitor the nearshore
assessed and documented. Common methods for environment for benthic pollution. If volun-
assessing plastic pollution in marine ecosystems, teers are not available, this method is highly
which can be grouped into sample gathering and cost intensive. Furthermore, the feasibility
visual observations, are: of scuba-diving depends on the accessibility
of the diving sites and is prone to subjective
Sample gathering: perceptions of the divers 5.
1) Beach litter monitoring is a widely used method Visual observations:

for measuring pollution in coastal environ-
ments. The method relies on area gridding, 5) Visual ship-based observations: To reduce costs,
manual collection, counting and categorising surveillance from a vessel does not have to
litter by size, material and/or type of items. be a specific activity but can be conducted in
The method is simple, cost-effective and helps parallel with other ongoing activities, such as
to understand the extent of littering, to raise fishing, research activities or tourist cruises 8.
awareness, and to facilitate preventive meas-
ures. It is less suitable for regularly cleaned 6) Visual off-shore observations can be carried out
beaches, poorly accessible shores, or extremely either directly from observation platforms, aer-
littered areas. Moreover, beach litter moni- oplanes, bridges, or indirectly through video
// 10

Introduction
equipment. Visual off-shore observations can be used in areas with dense vegetation or
be performed both stationary and dynami- rough terrain 5.
cally 9. In stationary observations, an observer/
device is stationed in a fixed spot, for example Plastic items can be transported over long dis-
on a bridge over a river, while dynamic riverine tances and accumulate over time in areas very
observations, similarly to ship-based marine far from their places of origin 12. Existing data
monitoring, rely on boats and use similar collection systems are limited and thus unable
protocols 10. This method is primarily suitable to answer fundamental questions, e.g. regarding
for macro-sized litter. the concentration of plastic waste in the sea and
its spatial and temporal dynamics. This is partly
7) Towed cameras: Compared to monitoring with due to the geographic scale of the problem 8, a
trawls and divers, monitoring benthic waste lack of harmonized methodological approaches
with the help of towed cameras is more accu- and non-comparable metrics 5, 11, and due to the
rate as it enables the avoidance of bias in sam- diversity of the waste items themselves. Efforts
pling and assessment and allows for post-sur- to fill the current monitoring gaps of marine
vey evaluation. The method is cost-effective litter have led to a growing interest in imaging or
and does not require a high level of expertise remote sensing technologies using boats, aircrafts
or expensive equipment. However, it is limited or satellites to monitor physical characteristics of
to areas with good water visibility and cannot the earths environmental components 13.
1.3 Remote sensing

in the context of environmental monitoring
According to the U.S. Geological Survey, “remote Specific uses of remotely sensed images of the
sensing is the process of detecting and monitoring the Earth in the context of environmental monitoring
physical characteristics of an area by measuring its include 13:
reflected and emitted radiation at a distance (typi-
cally from satellite or aircrafts)” 13 . › Mapping large forest fires;
Examples for the application of remote sensing › Tracking clouds for weather forecasting;
are 13:
› Monitoring erupting volcanoes;
› Cameras on satellites and aircrafts taking
images of large areas of the Earth’s surface; › Monitoring dust storms;
› Sonar systems on ships taking pictures of the › Mapping changes in farmland or forests over
ocean floor without having to sail or dive to several years or decades;
the bottom of the ocean;
› Discovering and mapping the rugged topog-
› Cameras on satellites can be used to take pic- raphy of the ocean floor, e.g. huge mountain
tures of temperature changes in the oceans 13. ranges, deep canyons;
› Monitoring animal populations and move-

ments of the herds 13.
// 11

Introduction
Figure 3: a ) Spatial distribution of the “MODIS burned areas” reference products and Copernicus EMS rapid mapping
activations (areas of interest are overlaid); (b) spatial distribution of the identified burned areas; (c) focus
map of the identified burned areas over the Calabria Region; d) focus map of the identified burned areas over
the Sicily region. (Derived from 14).
1.4 History on Remote Sensing of Plastics

The early recorded remote sensing of anthropo- tion Marine Debris Program and the Japanese
genic waste was conducted from land or on boats Government in the North Pacific Ocean as well
through visual inspection combined with taking as Alaskian waters revealing presence of plastic
images or photographs 15-18. Some of the early objects including styrofoam, ghostnets or derelict
coordinated extensive activities were through the fishing gear 19. With time, field surveys started
National Oceanic and Atmospheric Administra- using imaging sensors integrated on airborne and
// 12

Introduction
fixed platforms for wide aerial observations and evaluating ways to best leverage remote sensing
direct measurements whilst satellites provided technologies for monitor plastics. The European
indirect data 20-23. These manned airborne obser- environmental observation program Copernicus,
vations included trained human observers who coordinated by the European Commission, offers
counted visible floating and slightly submerged lit- information services drawn from satellite Earth
ter. Over time, the benefits of fixed, airborne and Observation (EO) and in-situ (non-space) data.
satellite sensors have been recognized by stake- Copernicus is served by a set of dedicated satellites
holders as reflected by the significant resources (the Sentinels) and further commercial and public
being made available to the scientific and industry satellites as well as in-situ systems such as ground
community to assess and propose remote sens- stations, which deliver data acquired by a multi-
ing-based solutions for plastic waste monitoring. tude of sensors on the ground, at sea or in the air.
The European Space Agency ESA is in charge of
Recently, satellite-based remote sensing technol- the space component, i.e. responsible for develop-
ogies have been increasingly explored through ing Sentinel satellites on behalf of the EU and data
with the support of space agencies interested in management services.
Table 1. Examples of space agency initiatives related to remote sensing of plastics.

Agency Programme/Initiative Goals
ESA Atlantic Regional Initiative Promote and explore the use of information about the environment
(e.g., water quality, oil spills, leakage plastic litter) generated
from remote sensing in the Atlantic region to support future oper-
ational monitoring and decision making.
Blue Worlds Task Force As a team of ESA Member States the task force has some focus
on the use of remote sensing to support the sustainable and
interdisciplinary monitoring of the maritime environment. The task
force provides recommendations that contribute to priority issues
for the ESA Council at Ministerial Level
Discovery Campaign Support innovative and disruptive scientific evidence-based re-

search to assess the application of remote sensing technologies
to monitor plastic marine litter.
EO4Society Support scientific evidence-based research to advance EOs, better

understand the changes or state of the environment and make
decisions based on gathered data.
General Support Technology Programme Evaluate and develop technologies for future space deployment
(e.g., testing remote sensing technologies suitable for remote
sensing of floating plastic waste).
Plastic-Less Society Explore the feasibility of commercializing information and servic-

es utilizing remote sensing technologies to address marine plastic
waste pollution (e.g., locating hotspot, pathways, sources, sinks).
NASA Research Opportunities in Space Explore and evaluate the advances in remote sensing of the green
and Earth Sciences Program Element and blue planet (e.g., check the prospects of remote sensing with
A.3 Ocean Biology and Biogeochemistry current or future satellite sensors).
Earth Science Technology Office Support the advancement of sensor technologies relevant to space
based remote sensing of the Earth system (e.g., testing new sen-
sors for remote sensing of plastics).
PSA AI Moonshot Challenge Assess the combines use of remote sensing technologies and
artificial intelligence to sustainably monitor the environment (e.g.,
marine plastic litter).
// 13

Introduction
ESA also funded two feasibility studies for the whilst some already gathered relevant information
projects RESMALI and OPTIMAL on remote from drones at a small scale. A list of scientific
sensing of marine litter 25. The extensive research research projects that have been conducted or cur-
findings of these ESA activities have been critical rently ongoing is presented in the annex (Annex A1.
in establishing an initial roadmap based on capa- Scientific Research Projects).
bilities and limitations of remote sensing technol-
ogies relevant for plastic waste monitoring 26. Several of these experimental scientific evi-
dence-based efforts to leverage current remote
National Aeronautics and Space Administration sensing technologies have revealed the importance
(NASA) initiatives include the “Research Oppor- for implementing imagery fusion to better explore
tunities in Space and Earth Sciences Program Ele- the detection, identification, quantification or
ment A.3 Ocean Biology and Biogeochemistry” tracking goals when monitoring floating plastic
and future investments planned with synergies litter because these sensors are not fully fit-for-
across NASA. Both ESA and NASA supported purpose or mature enough 26-28. These disruptive
several workshops that brought together interdis- experimental studies have unfortunately been
ciplinary international experts on the topic. limited to specific sensors, scenarios, times and
study areas a challenge resulting with available
Governments and environmental agencies have resources in a field that is still in the early stages of
also been investing in remote sensing research development.
1.5 Basics of Remote Sensing

Remote sensing is a way of getting information colours that the human eye can perceive and
about an object-of-interest by looking at it from distinguish. True colour images apparent to the
a distance. It therefore means no direct contact human eye are mainly composites of variable
between the observing sensor and targeted shades and tones of Red-Green-Blue (RGB)
object-of-interest to the observer. Information colours. Infrared (IR) the part of the electro-
recorded can be sound or light measured at magnetic waves or ambient light beyond the
specific temporal, geo-spatial pixel and spectral visible spectrum > 700 nm whereby the human
resolutions that is transformed into an image eye cannot easily distinguish light. Common
to derive shape, distance, size or colour details regions include Near IR (NIR, 700 – 1000
of the target using algorithms. An algorithm is nm), Shortwave IR (SWIR, 1000 – 2500 nm),
a set of commands or steps to be followed for Midwave IR (MWIR, 2500 – 6000 nm) and
handling data to generate an output that can be Thermal or Longwave IR (LWIR, 1000 – 14500
a qualitative or quantitative variable. nm). For example, the human eye acts as a sensor
that detects visible light that is transformed in
In ocean colour remote sensing, the derived the brain to infer qualitative details like shape,
parameters are used as proxies or indicators of tone, true apparent colour, size, form, texture or
optical active constituents of the aquatic envi- patterns of target objects.
ronment. Light or electromagnetic waves from
the sun possess energy at different parts of the Using manmade sensors both qualitative and
radiation spectrum. The spectrum includes high quantitative information can be derived from the
energy cosmic rays, x-rays, ultraviolet, visible signal obtained from targets of interest. Remote
to the low energy shortwave, thermal infrared, sensing technologies relevant for monitoring the
microwaves and radio waves. Visible light is the natural environment are grouped into active or
part of ambient light in the wavelength range passive sensors. These two types of sensors measure
from 400 to 700 nm that corresponds to the light that is then transformed into numerical num-
// 14

Introduction
bers relevant about environmental geological, bio- time during daylight. Sun synchronous satellites
logical and physical characteristics of object-of-in- have extensive applications in meteorological,
terest. Active sensors (e.g., laser scanner, sonar, ocean and terrestrial remote sensing.
LIDAR, radar) have to generate an energy signal
that is sent or directed towards an object-of-inter- Geostationary satellites move at the same speed as
est and then this produced pulse is measured after the Earth is rotating making it possible to image
interacting with the target. Passive sensors (e.g., the same location at all times and find application
imaging spectrometer, digital camera, thermal in real-time applications such as communication,
scanner, microwave radiometer) do not generate weather monitoring and tracking. Low orbit
own energy signal but only detect ambient sun- satellites include the International Space Station
light. Active sensors can be operated in any envi- and other sensors used for on-demand remote
ronment because of their ability to generate energy sensing application. Water quality monitoring is
sources whereas passive tools depend on signals widely conducted by using optical remote sensing
from other items like the sun. As passive remote technologies that obtain light signal in the visible
sensors relevant for environmental monitoring and IR spectrum 29, 30. The concept is based on
depend on energy generated by external sources the detecting and measuring artificial or natural
they are prone to environmental conditions (e.g., objects that possess a unique optical signature that
clouds, dust, rainfall) especially when operating at can be observed from optical remote sensing tech-
high altitudes (e.g., aircrafts, satellites). nologies. Objects that have any apparent colour
the human eye can determine can be considered
Remote sensing technologies have spectral, as optically active in the visible spectrum (e.g.,
geo-spatial pixel and temporal resolution charac- plastics, algae, suspended sediments, vegetation,
teristics parameters that are relevant to the mon- water). The colour of water is controlled by four
itoring of the environment. Spectral resolution optically active components (e.g., water itself,
is the amount of spectral information a remote algae, organic material, suspended solids, dis-
sensing instrument can resolve from a measured solved organic matter). The unique signature of
signal. The primary specifications are the num- these optically active materials can be visualized
ber of spectral wavebands or channels, spectral as the signal (e.g., radiance, reflectance) over the
response function and full-width at half-maxi- light spectrum measured. A simple way to dis-
mum of the instrument. A true colour composite tinguish the observed signal is to use the spectral
image requires three wavebands Red, Green reflectance shape and magnitude. The shape can
and Blue (RGB; R = 650 nm, G = 550 nm, B = be explained as the peaks representing local region
450 nm). Instruments can be designed as single, of very high reflectance or valleys meaning lowest
multispectral (< 20 channels) or hyperspectral (> magnitude in a given portion of the measured
100 channels) sensors. Spatial or geo-spatial pixel signal. Valleys in the spectral reflectance are also
resolution refers to the least qualitative ground known as absorption features.
sampling distance or pixel size that can be deter-
mined from an image. Remote sensing images can Remote sensing algorithms are used to produce
be reported as course (> 1000 m/pixel), moderate maps based on statistical correlation of the meas-
(< 500 m/pixel), high (< 30 m/pixel) and very high ured signal (e.g., reflectance, radiance) and any
(< 5 m/pixel) geo-spatial resolution. Temporal res- in-situ measured parameter with the matching
olution refers to the time interval data of interest is optical signature (e.g., vegetation). A relation-
gathered. Time intervals can range from seconds, ship equation is derived regressing diagnostic
hours, days to weeks. When monitoring anything wavebands to the in-situ measurements. The
the time resolution of the sensor or observing equation will be used to compute the proxy or
interval can be determined by how fast or slow the indicator of the environmental parameter (e.g.,
target physical/chemical/biological descriptors are amount of suspended sediments in the water)
changing. Satellite orbits explain how the sensor because from remote sensing we only measure
platforms move at different altitude, speeds, path- and infer the target material based on the signal.
ways relative to our Earth surface. Sun synchro- Correlation tests that have been widely used to
nous or polar orbiting travel from North to South derive an index or proxy include empirical linear
over the poles at an altitude range of 200 – 1000 regression, polynomial fits and semi-analytical
km collecting images over a fixed area at a specific methods.
// 15

Maturity and Advances
in Remote Sensing of Plastics

Maturity and Advances in Remote Sensing of Plastics
The field of remote sensing is still in the An understanding of the physical properties of the
early stages and research is still ongoing with plastics remote sensing technologies is ongoing
continued advances being reported. However, and has been explored of wide range of relevant
previous research has been revealing poten- materials (Figure 4). The size classes that have
tial applications of direct remote sensing in been of interest include micro and macro plastics.
the detection, distinguishing, quantify and Sampled materials were a combination of virgin
tracking of aquatic plastic litter. These research and weathered plastics 31-34. Microplastics were
efforts have been conducted in mostly con- investigated after being manually aggregated
trolled settings at various geographic location, to make a dense large target 35, the macro sized
examples provided in the following chapters objects as either scattered individual items 31, 34,
as well as in Annex A2. Application of Remote clustered to form large artificial targets 32, 36, 37 oth-
Sensing of Plastic Waste. erwise as a soup mixture of all size classes 38, 39.
Figure 4: E xamples of plastic samples of varying sizes that have been observed in remote sensing
experiments (a) ocean harvested microplastics 35, (b) soup of plastics in the Port of Antwerp 40,
(c) mixture of litter in windrows imaged by Caroline Power, (d-f) 0.5 × 0.5 m aggregated plastic
bottles and sheets 34, (g) 3 × 10 m target of floating plastic bottles 37 and (h-l) large floating
and submerged plastic structures. (Images provided by Konstantinos Topouzelis).
// 17

Sensors were integrated on different platforms in various platforms are presented below (Figure 5
space 36, 39, 41, 42, on aircrafts 20, 21, 43-46, unmanned and Table 2.1) and more detailed examples are
aerial systems or drones 34, 47, 48, air balloons 23, 49, showcased in (Annex A2.1 Platforms for Remote
floating vessels 33, 50, 51, handheld 52, fixed frames Sensing of Plastic Waste).
outdoors and indoors 22, 31, 53. Examples of the
Figure 5: Overview of remote sensing platforms relevant for plastic waste 65, examples from
(a) experiments over water using sensors (b) attached to balloons 49, (c-d) on moveable
frames and aircrafts 46, (e) fixe to ships 33, (f) fixed on bridges and (g) on drones 33.
As widely showcased by the vast scientific evi- potential application of remote sensing is limited
dence-based research projects, peer reviewed to the surface layer of floating and slightly sub-
literature, workshops as well as conferences (more merged plastics observed from nadir meaning the
detailed examples in the Annex section), the sensor will be looking straight downwards or per-
// 18

pendicular to the surface. Sensing from a viewing in a three-dimensional form is revealed. The cur-
angle perpendicular to the target area has limita- rent steps and advances in the detection, identifi-
tion in terms of abundance because only the sur- cation, quantification and tracking of plastic waste
face area coverage can be revealed from imagery from remote sensing technologies are summarized
but no data about the depth or underlying objects below (Figure 6);
Figure 6: (a) A simplified pipeline of the approach from image capture to deriving essential parameters from remote
sensing imagery. (b) Example of the pipeline used in Coastal Marine Litter Observatory (CMLO, https://cmlo.
aegean.gr/ ) and (c) CMLO maps of marine litter geolocations, types and densities. Courtesy and adapted
from 66.
(a) Remote Sensing Imagery
• Satellite missions
(e.g., Sentinel-2, WorldView-3, Validation and Verification
GeoEye-1, PlanetScope, SkySat, • Field observations
PRISMA, EnMAP)
• Known or artificial targets
• Airborne platforms
• Social media
(e.g., drones, aircrafts)
• Citizen science
• Near surface platforms
(e.g., boats, fixed poles,
bridges, handheld)
Image Preprocessing Analyses and Classifications Output

• Atmospheric correction • Pixel coverage analyses • Detection, quantification and
• Land masking • Spectral properties (shape and magnitude) identification end-products
• Cloud detection • Algorithms (e.g., band ratio, normalized (e.g., maps, amounts, polymer,
difference, band depth, neural networks, shapes, colour, sizes)
• Whitecaps detection
object edge detection, random forest) • Statistical metrics (e.g., precision,
• Sun glint correction
• Sensitivity tests pixel coverage, false positives,
• Labelling known targets false negatives, F-score)
(b)
(c)
// 19

Table 2.1: Examples of remote sensing platforms used to study floating and beached litter.
Platform Overview Study
Fixed A security camera was mounted to a customized frame with Automated river plastic monitoring
an extended arm on bridges in Indonesia to continuously moni- using deep learning and cameras 54
tor floating litter including plastics. The true colour RGB Dahua
Easy4ip IPC-HDBW1435EP-W sensor was mounted perpendicular
to observe directly downwards at the river surface. In this setup
the water surface and any floating material were captured in the
images.
A webcam was placed on a pole attached to attached to a frame A new technique for detecting colored
for the monitoring of a beach in Japan. True colour RGB images macro plastic debris on beaches using
were collected at an angle using a Vivotek IP7361 sensor. Images webcam images and CIELUV 55
captured the beach, water and part of the skyline.
Ship Floating litter was observed continuously from a transporter ves- Quantifying floating plastic debris at
sel in the Great Pacific Garbage patch using a GoPro camera. The sea using vessel-based optical data
GoPro Hero 6 sensor was attached to the side of the vessel and and artificial intelligence 33
viewed the ocean surface at an angle. Drones were also deployed
to survey a wide area of the ocean surface. Mantra net trawling
was also conducted to gather in-situ data.
Trained human observers counted floating litter from a research Floating macro- and microplastics
vessel around Antarctica. The observers used binoculars to aid around the Southern Ocean: Results
the survey from the bow or bridge of the vessel. from the Antarctic Circumnavigation
Expedition 51
Drone The work investigated the presence of marine litter with a focus Understanding through drone image
on the geomorphology and environmental characteristic of the ar- analysis the interactions between
eas it accumulated in Italy. The images used were captured using geomorphology, vegetation and marine
a DJI Phantom 4 Pro Obsidian DJI drone. debris along a sandy spit 56
A DJI Mavic 2 Pro quadcopter equipped with a video system was Eye in the sky’: Off-the-shelf unmanned
used to survey for floating litter in the Philippines. Floating litter aerial vehicle (UAV) highlights expo-
was revealed to be a potential threat to nearby turtles. sure of marine turtles to floating litter
(FML) in nearshore waters of Mayo Bay,
Philippines 57
Aircraft Hyperspectral sensors mounted on an aeroplane were used to Detection of sub-pixel plastic abun-
observe an artificial set of plastic targets in Switzerland. The dance on water surfaces using airborne
floating targets were built using Polyethylene terephthalate water imaging spectroscopy 44
bottles and information was collected from the visible to the
SWIR spectrum (400 - 2500 nm).
Three artificial targets of common plastics were investigated from Remote hyperspectral imaging acqui-
an aircraft with hyperspectral sensors (400 - 2500 nm). The tar- sition and characterization for marine
gets were low density polyethylene orange with used oyster spat litter detection 46
collectors, white plastic film and ropes.
Satellite Floating artificial targets were placed to investigate the feasibility Sentinel-2 detection of floating marine
of detecting litter using multispectral Sentinel-2 satellite data in litter targets with partial spectral
Greece. The targets were composed of high-density polyethylene, unmixing and spectral comparison with
wood and a mixture of plastic with wood. Additional remote sens- other floating materials 58
ing was achieved using a hyperspectral drone (400- 1000 nm).
Floating litter from the Japan 2011 Tsunami was observed by Dynamics and early post-tsunami
several satellites. Sensors in space included PALSAR Synthetic evolution of floating marine debris near
Aperture Radar and WorldVew-2, RapidEye, AVNIR-2. Fukushima Daiichi 59
Single to multipixel optical imaging of floating visible to thermal spectrum. The application of
or beach plastic litter has been achieved using Light Detection and Ranging (LIDAR) 20, 60, 61 as
single band 20, 34, RGB or digital camera sys- well as radar 36, 62-64 technologies have been show-
tems 20, 22, 33, 34, 45, 50, 54, multispectral 36, 39, 42, 59, ing promise in monitoring plastic waste.
hyperspectral sensors 31, 43, 46, 52 covering the
// 20

Tackling the plastic debris challenge at its source – Linking Earth Observation data
with multi-source in-situ data for modelling debris pathways from source to sink
Figure 7: L inking Earth Observation data with multi-source in-situ data

for modelling debris pathways from source to sink.
The project was funded by the European Space Agency’s within its program “Basic Activities in the framework dis-
covery campaign on Remote Sensing of Plastic Marine Litter” and implemented in 2020-2022. The objective of the
project was to link earth observation data from different platforms with in-situ data for modelling debris pathways
from source to sink. Knowledge about particle transportation is still limited as current practise of point measure-
ments cannot provide a full view on complex source-sink relationships. Hence, upscaling in-situ point data of litter
with EO and hydrodynamic models was the central concept of the project.
For the first time, a wide range of technologies were combined and tested, including satellite remote sensing, drone
technology, near-range cameras, various in-situ plastic sampling techniques, lab analytics and hydrodynamic mod-
elling. Each module was independently developed by the different partners, in order to test the readiness level of the
various technologies. The modules were finally linked in order to demonstrate the inter-linkages and the benefits of
the technologies as a whole.
Starting with low-cost camera systems operated from bridges above the Po River (Italy), to identify floating mac-
roplastic particles in the images using artificial intelligence. The number of observed particles were used as input to
seed a macroplastic transport model (HYDROMOD-Tracer) which provides transport pathways for the particles.
// 21

Figure 8: C amera-based macroplastic detection. Example of an original image (right) and the predicted detec-
tions of macroplastic particles (left). Particles with a blue and a red box indicate correct detections,
areas with a blue or red box only indicate false positive or false negative detections, respectively.
In addition to the macroplastic, in-situ microplastic data collected in the river itself as well as in transects along
the coastline, were used to calibrate and validate the hydrodynamic models (Regional Ocean Modelling System
and the Lagrangian model) to simulate the dispersion pathways of microplastics. The collected in-situ microplastic
samples were examined by laboratory analytics to quantify the abundance of microplastic and to qualify the plastic
type. In addition, near-shore fronts with a high abundance of microplastic debris could be identified. The locations
of these fronts have also been proved using satellite-based total suspended matter (TSM) information of Sentinel 3,
in combination with a newly developed application for generating high-resolution (10m) TSM maps based on Sen-
tinel-2 data and a machine learning approach.
Along the further transportation pathway, drones were used at potential hotspots along the coastline. By using
the LitterDrone software which employs a machine learning algorithm, based on parameters such as the colour
and shape of plastic objects and beach background, macroplastic particles (>2,5cm) could be identified in the very
high-resolution images. With the acquired drone imagery, these particles were automatically classified to quantify
the abundance of plastic litter.
// 22

Figure 9: O
rthomosaic (composed of hundreds of single pictures) of Barricata beach
with the detected and identified objects (left) by the LitterDrone software (right).
With additional research, development and test activities in the future, this monitoring concept can lead to an
operational system that provides continuous spatial information crucial for entities and stakeholders involved in the
mitigation and removal of plastics in the environment. Such monitoring system could then not only support pre-
vention and counteraction activities, but also spatially target interventions and help to make them more efficient.
Monitoring concepts such as demonstrated in this project can act as a significant contribution towards the crea-
tion of an urgently needed legal basis for avoiding water system impairment through plastic contamination, since
threshold monitoring will require a functioning measurement system.
2.1 Detection
The human eye and synthetic sensor technologies Essential Plastic Litter Descriptors. Essential
are primary tools in the direct detecting of plastic Plastic Litter Descriptors any qualitative and
litter using monochromatic, true RGB or false quantitative characteristic of plastic waste that
colour composite information. Detection is a sim- can be used to distinguish it from other materials
plified binary approach that can be exemplified including shape, form, colour, size/ dimensions,
by a zero suggesting absence and a one indicating polymer type, density or weight. However, apply-
presence of a target of interest. It can also be ing machine learning algorithms after training
achieved by looking for anomalies in imagery. the models has made it possible to automatically
Anomalies can be used to directly or indirectly detect plastic waste, thus reducing the tedious
detect plastics by using prior knowledge about effort by trained human experts. Therefore, the
the environment and relevant descriptors of accuracy of the most automated detection meth-
plastic waste. A trained observer can inspect true ods depends on continuous training of machine
colour images manually to detect and derive learning models.
// 23

Further examples of how machine learning has spectral reflectance at specific wavebands as input
been applied to the detection and quantification to derive a numerical number or index that is
of remote sensing imagery is presented in Annex used to infer presence or absence of plastic-based
A3. Machine Learning Application in Remote material. Deriving the index can be the ratio,
Sensing of Plastic Waste. There are some benefits difference or summation of the measured remote
that have been reported with respects to using true sensing wavebands in a linear or polynomial
colour images, is the capability for both manual equation. Some examples include the Advanced
as well as automated methods to extract details Plastic Greenhouse Index 71, Floating Debris
about shape, colour, size and form descriptors of Index 72, Hydrocarbon Indexes 35, 73, 74, Normal-
suspected plastic waste 42, 45, 47, 48, 55, 67-70. However, ized Difference Plastic Index 75, Plastic Green-
it is important to consider the fact that for the house Index 76 and Relative Band Depth Index 77.
accurate verification of suspected plastic waste, Uncertainties in derived distribution proxy maps
polymeric identification or in-situ sampling needs from such algorithms have been associated with
to be added in any related monitoring protocol. false positives. False positives can be resolved
by adapting the thresholds through detailed
Size makes a difference when considering in-situ validation exercises. Stakeholders tend to
the detection capabilities of various sensors. require the maps for various purposes, in some
Microplastics have been assumed to be detectable cases the degree of uncertainty is not a priority
in relatively large, aggregated patches that might as the goal is to know the location or potential
also consist of other size classes, whilst macroplas- size of the plastic litter (e.g., when, where, how
tics can be observed as individual or aggregated large). The alternative complementary detection
items using high to very fine geo-spatial pixel can be achieved by indirect mapping of essential
resolution sensors (< 10 m/pixel). Depending on ocean variables. Essential ocean variables are the
the geo-spatial capabilities of the sensors, further geophysical and biological properties of marine
descriptors can also be derived from images to bet- environment. Sea surface waves, currents, wind
ter classify the composition of the plastic waste. speed, direction and ocean colour end-products
are some of the essential ocean variables that have
Mapping of the surface distribution of plastics in demonstrated reasonable statistical correlations
various environments has been achieved through with waste accumulation dynamics and path-
spectral based algorithms. These algorithms use ways 21, 78, 79.
// 24

Investigating detection of floating plastic litter from orbit
Researchers from the University of Michigan used satellite data from eight microsatellites that are part of NASA’s
Cyclone Global Navigation Satellite System (CYGNSS) mission to map the concentration of ocean microplastic 144.
CYGNSS satellites receive signals reflected from the sea from GPS satellites to measure the roughness of the sea
surface. These roughness measurements are usually used to predict hurricanes. In the presences of plastic or debris
in the sea, the waves are attenuated, resulting in lower roughness than expected, while in clean waters there is a
high correspondence between sea roughness and wind speed. According to the researchers, the results showed that
the deeper they moved towards the Great Pacific Garbage Patch, the greater was the discrepancy between wind
speed measurements and surface roughness. Using this method in combination with plastic concentration data
from the literature, the researchers mapped daily concentrations of microplastics throughout the ocean (fig. 4). This
microplastic dataset was recently published in NASA’s Physical Oceanography Distributed Active Archive Center
(PO.DAAC).
Figure 10: NASA’s CYGNSS mission maps the concentration of ocean microplastic
with data from eight microsatellites
(Source: ESA investigating detection of floating plastic litter from orbit).
// 25

Satellite-based marine debris data set
Researchers from the Remote Sensing Laboratory of the National Technical University of Athens (NTUA) and the
National and Kapodistrian University of Athens (NKUA) have recently developed a marine debris data set based
on the multispectral data from the Copernicus program’s Sentinel-2 mission. The research project is one of those
ESA-funded projects working on remote sensing of marine litter based on Copernicus Sentinel-2 data. The study
developed the MARIDA data set which allows the discrimination of marine debris from other co-existing features
such as macroalgae, ships, waves and dissimilar water types. Based on the ground-truth events, the corresponding
images were acquired from Copernicus Hub (https://scihub.copernicus.eu/ ) for the exact reported dates and loca-
tions using a mean time window of 10 days. Additionally, for the regions that are significantly affected by plastic
pollution (such as river discharges), the seasonality and the periods of maximum plastic presence were examined.
Complemented by literature review and intensive image interpretation, MARIDA provides 3399 marine debris pix-
els for various countries and different seasons, years and sea state conditions. The images are freely available (here)
and could be used for evaluating existing detection methods and developing new techniques based on available
Sentinel-2 data.
Further examples of applying remote sensing to detect and identify marine litter are listed in the follow-
ing table.
Table 3.1: E xamples demonstrating the detection, identification and quantification of litter from remote sensing.
Fixed Floating litter was imaged from a bridge using a handheld Random forest-based river plastic
multispectral sensor in Italy. The sensor was the MAIA-WV2 that detection with a handheld multispectral
similar technical capabilities as the WorldView-2 satellite. Using camera 80
machine learning it was possible to implement semantic seg-
mentation of the floating plastics. Two classes were used that is
plastic or non-plastic.
Camera systems mounted on a bridge railing and drone were Targeting plastics: Machine learning
used to study plastic waste in Belgium. Imaging was done using applied to litter detection in aerial
a MicaSense RedEdge-M multispectral camera. Classification into multispectral images 81
11 groups of materials and quantification by area coverage was
demonstrated using machine learning approach.
Ship Floating litter was observed continuously from a transporter Quantifying floating plastic debris at
vessel in the Great Pacific Garbage patch using a GoPro Hero 6 sea using vessel-based optical data
action camera. Machine learning supported the object detection and artificial intelligence 33
and quantification of floating litter observed. Size distribution was
derived from the images captured during the survey.
Trained human observers counted floating litter from a research Floating macro- and microplastics
vessel around Antarctica. The observers used binoculars to aid around the Southern Ocean: Results
the survey from the bow or bridge of the vessel. Floating objects from the Antarctic Circumnavigation
were identified, classified and quantified. The main groups were Expedition 51
different types of plastics and non-plastics. Properties of the
litter recorded included colour, buoyancy, type of object, size and
possible function it was used for.
// 26

Drone Plastics waste surveys using DJI Phantom 3 Advanced (Adv) Use of unmanned aerial vehicles for
quadcopter were conducted in Saudi Arabia and the imagery cap- efficient beach litter monitoring 48
tured was analysed using machine learning techniques validated
by visual inspection. Quantification and classification of litter
objects was achieved. Counts of plastic objects and abundance
maps were generated
DJI Phantom 4 RTK drone images captured in Portugal were Detecting stranded macro-litter catego-
analysed using machine learning methods to detect, group and ries on drone orthophoto by a mul-
quantify litter. The outputs were abundance maps, item counts of ti-class Neural Network 82
waste, plastic bottles, fishing ropes, octopus pots, fragments.
Aircraft Hyperspectral images gathered from an airborne survey in Portu- Hyperspectral imaging zero-shot learn-
gal were analysed using machine learning classification ap- ing for remote marine litter detection
proaches. The imagers were the Specim FX10e and HySpex Mjolnir and classification 83
S-620 measuring from 400 nm to 2500 nm). Known and unknown
targets were water, orange target, white target, rope target, con-
crete pier, trees, and boats.
Human observers and imaging sensors aboard an aircraft ob- Evidence that the Great Pacific Garbage
served floating plastic litter in the Great Pacific Garbage Patch. Patch is rapidly accumulating plastic 60
True colour RGB imagery were automatically analyses to quan-
tify and classify the floating litter. Validation of the automated
approach was completed by information gathered by the trained
human observers. Dimension, shapes, colour and form of the
floating litter was derived from the imagery analyses.
Satellite Marine beached litter was quantified and detected in Chile using Anthropogenic marine debris over
WorldView-3 satellite imagery. The machine learning techniques beaches: Spectral characterization for
were used to make maps of the Expanded polystyrene or mixed remote sensing applications 84
plastics groups and density in the study area.
A training dataset was created and a machine learning approach Application of Copernicus EO data 85
was evaluated on the potential classification of plastic waste ar-
eas in Sentinel 1 and 2 satellite imagery. The groups were known
targets with various water, landcover classes including plastics
with sub-categories like greenhouses, plastic, tyres and waste
sites. Algorithms based on such a dataset might have the ability
to detect pixels matching the proposed classes.
2.2 Identification
Simple detection might not be sufficient to meet aircrafts 43, 44, 46, 74 and satellite mission 41, 89 to sense
the needs of stakeholders interested in plastic plastics. Similar to the laboratory methods, a
waste hence there is a need to further identify or best prediction of polymer composition in plastic
characterize the objects. The approach of validat- waste will be determined by matching the remote
ing what detection methods reveal as suspected or sensing signal of suspected plastics to an open-ac-
known plastic materials involves detailed analyses cess or a commercial encyclopaedia of known
by humans in the laboratory or outdoors. Key materials. Efforts to further expand well curated
ongoing research has been on scaling up identifi- open-access spectral libraries are ongoing but are
cation of polymers from laboratory technologies to still limited considering the diversity of virgin and
remote sensing tools in the natural environment weathered plastic bearing materials that continue
using the prospective application of hyperspectral to reach the natural environments as well as being
SWIR sensors on handheld or platforms 52, 86-88, produced by humans.
// 27

Detection of waste accumulation in Mumbai
The University of Georgia, in collaboration with the Indian Institute of Technology-Kharagpur, is carrying out
a case study in Mumbai, India, to explore plastic pollution patterns across the urban landscape and the associated
geographical and socioeconomic factors that often influence waste distribution. Mumbai has been chosen as a pre-
liminary study area for its mosaicked urban landscape, socio-economic fragmentation, tropical monsoon climate,
proximity to the ocean, and riverine environment. The work aims to create accurate, inexpensive, and scalable
plastic waste detection methods for an urban feature space in India using satellite-based imagery and hyper-local
socioeconomic data to elucidate waste accumulation patterns and to inform underlying challenges and successes of
waste management efforts in the city, helping communities develop more targeted waste management strategies.
Methods for a highly cognitive deep learning algorithm that fuses optical and RADAR modes of satellite data types
(WV3, PlanetScope, Sentinel-2, and ICEYE) with community-level data were developed to detect waste accumula-
tion sites comprised chiefly of plastic material.
Intensive testing has been conducted to date using PlanetScope’s 8-band 3m spatial resolution optical satellite
imagery. To address the spatial-spectral trade-off in PlanetScope data, instead of straightaway detecting waste,
a heuristic method was devised to detect and remove all non-waste classes (e.g., buildings and vegetation) from
the satellite image feature space. Remote sensing-based band ratios and indices that gave the highest separation
between waste and non-waste classes were applied to the satellite image. This helped heuristically reduce the search
space by 30% (95 million pixels to 66 million pixels). In this reduced search space, unsupervised k-means clus-
tering, an algorithm that uses unlabelled data and clusters features based on shared similarities, was performed
and then ranked based on how pure it was in terms of waste pixels. Validation sites were identified based on field
sampling and interviews. Top-ranked clusters were selected as model-detected waste accumulation sites, correctly
identifying 244 of the 247 ground-sampled waste sites. The preliminary results from the proof-of-concept image
processing framework look promising in studying the spatiotemporal distribution of plastic waste in megaci-
ties with complex land use patterns. However, more false positives offset the high number of true positives. This
requires further improvement of detection using multimodal data – optical, radar, lidar, and thermal. The resulting
model will perform a long-term analysis to evaluate sources, pathways, and drivers of plastic waste at the urban
land-water interface.
Figure 11:
A cropped image (RGB)
from Planet 3-m data
for Mumbai showing the
potential waste-positive
localities after applying
heuristic feature space
reduction and clustering.
Yellow pixels represent
plastic waste locations
obtained from the pre-
liminary analysis.
// 28

2.3 Quantification
An additional descriptor about the plastic waste derived counts or weight over a unit area observed 60,
after detection would be the surface abundance 84, 92
, pixel percent coverage or heat map of suspected
as observed from a nadir viewing angle. Manual plastic waste 36, 91, 93.
and automated methods have been used to provide
estimates of plastics within a region of interest of a Counts by trained observers combined with
captured using monochromatic, false or true colour statistical techniques tend to be used as validation
composite images. Machine learning has been tools and findings can be regressed to extrapolate
applying supervised and unsupervised algorithms to estimates of plastic waste in a selected region-of-in-
quantify the suspected plastics from remote sensing terest. In a similar way, net trawl data can be sta-
images 47, 48, 67, 70, 84, 90, 91. However, these related tistically correlated or regressed to remote sensing
studies have limitations because the in-situ surveys products (e.g., signal in the form of spectral radi-
do not cover the full extent of imaged regions or the ance or reflectance) better quantify better plastics
targets are artificially placed to assess performance in a wide geo-spatial area. Enumerating surface
of the quantifying algorithms. In-situ datasets are aggregated plastics can be biased and challenging
important in developing remote sensing algorithms especially for automated tools because they may
as they are used to statistically validate and verify rely on distinguishing edges of objects that tend to
the accuracy of the approach in estimating or indi- overlap as plastic waste litter of all sizes randomly
cating the abundance of observed target. Quantifi- aggregate in nature. Furthermore, a vertical profile
cation techniques have been revealed to be challeng- of the plastics once they begin aggregating can be
ing on water bodies compared to land surfaces. On difficult to obtain from above surface imaging sen-
water the plastic waste is constantly in motion yet sors. Without the side or vertical profile imaging
fixed on land likely trapped in vegetation or lying matching above water remote sensing constraints
on sand. The abundance efforts involved the task of the quantification details that can be derived about
detection, classification and quantification 48, 67, 82, plastic waste especially at sea.
2.4 Tracking
The pathways from source (e.g., land, river, sea) to in GOCI data linked to the extreme weather
sink (e.g., land, river, sea) of the plastic waste can events in 2011 namely East Japan tsunami and
be obtained by continuous detection and quantifi- Japan earthquake. When anomalies are detected
cation from a fixed point of reference. The images one problem is knowing the source or event gen-
can be taken from fixed poles, bridges, drones, erating such variations at the sea surface. During
ships, aircrafts, satellites over the same target area the tsunami extreme event these anomalies were
at repeated intervals. These intervals are controlled more likely a result of the floating litter that was
by how fast the plastics as target of interest move composed of natural and anthropogenic mate-
on water or accumulate on land. rials. Verification in-situ images were available
from social media, environmental agencies and
South Korean satellite Geostationary Ocean news outlets. However, during normal condi-
Color Imager (GOCI) has captured hourly tions in the region without the support of in-situ
images at 500 m/pixel of the Korean sea between images, it is challenging to verify or confirm
09:00 to 16:00 local time in the last decade. what would be causing anomalies if detected
Despite the coarse geo-spatial resolution ongoing in satellite imagery. Therefore, it highlights the
proof-of-concept have been exploring the detec- importance of in-situ datasets in remote sensing
tion and tracking of anomalies on the sea surface application.
// 29

Planet Labs PlanetScope offers very high geo-spa- tools with promising results 22, 54, 94, 95. Using a
tial pixel resolution of 3 m/pixel and SkySat 0.5 webcam fixed on a pole in Japan, beach litter
m/pixel. Imaging coverage is dependent on cloud dynamics were continuously tracked over several
cover and geographic location, some observations months. The method demonstrated the capabil-
can be obtained at least twice a day over a region ity to investigate the types and amounts of litter
of interest. Such capabilities can be useful for the that were reaching the beach whilst linking the
detection and tracking of plastic waste on land observations to environmental conditions 22. Litter
and over water bodies considering the flow rate of passing under bridges in Indonesia was monitored
the waste is relatively similar to the daily observa- using cameras mounted on bridges in Indonesia.
tions. The approach managed to detect, distinguish and
quantify the plastics based on imagery captured
Balloons, drones and low-cost fixed camera have by the camera system whilst validation of the find-
also been explored as monitoring and tracking ings was supported by visual inspection 54.
2.5 Online Platforms and Web-based Tools

Web-based tools related to plastic waste have port remote sensing application as sea-truth
gained a rising interest and usage in the last information. Examples of web-based tools
decade. The tools provide information about that provide a wide geographic coverage with
plastic litter characteristics that is either potential to support remote sensing application
gathered from remote sensing or could sup- are presented below.
// 30

2.5.1 Ocean Scan Zenodo online open-access repository allowing

permanent citation and crediting of data source.
A smartphone app ‘Ocean Scan’ is available for
An advanced online data repository or database data capturing during field survey. Users can
that is dedicated to remote sensing of plastic litter select in-situ datasets that have matching remote
applications with the ability of automatically sensing imagery. The system thrives to automat-
matchinguploaded field observations to availa- ically match any dataset to satellite observations
ble satellite data 96. High-quality match-ups of over the same location and time of field uploaded
relevant information enable the development of information. Additional information required
robust satellite algorithms that would have been when uploading data into Ocean Scan is consid-
validated using in-situ images with information ered important for future users. Such information
about presence of plastic waste. Simple techniques includes details about size of the patch observed
to sophisticated artificial intelligence methods during a survey, presence of plastics, time, geo-
can be used to analyse such well curated imagery. graphic location and types of objects. The level
Open-access to the interoperable datasets and of confidence in each dataset with regards to
uploads can be achieved via the user-friendly presence of plastics is provide by the validation
web interface (Figure 7), smartphone application options. Each dataset will carry a tag to indicate if
or application programming interface (API). there is in-situ information related to the dataset
Datasets in the repository are assigned a per- matching with remote sensing imagery reported
manent digital object identifier (DOI) through in Ocean Scan.
Figure 12: O
cean Scan graphic interface showing the dataset catalogue from various geographic locations
and additional options related to matching satellite to in-situ data, observation platform
for validation, time, type and size of litter.
// 31

2.5.2 Litterbase resource for remote sensing validation effort for

studies with information about geo-location, size
and physical descriptors than can also be derived
The open-access PostgreSQL 9.5. tool provides from imagery. Combining the satellite imagery
a global visual map of peer-reviewed and citi- information related to plastics statistical correla-
zen science studies related to plastic waste since tions can be explored by using matching in-situ
1960 97, 98. Litterbase combines a living bibliog- observations in Litterbase. Litterbase graphic
raphy that allows a number of interdisciplinary interface allows the user to group studies by the
analyses from the literature available in the plastic size class, habitat of collection, time and
database. The user-friendly platform allows the location. The tool has relevance to the possible
viewing of stored publication and related findings validation using in-situ data matching remote
about plastic waste such as biome, geoloca- sensing imagery of the near surface, floating and
tion, quantities, polymer type, environment 98. beached litter that might have been captured by
Litterbase could be a prospective complementary drones, aircraft or satellite.
Figure 13: E
xamples of the features and products available in Litterbase graphic interface.
Data layers allow user to select quantification units reported in literature, size classes,
the habitat samples were harvested from and year of study.
// 32

2.5.2 Global Partnership for Macroplastic Litter (https://wrd-forecast-

custom.azurewebsites.net/unep-plastics/ ). It can
on Marine Litter (GPML) predict up to 9 months in advance the dynamics
of waste in global rivers based on the advanced
The GPML datahub tool provides a global map United Nations Environmental Programme Center
and analytical functions. The maps are statistical on Water and Environment Global Hydrological
based distribution maps of estimated plastic waste Model and in-situ observations (Figure 9). Remote
in the natural environment whilst functionality sensing data (e.g., precipitation, evapotranspira-
ranges from modelling to data gathering application, surface area of rivers) is also incorporated
tions. Key characteristics are prediction maps from to infer plastic waste dynamics via the built-in
numerical solutions or statistical models. Numer- module called Forecasted plastic litter in rivers of
ical solutions are derived from mathematical com- GPML as it tracks environmental changes. The
putations aimed at solving theoretical scenarios tool includes a map viewer in the GMPL datahub
useful in hindcasting and forecasting applications. that shows the relationship of different environ-
One of the test phase functionality or modules mental parameters, such as transportation dynam-
in the GPML is the Risk and Warning System ics and plastic waste on the beach.
Figure 14: GPML Risk and Warning System for Macroplastic Litter user interface.
// 33

2.5.3 Global Plastic Watch the user interface do not highlight uncertainties
related to the detected litter abundance as well
as dump site area coverage. Such data could be
An online tool supported by artificial intelligence generated through ground-truthing or clean-up
algorithms generates detection and abundance surveys in future advancement of the map out-
maps from the ESA Copernicus programme Sen- puts. Suspected plastic litter zones can also be
tinel-2 satellite mission at a resampled 10 m/pixel verified to provide a probability metric for the
resolution 99. Visualization in the web-based tool detected regions. The platform provides a tool for
provides geo-location information and an esti- the potential detection of plastics over land targets
mated amount of suspected plastic litter as derived but as presented before the automated approaches
from satellite remote sensing (Figure 10). For require continuous datasets or validation to
each site additional information reported include perform better. Similar ongoing efforts are being
distance to a water body, a time series of site area explored using Planet Labs PlanetScope images to
coverage and soil type content. The statistics in detect floating litter at sea 100.
Figure 15: Example showing Global Plastic Watch web interface.
// 34

WWF Germany in cooperation with Earthrise Media has tested the feasibility to detect waste accumulation in
Vietnam with Global Plastic Watch tool.
Vietnam has a pronounced relief and a dense network of rivers and water bodies. Thus, topography and proximity
to the nearest water course influence the risk of plastic waste leaking from dumpsites and landfills into the marine
environment. The system was tested in a high-sensitivity configuration to identify waste sites in Vietnam. The waste
detection tool is composed of two convolutional neural networks that analyze and combine spectral, spatial, and
temporal signals from data collected by the Sentinel-2 satellite. This data has 12 spectral bands, which capture visi-
ble and IR light. The resolution of the data is 10 meters per pixel. Following detection, relevant metadata about the
geographic and physical setting is collected. The model was run across all of Vietnam between January 2019 and
June 2022. The results are promising: 198 garbage dumps were discovered with accumulations of waste, 20 percent
of which are within a sensitive 250 meters of the nearest body of water. With further improvement in detection,
waste (distribution) could potentially be measured rather than modelled, and the development of waste sites could
be monitored over time. The outputs of the tool can be seen at globalplasticwatch.org. This tool can help developing
effective intervention strategies to better manage the major sources of plastic waste and understand which of these
represent the greatest risk.
Figure 16: Potential illegal dumpsite close to a river. The site has changed significantly over time
(Global Plastic Watch).
// 35

2.6 Stakeholder Community Initiatives

Monitoring and ocean observing initiatives are rise in interest, the main aim has been to better
primary sources of the essential datasets needed locate hotspots for clean-up campaigns. A detailed
by stakeholders about plastic litter pathways overview of the current monitoring efforts by spe-
including Essential Plastic Litter descriptors. cific members states has been recently presented in
GPML and the Global Ocean Observing System literature 105.
(GOOS) have played major roles by establishing
platforms to bring together stakeholders, resources The Blue Planet Marine Litter working group
and strategies. GOOS aims to support a suite of was also founded by the Group on Earth Obser-
ocean observing platforms (e.g., in-water, above vation (GEO) with the main goal to investigate
water, fixed, mobile, autonomous and manned) the research gaps in technology and data needs of
that could also contribute towards remote sensing stakeholders though network platforms as well as
of plastic waste. Harmonisation of the in-situ workshops. The working group has held several
sampling and terminologies has been coordinated workshops that have been developing a roadmap
through the Joint Group of Experts on the Scien- on best practices in observing and numerical mod-
tific Aspects of Marine Environmental Protection elling of aquatic plastics. Efforts are also under-
(GESAMP) and G20/Ministry of Environment way to further develop and realize the proposed
Government of Japan. Integrated Marine Debris Observing System –
IMDOS 27 that integrates the use of remote sens-
Member States of the Association of Southeast ing as a data source.
Asian Nations (ASEAN) with the support of the
World Bank Group and PROBLUE initiative have More recently, the Task Force on Remote
Action Plans that emphasize the need for rigorous Sensing of Marine Debris and Litter was formed
monitoring strategies for plastic waste 101. Using under the auspice of the International Ocean
imaging technologies on drones or bridges to Colour Coordinating Group (IOCCG). The
observe water or land surface to perform analyses IOCCG Task Force provides a one-stop shop of
on the gathered high geo-spatial pixel resolution resources related to remote sensing of plastics
was showcased as an affordable monitoring strat- generated by four Core Topics dedicated to
egy in Vietnam 102. Artificial intelligence methods (i) technologies, (ii) algorithms and applications,
were used to quantify and classify plastics, but (ii) datasets and (iv) interdisciplinary aspects.
the challenge as reported is the limited number of Living resources include a bibliography of
datasets for training as well as validating the mod- peer-reviewed publications and datasets on the
els. It therefore means such algorithms are region topic of remote sensing of marine litter. Related
specific but can be transferred to other geographic research projects (Annex A1. Scientific Research
regions after extensive model training and tuning Projects) and events supported or coordinated
for the new study location of interest. (Annex A4. Conferences, Proceedings and
Workshops) by the Task Force are also pro-
European Union member states have been thriv- vided on the website. As remote sensing will be
ing to meet the proposed ideal scenarios outlined a component in IMDOS the Task Force also
in the Marine Strategy Framework Directive supports GOOS to address user needs, expecta-
2008/56/EC. To this end, the monitoring strate- tion and informing the community on capabil-
gies to determine changes in the aquatic environ- ities of current technology as well as what could
ment that have been utilized include human visual be improved in future sensors. The Task Force
observations, in-situ field sampling, airborne and supported UN initiatives such as (i) Ocean Con-
satellites remote sensing. Visual observations have ference through the expertise of remote sensing
been commonly applied to detect and quantify Integrating Marine Litter Monitoring to Inform
floating litter objects, at least easy to see for the Action event on 29 June 2022 in Portugal and
human eye from a distance, around European (ii) Ocean Decade Laboratory satellite activity
waters from research vessels e.g., 50, 103, 104. Routine by organising the Remote Sensing and Smart
use of airborne platforms has also been gaining a Tech for Marine Litter workshop.
// 36

Outlook

Outlook
Remote sensing of plastic waste is an emerging experiments using artificial plastic targets
approach that has been showing potential in and small area coverage studies. A prospective
meeting stakeholder needs to monitor the litter approach is the design and launching of proto-
qualitatively and quantitatively in all envi- type satellite sensors aimed at bridging the gaps
ronments. However, there is no market-ready and improve the monitoring of plastic waste.
application available yet, and current works are Such a prototype satellite will possess charac-
still in the early stages and aim to develop reliable teristics that would have been identified from
algorithms to improve the detection and meas- available extensive research on remote sensing of
urement of waste accumulation. Preliminary plastic waste using various imaging technologies.
scientific evidence-based studies (e.g., Annex A prototype satellite could also be an opportunity
material) have been reporting recommendation for the community to define and set standardized
on possibilities to scale up (e.g., handheld and methods that will allow monitoring of plastics
drone based remote sensing) from controlled assessments using well defined baselines.
3.1 Prototype Satellite

A team of experts at the Flemish Institute for High geo-spatial pixel resolution (≤ 10 m) satel-
Technological Research conducted a community lites with multispectral wavebands and specific
user needs survey that was complemented by the data models developed by multidisciplinary teams
current literature on remote sensing of plastic have been showing promise in detection of sus-
litter. The findings from the user needs survey pected plastic by inferring the inherent accumu-
were used to define specifications, capabilities lation zones of waste e.g., windrows, ocean gyres,
and characteristics for a suitable satellite that fronts. Examples of such satellite include GeoEye,
could be launched to further assess the space Pléiades, PlanetScope, RapidEye, Sentinel-2,
based monitoring of marine litter 106. The main Skysat. WorldView-2 and WorldView-3. The
regions of interest were identified as plastic waste value of these satellite sensors could be advanced
hotspots such as landfills, beaches and coastal by including hyperspectral SWIR information as
windrows. Technical specifications proposed this might allow a better identification of plastics
included have ~26 wavebands covering the visible among other optically active materials in the natu-
to short-waved infra-red (SWIR) spectrum with a ral environment or in the windrows.
very high geo-spatial pixel resolution (1 m in the
visible, 3 m in SWIR). It is crucial to prioritize Defining a sensor that is appropriate for remote
and dedicate resources for the proposed satellite as sensing of plastics might seem simple yet compli-
it could advance evidence-based knowledge about cated. The stakeholder needs based on detection,
the application of remote sensing technologies to identification, quantification and tracking capa-
monitor plastic waste. Governments, space agen- bilities of remote sensing technologies are different
cies and the industry have been recently reminded considering the various parameters than can be
and encouraged to invest in the efforts as it will resolved. For example, for a clean-up stakeholder
be a major step towards a more fit-for-purpose just knowing where the litter is present and area
plastic waste satellite sensor 107. Using the informa- coverage might be sufficient information mean-
tion that will likely be derived from the proposed ing the colour, polymer types or shape details are
concept satellite, remote sensing experts believe not needed. Governments might want to detect,
the invaluable scientific evidence-based knowl- identify and track the litter to locate the source or
edge will further allow a more refined assessment sinks. To meet any of the above capabilities with a
on the capabilities of space-based monitoring of single sensor requires trade-offs or different imag-
plastics. ing technologies because of the pixel size, spectral
wavebands and temporal revisit resolution that
must be integrated.
// 38

Outlook
3.2 Baselines and Standardized Monitoring

Baselines can be established if standardized of hyperspectral and very high-resolution imagery
monitoring is continuous and regular times series can be launched to support the detection, identi-
exist. Unfortunately, standardized monitoring fication, quantification and tracking of plastics.
methods are still not well established in interdisci- Low-cost small satellite known as cubesats could
plinary research on plastics as various studies use be explored as supporting imaging tools in space.
different metrics to detect, classify and quantify Additionally, it would be important to have
waste 108. Due to data gaps in satellite revisits of coordinated efforts for in-situ sampling matching
plastic sources and sinks, it is difficult to initi- satellite observation for validation and verification
ate clean-up campaigns solely based on remote tasks geared towards algorithm development as
sensing data. Ideally, a constellation and synergy well as fine tuning.
3.3 Stakeholder Community Discussions

Similar to the efforts of proposing a prospective and networking events. One priority has been to
satellite sensor for marine litter monitoring a engage stakeholders to better understand the user
coordinated interdisciplinary holistic approach needs whilst also communicating the capabilities
is critical for a harmonised observing tool for and limitations of remote sensing as a complemen-
plastics. One example is the IOCCG Task Force tary monitoring tool for plastics in the aquatic
on Remote Sensing of Marine Debris and Litter environment. At present efforts are underway to
composed of experts from academia, industry, provide status update on remote sensing of marine
civil service, non-profit organisation, govern- litter. The findings and scientific research gaps
ments and space agencies. The Task Force thrives identified in the recently ended projects are to be
to coordinate and support the advancement in assessed with the aim to bridge the gaps or addon
the topic especially through thematic workshops the knowledge through ongoing or future studies.
// 39

References

References
1. UNEP Marine plastic debris and microplastics – Global lessons and research to inspire action and guide
policy change; Nairobi, Kenya, 2016; p 274.
2. Harding, S. Marine debris: understanding, preventing and mitigating the significant adverse impacts
on marine and coastal biodiversity; Secretariat of the Convention on Biological Diversity: Montreal,
Quebec, Canada, 2016; p 78.
3. GIZ Marine litter prevention. Reducing plastic waste leakage into waterways and oceans through circu-
lar economy and sustainable waste management; Deutsche Gesellschaft für Internationale Zusamme-
narbeit (GIZ) GmbH: Eschborn, Germany, 2018; p 112.
4. Eremina, T.; Ershova, A.; Martin, G.; Shilin, M. In Marine litter monitoring: review for the Gulf of
Finland coast, 2018 IEEE/OES Baltic International Symposium (BALTIC), Klaipeda, Lithuania,
12-15 June, 2018; Klaipeda, Lithuania, 2018; pp 1-8.
5. Miliute-Plepiene, J.; Fråne, A.; Haikonen, K.; Youhanan, L. Overview of available methods to monitor
marine plastic litter: Incl. method for riverine litter monitoring developed within BLASTIC; IVL Swed-
ish Environmental Research Institute: Stockholm, Sweden, 2018; p 47.
6. GRID/Arendal Pathways and fluxes of plastics into the oceans. https://www.grida.no/

resources/6921 (Last 17 January 2023:
17 January 2023),
7. IMO Review of the current state of knowledge regarding marine litter in wastes dumped at sea under
the London Convention and Protocol. Final Report; International Maritime Organization (IMO):
London, United Kingdom, 2016; p 38.
8. Cheshire, A. C.; Adler, E.; Barbière, J.; Cohen, Y.; Evans, S.; Jarayabhand, S.; Jeftic, L.; Jung, R. T.;
Kinsey, S.; Kusui, E. T.; Lavine, I.; Manyara, P.; Oosterbaan, L.; Pereira, M. A.; Sheavly, S.; Tkalin,
A.; Varadarajan, S.; Wenneker, B.; Westphalen, G. UNEP/IOC Guidelines on survey and monitoring
of marine litter; United Nations Environment Programme: Nairobi, Kenya, 2009; p xii+120.
9. USEPA Marine debris in the north Pacific. A summary of existing information and identification of data
gaps; EPA-909-R-11-006; U.S. Environmental Protection Agency: San Francisco, USA, November
2011, 2011; p 23.
10. González, D.; Hanke, G.; Tweehuysen, G.; Bellert, B.; Holzhauer, M.; Palatinus, A.; Hohenblum,
P.; Oosterbaan, L. Riverine litter monitoring - Options and recommendations; European Union: Lux-
embourg: Publications Office of the European Union, 2016; p 52.
11. Serra-Gonçalves, C.; Lavers, J. L.; Bond, A. L., Global review of beach debris monitoring and future
recommendations. Environmental Science & Technology 2019, 53, (21), 12158-12167, doi:10.1021/
acs.est.9b01424
12. ESA-OSIP Remote sensing of plastic marine litter. Strategic Innovation Area: Discovery. https://
ideas.esa.int/servlet/hype/IMT?documentTableId=45087640620544725&userAction=Browse&tem
plateName=&documentId=ae460bb2ce3f682ea452f15f8bc3753f (Last 17 January 2023: 17 January
2023),
13. USGS What is remote sensing and what is it used for? https://www.usgs.gov/faqs/what-remote-
sensing-and-what-it-used#:~:text=Remote%20sensing%20is%20the%20process,sense%22%20
things%20about%20the%20Earth. (Last 17 January 2023: 17 January 2023),
// 41

References
14. Filipponi, F., Exploitation of Sentinel-2 time series to map burned areas at the national level: A case
study on the 2017 Italy wildfires. Remote Sensing 2019, 11, (6), 622, doi:10.3390/rs11060622
15. Carpenter, E. J.; Smith, K. L., Plastics on the Sargasso Sea Surface. Science 1972, 175, (4027), 1240-
1241, doi:10.1126/science.175.4027.1240
16. Horn, M. H.; Teal, J. M.; Backus, R. H., Petroleum lumps on the surface of the sea. Science 1970,
168, (3928), 245-246, doi:10.1126/science.168.3928.245
17. Carpenter, E. J.; Anderson, S. J.; Harvey, G. R.; Miklas, H. P.; Peck, B. B., Polystyrene Spherules in
Coastal Waters. Science 1972, 178, (4062), 749-750, doi:10.1126/science.178.4062.749
18. Venrick, E. L.; Backman, T. W.; Bartram, W. C.; Platt, C. J.; Thornhill, M. S.; Yates, R. E., Man-
made objects on the surface of the central North Pacific Ocean. Nature 1973, 241, (5387), 271-271,
doi:10.1038/241271a0
19. Shomura, R. S.; Godfrey, M. L. In Proceedings of the Second International Conference on Marine
Debris, 2-7 April 1989, Honolulu, Hawaii, USA, 1990; NOAA Tech. Mem. NMFS-SWFSC-154,
U.S. Department of Commerce, Silver Spring: 2-7 April 1989, Honolulu, Hawaii, USA, 1990; p
1274.
20. Pichel, W. G.; Veenstra, T. S.; Churnside, J. H.; Arabini, E.; Friedman, K. S.; Foley, D. G.; Brain-
ard, R. E.; Kiefer, D.; Ogle, S.; Clemente-Colón, P.; Li, X., GhostNet marine debris survey in the
Gulf of Alaska – Satellite guidance and aircraft observations. Marine Pollution Bulletin 2012, 65,
(1–3), 28-41, doi:10.1016/j.marpolbul.2011.10.009
21. Pichel, W. G.; Churnside, J. H.; Veenstra, T. S.; Foley, D. G.; Friedman, K. S.; Brainard, R. E.;
Nicoll, J. B.; Zheng, Q.; Clemente-Colón, P., Marine debris collects within the North Pacific
Subtropical Convergence Zone. Marine Pollution Bulletin 2007, 54, (8), 1207-1211, doi:10.1016/j.
marpolbul.2007.04.010
22. Kako, S. i.; Isobe, A.; Magome, S., Sequential monitoring of beach litter using webcams. Marine
Pollution Bulletin 2010, 60, (5), 775-779, doi:10.1016/j.marpolbul.2010.03.009
23. Nakashima, E.; Isobe, A.; Magome, S.; Kako, S. i.; Deki, N., Using aerial photography and in situ
measurements to estimate the quantity of macro-litter on beaches. Marine Pollution Bulletin 2011,
62, (4), 762-769, doi:10.1016/j.marpolbul.2011.01.006
24. Maximenko, N.; Chao, Y.; Moller, D., Developing a remote sensing system to track marine debris.
Eos Science News by AGU 2016, 97, doi:10.1029/2016EO061605
25. Martinez-Vicente, V.; Biermann, L.; Mata, A., Optical methods for marine litter detection (OPTI-
MAL) – Final Report. In Zenodo: Zenodo, 2020; p 54.doi:10.5281/zenodo.3748797
26. Martínez-Vicente, V.; Clark, J. R.; Corradi, P.; Aliani, S.; Arias, M.; Bochow, M.; Bonnery, G.;
Cole, M.; Cózar, A.; Donnelly, R.; Echevarría, F.; Galgani, F.; Garaba, S. P.; Goddijn-Murphy, L.;
Lebreton, L.; Leslie, H. A.; Lindeque, P. K.; Maximenko, N.; Martin-Lauzer, F.-R.; Moller, D.;
Murphy, P.; Palombi, L.; Raimondi, V.; Reisser, J.; Romero, L.; Simis, S. G. H.; Sterckx, S.; Thomp-
son, R. C.; Topouzelis, K. N.; van Sebille, E.; Veiga, J. M.; Vethaak, A. D., Measuring marine
plastic debris from space: Initial assessment of observation requirements. Remote Sensing 2019, 11,
(20), 2443, doi:10.3390/rs11202443
// 42

References
27. Maximenko, N.; Corradi, P.; Law, K. L.; Van Sebille, E.; Garaba, S. P.; Lampitt, R. S.; Galgani, F.;
Martinez-Vicente, V.; Goddijn-Murphy, L.; Veiga, J. M.; Thompson, R. C.; Maes, C.; Moller, D.;
Löscher, C. R.; Addamo, A. M.; Lamson, M. R.; Centurioni, L. R.; Posth, N. R.; Lumpkin, R.;
Vinci, M.; Martins, A. M.; Pieper, C. D.; Isobe, A.; Hanke, G.; Edwards, M.; Chubarenko, I. P.;
Rodriguez, E.; Aliani, S.; Arias, M.; Asner, G. P.; Brosich, A.; Carlton, J. T.; Chao, Y.; Cook, A.-M.;
Cundy, A. B.; Galloway, T. S.; Giorgetti, A.; Goni, G. J.; Guichoux, Y.; Haram, L. E.; Hardesty, B.
D.; Holdsworth, N.; Lebreton, L.; Leslie, H. A.; Macadam-Somer, I.; Mace, T.; Manuel, M.; Marsh,
R.; Martinez, E.; Mayor, D. J.; Le Moigne, M.; Molina Jack, M. E.; Mowlem, M. C.; Obbard, R.
W.; Pabortsava, K.; Robberson, B.; Rotaru, A.-E.; Ruiz, G. M.; Spedicato, M. T.; Thiel, M.; Turra,
A.; Wilcox, C., Toward the Integrated Marine Debris Observing System. Frontiers in Marine Science
2019, 6, (447), doi:10.3389/fmars.2019.00447
28. Hu, C., Remote detection of marine debris using satellite observations in the visible and near infra-
red spectral range: Challenges and potentials. Remote Sensing of Environment 2021, 259, 112414,
doi:10.1016/j.rse.2021.112414
29. IOCCG Why ocean colour? The societal benefits of ocean-colour technology; Reports of the Interna-
tional Ocean-Colour Coordinating Group, No. 7; Dartmouth, Canada, 2008; p 147.
30. CEOS Feasibility study for an aquatic ecosystem earth observing system. Report 2.0; Commonwealth
Scientific and Industrial Research Organisation (CSIRO), Australia: Canberra, ACT, Australia.,
2018; p 197.
31. Knaeps, E.; Sterckx, S.; Strackx, G.; Mijnendonckx, J.; Moshtaghi, M.; Garaba, S. P.; Meire, D.,
Hyperspectral-reflectance dataset of dry, wet and submerged marine litter. Earth Syst. Sci. Data,
2021, 13, (2), 713-730, doi:10.5194/essd-13-713-2021
32. Topouzelis, K.; Papageorgiou, D.; Karagaitanakis, A.; Papakonstantinou, A.; Arias Ballesteros, M.,
Remote sensing of sea surface artificial floating plastic targets with Sentinel-2 and unmanned aerial
systems (Plastic Litter Project 2019). Remote Sensing 2020, 12, (12), 2013, doi:10.3390/rs12122013
33. de Vries, R.; Egger, M.; Mani, T.; Lebreton, L., Quantifying floating plastic debris at sea using ves-
sel-based optical data and artificial intelligence. Remote Sensing 2021, 13, (17), 3401, doi:10.3390/
rs13173401
34. Goddijn-Murphy, L.; Williamson, B. J.; McIlvenny, J.; Corradi, P., Using a UAV thermal infra-
red camera for monitoring floating marine plastic litter. Remote Sensing 2022, 14, (13), 3179,
doi:10.3390/rs14133179
35. Garaba, S. P.; Dierssen, H. M., An airborne remote sensing case study of synthetic hydrocarbon
detection using short wave infrared absorption features identified from marine-harvested macro-
and microplastics. Remote Sensing of Environment 2018, 205, 224-235, doi:10.1016/j.rse.2017.11.023
36. Topouzelis, K.; Papakonstantinou, A.; Garaba, S. P., Detection of floating plastics from satellite and
unmanned aerial systems (Plastic Litter Project 2018). International Journal of Applied Earth Obser-
vation and Geoinformation 2019, 79, 175-183, doi:10.1016/j.jag.2019.03.011
37. Themistocleous, K.; Papoutsa, C.; Michaelides, S.; Hadjimitsis, D., Investigating Detection of
Floating Plastic Litter from Space Using Sentinel-2 Imagery. Remote Sensing 2020, 12, (16), 2648,
doi:10.3390/rs12162648
// 43

References
38. Cózar, A.; Aliani, S.; Basurko, O. C.; Arias, M.; Isobe, A.; Topouzelis, K.; Rubio, A.; Morales-
Caselles, C., Marine litter windrows: A strategic target to understand and manage the ocean plastic
pollution. Frontiers in Marine Science 2021, 8, (98), doi:10.3389/fmars.2021.571796
39. Kikaki, A.; Karantzalos, K.; Power, C. A.; Raitsos, D. E., Remotely sensing the source and transport
of marine plastic debris in Bay Islands of Honduras (Caribbean Sea). Remote Sensing 2020, 12, (11),
1727, doi:10.3390/rs12111727
40. Knaeps, E.; Strackx, G.; Meire, D.; Sterckx, S.; Mijnendonckx, J.; Moshtaghi, M., Hyperspectral
reflectance of marine plastics in the VIS to SWIR. In Data set. Available on-line, 4TU.Centre for
Research Data: 2020.doi:10.4121/12896312.v2
41. Kremezi, M.; Kristollari, V.; Karathanassi, V.; Topouzelis, K.; Kolokoussis, P.; Taggio, N.;
Aiello, A.; Ceriola, G.; Barbone, E.; Corradi, P., Pansharpening PRISMA data for marine
plastic litter detection using plastic indexes. IEEE Access 2021, 1-1, doi:10.1109/ACCESS.2021.
3073903
42. Park, Y.-J.; Garaba, S. P.; Sainte-Rose, B., Detecting the Great Pacific Garbage Patch floating plastic
litter using WorldView-3 satellite imagery. Optics Express 2021, 29, (22), 35288-35298, doi:10.1364/
OE.440380
43. Garaba, S. P.; Aitken, J.; Slat, B.; Dierssen, H. M.; Lebreton, L.; Zielinski, O.; Reisser, J., Sensing
ocean plastics with an airborne hyperspectral shortwave infrared imager. Environmental Science &
Technology 2018, 52, (20), 11699-11707, doi:10.1021/acs.est.8b02855
44. Hueni, A.; Bertschi, S. In Detection of sub-pixel plastic abundance on water surfaces using airborne
imaging spectroscopy, IGARSS 2020 – 2020 IEEE International Geoscience and Remote Sensing
Symposium, 26 Sept.-2 Oct, Waikoloa, HI, USA, 2020; Waikoloa, HI, USA, 2020; pp 6325-6328.
45. Garcia-Garin, O.; Monleón-Getino, T.; López-Brosa, P.; Borrell, A.; Aguilar, A.; Borja-Robalino,
R.; Cardona, L.; Vighi, M., Automatic detection and quantification of floating marine macro-litter
in aerial images: Introducing a novel deep learning approach connected to a web application in R.
Environmental Pollution 2021, 273, 116490(1-11), doi:10.1016/j.envpol.2021.116490
46. Freitas, S.; Silva, H.; Silva, E., Remote hyperspectral imaging acquisition and characterization for
marine litter detection. Remote Sensing 2021, 13, (13), 2536, doi:10.3390/rs13132536
47. Gonçalves, G.; Andriolo, U.; Pinto, L.; Bessa, F., Mapping marine litter using UAS on a beach-
dune system: a multidisciplinary approach. Science of The Total Environment 2020, 706, 135742,
doi:10.1016/j.scitotenv.2019.135742
48. Martin, C.; Parkes, S.; Zhang, Q.; Zhang, X.; McCabe, M. F.; Duarte, C. M., Use of unmanned
aerial vehicles for efficient beach litter monitoring. Marine Pollution Bulletin 2018, 131, 662-673,
doi:10.1016/j.marpolbul.2018.04.045
49. Kako, S. i.; Isobe, A.; Magome, S., Low altitude remote-sensing method to monitor marine and
beach litter of various colors using a balloon equipped with a digital camera. Marine Pollution Bulle-
tin 2012, 64, (6), 1156-1162, doi:10.1016/j.marpolbul.2012.03.024
50. Hanke, G.; González-Fernández, D., Longterm deployment of the JRC Sealittercam on the Western
Mediterranean Sea. In Policy-oriented marine Environmental Research for the Southern European
(PERSEUS) 2nd Scientific Workshop – Marrakesh 2014. Book of Abstracts, Giannoudi, L.; Streftaris,
N.; Papathanassiou, E., Eds. 2014; p 49.
// 44

References
51. Suaria, G.; Perold, V.; Lee, J. R.; Lebouard, F.; Aliani, S.; Ryan, P. G., Floating macro- and
microplastics around the Southern Ocean: Results from the Antarctic Circumnavigation Expedi-
tion. Environment International 2020, 136, 105494, doi:10.1016/j.envint.2020.105494
52. Guffogg, J. A.; Soto-Berelov, M.; Jones, S. D.; Bellman, C. J.; Lavers, J. L.; Skidmore, A. K.,
Towards the spectral mapping of plastic debris on beaches. Remote Sensing 2021, 13, (9), 1850,
doi:10.3390/rs13091850
53. Goddijn-Murphy, L.; Dufaur, J., Proof of concept for a model of light reflectance of plastics
floating on natural waters. Marine Pollution Bulletin 2018, 135, 1145-1157, doi:10.1016/j.marpol-
bul.2018.08.044
54. van Lieshout, C.; van Oeveren, K.; van Emmerik, T.; Postma, E., Automated river plastic monitor-
ing using deep learning and cameras. Earth and Space Science 2020, 7, (8), e2019EA000960(1-14),
doi:10.1029/2019EA000960
55. Kataoka, T.; Hinata, H.; Kako, S. i., A new technique for detecting colored macro plastic debris on
beaches using webcam images and CIELUV. Marine Pollution Bulletin 2012, 64, (9), 1829-1836,
doi:10.1016/j.marpolbul.2012.06.006
56. Corbau, C.; Buoninsegni, J.; Olivo, E.; Vaccaro, C.; Nardin, W.; Simeoni, U., Understanding through
drone image analysis the interactions between geomorphology, vegetation and marine debris along a
sandy spit. Marine Pollution Bulletin 2023, 187, 114515, doi:10.1016/j.marpolbul.2022.114515
57. Abreo, N. A. S.; Aurelio, R. M.; Kobayashi, V. B.; Thompson, K. F., ‘Eye in the sky’: Off-the-shelf
unmanned aerial vehicle (UAV) highlights exposure of marine turtles to floating litter (FML)
in nearshore waters of Mayo Bay, Philippines. Marine Pollution Bulletin 2023, 186, 114489,
doi:10.1016/j.marpolbul.2022.114489
58. Papageorgiou, D.; Topouzelis, K.; Suaria, G.; Aliani, S.; Corradi, P., Sentinel-2 detection of floating
marine litter targets with partial spectral unmixing and spectral comparison with other floating
materials (Plastic Litter Project 2021). Remote Sensing 2022, 14, (23), 5997, doi:10.3390/rs14235997
59. Matthews, J. P.; Ostrovsky, L.; Yoshikawa, Y.; Komori, S.; Tamura, H., Dynamics and early
post-tsunami evolution of floating marine debris near Fukushima Daiichi. Nature Geoscience 2017,
10, 598, doi:10.1038/ngeo2975
60. Lebreton, L.; Slat, B.; Ferrari, F.; Sainte-Rose, B.; Aitken, J.; Marthouse, R.; Hajbane, S.; Cunsolo,
S.; Schwarz, A.; Levivier, A.; Noble, K.; Debeljak, P.; Maral, H.; Schoeneich-Argent, R.; Brambini,
R.; Reisser, J., Evidence that the Great Pacific Garbage Patch is rapidly accumulating plastic. Scien-
tific Reports 2018, 8, (1), 4666, doi:10.1038/s41598-018-22939-w
61. Ge, Z.; Shi, H.; Mei, X.; Dai, Z.; Li, D., Semi-automatic recognition of marine debris on beaches.
Scientific Reports 2016, 6, 25759, doi:10.1038/srep25759
62. Simpson, M. D.; Marino, A.; de Maagt, P.; Gandini, E.; Hunter, P.; Spyrakos, E.; Tyler, A.; Telfer,
T., Monitoring of plastic islands in river environment using Sentinel-1 SAR data. Remote Sensing
2022, 14, (18), 4473, doi:10.3390/rs14184473
63. Davaasuren, N.; Marino, A.; Boardman, C.; Alparone, M.; Nunziata, F.; Ackermann, N.; Hajnsek, I.
In Detecting microplastics pollution in world oceans using SAR remote sensing, IGARSS 2018 – 2018
IEEE International Geoscience and Remote Sensing Symposium, 22-27 July, Valencia, Spain, 22-27
July 2018, 2018; Valencia, Spain, 2018; pp 938-941.
// 45

References
64. Evans, M. C.; Ruf, C. S., Toward the detection and imaging of ocean microplastics with a
spaceborne radar. IEEE Transactions on Geoscience and Remote Sensing 2021, 1-9, doi:10.1109/
TGRS.2021.3081691
65. NASEM Reckoning with the U.S. role in global ocean plastic waste; The National Academies of
Sciences, Engineering, and Medicine: Washington DC, USA, 2022; p 269.
66. Topouzelis, K.; Papageorgiou, D.; Suaria, G.; Aliani, S., Floating marine litter detection algorithms
and techniques using optical remote sensing data: A review. Marine Pollution Bulletin 2021, 170,
112675, doi:10.1016/j.marpolbul.2021.112675
67. Wolf, M.; van den Berg, K.; Garaba, S. P.; Gnann, N.; Sattler, K.; Stahl, F.; Zielinski, O., Machine
learning for aquatic plastic litter detection, classification and quantification (APLASTIC-Q). Envi-
ronmental Research Letters 2020, 15, (11), 114042, doi:10.1088/1748-9326/abbd01
68. Kylili, K.; Kyriakides, I.; Artusi, A.; Hadjistassou, C., Identifying floating plastic marine debris
using a deep learning approach. Environmental Science and Pollution Research 2019, doi:10.1007/
s11356-019-05148-4
69. Panwar, H.; Gupta, P. K.; Siddiqui, M. K.; Morales-Menendez, R.; Bhardwaj, P.; Sharma, S.; Sarker,
I. H., AquaVision: Automating the detection of waste in water bodies using deep transfer learn-
ing. Case Studies in Chemical and Environmental Engineering 2020, 2, 100026(1-5), doi:10.1016/j.
cscee.2020.100026
70. Martin, C.; Zhang, Q.; Zhai, D.; Zhang, X.; Duarte, C. M., Enabling a large-scale assessment of
litter along Saudi Arabian red sea shores by combining drones and machine learning. Environmental
Pollution 2021, 277, 116730(1-10), doi:10.1016/j.envpol.2021.116730
71. Zhang, P.; Du, P.; Guo, S.; Zhang, W.; Tang, P.; Chen, J.; Zheng, H., A novel index for robust and
large-scale mapping of plastic greenhouse from Sentinel-2 images. Remote Sensing of Environment
2022, 276, 113042(1-24), doi:10.1016/j.rse.2022.113042
72. Biermann, L.; Clewley, D.; Martinez-Vicente, V.; Topouzelis, K., Finding plastic patches in coastal
waters using optical satellite data. Scientific Reports 2020, 10, (1), 5364, doi:10.1038/s41598-020-
62298-z
73. Kühn, F.; Oppermann, K.; Hörig, B., Hydrocarbon Index – an algorithm for hyperspectral detec-
tion of hydrocarbons. International Journal of Remote Sensing 2004, 25, (12), 2467-2473, doi:10.108
0/01431160310001642287
74. Hörig, B.; Kühn, F.; Oschütz, F.; Lehmann, F., HyMap hyperspectral remote sensing to
detect hydrocarbons. International Journal of Remote Sensing 2001, 22, (8), 1413-1422,
doi:10.1080/01431160120909
75. Guo, X.; Li, P., Mapping plastic materials in an urban area: Development of the normalized dif-
ference plastic index using WorldView-3 superspectral data. ISPRS Journal of Photogrammetry and
Remote Sensing 2020, 169, 214-226, doi:10.1016/j.isprsjprs.2020.09.009
76. Yang, D.; Chen, J.; Zhou, Y.; Chen, X.; Chen, X.; Cao, X., Mapping plastic greenhouse with
medium spatial resolution satellite data: Development of a new spectral index. ISPRS Journal of
Photogrammetry and Remote Sensing 2017, 128, 47-60, doi:10.1016/j.isprsjprs.2017.03.002
// 46

References
77. Asadzadeh, S.; de Souza Filho, C. R., Investigating the capability of WorldView-3 superspectral data
for direct hydrocarbon detection. Remote Sensing of Environment 2016, 173, 162-173, doi:10.1016/j.
rse.2015.11.030
78. Mace, T. H., At-sea detection of marine debris: Overview of technologies, processes, issues, and
options. Marine Pollution Bulletin 2012, 65, (1–3), 23-27, doi:10.1016/j.marpolbul.2011.08.042
79. Maximenko, N.; Hafner, J. SCUD: Surface currents from diagnostic model. IPRC Technical Note No. 5;
International Pacific Research Center,: Hawaii, USA, 2010; p 17.
80. Cortesi, I.; Masiero, A.; De Giglio, M.; Tucci, G.; Dubbini, M., Random forest-based river plastic
detection with a handheld multispectral camera. The International Archives of the Photogrammetry,
Remote Sensing and Spatial Information Sciences 2021, XLIII-B1-2021, 9-14, 10.5194/isprs-ar-
chives-XLIII-B1-2021-9-2021
81. Iordache, M.-D.; De Keukelaere, L.; Moelans, R.; Landuyt, L.; Moshtaghi, M.; Corradi, P.; Knaeps,
E., Targeting plastics: Machine learning applied to litter detection in aerial multispectral images.
Remote Sensing 2022, 14, (22), 5820, doi:10.3390/rs14225820
82. Pinto, L.; Andriolo, U.; Gonçalves, G., Detecting stranded macro-litter categories on drone
orthophoto by a multi-class Neural Network. Marine Pollution Bulletin 2021, 169, 112594(1-9),
doi:10.1016/j.marpolbul.2021.112594
83. Freitas, S.; Silva, H.; Silva, E., Hyperspectral imaging zero-shot learning for remote marine litter
detection and classification. Remote Sensing 2022, 14, (21), 5516, doi:10.3390/rs14215516
84. Acuña-Ruz, T.; Uribe, D.; Taylor, R.; Amézquita, L.; Guzmán, M. C.; Merrill, J.; Martínez, P.;
Voisin, L.; Mattar B., C., Anthropogenic marine debris over beaches: Spectral characterization
for remote sensing applications. Remote Sensing of Environment 2018, 217, 309-322, doi:10.1016/j.
rse.2018.08.008
85. Lavender, S., Detection of waste plastics in the environment: Application of Copernicus earth obser-
vation data. Remote Sensing 2022, 14, (19), 4772, doi:10.3390/rs14194772
86. Balsi, M.; Esposito, S.; Moroni, M. In Hyperspectral characterization of marine plastic litters, 2018
IEEE International Workshop on Metrology for the Sea; Learning to Measure Sea Health Parame-
ters (MetroSea), Bari, Italy, 8-10 October 2018, 2018; IEEE: Bari, Italy, 2018; pp 28-32.
87. Masoumi, H.; Safavi, S. M.; Khani, Z., Identification and classification of plastic resins using near
infrared reflectance spectroscopy. World Acad. Sci. Eng. Technol. 2012, 6, (5), 141-148.
88. Moshtaghi, M.; Knaeps, E.; Sterckx, S.; Garaba, S.; Meire, D., Spectral reflectance of marine mac-
roplastics in the VNIR and SWIR measured in a controlled environment. Scientific Reports 2021,
11, (1), 5436, doi:10.1038/s41598-021-84867-6
89. Taggio, N.; Aiello, A.; Ceriola, G.; Kremezi, M.; Kristollari, V.; Kolokoussis, P.; Karathanassi,
V.; Barbone, E., A combination of machine learning algorithms for marine plastic litter detec-
tion exploiting hyperspectral PRISMA data. Remote Sensing 2022, 14, (15), 3606, doi:10.3390/
rs14153606
90. Gnann, N.; Baschek, B.; Ternes, T. A., Close-range remote sensing-based detection and identifica-
tion of macroplastics on water assisted by artificial intelligence: A review. Water Research 2022, 222,
118902(1-15), doi:10.1016/j.watres.2022.118902
// 47

References
91. Kataoka, T.; Murray, C. C.; Isobe, A., Quantification of marine macro-debris abundance around
Vancouver Island, Canada, based on archived aerial photographs processed by projective transfor-
mation. Marine Pollution Bulletin 2018, 132, 44-51, doi:10.1016/j.marpolbul.2017.08.060
92. Gonçalves, G.; Andriolo, U.; Pinto, L.; Duarte, D., Mapping marine litter with Unmanned Aerial
Systems: A showcase comparison among manual image screening and machine learning tech-
niques. Marine Pollution Bulletin 2020, 155, 111158(1-11), doi:10.1016/j.marpolbul.2020.111158
93. Fallati, L.; Polidori, A.; Salvatore, C.; Saponari, L.; Savini, A.; Galli, P., Anthropogenic marine
debris assessment with unmanned aerial vehicle imagery and deep learning: A case study along
the beaches of the Republic of Maldives. Science of The Total Environment 2019, 693, 133581,
doi:10.1016/j.scitotenv.2019.133581
94. Kako, S. i.; Isobe, A.; Kataoka, T.; Yufu, K.; Sugizono, S.; Plybon, C.; Murphy, T. A., Sequential
webcam monitoring and modeling of marine debris abundance. Marine Pollution Bulletin 2018,
132, 33-43, doi:10.1016/j.marpolbul.2018.04.075
95. Merlino, S.; Paterni, M.; Berton, A.; Massetti, L., Unmanned aerial vehicles for debris survey in coastal
areas: Long-term monitoring programme to study spatial and temporal accumulation of the dynamics
of beached marine litter. Remote Sensing 2020, 12, (8), 1260 (1 - 23), doi:10.3390/rs12081260
96. OceanScan Ocean plastic from space. https://www.oceanscan.org/ (Last access: 23 September
2023).
97. Bergmann, M.; Tekman, M. B.; Gutow, L., Sea change for plastic pollution. Nature 2017, 544,
(7650), 297-297, doi:10.1038/544297a
98. Tekman, M. B.; Gutow, L.; Peter, C.; Bergmann, M., LITTERBASE – Online Portal for Marine
Litter & Microplastics and their Implications for Marine Life. In 6th International Marine Debris
Conference, San Diego, USA, 2018.
99. Kruse, C.; Boyda, E.; Chen, S.; Karra, K.; Bou-Nahra, T.; Hammer, D.; Mathis, J.; Maddalene, T.;
Jambeck, J.; Laurier, F., Satellite monitoring of terrestrial plastic waste. ArXiv e-prints 2022, 1-14,
doi:10.48550/arXuc.2204.01485
100. Thomas, L.; Shah, A.; Ramasubramanian, M.; Priftis, G.; Maskey, M.; Ramachandran, R., Detect-
ing floating marine debris from commercial small satellite imagery with deep learning. In AGU Fall
Meeting, AGU: New Orleans, LA, USA. 13-17 December 2021, 2021; Vol. 2021, pp IN23A-04.
101. ASEAN Secretariat ASEAN Regional Action Plan for Combating Marine Debris in the ASEAN Mem-
ber States (2021-2025); Community Relations Division (CRD), The Association of Southeast Asian
Nations (ASEAN): Jakarta, Indonesia, 2021; p 54.
102. WBG Vietnam: Plastic pollution diagnostics; Report No. 127695; World Bank Group: Washington
D.C., USA, 2022; p 95.
103. Slabakova, V.; Zlateva, I.; Slavova, K., Initial assessment of composition, abundance, spatial
distribution and hotspots identification of floating macro-litter in the Bulgarian Black Sea waters.
EnviroRISKs 2020, 537-547, doi:10.48365/envr-2020.1.49
104. Campanale, C.; Suaria, G.; Bagnuolo, G.; Baini, M.; Galli, M.; de Rysky, E.; Ballini, M.; Aliani,
S.; Fossi, M. C.; Uricchio, V. F., Visual observations of floating macro litter around Italy (Mediter-
ranean Sea). Mediterranean Marine Science 2019, 20, (2), 271-281, doi:10.12681/mms.19054
// 48

References
105. Vighi, M.; Ruiz-Orejón, L. F.; Hanke, G. Monitoring of Floating Marine Macro Litter – State of the
art and literature overview; Publications Office of the European Union, Luxembourg, 2022; p 69.
106. Livens, S.; Knaeps, E.; Benhadj, I.; Bomans, B.; Dries, J., Feasibility assessment of a dedicated
satellite mission for monitoring marine macroplastics. In 4S Symposium, 16-20 May, Vilamoura,
Portugal, 2022; pp 1-8.
107. Martinez-Vicente, V., The need for a dedicated marine plastic litter satellite mission. Nature
Reviews Earth & Environment 2022, 3, (11), 728-729, doi:10.1038/s43017-022-00360-2
108. GESAMP Guidelines or the monitoring and assessment of plastic litter and microplastics in the ocean.
(IMO/FAO/UNESCO-IOC/UNIDO/WMO/IAEA/UN/UNEP/UNDP/ISA Joint Group of Experts on
the Scientific Aspects of Marine Environmental Protection). International Maritime Organization –
London, UK, 2019; p 130.
109. Veettil, B. K.; Hong Quan, N.; Hauser, L. T.; Doan Van, D.; Quang, N. X., Coastal and marine
plastic litter monitoring using remote sensing: A review. Estuarine, Coastal and Shelf Science 2022,
279, 108160, doi:10.1016/j.ecss.2022.108160
110. Andriolo, U.; Garcia-Garin, O.; Vighi, M.; Borrell, A.; Gonçalves, G., Beached and floating litter
surveys by unmanned aerial vehicles: operational analogies and differences. Remote Sensing 2022,
14, (6), 1336, doi:10.3390/rs14061336
111. Janssens, N.; Schreyers, L.; Biermann, L.; van der Ploeg, M.; Bui, T.-K. L.; van Emmerik, T., Riv-
ers running green: water hyacinth invasion monitored from space. Environmental Research Letters
2022, 17, (4), 044069, doi:10.1088/1748-9326/ac52ca
112. Gonçalves, G.; Andriolo, U., Operational use of multispectral images for macro-litter mapping
and categorization by unmanned aerial vehicle. Marine Pollution Bulletin 2022, 176, 113431(1-11),
doi:10.1016/j.marpolbul.2022.113431
113. Kikaki, K.; Kakogeorgiou, I.; Mikeli, P.; Raitsos, D. E.; Karantzalos, K., MARIDA: A bench-
mark for Marine Debris detection from Sentinel-2 remote sensing data. PLoS ONE 2022, 17, (1),
e0262247(1-20), doi:10.1371/journal.pone.0262247
114. Kremezi, M.; Kristollari, V.; Karathanassi, V.; Topouzelis, K.; Kolokoussis, P.; Taggio, N.; Aiello,
A.; Ceriola, G.; Barbone, E.; Corradi, P., Increasing the Sentinel-2 potential for marine plastic litter
monitoring through image fusion techniques. Marine Pollution Bulletin 2022, 182, 113974(1-20),
doi:10.1016/j.marpolbul.2022.113974
115. Savastano, S.; Cester, I.; Perpinyà, M.; Romero, L. In A first approach to the automatic detection of
marine litter in SAR images using artificial intelligence, 2021 IEEE International Geoscience and
Remote Sensing Symposium IGARSS, Brussels, Belgium, 11-16 July 2021, 2021; Brussels, Bel-
gium, 2021; pp 8704-8707.
116. Andriolo, U.; Gonçalves, G.; Sobral, P.; Bessa, F., Spatial and size distribution of macro-litter on
coastal dunes from drone images: A case study on the Atlantic coast. Marine Pollution Bulletin
2021, 169, 112490, doi:10.1016/j.marpolbul.2021.112490
117. Balsi, M.; Moroni, M.; Chiarabini, V.; Tanda, G., High-resolution aerial detection of marine plas-
tic litter by hyperspectral sensing. Remote Sensing 2021, 13, (8), 1557, doi:10.3390/rs13081557
// 49

References
118. Basu, B.; Sannigrahi, S.; Sarkar Basu, A.; Pilla, F., Development of novel classification algorithms
for detection of floating plastic debris in coastal waterbodies using multispectral Sentinel-2 remote
sensing imagery. Remote Sensing 2021, 13, (8), 1598, doi:10.3390/rs13081598
119. Ciappa, A. C., Marine plastic litter detection offshore Hawai’i by Sentinel-2. Marine Pollution
Bulletin 2021, 168, 112457, doi:10.1016/j.marpolbul.2021.112457
120. Merlino, S.; Paterni, M.; Locritani, M.; Andriolo, U.; Gonçalves, G.; Massetti, L., Citizen science
for marine litter detection and classification on unmanned aerial vehicle images. Water 2021, 13,
(23), 3349, doi:10.3390/w13233349
121. Andriolo, U.; Gonçalves, G.; Sobral, P.; Fontán-Bouzas, Á.; Bessa, F., Beach-dune morphodynam-
ics and marine macro-litter abundance: An integrated approach with Unmanned Aerial System.
Science of The Total Environment 2020, 749, 141474, doi:10.1016/j.scitotenv.2020.141474
122. Andriolo, U.; Gonçalves, G.; Bessa, F.; Sobral, P., Mapping marine litter on coastal dunes with
unmanned aerial systems: A showcase on the Atlantic Coast. Science of The Total Environment
2020, 736, 139632, doi:10.1016/j.scitotenv.2020.139632
123. Garcia-Garin, O.; Aguilar, A.; Borrell, A.; Gozalbes, P.; Lobo, A.; Penadés-Suay, J.; Raga, J. A.;
Revuelta, O.; Serrano, M.; Vighi, M., Who’s better at spotting? A comparison between aerial pho-
tography and observer-based methods to monitor floating marine litter and marine mega-fauna.
Environmental Pollution 2020, 258, 113680, doi:10.1016/j.envpol.2019.113680
124. Jakovljevic, G.; Govedarica, M.; Alvarez-Taboada, F., A deep learning model for automatic
plastic mapping using unmanned aerial vehicle (UAV) data. Remote Sensing 2020, 12, (9), 1515,
doi:10.3390/rs12091515
125. Kako, S. i.; Morita, S.; Taneda, T., Estimation of plastic marine debris volumes on beaches using
unmanned aerial vehicles and image processing based on deep learning. Marine Pollution Bulletin
2020, 155, 111127, doi:10.1016/j.marpolbul.2020.111127
126. Papachristopoulou, I.; Filippides, A.; Fakiris, E.; Papatheodorou, G., Vessel-based photographic
assessment of beach litter in remote coasts. A wide scale application in Saronikos Gulf, Greece.
Marine Pollution Bulletin 2020, 110684, doi>10.1016/j.marpolbul.2019.110684
127. Ryan, P. G., Using photographs to record plastic in seabird nests. Marine Pollution Bulletin 2020,
156, 111262, doi:10.1016/j.marpolbul.2020.111262
128. Goddijn-Murphy, L.; Williamson, B., On thermal infrared remote sensing of plastic pollution in
natural waters. Remote Sensing 2019, 11, (18), 2159, doi:10.3390/rs11182159
129. Geraeds, M.; van Emmerik, T.; de Vries, R.; bin Ab Razak, M. S., Riverine plastic litter monitoring
using unmanned aerial vehicles (UAVs). Remote Sensing 2019, 11, (17), 2045, doi:10.3390/rs11172045
130. Valdenegro-Toro, M., Deep neural networks for marine debris detection in sonar images. PhD
Thesis Heirot-Watt University, Edinburgh, United Kingdom. ArXiv e-prints – CoRR 2019, 1-241,
doi:10.48550/arXiv.1905.05241
131. Watanabe, J.-I.; Shao, Y.; Miura, N., Underwater and airborne monitoring of marine ecosystems
and debris. Journal of Applied Remote Sensing 2019, 13, (4), 044509, doi:10.1117/1.JRS.13.044509
// 50

References
132. Bao, Z.; Sha, J.; Li, X.; Hanchiso, T.; Shifaw, E., Monitoring of beach litter by automatic inter-
pretation of unmanned aerial vehicle images using the segmentation threshold method. Marine
Pollution Bulletin 2018, 137, 388-398, doi:10.1016/j.marpolbul.2018.08.009
133. Deidun, A.; Gauci, A.; Lagorio, S.; Galgani, F., Optimising beached litter monitoring protocols
through aerial imagery. Marine Pollution Bulletin 2018, 131, 212-217, doi:10.1016/j.marpol-
bul.2018.04.033
134. Aoyama, T. In Extraction of marine debris in the Sea of Japan using satellite images, Proceedings of
SPIE Volume: 10778, Remote sensing of the open and coastal ocean and inland waters, Hono-
lulu, Hawaii, USA, 2018; Frouin, R. J.; Murakami, H., Eds. Honolulu, Hawaii, USA, 2018; pp
107780R-1--8.
135. Moy, K.; Neilson, B.; Chung, A.; Meadows, A.; Castrence, M.; Ambagis, S.; Davidson, K., Map-
ping coastal marine debris using aerial imagery and spatial analysis. Marine Pollution Bulletin
2018, 132, 52-59, doi:10.1016/j.marpolbul.2017.11.045
136. Serranti, S.; Palmieri, R.; Bonifazi, G.; Cózar, A., Characterization of microplastic litter from
oceans by an innovative approach based on hyperspectral imaging. Waste Management 2018, 76,
117-125, doi:10.1016/j.wasman.2018.03.003
137. Hengstmann, E.; Gräwe, D.; Tamminga, M.; Fischer, E. K., Marine litter abundance and distri-
bution on beaches on the Isle of Rügen considering the influence of exposition, morphology and
recreational activities. Marine Pollution Bulletin 2017, 115, (1), 297-306, doi:10.1016/j.marpol-
bul.2016.12.026
138. Kako, S. i.; Isobe, A.; Kataoka, T.; Hinata, H., A decadal prediction of the quantity of plastic
marine debris littered on beaches of the East Asian marginal seas. Marine Pollution Bulletin 2014,
81, (1), 174-184, doi:10.1016/j.marpolbul.2014.01.057
139. Garcia-Garin, O.; Borrell, A.; Aguilar, A.; Cardona, L.; Vighi, M., Floating marine macro-litter
in the North Western Mediterranean Sea: Results from a combined monitoring approach. Marine
Pollution Bulletin 2020, 159, 111467, doi:10.1016/j.marpolbul.2020.111467
140. Cortesi, I.; Masiero, A.; Tucci, G.; Topouzelis, K., UAV-based river plastic detection with a mul-
tispectral camera. The International Archives of the Photogrammetry, Remote Sensing and Spatial
Information Sciences 2022, XLIII-B3-2022, 855-861, doi:10.5194/isprs-archives-XLIII-B3-2022-
855-2022
141. Cocking, J.; Narayanaswamy, B. E.; Waluda, C. M.; Williamson, B. J., Aerial detection of beached
marine plastic using a novel, hyperspectral short-wave infrared (SWIR) camera. ICES Journal of
Marine Science 2022, 10.1093/icesjms/fsac006
142. Escobar-Sánchez, G.; Haseler, M.; Oppelt, N.; Schernewski, G., Efficiency of aerial drones for mac-
rolitter monitoring on Baltic Sea beaches. Frontiers in Environmental Science 2021, 8, 560237(1-
18), 10.3389/fenvs.2020.560237
143. Gonçalves, G.; Andriolo, U.; Gonçalves, L.; Sobral, P.; Bessa, F., Quantifying marine macro litter
abundance on a sandy beach using unmanned aerial systems and object-oriented machine learning
methods. Remote Sensing 2020, 12, (16), 2599, doi:10.3390/rs12162599
144. NASA EARTHDATA Open Access for Open Science: Tracking Ocean Plastics from Space.
Accessed on 24th April 2023. https://www.earthdata.nasa.gov/learn/articles/ocean-plastic
// 51

Annex

Annex
A1. Scientific Research Projects

Information presented here has been sourced from updates/ ). The living list is not exhaustive but
Related Research Topics section of the IOCCG provides an overview of ongoing and concluded
Task Force on Remote Sensing of Marine Litter scientific projects that have been supported by
and Debris (https://ioccg.org/rsmld-news-and- various funding entities.
Title: A
dvancing Remote Sensing of Microplastics on the Surface Ocean, SQOOP
(March 2021 – March 2024)
Contact Person: Heidi Dierssen ([email protected])
Funding Agency: NASA Ocean Biology and Biogeochemistry
Summary: Since the 1950’s, positively buoyant plastic objects have been accumulating at the surface of
the oceans, transported by currents, wind and waves. Small millimetre-sized pieces (<4.75 mm), known
as microplastics, count in trillions at global scale and pose an increasing risk to marine biota. Floating
microplastics concentrate along convergence zones in the five major ocean basins, but a comprehen-
sive analysis of the spatial and temporal distributions is lacking, and the monitoring tools are not well
developed to assess global distributions. Thus far, remote sensing methods have focused on larger mac-
roplastics. Our specific objectives are to: 1) Evaluate geospatial and temporal trends in existing ocean
colour products across hot spots that may be related to enhanced reflectance from plastics; 2) Propagate
estimates of ocean surface hyperspectral reflectance using simple mixed pixel models to the Top of the
Atmosphere (TOA) under different microplastic concentrations and atmospheric conditions; 3) Sim-
ulate spaceborne ocean colour remote sensing observations for different microplastic and atmospheric
conditions using robust vector radiative transfer models for coupled ocean-atmosphere systems; 4) Assess
microplastic remote sensing detectability using statistical information content assessment in terms of
current and future instrument characteristics, microplastic quantity and nature, and external conditions,
such as observation geometry and atmospheric state; and 5) Evaluate how results from the above analyses
relate to our hypotheses and implications for the remote sensing of microplastic and provide recommen-
dations for new algorithms and instrument design.
// 53

Annex
Title: A
rtificial Intelligence and drones supporting the detection and mapping
of floating aquatic plastic litter, AIDMAP (September 2020 – March 2022)
Contact Person: Els Knaeps ([email protected])
Website: https://remotesensing.vito.be/case/marine-plastic-litter
Funding Agency: ESA, the European Space Agency (Discovery Element of the Basic Activities –
Campaign on Remote Sensing of Plastic Marine Litter)
Summary: The AIDMAP project proposes an Artificial Intelligence (AI)-based approach for the detec-
tion of FML in accumulation zones. Vertical integration of small drone and satellite data will be evalu-
ated for the detection of the marine litter at different spatial scales. These can be complemented by High
Altitude Pseudo-Satellite (HAPS) in the longer term to come to a long-term sustainable solution. The
AIDMAP project therefore responds to the quickly evolving EO landscape with an increasing emphasis
on modern, affordable and sustainable technologies, such as Artificial Intelligence, and the launch of
(constellations of) small satellites and non-orbiting platforms (such as HAPS) while also exploring the
added value of the current Copernicus Sentinel program. Here, study areas in Vietnam are selected to
demonstrate the proposed approach.
Title: A
irborne & satellite observation strategies for marine litter monitoring, AIR-SOS
(June 2020 – Sep 2021)
Collecting Multispectral data from floating debris using a seaplane over the Elbe river discharge area to
validate current algorithms and methodologies.
Contact Person: Irina Rammos ([email protected])
Website: https://skyflox.eu/, https://www.pml.ac.uk/, https://aufwind.aero/
Summary: Marine Litter is a global issue and can be found in all the seas from the equator to the poles,
and in freshwater systems, such as rivers and lakes. Most of the marine litter is plastic and, as plastic
production continues to increase, greater impacts are expected. Plastic marine litter dramatically affects
marine life and ecosystems and has a great economic impact on coastal communities, tourism, and
fisheries. It furthermore poses a concern for human health due to contamination of seafood with plastic
particles and associated pollutants. Urgent questions around marine plastic pathways into the ocean,
sinks, trends, and fate remain open, but cannot be answered satisfactorily using ground-based and mod-
el-based systems alone. The emerging field of remote sensing for plastic detection is promising for tack-
ling unknowns around marine monitoring, but reliable in situ validation data are required to improve
and optimise algorithms and approaches. The AIR-SOS (AIRborne & Satellite Observation Strategies
for marine litter monitoring) study aims to do just that, by collecting high-quality and high-resolution
data of floating objects in coastal waters near the mouth of the river Elbe. A seaplane will be used on
clear and still (low wind) days to collect data coincident with Copernicus Sentinel-2 satellites overpass.
In this way, the project will assess and demonstrate the value of the aircraft as a platform for validation
of Sentinel-2 validation. The ability to fly sensors on General aviation Aircraft at lower cost, at lower alti-
tudes (visual cross-checks) and the possibility to perform in situ measurements (sea-landings) makes this
a multi-functional ‘platform’ suitable to for systematic validation of satellite remote sensing detection of
marine litter.
// 54

Annex
Title: A
ssessment of the Effects of Marine Debris on Ocean Colour Signals
(2021 – 2024)
Contact Person: Robert Foster ([email protected])
Summary: The objective of this research is to conduct an investigation into the effects of marine debris
upon top of atmosphere (TOA) ocean colour signals. Since so little is currently known about the optics
of both floating and suspended marine debris, primarily marine plastic, we will first conduct theoretical
studies to examine how different types of debris affect changes in the TOA radiance. We will determine
the limits of detectability of debris from orbit for current and planned satellite ocean colour sensors,
followed by an analysis of current and historical remote sensing data from orbiting sensors such as the
Hyperspectral Imager for the Coastal Ocean (HICO) and the DESIS Earth Sensing Imaging Spectrom-
eter. This study will seek to address the following science questions: 1) In what ways does the presence
of macro- and micro- marine debris affect the top-of atmosphere ocean color signal?; 2) What are the
detectability limits for debris with current and planned satellite ocean colour missions?; 3) How do
percent coverage, debris reflectance, and degree of submersion affect the detectability in the open ocean?;
4) Does marine debris, particularly plastics, alter the polarization of the upwelling radiance?; 5) Do
existing remote sensing datasets support the conclusions?
Title: B
rillouin – backscatter – fluorescence LIDAR research
for Underwater Exploration of marine litter, BLUE
(September 2020 – August 2022)
Contact Person: Valentina Raimondi, CNR-IFAC – Italy ([email protected])
Website: http://www.ifac.cnr.it/
Summary: Our idea is to investigate the potential of LIDAR – space, airborne and ground based – to
address plastic marine litter. Recent studies have stressed how plastic litter at the sea surface represents
only a small fraction of plastics entering the sea. Hence, the major contribution of our idea would be to
investigate remote sensing methods with the potential to provide information on plastics distributed in
the water column and its identification. Until now, the contribution of LIDAR to ocean plastic remote
sensing has been almost unexplored, except for sporadic bathymetric data from airplane to detect large
items. Meanwhile, spaceborne elastic LIDAR has already been used to detect algal blooms in oceanic
waters, while fluorescence LIDAR has already been suggested for plastics characterisation in different
contexts. The proposed approach is four-fold and aims at investigating the potential of: (1) elastic backs-
catter and Brillouin LIDAR from space to detect changes in the optical properties of the water column
due to microplastics; (2) LIDAR bathymetric data from airplane to detect plastic items by processing
airborne LIDAR data acquired over the Great Pacific Garbage Patch; (3) fluorescence LIDAR to identify
plastic items from airborne, ship- or ground-based platform (e.g. at river outlets); (4) Raman spectros-
copy to identify plastic items and microplastic as for the material type. For the first time, this study will
provide an insight on the feasibility of using LIDAR for remote sensing of microplastics and for the char-
acterisation of plastic items in terms of material type.
// 55

Annex
Title: U
sing Deep Learning Methods For Plastic Litter Detection From Satellite Remote Sensor,
DL4PlasticLitter
(June 2020 – April 2021)
Contact Person: Delphine Nobileau ([email protected]) – CAPGEMINI
Summary: Artificial Intelligence (AI) could represent a powerful tool in support of the EO data process-
ing for the detection of marine litter, provided the availability of a sufficiently large dataset of satellite
images of marine litter accumulations. However, there are currently no such datasets publicly available
yet. To address this issue, in the frame of the DL4PlasticLitter project, we first create realistic synthetic
spectra of floating litter accumulations. We simulate combinations of reflectance spectra of seawater and
macro-plastic for different observation geometries, different concentrations of chlorophyll and other
substances present in the water, and we model the radiative transfer to the top of atmosphere and at sen-
sor resolution. Then, we train AI models to learn to differentiate the spectra of modelled accumulations
containing plastic. The last step consists in validating the developed AI models with real satellite images
of marine litter accumulations.
Title: D
etection of Ocean Litter Plastics with Hyper-to-multispectral Infrared Neural Networks,
DOLPHINN
(July 2020 – July 2021)
Contact Person: Yolanda Brown ([email protected])
Website: https://mda.space/en/
Summary: This study seeks to determine the feasibility of using AI to better and more accurately detect
and quantify ocean and beach plastics litter from space-borne multispectral data, particularly in the
SWIR range. MDA is developing a novel spectral fusion approach that learns to associate multispectral
data with full hyperspectral features so that single multispectral images can be used to detect plastics
more reliably. Multispectral sensors already on orbit could then better contribute to marine litter detec-
tion. Once hyperspectral space assets such as CHIME become available, they could be used for contin-
uous training of multispectral sensors to allow more coverage and revisit in ocean plastic detection and
monitoring.
// 56

Annex
Title: S pectrometer for Marine Litter, ESAPlastics

(January 2020 – May 2021)
Contact Person: Hugo Silva ([email protected])
Website: https://www.aircentre.org/esaplastics/
Funding Agency: ESA, the European Space Agency (GSTP)
Summary: The objective of the project is to study, characterize, acquire, and process data from oceanic
marine litter samples using heterogeneous sensors information. The project work is divided into three
main parts: (i) the in-situ acquisition of marine litter samples from an oceanic marine litter hotspot in
Faial Island Azores; (ii) the characterization and identification in laboratory environment of the indi-
vidual components that are present in the collected marine litter samples i.e., type of material and other
chemical elements contained, by using different sensors, e.g. Spectroscopy FTIR, Raman, and LIBS;
(iii) performing extensive dataset campaigns using manned and unmanned aerial platforms, for acquir-
ing remote hyperspectral imaging data of artificial marine litter concentrations, fostering the develop-
ment of automatic methods based on supervised learning approaches for the detection using spatial/
spectral information of marine litter concentrations from space.
Title: T ackling the plastic debris challenge at its source – Linking EO data with multi-source
in-situ data for modelling debris pathways from source to sink, From Source to Sink
(August 2020 – July 2022)
Contact Person: Jonas Franke, Remote Sensing Solutions ([email protected])
Website: http://www.remote-sensing-solutions.com/
Summary: Monitoring areas closer to plastic marine litter sources such as rivers and estuarine sys-
tems has the potential to improve mitigation strategies. Upscaling in-situ litter point data with earth
observation (EO) and hydrodynamic models is our central concept. Multi-type in situ data will be
collected at various points along the pollution pathway (in our demonstration site in the Po River delta
in Italy): Imagery from installed cameras on bridges is analysed to detect floating plastic in rivers using
deep-learning (in-situ type 1). Water samples from estuaries and coastal areas using manta trawls are
used to quantify plastic litter abundances (in-situ type 2) Drone imagery along the shoreline is acquired
for accumulation analyses (in-situ type 3) Beach samples through field surveys (in situ type 4) Sentinel-2
and -3, together with VHR data such as WorldView-3, are used to monitor discharging rivers and their
estuaries (water constituents and river plume detection). Integration of these situ-data, multi-scale EO
and hydrodynamic modelling serves as the development basis, allowing for the first time a monitoring
of real-world debris transport pathways. Such source-to-sink monitoring systems can be used to identify
environmental, economic, human health and safety-related impacts of plastic litter and would support
targeted efforts of both off- and onshore-based clean-up projects.
// 57

Annex
Title: G
lobal Monitoring of Microplastics using GNSS-Reflectometry, GLIMPS
(November 2020 – December 2021)
Contact Person: Dr. Clarizia ([email protected])
Website: https://elecnor-deimos.com/project/glimps/
Funding Agency: ESA, the European Space Agency (Discovery Element of the Basic Activities)
Summary: The goal of GLIMPS is to produce global maps of the microplastics concentration in the
oceans using GNSS-R data and algorithms based on machine learning. This will deliver information
about the location and distribution of microplastics, which will be complementary to that provided
by in situ measurements and ocean circulation models. The idea is built upon the assumption that
microplastics and associated surfactants dampen the waves, reducing ocean surface roughness; and that
this reduction in roughness can be sensed by satellite radar. GNSS-Reflectometry is a radar-based remote
sensing technique which only requires cheap, lightweight and low-power receivers to be implemented,
since it exploits existing GNSS transmitters of opportunity. The wealth of existing GNSS transmitters,
and the nature of GNSS-R receivers, makes it easy to build a constellation, addressing the need for high
space-time sampling that is crucial for monitoring microplastics from space efficiently.
Title: H
yperdrone: Development of spectroradiometric proxies of shoreline marine plastic debris
for satellite validation using remotely piloted aircrafts, HyperDrone
(June 2020 – December 2021)
Contact Person: Aser Mata ([email protected])
Summary: Developing instruments and algorithms for satellite remote sensing of ocean plastic needs
standardised global in situ observations. This project plans to collect hyperspectral data from plastic
targets to develop a standardised indicator for in situ radiometric detection of plastic debris on the shore-
line, with a view to being deployed globally on different platforms. Field campaigns will be carried out
that will include hand-held hyperspectral spectrometers (SVC) and state of the art hyperspectral imagers
(BaySpec OCI-F and the Headwall Co-aligned VNIR+SWIR sensors) mounted on drone platforms
flying at different altitudes. Spectra from different plastic targets will be collected on the shoreline in
real conditions meeting traceability standards and with uncertainty estimates for each dataset (to be
made freely available upon completion of the project). Taking advantage of the SWIR spectral features
of plastic materials, HyperDrone aims to develop proxies for plastic detection on the shoreline and assess
subpixel detection. Using an atmospheric radiative transfer model, we will simulate at-satellite sensor
radiances to provide guidance on sensor requirements as well as model signal unmixing for retrieval of
plastic pixel coverage.
// 58

Annex
Title: P rediction of plastic hot-spots in coastal regions using satellite derived plastic detection,
cleaning data and numerical simulations in a coupled system, LOCATE
(June 2020 – May 2021)
Contact Person: Dr José M. Alsina ( [email protected]); Dr Silvia Huber ([email protected])
Website: https://lim.upc.edu/en; https://www.dhi-gras.com/
Summary: The LOCATE project concerns the identification of plastic hotspots in coastal waters and on
the shore by using a coupled system integrating satellite derived information, regional coastal models,
the computation of lagrangian plastic trajectories and information from cleaning campaigns. Satellite
derived hydrodynamic information (Sentinel-1 SAR and Sentinel-3 altimeter) is being used to validate
eulerian hydrodynamic simulations in coastal waters and will be also used to produce model inputs and
data assimilation (bathymetry, hydrodynamic variables). Moreover, Sentinel-2 optical information is
used to derive water quality (turbidity) which potentially can be correlated to plastic inputs, cleaning
data and numerical simulations. Eulerian hydrodynamic simulations are produced using the numerical
model COAWST and validated with Satellite hydrodynamic information (wave height, water surface
elevation). Daily hydrodynamic forecasting outputs produced at the Catalan coast in Spain are stored in
a dedicate web page at three different coastal grid domains (with grid sizes of 2500m, 350m and 70m).
Simulations of plastic dispersion in the nested domains are obtained using the Parcels software. Plastic
accumulation regions will be then identified and tracked in space and time and contrasted with cleaning
data. The developed system will answer a high demand of a more efficient local to regional management
of coastal plastic pollution by helping to identify hotspots of plastic accumulations in time and space.
The developed system will answer a high demand of a more efficient local to regional management of
coastal plastic pollution by helping to identify hotspots of plastic accumulations in time and space. The
forecasting system will be made publicly available.
Title: M
arine Litter Aggregation Forecast
(Sept 2020 – Dec 2021)
Contact Person: Mario Castro de Lera ([email protected])
Website: https://deepblueglobe.eu/
Summary: The main goal of this project is to provide a new estimation of long-term marine litter accu-
mulation areas at global scale taking advantage of numerical reanalysis databases of historical met-ocean
conditions. Different machine learning techniques are applied to generate long-term climate-based series
of those environmental variables that may affect the drift of marine litter over the sea surface such as cur-
rents, wind and wave-induced Stokes drift. The generated series will feed a state-of-the-art Lagrangian
model in order to simulate the global long-term evolution of marine litter transport through the ocean
surface. As marine debris sources coastal cities, river outputs and shipping routes are considered. Besides
the main goal of the project, the proposed methodology has also the potential for interesting secondary
achievements, such as providing the estimation of global scale marine litter distribution for specific past
dates or predicting the expected future location and distribution of marine litter patches up to approxi-
mately six months ahead.
// 59

Annex
Title: M
arine Macro Litter Drift Forecasting Service
(May 2018 – December 2020)
Contact Person: Anne Vallette ([email protected])
Website: https://argans.co.uk/proj-littertep.html
Funding Agency: Mercator Ocean (CMEMS)
Summary: The LITTER-TEP (Thematic Exploitation Platform) will provide a macro-litter beaching
forecast service for local authorities, government agents, NGOs and environmental protection agencies.
The fate of marine macro-litter can be forecast by modelling drift using Lagrangian models and bulk
fluxes using Eulerian models. These will be parameterised using CMEMS products forecasting wave,
wind and current data and supplemented with models of settlement, sinking and resuspension. Results
of these simulations will be posted on a dedicated LITTER-TEP web portal along with a geo-browser
to locate litter events associated with an area-of-interest (AOI), historical records and statistics. The
expected results are maps of likely stretches of coastline that will be affected by marine-litter grounding
subsequent to storm events. These will enable local agencies to forecast when and where clean-up opera-
tions will be needed.
Title: A
full-range plastic marine litter monitoring service to support cleaning and
littering reduction actions by mapping hotspots, pathways and littering sources, MARLISAT
(June 2020 – December 2021)
Contact Person: Marc Lucas ([email protected])
Website: https://www.cls-telemetry.com/
Summary: A combined satellite-based solution is proposed to map plastic marine litter pathways and
accumulation areas relying onto three innovative developments: 1) Utilising machine learning along-
side an augmented land cover classification, accumulations of plastic litter across coastal and riparian
zones will be mapped. 2) Based on the MAR-GE/T beacons with GPS positions relayed through the
Argos satellite system, the proposed novel satellite tracker will be specifically designed to track plastic
litter pathways, with better precision and realistic behavior. This miniaturized device will have a reduced
environmental footprint. 3) Ocean surface currents and winds play a primary role in the transport and
dispersal of plastic marine litters. A multi-year observations of ocean currents from space, in a syner-
getic use of satellite sensors (altimetry, scatterometers, …) and in-situ buoys, will be computed and used
within a Lagrangian drift modelling tool to simulate the plastic litters pathways and map the hotspots.
The combination of these three components will constitute a full-range monitoring system of plastic
marine litter, from littering sources determination to hotspots and pathways identification, thus improv-
ing marine litter collection efforts.
// 60

Annex
Title: A
Simulator for Marine Litter Observation from Space, ML-OPSI
(June 2020 – September 2021)
Contact Person: Theodora Papadopoulou ([email protected])
Website: https://argans.co.uk/
Summary: A breadboard for end-to-end (E2E) Marine Litter Optical Performance Simulations
(ML-OPSI) is being designed in the frame of the ESA Discovery Campaign to support Earth Observa-
tion scientists with the design of computational experiments for Operations Research. The ML-OPSI
breadboard will estimate Marine Litter signal at Top-Of-Atmosphere (TOA) from a set of Bot-
tom-Of-Atmosphere (BOA) scenarios representing the various case studies by the community (e.g.,
windrows, frontal areas, river mouths, sub-tropical gyres), coming from synthetic data or from real
observations. It is a modular, pluggable and extensible framework, promoting re-use and be adapted for
different missions, sensors and scenarios.
The breadboard consists of the OPSI components for the simulation and the Marine Litter model
components for the detection of marine litter. It shall consider the changes caused in the water reflec-
tance and properties due to marine litter, exploiting gathered information of plastic polymers, different
viewing geometries, and atmospheric conditions as naturally occurring.
Marine Litter scenarios of reference shall be built based on in-situ campaigns, to reflect the true littering
conditions at each case, both in spatial distribution and composition. The breadboard shall be validated
over artificial targets at sea in field campaigns as relevant.
Title: M
ulti-Functional Lidar Measurements to Identify and Characterize Marine Debris
using Time-Resolved Fluorescence
(2022 - 2023)
Contact Person: Madeline Cowell ([email protected])
Funding Agency: NASA ESTO Instrument Incubator Program (IIP) Instrument/Measurement Con-
cept Demo (ICD)
Summary: Our goal is to characterize the laser induced fluorescence (LIF) return of marine debris
both in the spectral and time domain. We will include measurements from naturally occurring targets,
such as phytoplankton, to demonstrate sufficient differentiation in aquatic scenes between biogenic and
anthropogenic material. Time-correlated single photon counting (TCSPC) will provide a measure of
fluorescence lifetime of the various targets. After having success measuring fluorescence spectra and
fluorescence lifetime independently, we propose to expand upon this research to feed into this study. The
laboratory measurements will consist of a tunable pulsed laser as the excitation source, a photon-sensitive
fast detector, and spectral filters tuned to the target’s peak emission wavelength. To ensure high prob-
ability of classification, machine learning algorithms will be developed and tested. The results of this
study will define the sensitivity of the fluorescence return for future performance modeling necessary for
developing an effective space-based lidar system. Ultimately, this will inform a mission architecture to
achieve global coverage for marine debris identification, characterization, and monitoring.
// 61

Annex
Title: M
ulti-Model synthetic S2-HS (hyperspectral) data for marine/plastic debris characterization,
MUSS2
(December 2020 – June 2022)
Contact Person: Jonathan Cheung-Wai Chan ([email protected])
Summary: For effective identification and tracking of marine plastic, the sensor data we want is hyper-
spectral (HS) images including the wavebands at SWIR (1000-2500 nm) with a high spatial resolution
around 0.5 m. These characteristics are non-existent in current EO missions. We propose to use novel
spectral and spatial enhancement method to generate simulated EO HS SWIR from Sentinel 2 MSI
using spectral response function modelling. For spatial enhancement, we apply spatial superresolution
through a CNN (Convolutional Neural Network) based 2 branch feature extraction model for capturing
detail feature in spatial and spectral domain for plastic debris identification.
Title: C haracterization of light polarization properties of virgin and marine-harvested plastic litter
toward remote-sensing mapping of ocean plastics, Ocean Plastics Polarization Properties OP³
(August 2020 – July 2022)
Contact Persons: Tristian Harmel ([email protected]);

Shungu Garaba ([email protected])
Summary: Measurements of polarization state of water-leaving light have been shown to be a signifi-
cant tool to disentangle complex aquatic light signal to retrieve water constituents. On another hand,
subsurface plastic marine litter (PML) might induce surfactants from “bio-fouling” production. In turn,
those surfactants will smooth away capillary waves which can be detectable through polarimetric remote
sensing. In the context of the future launch the satellite mission PACE (NASA) and 3MI-Sentinel-5
(ESA, EUMETSAT), embarking hyperspectral radiometers and polarimeters, we propose to fully char-
acterize the polarization signature of PML in relation to other natural seawater constituents through: (i)
laboratory experimentation, (ii) in-situ measurements, (iii) existing polarization data from older satellite
missions (e.g., PARASOL). For this characterization up-to-date polarimetric sensors will be exploited
as well as theoretical modeling of light propagation in PML contaminated waters (accumulation zones
in the open ocean and estuarine systems) focusing on the water-leaving polarization and surface sur-
factants/roughness. This is foreseen as an important complementary effort along those of the Copernicus
framework of operational satellite missions Sentinel 1 and 2 which have been shown to have potential
monitoring application for PML.
// 62

Annex
Title: O
cean Scan: Marine litter database from Earth and space, Ocean Scan
(July 2020 – October 2021)
Contact Person: Laia Romero ([email protected])
Website: https://www.oceanscan.org/
Summary: Satellite remote sensing has demonstrated great potential to become a breakthrough in the
mapping of marine litter. One limiting factor for its full development is the access to reliable, exten-
sive, and consistent ground truth of plastic and litter occurrence in aquatic environments. During the
past two decades, the amount of in-situ data and information about marine litter has greatly increased,
especially during the last five years, however this information is sparsely located in different databases
and often lacks the needed requirements or remote sensing technology research. Ocean Scan was
designed to address this problem, by becoming the first inclusive global labelled database to integrate
in-situ observations of marine plastic and litter with satellite data. In Ocean Scan, data will be presented
on a global interactive map, where users can access information about marine litter in-situ observations
and associated EO products. A web portal and a mobile application will provide user friendly access to
the platform for data upload, consultation and download. Designed to maximise interoperability and
scalability and to ensure a consistent data format and schema to fit the requirements of remote sensing
technologies, the database will be free of charge and open to everybody, upon user registration. By ensur-
ing data provenance and different levels of privacy for uploaded observations, Ocean Scan will provide
a unified reference point for marine plastic and litter observations to support and promote international
collaboration and research.
Title: P lastic waste and the Black Sea, monitoring litter at sea and on the land from Sentinel-2 data,
Plastic Detection: Black Sea Test/PDBS (May 2020– May 2021)
Contact Person: Noelia ABASCAL-ZORRILLA ([email protected])
Website: https://argans.co.uk/proj-blackseaplastics.html
Summary: An EO processor for the detection of marine debris was originally developed and validated
in different areas which were known for the presence of big patches of plastic. The current Argans Ltd.
detector, based on the analysis of marine debris spectral reflectance for already well-known indices, was
developed for fainter signals and proves to be a semi-robust litter and plastics detector. However, an
assessment of the densities and volumes of plastics that could be detected in comparison to big patches,
needed to be performed. The Black Sea, a semi-enclosed basin with numerous litter inflows by huge
watershed rivers and with only a spillway at the Bosporus, is an ideal test area for the further develop-
ment of the marine detector and the development of a land-litter detector. Therefore, the objective of the
project is to be able to effectively use Sentinel-2 to identify both floating rafts of marine litter and sites of
unconsolidated waste on land, providing information on both source and output areas. The detector is
tuned to the probabilities of detection and false alarms, fixed by the operator. A Bayesian approach com-
bined with an assessment of the diagnosis ability of the detector (represented by a ROC curve) allows
an adjustment of the detector’s thresholds according to the environmental, viewing conditions and the
a-priori knowledge of plastics presence delivered by a litter drift model deployed in the Black Sea.
// 63

Annex
Title: D
etecting riverine plastic conglomerations, fluxes and pathways in Indonesia, Plastic Monitor
(April 2021 – March 2022)
Contact Person: Marieke Eleveld ([email protected]) – Deltares
Website: https://www.deltares.nl/nl/
Summary: The general objective of Plastic Monitor is to assess detection of heavy plastic pollution loads
in rivers by satellite imaging and demonstrate how it can enhance the quantification and monitoring of
plastic input into the marine environment. Although none of the satellite mission concepts were spe-
cifically designed for the detection of plastic debris, there is potential for some sensors to be used in the
detection of plastics. Therefore, our idea is centred around using a multi-sensor method, where satellite
images from different sensors are analysed. This monitoring method is applied to rivers, to detect plastic
litter before it reaches the oceans. This will be achieved together with a new plastic capture system,
which first concentrates and then removes plastic floating in rivers and can, in this way, bring added
value for monitoring of plastic fluxes. In the analysis we will develop advanced data science techniques
to get information about the aquatic environment. The plastic detection capacity of existing sensors will
lead to recommendations to inspire ESA’s future mission design.
Title: D
etecting water hyacinth patches as a proxy for riverine plastic transport, Plastic Plants
(September 2020 – August 2023)
Contact Person: Louise Schreyers ([email protected])
Website: https://plasticmonitoring.com/
Summary: This project aims to develop and implement an algorithm to automatically detect floating
macroplastic accumulation in rivers using remote sensing. It combines the detection of floating water
hyacinths using mainly Sentinel-2 with in-situ estimates of macroplastic amounts carried by this invasive
aquatic weed. Water hyacinths typically form large patches of several meters of width and length, and
can thus be detected from space. Preliminary results show that they can aggregate as much as 80% of
floating plastic debris in tropical rivers. The detection tool and field measurements will focus on the
Saigon river, Vietnam, a river highly invaded by hyacinths. Our main scope is to quantify floating mac-
roplastic transport and accumulation for the Saigon river over several years, using water hyacinths as a
proxy. We will rely on robust in-situ data collected over one year, characterizing the share of macroplastic
entangled in hyacinths and its spatiotemporal variability. The algorithms and spectral libraries may serve
for future applications, notably for plastic monitoring in other tropical river systems. Given that tropical
rivers invaded by hyacinths typically overlap with the highest plastic polluted waterways, this detection
system could serve for global monitoring purposes.
// 64

Annex
Title: C an the microbial communities in the oceans help satellites to monitor micro-plastic pollution?,
PLASTICSURF
(October 2020 – October 2023)
Contact Person: Armando Marino ([email protected])
Website: https://www.stir.ac.uk/about/faculties/natural-sciences/our-research/research-groups/
earth-and-planetary-observation-research-group/
Summary: We propose to tackle the problem of monitoring plastic from another perspective. We aim
at observing the effects of plastic on the microbial environment and, as a consequence, on ocean surface
characteristics. This project brings together three pieces of research: a) Plastic in the ocean is heavily col-
onised by microbes; b) Microbes in water produce substances (surfactants) that dampen small waves; and
c) Synthetic Aperture Radar (SAR) can identify surfactants as dark areas or stripes in images. We dis-
covered that several ESA Sentinel-1 satellite images acquired over the garbage patches (Atlantic, Indian
and Pacific oceans) present the same dark features we associate with surfactants. We observed that such
features are not correlated with high chlorophyll-a and therefore microbes naturally occurring in the
ocean (i.e. phytoplankton). Our hypothesis is that these dark patches are the signature of micro-plastics.
Our experiments will show whether or not plastic-dwelling microbes can produce enough surfactants to
be visible from space. We will use plastic submerged in fish cages in Scotland and also lab experiments. A
ground radar and Sentinel-1 images will be used to check for surfactants around cages while the ground
radar will be used with the lab tanks.
Title: P lastic Flux for Innovation and Business Opportunities in Flanders, PLUXIN
(September 2020 – September 2023)
Contact Person: https://pluxin.be/nl/contact
Website: https://pluxin.be/
Funding Agency: https://www.blauwecluster.be/
Summary: PLUXIN focuses on plastic at the source prior to reaching the marine environment in rivers
and canals by studying a critical knowledge gap about the whereabouts of plastics and about their flux
towards the marine environment. This information is crucial to fast track cost-efficient plastic remedi-
ation measures. A central objective in this project is to develop a two-dimensional-horizontal (2DH)
plastic dispersal model. The model will be calibrated and validated with experiments and field sampling
data. In this context, plastics will be identified from remote sensing data through image recognition
algorithms (‘Deep Learning’) captured from fixed cameras on the bridges and drone acquisitions, hence
resulting in an automated plastic detection method. This information in combination with in situ
sampling will validate the 2DH-model. Main object of the project is through remote sensing and in-situ
observations in combination with numerical models contribute to our understanding of the sources,
circulation patterns and fate of plastic in the aquatic environment.
// 65

Annex
Title: C rowdsourcing, Copernicus and Hyperspectral Satellite Data for Marine Plastic Litter Detection,
Quantification and Tracking, REACT
(June 2020 – June 2021)
Contact Person: Antonello Aiello ([email protected])
Website: https://www.planetek.it/eng/react
Summary: Earth Observation by satellite can contribute to marine plastic litter monitoring thanks to its
global synoptic point of view. However, remote sensing of marine plastic litter is in its infancy, and it is a
significant scientific and technological challenge. REACT is focused on presenting a Proof-of-Concept
on remote sensing of marine plastic litter. The project aims to develop a methodology to detect plastic lit-
ter onshore or close to the shoreline and offshore. The methodology exploits data fusion of multispectral
(i.e., Sentinel-2, WorldView) and hyperspectral satellite data (i.e., PRISMA), together with in situ data
collection, and takes advantage of two different approaches. The first one based on spectral signature
unmixing, and the second one on artificial intelligence methodologies. REACT aims at filling the gap
related to: 1) The fundamental relationships between marine plastic and the reflectance captured in sat-
ellite remotely-sensed imagery, with current and future remote sensing instruments; 2) The sensitivity of
existing sensors on marine plastic litters; 3) The identification of the satellite combination delivering the
most useful fused-data; 4) The definition of spectral features and spatial scales recommended for future
missions (CubeSat/small satellite missions) to be matched with major satellites as Sentinel-2.
Title: R
emote Sensing of Marine Debris: Potentials and Limitations
(June 2021 – May 2024)
Contact Person: Chuanmin Hu ([email protected])
Website: https://optics.marine.usf.edu
Summary: Despite several pioneering studies showing potential in remote detection of marine debris
using optical means, there are still technical challenges to be addressed. The project is to address these
challenges with the following objectives: (1) To compile a spectral library of various types of marine
debris as well as other floating matters. This will be through literature search, data mining, and labo-
ratory and field experiments; (2) To determine the resolution requirements (spatial, spectral, radiomet-
ric) and optimal bands as well as potentials/limitations of current and future sensors in mapping and
quantifying marine debris. This will be through radiative transfer simulations and sensitivity analysis
using the endmember spectra and realistic measurement conditions; (3) To develop and evaluate prac-
tical approaches for several sensors to maximize their potentials in mapping marine debris; (4) To make
recommendations on future satellite missions as well as on algorithms and approaches toward remote
sensing of marine debris.
// 66

Annex
Title: D
evelopment of a risk index for floating marine litter in coastal areas by combining optical and
SAR techniques with numerical models, Satellite FRONTs for detection of Anthropogenic plastic
Litter / FRONTAL (September 2020 – September 2022)
Contact Person: Victor Martinez Vicente ([email protected])
Website: PML: https://www.pml.ac.uk/, https://www.isardsat.cat/ca/, https://www.isardsat.cat/
Summary: Fronts in coastal and oceanic regions are hot-spots for rich and diverse marine life, where
floating marine debris also tends to accumulate. The goal of FRONTAL is to develop a prototype of a
risk index for the accumulation of marine plastic debris at fronts. The approach is to combine state-of-
the-art optical processing techniques of direct detection (from Copernicus Sentinel-2 MSI), validate the
retrieval (using existing in situ datasets) and combine the results with front detection algorithms applied
to thermal, optical and SAR satellite imagery. Opportunistically, we will take advantage of hyperspectral
satellite data to explore the improvement of algorithms with collocated datasets.
In addition to mapping the risk areas for accumulation, their connectivity to the pathways into the
ocean, through numerical dispersion models of coarse and high spatial resolution will be investigated. In
doing so, we aim to provide a tool to local and regional policy makers to identify areas where interven-
tion would be more effective.
As a case study, we are working in collaboration with local stakeholders in Da Nang (Vietnam).
Title: d iStributed AI systeM for mArine plastic debRis moniToring (SMART), SMART
(May 2021 – May 2023)
Contact Person(s): PI – Leonardo Azevedo ([email protected]);

CO-PIs João Tasso ( [email protected]) and Renato Mendes ([email protected])
Website: SMART project (under construction);

AI Moonshot challenge (https://www.moonshotchallenge.ai);
Funding Agency: Portuguese Space Agency – Portugal Space (https://ptspace.pt) in partnership with
FCT, ANI, ESA, Unbabel and the support of the Web Summit
Summary: SMART is an intelligent framework based on deep physics-informed learning, which

combines automatic identification and classification of floating plastic debris from satellite images,
spatiotemporal modelling of plastic accumulations with high-resolution numerical ocean modelling,
physics-guided machine learning and a distributed system of sensors mounted on low-cost marine
autonomous vehicles for long-term deployment and validation of the model results. This unique combi-
nation will allow to bypass the need of running full ocean numerical models at small-scale simulation
grids, which brings numerical instabilities and are unable to assess uncertainty about the spatiotemporal
predictions. Instead, the final outcome for the end-user will be a probability of plastic occurrence map at
any time step required in the past and in the future. The probability of plastic occurrence map will allow
authorities to devise strategies for ocean clean-up while making decisions under uncertainty. The project
will be developed using two pilot sites in the North Atlantic and the results obtained will be validated
in situ using low-cost marine autonomous vehicles, which will collect samples at key sensitive regions
predicted by the model.
// 67

Annex
Title: S pectral Properties of Submerged and Biofouled Marine Plastic Litter, SPOTS
(September 2020 – December 2021)
Contact Person: Robin de Vries, The Ocean Cleanup ([email protected])
Website: https://theoceancleanup.com/research/
Summary: The SPOTS project will take a closer look at the influence of biofouling and water depth on
the spectral reflectance of plastics. By varying the water depth and degree of biofouling in a systematic
way and a controlled lab and outdoor environment, we will gather a more detailed dataset and predictive
model about the influence of both these factors on the hyperspectral footprint of plastic litter. Besides
debris from the marine environment, we will also investigate coastal and riverine plastic litter.
Title: T hermal Infrared Sensing of marine Plastic Litter, TISPLALI

(September 2021 – March 2022)
Contact Person: Lonneke Goddijn-Murphy ([email protected])
Website: https://eri.ac.uk/research/major-projects/tisplali/
Summary: This project explores the potential of thermal infrared remote sensing for the detection of
floating marine plastic litter and how it could complement other remote sensing methods such as those
in the optical spectrum. For example, thermal infrared sensing does not depend on the presence of day-
light and can look through light snow and rain. Some plastic materials that are transparent in the optical
spectrum may appear opaque in the thermal spectrum. We focus on the consequences of the presence
of sunlight and of different air and sea temperatures on the thermal infrared signal of plastic floating in
water. The aim is to verify a thermal radiance model using imaging long-wave infrared (7.5 – 13.5 μm),
near-infrared (850 nm), and visible colour (RGB) cameras, a drone, and plastic targets deployed at sea.
This involves drone surveys during day- and night-time hours, and in summer as well as winter to cover
a range of conditions. We support our findings with experiments in the laboratory where we can create a
more controlled environment. One of these experiments is looking at plastic litter that has spent time in
marine water, to study the effect of biofouling of the plastic surface on thermal infrared radiance leaving
this surface.
// 68

Annex
Title: T RACE: Detection and tracking of large marine litter based on high-resolution remote sensing
time series, machine learning and ocean current modelling, TRACE
(August 2020 – January 2022)
Contact Person: Mathias Bochow ([email protected])
Website: https://www.gfz-potsdam.de/en/section/remote-sensing-and-geoinformatics/projects/trace
Summary: Using daily high-resolution optical (PlanetScope), SAR (Sentinel-1), and hyperspectral
(PRISMA) satellite data, this project aims to obtain precise and reliable data on large pieces of floating
litter, regarding their quantity, trajectories and accumulation zones, material properties, floating depth,
and sources. To achieve this goal we will develop a scalable (current test area: Adriatic Sea) fully-auto-
matic remote sensing based detection and tracking system of large marine litter and accumulation zones
and couple it with oceanographic forecasting. After being operationally online the derived information
will be published with a delay of 3-4 days on a web-map-server and may serve as a basis for the recovery
of floating litter, for the elimination of its sources, and for preventing its dispersal.
Title: M
apping Windrows as Proxy for marine litter monitoring from space, WASP
(May 2020 – October 2021)
Contact Person: Manuel Arias ([email protected])
Summary: WASP is a data processor, developed in the frame of the ESA Discovery Campaign, exploit-
ing Copernicus Sentinel-2 L1C images to detect and catalogue the presence of filaments of floating
marine debris with high probability of containing man-made litter. WASP takes advantage of the
prototype EO data processor developed in the frame of ESA project “Earth Observation (EO) Track for
Marine Litter (ML) in the Mediterranean Sea” that successfully proved for first time that Copernicus
Sentinel-2 data can detect the presence of marine litter accumulations as proxies of plastic litter content.
The entire Sentinel-2 archive over the Mediterranean Sea will be processed and following an in-depth
analysis, a database of the identified proxies will be created over the area. The final product will be a map
of sub-mesoscale marine debris concentrations in the Mediterranean Sea based on Copernicus Senti-
nel-2. The product will consist on a census of these structures for each processed tile for the Mediterra-
nean Sea, with potential for global scalability.
// 69

Annex
A2. Application of Remote Sensing

in Monitoring Plastic Waste
Table A2. List of research studies on remote sensing of plastic waste since 2011
indicating sensor technologies, geographic location and key findings. Adapted from 109.
Geographic Areas Methods and Material Time Type of Debris Findings Study
Multiple sites Different drones (DJI). Manual 2020 Beached and Mapping marine plastic litter from 110
in Spain and Machine learning algo- floating UAV data can be improved by adding
rithms. marine litter dynamics information.
Saigon, Vietnam Sentinel-2 and WorldView-2 2018 - 2020 Floating Showcasing the benefits of satellite 111
data. Spectral based methods detection and quantification of hya-

and Naïve Bayes classifier. cinths in a river.
Leirosa Beach, Portugal Multispectral data from UAV 2021 Beached Spectral information from 42 types of 112
(DJI Matrix 210 RTK V2). Spec- litter items used for characterizing
tral Angle Mapper classifier. litter types.
Global scale Sentinel-2 (MARIDA) data. 2015 - Floating A labeled dataset of satellite images 113
Machine learning algorithms. present with known or suspected litter targets

useful for machine learning and de-
tection algorithm development.
Lesvos Island, Greece Sentinel-2 and WorldView 2/3 2018 - 2021 Floating Data fusion techniques applied to 114
data. Image fusion algorithms. multispectral satellite data to improve

plastic litter mapping accuracy.
Balearic Islands, Spain SAR imagery and Sentinel-2 2020 - 2021 Floating Floating marine plastic litter mapping 115
data. AI-based approaches and using SAR data applying AI tech-

PCA analysis. niques.
Lesvos Island, Greece PRISMA hyperspectral data. 2020 Beached and 96% marine litter mapping accuracy 89
Machine Learning algorithms. floating achieved using hyperspectral data.
Quiaios Beach, Portugal DJI Matrix 210 RTK V2, Multi 2019 Beached Highlights evaluates the effects of 116
spectral Sentera AGX 710 12.3 coastal vegetation in mapping marine

MP imager. Multiresolution plastic litter using remote sensing.
segmentation. It quantified and classified beached
litter.
Sardinia, Italy DJI Matrice 600. Hyperspectral 2019 - 2020 Beached and Hyperspectral data from UAV used for 117
Bobcat 32 Xenics SWIR imager, floating mapping both floating and beached
data. Supervised classification. marine plastic litter.
Limassol, Cyprus Sentinel-2 data. 2018 - 2019 Floating Detecting floating plastic debris from 118
and Mytilene in Greece Clustering, Support Vector Sentinel-2 data.

Regression.
Hawai’i Big Island, USA Sentinel-2 imagery. Spectra 2020 Floating Floating marine plastic litter from 119
indices. Sentinel-2 data using spectral indices.
Delta de l’Ebre Partenavia P- 68 aircraft, 2017 - 2019 Floating Deep learning methods for automatic 45
and Cap de Creus, Spain DJI Mavic Pro, Topografia and detection of floating marine macro-
fixed-wing HP1 drones. Canon litter from aerial imagery. An R pro-
EOS REBEL SL1, FC220, Sony gramming language application was
Alpha 7 R and Sony ILCE- developed MARLIT web-based tool.
6000 cameras. Deep learning
algorithms.
Faial Island, Portugal Cessna F150L and fixed frame. 2020 Floating Hyperspectral characterization of 46
HySpex Mjolnir S-620 and plastic litter with accuracy between

Specim FX10e. Supervised 70-80% and an evaluation of airborne
machine learning methods. detection of floating plastics.
Lesvos Island, Greece PRISMA imagery. spectral 2020 Floating Spectral index-based marine plastic 41
indices. litter mapping using pansharpened

PRISMA data.
// 70

Annex
Arno River estuary, Italy UAV (DJI Phantom 4 Pro v2 2021 Beached Assessment of drone captured RGB 120
quadcopter drone. CMOS 20 MP images of beached litter through

camera. Manual inspection. citizen science classification.
Leirosa beach, Portugal DJI Phantom 4 RTK (DJI- 2020 Beached Color-based approach can be used to 82
P4RTK) drone. 20MP camera. improve the categorization of stranded

Manual inspection and Neural litter on UAS orthophotos.
Networrk machine learning.
Leirosa Beach, Portugal DJI Phantom 4 RTK (DJI- 2019 - 2020 Beached Beach dune erosion and marine litter 121
P4RTK). 20 MP camera. Manual density from UAV-based data.

inspection.
Quiaios Beach, Portugal DJI Matrix 210 RTK V2 drone. 2019 Beached Mapping marine litter along the Atlan- 122
RGB DJI Zenmuse X5S 20.8 MP, tic coast using UAVs.
Sentera AGX 710 12.3 MP Mul-
ti- Spectral imager. Manual
inspection.
Accra - Ghana, Sentinel-2 and visual observa- 2018 - 2019 Floating Use of multispectral satellite data 72
Da Nang - Vietnam, tion in social media. Machine (Sentinel-2) for mapping floating plas-
Gulf Islands - Canada, learning and spectral based tic patches in coastal waters around
Scotland -UK algorithms. the globe. Proposed the detection
algorithm Floating Debris Index – FDI.
Mediterranean Sea, Partenavia P-68 high-wing 2019 Floating Inter-comparison of trained human 123
Spain aircraft. Visual and Canon EOS observer and automated camera mon-
REBEL SL1 camera. itoring approach for floating litter.
Cabedelo Beach, DJI Phantom 4 Pro. 20 MP 2019 Beached UAV data for mapping marine litter in 47
Portugal CMOS camera. Random Forest Portugal using machine learning ap-
(RF) classifier. proaches. F-Test score 75% compared
to manual mapping.
Cabedelo Beach, DJI Phantom 4 Pro. 20 MP 2019 Beached Comparative analysis of manual and 92
Portugal CMOS camera. Visual inspec- machine learning approaches for

tion and machine learning mapping marine plastic litter from
algorithms. imagery acquired using UAVs.
Lake Balkana and DJI Mavic Pro. RGB camera. 2019 Floating A semantic segmentation algorithm 124
Crna Rijeka River, Convolutional Neural Network revealed accurate mapping of floating
Bosnia and Herzegovina algorithm. litter based on drone imagery.
Fukiage and DJI Phantom 4 drone. 4K 2018 - 2019 Beached Estimation of beached marine plastic 125
Sato Beach, Japan RGB camera. Neural Network debris abundance using UAV imagery
algorithms. assessed by deep learning methods.
Bay Islands, Honduras Landsat-8, Sentinel-2 and 2014 - 2019 Floating and Mapping source and transport of plas- 39
PlanetScope images. Spectral beached. tic litter from the sea to the beach
based and visual inspection. using satellite remote sensing and
spectral based evaluation of suspect-
ed or known litter material.
Northern Italian coast DJI Phantom 4 PRO v2 with 2019 - 2020 Beached Long-term monitoring of beached 95
20 MP digital camera. Visual marine debris using UAV surveys.

interpretation.
Saronikos Gulf, Greece Research Vessel. Nikon D80 2017 - 2018 Beached This study utilised regression anal- 126
10.2 MP Digital SLR camera. ysis between in situ daya and ves-
Regression analysis between in sel-based photographs for beached
situ and camera data. litter assessment. Image analyses
were done to reveal classes and
quantities of marine debris.
Ducie Atoll, Digital camera. 2019 Beached (sea- Detecting plastic litter in seabird 127
southeast Pacific Ocean bird nest) nests using digital photographs.
Limassol, Cyprus DJI Phantom 2 and Pro drone. 2018 Floating Evaluated the detection of artificial 37
Sentinel-2, Sony Exmor IMX206 floating plastic targets using the Sen-
multispectral imager, GoPro tinel-2 imagery and spectral based
and 20 MP RGB camera, SVC algorithms.
HR-1024 spectroradiometer.
Spectral based algorithms.
// 71

Annex
Tsamakia Beach, Greece DJI Phantom 4. 2019 Floating Detection of artificial floating plastic 32
Sentinel-2 and 12.4 MP RGB targets using Sentinel-2 and drone

camera. Spectral and object data. The Plastic Litter Project
identification-based algorithms. (PLP2019) was also showcased.
Aegean Sea, Greece ECMWF ERA5 data. 2018 Floating Theoretical approaches are presented 128
and global and prospect for thermal thermal IR

remote sensing application in detect-
ing floating plastics.
Alif Dhaalu and DJI Phantom 4 drone. 12.4 MP Not pro- Beached Automated method proposed for 93
Faafu Atolls, Maldives CMOS camera. Visual inspec- vided the detection and quantification of
tion and Convolutional Neural beached marine plastic litter using
Network algorithm. aerial data from UAVs by applying
convolutional neural networks.
Jalan Tengku Kalana DJI Phantom 4 Advanced 2019 Floating Quantification of riverine floating 129
bridge and Klang River, drone. 12 MP CMOS camera. (riverine) plastics and its transport using
Malaysia Visual inspection. UAV imagery supported by visual
inspection.
Tsamakia Beach S900 DJI Hexacopter drone. 2018 Floating Marine floating plastic litter detection 36
on Lesvos Island, Sony A5100 24.3 MP camera, using satellite and drone data. Spec-
Greece Slantrange 3P, Parrot Sequoia tral feature and atmospheric correc-
multispectral and FLIR Duo tion evaluation on top- and bottom-
R camera, Sentinel-1, 2 data. of-atmosphere data products.
Visual inspection and spectral
analyses.
Edinburg, Scotland Nessie AUV. SoundMetrics Floating and Submerged marine litter detection 130
ARIS Explorer 3000 acoustic submerged using sonar data with the support of
camera. Deep Neural Network dep neural network algorithms.
algorithms.
Hokkaido and Kanto, DJI Matrice 210 RTK. Go Pro 2018 - 2019 Floating and Deep learning approaches were de- 131
Japan Hero 6, Olympus Tough TG5, submerged veloped to detect and classify debris
iPhone ZENMUSE X5S and ZEN- and other objects using images from
MUSE XT). Deep learning object underwater and airborne observations.
detection algorithm.
Atlantic Ocean, PANalytical Boulder ASD Field- 2015 Beached, Shortwave infrared absorption fea- 35
Pacific Ocean, Spec 4 and AVIRIS data. Spec- ocean harvest- tures of synthetic hydrocarbons using
Westcoast and Hawai’i, tral indices, spectral mixing. ed and virgin airborne and spectroradiometer data.
USA
Chiloé Archipelago, Hyperspectral ASD, HyLogger-3 2017 Beached Spectral characterization of anthropo- 84
north-western spectrometer and WorldView-3. genic marine debris using spectrome-

Chilean Patagonia SVM classification and Linear ter and WorldView-3 data analysis.
Discriminant Analysis (LDA)
Fuzhou, China DJI PHANTOM 4 PRO quad- 2017 Beached Automatic mapping of beached marine 132
copter drones. 20MP camera. plastic litter using image segmenta-

Segmentation threshold method tion by applying a threshold. Effective
in quick monitoring in an area of
interest.
North-East Marine Aerial photographs from UAVs 2017 Beached Marine plastic litter monitoring using 133
Protected Area of the and GoogleEarth photographs. aerial photographs taken from UAVs.
Maltese Islands Generating point cloud and
(Malta, Comino, Gozo) texture map.
Pacific Ocean SWIR (SASI) and RGB imagery. 2016 Floating Airborne hyperspectral SWIR imagery 43
Lockheed C-130 Hercules for mapping ocean floating plastics

aircraft. based on spectral reflection prop-
erties.
Scotland, UK FieldSpec hyperspectral data. 2018 Floating Spectral properties of floating plastic 53
litter using spectroradiometer (350-

2500 nm).
// 72

Annex
Tsuruga Peninsula, WorldView-3. Spectral Angle 2014 Spectral analyses of floating litter 134
Japan Mapper. and plastic targets. Method proposed

to distinguish marine litter from
white-crested waves
Vancouver Island, Aerial photographs from air- 2014 - 2015 Beached This study quantified marine macro 91
Canada plane and GoogleEarth data debris around Vancouver Island in

Canada using aerial photographs from
airplane. Aerial photographs were
processed by projective transforma-
tion and by extraction of debris pixels.
Saudi Arabian DJI Phantom 4 mounted with 2017 Beached Mapping beached marine debris using 48
Red Sea Coast 20MP camera. Machine learn- photographs taken from UAV.
ing algorithms.
Hawaiian Islands, USA High-resolution aerial photo- 2015 Beached High-resolution aerial photographs for 135
graphs detecting coastal marine macro-de-

bris (> 0.05m2).
Arctic, Mediterranean, Hyperspectral data. PCA and 2014 - 2017 Floating/sub- Characterization of the polymeric 136
South Atlantic and Partial least squares-discrimi- merged composition of marine plastic litter
North Pacific nant analysis. using hyperspectral data (1000-2500
nm).
Fukushima Daiichi, ASTER, AVNIR-2, RapidEye, 2011 Floating Assessment of the litter dynamics af- 59
Japan PALSAR, RADARSAT-2, World- ter the tsunami event near Fukushima.
View-2, RGB camera
Isle of Rügen, Germany Field survey and photographs 2015 Beached Marine litter abundance and distribu- 137
from drones. tion on beaches using field data and

photographs from UAV. The accuracy
depends on vegetation.
Nanhui beach, Shanghai, RIEGL VZ-4000 terrestrial 2015 Beached Semi-automatic detection of beached 61
China laser scanner LiDAR data. SVM marine macroplastic debris using
classification. LiDAR data.
Coasts of East Asia Sequential webcam photo- 2010 - 2011 Beached This study estimated a 250-fold 138
graphs and particle tracking increase in beached plastic litters in

model. East Asia in the next 10 years using
sequential webcam photographs and
particle tracking model.
Ookushi Beach and Low-altitude aerial digital - Floating and Monitoring marine and beach litters 49
Ookushi Beach, Japan camera images from balloon. beached of assorted colours using photographs
taken from balloon equipped with a
digital camera.
Tobishima Island, Japan Vivotek IP7361 Webcam cam- 2010 - 2011 Beached Mapping coloured beached mac- 55
era. Colour references using a roplastic debris using webcam images

uniform colour space (CIELUV).
Ookushi Beach, Japan Helium Sky Catcher Balloon. 2009 Beached Quantification by weight of beached 23
Digital camera and visual plastic litter based on aerial photo-

inspection. Threshold based on graphs and in situ data.
lightness value.
// 73

Annex
A2.1 Platforms for Remote Sensing of Plastic Waste
Table A2.1 D
etailed examples of platforms that have been utilized to investigate the potential applications of remote
sensing in monitoring plastic litter. Adapted from 105.
Quantification of macroplastics larger than 50 cm using of action cameras placed on vessels of opportunity. Macroplastic density 33
is estimated through a twofold approach based on object detection and training object detection models. This study compared the
distributions of macroplastics with concentrations of micro- and mesoplastics collected with manta trawl nets.
Ship The Convolutional Neural Network approach is able to train itself on images of plastic objects larger than a few centimetres and 68
automatically predict the class of new images of macro plastic objects floating at sea. With the aid of a camera mounted on
board a marine vessel, the system would scan the sea surface, and detect and recognise litter. The system was trained on three
categories of plastic marine litter (bottles, buckets and straws) and the classifier was able to recognise these types of floating
objects at a success rate of � 86 %.
Floating plastics in aerial images were detected using deep learning models based on an algorithm that uses Convolutional Neu- 45
ral Networks capable of learning from unstructured data. This model was implemented in an application to detect and quantify
marine litter in the images.
Two combined methods to detect floating macro litter, visual observations and drone surveys, were compared. Both methods 139
proved equally effective at detecting floating plastics. Two different commercial drones were used for drone surveys, equipped
with a 12-megapixel camera. Flight height was set between 45 and 65 m to guarantee a ground sampling distance of 2 cm per
pixel.
RGB and hyperspectral short-wave infrared imagery were captured with equipment mounted on a C-130 aircraft surveying the 43
Aircraft great Pacific garbage patch at a height of 400 m and a speed of 140 knots. Position, size, colour and type (container, float,
Drone ghost net, rope and unknown) were recorded for every plastic piece identified in the RGB mosaics, and then the top 30 largest
items within each plastic type category (0.6−6.8 m in length) were selected to investigate spectral information obtained with a
SASI-600 imager (950−2450 nm). Analyses revealed unique spectral features common to plastics, with some variability probably
influenced by differences in the objects’ optical properties, water submersion and the atmosphere. Simulations confirmed that the
plastics’ absorption features have potential applications in detecting and quantifying ocean plastics from spectral information
obtained from airborne images.
Low-altitude remote sensing methods were used to monitor marine and beach litter with a remote-controlled digital camera sus- 49
pended from a balloon filled with helium gas, suspended at 0–500 m above sea level. Photographs were taken at various angles,
and images were processed to identify litter using colour differences between target objects and the background in the CIELUV
colour space. With the balloon suspended at 150 m, a pixel represented an area of 100 cm2.
Novel supervised and unsupervised clustering algorithms were developed to identify floating plastics using in situ validated 118
Sentinel-2 images with different size of deployed plastic targets (10 m × 10 m, 5 m × 5 m and 1 m × 10 m). Three different sets
of bands and indices were employed to develop the attributes for the classification process. The best-performing method, Support
Vector Regression based supervised classification, had an accuracy in the range of 96.9–98.4 %.
This study explores for the first time the use of satellite hyperspectral PRISMA images to detect floating marine plastic litter. 41
Thirteen pansharpening methods and denoising pre-processing techniques were employed (e.g., Bayesian, deep learning, compo-
nent substitution) along with three novel indices to detect floating plastic targets with a low number of false positives.
WorldView-3 imagery was used to detect known and suspected floating plastic materials in the North Pacific Garbage Patch. A 42
simplified approach was proposed that utilizes the spectral anomalies by assuming the ocean surface signal as the baseline or
background signature.
Satellite technology was employed to investigate the feasibility of monitoring marine surface floating plastic litter in coastal 72
areas.
High-resolution multispectral satellite images from Landsat-8, Sentinel-2 and Planet satellite missions and in situ observations 39
were employed to study the sources and trajectories of floating marine plastic litter in the Bay Islands of Honduras (Caribbean
Satellite Sea). The determination and discrimination of floating litter was carried out manually by photo interpretation experts.
A spectral signature for the polyethylene terephthalate targets was produced by modifying the US Geological Survey polyethylene 32
terephthalate signature, using an inverse spectral unmixing calculation to perform matched filtering processing on the Sentinel-2
images. The results provide evidence that, under suitable conditions, pixels with a polyethylene terephthalate abundance fraction
of at least as low as 25 % can be successfully detected.
Sentinel-2 satellite images were combined with multispectral aerial images acquired from an UAV to determine if plastic litter on 37
the sea surface can be detected using an artificial plastic target (3 m × 10 m). Images were processed using two newly devel-
oped indices, the plastic index and the Reversed Normalised Difference Vegetation Index.
This study explored the application of unmanned aerial systems and open-access satellite imagery in remote detection of floating 36
plastics in natural seawater, through a dedicated aquatic environment experiment using a set of three artificial floating plastic
targets placed in the coastal zone.
WorldView-3 images were combined with anthropogenic marine debris hyperspectral laboratory characterization to detect large 84
and highly reflective plastic items on beaches. A spectral library for the implementation of a digital classification method applied
to WorldView-3 satellite images was generated by collecting litter samples from the Chiloé Islands beaches (Chile) and assess-
ing their spectral signature.
// 74

Annex
A3. Machine Learning Application

in Remote Sensing of Plastic Waste
Table A3. List of example research studies that utilized machine learning algorithms to derive descriptors (e.g., pres-
ence, quantities, shapes, forms, colours) about plastic waste in imagery captured using remote sensing
technologies. Adapted from 90.
Methods and Materials
Image Source Algorithm Annotations Feature Space or Wavebands Findings Study
Drone. Segmentation and Object-based Blue, Green, Red, Red Edge, NIR Density map, material counts 112
Multispectral. spectral angle mapping of litter and plastics.
Drone. Random Forest Pixel-based Violet, Blue, Green, Red, Red Edge 1, Semantic segmentation of 140
Multispectral. Red Edge 2, NIR 1, NIR 2, NIR3 floating plastic.
Trolley. Feature selection, Pixel-based 8/36 wavebands in the range Semantic segmentation and 141
Hyperspectral. spectral angle and 1169-1233 nm and 1612-1677 nm detection of polyethylene and
correlation mapping polypropylene plastics.
Drone. Feature selection, Linear Pixel-based 10/320 bands in range Semantic segmentation and 117
Hyperspectral. Discriminant Analysis 900 – 1700 nm – Feature Selection detection of polyethylene and
polyethylene terephthalate
plastics.
Handheld. Support Vector Machine, Pixel-based Violet, Blue, Green, Red, Red Edge 1, Semantic segmentation 80
Multispectral. Random Forest, spectral Red Edge 2, NIR 1, NIR 2, NIR3 of floating plastics.
based indexes
Drone. Support Vector Machine, Object-based Red, Green, Blue Object-based litter 142
RGB. Random Forest, occurrence map litter.

Segmentation,
Maximum Likelihood
Aircraft. Support Vector Machine, Pixel-based Wavebands from 400 – 2500 nm Semantic segmentation and 46
Hyperspectral. Random Forest detection of 3 artificial ma-

rine litter targets.
Drone. Shallow Feed-Forward Image Density map, item counts 82

Red, Green, Blue and
RGB. Neural Network samples of litter, plastic bottles,
three colour spaces
fishing ropes, octopus pots,
(HSV CIELab, YCbCr): 12 features
fragments.
Drone. Segmentation, K-Nearest Object-based Red, Green, Blue and Object-based litter occur- 143
RGB. Neighbour, Support Vector three colour spaces rence map of beach litter.
Machine, Random Forest (HSV CIELab, YCbCr): 12 features
Drone. Random Forest Image sam- Red, Green, Blue and Density map, item counts of 47
RGB. ples three colour spaces beach litter.

(HSV CIELab, YCbCr): 12 features
Drone. Random Forest, Convolu- Image sam- Red, Green, Blue and Density map, item counts of 92
RGB. tional Neural Network ples three colour spaces bottle, fishing string, plastic
(HSV CIELab, YCbCr): 12 features pieces, octopus pot, cap,
boot/shoes,
polystyrene, pieces
Satellite. Support Vector Machine, Pixel-based 8 wavebands in VNIR Semantic segmentation, 84
Multispectral. Random Forest, Linear detection and quantification

Discriminant Analysis of beached litter including
Styrofoam.
Drone. Segmentation and Object-based Blue, Green Object-based litter occur- 132
RGB. Thresholding rence map of beach litter.
Drone. Random Forest Image sam- Histogram of oriented gradients Density map, item counts of 48
RGB. ples beach Litter, drink containers,

bottle caps, plastic bags.
// 75

Annex
A4. Conferences, Proceedings and Workshops
A4.1 Conference Proceedings 7th Marine Debris Conference,

Busan, South Korea (7IMDC, 18 – 23 September 2022)
The IOCCG Task Force on Remote Sensing of Marine Litter and Debris and the NOAA Marine Debris
Program supported and their representatives co-chaired the session TS-3.5 Satellite and Airborne
Remote Sensing of Marine and Coastal Litter at the 7IMDC (https://7imdc.exordo.com/programme/
session/110). Oral and poster presentations showcased advances and ongoing research on remote sens-
ing of plastic waste.
Oral Presentations
Title: Towards a framework for modelling marine litter detection from space.
Author: Delaney et al., 2022
Link: https://7imdc.exordo.com/programme/presentation/407
Abstract: To optimise mission success, one needs the sensor characteristics to match domain require-
ments. The algorithms and processing need to be prototyped and tested rigorously before launch and
during operation when introducing new features, and debugging are crucial to ensure longevity. Model-
ling and simulation enable engineers and scientists to understand how parts of a system interact and how
the system behaves. Towards this end, the European Space Agency (ESA) Discovery Element funded
the Marine Litter Operational Performance Simulator (ML-OPSI), a virtual breadboard to model the
marine litter (ML) domain and simulate the acquisition of optical and IR signals at top-of-atmosphere
(TOA) by user-defined EO sensors under varying environmental conditions. An aim is the assessment of
the optimal characteristics required of a sensor to detect ML, thus informing future missions carrying
the next generation of sensors. More generally, it supports EO scientists and engineers in designing and
implementing computational experiments by acting as a ‘virtual laboratory’ capable of benchmarking,
verifying and validating different algorithms. The objective is to enhance scientific evidence-based
knowledge about how varying the quantities, composition and location of marine litter governs detect-
able water-leaving reflectance. As a proof-of-concept, a standalone demonstrator to estimate reflectance
was developed, using the model of Goddijn et al., 2019 configured with laboratory measurements of ML
spectral signatures. The aim is to model test scenarios in which aggregates of ML form, such as wind-
rows, oceanic fronts and gyres, and river mouths, to provide a bottom-of-atmosphere (BOA) signal and,
using the Free and Open-Source Software (FOSS) 6SV Radiative Transfer Model (RTM), propagate
that signal to TOA and subsequent sensor detection. We designed ML-OPSI as a modular, pluggable
and extensible framework that promotes re-use and adapting to different missions, sensors and scenarios.
It is conceptually a component-based architecture composed of two packages, a Modeller and a Simula-
tor, themselves modular in design and user-customisable to represent one or more EO domains, such as
ML detection. The ML model comprises top-level modules responsible for scene generation, atmospheric
correction, instrument detection and retrieval, and the definition of inputs, outputs, and parameters
characterising the interface. The OPSI simulator comprises a GUI-based scenario/model builder and an
orchestrator that manages simulation runs, configuration management, and performance assessment.
We propose the sum of these parts is a framework for performing EO end-to-end simulations and, with
// 76

Annex
appropriate configuration and real-time updates from an operational sensor, acting as a ‘digital twin’ to
enable experiments and assessments to be performed on a simulated version of the physical asset. The
authors call to action is to engage the ML community in discussion to converge towards a common goal
considering the interface standards, models and data types available, implementation and deployment to
advance ML-OPSI as a standardised EO modelling and simulation framework.
Demo-Version of ML-OPSI to get a proof of concept
Title: T he development of a COTS-based multi-spectral imaging system

for the airborne and space-based detection and imaging of plastics.
Author: Hibbitts and Bekker, 2022
Abstract: The remote detection and mapping of plastics in the natural environments is challenging.
While plastics have diagnostic spectral reflectance features in the shortwave infrared, those features also
vary in strength and position with composition [e.g. 1]. However, spectral studies [e.g. 2] have shown
that marine debris plastics tend to be dominated by a few plastic varieties, even if most/all types are pres-
ent at some abundance. These most prevalent plastics consistently have spectral features near 1.2, 1.4,
and 1.7 microns. Spectral obscuration by telluric absorptions occurs near 1.4 microns due to atmospheric
water vapor, leaving the 1.2 and 1.7 micron absorption features for plastic detection. While hyperspec-
tral instruments can discriminate between these plastics, detection only requires multispectral measure-
ments. The modest spectral capability of a multispectral approach enables emphasizing high resolution
imaging for acquiring high-quality compositional images. Because much of marine debris will remain
subpixel in even the best aerial and especially space-based imaging, higher spatial resolution improves
performance by increasing the fractional pixel area associated with plastics for a stronger signal, enabling
even subpixel detection of plastics (such as by microplastics). Thus, spatial resolution is no longer driven
by the impractical requirement to spatially resolve plastic objects but by the signal level associated with
// 77

Annex
the fractional area the plastic debris covers. Beaches containing plastic debris can be several meters wide.
This resolution is easily achieved from an airborne instrument and is feasible from space. A spatial resolu-
tion of ~4 meters can be achieved with a 10-cm aperture at about 1.2 microns when observing from LEO
such as from the ISS or a CubeSat. However, it is the ground sampling distance of over 7m achieved with
even a relatively fast integration time of 1ms that will be the limiting factor. This resolution remains suf-
ficiently fine to resolve many coastlines as well oceanic garbage patches and is superior to that achieved
with spacebased hyperspectral imaging.We are developing a Distributed MultiSpectral Imaging System
(DMSIS) using this approach for the detection and imaging of marine debris and other plastics in the
natural environment. It is a reconfigurable multispectral imaging system consisting of four IR cameras
with tailorable wavelengths. The ‘imaging first’ approach maximizes spatial resolution and the four
modular cameras provide spectral information for detecting plastics. This modular approach leverages
unmodified COTS SWIR imagers for a low barrier to entry intended to enable wide-scale adoption.
A previous iteration using only two wavelengths has been demonstrated but the spectral sampling was
insufficient for the reliable detection of plastics [3]. A third band is needed to better constrain illumina-
tion geometry effects and a fourth band is needed for discriminating plastics from spectral distractors.
This approach of using separately filtered COTS cameras transfers the complexity and expense from
hardware to software and is enabled by parallel camera data acquisition and software-based image regis-
tration.
References: [1] Garaba, S.P. and H. M. Dierssen, 2018, Remote Sens. Env., 205, 224-235; [2] Guffogg,
J.A. et al., 2021, Remote Sens., 13, 4548; [3] Hibbitts, C.A. et al., 2019, Proc. SPIE 11012.
Title: E
lucidating Patterns of Urban Plastic Pollution in Mumbai, India
using Remote Sensing Technologies.
Author: Mathis et al., 2022
Abstract: Because of the large scale over which plastic waste is managed or leaked into the environment,
monitoring must also happen at a large scale to provide consistent and widespread assessments. How-
ever, reliable data on mismanaged waste remains scarce in the developing world, creating challenges for
communities to design, implement and monitor policy and interventions to improve waste management
practices. Large scale monitoring techniques like remote sensing can give communities the ability to
analyze the spatiotemporal patterns of waste transformations in their locality, predict scenarios, and
develop targeted solid waste management and intervention strategies. Further, fast, and accurate earth
observation approaches that classify waste accumulation in the environment in satellite imagery paired
with comprehensive local data can help to fill knowledge gaps, increase social connectivity, and detect
zones with elevated risk. We chose Mumbai, India as a preliminary study area for its mosaicked urban
landscape, socio-economic fragmentation, tropical monsoon climate, proximity to the ocean and its
riverine environment. While the focus of this work uses optical satellite imagery to detect waste accumu-
lation sites in Mumbai, it provides a baseline framework to develop accurate, inexpensive, and scalable
methods that fuses satellite data types with hyperlocal data (e.g., socio-economic data that influences
plastic patterns) that can help communities develop more targeted waste management strategies. ncept
model was developed to detect waste accumulation sites in Mumbai, India using PlanetScope’s 8-band
3m spatial resolution optical satellite imagery. To address the spatial-spectral trade-off in PlanetScope
data, instead of straightaway detecting waste, a heuristic method was devised to detect and remove all
non-waste classes (e.g., buildings and vegetation) from the satellite image feature space. Remote sens-
ing-based band ratios and indices that gave the highest separation between waste and non-waste classes
were applied to the satellite image. This helped heuristically reduce the search space by 30% (95M pixels
to 66M pixels). In this reduced search space, unsupervised k-means clustering was performed and then
// 78

Annex
ranked based on how pure it was in terms of waste pixels. Top-ranked clusters were selected as model-de-
tected waste accumulation sites, which ended up identifying 244 of the 247 ground sampled waste sites
correctly. The preliminary results from the proof-of-concept image processing framework looks prom-
ising in studying the spatiotemporal distribution of plastic waste in megacities with complex land use
patterns.
Title: Towards a spectral index for detecting marine plastics in beach environments.
Author: Guffogg et al., 2022
Abstract: Plastic pollution in marine and coastal environments pose environmental and economic
challenges. This marine plastic debris (MPD) often initiates from terrestrial sources and can move
through, and accumulate in several environment sinks: open ocean surface waters, the water column,
the benthic layer, coastal waters and beaches. While the bulk of MPD is in open oceans, movement of
MPD between sinks is fluid, and a global MPD accounting system should consider each potential sink.
Remote sensing is a cost effective and scalable solution that has gained traction over recent years for
MPD detection. Spectral signatures of plastics found in the marine environment (polypropylene, high
and low density polyethylene, and polystyrene) have been studied with the intent of using remote sensing
platforms to locate and quantify MPD. However detecting MPD on beaches presents several challenges;
MPD may co-occur with organic debris which is subject to similar forces of accretion, and underlying
sediments can vary significantly in colour, mineral composition and grain size. We evaluate spectral
reflection and several indices derived from experimentally generated mixed pixels containing plastics
and sands. Investigating these indices contributes to the development of a novel method for detecting
beached plastics using spectral absorption features in the VIS-SWIR (350-2500 nm) spectrum. We test
these indices along with previously developed plastic debris specific indices (FDI, PI) to determine their
usefulness for detecting beached MPD. Spectral libraries were developed from MPD collected from
the Cocos (Keeling) Islands in the Indian Ocean. Linearly mixed synthetic pixels were created from
endmembers of feldspar-silica, carbonate and basaltic sands, five types of MPD and three types of virgin
plastics. The mixed pixels were then resampled to the spectral resolution of several candidate satellites:
Sentinel-2, Landsat-8, Worldview-3 and EnMAP. Previously developed indices(PI, FDI) and general
index formula (Ratio index, difference index, normalised difference index and soil-adjusted normalised
difference index) were then generated for all possible 2-band combinations from the resampled pixels.
Simple linear regression was used to determine the strength of the relationship between these indices
and changes in MPD surface cover in the resampled pixels. Of the satellite sensors evaluated, Landsat-8
performed most poorly across all tested indices (r2 0.04-0.42) and EnMAP performed the best (r2 0.74-
0.91). The improved spectral resolution of the Worldview-3 sensor compared to Sentinel-2 and Landsat-8
also showed promise for the development of new indices. However, the sand substrate influenced these
results; For the ratio index, r2 values ranged from weak (r2=0.44) to strong (r2=0.81) when all values were
consistent except for the type of sand and for the soil-adjusted index, r2 varied from 0.39 to 0.91. It was
also found that the optimum value for the constant L in the soil-adjusted index differed depending on
sand. Several indices have been proposed detecting floating plastics; however they have not been tested
in beach environments. Our results progress the state of the art for detecting plastic debris in beach and
coastal environments. This contributes to a better understanding of the total MPD across marine sinks.
// 79

Annex
Spectral reflectance of various material as observed by selected satellite missions.
Title: M
onitoring technology combining drones and artificial intelligence
for efficient management of marine debris.
Author: Bak et al., 2022
Abstract: In order to strengthen the ability to collect and manage marine debris, it is important to
accurately understand the actual situation based on field survey information. Through the rapid collec-
tion of on-site information using drones and the analysis technology based on artificial intelligence, it
is possible to support objective decision-making required for the rapid collection and management of
marine debris. As of 2018, the annual inflow (generation) of marine debris in Korea is estimated to be
145,258 tons, and it is estimated that about 148,721 tons of garbage are present in the domestic ocean.
Monitoring, which is currently being conducted for marine debris management in Korea, is conducted
through visual inspection by ships and manpower, resulting in loss of time and manpower. In addition,
it is difficult to grasp the exact condition of occurrence because it is difficult for manpower to access or
there is a limit to investigating a wide range of space. In order to solve this problem, it was attempted
to collect more than 420,000 pieces of data for each type of marine debris pollutant and pollutant,
which are required to strengthen ICT-based marine debris collection and management capabilities,
through drones, etc. In addition, it was attempted to objectively quantify the amount of marine debris
by developing an AI detection algorithm using the collected AI learning data. In this study, the marine
debris detection model was implemented into two types: an object detection model using data labeled
in the form of a bounding box and a semantic segmentation model using data labeled in the form of
polygon segmentation. As a result, in the case of the object detection method model, there is a difference
in AP (Average Precicion) depending on the properties of marine debris, but in the IoU (Intesection over
Union: 0.5 standard) detection performance evaluation, the mAP (mean average precision) of the entire
// 80

Annex
class is higher than 0.85. In the case of the semantic segmentation model using polygon segmentation
labeling data, it showed a high level of accuracy such as 0.76 for beach litters and 0.85 for floating debris
based on mIoU. In the future, it is expected that this study will support rapid marine debris collection
and objective decision-making through a data-based scientific approach to the generation and distribu-
tion of marine debris.
Title: Plastic Litter Project 2022: first results on the detection of artificial floating marine litter targets.
Author: Topouzelis and Papageorgiou, 2022
Abstract: Accurate observations of the sources, composition and densities of FML in oceans are sparse
and lacking. Remote sensing can play a significant role in detecting and monitoring marine debris, and
to this extent, adequate in situ observations of calibration and validation data are essential. Since 2018
we have launched a series of experimental field campaigns, the Plastic Litter Projects, to enrich the
scientific community’s understanding of FMLs spectral properties and behaviour. By developing, con-
structing, and deploying artificial floating targets containing various types of FML, we aim to produce a
comprehensive remote sensing image database that can be used to develop, calibrate, and validate FML
detection algorithms. Throughout the years, we have used various types of marine litter items such as
PET bottles and HDPE bags and natural floating materials such as reeds in the construction of artificial
floating targets. During the first years of the PLPs, we formed small-scale re-deployable artificial floating
targets and made specific time-limited experiments. In the latter years, we moved on to long-term target
infrastructure. During the 2021 experiment, we constructed two large long-term deployment artificial
floating targets, which were deployed for a four-month acquisition period. The first target consisted of a
circular 28 m diameter HDPE pipe frame, with white HDPE mesh attached acting as a representative
target material, creating an effective target area of about 600 m 2. The second target was made from more
than 350 wooden planks representing natural floating marine debris, approximating the same effective
target area. One of the main project outcome was the production of an image database containing Senti-
// 81

Annex
nel-2 and very high resolution RGB aerial images of artificial marine debris targets. The drone data allow
for the estimation of the targets’ position in Sentinel-2 images and the derivation of pixel abundance
fractions. Partial unmixing is used for the detection and concentration estimation of the FML targets in
Sentinel-2 images. The PLP 2022 experiment is twofold; firstly, we plan to move on to a more replicable
scheme using floating plastic targets which are deployed for an extended period and secondly, to examine
the theoretical minimum plastic coverage that can be detectable from satellites. Spectral classification
methodologies such as partial unmixing algorithms are expected to successfully detect FML using
the mean spectral signature acquired during the PLP2021 deployment period. The present work will
showcase a satellite and drone imagery analysis to detect FML and discriminate from natural floating
materials or lookalike phenomena. Further analysis of the correlation between environmental and
detection-affecting parameters and the resultant signal of FML will be helpful in operational scenarios of
FML detection and monitoring.
Title: Enhanced Detection and Characterization of Shoreline Marine Debris using Polarimetric Imagery.
Author: Uhrin et al., 2022
Abstract: Polarimetric imaging is an emerging technology, which is proving useful in detection and
characterization of objects within a scene. While traditional multispectral and hyperspectral cameras
provide information about the reflectance of objects in different parts of the electromagnetic spectrum,
polarimetric imaging cameras go beyond this by providing information regarding the polarization state
of the received electromagnetic energy. This capability is valuable in obtaining information regard-
ing objects’ surface roughness and other characteristics. The combination of information acquired by
spectral and polarimetric imaging often reveals independently distinctive features about human-made
// 82

Annex
objects, aiding in their detection and identification. This type of imaging is particularly valuable when
discriminating weak target signatures from complicated surroundings under a range of illumination
conditions. For these reasons, there are potential applications of this imaging technology to support
marine debris detection on shorelines, where it is often intermixed with substrate, vegetation, or natural
(woody) debris. In 2020, a team of NOAA scientists and Oregon State University remote sensing experts
initiated a project to better understand the applicability polarimetric imaging to marine debris detection.
The specific goal of this study was to investigate whether polarimetric imagery improves both visual and
automated identification of debris objects on sand beaches compared to standard red-green-blue (RGB)
imagery, and whether debris classification accuracy improves by including polarimetric bands. Using
a FLIR Blackfly-S USB3 RGB polarimetric imaging camera, we created and evaluated three orthomo-
saic 8-band composite shoreline images comprised of three spectral (RGB) and five polarimetric bands.
Images were collected from Neptune State Scenic Viewpoint in Oregon (staged debris) and two sand
shoreline sites on San Jose Island, Texas (in situ field debris). Classifications were limited to seven debris
material types including plastic, rubber, glass, processed wood, buoy, tope and aluminum. Correlation
matrices revealed that bands representing the degree and angle of linear polarization were sufficiently
different from the RGB bands such that their inclusion in the classification algorithm was beneficial.
It was also shown quantitatively that the addition of the polarimetric image-derived bands improved
the separability of debris classes. When polarimetric image-derived bands were included in a K-nearest
neighbors machine learning algorithm, the overall accuracy of debris classification increased by 6.7 –
25.6 percentage points, depending on the site. Kappa statistics increased by 9.1 – 17.6 percentage points,
again, site dependent. There was also marked improvement in both producer’s and user’s accuracy for
most categories of debris when polarimetric image-derived bands were included. The ability to accu-
rately detect and identify marine debris enables a comprehensive assessment for understanding its likely
impacts while providing significant value in debris removal prioritization efforts. Our study provides
strong indication that polarimetric imaging is a useful asset in detection and classification of marine
debris found on sand shorelines when combined with RGB bands. Future work will focus on integration
of polarimetric imaging cameras on uncrewed aircraft systems (UxS) for efficient marine debris detec-
tion and characterization.
// 83

Annex
Title: E
ye in the sky: Drone highlights exposure of marine turtles to floating litter
in nearshore waters of Mayo Bay, Philippines.
Author: Abreo et al., 2022
Abstract: Man-made litter is threatening the marine environment with detrimental effects on wildlife.
However, limited marine litter data brought by funding and logistical challenges in developing coun-
tries, such as the Philippines, hampers the full understanding of the problem. Moreover, data from the
Philippines are not comparable due to non-standardized methodologies used by different studies. Here,
we employed commercially-available unmanned aerial vehicles (UAV) as a cost-effective tool to study
risk exposure of marine turtles to floating litter in nearshore waters adjacent to a nesting beach in Mayo
Bay, Davao Oriental, Philippines. A quadcopter drone was flown autonomously to a height of 30m at
a speed of 8 m/s, with on-board camera pointed at nadir to collect videos. The videos were subjected to
post-flight processing and still frames were extracted when either turtles or floating litter were detected.
The extracted frames were georeferenced in QGIS software. Spatial statistical packages in QGIS and
R programming language were used to analyse the relationships between the point patterns of floating
litter and marine turtles. Results showed that within 50m radius, 68.2% of marine turtles were found in
proximity to litter and 100% of turtles were in proximity to litter within a 150m radius. The mean num-
ber of floating litter surrounding turtles at 50m radius was 6.32 items; this increases to 79.27 items at
300m radius. The average minimum distance between turtle and litter was 40.06m (s.d. = 32.16), with
the shortest distance observed at 0.36m. Also, Cross K function revealed that floating litter occurrence is
clustered around marine turtle signifying spatial dependence between the two. A source of concern since
the overlap in marine turtle and floating litter occurrences in nearshore coastal waters likely increases the
risk of ingestion and entanglement, adding to the myriad of threats on the survival of these organisms
and can deleteriously affect the marine turtle population. The study highlights the effectiveness of off-
the-shelf UAVs in unraveling the relationships between litter and marine wildlife, especially in develop-
ing countries where it is costly and time consuming to collect data. Application of emerging technologies
can provide the needed fine scale resolution data and empirical evidence to substantiate findings of
marine litter modeling studies. Application of artificial intelligence and machine learning can further
improve methodology presented here and could be an essential tool for standardization of data collection
for comparability of marine litter studies. Finally, the cost-effectiveness of this methodology can allow
for regular FML monitoring.
// 84

Annex
// 85

Annex
A4.2 ESA Living Planet Symposium, Bonn, Germany

(25 - 26 May 2022)
The material presented here is adapted from the symposium report (https://ioccg.org/wp-content/
uploads/2022/07/esa-lps-report-tf-rsmld_july2022.pdf) prepared by the IOCCG Task Force on
Remote Sensing of Marine Litter and Debris. The coordinated activities by the Task Force included (i)
scientific session and (ii) networking event. These activities were aimed to provide a status update to
the stakeholders, provide a community two-way feedback and identify ways forward. The two scien-
tific sessions were (a) A8.08.1 Advances and EO Applications in Remote Sensing of Marine Litter and
Debris – 1, (b) A8.08.2 Advances and EO Applications in Remote Sensing of Marine Litter and Debris
– 2 and a related poster session. A nettworking event The event was aimed at giving an overview of the
Task Force to the ESA LPS attendees. Presentations about the proposed Simulator ML-OPSI, dedicated
database Ocean Scan, WASP, Marine Litter mission concept MARLISE and RESMALI. An overview of
the scientific session presentations and posters is provided below:
Oral Presentations
Time Title
08:30 FRONTAL: Satellite FRONTs for detection of Anthropogenic plastic Litter

V. Martinez-Vicente et al.
08:45 Observation of Marine Litter Windrows with Sentinel-2/MSI as a Strategic Target for Plastic Pollution
M. Arias et al.
09:00 Satellite remote sensing of marine litter floating in open ocean and coastal waters
Y.-J. Park et al.
09:15 Self-supervised learning for robust floating debris detection

J. Mifdal et al.
09:30 Unraveling the spatial heterogeneity of floating macroplastics at the sea surface
using Unmanned Aerial Vehicles (UAVs)
R. de Vries et al.
09:45 Polarization signatures of nano- and micro-plastics suspended in the water column
simulated at the water surface and top-of-atmosphere levels
T. Harmel et al.
10:40 Plastic Litter Project (PLP) 2021 – calibration and validation data for Sentinel-2 floating marine litter
remote detection
D. Papageorgiou et al.
10:55 Exploring spectral signature unmixing techniques and machine learning algorithms
on fused multi- and hyper-spectral data for plastic marine litter detection – the REACT project
A. Aiello et al.
11:10 Spectral responses meet AI to detect Marine Litter

M. Moshtaghi et al.
11:25 Presenter not available

Marine litter detected by UAV over Patagonian Pristine beaches
C. Mattar
11:40 Experimental tests for the detection and characterisation of Plastic Marine Litter
by means of fluorescence LIDAR technique
V. Raimondi et al.
11:55 Advancing Remote Sensing of Floating Marine Microplastics

H. Dierssen et al.
// 86

Annex
Poster Presentations 
Code Title
62710 Eyes on Plastic – flying high, diving deep to fight aquatic plastic litter
E. Haas et al.
62740 The Use of Spaceborne Radars to Image Ocean Microplastic Dynamics

C. Ruf et al.
63621 Detecting and recognising surface accumulations in Sentinel-2 imagery

T. Kutser et al.
63646 Monitoring of Large Plastic Accumulation Near Dams Using Sentinel-1 Polarimetric SAR Data
M. Simpson et al.
63653 Ocean Global Watcher: Detection of marine anomalies using radar or optical satellite images
A. Lagrange et al.
64067 Identifying macro plastics assisted by close-range hyperspectral remote sensing and deep learning
N. Gnann et al.
64415 Coastal Marine Litter Observatory: drone imagery and AI for marine litter detection
K. Topouzelis et al.
64611 Detection of Marine Plastic Source Locations using Machine Learning applied to Sentinel-1 & Sentinel-2 Data
S. Lavender et al.
64613 River plastics from space: Combining Sentinel-1 and UAV imagery to monitor mixed debris patches
L. Schreyers et al.
65419 Tackling the plastic debris challenge at its source – Linking EO data with multi-source in-situ data for modelling
debris pathways from source to sink
A. Brand et al.
65445 Using hyperspectral radiometry towards subpixel detection of plastic debris on rivers and shorelines.
A. Mata et al.
65637 Feasibility of a satellite mission for monitoring marine macroplastics

S. Livens et al.
66571 Assessment of Machine Learning for floating marine litter detection

L. Fronkova et al.
66724 River plastic monitoring from space: Fact of fiction?

T. van Emmerik et al.
66777 A review of the application of satellite mapping techniques for marine litter monitoring
M. King et al.
66853 PLASTIC MONITOR: Detecting riverine plastic conglomerations, fluxes and pathways in Indonesia
M. Eleveld et al.
66965 Using a drone-based thermal infrared camera to monitor floating plastic litter
L. Goddijn-Murphy et al.
67309 Finding Floating Plastics in Plant Patches using Worldview-3 Satellite Imagery
L. Biermann et al.
67385 Ocean Scan, a marine debris database from Earth and space
L. Romero et al.
67469 Airborne backscatter LIDAR data over the Great Pacific Garbage Patch and their processing
for the Detection of Marine Litter: pros and cons analysis
L. Palombi et al.
// 87

Deutsche Gesellschaft für
Internationale Zusammenarbeit (GIZ) GmbH
Registered offices
Bonn and Eschborn
Friedrich-Ebert-Allee 32 + 36 Dag-Hammarskjöld-Weg 1 – 5

53113 Bonn, Germany 65760 Eschborn, Germany
T +49 228 44 60-0 T +49 61 96 79-0
F +49 228 44 60-17 66 F +49 61 96 79-11 15
E [email protected]
I www.giz.de

Advances in Remote Sensing of Plastic Waste

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Advances in Remote Sensing of Plastic Waste

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Advances in Remote Sensing of Plastic Waste

Uploaded by

Copyright:

Available Formats

Advances in remote sensing

Authors: (in alphabetic order of last names)

Reviewers and Editors:

2 Maturity and Advances in Remote Sensing of Plastics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

Preliminary scientific evidence-based studies have reported recommendations on possibilities to scale up

1.1 Plastic leakage pathways

Figure 1: Pathways and fluxes of plastics into the oceans 6.

Figure 2: Process of monitoring marine litter (own delineation based on 5).

1) Beach litter monitoring is a widely used method Visual observations:

1.3 Remote sensing

› Monitoring animal populations and move-

1.4 History on Remote Sensing of Plastics

Table 1. Examples of space agency initiatives related to remote sensing of plastics.

Discovery Campaign Support innovative and disruptive scientific evidence-based re-

EO4Society Support scientific evidence-based research to advance EOs, better

Plastic-Less Society Explore the feasibility of commercializing information and servic-

1.5 Basics of Remote Sensing

Image Preprocessing Analyses and Classifications Output

Figure 7: L inking Earth Observation data with multi-source in-situ data

Investigating detection of floating plastic litter from orbit

(Source: ESA investigating detection of floating plastic litter from orbit).

Satellite-based marine debris data set

Platform Overview Study

Detection of waste accumulation in Mumbai

2.5 Online Platforms and Web-based Tools

2.5.1 Ocean Scan Zenodo online open-access repository allowing

2.5.2 Litterbase resource for remote sensing validation effort for

2.5.2 Global Partnership for Macroplastic Litter (https://wrd-forecast-

Figure 15: Example showing Global Plastic Watch web interface.

2.6 Stakeholder Community Initiatives

3.1 Prototype Satellite

3.2 Baselines and Standardized Monitoring

3.3 Stakeholder Community Discussions

6. GRID/Arendal Pathways and fluxes of plastics into the oceans. https://www.grida.no/

A1. Scientific Research Projects

Contact Person: Heidi Dierssen ([email protected])

Funding Agency: NASA Ocean Biology and Biogeochemistry

Contact Person: Els Knaeps ([email protected])

Contact Person: Irina Rammos ([email protected])

Website: https://skyflox.eu/, https://www.pml.ac.uk/, https://aufwind.aero/

Contact Person: Robert Foster ([email protected])

Funding Agency: NASA Ocean Biology and Biogeochemistry

Contact Person: Valentina Raimondi, CNR-IFAC – Italy ([email protected])

Contact Person: Delphine Nobileau ([email protected]) – CAPGEMINI

Contact Person: Yolanda Brown ([email protected])

Title: S pectrometer for Marine Litter, ESAPlastics

Contact Person: Hugo Silva ([email protected])

Funding Agency: ESA, the European Space Agency (GSTP)

Contact Person: Jonas Franke, Remote Sensing Solutions ([email protected])

Contact Person: Dr. Clarizia ([email protected])

Contact Person: Aser Mata ([email protected])

Contact Person: Dr José M. Alsina ( [email protected]); Dr Silvia Huber ([email protected])

Website: https://lim.upc.edu/en; https://www.dhi-gras.com/

Contact Person: Mario Castro de Lera ([email protected])

Contact Person: Anne Vallette ([email protected])

Funding Agency: Mercator Ocean (CMEMS)

Contact Person: Marc Lucas ([email protected])

Contact Person: Theodora Papadopoulou ([email protected])

Contact Person: Madeline Cowell ([email protected])

Figure 7: L inking Earth Observation data with multi-source in-situ data

Title: S pectrometer for Marine Litter, ESAPlastics

Title: T hermal Infrared Sensing of marine Plastic Litter, TISPLALI