Pub Date : 2026-02-01Epub Date: 2025-09-12DOI: 10.1016/j.ocemod.2025.102626
Jian Dong , Xu Qian , Huizan Wang
This work introduces a structure-preserving nonstaggered central scheme for the two-dimensional shallow water equations with wet–dry fronts and Coriolis force on triangular meshes. A key innovation of our approach is the development of a novel discretization method for source terms that exploits the geometric properties of the mesh within staggered cells. This method effectively overcomes the limitations of existing central schemes, which often exhibit a lack of well-balanced property in configurations that involve wet–dry fronts. In particular, the defined numerical fluxes not only utilize information from the central points but also from the vertex points. We rigorously show that the proposed numerical scheme maintains both positivity-preserving and well-balanced properties, essential attributes that ensure the physical validity and stability of the simulations. To verify our theoretical results, we conduct comprehensive numerical experiments that encompass a variety of scenarios. The results highlight the method’s exceptional performance in accurately modeling complex fluid dynamics associated with wet–dry fronts and Coriolis force.
{"title":"A structure-preserving nonstaggered central scheme for shallow water equations with wet–dry fronts and Coriolis force on triangles","authors":"Jian Dong , Xu Qian , Huizan Wang","doi":"10.1016/j.ocemod.2025.102626","DOIUrl":"10.1016/j.ocemod.2025.102626","url":null,"abstract":"<div><div>This work introduces a structure-preserving nonstaggered central scheme for the two-dimensional shallow water equations with wet–dry fronts and Coriolis force on triangular meshes. A key innovation of our approach is the development of a novel discretization method for source terms that exploits the geometric properties of the mesh within staggered cells. This method effectively overcomes the limitations of existing central schemes, which often exhibit a lack of well-balanced property in configurations that involve wet–dry fronts. In particular, the defined numerical fluxes not only utilize information from the central points but also from the vertex points. We rigorously show that the proposed numerical scheme maintains both positivity-preserving and well-balanced properties, essential attributes that ensure the physical validity and stability of the simulations. To verify our theoretical results, we conduct comprehensive numerical experiments that encompass a variety of scenarios. The results highlight the method’s exceptional performance in accurately modeling complex fluid dynamics associated with wet–dry fronts and Coriolis force.</div></div>","PeriodicalId":19457,"journal":{"name":"Ocean Modelling","volume":"199 ","pages":"Article 102626"},"PeriodicalIF":2.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145047496","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2026-01-02DOI: 10.1016/j.ocemod.2026.102677
Qingyu Zheng , Qi Shao , Guijun Han , Wei Li , Hong Li , Xuan Wang
Advances in ocean observation technology have significantly enhanced the accuracy of Earth system forecasting. Reconstructing missing information of nonlinear evolution processes from observational data is essential for investigating rapid changes in the marine environment and climate. However, traditional methods often struggle to extract unseen nonlinear processes from data. In fact, a large amount of dynamic evolution information hidden in historical data has not been effectively mined. To address this issue, we propose DeepDA, a latent space data assimilation approach based on deep learning. DeepDA employs a generative deep learning model to capture complex spatiotemporal multiscale features and nonlinear evolution processes in observations. By incorporating an attention mechanism, DeepDA effectively assimilates rich historical information of sea surface temperature. The results show that DeepDA remains highly stable in generating nonlinear evolution even with extensive data gaps and high noise levels. Notably, when only 10% (sparse sampling) of observation is available, the increase in error for DeepDA is limited to 40% compared to the case with complete data. Furthermore, DeepDA demonstrates effectiveness in multiscale reconstruction and analysis of climate variability, generating nonlinear patterns that are more physically consistent than linear methods. The nonlinear features extracted from the latent space exhibit multiscale structures, which may provide new insights into enhancing ocean data assimilation.
{"title":"Generating unseen nonlinear evolution in the ocean using deep learning-based latent space data assimilation model","authors":"Qingyu Zheng , Qi Shao , Guijun Han , Wei Li , Hong Li , Xuan Wang","doi":"10.1016/j.ocemod.2026.102677","DOIUrl":"10.1016/j.ocemod.2026.102677","url":null,"abstract":"<div><div>Advances in ocean observation technology have significantly enhanced the accuracy of Earth system forecasting. Reconstructing missing information of nonlinear evolution processes from observational data is essential for investigating rapid changes in the marine environment and climate. However, traditional methods often struggle to extract unseen nonlinear processes from data. In fact, a large amount of dynamic evolution information hidden in historical data has not been effectively mined. To address this issue, we propose DeepDA, a latent space data assimilation approach based on deep learning. DeepDA employs a generative deep learning model to capture complex spatiotemporal multiscale features and nonlinear evolution processes in observations. By incorporating an attention mechanism, DeepDA effectively assimilates rich historical information of sea surface temperature. The results show that DeepDA remains highly stable in generating nonlinear evolution even with extensive data gaps and high noise levels. Notably, when only 10% (sparse sampling) of observation is available, the increase in error for DeepDA is limited to 40% compared to the case with complete data. Furthermore, DeepDA demonstrates effectiveness in multiscale reconstruction and analysis of climate variability, generating nonlinear patterns that are more physically consistent than linear methods. The nonlinear features extracted from the latent space exhibit multiscale structures, which may provide new insights into enhancing ocean data assimilation.</div></div>","PeriodicalId":19457,"journal":{"name":"Ocean Modelling","volume":"200 ","pages":"Article 102677"},"PeriodicalIF":2.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145976937","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Submesoscale processes play a key role in re-stratifying the upper ocean through inducing strong vertical buoyancy flux (VBF). Because the prevailing climate and global ocean models are unable to resolve submesoscale processes, submesoscale VBF needs to be parameterized in models to reduce the associated simulation bias. Recently, Zhang et al. (2023) proposed a new VBF parameterization which simultaneously considers submesoscale baroclinic instability and strain-induced frontogenesis (Zhang23 parameterization hereafter). In this study, we implement the Zhang23 parameterization in a mesoscale-resolving (9-km) configuration of Regional Ocean Modeling System (ROMS) for the North Pacific, and assess its impact by comparing results with observations and a submesoscale-resolving (1-km) simulation. The parameterized VBFs have similar magnitudes and spatial patterns with those derived from the 1-km simulation, demonstrating the effectiveness of Zhang23 parameterization. Additionally, the Zhang23 parameterization yields significantly reduced mixed-layer depth (MLD) and strengthened upper-ocean stratification in winter compared with those in the control run without this parameterization. In the Kuroshio Extension region, the sensitivity run including the Zhang23 parameterization reduces the deep MLD bias by 94 % and yields an upper-ocean stratification in better agreement with a submesoscale-resolving simulation. These results show that the Zhang23 parameterization has a good potential to improve the simulation of upper-ocean processes in mesoscale-resolving models.
{"title":"Implementation and evaluation of a new parameterization of submesoscale vertical flux in a mesoscale-resolving model in the North Pacific","authors":"Zhe Feng , Zhiwei Zhang , Jinchao Zhang , Wenda Zhang , Man Yuan , Zhao Jing , Wei Zhao , Jiwei Tian","doi":"10.1016/j.ocemod.2025.102655","DOIUrl":"10.1016/j.ocemod.2025.102655","url":null,"abstract":"<div><div>Submesoscale processes play a key role in re-stratifying the upper ocean through inducing strong vertical buoyancy flux (VBF). Because the prevailing climate and global ocean models are unable to resolve submesoscale processes, submesoscale VBF needs to be parameterized in models to reduce the associated simulation bias. Recently, Zhang et al. (2023) proposed a new VBF parameterization which simultaneously considers submesoscale baroclinic instability and strain-induced frontogenesis (Zhang23 parameterization hereafter). In this study, we implement the Zhang23 parameterization in a mesoscale-resolving (9-km) configuration of Regional Ocean Modeling System (ROMS) for the North Pacific, and assess its impact by comparing results with observations and a submesoscale-resolving (1-km) simulation. The parameterized VBFs have similar magnitudes and spatial patterns with those derived from the 1-km simulation, demonstrating the effectiveness of Zhang23 parameterization. Additionally, the Zhang23 parameterization yields significantly reduced mixed-layer depth (MLD) and strengthened upper-ocean stratification in winter compared with those in the control run without this parameterization. In the Kuroshio Extension region, the sensitivity run including the Zhang23 parameterization reduces the deep MLD bias by 94 % and yields an upper-ocean stratification in better agreement with a submesoscale-resolving simulation. These results show that the Zhang23 parameterization has a good potential to improve the simulation of upper-ocean processes in mesoscale-resolving models.</div></div>","PeriodicalId":19457,"journal":{"name":"Ocean Modelling","volume":"199 ","pages":"Article 102655"},"PeriodicalIF":2.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145614618","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-10-31DOI: 10.1016/j.ocemod.2025.102649
Rui Li , Kejian Wu , Qingxiang Liu , Jin Liu , Shang-Min Long , Jian Sun , Alexander V. Babanin
The effect of the small-scale ocean surface waves on large-scale ocean climate has been usually neglected. The Stokes drift-induced water transport has the potential to contribute to ocean heat transport and the wave-induced heat transport (WHT) in the global ocean is quantified for the first time in this research. The magnitude of wave-induced water transport is found to be comparable to Ekman transport in the global ocean. Notably, both of the zonal and meridional surface Stokes drift exhibit a strong correlation with the El Niño-Southern Oscillation and Indian Ocean Dipole (IOD). We found that there is an anomalous increase in wave-induced heat transport towards the equator during El Niño events in the Pacific Ocean. Additionally, an increase in eastward WHT appears during eastern-type El Niño events. Moreover, the zonal WHT anomalies co-vary with IOD phases. The large-scale climate modes drive the ocean wave large-scale anomalies, and then the abnormal WHT leads to redistribution of global ocean heat, even exceeding the heat transport induced by Ekman transport.
{"title":"The wave-induced heat transport in the global ocean","authors":"Rui Li , Kejian Wu , Qingxiang Liu , Jin Liu , Shang-Min Long , Jian Sun , Alexander V. Babanin","doi":"10.1016/j.ocemod.2025.102649","DOIUrl":"10.1016/j.ocemod.2025.102649","url":null,"abstract":"<div><div>The effect of the small-scale ocean surface waves on large-scale ocean climate has been usually neglected. The Stokes drift-induced water transport has the potential to contribute to ocean heat transport and the wave-induced heat transport (WHT) in the global ocean is quantified for the first time in this research. The magnitude of wave-induced water transport is found to be comparable to Ekman transport in the global ocean. Notably, both of the zonal and meridional surface Stokes drift exhibit a strong correlation with the El Niño-Southern Oscillation and Indian Ocean Dipole (IOD). We found that there is an anomalous increase in wave-induced heat transport towards the equator during El Niño events in the Pacific Ocean. Additionally, an increase in eastward WHT appears during eastern-type El Niño events. Moreover, the zonal WHT anomalies co-vary with IOD phases. The large-scale climate modes drive the ocean wave large-scale anomalies, and then the abnormal WHT leads to redistribution of global ocean heat, even exceeding the heat transport induced by Ekman transport.</div></div>","PeriodicalId":19457,"journal":{"name":"Ocean Modelling","volume":"199 ","pages":"Article 102649"},"PeriodicalIF":2.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145465783","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-11-28DOI: 10.1016/j.ocemod.2025.102657
Junkai Qian , Qiang Wang , Wuhong Guo , Huier Mo , Hui Zhang
The Kuroshio, a powerful western boundary current in the North Pacific, exhibits multi-scale variability that profoundly affects regional weather, climate, marine ecosystems, and fisheries, rendering its accurate prediction indispensable. However, this variability is driven by complex multi-scale physical processes, necessitating high-resolution numerical models that are computationally expensive and often constrained by limited timeliness. In recent years, the emergence of data-driven models has opened new avenues for ocean forecasting, and the global ocean intelligent prediction systems are now approaching or even surpassing traditional numerical models across various metrics. Despite these advances, their performance in the Kuroshio region remains limited. To address this challenge, this study develops an eddy-resolving (1/12°) Kuroshio Intelligent Prediction System (KIPS) based on the Swin Transformer architecture. Specifically designed to capture Kuroshio dynamics, KIPS uses an autoregressive strategy to generate daily forecasts of three-dimensional temperature, salinity, current, and sea surface height, with a lead time of up to 10 days. KIPS achieves higher accuracy compared to existing numerical and AI-based prediction systems, while significantly reducing computational costs. In operational forecasts, KIPS successfully captures several recent eddy shedding and merging events in the southern Kuroshio region of Japan, demonstrating agreement with near-real-time satellite observations. These results underscore the value of integrating prior physical knowledge into region-specific forecast systems to improve fine-scale ocean prediction.
{"title":"Improving Kuroshio forecasts with an eddy-resolving AI prediction system","authors":"Junkai Qian , Qiang Wang , Wuhong Guo , Huier Mo , Hui Zhang","doi":"10.1016/j.ocemod.2025.102657","DOIUrl":"10.1016/j.ocemod.2025.102657","url":null,"abstract":"<div><div>The Kuroshio, a powerful western boundary current in the North Pacific, exhibits multi-scale variability that profoundly affects regional weather, climate, marine ecosystems, and fisheries, rendering its accurate prediction indispensable. However, this variability is driven by complex multi-scale physical processes, necessitating high-resolution numerical models that are computationally expensive and often constrained by limited timeliness. In recent years, the emergence of data-driven models has opened new avenues for ocean forecasting, and the global ocean intelligent prediction systems are now approaching or even surpassing traditional numerical models across various metrics. Despite these advances, their performance in the Kuroshio region remains limited. To address this challenge, this study develops an eddy-resolving (1/12°) Kuroshio Intelligent Prediction System (KIPS) based on the Swin Transformer architecture. Specifically designed to capture Kuroshio dynamics, KIPS uses an autoregressive strategy to generate daily forecasts of three-dimensional temperature, salinity, current, and sea surface height, with a lead time of up to 10 days. KIPS achieves higher accuracy compared to existing numerical and AI-based prediction systems, while significantly reducing computational costs. In operational forecasts, KIPS successfully captures several recent eddy shedding and merging events in the southern Kuroshio region of Japan, demonstrating agreement with near-real-time satellite observations. These results underscore the value of integrating prior physical knowledge into region-specific forecast systems to improve fine-scale ocean prediction.</div></div>","PeriodicalId":19457,"journal":{"name":"Ocean Modelling","volume":"200 ","pages":"Article 102657"},"PeriodicalIF":2.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145692888","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-11-22DOI: 10.1016/j.ocemod.2025.102656
Susmita Saha , Satyasaran Changdar , Soumen De
Solving the shallow water equations is essential in science and engineering for understanding and predicting geophysical phenomena such as atmospheric and oceanic flows. Physics-informed machine learning has emerged as a powerful alternative to traditional numerical methods, avoiding the complexities of grid generation and enabling mesh-free solutions to partial differential equations. In this study, we apply a sequential multi-model approach within a time-decomposed framework to solve the shallow water equations on a rotating sphere, in the context of meteorological applications. We employed advanced physics-informed neural networks integrated with deep learning, using diverse network architectures to conduct a detailed analysis of cosine bell advection across multiple orientations on the Earth. The results demonstrate high predictive accuracy, underscoring the method’s transformative potential for geophysical fluid dynamics. We also implemented a finite difference upwind scheme and a fully data-driven deep neural network to supplement the validation process and comparative analysis. Additionally, we perform a sensitivity analysis to examine the influence of physics-informed error terms on the training dynamics of the networks.
{"title":"Multi-model physics informed neural networks to the shallow water equations for cosine bell advection","authors":"Susmita Saha , Satyasaran Changdar , Soumen De","doi":"10.1016/j.ocemod.2025.102656","DOIUrl":"10.1016/j.ocemod.2025.102656","url":null,"abstract":"<div><div>Solving the shallow water equations is essential in science and engineering for understanding and predicting geophysical phenomena such as atmospheric and oceanic flows. Physics-informed machine learning has emerged as a powerful alternative to traditional numerical methods, avoiding the complexities of grid generation and enabling mesh-free solutions to partial differential equations. In this study, we apply a sequential multi-model approach within a time-decomposed framework to solve the shallow water equations on a rotating sphere, in the context of meteorological applications. We employed advanced physics-informed neural networks integrated with deep learning, using diverse network architectures to conduct a detailed analysis of cosine bell advection across multiple orientations on the Earth. The results demonstrate high predictive accuracy, underscoring the method’s transformative potential for geophysical fluid dynamics. We also implemented a finite difference upwind scheme and a fully data-driven deep neural network to supplement the validation process and comparative analysis. Additionally, we perform a sensitivity analysis to examine the influence of physics-informed error terms on the training dynamics of the networks.</div></div>","PeriodicalId":19457,"journal":{"name":"Ocean Modelling","volume":"199 ","pages":"Article 102656"},"PeriodicalIF":2.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145614617","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-09-26DOI: 10.1016/j.ocemod.2025.102623
Ali Abdolali , Tyler J. Hesser , Aron Roland , Martha Schönau , David A. Honegger , Jane McKee Smith , Héloïse Michaud , Luca Centurioni
Using the six-month hurricane season of 2022 as a case study and the spectral wave model WAVEWATCH III, this effort shows that wave parameters produced via a variable-resolution global mesh (5–30 km) agree with a diverse array of validating observational datasets at a level comparable to that of a constant-resolution mesh (3 km) that is six times more costly to run. The optimized variable-resolution, unstructured triangular mesh is faithful to land geometry and wave transformation gradients while relaxing focus in deeper regions where gradients are typically less pronounced. Wave parameters measured via satellite altimetry, stationary buoy networks, and drifting buoys are employed to demonstrate not only a substantial increase in performance over a coarse, constant-resolution grid (40 km), with RMSE reduced from 0.28 m to 0.14 m and Correlation Coefficient (CC) improved from 0.92 to 0.98 overall, but also a comparable level of performance to that of a mesh that has undergone a full convergence analysis. Performance comparisons isolated to shallow regions and near cyclonic storms highlight the importance of resolving relevant geometries. For nearshore data, RMSE improves from 0.29 m to 0.13 m and CC from 0.89 to 0.98; in shallow regions, RMSE from 0.29 m to 0.15 m and CC from 0.88 to 0.97; and under cyclonic conditions, RMSE from 0.62 m to 0.35 m and CC from 0.93 to 0.98. Wave model results using the variable-resolution mesh were further analyzed to provide a detailed summary of the wave climate, including wind-wave and swell partitions, over the six-month study period in the study area.
{"title":"Advancing multi-scale wave modeling: Global and coastal applications during the 2022 Atlantic hurricane season","authors":"Ali Abdolali , Tyler J. Hesser , Aron Roland , Martha Schönau , David A. Honegger , Jane McKee Smith , Héloïse Michaud , Luca Centurioni","doi":"10.1016/j.ocemod.2025.102623","DOIUrl":"10.1016/j.ocemod.2025.102623","url":null,"abstract":"<div><div>Using the six-month hurricane season of 2022 as a case study and the spectral wave model WAVEWATCH III, this effort shows that wave parameters produced via a variable-resolution global mesh (5–30 km) agree with a diverse array of validating observational datasets at a level comparable to that of a constant-resolution mesh (3 km) that is six times more costly to run. The optimized variable-resolution, unstructured triangular mesh is faithful to land geometry and wave transformation gradients while relaxing focus in deeper regions where gradients are typically less pronounced. Wave parameters measured via satellite altimetry, stationary buoy networks, and drifting buoys are employed to demonstrate not only a substantial increase in performance over a coarse, constant-resolution grid (40 km), with RMSE reduced from 0.28 m to 0.14 m and Correlation Coefficient (CC) improved from 0.92 to 0.98 overall, but also a comparable level of performance to that of a mesh that has undergone a full convergence analysis. Performance comparisons isolated to shallow regions and near cyclonic storms highlight the importance of resolving relevant geometries. For nearshore data, RMSE improves from 0.29 m to 0.13 m and CC from 0.89 to 0.98; in shallow regions, RMSE from 0.29 m to 0.15 m and CC from 0.88 to 0.97; and under cyclonic conditions, RMSE from 0.62 m to 0.35 m and CC from 0.93 to 0.98. Wave model results using the variable-resolution mesh were further analyzed to provide a detailed summary of the wave climate, including wind-wave and swell partitions, over the six-month study period in the study area.</div></div>","PeriodicalId":19457,"journal":{"name":"Ocean Modelling","volume":"199 ","pages":"Article 102623"},"PeriodicalIF":2.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221531","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-12-13DOI: 10.1016/j.ocemod.2025.102674
Guangjun Xu , Yucheng Shi , Xueming Zhu , Zhao Jing , Shuyi Zhou , Jiexin Xu , Huabing Xu , Guancheng Wang , Dongyang Fu , Changming Dong
Accurate, simultaneous prediction of three-dimensional (3D) ocean temperature, salinity, and current fields is vital for understanding ocean dynamics and informing marine applications. This study introduces a Fourier Neural Operator (FNO)-based model specifically designed for this 3D multi-variable task, leveraging Fourier transforms to efficiently capture complex multi-scale spatio-temporal dependencies within the ocean state. Evaluated on multi-year data from the South China Sea, the FNO model demonstrates strong predictive skill. Compared against the Copernicus Marine Environment Monitoring Service (CMEMS) operational forecast product, our model achieved significant average reductions in Root Mean Square Error (RMSE) by 43.07 % and Mean Absolute Error (MAE) by 46.18 % (averaged across all four variables and the full 10-day forecast horizon). The FNO particularly excels in short-term predictions (1–3 days), outperforming conventional deep learning benchmarks (such as U-Net) in accuracy for key variables. Spectral analysis reveals this outperformance is linked to FNO's superior ability to represent the energy of multi-scale oceanic features, indicating a more faithful capture of their structures, while also offering substantial computational efficiency compared to traditional numerical simulations. While forecast accuracy decreases over longer periods, this work highlights the considerable potential of FNOs as a scalable and effective data-driven approach for advancing 3D oceanographic forecasting.
{"title":"Prediction of three-dimensional ocean temperature, salinity and current fields based on fourier neural operators","authors":"Guangjun Xu , Yucheng Shi , Xueming Zhu , Zhao Jing , Shuyi Zhou , Jiexin Xu , Huabing Xu , Guancheng Wang , Dongyang Fu , Changming Dong","doi":"10.1016/j.ocemod.2025.102674","DOIUrl":"10.1016/j.ocemod.2025.102674","url":null,"abstract":"<div><div>Accurate, simultaneous prediction of three-dimensional (3D) ocean temperature, salinity, and current fields is vital for understanding ocean dynamics and informing marine applications. This study introduces a Fourier Neural Operator (FNO)-based model specifically designed for this 3D multi-variable task, leveraging Fourier transforms to efficiently capture complex multi-scale spatio-temporal dependencies within the ocean state. Evaluated on multi-year data from the South China Sea, the FNO model demonstrates strong predictive skill. Compared against the Copernicus Marine Environment Monitoring Service (CMEMS) operational forecast product, our model achieved significant average reductions in Root Mean Square Error (RMSE) by 43.07 % and Mean Absolute Error (MAE) by 46.18 % (averaged across all four variables and the full 10-day forecast horizon). The FNO particularly excels in short-term predictions (1–3 days), outperforming conventional deep learning benchmarks (such as U-Net) in accuracy for key variables. Spectral analysis reveals this outperformance is linked to FNO's superior ability to represent the energy of multi-scale oceanic features, indicating a more faithful capture of their structures, while also offering substantial computational efficiency compared to traditional numerical simulations. While forecast accuracy decreases over longer periods, this work highlights the considerable potential of FNOs as a scalable and effective data-driven approach for advancing 3D oceanographic forecasting.</div></div>","PeriodicalId":19457,"journal":{"name":"Ocean Modelling","volume":"200 ","pages":"Article 102674"},"PeriodicalIF":2.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145798078","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Langmuir turbulence in shallow-water coastal environments can reach the seafloor, developing into Langmuir supercells, which enhance size and mixing intensity. Two fundamental issues in coastal Langmuir turbulence remain unclear: (i) the energy cycle of the turbulence under different circumstances, and (ii) its effect on vertical mixing. We investigate these issues using large eddy simulations, considering aligned and opposing wind-wave and current directions. Results show that Langmuir supercells possess an intense full-column, narrow-band energetic mode, distinct from Langmuir turbulence in the energy spectrum. This mode occurs with aligned wind/wave and current directions but disappears when they oppose. In the latter case, only Langmuir and shear turbulence exist near surface and bottom boundaries; moreover, despite no stratification in simulations, their intensities are suppressed by a mid-layer barrier that limits surface-bottom interaction. When Langmuir supercells are present, the surface-bottom exchange of momentum is highly asymmetric between upwelling and downwelling limbs. Strong connections between surface and bottom turbulence, as indicated by the vortex-tube-connection events, can only be found in upwelling regions. As a result, the upwelling motions contribute considerably more to the momentum flux than the downwelling motions. All these results indicate that, despite the windrow pattern on the ocean surface from near-surface wind-wave interaction, whether full-column supercells can be activated or suppressed depends on different interactions between near-surface wind-wave forcing and near-bottom shear forcing. Once Langmuir supercells are activated, they differ significantly from Langmuir turbulence from the perspectives of energy and momentum transport; therefore, they cannot be simply treated as a “full column” version of Langmuir turbulence.
{"title":"The role of longitudinal alignment between surface and bottom forcing on the full-column turbulence mixing in the coastal ocean","authors":"Jiahao Huang , Marcelo Chamecki , Qing Li , Bicheng Chen","doi":"10.1016/j.ocemod.2025.102637","DOIUrl":"10.1016/j.ocemod.2025.102637","url":null,"abstract":"<div><div>Langmuir turbulence in shallow-water coastal environments can reach the seafloor, developing into Langmuir supercells, which enhance size and mixing intensity. Two fundamental issues in coastal Langmuir turbulence remain unclear: (i) the energy cycle of the turbulence under different circumstances, and (ii) its effect on vertical mixing. We investigate these issues using large eddy simulations, considering aligned and opposing wind-wave and current directions. Results show that Langmuir supercells possess an intense full-column, narrow-band energetic mode, distinct from Langmuir turbulence in the energy spectrum. This mode occurs with aligned wind/wave and current directions but disappears when they oppose. In the latter case, only Langmuir and shear turbulence exist near surface and bottom boundaries; moreover, despite no stratification in simulations, their intensities are suppressed by a mid-layer barrier that limits surface-bottom interaction. When Langmuir supercells are present, the surface-bottom exchange of momentum is highly asymmetric between upwelling and downwelling limbs. Strong connections between surface and bottom turbulence, as indicated by the vortex-tube-connection events, can only be found in upwelling regions. As a result, the upwelling motions contribute considerably more to the momentum flux than the downwelling motions. All these results indicate that, despite the windrow pattern on the ocean surface from near-surface wind-wave interaction, whether full-column supercells can be activated or suppressed depends on different interactions between near-surface wind-wave forcing and near-bottom shear forcing. Once Langmuir supercells are activated, they differ significantly from Langmuir turbulence from the perspectives of energy and momentum transport; therefore, they cannot be simply treated as a “full column” version of Langmuir turbulence.</div></div>","PeriodicalId":19457,"journal":{"name":"Ocean Modelling","volume":"199 ","pages":"Article 102637"},"PeriodicalIF":2.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145267922","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-11-07DOI: 10.1016/j.ocemod.2025.102646
Raisha Lovindeer , Elizabeth A. Fulton , Susan E. Allen , Javier Porobic , Douglas J. Latornell , Hem Nalini Morzaria-Luna , Alaia Morell
Biological risk assessment modelling for oil spills using whole-of-ecosystem models has the benefit of assessing species-specific toxicology and the chronic impact of oil spills by layering these impacts on top of the already-built ecosystem within the model. In deterministic models this approach requires tracking contaminants as they move throughout the biology of the ecosystem, from uptake to loss. Here we consolidate, modify, and add to existing equations to produce a synergistic set that can be used to define the impact of contaminants on biological groups throughout the food web. We demonstrate how these equations work, individually as well as in tandem, for oil-based contaminants by implementing them in a three-dimensional marine ecosystem model. We assess the sensitivity of parameters within these equations, showing the impact on the model outcome. Although we focus on oil-based contaminants in our examples, the equations presented can be applied to any contaminants in the aquatic or marine environment.
{"title":"Equations for modelling contaminant impacts throughout a marine ecosystem","authors":"Raisha Lovindeer , Elizabeth A. Fulton , Susan E. Allen , Javier Porobic , Douglas J. Latornell , Hem Nalini Morzaria-Luna , Alaia Morell","doi":"10.1016/j.ocemod.2025.102646","DOIUrl":"10.1016/j.ocemod.2025.102646","url":null,"abstract":"<div><div>Biological risk assessment modelling for oil spills using whole-of-ecosystem models has the benefit of assessing species-specific toxicology and the chronic impact of oil spills by layering these impacts on top of the already-built ecosystem within the model. In deterministic models this approach requires tracking contaminants as they move throughout the biology of the ecosystem, from uptake to loss. Here we consolidate, modify, and add to existing equations to produce a synergistic set that can be used to define the impact of contaminants on biological groups throughout the food web. We demonstrate how these equations work, individually as well as in tandem, for oil-based contaminants by implementing them in a three-dimensional marine ecosystem model. We assess the sensitivity of parameters within these equations, showing the impact on the model outcome. Although we focus on oil-based contaminants in our examples, the equations presented can be applied to any contaminants in the aquatic or marine environment.</div></div>","PeriodicalId":19457,"journal":{"name":"Ocean Modelling","volume":"199 ","pages":"Article 102646"},"PeriodicalIF":2.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145516669","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号