中国P站

The Scale and Sources of Marine Pollution

The oceans are facing growing threats from waste produced by humans. By 2025, projections indicate that the oceans could contain between 75 and 199 million tons of plastic waste, resulting in the annual death of over 100 million marine animals due to plastic debris [1]. It is estimated that over 8 million tons of plastic make their way into the ocean annually, with micro plastics discovered in the deepest ocean trenches and even within human bloodstreams [2]. Pollution comes from various sources, including poorly managed solid waste and litter on land, as well as oil spills, industrial discharges, and micro plastics. Currently, land-based solid waste systems are being improved to sort, classify, and minimize plastic leakage more effectively than ever before [3], highlighting the essential role that upstream interventions have in safeguarding marine environments. These pollutants present significant dangers to marine ecosystems and human health, highlighting an immediate necessity for more advanced monitoring and management approaches. Conventional approaches, like manual beach clean-ups and local water sampling, require significant effort and fall short in tackling pollution on a global level [4].

AI as a Technological Response

In this context, advanced technology has emerged as a promising ally. Advanced systems, featuring sophisticated sensors, computer vision, and learning algorithms, provide the capability to analyze extensive environmental datasets in real time, identify patterns or anomalies, and facilitate data-driven decisions for intervention [4]. This review is centered around a key inquiry: What effects does the implementation of AI have on the tracking and management of ocean waste?

In recent years, there has been a significant surge in interest in utilizing artificial intelligence (AI) to tackle issues related to marine pollution. The number of studies on AI applications for ocean waste has surged dramatically, surpassing 10,000 publications from 2004 to 2024, with a notable high of over 2,100 papers published in 2023 (Figure 1). This increase corresponds with improvements in AI technologies and worldwide demands for sustainable development strategies [4]. Researchers are exploring innovative methods powered by AI throughout the pollution life cycle-from enhancing recycling and waste management on land to tracking pollution in rivers and oceans and even assisting in the collection and removal of debris (Figure 1) (Institute for Operations Research and the Management Sciences, 2025).

Figure 1: Accuracy of marine pollution detection by method. Human analysts aloneachieved around 78% identification accuracy, an AI model achieved about 94%,and combining human expertise with AI yielded approximately 97% accuracy.This illustrates the performance improvements offered by AI in environmentalmonitoring. (Alt text: Bar chart comparing detection accuracy: Human ~78%, AI~94%, Human+AI ~97%.

This review is primarily supported by a few theoretical perspectives. Firstly, The Environmental Governance Theory and Policy Cycle which highlights that successful environmental management necessitates collaboration among various stakeholders, including governments, NGOs, industries, and communities. When it comes to enhancing ocean health through technology, collaboration is key; sharing data and forming partnerships across various sectors are crucial for the successful development and widespread implementation of these tools. Secondly, a perspective on the Technological Sustainability Theory highlights that although advanced solutions such as AI present possible advantages, they also entail environmental repercussions. Utilizing AI models often requires substantial energy and computing power, leading to increased carbon emissions and electronic waste. Therefore, it is essential to consider the overall environmental effects of AI-driven initiatives, making sure that all of them truly achieves a favourable outcome [5]. These viewpoints shape our examination of AI's role from a broader perspective; we seek not just technical efficiency, but also harmony with cooperative governance and enduring sustainability.

This literature review begins by outlining the research question; what effects does the implementation of AI have on the tracking and management of ocean waste? The following section details the process of selecting and analyzing relevant studies through different methodologies. The main body presents findings from empirical research on applications in marine pollution management, organized by key functional areas: monitoring, prediction, prevention, and clean-up. In the evaluation of the studies, we examine the strengths and weaknesses of the current research, including challenges related to data availability, barriers to implementation, and the generalizability of findings. We synthesize the insights to comprehensively address the research question, pinpointing how technology is transforming ocean waste management and highlighting the existing gaps. In conclusion, the review offers insights into potential future paths.

Methods

Literature Search and Selection

The employment of a structured search strategy to collect relevant academic literature on AI applications in marine pollution monitoring and management was done. Electronic databases (such as Web of Science, Scopus, and Google Scholar) were queried using keywords including “artificial intelligence,” “machine learning,” “marine pollution,” “ocean waste,” “plastic debris,” “oil spill,” and “waste management.” The search was restricted to publications between 2018 and 2025, capturing the recent surge in research interest, although earlier works were also considered for context. Many different sources were gathered, including peer-reviewed journal articles, conference proceedings, technical reports, and a few credible news releases or institutional reports for up-to-date statistics and real-world project information. These articles were selected between May of 2025 to September of 2025.

Screening and Inclusion Criteria

Screening of sources was done in two stages. First, titles and abstracts were reviewed to filter for relevance to ocean waste and AI; we excluded papers that dealt with AI in environmental domains unrelated to pollution (e.g., terrestrial wildlife conservation AI, unless they had transferable insights). Second, the remaining sources were read in full and assessed for fit to study criteria (or something of that sort). Preference was given to studies with clear methodologies (e.g., defined datasets, evaluation metrics for AI models) and to review papers that provided broad coverage of the field. We also included systematic reviews and bibliometric analyses for an overarching view of research trends [6]. Overall, the screening resulted in us finding many different categories of AI applications.

Results

Monitoring and Detection of Marine Pollution

AI technologies are revolutionizing the way marine pollution is monitored, enabling real-time, wide-area surveillance that manual methods cannot achieve. Tools like convolutional neural networks (CNNs), U-Net, and Mask R-CNN are used to analyze satellite and drone imagery to detect floating plastics, oil slicks, and other pollutants with high precision. These systems can distinguish between pollution and natural features such as algal blooms, improving the accuracy of alerts and early detection. In addition, AI-powered camera networks on rivers and boats are helping identify and quantify debris before it enters the ocean, allowing for targeted interventions. This shift toward automated, continuous monitoring has greatly increased the speed, scale, and responsiveness of marine pollution detection efforts.

Prediction and Decision-Support Systems

Beyond detection, AI is being used to forecast pollution events and support smarter decision-making. Predictive models integrate environmental data such as ocean currents, wind patterns, and historical pollution trends to anticipate future hotspots and guide interventions. For example, wastewater treatment facilities can use machine learning algorithms to predict pollutant surges following rainfall and adjust processes in real time, minimizing environmental discharge. These tools are also being integrated into long-term sustainability planning, helping policymakers simulate and evaluate the outcomes of different waste management strategies. Such capabilities make AI a valuable asset not only for real-time operations but also for building proactive and adaptive pollution management systems.

Clean-up Optimization and Pollution Management

AI contributes to clean-up efforts by optimizing logistics and improving the efficiency of debris collection. Projects like The Ocean Clean-up have used AI-based routing algorithms to guide collection vessels, resulting in up to 60% more plastic retrieved without additional fuel costs. AI is also being explored in experimental technologies such as autonomous drones and microbots that can detect and degrade micro plastics in situ. On land, AI is used to reduce ocean-bound waste through smart recycling systems that automate sorting and improve recovery rates. While some of these technologies are still in early development, they reflect a broader trend of integrating AI across the entire pollution lifecycle-from source prevention to final removal. These three subcategories were the primary categories of research articles within this subtopic (Figure 2).

Figure 2: Categories of AI applications in marine pollution based on a systematic review by Ning et al. (2024). 57% of AI applications focus on pollution monitoring (e.g., detecting oil spills or floating plastics), 24% on pollution management (e.g., optimizing waste collection or treatment processes), and 19% on prediction (for forecasting pollution events or hotspots). Key target areas for AI include oil spill detection, water quality monitoring, and identification of plastic debris.

Overview of Research Maturity and Strengths

It is essential to evaluate the quality and constraints of the body of research that holds this review up. Overall, the literature on ocean waste management is rapidly evolving but remains in relatively early stages, characterized by numerous proof-of-concept studies and rudimentary projects. Several systematic reviews and analyses enhance credibility by synthesizing results from numerous studies. For instance, Prakash and Zielinski (2025) synthesize 53 recent studies and emphasize that a fundamental group of models (Random Forest, U-Net, GANs, Mask R-CNN, YOLO) consistently demonstrate high success rates in detecting marine pollutants [7]. The alignment of findings from numerous independent studies-such as different research groups attaining over 90% accuracy in the classification of marine debris images through deep neural networks-enhances the reliability of these results. Bibliometric analyses further indicate that the research community has thoroughly peer-reviewed these methodologies over the last ten years [4].

Limitations

Nonetheless, noteworthy limitations exist. A significant number of studies are conducted using restricted datasets or localized programs, leading to concerns regarding their generalizability. A model developed to identify waste in the coastal waters of a specific area, such as the Catalonian coast in the MARLIT project [8], may exhibit reduced effectiveness in other environments unless it undergoes retraining. Researchers often highlight the challenges posed by data scarcity and bias, noting that satellite imagery and labelled datasets concerning ocean pollution are neither plentiful nor consistently gathered [7] This indicates that the performance metrics reported may appear overly favourable when evaluated on limited, carefully selected datasets, and may decline under varied, real-world conditions.

Importantly, the synthesis must acknowledge that AI is not a silver bullet. Many advanced AI-driven proposals (like autonomous drone fleets to collect trash, or AI-guided “plastic-eating” robots) remain at prototype levels. The review uncovered consistent limitations that temper optimism. One issue is that AI models, no matter how sophisticated, depend on data – and the ocean is a challenging data environment. Sparse sensor coverage, variability of conditions (weather, lighting, sea state) and novelty of pollution (new types of plastics or polymers) can all degrade AI performance [7] Thus, robust deployment of AI requires parallel investment in sensing infrastructure (satellites, drones, IoT ocean sensors) and big data management, as well as periodic retraining of models to handle changing distributions of inputs.

A further quality concern is that numerous AI applications have yet to be implemented beyond the stages of simulations or laboratory trials. Ditria et al. (2022) observe that while various automated monitoring tools appear promising in theory, they have yet to yield quantifiable conservation results in practice [9]. The authors highlight the necessity of transcending technical innovation to assess the actual impact of AI tools on ecological conservation initiatives in practice-a challenge that is reflected throughout the academic conversation [10]. In a similar sense, numerous "smart" clean-up concepts, such as swarms of cleaning robots or predictive re-routing of waste, are documented in the literature, yet lack comprehensive field validation at a full scale. The disparity between research findings and practical application indicates a need for careful consideration regarding the commonality of research findings compared to the data proven through experimentation on the topic vary greatly, with research findings covering majority of all literature. The reason for this is because the tools must undergo rigorous evaluation in marine environments over an extended period, their genuine influence remains ambiguous.

The Need for Collaboration

Ultimately, this field possesses an inherent interdisciplinary complexity that influences the quality of study. Effective solutions in this domain necessitate the integration of expertise from oceanography, engineering, and computer science. Not all studies are comprehensive in addressing every facet; a technically proficient AI model may neglect practical maritime limitations, while a marine science paper might employ AI methodologies in a less-than-optimal manner. Our observations indicate that contributions of the highest quality are often derived from collaborative teams or extensive multi-author reviews that integrate a variety of expertise. This highlights that, moving ahead, increased interdisciplinary collaboration and transparent sharing of data and methodologies will be crucial for enhancing the quality and impact of research in the field of marine pollution management.

Summary of AI Applications Across the Lifecycle

Examples of AI Applications in Marine Waste Management, illustrating the wide range of approaches across the pollution lifecycle. Each example is drawn from the literature and highlights the AI technique, its purpose, and any demonstrated impact (Table 1).

Table 1. illustrates the multi-faceted role of AI: from preventive measures to end-of-pipe solutions. It also highlights the varying maturity levels – some solutions are operational (e.g., route optimization), while others are in early research stages (e.g., microplastic microbots).

Discussion

Advancements in Monitoring and Analysis

Drawing together the theoretical framework, empirical evidence, and quality appraisal, we synthesize how AI is impacting ocean waste tracking and management. The most immediate contributions of AI are in the realm of monitoring and data analysis, effectively “augmenting the senses” of environmental managers. Instead of sporadic ship-based observations or manual aerial surveys, AI enables continuous, automated surveillance of vast ocean areas. For example, convolutional neural networks processing satellite images can detect features of pollution (oil sheen, debris fields) with high precision, flagging problems that human analysts might miss [7]. AI systems are also increasingly being used for predictive modeling and anomaly detection in maritime environments, allowing managers to anticipate pollution events before they escalate [11]. This improves early detection of events like oil spills or illegal waste dumping, allowing quicker response to contain damage. Notably, Multi-sensor AI models (combining optical imagery, radar, and even social media data) are being developed to improve reliability under different conditions [7]. For example, Edeh et al. (2024) developed a novel deep learning architecture that accurately predicts marine pollution patterns using environmental and satellite data inputs-showcasing how AI is evolving toward capabilities that support sustainable ocean management [12]. The speed and scalability of AI-driven monitoring mark a substantial improvement over conventional methods-tasks that once took weeks’ worth of human effort can be done in hours. As a result, agencies can move toward real-time pollution tracking, which is crucial for dynamic ocean phenomena (e.g., currents can spread oil or trash quickly). In sum, AI is shifting marine pollution monitoring from a reactive, patchy endeavour to a more proactive and comprehensive system.

Operational and Industrial Applications

Beyond monitoring, AI is starting to influence on-the-ground (and on-the-water) management actions. One prominent example is The Ocean Clean-up’s adoption of an AI-based routing algorithm to guide their collection vessels (Institute for Operations Research and the Management Sciences, 2025). This illustrates how AI can not only identify where pollution is but also optimize how to remove it. By crunching data on ocean currents, wind, and known garbage patch locations, the algorithm finds efficient pathways that human planners would struggle to compute quickly. The reported 60% boost in plastic removal efficiency is a tangible real-world impact-more waste is being collected per mission, accelerating progress toward cleaner oceans (Figure 3) (Institute for Operations Research and the Management Sciences, 2025). Another management application is wastewater treatment and water quality management. AI controllers (using machine learning predictions) can adjust treatment plant processes to better remove contaminants or predict when pollutant loads will spike (e.g., after heavy rains). These systems show marked improvements in efficiency, accuracy, and operational cost when compared to traditional techniques [13]. This helps prevent untreated wastewater from reaching marine environments, thus reducing one source of ocean pollution. While these industrial applications are less visible to the public, they form a critical part of a holistic ocean pollution strategy by cutting off pollution at the source.

Another recurring theme is the need for interdisciplinary and cross-sector collaboration. Successful case studies often involve partnerships: e.g., the Dutch NGO using AI in rivers combines tech experts with local environmental groups [9] the Plastics “Inventory” project compiled by engineers and conservationists identified 52 technologies against pollution [14]. Effective AI for ocean waste sits at the intersection of marine science, AI engineering, and policy. The Environmental Governance Theory is borne out in the literature-researchers repeatedly call for cooperation among scientists, policymakers, and stakeholders to share data and align AI solutions with regulatory frameworks [4]. For instance, one recommendation is to develop common data standards and open databases of marine litter, so that AI models worldwide can benefit from larger training sets and avoid duplicating effort. There is also recognition that legal and ethical guidelines will be needed for AI monitoring (e.g., using drones or satellite images raises privacy or sovereignty issues if not handled carefully) [4]. This aligns with [15], who argue that integrating AI into circular economy frameworks and marine governance will require new international standards for data sharing and government regulations. In summary, the impact of AI will be maximized when it is embedded in a cooperative, well-governed ecosystem of ocean stewardship, rather than pursued in isolated tech silos.

Sustainability Trade-offs and Long-Term Vision

From a long-term sustainability perspective, our synthesis finds an interesting picture. AI can undoubtedly help achieve sustainability goals related to oceans (UN SDG 14), by providing tools to reduce pollution and protect marine life. There are inspiring examples of this: AI vision systems that enforce sustainable fishing practices by recognizing illegal catch, or IBM’s AI-powered microscopes monitoring plankton health as an early warning for ocean toxins [9]. These illustrate AI’s potential as a guardian of marine ecosystems. Yet, echoing the Technological Sustainability Theory, authors caution against overlooking the environmental cost of AI itself (Khaukeral et al., 2018). Large AI computations (like training a deep neural network) can consume significant electricity, often from carbon-intensive sources, and hardware used in AI (sensors, data centers) eventually contributes to electronic waste. Wu et al. (2022) and others note that the net benefit of “AI for the environment” must be proven – if an AI solution’s carbon footprint outweighs its conservation benefits, then its implementation would be counterproductive. Fortunately, in the domain of marine pollution, many AI tasks (such as running inference on a small drone or processing satellite images periodically) are not as energy demanding as, say, training a massive language model. Additionally, many projects leverage cloud computing infrastructure that is increasingly being offset by renewable energy. Still, the field is beginning to account for this imbalance. Future designs might include more energy-efficient algorithms or the use of low-power edge AI devices for monitoring to minimize impact (Khakurel et al., 2018).

In essence, the implementation of AI is changing ocean waste management in meaningful ways, but its impact is a function of both technological efficacy and smart integration into broader systems. AI excels at handling the “data overload” – making sense of large-scale, complex, and dynamic environmental data that humans alone cannot process in real time. This strength is already improving situational awareness and guiding interventions. When it comes to physical outcomes (less waste in the water), AI is an enabler: a navigation algorithm can guide a ship, or a detection model can pinpoint clean-up targets, but humans and machines still need to do the actual removal and policy enforcement. Thus, current AI advances act as force-multipliers for human-led efforts, boosting efficiency (as seen with the 60% clean-up increase) and precision (targeting the 1000 most polluting rivers could dramatically reduce inputs) (Isabelle & Westerlund, 2022). If the remaining challenges – data sharing, cross-sector coordination, and sustainable AI practices – are addressed, AI’s role could expand from a set of isolated solutions to a global intelligent marine stewardship network. Such a network might continuously monitor ocean health, predict emerging pollution problems, and coordinate rapid responses, all while minimizing its own footprint. The literature suggests that achieving this vision will require continued innovation and vigilance, but the foundational pieces are falling into place.

Conclusion

In summary, the integration of artificial intelligence into the monitoring and administration of ocean waste is revolutionizing the way we address marine pollution. AI-driven tools facilitate earlier detection of pollutants, more efficient operations, and data-informed decision-making that were previously unattainable using manual methods alone. This review has emphasized a multitude of implementations, including smart optimization models that chart the quickest path to a plastic-free ocean [17] and deep learning algorithms that map floating garbage from aerial images [16]. The current body of evidence suggests that AI has the potential to substantially improve our ability to monitor and safeguard marine environments. In terms of practical application, these technologies have the potential to mitigate the overall burden of pollution by intercepting waste in rivers or enhancing recycling to prevent ocean entry, as well as to protect marine fauna by identifying and removing debris before it causes damage [18-20].

Nevertheless, the pursuit of AI-enhanced ocean stewardship requires a deliberate approach. The next critical step is to bridge the divide between pilot projects and large-scale deployment, as many AI applications are still in the nascent stages. To ensure that AI solutions are accessible and adaptable for use in various regions-including developing coastal nations that may lack advanced technical infrastructure-researchers, policymakers, and practitioners must collaborate. Additionally, it will be crucial to maintain a simple perspective on the costs and constraints of AI. The objective should be to integrate AI into a more comprehensive sustainability strategy, rather than treating it as a technological solution in isolation. This encompasses the development of comprehensive data ecosystems, regulatory frameworks for data governance and privacy, and capacity-building initiatives to enable local communities to interact with AI tools (e.g., citizen science programs that employ AI to report contamination).

The coexistence of AI and human society is both inevitable and expanding. It is an imperative endeavour to capitalize on this coexistence to benefit the planet. The literature reviewed in this article suggests that AI has the potential to be an effective ally in the restoration of ocean health if it is governed by sound environmental science and ethics. We are currently experiencing the initial phases of AI applications, which have the potential to result in near real-time global ocean monitoring and more intelligent interventions to maintain their cleanliness soon. It is probable that the extent to which these innovations contribute to the resolution of one of our most urgent environmental crises will be determined in the upcoming decade. Utilizing the potential of AI while simultaneously mitigating its risks could allow us to reverse the trend of ocean pollution, thereby guaranteeing healthier oceans for future generations.

References

1. Condor Ferries (2025). .
2. Fisher, C. (2023, 30 December). . Recycle Track Systems.
3. Ihsanullah I, Alam G, Jamal A, Shaik F (2022) Chemosphere, 309: 136631.

   , ,

4. Adeoba MI, Pandelani T, Ngwangwa H, Masebe T (2025) . Sustainability 17:3912.

   , ,

5. Khakurel J, Penzenstadler B, Porras J, Knutas A, Zhang W (2018) . Technologies 6: 100.

   , ,

6. Song T, Pang C, Hou B, Xu G, Xue J et al. (2023) .

   , ,

7. Prakash, N. and Zielinski, O. (2025) . Frontiers in Marine Science 12: 1486615.

   , ,

8. Alayón CL, Säfsten K, Johansson G (2022) .

   , ,

9. Khaukera DA, and Westerlund M, (2022) . Sustainability 14: 1979.

   , ,

10. Ditria EM, Buelow CA, Gonzalez-Rivero M, Connolly RM (2022) . Frontiers in Marine Science, 9, 918104.

    , ,

11. Agarwala N (2021) . Maritime Technology and Research 3: 120-136.

    , ,

12. Edeh MO, Dalal S, Alhussein M, Aurangzeb K, Seth B et al. (2024) . Peer J Computer Science, 10, e2482.

    , ,

13. Altowayti WAH, Shahir S, Othman N, Eisa TAE, Yafooz WMS (2022) . Processes, 10: 1832.

    ,

14. Nicholas School of the Environment (2022) .

15. Seyyedi S, reza Kowsari E, Ramakrishna S, Gheibi M, Chinnappan A (2023) . J Env Man 345: 118591.

    , ,

16. AZoRobotics (2021) . AZoRobotics.
17. (2025). AI-powered tech supercharges ocean clean-up, boosting plastic collection by 60%. Phys.org News.
18. Ning J, Pang S, Arifin Z, Zhang Y, Epa UPK, Qu M (2024) . J Marine Sci Eng 12: 1181.

    , ,

19. Schmaltz, E. (2020). Env Int144: 106067.

    , ,

20. Wu CJ, Raghavendra R, Gupta U, Acun B, Ardalani N et.al. (2022) .

    , ,