Inversion of the three-dimensional temperature structure of mesoscale eddies in the Northwest Pacific based on deep learning

Yu, Fangjie; Wang, Zeyuan; Liu, Shuai; Chen, Ge

doi:10.1007/s13131-021-1841-z

Cited by 11 publications

(5 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Among them, machine learning models such as MLR and RFR, have been previously employed in ocean temperature estimation (Guinehut et al, 2012;Su et al, 2018;Jeong et al, 2019). In addition, significant progress has been achieved in deep learning methodologies like the CNN architecture in this field (Su et al, 2021;Yu et al, 2021). These models are selected for comparison in esti mati ng ESTA, with parameters and variables adjusted accordingly.…”

Section: Compared Modelsmentioning

confidence: 99%

“…While MLR can fit the smooth characteristics of large-scale ocean temperature anomalies, the inherently nonlinear dynamics of mesoscale eddies impede the application of linear regressors such as MLR, leading to incapable of effectively capturing the nonlinear characteristics (Chelton et al, 2011b). In addition, compared to the RFR method applied in Case 6 (Su et al, 2018) and the CNNs method used in Case 7 (Yu et al, 2021), the performance of Cases 1 to 5 demonstrates lower RMSE and higher R values. As mentioned above, our approach, which employs the foundational architecture of convolutional neural networks, shares some similarities with the method used in Case 7.…”

Section: Feature Combinations and Evaluationsmentioning

confidence: 99%

“…However, compared to the estimation of temperature profiles in the global ocean, mesoscale eddies exhibit more intricate threedimensional structures, diverse shapes and pronounced nonlinearity (McGillicuddy, 2016). Yu et al (2021) devised a method applying the Eddy CNN method for ESTA with SLA and Argo observational profiles. But this network relies on SLA alone, which contains limited surface feature information, and the relatively simplistic structure of Eddy CNN poses challenges in dynamically adjusting attention weights and reducing noise to capture mesoscale eddy features.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Subsurface temperature estimation of mesoscale eddies in the Northwest Pacific Ocean from satellite observations using a residual muti-channel attention convolution network

Liu,

Zhang,

Zhang

et al. 2024

Front. Mar. Sci.

Self Cite

View full text Add to dashboard Cite

The mesoscale eddies are prevalent oceanic circulation phenomena, exerting significant influence on various aspects of the marine environment including energy transfer, material transport and ecosystem dynamics in the Northwest Pacific Ocean. However, due to sparse vertical observational data, the understanding of the three-dimensional temperature structure of individual cases of mesoscale eddies remains limited. In recent years, utilizing surface remote sensing observations to estimate subsurface temperature anomaly has been crucial for comprehending the intricate multi-dimensional dynamic processes in the ocean. Consequently, this paper proposes an eddy residual multi-channel attention convolution network (ERCACN) with the adaptive threshold and designs the combination of various surface features to estimate the eddy subsurface temperature anomaly (ESTA). By integrating results with climatic temperature, thermal structures containing 46 levels at depths up to 1000 m could be obtained, achieving excellent daily temporal resolution and 0.25° spatial resolution. Validation using independent Argo profiles from 2016 to 2017 reveals that the combination of multiple surface variables outperforms univariate methods, and the ERCACN model demonstrates superior performance compared to other approaches. Overall, with an 8% error deemed acceptable, the ERCACN model achieves a precision of 88.08% in estimating ESTA. This method provides a novel perspective for other essential oceanic variables, contributing to a better perception of the global climate system.

show abstract

Section: Compared Modelsmentioning

confidence: 99%

Section: Feature Combinations and Evaluationsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Subsurface temperature estimation of mesoscale eddies in the Northwest Pacific Ocean from satellite observations using a residual muti-channel attention convolution network

Liu,

Zhang,

Zhang

et al. 2024

Front. Mar. Sci.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Su et al [6] developed a random forest machine learning method, utilizing multi-source satellite observations such as surface height anomalies and in-situ Argo observational data, to explore global oceanic sea surface temperature anomalies. Yu et al [7] constructed a convolutional neural network (Eddy Convolution Neural, ECN) to invert the threedimensional structure of mesoscale eddies in the Northwest Pacific region. In existing studies, numerical simulation methods such as ROMS involve simulating the evolution process of mesoscale eddies using computers.…”

Section: Introductionmentioning

confidence: 99%

Temperature inversion of mesoscale eddies from satellite observations in the South China Sea

Huo,

Wang,

Geng

et al. 2024

Fifth International Conference on Geology, Mapping, and Remote Sensing (ICGMRS 2024)

View full text Add to dashboard Cite

As a common oceanic phenomenon, Mesoscale eddies play a significant role in oceanic heat exchange and energy transfer in the South China Sea. However, satellite remote sensing is primarily utilized for measuring surface ocean data, posing challenges in detecting subsurface eddies. This study proposes a data-driven deep learning approach to address the deficiency of subsurface oceanic data. This method leverages satellite surface data to establish a mapping model between surface satellite data and subsurface temperature. The proposed Unet-Eddy model can successfully invert the threedimensional temperature of mesoscale eddies at depths of 0-1000 m. Experimental results demonstrate the potential of this approach in inverting the three-dimensional subsurface temperature of mesoscale eddies and partially compensating for the shortage of internal oceanic data.

show abstract

“…Through deep ocean remote sensing (DORS) techniques combining satellite data and Argo data, it is possible to realize the expansion of satellite observations from the surface to the subsurface as well as the time sequence from the back to the front. So far, empirical statistical methods [18][19][20][21], and artificial intelligence data-driven methods [14,17,[22][23][24][25] have become the most common and popular methods for developing DORS techniques.…”

Section: Introductionmentioning

confidence: 99%

Subsurface Temperature Reconstruction for the Global Ocean from 1993 to 2020 Using Satellite Observations and Deep Learning

Jiang

Wang

et al. 2022

Remote Sensing

View full text Add to dashboard Cite

The reconstruction of the ocean’s 3D thermal structure is essential to the study of ocean interior processes and global climate change. Satellite remote sensing technology can collect large-scale, high-resolution ocean observation data, but only at the surface layer. Based on empirical statistical and artificial intelligence models, deep ocean remote sensing techniques allow us to retrieve and reconstruct the 3D ocean temperature structure by combining surface remote sensing observations with in situ float observations. This study proposed a new deep learning method, Convolutional Long Short-Term Memory (ConvLSTM) neural networks, which combines multisource remote sensing observations and Argo gridded data to reconstruct and produce a new long-time-series global ocean subsurface temperature (ST) dataset for the upper 2000 m from 1993 to 2020, which is named the Deep Ocean Remote Sensing (DORS) product. The data-driven ConvLSTM model can learn the spatiotemporal features of ocean observation data, significantly improves the model’s robustness and generalization ability, and outperforms the LighGBM model for the data reconstruction. The validation results show our DORS dataset has high accuracy with an average R2 and RMSE of 0.99/0.34 °C compared to the Argo gridded dataset, and the average R2 and NRMSE validated by the EN4-Profile dataset over the time series are 0.94/0.05 °C. Furthermore, the ST structure between DORS and Argo has good consistency in the 3D spatial morphology and distribution pattern, indicating that the DORS dataset has high quality and strong reliability, and well fills the pre-Argo data gaps. We effectively track the global ocean warming in the upper 2000 m from 1993 to 2020 based on the DORS dataset, and we further examine and understand the spatial patterns, evolution trends, and vertical characteristics of global ST changes. From 1993 to 2020, the average global ocean temperature warming trend is 0.063 °C/decade for the upper 2000 m. The 3D temperature trends revealed significant spatial heterogeneity across different ocean basins. Since 2005, the warming signal has become more significant in the subsurface and deeper ocean. From a remote sensing standpoint, the DORS product can provide new and robust data support for ocean interior process and climate change studies.

show abstract

Inversion of the three-dimensional temperature structure of mesoscale eddies in the Northwest Pacific based on deep learning

Cited by 11 publications

References 30 publications

Subsurface temperature estimation of mesoscale eddies in the Northwest Pacific Ocean from satellite observations using a residual muti-channel attention convolution network

Subsurface temperature estimation of mesoscale eddies in the Northwest Pacific Ocean from satellite observations using a residual muti-channel attention convolution network

Temperature inversion of mesoscale eddies from satellite observations in the South China Sea

Subsurface Temperature Reconstruction for the Global Ocean from 1993 to 2020 Using Satellite Observations and Deep Learning

Contact Info

Product

Resources

About