Predicting the 3-D ocean temperature field is a significant task that helps to understand global climate change and the state of ocean motion. Lots of numerical and data-driven models are used to predict ocean temperatures. However, these methods are restricted to the time-sequence prediction of discrete points or rely on convolutional layers to inefficiently capture local spatial dependencies for spatio-temporal prediction. In this letter, we propose a deep learning model named SA-PredRNN that combines attention mechanisms and predictive recurrent neural networks to capture global positional correlations and spatio-temporal features. Global gridded Argo temperature data with Barnes objective analysis (BOA-ARGO) are used to predict the future 3-D ocean temperature. The average RMSEs of the proposed model are promoted by at most 11% and 10%, which indicates that the SA-PredRNN model has better performance than the other baseline models.
15 July 2024 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Satellite remote sensing enables the extensive, long-term observation of oceanic changes. To achieve transparent ocean observation, artificial intelligence is innovatively used to reconstruct the three-dimensional ocean structure from remote sensing data with spatiotemporal attention. However, high-resolution satellite imagery imposes high computational demands. This study proposes Spatiotemporal Window Ocean (SWO), which is an efficient network that uses a special calculation strategy and spatiotemporal window attention to reduce complexity. The model integrates multiple satellite data sources, including sea surface temperature (SST), absolute dynamic topography (ADT), and sea surface salinity (SSS). Our experimental results demonstrate that SWO achieves lower computational costs and superior performance compared with recent commonly used spatiotemporal sequence models. Specifically, SWO requires a training time equivalent to 1/3 of SimVP, 1/4 of PredRNN, and 1/5 of SA-ConvLSTM while achieving a root mean square error (RMSE) index that is 13.8%, 20.3%, and 13.2% better, respectively. The computational advantages of SWO offer an important technical means for the high-resolution reconstruction of ocean phenomena in the future.
Determining the dynamic processes of the subsurface ocean is a critical yet formidable undertaking given the sparse measurement resources available presently. In this study, using the light gradient boosting machine algorithm (LightGBM), we report for the first time a machine learning strategy for retrieving subsurface velocities at 1000 dbar depth in the Southern Ocean from information derived from satellite observations. Argo velocity measurements are used in the training and validation of the LightGBM model. The results show that reconstructed subsurface velocity agrees better with Argo velocity than reanalysis datasets. In particular, the subsurface velocity estimates have a correlation coefficient of 0.78 and an RMSE of 4.09 cm/s, which is much better than the ECCO estimates, GODAS estimates, GLORYS12V1 estimates, and Ora-S5 estimates. The LightGBM model has a higher skill in the reconstruction of subsurface velocity than the random forest and the linear regressor models. The estimated subsurface velocity exhibits a statistically significant increase at 1000 dbar since the 1990s, providing new evidence for the deep acceleration of mean circulation in the Southern Ocean. This study demonstrates the great potential and advantages of statistical methods for subsurface velocity modeling and oceanic dynamical information retrieval.
28 November 2024 in IEEE Geoscience and Remote Sensing Letters
Ocean velocities are essential for understanding how the ocean influences and responds to climate dynamics, making their accurate reconstruction crucial for both climate modeling and predictions. However, reconstructing interior ocean velocities remains a significant challenge due to the sparse distribution of velocity observations and the ocean’s complex dynamics. In this study, we introduce an efficient methodology for reconstructing interior ocean velocities by combining sea surface satellite data—including sea surface height (SSH), temperature, wind, and current—with Argo velocity observations, using the dynamic mode decomposition (DMD) technique. DMD offers the advantage of reducing the dimensionality of interior velocity fields, helping to address the limitations caused by sparse observations. The reconstructed velocity for the Southern Pacific Ocean (SPO) was validated against Argo and acoustic Doppler current profiler (ADCP) velocities, showing a strong correlation than GLORYS12V1 velocities. In particular, the reconstructed velocities have a mean correlation coefficient of 0.78 for the zonal component and 0.74 for the meridional component above 1000 m. Additionally, the reconstructed flow field exhibits a coherent pattern that closely aligns with the eddies observed in SSH. This research significantly contributes to the Global Ocean Monitoring and Observing Program by enhancing both the accuracy and resolution of ocean velocity measurements.