Application of deep learning technique to the sea surface height prediction in the South China Sea

Song, Tao; Han, Ningsheng; Zhu, Yuhang; Li, Zhongwei; Li, Yineng; Li, Shaotian; Peng, Shiqiu

doi:10.1007/s13131-021-1735-0

Cited by 16 publications

(13 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Ocean (Atmosphere) reanalysis gridded datasets are able to reproduce historical oceanic (atmospheric) states by combining oceanic (atmospheric) observations from multiple sources with a state-ofthe-art numerical ocean (atmosphere) model using robust data assimilation techniques. The development of reanalysis products has provided an unprecedented golden opportunity for deep learning to explore time series statistical predictions methods (Song et al, 2021). With the development of numerical models and the increase in grid resolution, as well as the improvement of data assimilation skills, long sequential and higher quality reanalysis data products have begun to emerge to serve as indicators of global/local climate and ecological change.…”

Section: Introductionmentioning

confidence: 99%

Monthly extended ocean predictions based on a convolutional neural network via the transfer learning method

Miao

Zhang

et al. 2023

Front. Mar. Sci.

View full text Add to dashboard Cite

Sea surface temperature anomalies (SSTAs) and sea surface height anomalies (SSHAs) are indispensable parts of scientific research, such as mesoscale eddy, current, ocean-atmosphere interaction and so on. Nowadays, extended-range predictions of ocean dynamics, especially in SSTA and SSHA, can provide daily prediction services in the range of 30 days, which bridges the gap between synoptic-scale weather forecasts and monthly average scale climate predictions. However, the forecast efficiency of extended range remains problematic. With the development of ocean reanalysis and satellite remote sensing products, large amounts datasets provide an unprecedented opportunity to use big data for the extended range prediction of ocean dynamics. In this study, a hybrid model, combing convolutional neural network (CNN) model with transfer learning (TL), was established to predict SSTA and SSHA at monthly scales, which makes full use of these data resources that arise from delayed gridding reanalysis products and real-time satellite remote sensing observations. The proposed model, where both ocean and atmosphere reanalysis datasets serve as the pretraining dataset and the satellite remote sensing observations are employed for fine-tuning based on the transfer learning (TL) method, can effectively capture the evolving spatial characteristics of SSTAs and SSHAs with low prediction errors over the 30 days range. When the forecast lead time is 30 days, the root means square errors for the SSTAs and SSHAs model results are 0.32°C and 0.027 m in the South China Sea, respectively, indicating that this model has not only satisfactory prediction performance but also offers great potential for practical operational applications in improving the skill of extended-range predictions.

show abstract

Section: Introductionmentioning

confidence: 99%

Monthly extended ocean predictions based on a convolutional neural network via the transfer learning method

Miao

Zhang

et al. 2023

Front. Mar. Sci.

View full text Add to dashboard Cite

show abstract

“…Sea surface height anomaly (SSHA) is considered one of the features for the surface and subsurface dynamics of the ocean and directly or indirectly reflects information on the main dynamic processes, including mesoscale eddies, waves, currents, and tides (Tanajura et al, 2016;Song et al, 2021). The SSHA has provided a wealth of information about ocean circulation and atmosphere-ocean interactions (Tandeo et al, 2014).…”

Section: Introductionmentioning

confidence: 99%

“…The prediction of SSHA has Is been a challenge (Song et al, 2021). For decades, numerical models based on dynamical/physical equations have played a dominant role in ocean predictions and are also often used to predict SSHA.…”

Section: Introductionmentioning

confidence: 99%

“…Moreover, numerical models involve a lot of complex physical processes, and their corresponding integration requires a lot of computational resources. Then, if rich SSHA data provided by satellite altimeters are assimilated, it will need relatively high computational cost and even lead to an overfitting situation (Li et al, 2010;Song et al, 2021), which, together with the effect of the model errors, causes this assimilation to not necessarily provide positive effects on predictions. To address these embarrassments, it is essential to thin the data and devise an appropriate assimilating strategy to highly and effectively initialize numerical models associated with SSHA forecasting (Li et al, 2010;Zanna et al, 2018;Fraser et al, 2019).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Evaluation of the sensitivity on mesoscale eddy associated with the sea surface height anomaly forecasting in the Kuroshio Extension

et al. 2023

View full text Add to dashboard Cite

The sensitivity of the sea surface height anomaly (SSHA) forecasting on the accuracy of mesoscale eddies over the Kuroshio Extension region, which was determined by the conditional non-linear optimal perturbation (CNOP) method using a two-layer quasigeostrophic model, is evaluated by adopting multiply realistic marine datasets through an advanced particle filter assimilation method. It is shown that, if additional observations are preferentially assimilated to the sensitive area of mesoscale eddies identified by the CNOP, where the eddies present a clear high- to low-velocity gradient along the eddy rotation, the forecasting skill of the SSHA can be more significantly improved. It is also demonstrated that the forecasts of the SSHA in the region where the large-scale mean flow possesses much stronger barotropic and/or baroclinic instability tend to exhibit stronger sensitivity to the accuracy of the initial field in the sensitive area of mesoscale eddies. Therefore, more attention should be preferentially paid to the assimilation of the additional observations of the mesoscale eddies for the SSHA forecast in the region with a strong velocity shear of ocean circulation. The present study verifies the sensitivity on mesoscale eddies of SSHA forecasts derived by the two-layer quasigeostrophic model using multiply sets of realistic oceanic data, especially including observation and reanalysis data, which further additionally demonstrates the importance of targeted observations of mesoscale eddies to the SSHA forecast in the regions of strong velocity shear of ocean circulation and provides a more credible scientific basis for the field campaign of the targeted observations for mesoscale eddies associated with the SSHA forecasting.

show abstract

“…

In recent years, deep learning has been developed, and some technologies [1][2][3][4][5][6] to improve the performance of deep learning have also emerged. However, deep learning often requires large amounts of labelled data.

…”

mentioning

confidence: 99%

A framework for self-supervised federated domain adaptation

Wang¹,

et al. 2022

J Wireless Com Network

View full text Add to dashboard Cite

Unsupervised federated domain adaptation uses the knowledge from several distributed unlabelled source domains to complete the learning on the unlabelled target domain. Some of the existing methods have limited effectiveness and involve frequent communication. This paper proposes a framework to solve the distributed multi-source domain adaptation problem, referred as self-supervised federated domain adaptation (SFDA). Specifically, a multi-domain model generalization balance is proposed to aggregate the models from multiple source domains in each round of communication. A weighted strategy based on centroid similarity is also designed for SFDA. SFDA conducts self-supervised training on the target domain to tackle domain shift. Compared with the classical federated adversarial domain adaptation algorithm, SFDA is not only strong in communication cost and privacy protection but also improves in the accuracy of the model.

show abstract

Application of deep learning technique to the sea surface height prediction in the South China Sea

Cited by 16 publications

References 24 publications

Monthly extended ocean predictions based on a convolutional neural network via the transfer learning method

Monthly extended ocean predictions based on a convolutional neural network via the transfer learning method

Evaluation of the sensitivity on mesoscale eddy associated with the sea surface height anomaly forecasting in the Kuroshio Extension

A framework for self-supervised federated domain adaptation

Contact Info

Product

Resources

About