Satellite tool
the POS2IDON tool
START
Detecting meter-sized floating macroplastic aggregations in satellite Imagery: the POS2IDON tool
Recent estimates indicate that a total of 3,200 kilotonnes of plastic floats in the ocean, with nearly all mass (95%) contained in macroplastics (>25 mm) (Kaandorp et al. 2023). Floating macroplastics originate from various sources, including mismanaged waste disposal, industrial spills, and abandoned fishing gear. However, extreme weather events and natural disasters can drastically amplify this issue. Heavy rainfall, flooding, and hurricanes can mobilize vast amounts of macroplastics from land, transporting it via rivers into the ocean. This makes it urgent to develop monitoring systems, in particular of floating macroplastics due to their high global mass and consequential hazards.
Floating macro plastic: the plastic tsunamis in Honduras
Observing and tracking floating macroplastics is crucial for mitigation efforts. Modern satellite technology can detect floating features in the sea surface, offering valuable insights into pollution hotspots and transport dynamics. Detecting floating plastics remains a challenge, as plastics tend to disperse. However, when they aggregate into patches influenced by ocean currents and winds, satellites play a key role in their monitoring and detection. Satellite-based detection enables large-scale tracking, helping to distinguish these plastic accumulations from natural ocean materials.
An unprecedent satellite data source: Sentinel-2
Satellite data from the EU Copernicus Sentinel-2 mission offers an unprecedented opportunity due to its higher spatial resolution and open-access availability.Sentinel-2, with its high-resolution optical imagery and unique spectral capabilities, has proven effective in detecting floating marine debris, offering a cost-free, global, and frequent monitoring solution for ocean pollution despite not being originally designed for this purpose.
Info
Spectral Characteristics for differentiating floating materials
Floating macroplastics appear in satellite images as highly reflective and bright pixels compared to the surrounding water, which mostly absorbs solar radiation. However, distinguishing plastic from other floating features—such as algae, sea snot and many more—remains a key challenge due to their similar reflective properties.
Info
Marine Debris
Microalgae Sargassum
Microalgae Noctiluca
Sea Snot
Caroline Power
NOAA
Andrea Giusti
Ceylan Yüceoral (VOA)
spectral curves of different features
Applying AI Machine Learning (ML) models to libraries of spectral signatures help differentiate various objects. We followed a machine learning approach based on the MARIDA library (Kikaki et al., 2022), as well as augmented and consistent versions of this library, and augmented, and consistent, versions of this library. To the verified debris events already existing in MARIDA we added other features with plastic-like signatures such as spume and scum-forming phytoplankton blooms.
By training and testing different ML models our AI-based methodology predicts suspected locations of marine plastic debris accumulations in Sentinel-2 satellite images.
Machine Learning models for differentiating floating materials
Leveraging Spectral Characteristics and Machine Learning for differentiating floating materials
- AI models detect plastic based on its spectral signature and texture.
- Different materials reflect light in unique ways, allowing for differentiation.
- This method improves accuracy compared to traditional visual inspection.
RESULTS: POS2IDON Pipeline for Ocean Features Detection with Sentinel-2
POS2IDON uses AI and Sentinel-2 data to detect marine plastic, providing an open-access, customizable tool for long-term analyses and monitoring of suspected marine plastic debris accumulations (> 10 m) and other ocean features, such as floating macroalgae sargassum.
Region of Interest
FeLs
POS2IDON allow for automatic search and download of Sentinel-2 imagery from the latest repositories.
Download Sentinel-2 L1C Products
Sesing Period
...
CDSE Credentials
Copernicus Data Ecosystem (CDSE)
CDSETool
Others
Free Data Provider Platforms
Python Interfaces
User Imputs
RESULTS: POS2IDON Pipeline for Ocean Features Detection with Sentinel-2
POS2IDON uses AI and Sentinel-2 data to detect marine plastic, providing an open-access, customizable tool for long-term analyses and monitoring of suspected marine plastic debris accumulations (> 10 m) and other ocean features, such as floating macroalgae sargassum.
From: Top of Atmosphere
Different pre-processing and masking steps allow for refinement of the data
To: Rayleigh Atmospheric Corrected
Land Mask ESA World Cover 2021 TerraScope
Water Mask NDWI
Atmospheric Correction ACOLITE
RESULTS: POS2IDON Pipeline for Ocean Features Detection with Sentinel-2
POS2IDON uses AI and Sentinel-2 data to detect marine plastic, providing an open-access, customizable tool for long-term analyses and monitoring of suspected marine plastic debris accumulations (> 10 m) and other ocean features, such as floating macroalgae sargassum.
Sentinel-2 Masked Product
Machine learning models analyze satellite imagery to identify plastic waste.
Random Forest
XGBOOST
U-Net
Split & Mosaic
...
ML Algorithms and Models
Others
User Inputs
Classification Map
Available Machine Learning Algorithms
read the content of the table
RESULTS: POS2IDON models
Several models were trained to analyze different floating materials, aiming to achieve the best performance and representativeness in marine debris detection.Four models were trained—two Random Forest, one XGBoost, and one U-Net—using the MARIDA dataset (Kikaki et al., 2021) with the U-Net architecture achieving the best overall performance. Two decision tree-based models (Random Forest and XGBoost) were trained using an augmented MARIDA spectral library, incorporating new classes like foam and phytoplankton blooms to improve plastic differentiation.
Field Campaign June 2023 (LABPLAS project)
RESULTS: POS2IDON testing
Testing was conducted using artificially deployed targets at sea to evaluate the detection methodology, yielding satisfactory results. Additionally, the acquisition of Very High-Resolution (VHR) images and field observations during the LABPLAS campaign provided valuable insights for refining models and improving accuracy.
+info
RESULTS: POS2IDON applications
Post-disaster management: flooding events in heavily polluted regions is crucial for directing clean-up efforts and assessing debris input into the ocean. POS2IDON’s application to a major plastic debris event in the Gulf of Honduras on the 18 September 2020, where likely marine plastics debris (MD) is detected as red pixels and dots along the river front. Red stars with white borders indicate clusters of at least 10 MD pixels within a 100-meter range.
RESULTS: POS2IDON applications
Post-disaster management: On December 8th 2023, several tonnes of plastic pellets (along with other debris) were released from containers on a ship near the Portugal-Spain border. Strong south winds carried the pellets to Galicia, prompting analysis using Sentinel-2 data from the 13th, 18th, and 23rd with POS2IDON. Suspicious features were detected on the 18th, with drift model simulations aiding validation. Given the vast affected area, pinpointing specific locations with higher pellet concentrations remains a priority.
+info
RESULTS: POS2IDON applications
Long-term seasonal analysis leveraging Sentinel-2’s regular 5-day imagery since 2018 helps reveal patterns and trends in marine debris. Using POS2IDON, 72 images from the Gulf of Honduras in 2020 were analyzed. Results show that drier seasons (winter and spring) have fewer debris detections, aligning with the role of rivers as a major pollution source.
RESULTS: POS2IDON applications
POS2IDON applied in the less polluted LABPLAS region in the German Bight (North Sea), where we had ground-truth vessel-based information. When using MARIDA-only models, we suspect recurrent confusion of plastic debris with filaments of Noctiluca scintillans, which was circumvented with a specific XGBoost model trained with Noctiluca scintillans signatures.
Comparison between RGB image (left), and the U-Net (MARIDA-only, middle) and XGBoost (MARIDA-augmented, right) models for the 16 July 2020 in German Bight. Note that U-Net classified the filaments as Marine Plastic Debris (MD, large red points), whereas the XGBoost classified the filaments as Noctiluca Blooms (small purple points).
+info
conclusions
In summary, POS2IDON demonstrates the potential of combining satellite data and machine learning to detect suspected plastic pollution across marine environments. By offering an open-source, scalable solution, this tool contributes to the global effort to monitor and mitigate plastic pollution from space.
+info
References and Further Reading
Valente, E. Castanho, A. Giusti, J. Pinelo and P. Silva, "An Open-Source Data Pipeline Framework to Detect Floating Marine Plastic Litter Using Sentinel-2 Imagery and Machine Learning," IGARSS 2023 - 2023 IEEE International Geoscience and Remote Sensing Symposium, Pasadena, CA, USA, 2023, pp. 4108-4111, doi: 10.1109/IGARSS52108.2023.10281415.
Kikaki K, Kakogeorgiou I, Mikeli P, Raitsos DE, and Karantzalos K (2022), MARIDA: A benchmark for marine debris detection from Sentinel-2 remote sensing data. PLoS One 17:e0262247. doi: 10.1371/journal.pone.0262247.
Hu C (2021), Remote detection of marine debris using satellite observations in the visible and near-infrared spectral range: Challenges and potentials. Remote Sensing of Environment 259:112414. doi: 10.1016/j.rse.2021.112414.
Topouzelis K, Papageorgiou D, Suaria G, and Aliani S (2021), Floating marine litter detection algorithms and techniques using optical remote sensing data: A review. Mar. Pollut. Bull. 170:112675. doi:10.1016/j.marpolbul.2021.112675.
Biermann L, Clewley D, Martinez-Vicente V, and Topouzelis K (2020), Finding plastic patches in coastal waters using optical satellite data. Sci. Rep. 10:5364. doi: 10.1038/s41598-020-62298-z.
Martinez-Vicente V (2022), The need for a dedicated marine plastic litter satellite mission. Nature Reviews Earth & Environment 3:728–729. doi: 10.1038/s43017-022-00360-2.
Topouzelis K, Papakonstantinou A, and Garaba SP (2019), Detection of floating plastics from satellite and unmanned aerial systems (Plastic Litter Project 2018). Int. J. Appl. Earth Obs. Geoinf. 79:175–183. doi:10.1016/j.jag.2019.03.011.
Kaandorp, M.L.A., Lobelle, D., Kehl, C. et al. Global mass of buoyant marine plastics dominated by large long-lived debris. Nat. Geosci. 16, 689–694 (2023)
FOLLOW US!
🛰️Sentinel-2 provides high-resolution optical imagery for environmental monitoring of larger areas, as well as remotely, hard-to-reach areas 🛰️Multi-Spectral Instrument covers wavelengths of vis-NIR-SWIR, suitable for detecting and differentiating small floating matters. 🛰️ provides a trade between spatial resolution (10–20 m) and revisit frequency (5 days) for most coastal waters.
(Kikaki et al., 2020 Photos: Caroline Power)
Testing on PLP (Topouzelis et al, 2021) artificial targets
While macroplastic aggregations can be detected from space, regional models and better ground-truth data (e.g. recurrency and macroplastic density in these meter-sized surface aggregations of macroplastics) are essential to advance.
RESULTS: POS2IDON applications
Scan the QR code and discover much more
Satelite Tool
Estibaliz Garmendia
Created on March 26, 2025
Start designing with a free template
Discover more than 1500 professional designs like these:
View
Florida Neon Presentation
View
Blackboard Presentation
View
Genial Storytale Presentation
View
Historical Presentation
View
Psychedelic Presentation
View
Memories Presentation
View
Animated Chalkboard Presentation
Explore all templates
Transcript
Satellite tool
the POS2IDON tool
START
Detecting meter-sized floating macroplastic aggregations in satellite Imagery: the POS2IDON tool
Recent estimates indicate that a total of 3,200 kilotonnes of plastic floats in the ocean, with nearly all mass (95%) contained in macroplastics (>25 mm) (Kaandorp et al. 2023). Floating macroplastics originate from various sources, including mismanaged waste disposal, industrial spills, and abandoned fishing gear. However, extreme weather events and natural disasters can drastically amplify this issue. Heavy rainfall, flooding, and hurricanes can mobilize vast amounts of macroplastics from land, transporting it via rivers into the ocean. This makes it urgent to develop monitoring systems, in particular of floating macroplastics due to their high global mass and consequential hazards.
Floating macro plastic: the plastic tsunamis in Honduras
Observing and tracking floating macroplastics is crucial for mitigation efforts. Modern satellite technology can detect floating features in the sea surface, offering valuable insights into pollution hotspots and transport dynamics. Detecting floating plastics remains a challenge, as plastics tend to disperse. However, when they aggregate into patches influenced by ocean currents and winds, satellites play a key role in their monitoring and detection. Satellite-based detection enables large-scale tracking, helping to distinguish these plastic accumulations from natural ocean materials.
An unprecedent satellite data source: Sentinel-2
Satellite data from the EU Copernicus Sentinel-2 mission offers an unprecedented opportunity due to its higher spatial resolution and open-access availability.Sentinel-2, with its high-resolution optical imagery and unique spectral capabilities, has proven effective in detecting floating marine debris, offering a cost-free, global, and frequent monitoring solution for ocean pollution despite not being originally designed for this purpose.
Info
Spectral Characteristics for differentiating floating materials
Floating macroplastics appear in satellite images as highly reflective and bright pixels compared to the surrounding water, which mostly absorbs solar radiation. However, distinguishing plastic from other floating features—such as algae, sea snot and many more—remains a key challenge due to their similar reflective properties.
Info
Marine Debris
Microalgae Sargassum
Microalgae Noctiluca
Sea Snot
Caroline Power
NOAA
Andrea Giusti
Ceylan Yüceoral (VOA)
spectral curves of different features
Applying AI Machine Learning (ML) models to libraries of spectral signatures help differentiate various objects. We followed a machine learning approach based on the MARIDA library (Kikaki et al., 2022), as well as augmented and consistent versions of this library, and augmented, and consistent, versions of this library. To the verified debris events already existing in MARIDA we added other features with plastic-like signatures such as spume and scum-forming phytoplankton blooms. By training and testing different ML models our AI-based methodology predicts suspected locations of marine plastic debris accumulations in Sentinel-2 satellite images.
Machine Learning models for differentiating floating materials
Leveraging Spectral Characteristics and Machine Learning for differentiating floating materials
RESULTS: POS2IDON Pipeline for Ocean Features Detection with Sentinel-2
POS2IDON uses AI and Sentinel-2 data to detect marine plastic, providing an open-access, customizable tool for long-term analyses and monitoring of suspected marine plastic debris accumulations (> 10 m) and other ocean features, such as floating macroalgae sargassum.
Region of Interest
FeLs
POS2IDON allow for automatic search and download of Sentinel-2 imagery from the latest repositories.
Download Sentinel-2 L1C Products
Sesing Period
...
CDSE Credentials
Copernicus Data Ecosystem (CDSE)
CDSETool
Others
Free Data Provider Platforms
Python Interfaces
User Imputs
RESULTS: POS2IDON Pipeline for Ocean Features Detection with Sentinel-2
POS2IDON uses AI and Sentinel-2 data to detect marine plastic, providing an open-access, customizable tool for long-term analyses and monitoring of suspected marine plastic debris accumulations (> 10 m) and other ocean features, such as floating macroalgae sargassum.
From: Top of Atmosphere
Different pre-processing and masking steps allow for refinement of the data
To: Rayleigh Atmospheric Corrected
Land Mask ESA World Cover 2021 TerraScope
Water Mask NDWI
Atmospheric Correction ACOLITE
RESULTS: POS2IDON Pipeline for Ocean Features Detection with Sentinel-2
POS2IDON uses AI and Sentinel-2 data to detect marine plastic, providing an open-access, customizable tool for long-term analyses and monitoring of suspected marine plastic debris accumulations (> 10 m) and other ocean features, such as floating macroalgae sargassum.
Sentinel-2 Masked Product
Machine learning models analyze satellite imagery to identify plastic waste.
Random Forest
XGBOOST
U-Net
Split & Mosaic
...
ML Algorithms and Models
Others
User Inputs
Classification Map
Available Machine Learning Algorithms
read the content of the table
RESULTS: POS2IDON models
Several models were trained to analyze different floating materials, aiming to achieve the best performance and representativeness in marine debris detection.Four models were trained—two Random Forest, one XGBoost, and one U-Net—using the MARIDA dataset (Kikaki et al., 2021) with the U-Net architecture achieving the best overall performance. Two decision tree-based models (Random Forest and XGBoost) were trained using an augmented MARIDA spectral library, incorporating new classes like foam and phytoplankton blooms to improve plastic differentiation.
Field Campaign June 2023 (LABPLAS project)
RESULTS: POS2IDON testing
Testing was conducted using artificially deployed targets at sea to evaluate the detection methodology, yielding satisfactory results. Additionally, the acquisition of Very High-Resolution (VHR) images and field observations during the LABPLAS campaign provided valuable insights for refining models and improving accuracy.
+info
RESULTS: POS2IDON applications
Post-disaster management: flooding events in heavily polluted regions is crucial for directing clean-up efforts and assessing debris input into the ocean. POS2IDON’s application to a major plastic debris event in the Gulf of Honduras on the 18 September 2020, where likely marine plastics debris (MD) is detected as red pixels and dots along the river front. Red stars with white borders indicate clusters of at least 10 MD pixels within a 100-meter range.
RESULTS: POS2IDON applications
Post-disaster management: On December 8th 2023, several tonnes of plastic pellets (along with other debris) were released from containers on a ship near the Portugal-Spain border. Strong south winds carried the pellets to Galicia, prompting analysis using Sentinel-2 data from the 13th, 18th, and 23rd with POS2IDON. Suspicious features were detected on the 18th, with drift model simulations aiding validation. Given the vast affected area, pinpointing specific locations with higher pellet concentrations remains a priority.
+info
RESULTS: POS2IDON applications
Long-term seasonal analysis leveraging Sentinel-2’s regular 5-day imagery since 2018 helps reveal patterns and trends in marine debris. Using POS2IDON, 72 images from the Gulf of Honduras in 2020 were analyzed. Results show that drier seasons (winter and spring) have fewer debris detections, aligning with the role of rivers as a major pollution source.
RESULTS: POS2IDON applications
POS2IDON applied in the less polluted LABPLAS region in the German Bight (North Sea), where we had ground-truth vessel-based information. When using MARIDA-only models, we suspect recurrent confusion of plastic debris with filaments of Noctiluca scintillans, which was circumvented with a specific XGBoost model trained with Noctiluca scintillans signatures.
Comparison between RGB image (left), and the U-Net (MARIDA-only, middle) and XGBoost (MARIDA-augmented, right) models for the 16 July 2020 in German Bight. Note that U-Net classified the filaments as Marine Plastic Debris (MD, large red points), whereas the XGBoost classified the filaments as Noctiluca Blooms (small purple points).
+info
conclusions
In summary, POS2IDON demonstrates the potential of combining satellite data and machine learning to detect suspected plastic pollution across marine environments. By offering an open-source, scalable solution, this tool contributes to the global effort to monitor and mitigate plastic pollution from space.
+info
References and Further Reading
Valente, E. Castanho, A. Giusti, J. Pinelo and P. Silva, "An Open-Source Data Pipeline Framework to Detect Floating Marine Plastic Litter Using Sentinel-2 Imagery and Machine Learning," IGARSS 2023 - 2023 IEEE International Geoscience and Remote Sensing Symposium, Pasadena, CA, USA, 2023, pp. 4108-4111, doi: 10.1109/IGARSS52108.2023.10281415. Kikaki K, Kakogeorgiou I, Mikeli P, Raitsos DE, and Karantzalos K (2022), MARIDA: A benchmark for marine debris detection from Sentinel-2 remote sensing data. PLoS One 17:e0262247. doi: 10.1371/journal.pone.0262247. Hu C (2021), Remote detection of marine debris using satellite observations in the visible and near-infrared spectral range: Challenges and potentials. Remote Sensing of Environment 259:112414. doi: 10.1016/j.rse.2021.112414. Topouzelis K, Papageorgiou D, Suaria G, and Aliani S (2021), Floating marine litter detection algorithms and techniques using optical remote sensing data: A review. Mar. Pollut. Bull. 170:112675. doi:10.1016/j.marpolbul.2021.112675. Biermann L, Clewley D, Martinez-Vicente V, and Topouzelis K (2020), Finding plastic patches in coastal waters using optical satellite data. Sci. Rep. 10:5364. doi: 10.1038/s41598-020-62298-z. Martinez-Vicente V (2022), The need for a dedicated marine plastic litter satellite mission. Nature Reviews Earth & Environment 3:728–729. doi: 10.1038/s43017-022-00360-2. Topouzelis K, Papakonstantinou A, and Garaba SP (2019), Detection of floating plastics from satellite and unmanned aerial systems (Plastic Litter Project 2018). Int. J. Appl. Earth Obs. Geoinf. 79:175–183. doi:10.1016/j.jag.2019.03.011. Kaandorp, M.L.A., Lobelle, D., Kehl, C. et al. Global mass of buoyant marine plastics dominated by large long-lived debris. Nat. Geosci. 16, 689–694 (2023)
FOLLOW US!
🛰️Sentinel-2 provides high-resolution optical imagery for environmental monitoring of larger areas, as well as remotely, hard-to-reach areas 🛰️Multi-Spectral Instrument covers wavelengths of vis-NIR-SWIR, suitable for detecting and differentiating small floating matters. 🛰️ provides a trade between spatial resolution (10–20 m) and revisit frequency (5 days) for most coastal waters.
(Kikaki et al., 2020 Photos: Caroline Power)
Testing on PLP (Topouzelis et al, 2021) artificial targets
While macroplastic aggregations can be detected from space, regional models and better ground-truth data (e.g. recurrency and macroplastic density in these meter-sized surface aggregations of macroplastics) are essential to advance.
RESULTS: POS2IDON applications
Scan the QR code and discover much more