Abstract:Accurate prediction of global sea surface temperature at sub-seasonal to seasonal (S2S) timescale is critical for drought and flood forecasting, as well as for improving disaster preparedness in human society. Government departments or academic studies normally use physics-based numerical models to predict S2S sea surface temperature and corresponding climate indices, such as El Niño-Southern Oscillation. However, these models are hampered by computational inefficiencies, limited retention of ocean-atmosphere initial conditions, and significant uncertainty and biases. Here, we introduce a novel three-dimensional deep learning neural network to model the nonlinear and complex coupled atmosphere-ocean weather systems. This model incorporates climatic and temporal features and employs a self-attention mechanism to enhance the prediction of global S2S sea surface temperature pattern. Compared to the physics-based models, it shows significant computational efficiency and predictive capability, improving one to three months sea surface temperature predictive skill by 13.7% to 77.1% in seven ocean regions with dominant influence on S2S variability over land. This achievement underscores the significant potential of deep learning for largely improving forecasting skills at the S2S scale over land.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the problem of accurately predicting the global sea - surface temperature (SST) on the sub - seasonal to seasonal (S2S) time scale. Specifically: 1. **Key background**: - Accurately predicting the global sea - surface temperature is crucial for drought and flood forecasts and for improving the disaster - response capabilities of human society. - Currently, government and academic research institutions usually use physics - based numerical models to predict S2S sea - surface temperatures and corresponding climate indices (such as El Niño - Southern Oscillation, ENSO). However, these models have problems of low computational efficiency, limited retention of initial conditions, high uncertainty, and large bias. 2. **Existing challenges**: - Physical models (such as general circulation models GCMs) face huge challenges in predicting SST on the sub - seasonal to seasonal scale, especially in temperate regions, because these models have difficulty fully representing air - sea interactions and atmospheric processes. - Traditional methods rely on partial differential equations and physical parameterizations to represent the coupled ocean - atmosphere system, which leads to huge computational requirements, limits their spatial resolution and ensemble size, and thus reduces the S2S SST prediction skills. 3. **Proposed new method**: - The authors introduce a new three - dimensional deep - learning neural network model - CAS - Canglong, for modeling nonlinear and complex coupled atmosphere - ocean weather systems. - This model combines climate and temporal features and adopts a self - attention mechanism to enhance the prediction ability of global S2S sea - surface temperature patterns. - Compared with physics - based models, CAS - Canglong shows significant computational efficiency and prediction ability, with improvement amplitudes ranging from 13.7 ± 3.0% to 77.1 ± 0.6% in S2S predictions in seven major ocean regions. 4. **Objectives and significance**: - Improving the prediction accuracy of the global sea - surface temperature on the S2S time scale helps to improve the early warning of extreme weather events and disaster preparedness. - By more accurately predicting SST, it is possible to better simulate and predict climate variability patterns related to SST, such as ENSO events, thus providing earlier extreme - event warnings for areas affected by ENSO. - The successful application of the CAS - Canglong model shows that deep learning has great potential in enhancing S2S - scale prediction skills, especially in the context of increasingly frequent extreme weather events under climate change. In summary, by developing and validating the CAS - Canglong model, this paper aims to solve the problems of low computational efficiency and insufficient accuracy in current S2S SST predictions, and provides a new way to improve the prediction ability of S2S weather and climate events.

CAS-Canglong: A skillful 3D Transformer model for sub-seasonal to seasonal global sea surface temperature prediction

Deep Learning-Based Seasonal Forecast of Sea Ice Considering Atmospheric Conditions

SST Forecast Skills Based on Hybrid Deep Learning Models: With Applications to the South China Sea

A spatiotemporal deep learning model for sea surface temperature field prediction using time-series satellite data

A Deep Learning Model For Forecasting Sea Surface Height Anomalies And Temperatures In The South China Sea

Predicting temporal and spatial 4-D ocean temperature using satellite data based on a novel deep learning model

A Deep Learning Model for Joint Prediction of Three-Dimensional Ocean Temperature, Salinity and Flow Fields

A Method for Predicting High-Resolution 3D Variations in Temperature and Salinity Fields Using Multi-Source Ocean Data

A Multivariable Convolutional Neural Network for Forecasting Synoptic-Scale Sea Surface Temperature Anomalies in the South China Sea

Short-Term Prediction of Global Sea Surface Temperature Using Deep Learning Networks

Sea Surface Temperature Prediction Using ConvLSTM-Based Model with Deformable Attention

Prediction of seasonal sea surface temperature based on temperature and salinity of subsurface ocean using machine learning

Prediction of long lead monthly three-dimensional ocean temperature using time series gridded Argo data and a deep learning method

Sea Surface Temperature and Marine Heat Wave Predictions in the South China Sea: A 3D U-Net Deep Learning Model Integrating Multi-Source Data

A self-attention–based neural network for three-dimensional multivariate modeling and its skillful ENSO predictions

The CNN-GRU model with frequency analysis module for sea surface temperature prediction

A deep learning model for forecasting global monthly mean sea surface temperature anomalies

Sea Surface Temperature Prediction Approach Based on 3D CNN and LSTM with Attention Mechanism

Seven-day sea surface temperature prediction using a 3DConv-LSTM model

Time Domain Convolutional Network Ocean Surface Temperature Prediction Model Based on Multi-Feature Fusion Attention Mechanism

Deep Learning Improves Sub-Seasonal Marine Heatwave Forecast