This review examines publications on digital technology for the decarbonizing maritime sector. It picked the Web of Science (WoS) database for in-depth analysis to understand the newest academic advancements, which is essential for capturing worldwide scholarship (Li et al., 2018). This review’s dataset is sourced from the two primary indexes in WoS: the Science Citation Index Expanded (SCI-EXPANDED) and the Social Science Citation Index (SSCI). It is confined to English-language literature, encompassing both journal articles and reviews, which cover the period from 2005 to 2024, as of November 12, 2024. English is the main language of international academic exchange, and the relevant literature is more representative. Considering the emergence of digital technologies, their application in the shipping industry, and the growing global focus on decarbonization, we set the starting point of our data as 2005. To ensure the publication search is both precise and exhaustive, thereby preventing the omission of critical information or the inclusion of irrelevant publications, we employed a strategy of selecting keywords that effectively articulate and explore the role of digital technologies in decarbonizing shipping. We examined research through four distinct processes, including a preliminary evaluation of keywords, broadening the spectrum of keywords, removing extraneous terms, iterative refinement of search terms, and removing predatory references. The publication search process is shown in Figure 1 . The search criteria are shown in Table 1 .

Bibliometrics is a field that elucidates scientific trends and structures via the quantitative analysis of literary data. This entails employing mathematical and statistical techniques to examine data in publications, including authors, keywords, and citations, to elucidate the distribution of scientific knowledge, collaborative networks, research hotspots, and trends (Dalle Lucca Tosi and dos Reis, 2020). Moreover, the advantage of bibliometrics compared to other literature review methodologies is its capacity to yield more objective and dependable outcomes, hence furnishing scholars with comprehensive insights into the evolution of a discipline (Wallin, 2005). Investigations in this domain can assist researchers, policymakers, and academic institutions in comprehending the dynamics of disciplines, evaluating research impact, and directing future research trajectories. Bibliometrics possesses diverse applications, enabling the evaluation of research production and impact of a country or institution, and the identification of research frontiers and seminal literature within a specific domain (Ninkov et al., 2022). Moreover, citation analysis enables bibliometrics to elucidate knowledge flows and the distribution pathways of scholarly concepts (Markus, 2003). The advancement of big data is propelling the evolution of bibliometric research methods. Bibliometric analysis has been employed across various industry sectors to detect research trends, difficulties, and possibilities within a certain topic (Muhuri et al., 2019). This publication aims to identify the current focal points of digital technology in shipping decarbonization applications via bibliometric analysis. To explore the research hotspots in the relevant field, we employed VOSviewer, a complimentary Java-based software created by the Centre for Technology Research at Leiden University in the Netherlands, to visualize and analyze countries, authors, and institutions, culminating (Bui et al., 2020).

The H-index, or Hirsch index, quantifies the academic effect of an individual scholar, researcher, or research organization. Introduced by American physicist Jorge Hirsch in 2005, the fundamental concept of the H-index posits that a scholar’s H-index corresponds to the number of instances in which at least h of their publications have been mentioned no fewer than h times, while citations for any additional publication do not surpass h (Hirsch, 2005). The H-index possesses multiple advantages: firstly, it amalgamates the number of publications and citations of researchers, thereby reflecting their academic impact; secondly, the H-index exhibits relative stability, remaining unaffected by atypically high citations of individual publications, rendering it a more objective metric for academic assessment; lastly, the H-index is straightforward to comprehend and compute, contributing to its widespread application in academic circles. This review uses the H-index to evaluate the research performance of institutions and nations.

Analyzing the fluctuations in publication volume within a specific domain can elucidate the evolving patterns and focal points of research in that area (Adnan et al., 2023; Shi et al., 2023). The annual number of publications from 2015 to 2024 is shown in Figure 2 . This review examines the trend in publications on digital technology used for decarbonizing shipping during the past decade. The writing on this topic commenced in 2015, with a single publication, which may be attributed to the fact that in the early 21st century, there were fewer studies on digital technologies such as AI in shipping due to the slow development of technologies such as AI (Xiao and Xu, 2024; Xiao et al., 2024b). Conversely, the total number of annual publications in 2023 is 55, representing a 54-fold rise in the overall count. The period from 2015 to 2020 marks the beginning of a growing publication volume within the discipline, culminating in the release of 28 publications. This results from the groundbreaking research conducted by scholars in this domain, including Gkerekos et al. (2019), who investigated the efficacy of various multiple regression algorithms for predicting the Fuel Oil Consumption (FOC) of a vessel using diverse data sampling frequencies. From 2020 to 2023, the quantity of publications in this domain increased substantially, by almost threefold. Furthermore, 45 publications in this domain were published between January and November 12, 2024, indicating a foreseeable increase. This signifies that an increasing number of scholars have focused on studying the application of digital technology in decarbonizing shipping in recent years.

Co-occurrence analysis is a content analysis method that accurately represents the intensity of correlation between information items in textual data (Wang et al., 2012). The co-occurrence of literature terms is studied to assess the intensity of association among keywords and ascertain the composition, research focus, and future trends. We employed VOSviewer to create the keyword co-occurrence network, depicted in Figure 6 . Keyword co-occurrence denotes the concurrent emergence of two keywords; each item signifies a keyword, with larger nodes indicating a higher frequency of keyword occurrence. The proximity between two elements (i.e., keywords) signifies the intensity of their relationship, with a reduced distance reflecting a more robust association. The color of the items signifies the clustering outcomes, with items sharing the same color denoting membership in the same cluster. This keyword co-occurrence network has five distinct clustering groups, as Figure 6 illustrates, which results in five distinct domains.

Figure 7A depicts the application of digital technology in optimization, and prediction, encompassing ship speed optimization, trim optimization, oil consumption forecasting, energy consumption forecasting, and hull design. Emissions from vessels in seaports are progressively eliciting concerns regarding environmental sustainability. Precise emission forecasts can facilitate the monitoring and resolution of this issue, hence enhancing the sustainability of maritime transportation (Liu et al., 2023; Chen et al., 2024a). To attain optimal environmental sustainability, the selected energy technology on board must be integrated with the operational mode within specified limits, such as routes, operations, and speed (Elg et al., 2023). In this domain, machine algorithms have been utilized most frequently, with 40 instances, links to 40 keywords, and an overall cooperation intensity of 107, the highest among all keywords. Due to their significance, we enumerated the applications of machine learning and deep learning in Table 6 . Simultaneously, machine learning, optimization, speed optimization, prediction, and oil consumption prediction have connection strengths of 8, 7, 4, and 7, respectively, making them the most interconnected keywords to these four terms.

Figure 7B emphasizes the application of digital technology in maritime emissions, encompassing emission prediction, inventory management, and air quality assessment (He et al., 2022). Considering global warming, effectively measuring and reducing ship emissions has emerged as a significant problem for the transportation industry and society. In recent years, a significant advancement in emission accounting has arisen, specifically using big data, particularly data derived from AIS (Yin et al., 2021).

Figure 7C emphasizes the application of digital technologies in logistics and supply chain management, encompassing ship energy efficiency optimization and emission forecasting, using technologies such as blockchain, big data, and machine learning. Maritime supply chain management is critical to ensuring the smooth flow of global trade, reducing transport costs, and improving logistics efficiency (Xu et al., 2021, 2022). For instance, Perera, etc. introduced a Machine Intelligence (MI)-based data processing framework for ship performance and navigation data to formulate suitable navigation strategies (Perera and Mo, 2017).

Figure 7D emphasizes model development and optimization, centering on the ship FCP model, which incorporates digital technologies, including blockchain, IoT, machine learning, and AIS. The FCP in ships is crucial to decarbonization initiatives, enhancing energy efficiency and reducing GHG emissions from international shipping (Nguyen et al., 2023). Figure 7E illustrates additional applications of digital technology in carbon reduction within shipping, encompassing ship-port interfaces, identification systems, and route optimization.

A burst keyword is a phenomenon where a keyword exhibits a high frequency of occurrence within a designated timeframe. This information illustrates the history of research hotspots, indicates recent research trends, and aids in forecasting future research directions (Lyu et al., 2023). To better understand the evolution of research trajectories, the burst word analysis is presented in Table 7 . The years during which the 12 burst words emerged are 2016–2024. The term “algorithm” emerged as a prominent buzzword from 2016 to 2018. Oil depletion, environmental change, condition monitoring, and blockchain have been extensively examined since 2020. Air quality, AIS, and autonomous ships emerged as prominent research topics in 2021–2022. Since 2022, ANN, speed optimization, and neural networks have emerged as prominent research focal points.

The relationship between FCP and speed is not linear. FCP is considered to be proportional to the third power of the ship’s speed, so a small decrease in speed implies a significant decrease in FCP (Alvarez et al., 2010; Fagerholt et al., 2010). Shipping companies have widely used digital technology for speed optimization. For example, global shipping company Maersk uses a speed optimization system based on big data and machine learning, which has helped Maersk reduce FCP by about 10-15% on several routes (Luo et al., 2024b). Several researchers have offered valuable insights and technical solutions through various methodologies and viewpoints in speed optimization, they have continued to refine their research from the perspective of using weather archive data and route segmentation methods and combining longitudinal inclination optimization. Kim and Lee (2018) centered on employing Dynamic Programming (DP) techniques to develop speed adjustment strategies for ships, targeting energy conservation and emission reduction by efficiently managing external factors. Their research integrated AIS data with marine environmental data through MapReduce to analyze and quantify speed variations caused by external factors. Previous literature has included theoretical FCP functions in speed optimization models; however, these functions are constrained by meteorological circumstances encountered during navigation. Lee et al. (2018) adopted an alternative strategy by creating a decision support system to enhance speed by integrating weather archive data and a Particle Swarm Optimization (PSO) methodology. Li et al. (2022) utilized the route segmentation approach to develop a machine-learning model for predicting FCP and shaft speed. They introduced an iterative strategy for optimizing route segmentation, weather loading, and speed. (Xie et al., 2023a) integrated the FCP prediction model from the prior study. They introduced optimization techniques for speed, longitudinal gradient, and nodes across various wind and wave conditions in each segment.

Summary and prospects: Speed affects only a quarter of total supply chain costs, but it impacts half of delivery times and more than 70% of the carbon footprint. To reduce ship emissions, scholars have proposed rational control of ship speed, acceleration, engine speed, and torque (Fan et al., 2024). Future research directions of digital technology in speed optimization can focus on the in-depth exploration of multi-objective optimization methods, such as real-time balance between economic benefits and environmental impacts by introducing dynamic weight adjustment strategies (Li et al., 2024a). In addition, the FCP prediction model is optimized by combining more advanced machine learning techniques to improve prediction accuracy and adaptability. Intelligent algorithms are further developed to solve multivariate dynamic speed optimization problems in complex environments, such as integrating wind and wave conditions, shiploads, and route adjustments. At the same time, practitioners and people who develop digital technologies can promote deeper integration between digital technologies, such as synergizing machine learning with IoT, big data analytics, and real-time sensing technologies to enhance comprehensive optimization capabilities. Finally, global optimization methods based on big data and real-time information processing are explored to provide more efficient solutions for ship decarbonization.

The shipping industry, a significant contributor to GHG and pollutant emissions, has garnered international attention about emission reduction due to the escalating global concern over climate change and environmental protection (Xiao et al., 2023; Luo et al., 2024a). In the area of emission prediction, researchers have used techniques such as machine learning and big data to improve their research in terms of optimization of prediction models, data sources, and expansion of prediction methods.

Initially, researchers created numerous machine learning models to enhance predictive accuracy regarding emission prediction. Ren et al. (2022) developed an algorithmic model based on ridge regression through data fusion techniques, combining dynamic information of the ship and meteorological conditions. Papandreou and Ziakopoulos (2022) developed an FCP model using machine learning algorithms, especially XGBoost. Some researchers also improved the study by combining machine learning, big data, and neural networks, in terms of optimizing the FCP prediction model, and data sources and extending the prediction methods. Hu et al. (2019) developed machine learning to predict the FCP of ships under the influence of different marine environmental factors, in particular Back Propagation Neural Networks (BPNN) and Gaussian Process Regression (GPR). Tran (2021) applied a combination of Monte Carlo (MC) simulation methods and ANN to develop a probabilistic model based on machine learning, for main diesel engine FCP prediction. Using meteorological data and machine learning, Jaques et al. (2023) developed a ridge regression-based algorithmic model to automate FCP estimation under different weather conditions through data fusion techniques, combining the ship’s dynamic information and meteorological conditions. Cao et al. (2021) suggested a methodology that integrates optical remote sensing with deep learning approaches to forecast pollutant concentrations and assess fuel sulfur content by analyzing ship exhaust plume images. He et al. (2022) devised a rapid approach utilizing data panning, which markedly enhances the computing efficiency of emission dispersion concentrations from many vessels over extended durations and distances while maintaining accuracy. Nonetheless, these publications have also uncovered several hurdles to emission reduction within the maritime sector. The initial issue is the quality and accessibility of data. Su et al. (2024) demonstrated that the CatBoost algorithm excelled in estimating the FCP of PCTC ships, although random forests or neural networks may be more appropriate in alternative investigations. The research conducted by Yuan et al. (2024) uncovered the unintended consequences of emission reduction strategies, including the IMO 2020 fuel sulfur regulation, which may exacerbate global warming even as air quality improves, indicating the necessity for a thorough evaluation of the long-term effects of such measures.

Summary and prospects:

Autonomous ships can significantly reduce shipping carbon emissions by optimizing operations to improve fuel efficiency while integrating alternative power systems such as methanol and electricity (Yan et al., 2024). These technologies have demonstrated clear environmental and economic advantages in short- and medium-haul routes. In light of the swift advancement of the global transportation sector and increasing apprehensions regarding climate change, the marine industry is diligently investigating alternative energy sources and using innovative technology to improve safety efficiency and facilitate decarbonizing shipping (Xiao et al., 2024a; Xiao and Xu, 2024). Autonomous ships and navigation facilitated by digital technology have emerged as a prominent research focus. In this area, the study advances from optimizing the establishment and optimization of alternative fuel and power systems, efficient communication systems, and precise navigation systems.

Jovanovic et al. (2022) studied the feasibility of unmanned ro-ro passenger ships in the Adriatic Sea region, analyzing the life cycle, environmental, and economic performance of three ships on short, medium, and long routes. They found that combining automation with alternative fuels can bring environmental and economic benefits in specific cases. Of these, methanol-powered and electric vessels perform best. In addition, the economic advantages of automated vessels are affected by the price volatility of fossils and alternative fuels, highlighting the environmental advantages and economic uncertainties in the development of autonomous vessels. The enhancement of marine communication systems has emerged as a critical concern concurrent with the advancement of autonomous ships, and it is this challenge that the research by Jurdana et al. (2021) offers a unique solution to. These systems help to optimize route and traffic management by supporting real-time navigation, monitoring, and operational control of autonomous vessels, thereby reducing FCP. Current maritime communication systems encounter limitations, including low data rates, high costs, and restricted capacity, particularly in the Global Maritime Distress and Safety System (GMDSS) and AIS. Jurdana et al. proposed a variable record length data compression technique utilizing differential binary coding to mitigate this issue. This approach accomplished compression by transmitting solely the data items that have altered since the previous transmission, thus markedly diminishing the volume of data sent. A recent publication examines the enhancement of route planning and navigational safety for MASS in maritime environments. Li and Yang (2023) introduced a machine learning methodology leveraging AIS data, employing AADTW, SCAF, and a novel route optimization algorithm for MASS to facilitate feature extraction and unsupervised route planning. The strategy significantly enhanced route planning for real-time collision avoidance by emphasizing safety at critical hotspots at intersections of established routes through a temporal analysis approach.

Summary and prospects: These publications indicate several critical avenues for the advancement of autonomous ships:

In addition, Given the increasing global emphasis on reducing carbon emissions, changes in relevant policies and regulations will play a substantial role in shaping the trajectory of decarbonization efforts within the shipping industry (Chen et al., 2023, 2024b). The IMO has implemented several regulations to reduce the carbon footprint of maritime transportation. Two important emission regulations are the EU Monitoring, Reporting and Verification (EU-MRV) and the IMO Data Collection System (IMO DCS), both of which are part of the Ship Energy Efficiency Management Plan (SEEMP) (Kanberoglu and Kökkülünk, 2021). To reduce CO₂ emissions, the Chinese government has established a carbon trading mechanism that allocates free emission allowances to industrial companies, while companies exceeding the emission allowances are not allowed to use CO₂ in their operations (Wang et al., 2023). Researchers should continually monitor and incorporate these evolving policies to ensure that their findings are both consistent with market demand and relevant to regulatory expectations.