论文信息

Mona I. A. Kaamoush / Ahmed El-Zeiny

Discover Sustainability

论文解读

AI 辅助整理
方法
综述计算机视觉、深度学习、CNN、Transformer及IoT与无人机遥感融合
发现
AI结合遥感与IoT可高精度检测油污和塑料垃圾并实现实时监测。
机会点
可借鉴其遥感+深度学习框架,迁移至农业面源污染与水体富营养化监测。

论文摘要

摘要:在全球范围内,由石油泄漏、塑料垃圾和工业废弃物造成的海洋污染对环境、经济和人类健康构成了重大威胁。现场调查、光学传感和人工采样等传统监测与缓解技术,往往因效率低下、精度不足和覆盖范围有限而受到制约,在大尺度或偏远环境中尤为明显。人工智能(AI)的最新进展为海洋污染的检测、管理和监测带来了革命性潜力。基于AI的方法,如计算机视觉、深度学习和机器学习,已与无人机影像、遥感和原位监测有效结合,能够以极高的精度检测、分类和量化塑料污染物与石油泄漏。卷积神经网络(CNN)、基于Transformer的模型、模糊系统和神经网络均改进了污染物检测、热点制图和预测建模,同时应对了噪声数据集和多样环境条件等挑战。此外,融合机器人机制和物联网(IoT)传感器的AI驱动自主系统,正在催生可扩展的实时海洋废弃物检测与处置解决方案。本研究全面考察了AI在海洋污染物识别中的应用现状,重点阐述了方法学进展,并详细展望了将AI融入可持续海洋监测与污染管理策略的未来潜力。

全文

Discover Sustainability

Discover SustainabilityKaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
https://doi.org/10.1007/s43621-026-04956-2
REVIEW Open Access
Artificial intelligence for marine pollution
monitoring and control
Mona Kaamoush1* and Ahmed El-Zeiny2
*Correspondence:
Mona Kaamoush
monakaamoush@aast.edu
1Environmental Protection and
Crises Management Department,
Simulators Complex, Arab Academy
for Science, Technology and
Maritime Transport, Alexandria,
Egypt
2Environmental Studies
Department, National Authority
for Remote Sensing and Space
Sciences (NARSS), Cairo, Egypt
© The Author(s) 2026. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use,
sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article
are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the
article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to
obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Abstract Globally, marine pollution generated by oil spills, plastic litter, and industrial
waste poses major threats to the environment, the economy, and human health. Field
surveys, optical sensing, and manual sampling are examples of traditional monitoring
and mitigation techniques that are frequently hampered by inefficiency, poor
precision, and restricted coverage, particularly in large-scale or remote circumstances.
Recent advances in artificial intelligence (AI) hold revolutionary potential for detecting,
managing, and monitoring maritime pollution. AI-based methods such as computer
vision, deep learning, and machine learning have been effectively combined with
drone imagery, remote sensing, and in situ monitoring to detect, classify, and measure
plastic contaminants and oil spills with remarkable precision. Convolutional Neural
Networks (CNNs), transformer-based models, fuzzy systems, and neural networks
have all improved pollutant detection, hotspot mapping, and predictive modeling
while addressing challenges including noisy datasets and diverse environmental
conditions. Furthermore, scalable solutions for real-time maritime waste detection and
disposal are emerging from AI-driven autonomous systems that incorporate robotic
mechanisms and Internet of Things (IoT) sensors. This study fully examines the current
state of AI applications for marine pollutant identification, highlighting methodological
advancements and detailing future potential for integrating AI into sustainable ocean
monitoring and pollution management strategies.
Keywords Artificial intelligence, Marine pollution, Oil spill, Plastic pollution, Remote
sensing, Sustainable ocean monitoring
1 Introduction
Globally, marine pollution is steadily on the rise and poses a serious threat to the sus -
tainable use of seas. The management, reduction, and eradication of marine pollution
are currently the focus of multifaceted and sustainable ocean administration. Monitor -
ing accounts for 57% of AI applications in marine pollution, management for 24%, and
prediction for 19% [ 1]. AI is predicted to completely transform how people work, play,
and live in the future. AI has the potential to be used in several fields that are important
to sustainable development [2, 3].
In addition, both natural processes and human activities contribute to marine pol -
lution. Major human-generated marine pollutants include plastics and atmospheric

Page 2 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
aerosols, while human activities such as mining and exploration, urbanization and
industrialization, and agricultural practices represent important sources of these pol -
lutants entering the marine environment [ 4]. Although natural processes also contrib -
ute to contamination of aquatic environments, their impact is generally less significant
than that of anthropogenic activities [5]. According to earlier research, contamination of
the marine environment has adversely affected many facets of both human and animal
lives. Therefore, controlling this pollution is essential (Fig.  1). Achieving these objectives
depends on the accurate and timely identification of marine pollutants, which provides
the foundation for effective monitoring, risk assessment, and environmental manage -
ment [3].
Conventional monitoring techniques alone are frequently inadequate for quick and
extensive environmental assessment due to the increasing complexity and geographi -
cal scope of maritime pollution. As a result, artificial intelligence (AI) has become a
potent instrument for enhancing marine pollution detection, monitoring, prediction,
and management. AI gives academics sophisticated tools to detect pollution sources,
evaluate intricate environmental data, forecast pollution patterns, and assist in making
evidence-based decisions. Additionally, through intelligent modelling and automated
data analysis, AI speeds up the creation of efficient mitigation methods. In recent years,
there has been a significant increase in the use of AI technology for monitoring and con-
trolling marine pollution. Nevertheless, despite these developments, comparatively few
thorough review studies have methodically assessed the present situation and potential
of AI applications in maritime pollution. Simultaneously, cloud computing has made
it possible for various stakeholders, such as governmental organizations and academic
institutions, to effectively share computational power and storage resources, facilitating
extensive AI-based environmental monitoring [6].
Water quality evaluation and prediction is one of the most significant uses of AI in
the management of marine pollution. For water authorities to analyze environmental
Graphical abstract

Page 3 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
risks and make well-informed decisions, accurate water quality estimation is essential.
By enhancing data interpretation, knowledge integration, and prediction capabilities,
AI-driven decision support systems improve water quality management [ 7]. Numer -
ous studies have successfully developed predictive models for identifying Hg 2+, bacterial
concentrations, nitrogen, and ammonia in marine environments [ 8]. For instance, Pen -
nacchio et al. [ 9] have developed a portable, user-friendly biosensor that uses machine
learning to determine Hg 2+ levels in seawater, addressing the serious threat that mer -
cury contamination poses to both people and ecosystems. Guo et al. [ 10] and El-Zeiny
et al. [ 11] were capable of to identify and forecast levels of bacteria in seashore water
around-the-clock using water quality predicting models, which helped them make deci -
sions for the protection of public health [ 10]. Using a machine learning model, Ghayas
et al., [8] assessed the water quality along the Karachi coastline by measuring ammonia
concentrations. They provided decision support tools to help policymakers create solu -
tions for particular water quality issues [ 12]. The development of environmental regula -
tions will be possible after these accomplishments. More comprehensive and in-depth
insights are obtained by decision makers that integrate AI in their water quality control
actions (Fig. 2), which helps them better handle rapidly transforming water quality envi -
ronments, in addition to increased decision-making effectiveness and precision [ 1].
One of the worst types of marine pollution is oil spills, which have serious negative
effects on the environment, the economy, and society. The main causes of significant oil
spills (> 7 tons) observed between 1970 and 2023 have been determined to be collisions,
hull failures, operating events, equipment faults, fires, and explosions [ 13, 14]. Oil spills
have both immediate and long-term effects on ecosystems. Fish, sea turtles, seabirds,
and plankton are among the marine creatures that can suffer significant short-term
death from direct oil exposure [ 15, 16]. The oil's effects on the ecosystem are prolonged
over time by weathering processes that are fueled by wind, waves, and saltwater. Loss
of biodiversity, disturbance of marine food webs, deterioration of coastal ecosystems,
Fig. 1 Comparison between traditional marine pollution monitoring methods and AI-driven approaches based
on the systematic literature review of Ning et al. [1]

Page 4 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
and ongoing financial difficulties for coastal people are some of these long-term conse -
quences [17]. Therefore, it is crucial to identify oil spills quickly and accurately in order
to minimize environmental damage, improve the efficacy of maritime pollution manage-
ment, and enable prompt reaction actions.
One issue is the dearth of reasonably priced equipment for operationally monitoring
such local situations. Due of their expense and spatial resolution, the techniques and
equipment utilized for global disasters cannot be applied in this situation. The MARPOL
Convention states that bilge waters can include no more than 15 ppm of oil products.
But if levels of dissolved oil products surpass 15 ppm, The surface of the sea above the
water release point is not usually where an oil slick forms. The formation of a slick is
determined by the concentration of oil products in the bilge fluids as well as the hydrol -
ogy properties of the saltwater into which the spillage is made. Therefore, in addition
to ensuring that oil product slicks are detected on the sea surface, measuring their vol -
ume, and allowing the measurement of the concentration of dissolved oil products in
the subsurface layer, monitoring tools should also identify the varieties of oil products
present in slicks or methods in order to pinpoint the root cause of pollution. Above the
water discharge point, an oil slick does not usually form on the water's surface. Both the
concentration of oil products in the bilge fluids and the hydrological properties of the
saltwater into which the release is made influence the formation of a slick [18].
The ongoing release of massive amounts of plastic debris into the marine environment
poses a growing threat to our seas and the marine life they support. The health and sus -
tainability of marine ecosystems are seriously threatened by the estimated eight million
metric tons of plastic that enter the oceans annually. Marine biodiversity is put at greater
risk by plastic pollution, which intensifies already-existing environmental pressures
including overfishing, climate change, and other human disturbances [ 19]. The subject
of interest is the identification of anthropogenic garbage produced in coastal and marine
habitats using aerial and underwater drones, utilizing computational intelligence (CI),
deep learning (DL), machine learning (ML), and computer vision (CV) [ 20].
Fig. 2 Using AI to determine marine water quality for decision support

Page 5 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
The enterprises engaged in coastal development suffer from a number of problems
brought on by marine debris in the maritime environment. Marine litter is defined as
man-made materials that are dumped into the ocean or find their way there through
rivers, household or business effluent, or both. Human health and coastal ecosystems
are seriously threatened by plastic waste on beaches. The entanglement and consump -
tion of plastic trash by marine species can cause harm, decreased reproductive success,
and even death, all of which have a negative impact on beach biodiversity. Furthermore,
microplastics and related contaminants may enter the marine food web and present
health risks to humans through seafood consumption, while plastic waste deteriorates
coastal environments and increases beach goers' risks due to the presence of sharp or
dangerous debris [21]. The need of efficient monitoring and management techniques to
lessen plastic pollution in marine and coastal ecosystems is underscored by these envi -
ronmental and public health issues [22].
Recent breakthroughs in artificial intelligence have altered maritime environmental
monitoring by allowing for automated pollutant detection, classification, forecast, and
decision assistance. Machine learning algorithms, such as Support Vector Machines
(SVMs), Random Forests (RFs), Artificial Neural Networks (ANNs), and fuzzy systems,
have been widely used for environmental prediction and water quality monitoring. Deep
learning techniques, particularly Convolutional Neural Networks (CNNs), You Only
Look Once (YOLO) object detection models, and transformer-based architectures, have
recently demonstrated remarkable capabilities for detecting oil spills, floating plastics,
marine litter, and microplastics in satellite imagery, unmanned aerial vehicle (UAV)
images, underwater cameras, and hyperspectral data. Additionally, real-time environ -
mental monitoring, predictive modeling, and intelligent decision support for sustainable
marine pollution management are increasingly being provided by AI-integrated Internet
of Things (IoT) sensor networks, autonomous surface and underwater vehicles, com -
puter vision systems, and digital twins [23–26].
Recent reviews highlight the increasing role of Artificial Intelligence (AI) in addressing
marine pollution arising from rapid industrial growth. While remote sensing and moni -
toring technologies have improved marine environmental assessment, the large volume
and complexity of collected data remain a major challenge. AI offers effective tools for
processing, classifying, and interpreting big environmental datasets, thereby enhancing
the accuracy and efficiency of marine pollution monitoring [ 27]. Artificial intelligence
(AI) applications have shown great promise for detecting pollution trends and assist -
ing with the sustainable management of marine habitats. For instance, AI methods have
been used to analyse data from environmental monitoring and remote sensing, find pol -
lution hotspots, locate marine debris and oil spills, and assist in real-time environmen -
tal evaluation and decision-making. Compared to traditional monitoring techniques,
these capabilities allow for the quicker and more reliable analysis of huge and complex
datasets, improving pollution surveillance and offering important scientific evidence to
promote efficient management of maritime pollution and environmental protection [ 28,
29].
Recent advances in edge artificial intelligence have enabled efficient and real-time
classification of marine pollution. Palanisamy et al. [ 25, 30] proposed a customized con -
volutional neural network (CNN) for underwater image-based marine pollution clas -
sification, achieving superior performance compared with established deep learning

Page 6 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
architectures such as VGG, Inception, and DenseNet models. Applications of artificial
intelligence (AI) have demonstrated significant potential in identifying patterns of pol -
lution and supporting the sustainable management of maritime environments. Recent
advancements include the integration of AI with edge computing, deep learning, and
remote sensing to allow rapid and accurate monitoring of maritime pollution. This
model, which is characterized by a customized convolutional neural network (CNN)
that was successfully implemented on an Edge AI platform and obtained very high clas -
sification accuracy for marine pollution identification, serves as an example of its use
for real-time environmental monitoring. These advancements demonstrate how AI-
based methods can overcome the limitations of conventional monitoring techniques
by improving the speed, accuracy, and automation of pollution identification while pro -
viding timely decision support for maritime environmental management and pollution
mitigation.
Marine pollution from plastics, oil spills, and industrial waste presents significant eco -
logical challenges, with conventional cleanup methods often limited in efficiency and
adaptability. Artificial intelligence (AI) applications have shown great promise for detect-
ing pollution trends and assisting in the sustainable management of marine habitats [ 31,
32, 33]. In order to enable intelligent maritime waste monitoring and disposal, recent
developments have combined AI with autonomous robotic systems, computer vision,
and Internet of Things (IoT) technology. In order to facilitate effective marine trash
detection and collection, Subramani et al. [ 26] created an AI-integrated autonomous
marine cleanup system that integrates intelligent waste classification with real-time pic -
ture processing. High accuracy in pollution identification and categorization was shown
in the performance evaluation, demonstrating the potential of AI-driven autonomous
systems to offer dependable real-time monitoring while lowering the requirement for
ongoing human interaction. These results show that by increasing detection efficiency,
facilitating scalable marine trash management, and enabling proactive, data-driven envi-
ronmental decision-making, AI-based systems can overcome many of the drawbacks of
traditional monitoring techniques.
The aim of this review is to thoroughly investigate AI-based approaches for the detec -
tion, categorization, tracking, and control of man-made marine pollution. It outlines
future research directions for integrating AI-based methodologies into sustainable
marine pollution monitoring and environmental management, highlights recent meth -
odological advancements, assesses current applications across major pollutant catego -
ries, and discusses current challenges and limitations.
1.1 Novelty statement
This review offers a thorough synthesis of AI-based approaches for the identification,
classification, monitoring, prediction, and management of various anthropogenic marine
pollutants, in contrast to earlier reviews that mainly focus on individual marine pollut -
ants or particular AI techniques. In a variety of pollution scenarios, it compares com -
puter vision, machine learning, deep learning, transformer-based models, IoT-enabled
monitoring, autonomous robotic systems, and remote sensing technologies. Further -
more, the analysis specifically links AI-driven marine pollution management to the Sus -
tainable Development Goals (SDGs) of the United Nations while highlighting existing
research gaps, implementation issues, useful industry applications, and future research

Page 7 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
objectives. Researchers, decision-makers, and maritime stakeholders may advance intel -
ligent marine environmental monitoring and sustainable ocean governance with the
help of this integrated perspective.
1.1.1 Importance of the marine environment and the maritime industry
Maintaining the global ecological balance and sustaining important maritime industries
like shipping, fishing, offshore energy, and coastal tourism depend heavily on the marine
environment. Oceans maintain marine biodiversity, control the climate, and offer eco -
system services that support human welfare and economic growth. The economic sig -
nificance of healthy marine ecosystems is highlighted by the fact that almost 90% of the
world's trade is carried out by sea. However, growing industrial activity, coastal expan -
sion, and marine traffic have increased environmental pressures, leading to deteriorating
environmental quality and growing hazards to the sustainability of maritime opera -
tions [34]. According to Aretoulaki et al. [35], marine pollution also results in significant
socioeconomic costs due to decreased fisheries production, decreased tourism income,
higher port and beach cleanup expenses, and harm to coastal infrastructure and mari -
time enterprises.
Due to its extensive distribution, lengthy environmental persistence, and detrimental
effects on marine life and coastal ecosystems, plastic litter has become one of the most
enduring environmental problems among the main causes of marine pollution. Oil spills
provide a serious concern in addition to plastic pollution, resulting in both acute ecolog-
ical harm and long-term destruction of marine and coastal environments. Additionally,
chemical pollutants discharged from maritime operations, such as antifouling biocides
like Irgarol 1051, can affect marine microalgae growth, photosynthetic efficiency, and
metabolism, which can disrupt primary productivity and jeopardize the stability of
marine ecosystems [36]. Collectively, these diverse pollution sources highlight the com -
plexity of marine environmental degradation and the need for integrated monitoring,
early detection, and sustainable management strategies to protect marine ecosystems
and maritime industries' long-term resilience.
Marine pollution such as plastics, oil spills, medications, and heavy metals have a
direct influence on port operations, navigation safety, and ecosystem health. Macrod -
ebris and floating plastics endanger ship propulsion systems, disrupt marine ecosystems,
and increase port and shipping route maintenance costs [ 37]. AI-based monitoring
research have shown that advanced machine learning approaches can reliably estimate
and trace the movement of macroplastic trash from rivers to coastal waterways. Kata -
oka et al. [ 38] created an AI-assisted monitoring framework that identified macroplas -
tic migration channels from riverine systems to the marine environment, providing vital
information on the spatial distribution and transit dynamics of plastic debris. These find-
ings contribute to evidence-based decision-making by identifying pollution hotspots and
influencing mitigation plans for coastal infrastructure, navigation routes, and marine
ecosystems Cernian and Iliuň Rodríguez’s [ 39] DEEP-PLAST framework uses artificial
intelligence and satellite data to automatically detect floating marine plastic waste on
vast scales. By allowing for the rapid identification of plastic accumulation zones, the
framework promotes early intervention, improves the effectiveness of marine pollution
surveillance, and improves risk assessment. These capabilities provide maritime authori-
ties with timely information for prioritizing cleaning operations, safeguarding coastal

Page 8 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
infrastructure, increasing navigation safety, and promoting evidence-based decision-
making for long-term marine pollution management.
Oil contamination is still one of the most serious threats to marine ecosystems and
maritime activities due to its widespread environmental and socioeconomic conse -
quences. In addition to causing extensive ecological damage, major oil spills can disrupt
fishing activities, reduce fishery-related income, and harm coastal communities' liveli -
hoods by destroying marine resources and limiting fishing operations [ 40]. These conse-
quences underline the need of early oil spill identification and effective response tactics
in minimizing environmental damage and reducing economic losses across maritime
industries. The ability to monitor and respond to oil spills has been greatly enhanced
by the combination of artificial intelligence with remote sensing technology. Convolu -
tional neural networks used to Sentinel-1 SAR imaging can identify oil spills with high
accuracy, allowing for quick diagnosis and mitigation before damages worsen, as dem -
onstrated by Kalogirou et al. [ 41]. According to Dong et al. [ 42], AI-enabled remote
sensing enables near-real-time oil spill detection and continuous environmental moni -
toring, allowing for the quick identification of spill occurrences and prompt decision-
making for environmental protection and marine safety. The suggested system exhibits
AI's operational capability to efficiently analyze massive volumes of remote sensing data,
boosting the effectiveness of oil spill surveillance and emergency response.
Marine ecosystems and, indirectly, maritime industries that rely on healthy coastal
waters are being threatened by chemical contaminants. Santhosh et al. [ 43] identi -
fied pharmaceutical residues in Indian marine ecosystems and proved their negative
effects on microbial communities, trophic transfer, and fish health. These data suggest
that pharmaceutical contamination can impair ecosystem functioning and endanger
fisheries, aquaculture production, and coastal economies. Building on these findings,
AI-powered monitoring and prediction systems could be a useful tool for integrat -
ing environmental data, identifying contamination patterns, and facilitating early risk
assessment and evidence-based management of pharmaceutical pollution in marine
environments.
2 Artificial intelligence applications in oil spill monitoring and marine
environmental protection
Aquatic oil pollution continues to be a significant environmental problem with detri -
mental effects on the environment, human health, and the economy. The requirement
for quick identification and monitoring is highlighted by thorough assessments that
highlight how environmental factors have a significant impact on oil behavior in water,
including its spread, persistence, and toxicity. Current methods for monitoring oil spills
may be generally divided into three categories: wireless sensing employing sensor net -
works and Internet of Things (IoT) technology, optical sensing, and distant sensing. The
increasing relevance of artificial intelligence, particularly machine learning and deep
learning approaches, in enhancing detection accuracy, predicting spill dynamics, and
assisting effective response tactics is further highlighted by recent developments. When
combined, these integrated monitoring systems offer a more thorough framework for
surveillance of oil contamination in aquatic settings. According to Samsuria et al. [ 44],
the detection, categorization, and ongoing evaluation of oil pollution are improved when
traditional monitoring techniques like remote sensing, optical sensing, and sensor-based

Page 9 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
monitoring are combined with AI-driven data analysis. These integrated approaches
facilitate prompt decision-making and help manage oil pollution more effectively and
sustainably by enabling more effective monitoring and enhancing the understanding of
complicated environmental statistics.
For tracking maritime oil pollution and comprehending its spatiotemporal dynamics,
remote sensing and machine learning techniques have shown great promise. The Nor -
malized Difference Water Index (NDWI) and the Oil and Suspended Solids Index (OSSI)
are two satellite-derived spectral indices that have been evaluated recently. NDWI
consistently performs better (R 2 = 0.81) and has higher spectral sensitivity than OSSI.
K-means clustering significantly outperformed Self-Organizing Maps (SOM) in the
evaluation of unsupervised classification algorithms for mapping contaminated zones,
achieving higher overall accuracy (95.0%), precision (94.2%), recall (93.8%), and F1score
(94.0%) in isolating high-pollution coastal areas. For precise and scalable maritime oil
spill monitoring, our comparative evaluation measures highlight the need to combine
strong machine learning algorithms with high-precision remote sensing indices [ 45].
For efficient pollution response and environmental management, accurate modeling of
marine oil spill drift is crucial. In order to simulate the movement and development of
maritime oil spills under various environmental circumstances, Popa et al. [ 46] intro-
duced a computationally effective framework that incorporates important physical pro -
cesses, such as advection, diffusion, and degradation. In addition to helping emergency
response and coastal protection, these physically based models offer useful decision-
support tools for predicting spill paths. By incorporating real-time oceanographic obser-
vations, satellite remote sensing data, and sophisticated learning algorithms to improve
spill trajectory forecasting and environmental risk assessment, AI-based approaches
have the potential to further improve predictive performance by building upon these
modeling frameworks.
Although conventional techniques for detecting and tracking oil spills, such as buoy
monitoring [47], optical detection [ 48], and manual patrols [ 49], are fairly successful in
identifying oil contamination, but they have limitations in terms of efficacy, accuracy,
significant environmental impact, and coverage. These limitations make them inade -
quate to manage sudden or large-scale oil spills [ 50]. Computer vision technology has
been used to monitor crude oil slicks on the sea surface by automatically recognizing and
classifying oil-contaminated spots in optical images. Compared to traditional monitor -
ing methods, which frequently rely on manual picture interpretation or labour-intensive
field observations, computer vision allows for faster, more consistent, and large-scale
analysis while reducing human interaction. To improve oil spill characterisation, the
laser-induced fluorescence (LIF) approach was developed to identify distinct types of oil
products and quantify their quantities using their fluorescence properties. According to
Bukin et al. [47], combining active laser-induced fluorescence spectroscopy with passive
optical monitoring via computer vision offers complementary capabilities for precise oil
slick detection and in-depth oil type and concentration characterization, improving the
efficacy of environmental response and oil spill monitoring.
According to Zare et al. [ 5], machine learning's capacity to efficiently analyze com -
plex chemical datasets, identify distinctive patterns in a variety of petroleum samples,
and improve the classification and source identification of oil contaminants has made
it a valuable tool for oil fingerprinting. By processing vast amounts of analytical data

Page 10 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
more rapidly and reliably than conventional data analysis techniques, machine learning
enhances the accuracy and dependability of oil fingerprinting, enabling more effective
environmental monitoring and pollutant source tracking. Using predictive algorithms
(which are essential for functional machine learning) allows it to be independent of pre -
vious information about subjects, successfully classify problems, be built for massive
data analysis, and assess numerous inputs and samples with intricate linkages without
having to confirm presumptions.
Currently, a lot of people employ synthetic aperture radar (SAR) imaging to find and
evaluate oil spills [ 51–53]. Synthetic Aperture Radar (SAR) allows for high-resolution,
all-weather detection of oil spills by separating oil slicks from surrounding seawater
based on texture, form, pattern, and contrast. When integrated with machine learn -
ing, SAR imagery may be automatically evaluated to enhance the speed, accuracy, and
consistency of oil spill detection while eliminating the need for manual interpretation,
making it a valuable tool for large-scale marine pollution monitoring. Using the SAR
technique to detect oil spills involves three steps: dark spot identification, recognition,
and categorization [ 54]. In actuality, however, images from remote sensing are influ -
enced by a variety of elements, including the atmosphere, fog, clouds, sunlight exposure,
and sensor noise, which can result in noise, distortion, and abnormalities [ 55]. Remote
sensing images must be denoised to reduce the impact of interfering factors and improve
the accuracy of image segmentation (Table 1). Araújo et al. [57] used the à trous method
to reduce noise in photos. The undecimated wavelet transform can be used to smooth
SAR picture speckle noise while also enhancing the image's edges. Aside from calibrat -
ing to reduce the effects of antenna incidence angle fluctuations, Lee et al. [ 58] used
mean filters to reduce speckle noise in SAR images [59].
The SOS dataset incorporates a variety of additional noise types, for example, multipli-
cative coherent speckle noise in synthetic aperture radar (SAR) images, to simulate how
an oil spill would affect pictures. Additionally, researchers have improved the use of the
Table 1 Representative AI-based studies for marine oil pollution detection and monitoring
Pollutant
type
AI method Sensing
platform
Key per-
formance
metrics*
Major advantages Limitations Refer-
ences
Oil spill Hybrid CNN–
Vision Trans-
former (ViT)
Sentinel-1
SAR imagery
Best F1-score
= 78.48%
Automatic oil-spill
detection and
segmentation
Requires large
annotated SAR
datasets and
computational
resources
Deh-
ghani
et al.
[56]
Oil slick Computer
Vision +
Laser-Induced
Fluorescence
(LIF)
Optical imag-
ing + LIF
Not reported Simultaneous oil de-
tection and oil-type
characterization
High equipment
cost; limited field
deployment
Bukin
et al.
[47]
Oil spill AI-assisted Re-
mote Sensing
Satellite
imagery
Near
real-time
monitoring
Rapid environmental
assessment and deci-
sion support
Dependent on
satellite revisit
frequency
Dong
et al.
[42]
Oil spill AI-enhanced
numerical
modelling
Remote
sensing +
Oceano-
graphic data
Not reported Predictive spill trajec-
tory modelling
Requires accu-
rate environ-
mental input
data
Popa
et al.
[46]
Oil
pollution
AI-integrated
monitoring
framework
Remote
sensing, IoT
sensors
Not reported Integrates mul-
tiple monitoring
technologies
Limited
industrial-scale
validation
Sams-
uria et
al. [44]

Page 11 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
FFDNet model to lessen additive and multiplicative noise. FFDNet is a quick and effec -
tive image denoising algorithm that uses fast fourier transform (FFT) [ 60], non-local
means filters, and deep convolutional neural networks (CNNs) [ 61]. Because CNNs use
varied training approaches and residual learning, FFDNet can effectively denoise a range
of image noise types, including Gaussian, impulse, and mixed noise. Candidate pictures
can be denoised using the improved FFDNet model. A number of pre-segmentation
image processing methods have either optimized classical algorithms based on observa -
tions or employed specific algorithms. These approaches are based on different algorith -
mic ideas. These advancements and enhancements will make SAR processing of images
more dependable and effective in addressing noise problems [ 1]. For applications like
image-guided remote monitoring, quantitative analysis, and visualization, SAR image
segmentation is essential. It increases the efficiency of further research and applications
by making it simpler to identify and differentiate oil from similar substances in images
[12]. Convolutional Neural Networks (CNNs) were used to classify oil kinds (mineral,
biogenic, emulsified), estimate thickness, and delineate oil spills using SAR images [41].
Monitoring mission autonomy is made possible by the deployment of unmanned aer -
ial vehicles (UAVs), which are increasingly adapted from general coastal water quality
assessment to targeted oil spill surveillance in remote marine environments. By integrat-
ing dedicated carrying platforms with real-time payload navigation and multi-sensor
payloads, these systems establish integrated hardware and software framework (HSFW)
capable of both active and passive contamination mapping. Owing to operational com -
plexity, such systems are structured into three interconnected sub-frameworks dedi -
cated to flight control, autonomous navigation, and active optical monitoring, thereby
providing the scalable autonomy required for rapid coastal oil spill response [62].
The potential of GeoAI and large language models (ChatGPT) for evaluating the
dynamics of coastal water quality throughout the southeast Red Sea coast (Halayeb–
Shalateen, Egypt) was shown by Aziz et al. [ 63]. In order to monitor important physico -
chemical parameters, the study combined data from the Copernicus Marine Service and
Landsat imagery. Additionally, ChatGPT was used to create Python scripts that were
incorporated into ArcGIS to create optimal geographical distribution maps. The findings
highlighted the impact of terrestrial runoff and flash floods on nutrient and sediment
enrichment in this semi-enclosed marine environment by revealing significant spatial
variability in chemical characteristics close to the coastline, especially around the out -
lets of Wadi Hudain and Wadi Da'eb. Physical factors, on the other hand, showed no
consistent spatial pattern, suggesting the combined influence of local topography, mon -
soonal winds, and regional climate. These findings highlight how combining generative
AI code generation with spatial analysis stream-lines coastal water quality mapping and
enhances decision support for marine environmental management (Fig. 3a, b).
Monitoring vessel-generated atmospheric emissions represents a critical dimension of
holistic marine environmental protection, as airborne shipping pollutants directly affect
coastal air and water quality via atmospheric deposition. El-Zeiny et al. [ 34] investigated
the impact of maritime and industrial activities along the Egyptian Mediterranean coast
using AI-assisted, multi-source satellite observations. By integrating data from Aqua,
Terra, and Sentinel-5P , the study evaluated PM₂.₅, CO, NO₂, and SO₂ concentrations
across six major commercial harbors (Port Said, Damietta, Abo Qir, El-Mex, Gargob,
and Arish) throughout 2023. Results revealed pronounced spatiotemporal and seasonal

Page 12 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
variability, with elevated PM₂.₅, in spring (up to 19.41  μg/m 3) and localized NO₂, and
SO₂ hotspots closely tied to shipping density and port industrial activity (Fig.  4). These
findings demonstrate how Earth observation and geospatial analytics can track ship-
borne pollution loads, linking maritime emission monitoring to sustainable port opera -
tions and broader marine ecosystem management.
Beyond oil spill surveillance, comprehensive marine environmental protection relies
on multi-hazard remote sensing tools capable of assessing biological water quality
degradation in vulnerable coastal zones. El-Zeiny et al. [ 11] developed an AI-assisted
remote sensing and spatial modeling framework to assess bacterial contamination risk
Fig. 4 Spatial distribution of CO (μg/m3) along the Egyptian Mediterranean coast [34]

Fig. 3 Spatial dynamicity of chemical (a) and physical characteristics (b) of coastal water quality [63]

Page 13 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
in the northern coastal Nile Delta, Egypt. Multispectral Landsat imagery was used to
derive key marine water quality indicators (TP , TN, DO, SS, and BOD), chlorophyll
indices (NDVI), and land surface temperature (LST). Integrated into an AI-informed
spatial model, the framework accurately mapped bacterial contamination hotspots in
Manzala and Malahaat Lakes resulting from wastewater discharges (Fig.  5). These find -
ings demonstrate how AI-enhanced satellite observation provides a scalable early-warn -
ing and decision-support tool for coastal ecosystem protection and microbial hazard
monitoring.
3 Integration of AI in plastic pollution identification and monitoring
Since plastics were commercially available in the 1950s, the amount of plastic produced
annually has expanded dramatically [ 64]. By 2020, it had climbed to an incredible 367
million tons, increased from the 1950s' 1.5 million tons [ 65]. The rapid growth in plas -
tic production and consumption has resulted in a substantial increase in plastic debris
entering marine environments, leading to the widespread accumulation of plastic waste
across the world's oceans [ 66]. Plastic debris is one of the most common and persis -
tent components of marine litter, with its abundance increasing dramatically in recent
decades due to increased plastic production, insufficient waste treatment, and contin -
ual inputs from land- and sea-based sources. Long-term observations have indicated
a significant increase in ocean plastic pollution over the last 60 years, emphasizing the
growing buildup of plastic trash in marine settings and its escalating threat to marine
ecosystems. The COVID-19 pandemic exacerbated marine plastic pollution by signifi -
cantly increasing the use of single-use plastic products, particularly hospital medical
waste, personal protective equipment (PPE), testing materials, and packaging. In many
areas, insufficient collection and disposal of these materials led in enormous amounts
of improperly managed plastic debris entering rivers and eventually the marine environ -
ment. Peng et al. [ 67] calculated that 8.4 ± 1.4 million tons of pandemic-related plastic
garbage were generated globally, with 25.9 ± 3.8 thousand tons released into the global
ocean, accounting for approximately 1.5 ± 0.2% of annual riverine plastic discharge.
These findings emphasize the significance of improving medical waste management sys -
tems to reduce future plastic contamination during global public health catastrophes.
Estimating the quantity and analysing the different kinds of plastic trash in marine habi -
tats are still difficult tasks [1].
Fig. 5 Bacterial Contamination along the Mediterranean Sea [11]

Page 14 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
Marine plastic litter can be found and categorized using a variety of techniques, such
as remote sensing and traditional methods. In situ monitoring [ 68], direct visual obser -
vation [ 69], and net sampling [ 70] are the primary manual approaches. Due to their
labour-intensive and nature-consuming character, large-scale and coordinated surveys
are challenging to conduct in distant places [ 71]. These techniques depend on human
visual identification and manual classification of marine plastics, while also taking into
consideration factors like ship speed, observer isolation from the garbage, and weather
[72]. The main tools utilized for remote sensing [ 73]to identify and categorize marine
plastic debris are satellites and drones [ 74]. However, this approach has many draw -
backs, such as weather and flying circumstances, and its limited resolution, which makes
it impossible to identify microplastics effectively [75].
Modern developments in AI-based monitoring have overcome many of the con -
straints of traditional field surveys and visual inspections, allowing for faster, more
accurate, and automated identification of marine plastic pollution. Recent advances in
remote sensing (RS) and artificial intelligence (AI) have significantly improved the detec-
tion of floating and beached plastic debris by exploiting their spatial and spectral charac-
teristics. High-resolution airborne, satellite, and UAV imagery, combined with machine
learning (ML) and deep learning (DL) algorithms such as random forest, support vec -
tor machines, and convolutional neural networks, enable automated and scalable plastic
detection. Although AI-based methods are still evolving, their integration with RS data
shows strong potential for enhancing marine plastic pollution monitoring and support -
ing future operational frameworks [76].
Marine plastic pollution remains a major threat to aquatic ecosystems, necessitating
efficient detection and management strategies. Recent studies have demonstrated the
potential of integrating artificial intelligence (AI) with spatial analysis for identifying
plastic pollution hotspots. For instance, Ayana et al. [ 77] used YOLO object recogni -
tion, drone images, and GIS-based spatial analysis to map plastic pollution in Vembanad
Lake, India, successfully finding 17 main hotspots. Compared to traditional field surveys,
an AI-driven system allows for automatic identification over vast spatial areas, improves
monitoring efficiency, and creates high-resolution spatial data for hotspot mapping. The
combination of AI and GIS yields actionable knowledge for prioritizing cleanup opera -
tions, optimizing resource allocation, and promoting sustainable management of coastal
and freshwater ecosystems.
Superior resolution drone-based imagery was used to supplement satellite data, allow -
ing for the identification of smaller-scale floating plastics and trash. Field pictures were
taken at predefined intervals over 59 locations in Central Vembanad Lake, India [ 77].
Autonomous Underwater Vehicles (AUVs) equipped with hyperspectral imaging and
AI-assisted optical sensors were used to identify microplastic at constant depths [78].
Fourier Transform-Infrared (FT-IR) spectroscopy is the most crucial technological
method for identifying the chemical composition of plastic debris in the ocean. Smart
contract systems, digital wallets, and distributed records can successfully replace the
current recycling chain, and a digital token system and identity recognition mechanism
based on BCT can raise awareness about marine plastic waste governance. Addition -
ally, by utilizing high transparency and blockchain technology's traceability, an efficient
worldwide network for recycling marine plastic can be created. To detect microplas -
tic pollution in the marine ecosystem and to improve the ongoing monitoring of

Page 15 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
microplastic contamination in the marine environment, the correlation extraction of
features strategy was employed [ 79]. The efficiency and accuracy of marine plastic trash
detection and classification have been increased by the use of AI techniques. Marine
contamination is caused by both macro plastics and micro plastics. In general, macro
plastics include big, readily visible bits of plastic garbage, like toys, bags, and bottles [80].
Conversely, micro plastic fragments with a diameter of five millimetres or less. Although
they are invisible to the human eye, they are frequently found in aquatic habitats [ 1].
Ingesting or becoming entangled in macro plastics can cause gastrointestinal obstruc -
tions, suffocation, or even death for marine life. Because microplastics are easier to hide,
benthic species frequently eat them, which allows them to go into the food chain and
impact human health in the long run [81]. Therefore, it is imperative to properly identify,
monitor, and control marine macro plastics and microplastics. Traditional techniques of
monitoring marine plastics are labour-intensive and time-consuming, and they create
accuracy and efficiency difficulties [82].
Deep learning-based computer vision techniques, specifically convolutional neural
networks (CNNs), have been widely used to detect, classify, and quantify floating plastic
debris from optical and aerial imagery, allowing for automated and large-scale monitor -
ing of marine plastic pollution [ 23]. While these methods work well for detecting visible
plastic debris, characterisation of microplastics necessitates the use of additional ana -
lytical techniques. In this context, Fourier Transform Infrared Spectroscopy (FTIR) has
become a standard approach for detecting polymer composition using spectral finger -
prints. Lin et al. [ 24] created a modified Transformer neural network in which the con -
ventional decoder was replaced with a feedforward neural network, resulting in better
automated classification and identification of microplastics from FTIR spectra. Käppler
et al. [ 83] showed an F1-score and identification accuracy of more than 90% for micro -
plastic identification utilizing FTIR-based analysis. The proposed silicon filter substrate
increased polymer discrimination and identification reliability by allowing transmission
measurements below 1300 cm ⁻1, compared to standard FTIR techniques. These meth -
odological changes helped to achieve high identification performance and revealed the
promise of advanced FTIR-based approaches for precise microplastic characterization.
The unbalanced and inadequate training datasets for microplastics were addressed by
Yan et al. [84]. Critical FTIR spectral areas, or "fingerprint regions," were found for MP
categorization using Gaussian mixture models (GMMs) and artificial intelligence that
can be explained (XAI). To expand the MP dataset, a data augmentation method known
as Fingerprint Region based Data Augmentation (FRDA) was developed based on these
fingerprint locations. By strengthening resilience of machine learning models, the FRDA
approach provides a practical means of addressing the problem of data imbalance.
Marine plastic pollution spreads from coastal seas to the deep sea, endangering marine
ecosystems [ 85]. Because no single monitoring technique is appropriate for all marine
settings, numerous sensing platforms have been combined with AI to improve detection
at various spatial scales. Satellite photography offers broad regional coverage for tracking
floating debris across huge oceanic areas, although it is limited by spatial resolution and
atmospheric circumstances [ 54]. Unmanned aerial vehicles (UAVs) provide high-reso -
lution views of coastal and nearshore habitats, allowing for more thorough mapping of
plastic accumulation zones [ 86]. Shipborne cameras offer flexible, close-range observa -
tions that supplement satellite and UAV surveys, particularly in congested marine areas.

Page 16 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
AI-driven computer vision and deep learning techniques automate the identification and
classification of floating plastics across these image-based platforms, decreasing manual
interpretation and increasing detection efficiency and consistency [ 23]. Monitoring sub-
merged plastic litter provides extra complications since optical vision reduces as water
depth increases. Underwater holographic imaging can collect precise three-dimensional
information about submerged particles, while AI-based image processing improves
underwater plastic identification and classification by extracting complex morphologi -
cal traits [87]. Sonar technologies, while typically created for fish detection and fisheries
assessment [88], offer an alternative for low-visibility conditions since sonic signals may
penetrate muddy water, where optical approaches are ineffective. Recent breakthroughs
in deep learning have improved sonar data interpretation by automating macro plastic
litter recognition and minimizing the need for manual analysis, showing the emerging
possibility for large-scale underwater plastic monitoring [ 89]. Together, these comple -
menting AI-enhanced monitoring technologies allow more extensive, scalable, and accu-
rate monitoring of marine plastic pollution from the sea surface to deep-water settings
(Tables 2 and 3).
4 Integration of AI into marine litter prediction
Marine litter prediction remains difficult because beach morphology, litter sources, and
environmental variables differ greatly throughout coastal locations. As a result, conven -
tional field surveys alone frequently give little predictive capability while requiring sig -
nificant time, labor, and financial resources. Artificial intelligence (AI)-based approaches
have emerged as viable alternatives, allowing for the investigation of complicated, non -
linear interactions within marine litter datasets while enhancing forecast efficiency and
promoting evidence-based beach management. Fuzzy algorithms were developed as one
of the first artificial intelligence tools to address the uncertainty associated with marine
litter assessment. Unlike traditional statistical models, fuzzy systems combine expert
Table 2 Representative AI-based studies for marine plastic and marine litter detection
Pollutant
type
AI method Sensing
platform
Key per-
formance
metrics*
Major
advantages
Limitations Ref-
er-
ences
Floating
plastics
Deep CNN RGB Images High detection
performance
reported
Automated
floating plas-
tic detection
Sensitive to environ-
mental illumination
Kylili
et al.
[23]
Beach plas-
tic litter
Deep
Learning
UAV imagery Not reported Rapid
large-area
monitoring
Weather-dependent Kako
et al.
[86]
Plastic
pollution
YOLO + GIS UAV imagery 17 pollution
hotspots
identified
Automated
hotspot
mapping
Performance metrics
not reported
Ayana
et al.
[77]
Marine litter Random For-
est (RF) and
Convolu-
tional Neural
Network
(CNN)
UAV
orthophotos
RF: F-score
70%; CNN:
F-score 60%
Automated
mapping
and density
estimation of
beach marine
litter
Performance affected
by false positives and
image resolution
Gon-
çalves
et al.
[20]
Marine
plastics
Random For-
est (RF) and
Support Vec-
tor Machine
(SVM)
Sentinel-2 satel-
lite imagery
~80–90%
detection
accuracy
Automated
large-scale
detection of
floating ma-
rine plastics
Requires high-quality
satellite imagery and
representative train-
ing data
San-
ni-
grahi
et al.
[90]

Page 17 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
knowledge via language "if–then" rules, allowing qualitative data and human experi -
ence to be added into the prediction process. This capacity makes fuzzy systems ideal for
modeling extremely uncertain environmental circumstances if monitoring data is insuf -
ficient or system variables cannot be properly quantified [92].
Moshgbar et al. [93] proved that fuzzy logic provides a versatile framework for analyz -
ing, forecasting, and managing complex environmental systems with high uncertainty
and nonlinear behaviour.
While fuzzy methods generally deal with ambiguity via expert knowledge, artificial
neural networks (ANNs) improve marine litter prediction by learning complicated cor -
relations directly from observational data. Balas et al. [ 94] created a neural network
model to forecast beach litter categories and rank Antalya beaches based on the pres -
ence of litter. By learning the association between the general litter category and other
litter classes, the ANN approximated beach conditions quickly while significantly mini -
mizing the need for extensive field inspections. Such computerized prediction is espe -
cially useful in beach management, environmental monitoring, and health and safety
assessments, where rapid decisions are required.
Building on this work, Balas et al. [95] conducted a comparison of neural networks and
fuzzy AI approaches for assessing beach litter. The scientists proved that, while neural
networks are excellent for high-speed prediction from monitoring data, fuzzy systems
offer value by incorporating qualitative expert knowledge about beach characteristics,
trash sources, and environmental variability. These complementing AI strategies, which
combine data-driven learning with uncertainty modeling, provide more robust predic -
tion capabilities than traditional assessment methods, increasing decision support for
coastal litter control.
More recently, advancements in computer vision have expanded AI applications
beyond prediction to automated litter identification. Gonçalves et al. [ 20] created an AI-
based framework for automatically identifying beach litter using aerial photogrammetric
Table 3 AI techniques for microplastic identification and water quality assessment
Pollutant
type
AI method Analytical/
sensing
platform
Key perfor-
mance metrics*
Major
advantages
Limitations Refer-
ences
Microplastics Modified
Transformer
Neural
Network
FTIR
Spectroscopy
Improved
classification
performance
Automated
FTIR spectrum
interpretation
Requires
labelled spectral
datasets
Lin et
al. [24]
Microplastics 1D-CNN,
2D-CNN,
Random
Forest and
Decision
Tree
FTIR
spectroscopy
1D-CNN accura-
cy: 96.43–97.44%
Automated and
rapid polymer
classification;
effective with
large spectral
datasets
Performance
depends on da-
taset composi-
tion and spectral
quality
Liu et
al. [91]
Hg2⁺ Machine
Learning
Biosensor Not reported Portable
real-time heavy
metal detection
Limited pollut-
ant specificity
Pen-
nac-
chio et
al. [9]
Marine
bacteria
Machine
Learning
Water-quality
monitoring
Not reported Continu-
ous bacterial
prediction
Site-specific cali-
bration required
Guo et
al. [10]
Ammonia Machine
Learning
Water-quality
data
Not reported Supports coast-
al water-quality
management
Requires long-
term monitoring
data
Ghayas
et al.
[8]

Page 18 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
photos captured by unmanned aerial vehicles (UAVs). Compared to traditional field-
based monitoring, UAV-assisted image analysis allows for faster assessment of huge
coastal areas while reducing manual inspection and boosting geographic coverage. Fur -
thermore, combining AI and digital twin technology enables realistic simulations of UAV
and marine drone operations under a variety of weather circumstances, sensor setups,
and navigation techniques. These digital environments offer useful tools for enhancing
monitoring tactics prior to field deployment, increasing the efficiency and reliability of
AI-assisted marine litter surveillance. The AI-driven marine pollution monitoring archi -
tecture functions as an integrated, three-tiered workflow: Data Collection, AI Process -
ing & Analysis, and Results & Actions, as shown in Fig.  6. The primary computational
engine receives data from several platforms, including underwater autonomous vehicles
(AUVs), physical in-situ sampling, targeted UAV aerial surveys, and large-scale satel -
lite imaging. Sequential computer vision and analytical activities, such as object detec -
tion, semantic image segmentation, predictive neural modeling, and hydrodynamic drift
trajectory analysis, are carried out by this AI core. In order to produce high-resolution
pollution maps, mass/count quantifications, and geographic hotspot identifications to
guide focused mitigation and coastal cleanup strategies, this operational pipeline ulti -
mately converts raw multi-source spatial data into actionable environmental manage -
ment deliverables (Fig. 6).
The transition from fuzzy systems to artificial neural networks and AI-powered com -
puter vision indicates the steady growth of AI approaches for marine litter prediction
and monitoring. Fuzzy methods successfully resolve ambiguity with expert knowledge,
neural networks give efficient data-driven prediction of litter categories, and UAV-based
computer vision allows for automated large-scale litter detection. Together, these com -
plementing AI technologies transcend many of the constraints of traditional monitoring
systems, resulting in more scalable, efficient, and informed solutions for marine trash
management (Table 4).
Marine pollution monitoring is moving from experimental research to operational
deployment due to recent improvements in artificial intelligence. For complicated
Fig. 6 Workflow of AI-enabled marine pollution monitoring

Page 19 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
remote sensing and hyperspectral datasets, transformer-based deep learning architec -
tures have shown enhanced feature extraction capabilities, allowing for more reliable
pollutant identification in dynamic environmental settings. Similarly, edge AI reduces
communication latency and permits real-time environmental monitoring by enabling
onboard data processing in UAVs, autonomous surface vehicles (ASVs), and autono -
mous underwater vehicles (AUVs) [25]. By integrating in situ observations with satellite
remote sensing and cloud-based analytics, the integration of AI with IoT sensor net -
works significantly facilitates ongoing monitoring of maritime environments. Further -
more, environmental agencies may simulate contamination scenarios, refine cleanup
plans, and assess mitigation measures prior to field implementation thanks to digital
twin technologies, which offer virtual representations of marine ecosystems. Together,
these cutting-edge technologies enhance operational scalability, decision-making, and
monitoring effectiveness, indicating the growing maturity of AI-driven marine pollution
management systems [26, 46].
The capacity of AI–IoT integration to create ongoing environmental surveillance as
opposed to event-based monitoring is what makes it valuable in the real world. Water-
quality factors might be continuously measured by a coastal monitoring network, which
would then integrate these observations with meteorological data and satellite imagery.
Using an autonomous platform, an AI model may detect an anomalous pattern, compare
it to past conditions, calculate the likelihood of a pollution incident, and initiate tar -
geted verification. Because human resources may be allocated to areas where AI detects
the most environmental concern, this would significantly increase the effectiveness of
field sampling. For instance, Subramani et al. [ 26] created an automated marine cleanup
Table 4 Comparative analysis of AI methodologies for marine pollutant identification
AI technique Typical applications Main strengths Major limitations Repre-
sentative
primary
research
Convolutional
Neural Net-
works (CNNs)
Oil spills, floating plastics,
marine litter, microplastic
identification
Effective automated image/
spectral feature extraction
and classification
Requires representa-
tive training datasets;
performance can decline
under domain shift
Dehghani
et al. [56],
Liu et al.
[91]
Transformer
networks
Oil-spill detection; spec-
tral classification
Strong feature extraction
and ability to model com-
plex relationships
Computationally
demanding; requires
substantial training data
Dehghani
et al. [56]
YOLO object
detection
Floating plastics and
beach litter
Fast object detection
suitable for near-real-time
monitoring
Sensitive to image qual-
ity, object scale and envi-
ronmental conditions
Gonçalves
et al. [20]
Machine
learning (RF,
SVM, ANN)
Marine plastic clas-
sification; water-quality
prediction; pollutant
identification
Handles nonlinear relation-
ships and heterogeneous
environmental data
Strongly dependent on
training-data quality and
representativeness
Sanni-
grahi et al.
[90]
Deep learning
for spectral
analysis
Microplastic identification
using FTIR/Raman
Automated polymer
discrimination with high
classification accuracy
Requires labelled spec-
tral datasets and careful
preprocessing
Liu et
al. [91],
Michel et
al. [96]
IoT + AI Continuous marine envi-
ronmental monitoring
Enables continuous data
acquisition and automated
analysis
Sensor maintenance,
connectivity and calibra-
tion requirements
Ananthi
et al. [97]
Computer
vision
Oil-slick and marine-litter
detection
Automated visual inter-
pretation and scalable
monitoring
Performance affected by
illumination, visibility and
image quality
Gonçalves
et al. [20],
Michel et
al. [96]

Page 20 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
system with AI integration that integrates robotic mechanisms, real-time image process-
ing, and pollution classification. The approach shows how AI can potentially increase the
scalability of maritime trash management while reducing ongoing human intervention.
Quantitative literature and empirical benchmarks show high model efficacy, ranging
from UAV-based YOLO detection and regional satellite classification (80–90% accuracy)
for macroplastics [ 77, 90] to deep learning spectral discrimination (> 96% accuracy) for
microplastics [ 24, 91], SAR/optical segmentation (F1 up to 85.2%, and 95.0% accuracy
via K-means NDWI) for oil slicks [ 41, 45, 56], and multispectral/biosensor modeling
for dissolved chemical, microbial, and port atmospheric hazards [ 8, 9, 11, 34]. How -
ever, translating these algorithms to open marine environments remains constrained by
hydrodynamic wave action and biofouling, atmospheric interference and false positives
caused by sun glint or whitecaps, extreme class imbalances and regional domain shifts,
and strict edge-computing hardware limits on autonomous platforms.
5 Contribution of the study to the sustainable development goals
5.1 SDG 14: life below water (primary goal)
This review directly supports SDG 14 (Life Below Water) by showing how AI-based
approaches help to avoid and reduce marine pollution (Target 14.1) while also protect -
ing and restoring marine ecosystems (Target 14.2). The papers examined reveal that
deep learning algorithms, computer vision, convolutional neural networks (CNNs),
and transformer-based models can identify, classify, and quantify marine contaminants
such as oil spills, floating plastics, marine litter, and microplastics. AI-integrated satellite
imaging, unmanned aerial vehicles (UAVs), autonomous underwater vehicles (AUVs),
and IoT-enabled monitoring systems increase hotspot detection, pollutant tracking, and
predictive modeling, allowing for early response before pollution spreads. These tech -
nologies give scientific knowledge that helps to restore ecosystems, conserve biodiver -
sity, and manage marine resources sustainably.
5.2 SDG 12: responsible consumption and production
The assessed AI apps help to achieve SDG 12 by fostering more sustainable waste man -
agement and pollution prevention throughout the marine environment. Computer
vision and machine learning algorithms detect illegal garbage dumping, quantify plastic
accumulation, and track litter dispersion, allowing authorities to improve waste collec -
tion and prioritize cleanup efforts. AI-based prediction models contribute to circular
economy initiatives by identifying pollution sources and assessing the efficacy of waste
reduction efforts, minimizing plastic leaks into marine environments and encouraging
more responsible manufacturing and consumption behaviors.
5.3 SDG 9: industry, innovation and infrastructure
This review highlights how AI technology might help achieve SDG 9 by developing intel-
ligent environmental monitoring infrastructure. The combination of remote sensing,
IoT sensors, autonomous drones, underwater robots, digital twins, and cloud comput -
ing creates scalable monitoring systems capable of continuous environmental assess -
ment. AI-powered predictive models increase operational efficiency, cut monitoring
costs, and make it easier to shift from traditional field surveys to automated, real-time

Page 21 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
environmental observation. These advancements improve the technological capability
for sustainable maritime infrastructure and environmental protection.
5.4 SDG 6: clean water and sanitation
The evaluated research contribute to SDG 6 by illustrating the importance of AI in mon-
itoring water quality and tracing pollutant transfer from rivers and estuaries to coastal
and marine habitats. Machine learning models have been effectively used to predict con-
taminants such as mercury, ammonia, and bacterial concentrations, while AI-assisted
remote sensing allows for continuous monitoring of pollution dynamics. These appli -
cations enable early detection of water quality deterioration, promote integrated water -
shed-coastal management, and help decision-makers apply effective pollution mitigation
techniques to protect aquatic ecosystems.
5.5 SDG 3: good health and well-being
The evaluation also supports SDG 3 by emphasizing AI applications that limit human
exposure to marine pollution. AI-powered early warning systems help detect oil spills,
plastic pollution, microbiological infection, and hazardous environmental changes that
endanger seafood safety and coastal communities. Predictive modeling and real-time
monitoring enable timely environmental solutions, lowering the public health hazards
associated with contaminated marine resources while safeguarding fisheries, recre -
ational waters, and coastal communities.
The examined AI approaches show that artificial intelligence is more than just an
upgraded environmental monitoring tool; it is also an enabling technology for reach -
ing many Sustainable Development Goals at once. AI offers scalable and evidence-based
solutions for sustainable ocean governance, environmental protection, and resilient
coastal development by combining automated detection, predictive analytics, remote
sensing, autonomous monitoring platforms, and intelligent decision-support systems.
6 Conclusion
Marine pollution arises from an intricate interplay of natural processes and escalating
anthropogenic pressures across coastal, offshore, and port environments. This com -
prehensive review demonstrates that integrating artificial intelligence (AI) with Earth
observation platforms, autonomous robotic mechanisms, and in-situ sensor networks
significantly improves the precision, operational speed, and scalability of marine envi -
ronmental surveillance compared to conventional monitoring techniques. Methodologi-
cal evidence across the literature reveals that deep learning segmentation architectures
and advanced spectral indices yield substantially higher reliability in discriminating oil
slicks from natural biogenic look-alikes, while computer vision frameworks deployed on
aerial and underwater drones automate the mapping of macro- and microplastic accu -
mulation zones across heterogeneous marine terrains.
Furthermore, combining machine learning classifiers with multi-source satellite data
facilitates broad-scale, continuous tracking of coastal water degradation and vessel-
generated atmospheric plumes, successfully translating complex geospatial datasets
into actionable risk maps and microbial early-warning indicators. The transition from
episodic manual surveys to autonomous edge computing, predictive neural modeling,
and digital twin simulations equips environmental regulators and maritime stakeholders

Page 22 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
with the proactive decision-support systems necessary to enforce international stan -
dards and protect fragile aquatic biomes. Ultimately, realizing the full potential of AI for
sustainable ocean governance depends on moving beyond laboratory prototypes to scal -
able, field-deployed systems capable of navigating the dynamic physical, environmental,
and computational constraints of open-water marine ecosystems.
6.1 Future research
Despite these technological advances, existing AI applications confront significant oper -
ational constraints, such as geographic domain reliance, data scarcity in extreme sea sit -
uations, and a slow transition from experimental prototypes to field-deployed systems.
To fill these deficiencies, future research should concentrate on:

  1. Converting Experimental AI to Industrial and Field Applications: To ensure scalable
    deployment in a variety of marine environments, cross-disciplinary integration
    between computer science, physical oceanography, fluid dynamics, and environmental
    engineering is necessary when converting lab-scale models into operational marine
    governance.
  2. Measuring Economic and Operational Benefits: Responding to marine pollution
    requires a lot of resources. Future research should measure the cost–benefit ratios of
    implementing AI and show how automated detection lowers response latency, search
    time, and operational costs.
  3. Integrating AI into Marine Policy and Governance: To enable evidence-based
    policymaking, automated emission compliance, and proactive environmental
    governance, regulatory frameworks must incorporate AI-generated spatial analytics
    and real-time risk maps.
  4. Bridging AI-driven technology with policy and field deployment: It will result in
    proactive, scalable, and long-term interventions to protect global marine ecosystems.
  5. Develop AI for Integrated SDG-Oriented Marine Sustainability: Move from single-
    pollutant detection toward integrated assessment of multiple anthropogenic pressures.
    Combine AI with satellite data, in-situ observations, autonomous platforms, and
    numerical ocean models. Develop AI-enabled digital twins for pollution forecasting
    and management scenarios. Strengthen SDG 6, SDG 9, SDG 12, SDG 13, and
    particularly SDG 14.
    Author contributions
    M.K. and A.E.Z. Contributing the manuscript preparation, Writing original draft, validation and supervision, A.E.Z. Editing
    and Supervision, M.K. Writing review and editing, validation, and formal analysis. M. K. and A.E.Z. Conceptualization, Data
    curation, M.K. and A.E.Z. read and agreed on the submitted paper.
    Funding
    Open access funding provided by The Science, Technology & Innovation Funding Authority (STDF) in cooperation with
    The Egyptian Knowledge Bank (EKB). Not applicable. This research received no specific funding.
    Data availability
    No datasets were generated or analysed during the current study.
    Declarations
    Ethics approval and consent to participate
    Not applicable. This study does not involve human participants, human data, or animal experiments. Consent to
    participate: Not applicable. The study does not involve human participants.
    Consent for publication
    Not applicable. The manuscript does not contain individual-level personal data or identifiable information requiring
    consent for publication.

Page 23 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
Competing interests
The authors declare that they have no competing interests.
Received: 9 July 2026 / Accepted: 5 October 2026
References

  1. Ning J, Pang S, Arifin Z, Zhang Y, Epa UPK, Qu M, et al. The diversity of artificial intelligence applications in marine pollu-
    tion: a systematic literature review. J Mar Sci Eng. 2024;12(7):1181. https://doi.org/10.3390/jmse12071181.
  2. Vinuesa R, Azizpour H, Leite I, Balaam M, Dignum V, Domisch S, et al. The role of artificial intelligence in achieving the
    Sustainable Development Goals. Nat Commun. 2020;11(1):233. https://doi.org/10.1038/s41467-019-14108-y.
  3. Wang P , Yao J, Wang G, Hao F, Shrestha S, Xue B, et al. Exploring the application of artificial intelligence technology for
    identification of water pollution characteristics and tracing the source of water quality pollutants. Sci Total Environ.
    2019;693:133440. https://doi .org/10.101 6/j.scitote nv.2019. 07.246.
  4. Kaamoush M, El-Sayeh M, Omar MY. Observing air quality in Egypt’s Alexandria port based on the consequences of the
    COVID-19 lockdown. Discov Sustain. 2025. https://doi.org/10.1007/s43621-025-01016-z.
  5. Zare A, Ablakimova N, Kaliyev AA, Mussin NM, Tanideh N, Rahmanifar F, et al. An update for various applications of Artificial
    Intelligence (AI) for detection and identification of marine environmental pollutions: a bibliometric analysis and system-
    atic review. Mar Pollut Bull. 2024;206:116751. https://doi .org/10.101 6/j.marpolb ul.2024. 116751.
  6. Anthony D, Siriwardana H, Ashvini S, Pallewatta S, Samarasekara SM, Edirisinghe S, et al. Trends in marine pollution mitiga-
    tion technologies: scientometric analysis of published literature (1990–2022). Reg Stud Mar Sci. 2023;66:103156. h t t p s : / / d o
    i . o r g / 1 0 . 1 0 1 6 / j . r s m a . 2 0 2 3 . 1 0 3 1 5 6 .
  7. Polidoro BA, Comeros-Raynal MT, Cahill T, Clement C. Land-based sources of marine pollution: pesticides, PAHs and
    phthalates in coastal stream water, and heavy metals in coastal stream sediments in American Samoa. Mar Pollut Bull.
    2017;116(1–2):501–7. https://doi .org/10.101 6/j.marpolb ul.2016. 12.058.
  8. Ghayas S, Siddiquie JS, Safdar S, Mansoor A. Neural network implementations on the coastal water quality of Manora
    channel for the years 1996 to 2014. Int J Circuits Syst Signal Process. 2020;14:996–1004. h t t p s : / / d o i . o r g / 1 0 . 4 6 3 0 0 / 9 1 0 6 . 2 0 2
    0 . 1 4 . 1 2 7 .
  9. Pennacchio A, Giampaolo F, Piccialli F, Cuomo S, Notomista E, Spinelli M, et al. A machine learning-enhanced biosensor for
    mercury detection based on an hydrophobin chimera. Biosens Bioelectron. 2021;196:113696. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . b i o
    s . 2 0 2 1 . 1 1 3 6 9 6 .
  10. Guo J, Lee JHW. Development of predictive models for “very poor” beach water quality gradings using Class-Imbalance
    Learning. Environ Sci Technol. 2021;55(21):14990–5000. https://doi.org/10.1021/acs.est.1c03350.
  11. El-Zeiny A, Hassan E, Kaamoush M. Environmental variables-based spatial model for locating bacterial contamination in
    the Nile coastal region. Euro-Mediterr J Environ Integr. 2026. https://doi.org/10.1007/s41207-025-00999-5.
  12. Uddin MG, Nash S, Mahammad Diganta MT, Rahman A, Olbert AI, Uddin MG, et al. Robust machine learning algorithms for
    predicting coastal water quality index. J Environ Manage. 2022;321:115923. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . j e n v m a n . 2 0 2 2 . 1 1 5 9 2
    3 .
  13. Dawoud MA. Environmental impacts of seawater desalination: Arabian Gulf case study. Int J Environ Sustain. 2012. h t t p s : / /
    d o i . o r g / 1 0 . 2 4 1 0 2 / i j e s . v 1 i 3 . 9 6 .
  14. Ye X, Chen B, Lee K, Storesund R, Zhang B. An integrated offshore oil spill response decision making approach by human
    factor analysis and fuzzy preference evaluation. Environ Pollut. 2020;262:114294. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . e n v p o l . 2 0 2 0 . 1 1 4
    2 9 4 .
  15. Rattner BA. History of wildlife toxicology. Ecotoxicology. 2009;18(7):773–83. https://doi.org/10.1007/s10646-009-0354-x.
  16. Wiese FK, Robertson GJ. Assessing seabird mortality from chronic oil dischages at sea. J Wildl Manag. 2004;68:627–38.
    https://doi.org/10.2193/0022541X(2004)068[0627:ASMFCO]2.0.CO;2.
  17. García-Borboroglu P , Boersma PD, Ruoppolo V, Reyes L, Rebstock GA, Griot K, et al. Chronic oil pollution harms Magellanic
    penguins in the Southwest Atlantic. Mar Pollut Bull. 2005;52(2):193–8. https://doi .org/10.101 6/j.marpolb ul.2005. 11.004.
  18. Julian M. MARPOL 73/78: the International convention for the prevention of pollution from ships. Marit Stud. 2000;113:16–
  19. https://doi .org/10.108 0/07266472. 2000.108 78605.
  20. Campbell LM, Gray NJ, Fairbanks L, Silver JJ, Gruby RL, Dubik BA, et al. Global oceans governance: new and emerging
    issues. Annu Rev Environ Resour. 2016;41(1):517–43. https://doi .org/10.114 6/annurev-e nviron-1 02014-021121.
  21. 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 techniques. Mar Pollut Bull. 2020;155:111158. h t t p s : / / d o i . o r g / 1 0 . 1 0
    1 6 / j . m a r p o l b u l . 2 0 2 0 . 1 1 1 1 5 8 .
  22. Sharma S, Chatterjee S. Microplastic pollution, a threat to marine ecosystem and human health: a short review. Environ Sci
    Pollut Res. 2017;24(27):21530–47. https://doi.org/10.1007/s11356-017-9910-8.
  23. Williams AT, Tudor DT, Randerson P . Beach litter sourcing in the Bristol Channel and Wales, U.K. Water Air Soil Pollut.
    2003;143(1–4):387–408. https://doi.org/10.1023/a:1022808908500.
  24. Kylili K, Kyriakides I, Artusi A, Hadjistassou C. Identifying floating plastic marine debris using a deep learning approach.
    Environ Sci Pollut Res. 2019;26(17):17091–9. https://doi.org/10.1007/s11356-019-05148-4.
  25. Lin J, Liu H, Zhang J. Recent advances in the application of machine learning methods to improve identification of the
    microplastics in environment. Chemosphere. 2022;307(Pt 4):136092. https://doi .org/10.101 6/j.chemosp here.202 2.136092.
  26. Palanisamy S, Bonny T, Nasir N, Al Shabi M, Al Shammaa A. Edge AI-powered marine pollution classification with custom-
    ized CNN model. Neural Comput Appl. 2025;37(9):6449–63. https://doi.org/10.1007/s00521-024-10959-9.
  27. Subramani P , Rajasree R, Krishnan RS, Mathias A, Saranya T, Kanthimathi N. Smart ocean cleanup: an AI-integrated autono-
    mous system for marine waste management. In 2025 7th International conference on inventive material science and
    applications (ICIMA) (pp. 1258–1265) 2025. IEEE. https://doi .org/10.110 9/ICIMA6486 1.2025.1 1074176.

Page 24 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
27. Raihan A, Rahman SM, Rahman M, Rahman M, Rahman MM, Rahman MM, et al. Artificial intelligence and geospatial
approaches advancing environmental and marine science. Earth Syst Environ. 2025;10(5):6663–80. h t t p s : / / d o i . o r g / 1 0 . 1 0 0 7
/ s 4 1 7 4 8 - 0 2 5 - 0 0 9 7 1 - x .
28. Fazri MF, Kusuma LB, Rahmawan RB, Fauji HN, Camille C. Implementing artificial intelligence to reduce marine ecosystem
pollution. IAIC Trans Sustain Digit Innov. 2023;4(2):101–8. https://doi.org/10.34306/itsdi.v4i2.579.
29. Skognes K, Johansen Ø. Statmap—a 3-dimensional model for oil spill risk assessment. Environ Model Softw. 2003;19(7–
8):727–37. https://doi .org/10.101 6/j.envsoft .2003.08 .010.
30. Dong X, Li J, Li B, Jin Y, Miao S. Marine oil spill detection from low-quality SAR remote sensing images. J Mar Sci Eng.
2023;11(8):1552. https://doi.org/10.3390/jmse11081552.
31. Ghani SA, Hassan SWM, Shobier AH. The role of artificial intelligence in marine pollution: a review. Blue Economy. h t t p s : / / n
i o f - e g . r e s e a r c h c o m m o n s . o r g / b l u e - e c o n o m y / v o l 4 / i s s 1 / 6 / .
32. Lan D, Liang B, Bao C, Ma M, Xu Y, Yu C. Marine oil spill risk mapping for accidental pollution and its application in a coastal
city. Mar Pollut Bull. 2015;96(1–2):220–5. https://doi .org/10.101 6/j.marpolb ul.2015. 05.023.
33. Ostle C, Thompson RC, Broughton D, Gregory L, Wootton M, Johns DG. The rise in ocean plastics evidenced from a 60-year
time series. Nat Commun. 2019;10(1):1622. https://doi.org/10.1038/s41467-019-09506-1.
34. El-Zeiny A, Nagy A, Kaamoush M. Impact of maritime and industrial activities on air quality along the Egyptian Mediter-
ranean coast using data observation techniques. Euro-Mediterr J Environ Integr. 2025;10(4):2883–96. h t t p s : / / d o i . o r g / 1 0 . 1 0
0 7 / s 4 1 2 0 7 - 0 2 5 - 0 0 8 0 3 - 4 .
35. Aretoulaki E, Ponis S, Plakas G, Agalianos K, National Technical University Athens. Marine plastic littering: a review of socio
economic impacts. J Sustain Sci Manag. 2021;16(3):276–300.
36. Kaamoush MIA, Scopa A, Drosos M, El-Zeiny AM. Environmental impact of Irgarol 1051, a biocide, on marine microalgae
metabolism: a case study of Chlorella salina and Dunaliella bardawil. J Mar Sci Eng. 2025;13(4):695. h t t p s : / / d o i . o r g / 1 0 . 3 3 9 0 /
j m s e 1 3 0 4 0 6 9 5 .
37. Kaamoush M, El-Agawany N, El-Zeiny A. Heavy metal pollution and algal sustainability in Manzala lake: an integrated
ecotoxicological and geospatial approach. Discover Sustain. 2026. https://doi.org/10.1007/s43621-025-02576-w.
38. Kataoka T, Yoshida T, Sasaki K, Kosuge Y, Suzuki Y, van Emmerik THM. RiSIM: river surface image monitoring software for
quantifying floating macroplastic transport. Water Res. 2026;288:124678. https://doi.org/10.1016/j.watres.2025.124678.
39. Cernian A, Iliuţǎ M-E. Empowering sustainability through AI-driven monitoring: The DEEP-PLAST approach to marine
plastic detection and trajectory prediction for the Black Sea. Water. 2025;17(22):3318. https://doi.org/10.3390/w17223318.
40. De Oliveira Estevo M, Lopes PF, De Oliveira Júnior JGC, Junqueira AB, De Oliveira Santos AP , Da Silva Lima JA, et al.
Immediate social and economic impacts of a major oil spill on Brazilian coastal fishing communities. Mar Pollut Bull.
2021;164:111984. https://doi .org/10.101 6/j.marpolb ul.2021. 111984/0.
41. Kalogirou E, et al. Oil spill detection using convolutional neural networks and Sentinel-1 SAR imagery. ISPRS Arch.
2025;48(757–764):2025. https://doi .org/10.519 4/isprs-arc hives-XL VIII-G-2025-757-2025.
42. Dong S, Feng J, Gu Z, Yin K, Long Y. A review of artificial intelligence and remote sensing for marine oil spill detection, clas-
sification, and thickness estimation. Remote Sens. 2025;17(22):3681. https://doi.org/10.3390/rs17223681.
43. Santhosh K, et al. Exploring the silent threats of pharmaceutical contaminants in Indian seas. Chemosphere.
2025;392:144750. https://doi .org/10.101 6/j.chemosp here.202 5.144750.
44. Samsuria NNC, Ismail WZW, Nazli MNWM, Aziz NAA, Ghazali AK. Problems, effects, and methods of monitoring and sens-
ing oil pollution in water: a review. Water. 2025;17(9):1252. https://doi.org/10.3390/w17091252.
45. Mokarram M, Pham TM. Environmental monitoring of oil pollution in the marine waters using machine learning and
remote sensing. Adv Space Res. 2025;75(8):6173–87. https://doi.org/10.1016/j.asr.2025.01.062.
46. Popa C, Atodiresei D, Toma A, Dobref V, Vatamanu J. Solutions for modelling the marine oil spill drift. Environments.
2025;12(4):132. https://doi.org/10.3390/environments12040132.
47. Bukin O, Proschenko D, Korovetskiy D, Chekhlenok A, Yurchik V, Bukin I. Development of the artificial intelligence and opti-
cal sensing methods for oil pollution monitoring of the sea by drones. Appl Sci (Basel). 2021;11(8):3642. h t t p s : / / d o i . o r g / 1 0 .
3 3 9 0 / a p p 1 1 0 8 3 6 4 2 .
48. Lassalle G, Fabre S, Credoz A, Dubucq D, Elger A. Monitoring oil contamination in vegetated areas with optical remote
sensing: a comprehensive review. J Hazard Mater. 2020;393:122427. https://doi .org/10.101 6/j.jhazmat .2020.12 2427.
49. Jiao Z, Jia G, Cai Y. A new approach to oil spill detection that combines deep learning with unmanned aerial vehicles.
Comput Ind Eng. 2018;135:1300–11. https://doi.org/10.1016/j.cie.2018.11.008.
50. Murvay P , Silea I. A survey on gas leak detection and localization techniques. J Loss Prev Process Ind. 2012;25(6):966–73.
https://doi.org/10.1016/j.jlp.2012.05.010.
51. Cheng Y, Li X, Xu Q, Garcia-Pineda O, Andersen OB, Pichel WG. SAR observation and model tracking of an oil spill event in
coastal waters. Mar Pollut Bull. 2010;62(2):350–63. https://doi .org/10.101 6/j.marpolb ul.2010. 10.005.
52. Fiscella B, Giancaspro A, Nirchio F, Pavese P , Trivero P . Oil spill detection using marine SAR images. Int J Remote Sens.
2000;21(18):3561–6. https://doi.org/10.1080/014311600750037589.
53. Zhu Q, Zhang Y, Li Z, Yan X, Guan Q, Zhong Y, et al. Oil spill contextual and boundary-supervised detection network based
on marine SAR images. IEEE Trans Geosci Remote Sens. 2021;60:1–10. https://doi.org/10.1109/tgrs.2021.3115492.
54. Topouzelis KN. Oil spill detection by SAR images: dark formation detection, feature extraction and classification algo-
rithms. Sensors (Basel). 2008;8(10):6642–59. https://doi.org/10.3390/s8106642.
55. Fan J, Zhang F, Zhao D, Wang J. Oil spill monitoring based on SAR remote sensing imagery. Aquat Procedia. 2015;3:112–8.
https://doi.org/10.1016/j.aqpro.2015.02.234.
56. Dehghani-Dehcheshmeh S, Akhoondzadeh M, Homayouni S. Oil spills detection from SAR Earth observations based on a
hybrid CNN transformer networks. Mar Pollut Bull. 2023;190:114834. https://doi .org/10.101 6/j.marpolb ul.2023. 114834.
57. Araújo RTS, de Medeiros FNS, Costa RCS, Marques RCP , Moreira RB, Silva JL. Locating oil spill in SAR images using wavelets
and region growing. In B Orchard C. Yang, M Ali (Eds.), Innovations in applied artificial intelligence 2004;(Lecture Notes in
Computer Science, Vol. 3029, pp. 1184–1193). Springer. https://doi .org/10.100 7/978-3-540 -24677-0 _121.
58. Lee J. Speckle analysis and smoothing of synthetic aperture radar images. Comput Graph Image Process. 1981;17(1):24–
32. https://doi.org/10.1016/s0146-664x(81)80005-6.
59. Erkan U, Thanh DNH, Hieu LM, Enginoglu S. An iterative mean filter for image denoising. IEEE Access. 2019;7:167847–59.
https://doi.org/10.1109/access.2019.2953924.

Page 25 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
60. Mera D, Cotos JM, Varela-Pet J, Rodríguez PG, Caro A. Automatic decision support system based on SAR data for oil spill
detection. Comput Geosci. 2014;72:184–91. https://doi.org/10.1016/j.cageo.2014.07.015.
61. Liu B, Li Y, Li G, Liu A. A spectral feature based convolutional neural network for classification of sea surface oil spill. ISPRS
Int J Geo-Inf. 2019;8(4):160. https://doi.org/10.3390/ijgi8040160.
62. Lewis A. The development and use of the Bonn Agreement oil appearance code (BAOAC). In Proceedings of the Interspill
Conference, Cedre, France, 12–14 May 2009. https://www .interspill .org/downlo ads/arch ive/1100_lewis.pdf
63. Aziz MA, El-Zeiny A, Hassan FM, Naguib DM. Coastal water quality dynamics of the Red Sea, southeast coast of Egypt using
GeoAI and ChatGPT. J Afr Earth Sci. 2024;219:105409.
64. Yu J, Kim J, Li X, Jong Y, Kim K, Ryang G. Water quality forecasting based on data decomposition, fuzzy clustering and deep
learning neural network. Environ Pollut. 2022;303:119136. https://doi.org/10.1016/j.envpol.2022.119136.
65. Partalas I, Tsoumakas G, Hatzikos EV, Vlahavas I. Greedy regression ensemble selection: theory and an application to water
quality prediction. Inf Sci. 2008;178(20):3867–79. https://doi.org/10.1016/j.ins.2008.05.025.
66. Barnes DKA, Galgani F, Thompson RC, Barlaz M. Accumulation and fragmentation of plastic debris in global environments.
Philos Trans R Soc B Biol Sci. 2009;364(1526):1985–98. https://doi.org/10.1098/rstb.2008.0205.
67. Peng Y, Wu P , Schartup AT, Zhang Y. Plastic waste release caused by COVID-19 and its fate in the global ocean. Proc Natl
Acad Sci USA. 2021. https://doi.org/10.1073/pnas.2111530118.
68. Cózar A, Echevarría F, González-Gordillo JI, Irigoien X, Úbeda B, Hernández-León S, et al. Plastic debris in the open ocean.
Proc Natl Acad Sci U S A. 2014;111(28):10239–44. https://doi.org/10.1073/pnas.1314705111.
69. Hardesty BD, Lawson T, Van Der Velde T, Lansdell M, Wilcox C. Estimating quantities and sources of marine debris at a
continental scale. Front Ecol Environ. 2016;15(1):18–25. https://doi.org/10.1002/fee.1447.
70. Van Der Velde T, Milton DA, Lawson T, Wilcox C, Lansdell M, Davis G, et al. Comparison of marine debris data collected by
researchers and citizen scientists: Is citizen science data worth the effort? Biol Cons. 2016;208:127–38. h t t p s : / / d o i . o r g / 1 0 . 1
0 1 6 / j . b i o c o n . 2 0 1 6 . 0 5 . 0 2 5 .
71. 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. Sci Total Environ.
2019;693:133581. https://doi .org/10.101 6/j.scitote nv.2019. 133581.
72. Armitage S, Awty-Carroll K, Clewley D, Martinez-Vicente V. Detection and classification of floating plastic litter using a
vessel-mounted video camera and deep learning. Remote Sens. 2022;14(14):3425. https://doi.org/10.3390/rs14143425.
73. Lebreton L, Slat B, Ferrari F, Sainte-Rose B, Aitken J, Marthouse R, et al. Evidence that the Great Pacific Garbage Patch is
rapidly accumulating plastic. Sci Rep. 2018;8(1):4666. https://doi.org/10.1038/s41598-018-22939-w.
74. 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. Mar Pollut Bull. 2020;159:111467. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . m a r p o l b u l .
2 0 2 0 . 1 1 1 4 6 7 .
75. Angelini Z, Kinner N, Thibault J, Ramsey P , Fuld K. Marine debris visual identification assessment. Mar Pollut Bull.
2019;142:69–75. https://doi .org/10.101 6/j.marpolb ul.2019. 02.044.
76. Perera IJJUN, Sandaruwan RDC, Bellanthudawa BKA. Coastal and marine plastic pollution monitoring and control using
remote sensing (RS) and artificial intelligence (AI) technologies. Coastal and Marine Pollution: Source to Sink, Mitigation
and Management 2025. https://doi .org/10.100 2/978139423 7029.ch1 7Digital.
77. Ayana PP , Sethu MR, Nirmal A, Swetha MK, Mol VL. Identifying plastic pollution hotspots of Vembanad lake, Kerala, India:
an integrated approach using artificial intelligence and spatial analysis. Wetl Ecol Manag. 2025;33(4):49. h t t p s : / / d o i . o r g / 1 0 .
1 0 0 7 / s 1 1 2 7 3 - 0 2 5 - 1 0 0 6 2 - 3 .
78. Zhong Q, Bose N, Hwang J, Zou T. Exploring the potential of autonomous underwater vehicles for microplastic detection
in marine environments. Drones. 2025;9(8):580. https://doi.org/10.3390/drones9080580.
79. Kanhai LDK, Gårdfeldt K, Lyashevska O, Hassellöv M, Thompson RC, O’Connor I. Microplastics in sub-surface waters of the
Arctic Central Basin. Mar Pollut Bull. 2018;130:8–18. https://doi .org/10.101 6/j.marpolb ul.2018. 03.011.
80. Hartmann NB, Hüffer T, Thompson RC, Hassellöv M, Verschoor A, Daugaard AE, et al. Are we speaking the same language?
Recommendations for a definition and categorization framework for plastic debris. Environ Sci Technol. 2019;53(3):1039–
47. https://doi.org/10.1021/acs.est.8b05297.
81. McCauley SJ, Bjorndal KA. Conservation implications of dietary dilution from debris ingestion: sublethal effects in post‐
hatchling loggerhead sea turtles. Conserv Biol. 1999;13(4):925–9. https://doi .org/10.104 6/j.1523-17 39.1999. 98264.x.
82. Guerrera MC, Aragona M, Porcino C, Fazio F, Laurà R, Levanti M, et al. Micro and nano plastics distribution in fish as model
organisms: histopathology, blood response and bioaccumulation in different organs. Appl Sci. 2021;11(13):5768. h t t p s : / / d
o i . o r g / 1 0 . 3 3 9 0 / a p p 1 1 1 3 5 7 6 8 .
83. Käppler A, Windrich F, Löder MGJ, Malanin M, Fischer D, Labrenz M, et al. Identification of microplastics by FTIR and Raman
microscopy: a novel silicon filter substrate opens the important spectral range below 1300 cm−1 for FTIR transmission
measurements. Anal Bioanal Chem. 2015;407(22):6791–801. https://doi.org/10.1007/s00216-015-8850-8.
84. Yan X, Cao Z, Murphy A, Ye Y, Wang X, Qiao Y. FRDA: fingerprint region based data augmentation using explainable AI for
FTIR based microplastics classification. Sci Total Environ. 2023;896:165340. https://doi .org/10.101 6/j.scitote nv.2023. 165340.
85. Kikaki A, Karantzalos K, Power CA, Raitsos DE. Remotely sensing the source and transport of marine plastic debris in Bay
Islands of Honduras (Caribbean Sea). Remote Sens. 2020;12(11):1727. https://doi.org/10.3390/rs12111727.
86. Kako S, Morita S, Taneda T. Estimation of plastic marine debris volumes on beaches using unmanned aerial vehicles and
image processing based on deep learning. Mar Pollut Bull. 2020;155:111127. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . m a r p o l b u l . 2 0 2 0 . 1 1 1
1 2 7 .
87. Bianco V, Pirone D, Memmolo P , Merola F, Ferraro P . Identification of microplastics based on the fractal properties of their
holographic fingerprint. ACS Photon. 2021;8(7):2148–57. https://doi.org/10.1021/acsphotonics.1c00591.
88. McCann E, Li L, Pangle K, Johnson N, Eickholt J. An underwater observation dataset for fish classification and fishery
assessment. Sci Data. 2018;5(1):180190. https://doi.org/10.1038/sdata.2018.190.
89. Jia T, Kapelan Z, De Vries R, Vriend P , Peereboom EC, Okkerman I, et al. Deep learning for detecting macroplastic litter in
water bodies: a review. Water Res. 2023;231:119632. https://doi.org/10.1016/j.watres.2023.119632.
90. Sannigrahi S, Basu B, Basu AS, Pilla F. Development of automated marine floating plastic detection system using Sentinel-2
imagery and machine learning models. Mar Pollut Bull. 2022;178:113527. https://doi .org/10.101 6/j.marpolb ul.2022.
113527.

Page 26 of 26
Kaamoush and El-Zeiny Discover Sustainability (2026) 7:1963
91. Liu Y, et al. Spectral classification of large-scale blended (micro)plastics using FT-IR raw spectra and image-based machine
learning. Environ Sci Technol. 2023;57(16):6656–63. https://doi.org/10.1021/acs.est.2c08952.
92. Zadeh LA. Toward a theory of fuzzy information granulation and its centrality in human reasoning and fuzzy logic. Fuzzy
Sets Syst. 1997;90(2):111–27. https://doi.org/10.1016/s0165-0114(97)00077-8.
93. Moshgbar M, Parkin R, Bearman R. Application of fuzzy logic and neural network technologies in cone crusher control.
Miner Eng. 1995;8(1–2):41–50. https://doi.org/10.1016/0892-6875(94)00101-h.
94. Balas CE, Ergın A, Williams AT, Koc L. Marine litter prediction by artificial intelligence. Mar Pollut Bull. 2003;48(5–6):449–57.
https: //d oi. or g/10 . 1016/j.ma rpol bu l.2003.08.020.
95. Balas CE, Koç ML, Tür R. Artificial neural networks based on principal component analysis, fuzzy systems and fuzzy neural
networks for preliminary design of rubble mound breakwaters. Appl Ocean Res. 2010;32(4):425–33. h t t p s : / / d o i . o r g / 1 0 . 1 0 1
6 / j . a p o r . 2 0 1 0 . 0 9 . 0 0 5 .
96. Michel APM, Morrison AE, Presto VL, Marx CT, Colson BC, White HK. Rapid identification of marine plastic debris via
spectroscopic techniques and machine learning classifiers. Environmental Science & Technology, 2020;54(17), 10630–
10637.https://doi.org/10.1021/acs.est.0c02099.
97. Ananthi M, Lakshmana Kumar R, Muthu B, Punitha P . Adaptive marine intelligence and sensing architecture for autono-
mous underwater ecosystem monitoring using AI and IoT integration. Intell Data Anal. 2025. h t t p s : / / d o i . o r g / 1 0 . 1 1 7 7 / 1 0 8 8
4 6 7 X 2 5 1 3 3 9 2 7 1 .
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

论文原文与出版信息

Discover Sustainability
查看论文原文

继续探索