Skip to main content

OPINION article

Front. Mar. Sci., 18 May 2022
Sec. Marine Conservation and Sustainability
Volume 9 - 2022 |

The Role of Artificial Intelligence Algorithms in Marine Scientific Research

Min Jiang1,2 Zhiyuan Zhu1*
  • 1Chongqing Key Laboratory of Nonlinear Circuits and Intelligent Information Processing, College of Electronic and Information Engineering, Southwest University, Chongqing, China
  • 2Guangdong Key Laboratory of Intelligent Information Processing and Shenzhen Key Laboratory of Media Security, Shenzhen, China


The study of marine science is vital to the survival and development of humanity. On the one hand, the oceans act as a global climate regulator, supplying 70% of oxygen and 87.5% of water vapor to the atmosphere while storing large amounts of heat (Petterson et al., 2021). On the other hand, the oceans act as an important part of the global physical system, in which changes in mass-energy, biological, and geological processes can have a significant impact on marine and terrestrial life (Du et al., 2021). However, lacking knowledge of important areas such as the deep sea and polar regions, humans cannot yet decipher certain specific phenomena and patterns in the oceans.

Artificial Intelligence (AI) algorithms are trained on mathematical models with a specific structure using a large amount of statistical data to obtain a fitter that contains the statistical features inherent in the training data. It can be applied to solve optimization problems. As a result, AI algorithms have been very successful in a number of scientific fields, such as autonomous driving (Khan et al., 2021), medical imaging (Hickman et al., 2022), geophysics (Yu and Ma, 2021), and nanoscience (Jiang et al., 2022).

As marine scientific research enters a new era of intelligence and constantly improving marine data, AI can effectively tap into the potential information contained in vast amounts of data. As a result, it is also gaining more and more attention from marine researchers (Logares et al., 2021). The integration of AI technology with traditional models to improve marine safety has also been proven (Khayyam et al., 2020). In addition, data processing problems in marine pollution (Agarwala, 2021), wind and wave energy (Gu and Li, 2022) research can be solved using AI algorithms.

Therefore, this paper reports on the prospects of applying AI algorithmic methods in marine scientific research, mainly monitoring marine biodiversity, deep-sea resource modelling, and predicting SST, tide level, sea ice, and climate. In addition, the paper discusses the current problems of AI algorithms in processing marine data and building predictive models.

Research and Application of AI in Marine Scientific Research

Monitoring Marine Biodiversity

Due to the complexity and specificity of marine ecology, the use of only traditional manual solutions poses great challenges for monitoring the ecological dynamics of species. Therefore, researchers have been working to build monitoring solutions for automatic identification, classification, and prediction using AI algorithms. Allen et al. (Allen et al., 2021) trained a deep convolutional neural network to identify humpback whale songs in over 187,000 hours of acoustic data collected over a 14-year period at 13 different monitoring sites in the North Pacific (Figure 1A). The model successfully detected 75s audio segments containing humpback whale songs with an average accuracy of 0.97 and an average area under the receiver operating characteristic curve of 0.992. These results validate the feasibility of applying deep learning models to identify highly variable signals over a wide range of spatial and temporal scales. In order to solve the problem of classifying species in the images detected in the camera system, Allken et al. (Allken et al., 2019) introduced a deep learning network and developed a new training method. In experiments to classify blue whiting, Atlantic herring, and Atlantic mackerel, the results showed a classification accuracy of 94%. As shown in Figure 1B, Vanesa et al. combined machine learning and deep learning networks to integrate video/image annotation tools for a pipeline of visual data analysis (Lopez-Vazquez et al., 2022). Their experimental results showed that the scheme was able to obtain comprehensive information on the spatial distribution and temporal dynamics of the ocean in time, with an accuracy of 76.18% of their results.


Figure 1 (A) Map of the 13 acoustic recording locations. (Allen et al., 2021). (B) Image processing pipeline (Lopez-Vazquez et al., 2022). (C) Architecture of the All-season CNN (A_CNN) model for ENSO forecasts (Ham et al., 2021).

AI algorithms can provide a large amount of multi-parameter information in monitoring complex marine ecosystems. It provides an effective solution for building a robust and widely distributed marine ecological monitoring network.

Modeling of Deep-Sea Resources

The modeling of seabed resources has always been an important topic in marine scientific research. The combination of artificial intelligence algorithms to build a novel 3D modeling approach has an important role in the re-transportation of deep-water sediments and the reconstruction of the bottom shape of the seabed. Neettiyath et al. (Neettiyath et al., 2019) described a method to estimate the volume distribution of cobalt-rich manganese crusts (Mn-crusts) by combining multi-modal sensor data collected using an autonomous underwater vehicle (AUV). The AUV uses a sub-bottom sonar to calculate the thickness of Mn-crusts and a light-profile mapping system to generate 3D color reconstructions of the seafloor. A machine learning classifier was used to classify the 3D map into one of 3 types of seafloor-crust, sediment, and nodules. Percentage coverage of Mn-crusts and estimates of the mass of Mn-crusts per unit area were determined along AUV transects based on inferred thicknesses. This approach provides a novel method to estimate the distribution of Mn-crusts over large areas. De et al. (De La Houssaye et al., 2019) combined modern programming techniques in computer vision, machine learning, and deep learning applications with traditional geoscientific linear regression architectures. Then the data were trained end-to-end to predict the O18/O16 isotope ratio as an independent global proxy for the geological age of marine sediments over time. Ratto et al. proposed an alternative approach to ocean ray tracing through a combination of machine learning and generative adversarial networks (GAN) (Wang et al., 2017). Their experimental results demonstrate that GANs trained on thousands of small scenes generated by ray-tracing models can be used to generate megapixel scenes faster with a consistent spectrum and minimal processing artifacts (Ratto et al., 2019).

The use of AI algorithms to overlay and display different ocean state data on top of a 3D platform enables the management, integration, and analysis of ocean basic, target, and derived data. It provides a common and standardized tool for activities such as studying the structure and function of ocean systems and revealing the various patterns of ocean phenomena.

Prediction of Ocean Parameters

Sea surface temperature (SST), tide level, and sea ice are crucial parameters of the global ocean that can profoundly affect climate and marine ecosystems. To achieve accurate and comprehensive predictions for the short and medium-term domains of SST, Xiao (Xiao et al., 2019) proposed a spatiotemporal deep learning model that captures the correlation of SST across time and space. The model uses convolutional long and short-term memory as the building block and is trained in an end-to-end manner. Experiments using 36 years of satellite SST data in the East China Sea sub-region show that the proposed model outperforms persistence models when judged using multiple methods. It means that the proposed model is highly promising for accurate, convenient short, and medium-term daily SST field predictions. Riazi (Riazi, 2020) proposed a simple and effective deep neural network that can estimate future tidal levels based on the forces and factors affecting the tidal range. The main advantages of the proposed method are that the input layer can be kept as small as possible and the input data can be accessed. In addition, one network can be used for different beaches as the input data is based on the primary cause of tides. Gregory (Gregory et al., 2020) predicted the September mean sea ice extent in the Arctic region. They used a complex network statistics method that was developed by constructing spatiotemporal homogeneity regions and subsequently deriving remote connections between them. Ultimately, high prediction results are obtained in the Gardiner Islands, Eastern Siberia, and the Kara Sea by exploiting relationships in climatic event series data. Ham et al. developed an all-season convolutional neural network (A_CNN) model. The model predicts the El Niño/Southern Oscillation (ENSO) phenomenon by extracting various complex network metrics (Ham et al., 2021). The results reveal a combination of various climate precursors of ENSO events that act differently over time, thus demonstrating the potential of the A_CNN model as a diagnostic tool (Figure 1C).

The above case provides an idea to collect environmental information such as marine meteorology, sea temperature, and sea ice and combine them with predictive analysis by AI algorithms (e.g., support vector machines, long and short-term memory, and convolutional neural networks). And then, the correlation and basic laws of climate and environmental parameters are obtained to predict future trends. This provides support for information services for various marine activities such as marine resource utilization, fisheries, and shipping.


In recent years, AI algorithms have been widely used in marine science. A lot of research has been done on AI algorithms in processing marine biodiversity and deep-sea resource data information. Moreover, the research on predicting climate change based on ocean information parameters is also deepening. However, there are still many challenges in using AI algorithms for marine science due to the multi-source nature of current marine data and the complexity of algorithmic models.

Many marine data have been accumulated with the robust operation of various sensors such as satellite remote sensing, buoys, and shipborne instruments. That provides a large amount of sample space for AI algorithms. However, it is difficult to form uniform selection criteria in terms of the quality of sample data and sample richness, which affects the quality of model construction.

Therefore, multiple methods can be used to build data with different dimensions and characteristics during the training data construction process. In addition, as the processing methods mainly depend on the needs of different applications and the cognitive background of the users, the pre-processing data methods have an important influence on the construction of the models. The future development of AI algorithms should be geared toward multiple data formats with a view to reducing the significant amount of time spent on data preparation and thus improving the overall model building efficiency.

There are numerous artificial intelligence algorithmic models, and there is no consensus on which model is more suitable for solving a specific marine data mining problem. Detailed comparisons and conclusions have not yet been drawn, making practical applications difficult. Multiple training strategies are often used, including adjustments to parameters and training methods, and detailed comparisons are made during model construction through a large number of experimental comparisons to build a more accurate model.

AI algorithms in marine science not only improve the efficiency of marine data processing and break the traditional technical bottleneck. Moreover, it is of great significance for discovering marine laws, cognitive marine environment, revealing its interaction mechanism, protecting and using marine resources, marine disaster prevention, and mitigation.

However, as marine data enters the era of information-rich big data, it has the characteristics of diversity and complexity. The current big data analysis methods mainly rely on conventional standard data types and lack the integrated analysis system of scientific data. Therefore, “smart sensors” with high-efficiency data transmission should be continuously developed for automatic sampling. Secondly, an existing marine data management infrastructure should be used to enhance data processing and management to make data machine-readable. Finally, with the in-depth integration of marine science and big data technology, the application and intelligent management of marine information will gradually improve.

Author Contributions

MJ: original draft preparation and editing. ZZ: writing— review and project administration. All authors contributed to the article and approved the submitted version.

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s Note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.


Agarwala N. (2021). Managing Marine Environmental Pollution Using Artificial Intelligence. Maritime Technol. Res. 3 (2), 120–136. doi: 10.33175/mtr.2021.248053

CrossRef Full Text | Google Scholar

Allen A. N., Harvey M., Harrell L., Jansen A., Merkens K. P., Wall C. C., et al. (2021). A Convolutional Neural Network for Automated Detection of Humpback Whale Song in a Diverse, Long-Term Passive Acoustic Dataset. Front. Marine Sci. 8, 165. doi: 10.3389/fmars.2021.607321

CrossRef Full Text | Google Scholar

Allken V., Handegard N. O., Rosen S., Schreyeck T., Mahiout T., Malde K., et al. (2019). Fish Species Identification Using a Convolutional Neural Network Trained on Synthetic Data. ICES J. Marine Sci. 76 (1), 342–349. doi: 10.1093/icesjms/fsy147

CrossRef Full Text | Google Scholar

De La Houssaye B., Flaming P., Nixon Q., Acton G.. (2019). Machine Learning and Deep Learning Applications for International Ocean Discovery Program Geoscience Research. SMU Data Sci. Rev. 2 (3), 9.

Google Scholar

Du Y., Dong X., Jiang X., Zhang Y., Zhu D., Sun Q., et al. (2021). Ocean Surface Current Multiscale Observation Mission (OSCOM): Simultaneous Measurement of Ocean Surface Current, Vector Wind, and Temperature. Prog. Oceanography 193, 102531. doi: 10.1016/j.pocean.2021.102531

CrossRef Full Text | Google Scholar

Gregory W., Tsamados M., Stroeve J., Sollich P. (2020). Regional September Sea Ice Forecasting With Complex Networks and Gaussian Processes. Weather Forecasting 35 (3), 793–806. doi: 10.1175/WAF-D-19-0107.1

CrossRef Full Text | Google Scholar

Gu C., Li H. (2022). Review on Deep Learning Research and Applications in Wind and Wave Energy. Energies 15 (4), 1510. doi: 10.3390/en15041510

CrossRef Full Text | Google Scholar

Ham Y. G., Kim J. H., Kim E. S., On K. W., et al. (2021). Unified Deep Learning Model for El Niño/Southern Oscillation Forecasts by Incorporating Seasonality in Climate Data. Sci. Bull. 66 (13), 1358–1366. doi: 10.1016/j.scib.2021.03.009

CrossRef Full Text | Google Scholar

Hickman S. E., Woitek R., Le E. P. V., Im Y. R., Mouritsen Luxhøj C., Aviles-Rivero A. I., et al. (2022). Machine Learning for Workflow Applications in Screening Mammography: Systematic Review and Meta-Analysis. Radiology 302 (1), 88–104. doi: 10.1148/radiol.2021210391

PubMed Abstract | CrossRef Full Text | Google Scholar

Jiang M., Li B., Jia W., Zhu Z.. (2022). Predicting Output Performance of Triboelectric Nanogenerators Using Deep Learning Model. Nano Energy 93, 106830. doi: 10.1016/j.nanoen.2021.106830

CrossRef Full Text | Google Scholar

Khan Q. T. A., Abbas S., Khan M. A., Areej F., Saad A., Nouh S. E., et al. (2021). Modelling Intelligent Driving Behaviour Using Machine Learning. Comput. Mater. Con. 68 (3), 3061–3077 doi: 10.32604/cmc.2021.015441

CrossRef Full Text | Google Scholar

Khayyam H., Javadi B., Jalili M., Jazar R. N. (2020). “Artificial Intelligence and Internet of Things for Autonomous Vehicles”, In: Jazar R., Dai L.. Nonlinear Approaches in Engineering Applications (Cham: Springer), 39–68. doi: 10.1007/978-3-030-18963-1_2

CrossRef Full Text | Google Scholar

Logares R., Alos J., Catalán I., Solana A. C., del Campo F. J., Ercilla G., et al. (2021). Oceans of Big Data and Artificial Intelligence.

Google Scholar

Lopez-Vazquez V., Lopez-Guede J. M., Marini S., Fanelli E., Johnsen E., Aguzzi J. (2020). Video Image Enhancement and Machine Learning Pipeline for Underwater Animal Detection and Classification at Cabled Observatories. Sensors 20 (3), 726. doi: 10.3390/s20030726

CrossRef Full Text | Google Scholar

Neettiyath U., Thornton B., Sangekar M., Nishida Y., Ishii K., Sato T., et al. (2019). An AUV Based Method for Estimating Hectare-Scale Distributions of Deep Sea Cobalt-Rich Manganese Crust Deposits (OCEANS 2019-Marseille: IEEE), 1–6.

Google Scholar

Petterson M. G., Kim H. J., Gill J. C. (2021). “Conserve and Sustainably Use the Oceans, Seas, and Marine Resources,” in Geosciences and the Sustainable Development Goals (Cham: Springer), 339–367.

Google Scholar

Ratto C., Szeto M., Slocum D., Benne K. D. (2019). “OceanGAN: A Deep Learning Alternative to Physics-Based Ocean Rendering,” in ACM SIGGRAPH , 2019 Posters. New York: Association for Computing Machinery. 1–2.

Google Scholar

Riazi A. (2020). Accurate Tide Level Estimation: A Deep Learning Approach. Ocean Eng. 198, 107013. doi: 10.1016/j.oceaneng.2020.107013

CrossRef Full Text | Google Scholar

Wang K., Gou C., Duan Y., Wang F-Y.. (2017). Generative Adversarial Networks: Introduction and Outlook. IEEE/CAA J. Automatica Sin. 4 (4), 588–598. doi: 10.1109/JAS.2017.7510583

CrossRef Full Text | Google Scholar

Xiao C., Chen N., Hu C., Wang K., Xu Z., Cai Y., et al. (2019). A Spatiotemporal Deep Learning Model for Sea Surface Temperature Field Prediction Using Time-Series Satellite Data. Environ. Modelling Software 120, 104502. doi: 10.1016/j.envsoft.2019.104502

CrossRef Full Text | Google Scholar

Yu S., Ma J. (2021). Deep Learning for Geophysics: Current and Future Trends. Rev. Geophysics 59 (3), e2021RG000742. doi: 10.1029/2021RG000742

CrossRef Full Text | Google Scholar

Keywords: artificial intelligence algorithms, machine learning, deep learning, marine science, marine data

Citation: Jiang M and Zhu Z (2022) The Role of Artificial Intelligence Algorithms in Marine Scientific Research. Front. Mar. Sci. 9:920994. doi: 10.3389/fmars.2022.920994

Received: 15 April 2022; Accepted: 25 April 2022;
Published: 18 May 2022.

Edited by:

Ahmad Salman, National University of Sciences and Technology (NUST), Pakistan

Reviewed by:

Nitin Agarwala, National Maritime Foundation, India

Copyright © 2022 Jiang and Zhu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Zhiyuan Zhu,