MS-SSTNet: A Scale-Aware Spatiotemporal Learning Framework for Satellite SST Forecasting via Iterative Multiscale Decomposition and Dual-Window Modelling
Aug 2026· Remote Sensing· Vol 18, pp. 2803· 0 citations· 57 references
TL;DR
MS-SSTNet is introduced, a scale-aware framework designed for spatiotemporal SST forecasting that leverages iterative multiscale decomposition and a dual-window temporal module is integrated to characterize the coupling between long-term persistent trends and short-term stochastic fluctuations.
Abstract
Accurate sea surface temperature (SST) forecasting underpins operational oceanography and climate surveillance, yet it remains constrained by the inherent multiscale spatiotemporal variability. Existing deep learning methods often struggle to identify the physical hierarchies of SST fields, typically treating them as homogeneous inputs and thus failing to decouple large-scale coherent structures from transient features effectively. To bridge this gap, this study introduces MS-SSTNet, a scale-aware framework designed for spatiotemporal SST forecasting that leverages iterative multiscale decomposition. By iteratively distilling SST fields into hierarchical spatial modes and their principal components (PCs), the architecture facilitates a rigorous scale-decoupling representation of ocean dynamics. A dual-window temporal module is then integrated to characterize the coupling between long-term persistent trends and short-term stochastic fluctuations. Evaluations using satellite-derived SST products over the South China Sea (SCS) demonstrate that MS-SSTNet achieves robust 10th day forecast skill, yielding an overall spatiotemporal average MAE of 0.3031 °C, an average RMSE of 0.4062 °C, and an average ACC of 0.7973 across the entire 1–10 day forecast horizon. Ablation studies further underscore the indispensability of multiscale decomposition and dual-window integration in enhancing forecast fidelity across diverse spatiotemporal scales.
This study proposes a novel deep learning framework that uses a U-Net to generate an initial high-resolution SST estimate, which is subsequently refined using a residual corrective approach, and progressively refines initial U-Net predictions by incorporating dynamically scaled residuals at each step, enabling accurate capture of broad patterns and fine-grained features such as eddies and fronts.
Onkar Jadhav, Tim French, I. Janeković et al.· 1 citation
Sea surface temperature (SST), as a key variable in the ocean-climate system, plays a crucial role in global climate evolution, the occurrence of extreme weather events, and changes in marine ecosystems. Accurate SST prediction is of great significance for improving medium- and long-term climate forecasting capabilities and supporting effective marine environmental management. In recent years, deep learning methods have been widely applied to SST prediction tasks and have achieved promising results. However, existing approaches often suffer from inadequate modeling of temporal periodic structures and multilevel spatial features, making it difficult to effectively capture the multiscale dynamic variations of SST under complex spatiotemporal contexts. To address this issue, this article proposes a multiscale periodic spatiotemporal graph convolutional network (MPSGCN) to model the multiscale periodic spatiotemporal dependencies of SST. The proposed model explicitly captures periodic patterns in SST sequences through a periodic modeling mechanism, integrates multiscale adaptive graph convolution to dynamically learn regional dependencies at varying spatial scales, and employs multilayer spatial convolutions to mitigate the over-smoothing issue introduced by spectral-based graph convolution. Extensive experiments conducted on three representative marine regions with distinct climatic characteristics demonstrate that the proposed model significantly outperforms existing state-of-the-art methods across multiple evaluation metrics, validating the effectiveness and robustness of MPSGCN in modeling the complex spatiotemporal evolution of SST. Code available at https://github.com/xiaoling-0621/MPSGCN
Ling Xiao, Peihao Yang, Lang He et al.· IEEE Transactions on Geoscie...· 0 citations
Seasurface temperature (SST) serves as an essential parameter of the ocean-atmosphere system, and its accurate prediction is of great significance for meteorological disaster warning and marine ecosystem research. To address the inadequate extraction of multiscale temporal evolution and frequency-domain features in existing methods, this article proposes an SST prediction method based on ConvLSTM. By synergizing multiscale temporal dynamics with frequency-spatial feature enhancement, a deep learning architecture named the multiscale temporal frequency-spatial ConvLSTM (MTFS-ConvLSTM) is proposed. The model captures multiscale temporal dependencies in the time dimension via the circular dilated-time module. Simultaneously, it integrates a frequency channel spatial attention module to map the SST field to the frequency domain using discrete cosine transform. This mechanism realizes dynamic filtering of key multifrequency information and spatial feature enhancement, and incorporates spatial attention to improve the modeling capability for complex spatiotemporal patterns of the SST field. Experimental results demonstrate that compared to models, including ConvLSTM, U-Net, SimVP, Swin-Transformer, adaptive Fourier neural operator (AFNO), and WMSR, the proposed MTFS-ConvLSTM model reduces the root-mean-square error of the 20th-day lead time SST prediction by 16.6%, 13.2%, 14.1%, 13.1%, 7.1%, and 9.8%, and the mean absolute error by 17.3%, 20.7%, 28.1%, 20.7%, 4.9%, and 12.8%, respectively, significantly mitigating the error accumulation phenomenon in SST prediction. This study validates the effectiveness of joint frequency and spatiotemporal domain modeling in improving SST prediction accuracy, providing a novel technical approach for marine environmental monitoring.
Zhixuan Zhou, Weifu Sun, Yuhao Zhang et al.· IEEE Journal of Selected Top...· 0 citations
Sea ice thickness (SIT) is a key variable in improving the accuracy and lead time of sea ice forecasts. However, obtaining continuous gridded Arctic spatiotemporal SIT fields remains a significant challenge. This study proposes a machine learning framework (ICE-3D), which directly produces daily gridded pan-Arctic SIT. ICE-3D establishes a spatiotemporal, multiscale, perception learning-based estimation paradigm. We design an innovative training strategy called spatiotemporal multiscale fusion window (STMFW). The strategy fuses physical covariations at different scales to improve targeted inference, especially for summer SIT. Validated against CryoSat-2 satellite altimetry and BGEP in situ measurements, ICE-3D achieves an overall correlation coefficient exceeding 0.75 for basin-scale SIT. Its summer SIT MAE is less than 0.38 m, outperforming TOPAZ4 reanalysis and showing comparable precision to PIOMAS within validated regions. Compared with static offline training, the online learning module reduces SIT overestimation bias in marginal ice zones by approximately 16%, and the model reliably captures high-frequency regional ice variations and seasonal sea ice volume (SIV) cycles. Our derived SIV results reveal a 38% loss of pan-Arctic sea ice volume from 1991 to 2020, accompanied by a stabilized melting rate over the past decade, indicating severely depleted perennial ice reserves and largely irreversible Arctic sea ice decline under the current climate regime. The ICE-3D SIT with quantified uncertainties supports multiscale Arctic climate analysis and provides critical data support for risk assessment during the peak Arctic shipping season.
The MTPV-HDRNet model is proposed as a fixed-site, multi-level intelligent forecasting model driven by ECMWF ERA5 pressure-level reanalysis data, adopting a hybrid encoder–decoder architecture that explicitly represents temporal evolution and vertical stratification in a decoupled but complementary manner.
Xue-cai Zhang, Zonghua Ding, Shuji Sun et al.· Frontiers in Astronomy and S...· 0 citations
Predicting weather and climate hazards typically relies on computationally expensive kilometer‐scale numerical models. This study introduces a U‐Net‐based deep learning framework, the Joint Atmospheric fields Downscaling Network (JADNet), for rapid, joint downscaling of multiple atmospheric variables to kilometer resolution. JADNet is trained to map interpolated 25‐km ERA5 (IERA5) reanalysis fields to 4‐km surface fields dynamically downscaled from the latter using the Weather Research and Forecasting model. The proposed model is evaluated against multiple extreme events over the Arabian Peninsula in 2014, including heatwaves, extreme precipitation, extreme surface winds, and a localized Tokar Gap Jet. These case studies indicate that JADNet consistently outperforms IERA5 in reproducing fine‐scale intensity, timing, and spatial patterns. Our results demonstrate that JADNet is a computationally efficient and robust surrogate for traditional dynamical downscaling, offering a powerful tool for scalable high‐resolution weather and climate applications.
Unknown authors· Journal of Geophysical Resea...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
