Deep Learning Enhances Near-Coastal Sea Surface Temperature

Deep Learning Enhances Near-Coastal Sea Surface Temperature Forecasting

Accurate prediction of near-coastal sea surface temperature (SST) is vital for effective coastal management, yet it poses significant challenges due to the limitations of traditional numerical ocean models.

March 27, 20263 min readFrontiers in Marine Science | New and Recent Articles

Deep Learning Enhances Near-Coastal Sea Surface Temperature Forecasting

World Data Ocean recognizes the critical importance of precise near-coastal sea surface temperature (SST) forecasting for a multitude of applications, from marine safety to effective coastal management. Traditional numerical ocean models, while foundational, often struggle to fully capture the intricate, fine-scale dynamics characteristic of nearshore environments. This inherent limitation can lead to significant forecasting challenges. Our work underscores the power of technological innovation to address these gaps, as exemplified by the study detailing a novel deep learning framework designed to overcome these very obstacles. By focusing on station-level predictions, this research offers a more granular and responsive approach to understanding and anticipating SST fluctuations in these vital regions.

The proposed deep learning framework introduces several key advancements that enhance predictive accuracy. Its seasonal stratified sampling strategy is a sophisticated method for ensuring that the model learns robust thermodynamic patterns across the entire annual cycle, crucially preventing issues arising from shifts in temporal data distribution. Furthermore, by addressing an identified information compression bottleneck within the Segment Recurrent Neural Network (SegRNN) architecture, the development of an Attention-Enhanced Parallel Multi-step Forecast (Attn-PMF) strategy represents a significant leap forward. This strategy's ability to directly retrieve high-variance features from historical sequences through global attention mechanisms allows for the preservation of crucial high-frequency variability, a characteristic often smoothed out by conventional forecasting methods.

The empirical validation of this framework, utilizing four years of hourly observations from 31 coastal stations in the East China Sea, provides compelling evidence of its superior performance. Notably, its advantage becomes increasingly pronounced for lead times beyond 48 hours, a critical window for actionable planning in marine operations and safety. This research clearly demonstrates that the Attn-PMF strategy effectively mitigates forecast degradation, delivering reliable predictions that can inform critical decision-making processes. At World Data Ocean, we believe that such advancements in data-driven forecasting are instrumental in fostering greater ocean intelligence and supporting our collective commitment to ocean stewardship. This study is a prime example of how innovative technology, grounded in scientific rigor, can translate into tangible benefits for coastal communities and the marine environment.

From Frontiers in Marine Science | New and Recent Articles

Accurate near-coastal sea surface temperature (SST) prediction remains challenging due to the limitations of numerical ocean models in resolving fine-scale coastal dynamics. This study proposes a novel deep learning framework specifically designed for station-level SST forecasting in nearshore regions. The framework employs a seasonal stratified sampling strategy to capture thermodynamic patterns across the annual cycle while preventing temporal distribution shift. Building upon the Segment Recurrent Neural Network (SegRNN) architecture, we identify a fundamental information compression bottleneck that causes forecast smoothing. To address this limitation, an Attention-Enhanced Parallel Multi-step Forecast (Attn-PMF) strategy is developed, enabling the model to directly retrieve high-variance…

Integrating Physics and AI Improves Sea Surface Temperature Forecasts

Accurate prediction of sea surface temperature (SST) is essential for marine environmental monitoring and climate forecasting. However, most existing deep-learning-based approaches rely heavily on data-driven methodologies and lack sufficient integration of physical mechanisms, thereby limiting their physical consistency and interpretability. To overcome this limitation, this study introduces a multi-source coupled prediction neural network (MSCPNN), which incorporates temperature, salinity, and current dynamics into a multi-scale feature learning framework. Built upon the multi-feature physical neural network (MFPNN), the proposed model integrates a convolutional block attention module (CBAM), where channel attention adaptively models the multi-factor coupling among temperature, salinity, and currents, and spatial attention captures multi-scale spatial patterns in SST. Using high-resolution reanalysis data from the South China Sea spanning 2011 to 2020, comprehensive experiments were conducted comparing MSCPNN with MFPNN and PCL-MFPNN. The results demonstrate that MSCPNN significantly outperforms the baseline models across multiple evaluation metrics—including RMSE, correlation coefficient, PSNR, and SSIM—achieving an average reduction in RMSE of 17% and an increase in correlation coefficient of 0.035, which reflects higher predictive accuracy and improved physical consistency. Ablation studies further validate the superiority of multi-factor coupling over single-factor alternatives and clarify the distinct contributions of salinity and currents to SST prediction. Overall, MSCPNN advances the accuracy and stability of long-term SST forecasting while providing a more interpretable framework for intelligent ocean prediction.