Toward digital twin of the ocean: from digitalization to cloning

The forthcoming wave of progress in oceanographic technology is the digital twin of the ocean, a concept that integrates marine big data and artificial intelligence (AI). This development is a logical consequence of combining data science and marine science and is considered superior to previous models, such as the digital ocean, transparent ocean, and smart ocean. Amid the swift advancement of next-generation information technology, the conditions are favorable for developing a prototype digital twin of the ocean, which will integrate various functionalities—data fusion, situation presentation, phenomenon mining, autonomous learning, and intelligent prediction. The salient distinction between a digital twin of the ocean and traditional forms of virtual or augmented reality is because of the intelligence beyond digitalization exhibited by the former, primarily facilitated by AI-based cloning. Hence, herein, we initially propose a structured architecture for the generative digital twin ocean, encompassing elements from real-time data pools to key technologies and proof-of-concept applications. The core components of this prototype system include a data pool, an AI-based oceanographic model, and three-dimensional visualization interactions. Future research and objectives for the digital twin ocean will principally focus on the following: four-dimensional (comprising three-dimensional space along with time) digital cloning and real-time mapping of global ocean parameters, cooperative observation coupled with human–computer interactions, and intelligent prediction along with cutting-edge applications. Prospectively, this transformative technology holds the potential to considerably enhance our understanding of the ocean, yielding groundbreaking discoveries that will profoundly influence the marine economy and sustainable development.


Introduction
In 2002, the concept of the digital twin (DT) was initially presented as the 'information mirroring model' by Michael Grieves, a professor at the University of Michigan, and has been described as digital siblings or digital mappings (Grieves 2005(Grieves , 2023)).The term 'digital twin' was formally adopted in 2010 when the National Aeronautics and Space Administration (NASA) of the United States provided a detailed conceptualization.This description specifically refers to the comprehensive utilization of physical models, sensors, historical operational data, and additional relevant information to simulate multidisciplinary and multiscale processes.The goal is to create a virtual representation that can mirror the entire lifecycle of its corresponding physical entity (Glaessgen and Stargel 2012).Essentially, the proposed DT model comprises three core components: the physical product, the virtual product, and the connection between these two products.
Although the DT concept had initially emerged in the fields of aerospace and manufacturing, it has since evolved, extending its application to intangible processes, such as smart cities (Qi et al. 2022), precision healthcare (Torkamani et al. 2017; Rozenberg and Greenbaum 2020; Barat et al. 2021), digital earth, and even the metaverse (Tao et al. 2019;Huang et al. 2020;Shen et al. 2020;Bauer et al. 2021;Wu et al. 2021).In 2019, DT was listed as one of the top ten key technological trends in the digital industry by Gartner (Gartner 2019).A striking example of DT applications is the 'Destination Earth' project, initiated by the European Union in March 2021.This endeavor aims to construct an extremely accurate DT of the earth for the continuous and precise monitoring of planetary health.Moreover, it aims to facilitate highaccuracy predictions of natural disasters and anthropogenic environmental degradation, as well as to examine the socioeconomic ramifications of climate changes and severe natural disasters (Bauer et al. 2021).
With the advancement of DT technology, its range of applications has gradually been extended to include marine environments, thereby promoting the advent of the digital twin of the ocean (DTO).To date, multiple applications of DT have been implemented across various scenarios (Qu et al. 2019).For instance, real-time DT systems for maritime vessel operations have been engineered through wave reconstruction and ship motion prediction (Lee et al. 2022).Moreover, high-fidelity simulation systems have been constructed for bird wingblending unmanned gliders by applying DT technology to hybrid wing gliders (Hu et al. 2023).
In the field of disaster management, a lightweight model for coastal flood prediction has been devised through neural operators to enable accurate forecasting of sea-surface heights in coastal regions (Jiang et al. 2021).The technology has been utilized in the engineering of autonomous submersibles designed to detect underwater gas leaks (Hu et al. 2022).Furthermore, by combining deep learning techniques, probabilistic realtime models detailing the consequences of natural gas jet fires have been developed.This has promoted the establishment of a DT system for emergency management concerning fire and explosion incidents on offshore platforms (Xie et al. 2023).In the energy sector, the application of DT technology to offshore wind farms has yielded machine learning-based methods to monitor and evaluate the impact of such installations.This provides valuable insights for future marine energy infrastructure projects (Schneider et al. 2023).A complex physical-biogeochemical model has been simulated by integrating the DT technology, aiming to forecast marine oxygen levels, aquacultural and fishery hypoxia, and the oxygen content in marine environments (Skakala et al. 2023).
In ecological research, DT technology has been employed to examine the correlation between vegetation coverage and the urban heat island phenomenon in coastal cities.This development has realized the technology requisite for marine ecosystems through social and ecological integration to support sustainable marine governance (Qi et al. 2022).Additionally, a cooperative underwater network of ocean observation systems has been established using the DT prototype methodology, facilitating the study of mobile ad-hoc networks for oceanic and coastal monitoring (Barbie et al. 2021).
We claim that the connotation of DTO represents the fourth developmental phase, evolving from the digital ocean (1.0), transparent ocean (2.0), and smart ocean (3.0).The digital ocean-an extension of the Digital Earth concept-primarily involves the application of contemporary information technology to marine domains.This stage focuses on the digitization of extensive multiresolution and multitemporal data to render the ocean in a digital format by employing technologies such as remote sensing, geographic information systems, and global positioning systems (3S).Advancing from the digital ocean, the concept of the Transparent ocean establishes a three-dimensional marine observational system and a real-time predictive framework.It aims to achieve comprehensive marine environmental awareness and the goal of transparency of state (clear observation), processes (deep understanding) and changes (accurate predictions).Smart Ocean refers to the comprehensive utilization of advanced technologies, such as ocean perception, real-time information transmission, big data, cloud computing, and knowledge mining.It aspires to construct an information infrastructure comprising the Ocean Integrated Perception Network, Ocean Information and Communication Network, and Ocean Big Data Cloud Platform.Smart Ocean targets the development of intelligent applications and services in marine information, representing a deep integration of industrialization and informatization in the marine sector while serving as an initial comprehensive solution to enhance ocean management capabilities.Therefore, DTO incorporates the static real-time displaying capability of the digital ocean (stage 1.0: expression), dynamic process understanding and prediction capability of the transparent ocean (stage 2.0: mechanism), and comprehensive supporting capability of the smart ocean across the entire data perception, information communication, intelligent application, and decision supporting chain (stage 3.0: service).Moreover, DTO emphasizes the application of big data and AI to empower the ocean to develop excellent knowledge discovery ability of AI oceanography, enabling accurate predictions and virtual deduction of the ocean's past and present, which represent a more advanced stage of marine development (stage 4.0: integration and upgrading).
However, the development of DTO is in its initial stage, and developing key technologies and prototype systems holds significant potential in ocean cross-layer coupling based on the combination of artificial intelligence (AI) oceanography and multidisciplinary frontier applications, which represents the current direction and focus of development.Thus, the overall architecture, major components, key technologies, and exploratory applications are proposed in this study.

Overall architecture of DTO
The new generation of information technologies-represented by big data, cloud computing, 5G communication, virtual reality, supercomputing, and AI-constitute the most innovative, pervasive, and influential field in the contemporary world.It is inciting an upheaval in the global technological landscape and transforming into actual productivity at an unprecedented rate, thereby expediting profound shifts in science, technology, economics, and society.The application of these nascent technologies to fields such as ocean exploration, conservation, and governance, and the establishment of a DT system via 'thorough information perception, ubiquitous communication, extensive data sharing, and intelligent application services' is imperative for achieving a collectively envisioned future of the oceans and sustainable development objectives.To facilitate full-cycle integration and serve as an invaluable research tool, the construction of the DTO system should contain five major components: the data pooling system, dome system, virtual reality (VR) interaction system, intelligent control system, and intelligent prediction system.The desired effect of each part is illustrated in Fig. 1.
In contrast to conventional virtual or augmented reality, the distinguishing aspect of DTO lies in its superadded intelligence, primarily derived from AI-based replication.During the construction of the DTO system, we suggest a structured architecture tailored to address marine scientific challenges, as outlined in Fig. 2.

Data pool
The data pool serves as a pivotal element of the DTO system, functioning as the catalyst for AI applications in marine contexts and the foundation for realizing transparent and integrated twins (Wu et al. 2021).It also serves as a centralized storage repository that consolidates raw data from diverse sources and types into a standardized format.This unified storage enables efficient management and organization of data, regardless of its original form or source.Through the establishment of a centralized data storage repository, the data pool enables effortless data access, retrieval, and processing, Fig. 1 Digital twin of the ocean thereby promoting effective data management and analysis within the system.The data repository possesses the requisite computational prowess and ample storage capacity to facilitate rapid cross-analysis of multisource data (Lv et al. 2017).The establishment of the data pool involves three primary components: data collection, data management, and data services.

Data collection
The data pool for the digital twin of the ocean (DTO) system assimilates data from an array of sources, including global ocean remote sensing data sets, in-situ data sets, model data sets, and reanalysis products (Atkinson et al. 2014;Riser et al. 2016;Zheng et al. 2020;Gettelman et al. 2022).The accrued data are classified into three categories based on temporal immediacy: historical data, quasireal-time data, and real-time data.Historical data include parameters such as global climate state temperature, salt field data, reanalysis and remote sensing data.Quasi-realtime data, with a maximum latency of seven days, include global watercolor, hyperspectral, and microwave remote sensing data, along with high-definition regional model data released in a near-real-time fashion.Real-time data, limited to a maximum one-day delay, consist of observations, models, and predictions from the twin ocean system, which are directly accessed in real time.Real-time data acquisition refers to the ability of the data pool to promptly retrieve the latest data from each data source, ensuring the availability of data updates.The data acquisition process primarily relies on automated systems.

Data management
Currently, ocean big data encounters challenges related to multisource heterogeneity (i.e., multiplatform, multiparameter, multistructure, multiresolution, etc.) and the coexistence of large and small data (In terms of data volume, remote sensing and model are big data while in-situ observation data are small data.In terms of vertical distribution, sea-surface data is big data, and deepsea data is small data).To mitigate the complexities of multisource heterogeneous marine data, a unified spatial coordinate system and a spatiotemporal database have been implemented.This standardization ensures query consistency and enables seamless data source integration.Additionally, a four-dimensional marine data engine employing core technologies, such as ocean Fig. 2 Architecture of the digital twin of the ocean vector retrieval and elemental correlation calculation, has been developed.This engine facilitates distributed and extended storage of marine data, accommodating large and small datasets.Through these technological interventions, challenges associated with the storage of highgranularity and high-density ocean data are effectively addressed to achieve efficient and reliable data storage.

Data services
Multisource twin ocean data pools are endowed with real-time streaming capabilities, thus enabling instantaneous data stream processing.One strategy to augment data transmission efficiency involves partitioning the data stream into smaller segments, thereby minimizing their size (Peterson et al. 2001).Such segmentation enhances data transmission and reception efficacy.Concurrently, the establishment of unified standards for data formats and spatial references in marine data expedites the integration of data from multiple origins.By complying with these norms, the data pool can proficiently process and amalgamate data from varied sources.Furthermore, capitalizing on the data pool to construct a cloud service platform and employing in-situ cloud computing online allows the migration of processing algorithms and tools to the data platform.This computational shift in marine data management facilitates swift responses to user demands for large-scale ocean data (Gu and Grossman 2007), ensuring a seamless flow of real-time data to effectively satisfy user requirements.

Key technologies
The data pool ensures the availability of valid data essential for constructing the DTO system.To vividly present this data while facilitating user interaction, several key technologies have been incorporated, including marine Internet of Things (IoT), distributed computing, AI, and interactive technology.

Marine internet of things
A salient feature of the DTO system is its real-time connectivity between virtual and actual entities, which requires not only a sophisticated data flow mechanism within the data pool but also real-time data acquisition to guarantee minimal latency and high fidelity.In this context, marine IoT, which is one of the most robust and efficient communication systems, is essential for fluid data exchange between the physical and virtual environments.Marine IoT incorporates a network of interconnected sensors, devices, and systems that are deployed across marine landscapes (Xu et al. 2019).These IoT-enabled devices gather real-time data from multiple sources, such as ships, buoys, sensors, satellites, and underwater platforms, encompassing various parameters, such as water temperature, salinity, currents, weather conditions, pollutant levels, and marine fauna behaviors.Such comprehensive data serve as the foundation for the DTO, facilitating real-time monitoring, analysis, and prediction.

Distributed computing
Within the data pool, substantial volumes of data are amassed, processed, managed, analyzed, and disseminated.These data are integral for tasks like threedimensional modeling, model training, and DTO system construction.To manage the computational requirements of these processes, the system employs distributed computing.This approach of distributed computing involves partitioning extensive computational tasks into smaller sub-tasks and allocating them across multiple computational nodes.This strategy of parallel execution remarkably enhances computational speed and system scalability, thereby allowing for efficient processing of large-scale data and intricate computational challenges.This approach proves particularly beneficial for DTOrelated tasks where massive data volumes are involved.
Moreover, distributed computing facilitates large-scale AI training by streamlining the processing of extensive datasets and expediting the training regimen.It also accelerates the visualization rendering process, thereby hastening the generation of visual data representations.Advancements in distributed computing have been instrumental in enabling the DTO system to furnish near real-time oceanic conditions.By capitalizing on the computational prowess and scalability offered by distributed computing, the DTO system can expeditiously process and analyze data, thereby delivering timely and accurate oceanic information.

Artificial intelligence
AI has emerged as a powerful and contemporary approach for constructing and training DTO models.In particular, deep learning possesses the ability to simulate and predict the behavior and performance of real systems.Deep learning models, characterized by their multiple processing layers, excel in learning hierarchical data representations, thereby enabling the abstraction of complex features.The notable advancements in deep learning, such as computer vision, neural language processing, and scientific applications, have considerably benefited the DTO system.Furthermore, incorporating existing knowledge and theories from oceanography into AI models has led to the genesis of specialized oceanographic AI models that combine mechanistic constraints with data-driven techniques, thereby enhancing data interpretability and facilitating ocean data exploration and knowledge discovery within DTOs.Furthermore, the employment of lightweight models for multimodal prediction can improve computational efficiency without compromising accuracy, thereby enabling efficient processing and analysis of DTO data.
The rapid advancement of GPU computing clusters has led to the considerable expansion of the parameterization of neural networks.This technological advancement, in conjunction with transformer architecture and pretraining methodologies, has realized the development of ChatGPT-a significant milestone in AI research.Chat-GPT has substantially contributed to the evolution of the DTO system by enhancing its capabilities for humanmachine interactions.Utilizing its proficiency in language understanding and generation, ChatGPT facilitates effective user engagement, thereby enabling seamless communication and interaction.Additionally, it helps users to explore and gain insights into the various phenomena and dynamics of the ocean.

Interactive technology
Visualization technologies such as VR, augmented reality (AR), and mixed reality (MR) are integral to the deployment of DTO application systems across diverse scenarios.VR technology, an avant-garde digital innovation, offers invaluable perspectives for comprehending real-time oceanic changes and studying oceanic patterns within the DTO context (Anthes et al. 2016).The utilization of VR enables the transformation of the digital twin into a visual twin, thereby supporting real-time mapping and fostering a deeper connection between the physical ocean and its digital analog (Rokhsaritalemi et al. 2020).Conversely, AR technology synergizes the real world with computer-generated elements to augment user perception and interaction.MR, amalgamating features of both VR and AR, allows users to manipulate virtual objects while retaining a connection with the real world.In MR, virtual and physical entities are seamlessly integrated, thus creating a symbiotic experience in real-time.Leveraging these visualization technologies, the DTO system finds applicability in an array of contexts, including marine experiments, deep-submarine navigation, laboratory research, and scientific education.

DTO applications
At the intersection of distributed computing, AI, VR, and other similar technologies, the development of a DT by integrating data, algorithms, visualization, and interaction is imminent.This technology must cohesively consolidate data, algorithms, visualization, and interactive elements.The key considerations include the design of modular components, packaging of diverse elements, system integration, and the subsequent deployment of an advanced prototype of DTO.Furthermore, a preeminent requirement exists for devising a user-friendly interface that improves the efficiency of human-computer interaction.
The realization of a fully integrated twin ocean system necessitates the establishment of adept data-scheduling frameworks, the construction of distributed indexing structures suitable for spatiotemporal data, and the implementation of specialized scheduling algorithms for the ocean's four-dimensional spatial data engine.Additionally, incorporating algorithms that address ocean dynamics and ecological parameter inversion as well as recognize typical ocean phenomena within the AI models is important.These AI models should facilitate the coordinated, interactive visual analysis of various oceanic elements by constructing a highly perceptive twin ocean prototype system.The system should provide visualizations and interactive capabilities in two and three dimensions.Moreover, this system should adhere to workflows progressing from data fusion to current situation depiction, phenomenon mining, autonomous learning, and, ultimately, intelligent prediction.
A significant challenge lies in the development of panoramic, immersive, and dynamically interactive visual twin technology that utilizes data from multiple sources.The primary objective is to forge seamless connections between numerical simulations, efficient information extraction, and dynamic oceanic display.Achieving this objective requires surmounting challenges related to immersive experiences, low latency, high fidelity, and the spatiotemporal continuous visualization of 3D dynamic ocean data.
To augment the DTO system further, it is recommended to construct a panoramic virtual ocean environment that harnesses data from a multitude of oceanic sources.Such an environment would offer users a comprehensive, immersive visualization of the ocean, enabling them to navigate and explore various oceanic regions interactively.This interactive modality allows for real-time data analysis and interpretation, deepening the user's understanding of oceanic complexities.Additionally, the transition from digital to visual twins within the twin ocean system becomes feasible through this panoramic virtual ocean environment, thereby transcending physical constraints and enriching the tools available for oceanic study and comprehension.

From digitalization to cloning
The DTO accelerates digitization and enables cloning beyond intelligence.This enhances the predictability of outcomes and facilitates the identification of the underlying mechanisms that help to improve the understanding of marine science.

AI-oceanographic model for intelligence
The AI-based oceanographic model constitutes a computational algorithm that leverages AI techniques to simulate, analyze, and predict various oceanic aspects and processes.This model is applied to a wide range of oceanographic research, encompassing areas such as ocean circulation and currents, climate modeling and forecasting, marine ecosystem analysis (Benway et al. 2019), and marine environmental monitoring.Furthermore, AI technology excels in discerning correlations among disparate data elements and is notably proficient in inverting marine elements, including variables like tropical instability waves, internal waves, sea-surface temperature, wave patterns, and the ENSO index.
Contemporary marine science relies on cognitive experimentation based on observational data, which is pivotal in the design of AI models geared for scientific purposes.The ensuing outcome should epitomize a collaborative synergy between data-driven and strongly mechanistic models.Specifically, the design of the AI model, in congruence with the existing oceanic knowledge, should adhere to marine laws or symmetries to implement the intelligence of DTO effectively.Underpinning this design strategy is a conceptual continuum that transitions from a statistical, data-driven model to a robustly mechanistic model, as shown in Fig. 3.The hybrid DTO systems should be constrained by physical feasibility while concurrently exploiting data-derived insights.
The AI-based oceanographic model unveils the intrinsic nature of phenomena embedded within the data and demystifies the governing laws at both theoretical and applied strata.The integration of oceanic physical mechanisms enhances the interpretability of these datadriven AI methods and aids in the extraction of valuable knowledge embedded within the data set.Consequently, the digital twin's strengths in data mining and knowledge discovery can be leveraged to unveil intrinsic correlations and spatiotemporal laws among various elements in the oceanic system to the greatest extent possible.Such revelation paves the way for the incorporation of advanced AI technologies, including deep learning, into the DTO system, thereby maximizing its potential.

Visualization technology for cloning
Visualization serves as a pivotal conduit between the real ocean and its digital clone, facilitating the transition from a digital ocean to a visual twin.The development of a DTO system requires the support of transparent twin technology based on multisource data.This technology endows the system with panoramic, immersive, dynamically interactive, and highly perceptive capabilities.The implementation of this technology is reliant on three key technologies.First, the establishment of a panoramic virtual ocean environment contributes to low latency and increased perceptive faculties across various oceanic applications.This involves creating a holistic virtual representation of the ocean, thereby permitting users to explore and interact with the digital twin in a realistic and immersive context.Second, the real-time extraction of characteristic features and discovery of knowledge regarding typical oceanic phenomena is essential for offshore activities.This includes the development of standardized transfer function patterns for typical oceanic phenomena, such as eddies and fronts.These patterns facilitate interactive exploration, feature extraction, and a highly perceptive visual analysis of oceanic phenomena, thereby unveiling the underpinning physical mechanisms.Lastly, immersive interaction between the actual and virtual oceanic environments and phenomena is accomplished.Through the employment of VR headset hardware, a comprehensive spatiotemporal three-dimensional DTO is constructed.The procedural sequence for DTO visualization can be delineated in Fig. 4.

Prototype system integration for extensive applications
The development of a DTO prototype system encompasses several key dimensions: (1) the amalgamation of heterogeneous, multisource marine big data culled from data repositories; (2) the incorporation of AI models adept in ocean dynamics, ecological parameter inversion algorithms, and the identification of salient marine phenomena; (3) the inclusion of high-fidelity visualization coupled with immersive interactive capabilities to augment oceanic perception.As delineated in Fig. 5, the acronyms RSD, OPD, ORD, and OSD signify remote sensing data, ocean pattern data, ocean reanalysis data, and on-site data, respectively.This integration aims to achieve a hardware-software integration of a novel twin marine prototype system.On the software side, the development of a distributed indexing structure that can efficiently handle spatiotemporal data is important.This indexing structure enables quick and optimized data retrieval based on spatiotemporal attributes.Additionally, the implementation of a scheduling algorithm is necessary to manage the competent processing and scheduling of four-dimensional ocean data, thereby maximizing computational resource utilization.The inversion algorithm assumes a critical role in precisely estimating oceanic dynamics or ecological parameters based on observational data.Furthermore, an AI model needs to be developed to analyze and recognize typical oceanic phenomena, providing valuable insights into their intrinsic properties and behaviors.
On the hardware side, detailed module design and component packaging are vital for optimizing performance and achieving seamless integration within the twin ocean system.Procedures for system integration and deployment must be meticulously executed to guarantee smooth interoperation and compatibility among different hardware components.Additionally, the construction of a user-friendly human-machine interface is required to enhance user interaction and the efficient employment of system functionalities.
Through the synergistic integration of hardware and software components, the DTO prototype system is poised to offer a holistic digital twin functionality that encompasses the entire chain of 'data fusion → current situation presentation → phenomenon mining → autonomous learning → intelligent prediction.' This system is designed for application in regional and global industrial contexts, as well as in avant-garde scientific research, thereby showcasing its multifaceted utility.

Exploratory application examples of the DTO
The present reconstruction and 3D identification of mesoscale eddies are utilized as proof-of-concept applications to demonstrate the feasibility of the DTO framework.

Ocean-current reconstruction
Oceanic surface circulation, characterized by its multiscale characteristics, plays an important role in regulating climate, responding to atmospheric forcing, and maintaining the balance among marine material, heat, and energy.Because of the limitations imposed by sampling rates and spatial distribution on the observed record, as well as the coarse and weather-dependent accuracy of remote sensing estimates, a methodology for the efficient reconstruction of ocean currents based on multiscale physical processes is proposed.Combined with geostrophic and Ekman currents, TPXO9 tidal currents, and wave-induced Stokes drift, we analyze the coupling relations between ocean surface components associated with disparate physical processes and actual ocean currents observed via drifting buoys.Subsequently, we establish comprehensive and precise oceancurrent datasets with extensive spatiotemporal coverage.This methodology is implemented in accordance with the DTO framework, as depicted in Fig. 6.
To effectively visualize the velocity, direction, and depth of the flow field in the Distributed Temperature Observing System (DTOS), particle-based trace elements are employed as a visualization technique within a continuous spatiotemporal framework.These particles serve to represent the flow field and afford the DTOS user a lucid understanding of its various aspects.This methodological approach enables users to distinctly perceive variations in the flow field, thereby facilitating comprehension of its dynamic characteristics.

3D identification of oceanic eddy
Mesoscale eddies are pervasive across the global ocean, accounting for nearly half of its area.Despite this prevalence, it is posited that approximately 90% of these eddies, particularly the smaller ones, remain undetected due to inadequate sampling.To address this challenge, we propose a DTO-based methodology aimed at enhancing altimetry identification capabilities, as illustrated in Fig. 7.This approach involves developing an independent identification scheme that correlates surface features, such as radius amplitude, with their internal properties.This is based on the strong correlation observed between

Conclusions and discussions
The ocean is intricately connected to mankind as it influences climate regulation, food supply, and sustainable development.Therefore, it is crucial to explore how we can better take care of, understand, and strategically manage the ocean.The emergence of nextgeneration information technology, exemplified by big data and AI, offers unprecedented means to fulfill these objectives.The DTO framework serves as a vital bridge between current digitalization and future intelligence.A comprehensive DTO system, ranging from 'data fusion to situation presentation, phenomenon mining, autonomous learning, and intelligent prediction, ' typically encompasses a data pool that integrates real-time and historical data from multiple sources.Furthermore, an AI-based oceanographic model with generative, physically interpretable and lightweight features, as well as a visualization module with high perception, immersion, and interactivity, needs to be supplemented by requisite hardware and software components.Utilization of the DTO system promises efficacious applications across marine science, ocean engineering, and tourism and management, thereby benefiting society at large.
Projecting forward into the upcoming 5-10 years, the research priorities and objectives for DTO are centered on three key domains: (1) Four-dimensional digital cloning and real-time mapping of global ocean parameters.
The burgeoning volume of marine data, propelled by advancements in ocean remote sensing, in-situ observations, and modeling technologies, will propel the current data pool toward a more advanced data lake.Additionally, the real-time capability of multisource data will be further improved by the continuous advancements in satellite communication technology.Moreover, the DTO system will integrate autonomously learned new data and insights, yielding a spatiotemporal data space that merges virtual and real dimensions.The imminent DTO framework is anticipated to serve as an intricate four-dimensional digital clone of the global ocean.Fundamental to this advancement will be a user-friendly, secure cloud platform that facilitates multiparameter data processing of global oceanic metrics while also providing real-time (2) Synergistic perception of multiple elements and natural human-computer interaction within the DTO framework.
The development of highly-perceived spatiotemporal continuous marine visualization technology is expected to progress concomitantly with advancements in VR and MR technologies for improved natural human-computer interactions.Concurrently, the DTO system will improve in the areas of scene perception, data acquisition, and information mining.To accommodate the distribution characteristics of the data elements associated with marine phenomena, a standard transfer function will be established to facilitate immersive and perceptually rich three-dimensional visualizations of scalar, vector, and tensor fields.Furthermore, innovations in VR/MR technologies are projected to enhance multiple perceptual sources, immersion levels, and natural human-computer interactions.These developments will enable users to observe and analyze marine phenomena at various scales from a first-person perspective.This facilitates collaborative feature extraction for multiple hydrological elements and analysis of typical marine phenomena through natural human-computer interactions, which will transition the DTO system from a DT to a visual twin stage.
(3) Intelligent prediction and cutting-edge application of DTO.
In the future, the DTO system will assume a central role in cutting-edge scientific research, ocean engineering, and ocean management.On the one hand, the robust information extraction capabilities of DTO will be utilized to explore the comprehensive verification, local adjustment, and regional correction of ocean dynamics theory with the objective of addressing the current challenges related to the local applicability and regional invalidity of ocean dynamic theories in the field of physical oceanography.To resolve interdisciplinary constraints, an integrated modeling approach will be adopted for the collaborative prediction of ocean processes involving biotic and abiotic tracers, thereby enabling marine knowledge discovery across different ocean layers and opening a new path for ocean dynamics research and interdisciplinary exploration.On the other hand, the next-generation DTO system will provide multidimensional, integrated, intelligent perception capabilities.This will supply relevant entities with reliable, timely, and comprehensive marine data resources and predictive services for marine environment, process changes, and extreme natural events, such as typhoons and algal blooms.Consequently, the DTO system will maximize its support capacity for global ocean governance, enhancing decision-making across multiple sectors, including marine resource development, comprehensive management, and disaster prevention.
In summary, the DTO framework offers a comprehensive and holistic approach to understanding and managing the marine environment.It empowers datadriven governance, advances sustainable development objectives, and contributes substantially to the ongoing and future well-being and resilience of oceanic systems for the benefit of present and future generations.

Fig. 3
Fig. 3 Design strategy for the hybrid AI-oceanographic model

Fig. 6
Fig. 6 Ocean-current reconstruction under the DTO framework

Fig. 7
Fig. 7 3D identification of eddy under the DTO framework