We present a direct and proxy-based approach to qualitatively and semi-quantitatively observe floating plastic litter in the Great Pacific Garbage Patch (GPGP) based on a survey in 2018 using very high geo-spatial resolution 8-waveband WorldView-3 imagery. A proxy for the plastics was defined as a waveband difference for anomalies in the top-of-the-atmosphere spectra. The anomalies were computed by subtracting spatially varying reflectance of the surrounding ocean water as background from the top-of-the-atmosphere reflectance. Spectral shapes and magnitude were also evaluated using a reference target of known plastics, The Ocean Cleanup System 001 Wilson. Presence of ‘suspected plastics’ was confirmed by the similarity in derived anomalies and spectral shapes with respect to the known plastics in the image as well as direct observations in the true color composites. The proposed proxy-based approach is a step towards future mapping techniques of suspected floating plastics with potential operational monitoring applications from the Sentinel-2 that recently started regular imaging over the GPGP that will be supported or validated by numerical solutions and net trawling survey.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
The Great Pacific Garbage Patch (GPGP) has been identified as one of the largest accumulation zones of floating plastic litter [1, 2]. Continued efforts by environmental agencies, citizens, non-governmental entities and industry are working towards strategies to mitigate the plastics reaching the open ocean. Furthermore, these stakeholders have also been aiming to rid the ocean of this litter through dedicated and opportunity based clean-up activities as well as awareness campaigns. Monitoring during these clean-up or net trawling efforts is mostly based on visual inspection from a vessel, ship or sailing boat that can sometimes be supported by unmanned aerial surveys. These approaches are useful but tend to have limited geo-spatial as well as temporal coverage to allow baselines for operational continuous monitoring to be established.
Imaging floating ocean plastics from space has been gaining raised interest among the key stakeholders of the blue economy. Key objectives behind remote sensing of plastics has been to detect, distinguish, quantify and track these generally harmful aggregated anthropogenic materials [3,4]. Information about the presence of litter has some prospects in improving the efficacy of clean-up activities as well as finding hotspots or accumulation zones at sea. Furthermore, scientific evidence-based knowledge from remote sensing can complement numerical solutions and in-situ surveys. Already, the potential use of satellite sensors has been highlighted in several experimental evidence-based studies in coastal aquatic zones using direct observations [5–7] and garbage patches using indirect observations . However, in the open ocean ongoing work has not yet reported much findings related to direct and optical space based observations of actual verified or suspected plastic litter.
In this work, we present an analyses of very high geo-spatial resolution Maxar Technologies WorldView-3 satellite imagery captured over the GPGP in the visible-near infrared (VNIR) spectrum. The image captured had various targets with distinct spectral features that were considered useful in distinguishing ocean plastics from the ocean itself especially in the GPGP. We also provide an assessment of the capabilities of the WorldView-3 satellite mission in the future monitoring of ocean plastics. The region of interest was selected as it contained System 001 Wilson (600 m long, ∼250 m span) a device designed to collect floating and slightly submerged plastic by the The Ocean Cleanup. Furthermore, anomalies were assessed to infer presence of suspected plastic litter and showcase a proxy-based mapping approach.
2. Methods and materials
2.1 Satellite images
A set of three very high geo-spatial resolution images were captured from the Maxar Technologies WorldView-3 (WV-3) satellite over the GPGP (Table 1). True color (Red-Green-Blue, RGB) composites were generated for each image to visually inspect for signal noise, cloud cover and sea surface state. Further analyses were completed using data captured over calm sea state on 17 September 2018 Image ID:18SEP17195115-S2AS_R3C3-058429719070_01_P001. The other image of the same day covers only part of the System Wilson and provides the same information, which is not shown in this paper.
2.2 Satellite image processing and analyses
Standard orthoready Level 2A GeoTIFF files were retrieved from a dedicated Maxar Technologies WV-3 file repository. These multispectral images had been pansharpened following the University of New Brunswick standard. Each pansharpened image was composed of 8 VNIR spectral wavebands at 0.3 × 0.3 m ground sampling distance (GSD). Note that the GSD of the VNIR bands before pansharpening is typically 1.24 × 1.24 m as reported Maxar Technologies https://gbdxdocs.digitalglobe.com/docs/worldview-3. The wavebands were coastal (427.4 nm), blue (481.9 nm), green (547.1 nm), yellow (604.3 nm), red (660.1 nm), red edge (722.7 nm), NIR-1 (824.0 nm) and NIR-2 (913.6 nm). No contrast, color and atmospheric correction were performed on the delivered imagery but they were radiometrically corrected.
The satellite data was processed through a few steps including conversion of radiance to reflectance, masking clouds, computation of local mean of water background reflectance and computation of anomaly reflectance from the mean. At-sensor or top-of-the-atmosphere (TOA) radiance and was converted into TOA reflectance units as recommended by Maxar Technologies . Masking clouds was done using constant threshold of 0.1 for reflectance at the red waveband. This masking step filtered not only clouds but also boats and brightest parts of the System 001 Wilson. Computing background water reflectance was a challenging procedure since we had no a priori information as to whether a pixel was influenced by floating litter or not. Here we excluded any potential pixel of floating matter as much as possible. We tested each pixel if it had higher value in the NIR-1 - red reflectance than any two neighboring pixels up to 9 pixels apart from the center pixel in the horizontal direction. We do the same in the vertical direction. This test effectively picked up any localized bright pixels including floating litter. For the remaining pixels considered as pure seawater, we computed mean and standard deviation with a kernel size of 5 × 5 pixels. Finally, the mean or background water reflectance was subtracted from the TOA reflectance for each pixel to produce anomaly reflectance spectra. These processing steps were performed using Python 3.
Imagery was further visualized and analyzed using L3Harris Technologies ENVI version 5.6 and spectral analyses were conducted in Mathworks MATLAB version R2016a. Additional, automated anomaly detection was performed in ENVI using a Reed-Xiaoli detector (RXD) algorithm , with a mean spectrum calculation based on the global full dataset with a kernel size of 9 × 9 pixels and a 5% threshold. Best representative spectral information was derived from a 3 × 3 or 1 × 1 pixel window depending on the size of the target. We computed spatial anomalies from the derived TOA spectra as the difference between the signals of floating plastics and seawater as a reference background. We defined this as spatial anomaly of band difference NIR-1 and green. The two wavebands were chosen since the signal of floating plastic is large in the near-infrared wavebands, and the variability of the reflectance of background water is smaller in the green band than in the blue waveband. The assumption was that a pixel with suspected floating plastics had to exhibit a sufficiently large relative anomaly or difference to local standard deviation.
2.3 Optically active targets
We determined three main targets with unique optical signatures in our WV-3 imagery collected over the GPGP area after extensive visual inspection. These targets included the ocean surface, plastics and clouds. GPGP is typically oligotrophic waters suggesting the apparent color of water is blue and reflectance in the NIR is typically negligible [11,12]. The plastics targets we are aware of in this region are based on net trawl counts and model simulations , but from remote sensing platforms in space can be challenging to easily detect and identify hence we defined them as suspected plastics. Net trawling campaigns and other investigations of these floating plastics have revealed they are of varying size classes of different colors composed of mainly yellowish-whitish objects with other color appearing faded due to weathering [13–16]. However, we also had a known plastic target, System 001 Wilson that has been found to be collecting floating litter mainly composed of plastics. Clouds create bright white targets in imagery but the shadows and edge tend to have less reflective spectral characteristics.
3. Results and discussion
3.1 Feature detection in true and false color composite images
The System 001 Wilson, our known plastic target, was captured by the WV-3 satellite on 17 September 2018 at ∼0.3 m geo-spatial resolution (Fig. 1). Visual analyses of the true color composite (R = 660.1 nm, G = 547.1 nm, B = 481.9 nm) proved challenging to easily reveal all optically active targets in the image without prior knowledge of where to find known and suspected plastics (Fig. 1(a)). System 001 Wilson is highlighted in yellow and the region marked in red was therefore assumed to have relatively high amounts of suspected plastics.
A false color composite (R = 824 nm, G = 547.1 nm, B = 481.9 nm) was generated to enhance visualization of the ocean, confirmed plastics, cloud shadow, edge and clouds (Fig. 1(b)). The false color composite findings were also found to be consistent with the product generated by the automated RXD algorithm (Fig. 1(c)). The clouds remained as bright white targets with sharp edges that were initially not clear in the true color composite. Cloud shadows were apparent on the ocean surface that exhibited some surface reflected glint patches. The ocean surface did look bluish despite the false color manipulation. Without highlighting System 001 Wilson attached to a boat, it was more visible in the false color composite as bright yellowish target that was attached to a floating platform and partly in a cloud shadow (Fig. 1(d)). It is evident that for detection activities true color RGB imagery can mislead an observer especially if clouds are captured within a region of interest. It becomes even more complicated with the cloud shadow or thin clouds that might not be distinguishable by the human eye in the visible spectrum. Spectral information is therefore required to further quantitatively identify various targets. As applied in machine learning approaches relevant for plastic litter that ingest three waveband RGB imagery, it is probable that if imagery is gathered in cloud conditions strict measures have to implemented to mitigate false positives in the detection and classification of floating or beached objects.
3.2 Spectral anomalies associated with floating plastics
After distinguishing the suspected floating plastic litter pixels, we obtained a band difference end-product that could be applied as an anomaly proxy to potentially detect and quasi-quantify floating plastics (Fig. 2). The generated map revealed several suspected floating plastic litter object that we further investigated to understand the spectral properties.
A transect was also studied to further explore the prospects of implementing anomaly detection techniques. Here we determined good findings by subtracting the green from the NIR-1 waveband reflectance (Fig. 3). It was noted that the band differences fluctuated spatially corresponding to presence of known and suspected floating plastic litter along the transect highlighted by the white line, with the highest anomalies centered around the System 001 Wilson (Fig. 3). Most of the pixels of the System 001 Wilson showed the NIR1-green band difference (proxy for floating litter) values between 0.001 and 0.005, which is much greater than the noise (variability of background water) of 0.0006, so that the system is clearly visible in the proxy image (Fig. 2).
Without temporal matching sea-truth information we have to assume the anomaly peaks along the transect are indeed suspected floating plastics or oceanic features with spectral characteristics similar to the known plastics. Our assumption about these anomalies is also based on very high geo-spatial resolution (< 0.1 × 0.1 cm pixel) true color images that have been gathered around System 001 Wilson, and newer versions of it, confirming that large amounts of floating and slightly submerged litter accumulates in this area and are being harvested (Fig. 4).
3.3 Spectral characteristics
Spectral analyses of these targets observed in the false color composite (Fig. 1(b)) confirmed some known information about optical characteristics of clouds, seawater and plastics (Fig. 5). Clouds as white colored targets had the highest reflectance as seen by WV-3 whilst the shadow had the lowest signal. Seawater itself is a strong absorber of sunlight in the NIR , had slight higher reflectance compared to the cloud shadow regions, both had almost negligible signal in the NIR. Considering atmospheric reflectance at the NIR, this might indicate a calibration offset. However, it will have no impact on anomaly spectra used in this study. Cloud shadows were observed to exhibit lower reflectance only in the blue-green spectral range but nearly similar to that of seawater reflectance in the NIR. These differences in the blue-green wavebands between the seawater and the seawater covered by cloud shadow confirm that less radiation in the spectrum is reaching the ocean surface.
The suspected plastics are presented here as an example to showcase potential capabilities of detecting plastics from various locations without prior knowledge or sea-truth validation but rather based on spectral analyses. The suspected plastic litter was determined to have relatively high reflectance with respects to the surrounding water but lower than the clouds we assumed as the brightest targets. All TOA reflectance spectra were highest in the blue spectrum and lowest in the NIR influenced by atmospheric Rayleigh scattering, the shape seemed to flatten out as well in the NIR. The cloud pixels had two peak in the red and NIR-1 wavebands but the suspected plastic litter only had the red peak (Fig. 5). The spectral shapes of various targets are discussed further below with the spatial anomaly spectrum, which is TOA reflectance after subtraction of locally varying reflectance of the water background. System 001 Wilson bright portions were noted as having the second highest signal of the selected examples and the dark part also had higher reflectance compared to the surrounding seawater and the cloud shadow region.
Further analyses were conducted by focusing the spectral properties of suspected and known plastic targets (red region in Fig. 1(a)). As discussed earlier, spatial anomaly reflectance per pixel was computed by subtracting local mean of water reflectance from the TOA reflectance. Figure 6 shows example spectra of the spatial anomaly reflectance for various targets. It was evident that the bright parts of the System 001 Wilson had the highest reflectance in comparison the dark parts of the object in the spatial anomaly reflectance (Fig. 6). The suspected plastic litter items were determined to be relatively more reflective than the dark parts of the System 001 Wilson. This spatial anomalous reflectance tends to gradually increase with wavelength, and shows slightly low value in Yellow (604.3 nm, 580-629 nm), Red Edge (722.7 nm, 698-749 nm), and NIR-2 (913.6, 857-1039 nm) bands for particularly bright targets (Fig. 6). Since the anomaly spectra are influenced by both the atmospheric transmittance and the inherent reflectance of the target, it is complicated to determine the surface-level reflectance of plastic debris. At the same time, it does not contradict the results of other studies of the spectral shape of plastics e.g. [3,4,18] whereby the reflectance values increase from the blue wavebands toward the infrared spectrum. Due to the wide bandwidth of the WV-3 sensor, absorption of water vapor, ozone and oxygen affects TOA reflectance. Absorption features are salient in all the spectra located at the Yellow, Red Edge and NIR-2 (Fig. 6). These are consistent with the absorption contribution of atmospheric ozone centered around 600 nm, and water vapor in near infrared .
It has to be noted that detailed spectral analysis are likely to be affected by the presence of the cloud shadow in the captured imagery, hence evaluating the magnitude of the targets can be challenging. A simplified approach to enhance the spectral signatures and shapes in the image could be applying a background correction assuming the ocean or surrounding pixels contribute to the bulk signal observed by the satellite. The approach of a background correction has been widely used [20,21] and would be applicable when calculating the relative reflectance assuming the pixel coverage by each target is known [4,22].
3.4 Prospects of WorldView-3 applications in monitoring floating plastics
In this work, the spectral analyses and a proxy-based mapping approach were showcased using VNIR multispectral imagery as a steps towards exploring the prospects of using other satellite missions (e.g. Landsat-8, Sentinel-2, PlanetScope) with similar capabilities in detect anomalies of suspected plastics. Of course, depending on the size of plastic patches, visual inspection of the true color RGB composites can be used as a human-in-the-loop verification procedure for the proxy-based end-products. The proposed approach was based on TOA spectra which eliminates the need for atmospheric correction as each technique in use has several assumptions that can introduce uncertainties in end-products .
Already, extensive experimental-based studies have characterized the hyperspectral optical properties of virgin and ocean harvested plastic and have revealed diagnostic features primarily exist in the short to longwave infrared (> 1000 nm) spectrum [3–5,24–26]. To this end, detailed analyses related to plastics using multispectral VNIR sensor technologies is challenging , but as presented in this study can be used to provide indicator or proxy information. Enhancing the information that can be harvested from spectral data would require way more than 8 wavebands in VNIR as well as in the shortwave infrared. Alternatively, for mapping purposes at most four wavebands at specific diagnostic absorption features have proven sufficient for algorithms useful plastics litter [23,28,29], suggesting the need for shortwave infrared wavebands located at appropriate parts of the spectrum in satellite mission. At higher geo-spatial resolution (< 10 m), visual analyses and machine learning algorithms can be implemented for object detection generating shape, color or size distribution for the individual macrosized (diameter > 5 mm) plastics or aggregated floating patches. Likewise, for the high to moderate geo-spatial resolution (> 10 m) the goal will be focused on investigating mostly aggregated patches instead of single plastic object.
Mapping of the ocean using very high geo-spatial resolution imagery as proposed here does show good promise although verification would be critical in further studies taking advantage of the GPGP could be useful as it is an accumulation zone that has some already existing known targets. Indeed, the use of WorldView-3 comes at a cost which can be a limiting factor for stakeholders and thus the need to extend this approach to other open-access satellites or the need for space agencies to consider future sensors that combine water quality ocean color remote sensing wavebands including the requirements for plastics. These technical aspects are already under discussion through the efforts of the IOCCG Task Force on Remote Sensing of Marine Litter and Debris 
4. Conclusions and outlook
Observing the ocean surface from optical satellite missions has proven to be useful in monitoring water quality and initial efforts to better understand floating plastics have been explored in this study. Based on this study, it is evident that under clear sky conditions and relatively calm sea surfaces spectral based approaches can generate proxy information that can also be augmented by RGB true color imagery. It is likely that remote sensing supported by numerical models and sea-truth information future imagery captured is expected to advance the prospects of similar 8 waveband information in anomaly detection as a complementary tool for detecting plastics in the GPGP. Of course, extending the imagery capture beyond the VNIR is expected to be more valuable as unique spectral features of plastics are found in the shortwave infrared spectrum (SWIR, >1000 nm). Land-based mapping of objects containing plastics using the VNIR-SWIR wavebands of Maxar Technologies WorldView-3 mission has shown promise [28,29] although little has been reported in the open ocean.
Assessing the geo-spatial anomalies at 0.30 m pixel resolution is assumed to be useful especially in understanding the effects of plastic litter size and pixel coverage that contributes to the bulk signal detected by a sensor. In experimental studies [4,26], the pixel coverage by plastic litter has been shown to be a correlated to the magnitude of reflectance that can be detected using remote sensing tools. Here, we are confident that the full pixels were matching the parts with highest anomalies (Fig. 2) also consistent with our known target of the System 001 Wilson.
Developing operational approaches with the support of other satellite missions is expected to advance since ESA Sentinel-2 imagery capture started over the GPGP in 2021. The future approaches have to be able to distinguish floating plastics in rough sea conditions as this could be much more difficult due to the presence of whitecaps, breaking waves or surface reflected glint . As one of the first reports to showcase the prospects of mapping plastics in the open ocean, the proposed approach is a step towards sustainable remote sensing solutions to the suite of monitoring strategies relevant for ocean plastic litter or the plastic leakage problem. Ongoing efforts are therefore expected to provide detailed spectral based analyses of vast amounts of Sentinel-2 imagery captured over the GPGP with known or verified plastic presence combined with inter-comparison of prior observations from artificial targets or near shore observations of floating or submerged plastic litter.
Ministry of Oceans and Fisheries (20180456); European Space Agency (4000132037/20/NL/ GLC); Deutsche Forschungsgemeinschaft (417276871).
We are grateful to the donors of The Ocean Cleanup. Y-JP was partly supported by the Korea Ministry of Oceans and Fisheries project no. 20180456, “Technology development for Practical Applications of Multi-Satellite data to maritime issues”. SPG was funded through Deutsche Forschungsgemeinschaft grant no. 417276871 and Discovery Element of the European Space Agency’s Basic Activities contract no. 4000132037/20/NL/ GLC.
The authors declare no competing financial interest. The information in this manuscript reflects the views of the authors and does not reflect the official positions of the KIOST and The Ocean Cleanup Foundation.
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
1. L. Lebreton, B. Slat, F. Ferrari, B. Sainte-Rose, J. Aitken, R. Marthouse, S. Hajbane, S. Cunsolo, A. Schwarz, A. Levivier, K. Noble, P. Debeljak, H. Maral, R. Schoeneich-Argent, R. Brambini, and J. Reisser, “Evidence that the Great Pacific Garbage Patch is rapidly accumulating plastic,” Sci. Rep. 8(1), 4666 (2018) [CrossRef] .
2. M. Egger, F. Sulu-Gambari, and L. Lebreton, “First evidence of plastic fallout from the North Pacific Garbage Patch,” Sci. Rep. 10(1), 7495 (2020) [CrossRef] .
3. S. P. Garaba and H. M. Dierssen, “Hyperspectral ultraviolet to shortwave infrared characteristics of marine-harvested, washed-ashore and virgin plastics,”,” Earth Syst. Sci. Data 12(1), 77–86 (2020) [CrossRef] .
4. S. P. Garaba, M. Arias, P. Corradi, T. Harmel, R. de Vries, and L. Lebreton, “Concentration, anisotropic and apparent colour effects on optical reflectance properties of virgin and ocean-harvested plastics,” J. Hazard. Mater. 406, 124290 (2021) [CrossRef] .
5. K. Topouzelis, A. Papakonstantinou, and S. P. Garaba, “Detection of floating plastics from satellite and unmanned aerial systems (Plastic Litter Project 2018),” Int. J. Appl. Earth Obs. Geoinformation 79, 175–183 (2019) [CrossRef] .
6. K. Topouzelis, D. Papageorgiou, A. Karagaitanakis, A. Papakonstantinou, and M. Arias Ballesteros, “Remote sensing of sea surface artificial floating plastic targets with Sentinel-2 and unmanned aerial systems (Plastic Litter Project 2019),” Remote Sens. 12(12), 2013 (2020). [CrossRef]
7. J. P. Matthews, L. Ostrovsky, Y. Yoshikawa, S. Komori, and H. Tamura, “Dynamics and early post-tsunami evolution of floating marine debris near Fukushima Daiichi,” Nat. Geosci. 10(8), 598–603 (2017). [CrossRef]
8. M. C. Evans and C. S. Ruf, “Toward the detection and imaging of ocean microplastics with a spaceborne radar,” IEEE Trans. Geosci. Remote Sens.,1–9 (2021).
9. M. Kuester, “Radiometric use of WorldView-3 imagery. Technical Note,” Digital Globe, USA, (2016).
10. I. S. Reed and X. Yu, “Adaptive multiple-band CFAR detection of an optical pattern with unknown spectral distribution,” IEEE Trans. Acoust., Speech, Signal Process. 38(10), 1760–1770 (1990). [CrossRef]
11. A. Morel, B. Gentili, H. Claustre, M. Babin, A. Bricaud, J. Ras, and F. Tièche, “Optical properties of the “clearest” natural waters,” Limnol. Oceanogr. 52(1), 217–229 (2007). [CrossRef]
12. M. R. Wernand and H. J. van der Woerd, “Ocean colour changes in the North Pacific since 1930,” JEOS:RP 5, 10015s (2010). [CrossRef]
13. M. Filella, “Questions of size and numbers in environmental research on microplastics: methodological and conceptual aspects,” Environ. Chem. 12(5), 527–538 (2015). [CrossRef]
14. D. G. Shaw and R. H. Day, “Colour- and form-dependent loss of plastic micro-debris from the North Pacific Ocean,” Mar. Pollut. Bull. 28(1), 39–43 (1994). [CrossRef]
15. R. H. Day and D. G. Shaw, “Patterns in the abundance of pelagic plastic and tar in the north pacific ocean, 1976–1985,” Mar. Pollut. Bull. 18(6), 311–316 (1987). [CrossRef]
16. E. Martí, C. Martin, M. Galli, F. Echevarría, C. M. Duarte, and A. Cózar, “The colors of the ocean plastics,” Environ. Sci. Technol. 54(11), 6594–6601 (2020). [CrossRef]
17. L. Kou, D. Labrie, and P. Chylek, “Refractive indices of water and ice in the 0.65- to 2.5-μm spectral range,” Appl. Opt. 32(19), 3531–3540 (1993). [CrossRef]
18. M. Moshtaghi, E. Knaeps, S. Sterckx, S. Garaba, and D. Meire, “Spectral reflectance of marine macroplastics in the VNIR and SWIR measured in a controlled environment,” Sci. Rep. 11(1), 5436 (2021). [CrossRef]
19. B.-C. Gao, K. B. Heidebrecht, and A. F. H. Goetz, “Derivation of scaled surface reflectances from AVIRIS data,” Remote Sens. Environ. 44(2-3), 165–178 (1993). [CrossRef]
20. C. Hu, “A novel ocean color index to detect floating algae in the global oceans,” Remote Sens. Environ. 113(10), 2118–2129 (2009). [CrossRef]
21. J. Gower, C. Hu, G. Borstad, and S. King, “Ocean color satellites show extensive lines of floating Sargassum in the Gulf of Mexico,” IEEE Trans. Geosci. Remote Sensing 44(12), 3619–3625 (2006). [CrossRef]
22. B. Russell, H. Dierssen, T. LaJeunesse, K. Hoadley, M. Warner, D. Kemp, and T. Bateman, “Spectral reflectance of Palauan reef-building coral with different symbionts in response to elevated temperature,” Remote Sens. 8(3), 164 (2016). [CrossRef]
23. S. P. Garaba and H. M. Dierssen, “An airborne remote sensing case study of synthetic hydrocarbon detection using short wave infrared absorption features identified from marine-harvested macro- and microplastics,” Remote Sens. Environ. 205, 224–235 (2018). [CrossRef]
24. E. Knaeps, S. Sterckx, G. Strackx, J. Mijnendonckx, M. Moshtaghi, S. P. Garaba, and D. Meire, “Hyperspectral-reflectance dataset of dry, wet and submerged marine litter,”,” Earth Syst. Sci. Data 13(2), 713–730 (2021). [CrossRef]
25. S. P. Garaba, T. Acuña-Ruz, and C. B. Mattar, “Hyperspectral longwave infrared reflectance spectra of naturally dried algae, anthropogenic plastics, sands and shells,”,” Earth Syst. Sci. Data 12(4), 2665–2678 (2020). [CrossRef]
26. L. Goddijn-Murphy, S. Peters, E. van Sebille, N. A. James, and S. Gibb, “Concept for a hyperspectral remote sensing algorithm for floating marine macro plastics,” Mar. Pollut. Bull. 126, 255–262 (2018). [CrossRef]
27. V. Martínez-Vicente, J. R. Clark, P. Corradi, S. Aliani, M. Arias, M. Bochow, G. Bonnery, M. Cole, A. Cózar, R. Donnelly, F. Echevarría, F. Galgani, S. P. Garaba, L. Goddijn-Murphy, L. Lebreton, H. A. Leslie, P. K. Lindeque, N. Maximenko, F.-R. Martin-Lauzer, D. Moller, P. Murphy, L. Palombi, V. Raimondi, J. Reisser, L. Romero, S. G. H. Simis, S. Sterckx, R. C. Thompson, K. N. Topouzelis, E. van Sebille, J. M. Veiga, and A. D. Vethaak, “Measuring marine plastic debris from space: Initial assessment of observation requirements,” Remote Sens. 11(20), 2443 (2019). [CrossRef]
28. X. Guo and P. Li, “Mapping plastic materials in an urban area: Development of the normalized difference plastic index using WorldView-3 superspectral data,” J. Photogramm. Remote Sens. 169, 214–226 (2020). [CrossRef]
29. S. Asadzadeh and C. R. de Souza Filho, “Investigating the capability of WorldView-3 superspectral data for direct hydrocarbon detection,” Remote Sens. Environ. 173, 162–173 (2016). [CrossRef]
30. S. Garaba, M. Arias, L. Biermann, P. Corradi, L. Lorenzoni, F.-R. Martin-Lauzer, V. Martinez-Vicente, D. Mitra, and H. Murakami, “IOCCG Task Force on Remote Sensing of Marine Litter and Debris - Terms of Reference,” https://doi.org/10.5281/zenodo.4446238(Zenodo, 2021), p. 6.
31. H. M. Dierssen and S. P. Garaba, “Bright Oceans: Spectral Differentiation of Whitecaps, Sea Ice, Plastics, and Other Flotsam,” in Recent Advances in the Study of Oceanic Whitecaps: Twixt Wind and Waves, P. Vlahos and E. C. Monahan, eds. https://doi.org/10.1007/978-3-030-36371-0_13(Springer International Publishing, Cham, 2020), pp. 197–208.